Vol 51 | Issue 5 | October 2020
Hydrology Research Advances in Eco-hydrology and Watershed Water Resources Management in China
ISSN Online 1234-5678 iwaponline.com/hr
Editorial Board Editors-in-Chief
US Mail Identification Statement
Bjørn Kløve Water, Energy and Environmental Engineering Research Unit, Faculty of Technology, 90014 University of Oulu, Finland E-mail: bjorn.klove@oulu.fi
Hydrology Research is published six times a year, in February, April, June, August, October and December. IWA Publishing is a company registered in the UK, no. 3690822.
Nevil Wyndham Quinn Department of Geography & Environmental Management, Faculty of Environment & Technology, University of the West of England, Bristol, Bristol BS16 1QY, UK E-mail: Nevil.Quinn@uwe.ac.uk
US Publishing office: Hydrology Research, c/o Mercury Airfreight International Ltd, 365 Blair Road, Avenel, NJ 07001, USA Periodicals postage paid at Rahway, NJ, USA Postmaster: send address corrections to Hydrology Research, c/o Mercury Airfreight International Ltd, 365 Blair Road, Avenel, NJ 07001, USA
Editor Emeritus Chong-Yu Xu Department of Geosciences, University of Oslo, Sem Sælands vei 1, 0316 Oslo, Norway E-mail: c.y.xu@geo.uio.no
Associate Editors Andrew Barton, Federation University, Australia Jie Chen, Wuhan University, China Harry Dixon, UK Centre for Ecology and Hydrology, UK Michał Wojciech Kawecki, UK Emmanouil Varouchakis, Technical University of Crete, Greece Elena Volpi, Università Roma Tre, Italy Przemysław Wachniew, AGH University of Science and Technology, Poland Ke Zhang, Hohai University, China
Young Career Researcher Editorial Board Tanveer Adyel, Monash University, Australia Zheng Duan, Lunds Universitet, Sweden Masoud Irannezhad, Southern University of Science and Technology, China Mahboobeh Kiani-Harchegani, Yazd University, Iran Hong Li, Oslo Metripolitan University, Norway Kabir Rasouli, Environment and Climate Change Canada, Canada Prabin Rokaya, University of Saskatchewan, Cananda Hossein Tabari, KU Leuven, Belgium Cristina Valhondo, University de Montpellier, France
Editorial Board Azadeh Ahmadi, Isfahan University of Technology, Iran Paul Bates, University of Bristol, UK Daniela Biondi, Università della Calabria, Italy Andrew R. Black, University of Dundee, UK Günter Blöschl, Technical University of Vienna, Austria Michael Bruen, University College Dublin, Ireland Markus C. Casper, University of Trier, Germany Lu Chen, Huahzong University of Science and Technology, China Hannah Cloke, University of Reading, UK Barry F.W. Croke, The Australian National University, Australia Christophe Cudennec, INRA, France Ravinesh Deo, University of Southern Queensland, Australia Weili Duan, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, China Guangtao Fu, University of Exeter, UK Efi Foufoula-Georgiou, University of Minnesota, USA Sigurdur M. Gardarsson, University of Iceland, Iceland Rien van Genuchten, Federal University of Rio de Janeiro, Brazil Jamie Hannaford, Centre for Ecology and Hydrology, UK Nilgun B. Harmancioglu, Dokuz Eylul University, Turkey Denis Hughes, Rhodes University, South Africa Thomas Kjeldsen, University of Bath, UK Ozgur Kisi, Ilia State University, Georgia Harri Koivusalo, Aalto University, Finland George Kuczera, The University of Newcastle, Australia Ian Littlewood, IGLEnvironment, UK Mario Martina, School of Advanced Studies (IUSS), Italy Neil McIntyre, The University of Queensland, Australia Eva Nora Paton, TU Berlin, Germany Domenico Pianese, Universita' degli studi di Napoli, Italy John W. Pomeroy, University of Saskatchewan, Canada Roberto Ranzi, University of Brescia, Italy Liliang Ren, Hohai University, China Noori Roohollah, University of Tehran, Iran Dan Rosbjerg, Technical University of Denmark, Denmark K.D. Sharma, National Institute of Hydrology, India Kuniyoshi Takeuchi, ICHARM, PWRI, Japan Valery Vuglinsky, State Hydrological Institute, Russia Guoqiang Wang, Beijing Normal University, China Howard Wheater, University of Saskatchewan, Canada Stefanos Xenarios, Nazarbayev University, Kazakhstan Xing Yuan, Nanjing University of Information Science and Technology, China Qi Zhang, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, China
Aims and Scope It is the principal aim of the Journal to conform to the highest possible scientific and technical standards. Hydrology Research publishes articles within hydrology in its widest sense, with an emphasis on water quantity and quality aspects of the hydrological cycle. Papers that draw upon adjoining sciences are embraced as well. Hydrology Research is intended to be a link between basic hydrological research and practical application of scientific results within the broad field of water management.
Submission All manuscripts should be submitted electronically through our Online Submission and Peer Review System: http://www.editorialmanager.com/hydrology. Authors should prepare their papers in accordance with the current Instructions for Authors, available at https://iwaponline.com/hr/pages/Instructions_for_authors or from the Editorial Office at the address given inside the back cover.
Published in partnership with: Nordic Association for Hydrology Nordic Association for Hydrology (NHF) is an independent body aiming at promoting hydrology as a science, increasing the understanding of hydrology, and enhancing the application of hydrological methods within applied sciences and water management, generally and in the North. The Association strongly supports international hydrological co-operation. More information is available at http://www.nhf-hydrology.org/
British Hydrological Society The British Hydrological Society (BHS) was formed in 1983 in response to a clear need in the UK for a new, broad-based, national society for the advancement of hydrology. The Society caters for all those with an interest in the inter-disciplinary subject of hydrology. It aims to promote interest and scholarship in scientific and applied aspects of hydrology, and to foster the involvement of its members in national and international activities. More information is given at http://www.hydrology.org.uk/
Official Journal of: The Italian Hydrological Society (IHS): http://www.sii-ihs.it/ The German Hydrological Society (DHG): http://www.dhydrog.de/
Hydrology Research An International Journal volume 51 |
issue 5 | October 2020
Contents Advances in Eco-hydrology and Watershed Water Resources Management in China 833 Editorial: Advances in Eco-hydrology and Watershed Water Resources Management in China Guoqiang Wang 834 An improved routing algorithm for a large-scale distributed hydrological model with consideration of underlying surface impact Jingjing Li, Haoyuan Zhao, Jun Zhang, Hua Chen, Chong-Yu Xu, Lu Li, Jie Chen and Shenglian Guo 854 Identification of regional water security issues in China, using a novel water security comprehensive evaluation model Jiping Yao, Guoqiang Wang, Baolin Xue, Gang Xie and Yanbo Peng 867 Copula-based drought severity-area-frequency curve and its uncertainty, a case study of Heihe River basin, China Zhanling Li, Quanxi Shao, Qingyun Tian and Louie Zhang 882 Effects of antecedent soil water content on infiltration and erosion processes on loessial slopes under simulated rainfall Lan Ma, Junyou Li and Jingjing Liu 894 Assessment of hydrological drought based on nonstationary runoff data Xueli Sun, Zhanling Li and Qingyun Tian 911 Spatiotemporal distributions and ecological risk assessment of pharmaceuticals and personal care products in groundwater in North China Jin Wu, Jingchao Liu, Zenghui Pan, Boxin Wang and Dasheng Zhang 925 Impacts of bias nonstationarity of climate model outputs on hydrological simulations Yu Hui, Yuni Xu, Jie Chen, Chong-Yu Xu and Hua Chen 942 Drought prediction models driven by meteorological and remote sensing data in Guanzhong Area, China Jianzhu Li, Siyao Zhang, Lingmei Huang, Ting Zhang and Ping Feng
ii
Contents
Hydrology Research
|
51.5
|
2020
959 The influences of sponge city construction on spring discharge in Jinan city of China Kangning Sun, Litang Hu and Xiaomeng Liu 976 Quantification of climate change and land cover/use transition impacts on runoff variations in the upper Hailar Basin, NE China Yuhui Yan, Baolin Xue, Yinglan A, Wenchao Sun and Hanwen Zhang 994 Analysis for spatial-temporal matching pattern between water and land resources in Central Asia Ying Zhang, Zhengxiao Yan, Jinxi Song, Anlei Wei, Haotian Sun and Dandong Cheng 1009 Different runoff patterns determined by stable isotopes and multi-time runoff responses to precipitation in a seasonal frost area: a case study in the Songhua River basin, northeast China Jie Li, Wei Dai, Yang Sun, Yihui Li, Guoqiang Wang and Yuanzheng Zhai 1023 Water balance changes in response to climate change in the upper Hailar River Basin, China Junfang Liu, Baolin Xue, Yinglan A, Wenchao Sun and Qingchun Guo 1036 The spatial pattern of periphytic algae communities and its corresponding mechanism to environmental variables in the Weihe River Basin, China Yixin Liu, Jiaxu Fu, Dandong Cheng, Qidong Lin, Ping Su, Xinxin Wang and Haotian Sun 1048 Spatiotemporal variation and tendency analysis on rainfall erosivity in the Loess Plateau of China Yongsheng Cui, Chengzhong Pan, Chunlei Liu, Mingjie Luo and Yahui Guo 1063 Potential impact of water transfer policy implementation on lake eutrophication on the Shandong Peninsula: a difference-in-differences approach Jia He, Jiping Yao, Aihua Li, Zhongxin Tan, Gang Xie, Huijian Shi, Xuan Zhang, Wenchao Sun and Peng Du 1077 Succession of phytoplankton in a shallow lake under the alternating influence of runoff and reverse water transfer Qing Li, Guoqiang Wang, Zhongxin Tan and Hongqi Wang 1091 A framework for event-based flood scaling analysis by hydrological modeling in data-scarce regions Jianzhu Li, Kun Lei, Ting Zhang, Wei Zhong, Aiqing Kang, Qiushuang Ma and Ping Feng 1104 Response of redox zonation to recharge in a riverbank filtration system: a case study of the Second Songhua river, NE China Xiaosi Su, Yaoxuan Chen, Hang Lyu, Yakun Shi, Yuyu Wan and Yiwu Zhang 1120 Coincidence probability of streamflow in water resources area, water receiving area and impacted area: implications for water supply risk and potential impact of water transfer Xingchen Wei, Hongbo Zhang, Vijay P. Singh, Chiheng Dang, Shuting Shao and Yanrui Wu 1136 The exploration of a Temporal Convolutional Network combined with Encoder-Decoder framework for runoff forecasting Kangling Lin, Sheng Sheng, Yanlai Zhou, Feng Liu, Zhiyu Li, Hua Chen, Chong-Yu Xu, Jie Chen and Shenglian Guo
iii
Contents
Hydrology Research
|
51.5
|
2020
1150 Impact of urbanization on variability of annual and flood season precipitation in a typical city of North China Peijun Li, Depeng Zuo, Zongxue Xu, Xiaoxi Gao, Dingzhi Peng, Guangyuan Kan, Wenchao Sun, Bo Pang and Hong Yang 1170 Effect of initial plant density on modeling accuracy of the revised sparse Gash model: a case study of Pinus tabuliformis plantations in northern China Yiran Li, Xiaohua Liu, Chuanjie Zhang, Zedong Li, Ye Zhao and Yong Niu 1184 Entropy weight method coupled with an improved DRASTIC model to evaluate the special vulnerability of groundwater in Songnen Plain, Northeastern China Bin Wang, Yanguo Teng, Huiqun Wang, Rui Zuo, Yuanzheng Zhai, Weifeng Yue and Jie Yang 1201 A field investigation on rill development and flow hydrodynamics under different upslope inflow and slope gradient conditions Pei Tian, Chengzhong Pan, Xinyi Xu, Tieniu Wu, Tiantian Yang and Lujun Zhang
833
© 2020 The Author Hydrology Research
|
51.5
|
2020
Editorial: Advances in Eco-hydrology and Watershed Water Resources Management in China
The past decades have witnessed China’s rapid economic
research and management of hydrology, water resources,
development, urbanization and the improvement of people’s
and the water eco-environment. This Special Issue is
well-being. The water shortages caused by population
composed of 25 original articles contributed by the
growth, water pollution and/or climate extremes (drought
participants
and flood), as well as the degradation of aquatic ecosystems,
achievements in the field of ‘Advances in Eco-hydrology
who
officially
presented
their academic
place pressure on the social development of the whole
and Watershed Water Resources Management in China’.
country. In order to fuel sustainable development, it is
These articles consist of various topics including climate
necessary and urgent to introduce feasible strategies and
change influences on hydrological processes, evaluation and
policies concerning effective water resources management.
predication of drought and flooding, water pollution control
Therefore, the China Water Forum gathered professional
(surface and ground water), water security and water
brains to motivate sparkling ideas which are recorded in
resources shortages, as well as water environmental man-
this Special Issue.
agement modeling by process-based and deep learning
The China Water Forum, organized by the China
methods. The results and conclusions are expected to
Society of Natural Resources, is one of the most useful
guide global water quantity-quality-ecology management
platforms for scientists to discuss current water-related
and climate change adaptation.
problems in China. The 17th China Water Forum hosted by Beijing Normal University (BNU), China, was held on
Guest Editor
8‒10 November 2019. Over 1,000 professors, engineers
Guoqiang Wang
and graduate students from more than 80 domestic univer-
College of Water Sciences, Beijing Normal University,
sities/institutes attended this forum, who are related to the
Beijing 100875, China
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.200
834
© 2020 The Authors Hydrology Research
|
51.5
|
2020
An improved routing algorithm for a large-scale distributed hydrological model with consideration of underlying surface impact Jingjing Li, Haoyuan Zhao, Jun Zhang, Hua Chen, Chong-Yu Xu
, Lu Li,
Jie Chen and Shenglian Guo
ABSTRACT Large-scale hydrological models are important tools for simulating the hydrological effect of climate change. As an indispensable part of the application of distributed hydrological models, large-scale flow routing methods can simulate not only the discharge at the outlet but also the temporal and spatial distribution of flow. The aggregated network-response function (NRF), as a scale-independent routing method, has been tested in many basins and demonstrated to have good runoff simulation performance. However, it had a poor performance and produced an unreasonable travel time when it was applied to certain basins due to a lack of consideration of the influence of the underlying surface. In this study, we improve the NRF routing method by combining it with a velocity function using a new routing parameter b to reflect the wave velocity’s sensitivity to slope. The proposed method was tested in 15 catchments at the Yangtze River basin. The results show that it can provide better daily runoff simulation performance than the original routing model and the calibrated travel times in all catchments are more reasonable. Therefore, our proposed routing method is effective and has great potential to be applied to other basins. Key words
| DEM-based routing method, distributed hydrological model, NRF routing method, velocity function, WASMOD, Yangtze River basin
HIGHLIGHTS
• • • • •
This study coupled the aggregated network-response function (NRF) flow routing method with a velocity function. The improved NRF method reflects the sensitive wave velocity to slope and gets better runoff simulation performance. The calibrated travel time is closer to benchmark value after improvement. The improved NRF method adapts to basins of various underlying surfaces. The improved NRF method gets more reasonable wave velocity after improvement.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.170
Jingjing Li Haoyuan Zhao Hua Chen (corresponding author) Jie Chen Shenglian Guo State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University, Wuhan, China E-mail: chua@whu.edu.cn Jun Zhang Bureau of Hydrology, Changjiang Water Resources Commission, Wuhan, China Chong-Yu Xu Department of Geosciences, University of Oslo, Oslo, Norway Lu Li NORCE Norwegian Research Centre, Bjerknes Centre for Climate Research, Jahnebakken 5, 5007 Bergen, Norway
835
J. Li et al.
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
INTRODUCTION Climate change has received widespread attention from the
water to the total land area can be calculated based on the
scientific community and the public, as it has caused a series
water storage in each cell, it will contribute to land–atmosinteraction
simulation
(Olivera
).
of serious problems, including natural disasters and extreme
phere
climate events (Kharin et al. ; Chen et al. ; Lu & Qin
However, the choice of a proper spatial resolution is a
; Ragettli et al. ). As tools for estimating regional
dilemma for the cell-to-cell routing method. In distributed
and global water resources and predicting the hydrological
hydrological models, high resolution can reflect the hetero-
response
et
al.
hydrological
geneity of the hydrological characteristics of the basin and
models have become a hot topic recently (Müller Schmied
improve the accuracy of calibrated parameters. On the
& Döll ; Wang et al. ; Abou Rafee et al. ; Li
other hand, higher-resolution DEM will lead to a significant
et al. ). A large-scale routing method is an important
increase in the amount of required computations, especially
part of a large-scale hydrological model, with which a dis-
when coupled with a meteorological model because the
tributed hydrological model can simulate the change in
meteorological model has vast computational cost. However,
to
climate change,
large-scale
runoff at temporal and spatial scales (Beven ). Many
a coarser resolution may lead to a decrease in model perform-
routing methods based on a digital elevation model (DEM)
ance due to not considering routing within the cell (Arora
have been developed (Wen et al. ; Li et al. ; Lu
et al. ; Gong et al. ). Besides this, there may be a
et al. ; Huang & Lee ; Ling et al. ; Fan et al.
great difference between the network that is generated at a
). The main steps of runoff routing in distributed hydro-
coarse resolution and the true one, which is another disad-
logical models are (1) extracting the flow network and (2)
vantage of the cell-to-cell routing method. Attempts have
calculating the runoff routing according to the flow path
been made to bring a network that is generated at a coarse
(Wen et al. ). Generally, runoff routing based on a
resolution closer to the real one. However, a gap still exists
DEM is calculated in one of two ways (Olivera et al. ):
(Fekete et al. ; Guo et al. ).
cell-to-cell and source-to-sink. The cell-to-cell method calcu-
Some of the source-to-sink routing methods have the
lates the water movements from each grid to its adjacent
same disadvantages as the cell-to-cell method, including
downstream grid until the outlet grid of the basin is reached.
the routing models by Ducharne et al. () and Guo
The outflow of each cell is calculated based on the inflow and
et al. (), which will provide worse performance as the
the river channel’s storage. The source-to-sink method, under
cell resolution becomes coarser. The routing model by
the assumption that the water is rigid, directly calculates the
Ducharne et al. () uses the mean travel time of each
water’s movement from the grid where runoff is generated to
large routing cell, so the water in a cell will arrive at the
the outlet of the basin.
outlet in the same day. This can be unreasonable, especially
The first of these two kinds of routing methods is based
when the cell size is large. The routing model by Guo et al.
on the mass conservation equation and the channel storage
() uses the mean flow path length for each cell, making
function equation. It is widely used due to its simplicity
the routing method scale-dependent. Therefore, it would
(Miller et al. ; Sausen et al. ; Arora & Boer ;
benefit a hydrological modeler to find a routing method
Huang & Lee ). In addition to its simplicity, the cell-
with the capability to be scale-independent, which will pre-
to-cell routing method has other advantages. It has the
serve the spatially distributed travel time information from
potential to include the nonlinear losses from the surface
a finer-resolution DEM for coarser resolution cells in the
water to the groundwater because it considers the feedback
large-scale hydrological model. The disadvantages of some
between the flow and the water storage in each cell (Naden
source-to-sink routing methods have been overcome by
et al. ). Besides this, with the cell-to-cell routing method,
some model developers. For instance, Wen et al. () pro-
the water storage in any cell of interest can be queried at
posed a multiscale routing framework that can reduce the
each time step. If the ratio of land area covered by surface
impact of the spatial scale by using histograms for the flow
836
J. Li et al.
|
The improvement study of the NRF flow routing method
path lengths. The aggregated network-response function
Hydrology Research
|
51.5
|
2020
STUDY AREA AND DATA
(NRF) routing method (Gong et al. ) can preserve all of the spatially distributed travel time information in
Study area
high-resolution Hydro1 k DEM data for cells with different resolutions by aggregating the pixel-response function to
The study area is the Yangtze River basin (24 270 –35 540 N,
the cell-response function (CRF). Thus, NRF has the
90 330 –122 190 E). It has a drainage area of 1,800,000 km2.
unique advantage of being able to obtain a stable result
As shown in Figure 1, the Yangtze River is the longest river
when moving from a finer to a coarser resolution, making
in China and the third longest river in the world. It has a
it a preferable routing method for large-scale hydrological
total length of 6,397 km.
models. Another advantage of source-to-sink methods is
The climate, terrain, land cover and land use, and runoff
the computation efficiency (Naden et al. ; Olivera
characteristics vary greatly among the catchments in the
et al. ; Gong et al. ), which enables them to be
Yangtze River basin. Fifteen catchments in the Yangtze
widely applied (Olivera et al. ; Ducharne et al. ;
River basin were selected in this study in order to evaluate
Gong et al. ; Lu et al. ).
the performance of the improved routing method over a
NRF, as a source-to-sink routing method, has been
wide range of representative catchments. The characteristics
applied in basins of China, North America, and Africa
of the terrain in those catchments are shown in Table 1.
(Gong et al. , ; Li et al. , , ; Ngongondo
‘Elevation difference’ means the height difference between
et al. ). The results showed that NRF provides good daily
the highest point and the lowest point in a catchment.
runoff simulation performance. However, in some cases, the
Table 1 also presents the shape factor and the river length
calibrated wave velocity is too large to be physically realistic,
of the catchments. The basin shape factor (Morisawa )
and NRF will fail to provide good daily runoff simulation
is the ratio of the actual basin length to the perimeter of a
performance if the wave velocity becomes smaller (Gong
circle whose area is the same as that of the basin, and
et al. ; Li et al. ). In this method, the underlying fac-
measures the general runoff concentration behavior of the
tors that affect the routing process are not comprehensively
basin. Only the first-order stream, second-order stream,
considered. In this respect, the NRF routing method needs
and third-order stream (Wang et al. ) were included in
to be improved.
the stream length in this study. A drainage map of the
This study aims to improve NRF by considering the
river system and drainage stations can be found in Figure 1.
influence of the underlying surface in order to obtain a
The Yangtze River basin has very complex terrain, as it
more reasonable wave velocity and travel time and better
flows through mountains, plateaus, hills, and plains. The
daily runoff simulation performance. To achieve this aim,
overall distribution of elevation at the Yangtze River is
the original NRF routing method is coupled with a velocity
high in the west and low in the east. The mean elevation
function proposed by Sircar et al. () in order to take the
of the source of the Yangtze River, which is located in the
underlying surface’s influence into consideration, in which a
hinterland of the Qinghai–Tibet Plateau, is more than
new routing parameter b is added to reflect how sensitive
4,000 m above sea level (a.s.l.). The middle reaches lie
the wave velocity is to slope. The remainder of this paper
mostly in low mountains areas and the downstream lies
is organized as follows. Following a description of the
mostly in a hilly plain area.
study area, we present the original and improved NRF rout-
The mean annual precipitation is 1,110 mm. Due to the
ing methods. Then, the results of a daily runoff and peak
vast amount of territory, the complex terrain, and the typical
flow simulation by the model before and after the improve-
monsoon climate, the spatial distribution and the temporal
ment
routing
distribution of annual precipitation in the Yangtze River
parameters and the factors that influence them are analyzed
are very uneven (Xu et al. ). The main source of river
and compared in different catchments to explore the
runoff in the basin is the rainfall, but snowmelt and ground-
relationship between catchment characteristics and model
water are also important sources of runoff in the upper
parameters. Finally, we present our conclusions.
reaches of the Yangtze River basin (Yao et al. ). The
are
compared.
The
travel
time
and
837
Figure 1
J. Li et al.
|
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
Location of the sub-basins and discharge stations in the Yangtze River basin.
annual mean air temperature in the Yangtze River basin is
Geological Survey’s Earth Resources Observing System.
generally high in the east, low in the west, high in the
The projection of HYDRO1 k is Lambert azimuthal equal-
south, and low in the north. The source area of the basin
area projection, whose transformation equations (Weisstein)
experiences the lowest temperatures (Su et al. ).
enable each grid to have an equal area.
It is of great significance to simulate the runoff in the
Flow direction and flow accumulation databases based
Yangtze River basin, as it accounts for 18% of the total
on Hydro1 k are also available, which can be used with
area of China, and has more than 400 million people
the discharge stations to obtain the basin’s boundary.
living in it. The Yangtze River basin is also the most impor-
Then, the estimated basin boundary can be evaluated and
tant agricultural production base in China because of its
corrected based on the basin boundary that was delineated
warm and humid climate, which provides good conditions
using the finer DEM at the resolution of 90 × 90 m.
for crop growth (Tian et al. ).
Daily precipitation, temperature, and relative humidity data for the period 1969–2008 were taken from the daily data-
Data and processing
set of surface climatological data of China (V 3.0) in National Meteorological Information Center (http://data.cma.cn/
Hydro1 K (USGS ) was used in this study to delineate
data/cdcdetail/dataCode/SURF_CLI_CHN_MUL_DAY_V3.
the basin’s boundary. It is a global DEM with a resolution
0.html). The meteorological data were interpolated using the
of 1 × 1 km and was produced by the United States
inverse distance weighted method (Shepard ) to be
J. Li et al.
838
Table 1
|
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
Characteristics of the study area
Mean discharge (m3/s)
Area (km2)
Jinsha River basin Downstream of Shigu Pingshan Upstream of Shigu Shigu
4,394.49 1,275.61
Mintuo River basin
Fushun Gaochang
Jialing River basin Jialing River Wu River basin
Wujiang
Han River basin Dongting Lake basin Poyang Lake basin
Sub-basin
Mean Slope elevation (m)
Elevation difference (m)
River length (km)
252,217 3.10 213,003 3.66
7.45 3,042 4.87 4,569
5,620 4,124
5,430 3,559
360.91 2,750.02
23,096 2.39 134,622 2.47
1.98 590 8.88 2,949
4,523 7,087
724 2,924
Beibei
1,841.34
157,478 2.44
6.29 1,148
2,948
3,124
Wulong
1,415.63
80,099
2.71
3.91 1,104
2,628
1,237
Hanjiang
Baihe
686.30
54,085
2.06
6.29 1,148
2,948
1,462
Xiangjiang Yuanjiang Lishui
Xiangtan Taoyuan Shimen
1,990.02 2,028.19 456.86
78,659 86,172 15,118
2.19 2.55 1.96
3.02 338 3.32 622 5.12 722
1,857 2,378 2,204
2,200 2,210 332
Gangjiagn Fuhe Xinjiang Leanhe Changhe
Waizhou Lijiadu Meigang Shizhenjie Dufengkeng
2,032.18 376.80 557.69 296.61 155.62
83,532 16,023 15,592 8,294 4,938
2.27 2.04 2.08 1.78 1.73
2.46 2.33 3.17 2.74 3.01
2,026 1,415 2,040 1,421 1,361
1,454 432 323 224 156
River
Tuojiang Minjiang
Station
Shape factor
287 227 279 205 246
gridded data, which were used to drive the hydrological
described in Gong et al. (), the main components of
model at resolutions of 3, and 10 arc-minutes in the study.
WASMOD are the computation of snow accumulation and
The observed daily discharge during 1964–2008 of all catch-
melting, evapotranspiration, slow flow, fast flow, and the
ments, which was used to calibrate and validate the model,
total runoff. Compared with the equations in Gong et al.
was provided by the Bureau of Hydrology of the Changjiang
(), we set two more parameters in this study: c5 is set
Water Resources Commission.
to correct the precipitation bias, which results from the
The travel time of a flood between major hydrological
lack of snow measurements; c4 is set in the potential evapor-
stations in the Yangtze River basin was taken from Zhang
ation equation. As shown in Table 2, there are five to seven
et al. (), which was calculated based on a great deal of
parameters with or without the snow module. The snow
data and some proven methods. In this study, it was used
modules are only applied to Jialing River, Jingsha River,
as a benchmark to compare the travel time calculated by
and Mintuo River where snowmelting is one of the most
the original routing method with the travel time calculated
important sources of runoff. If the snow module is not
by the improved routing method.
taken into consideration, parameters a1 and a2 will not be
METHODS
Table 2
|
The parameters of the runoff-generation model and their prior range for calibration
Water and Snow Balance Modeling The Water and Snow Balance Modeling (WASMOD) system is a conceptual modeling system that was developed by Xu (). Different versions of WASMOD have been used in different regions of the world (Widén-Nilsson et al. ; Gong et al. ; Kizza et al. ; Li et al. ; Yang et al. , , ). In this study, the daily version of WASMOD was used as a runoff-generation model. As
Parameter
About
Prior range for calibration
a1 ( C)
Snowfall
[0 to 6]
a2 ( C)
Snowmelt
[ 6 to 0]
a4 (–)
Actual evaporation
[0.1 to 0.999]
c1 (1/mm)
Fast runoff
[0 to 0.1]
c2 (1/mm)
Slow runoff
[0 to 0.1]
c4 (mm day 1 C 2) Potential evaporation [6 × 10 6 to 6 × 10 5] c5 (–)
Precipitation
[0.5 to 1.5]
839
J. Li et al.
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
For a given grid, the travel time can be calculated as
calibrated in the hydrological model.
follows: pt ¼ c 5 ptd
(1)
where ptd is precipitation data described in the Study area and data section and pt is the corrected precipitation. pett ¼ c4 ( max (tmpt þ 25, 0))2 (100 rht )
8 k P > > > <
li
0:5 i¼1 v45 (tan (ci )) ti ¼ k > P li > > : 0:5 i¼1 v45 (tan (c0 ))
ci > c0 (5) ci c0
(2) where ci is the slope, vi is the wave velocity, ti is the travel
where pett is the potential evaporation, tmpt is the temperature, and rht is the relative humidity.
time of the grid, c0 is the slope threshold, which is set to be a constant to prevent ti from being infinite, and k is the number of grids through the flow path. The slope threshold is 0.1 in this study.
The original routing method
Other details about the NRF routing methods can be Gong et al. () developed a routing method, named the aggregated NRF, which was developed with the source-tosink concept. It builds the relationship between each grid in DEM and the outlet grid. There are five steps in the improved NRF routing method: (1) extract the flow direction and flow net based on DEM; (2) calculate the travel time ti between each grid and the outlet of the basin according to the flow
seen in Gong et al. (). In this study, the computation cell size in each catchment was determined by its area to ensure a modest number of cells. The cell size in a larger catchment is larger. Conversely, the cell size is smaller in a smaller catchment. The cell size was set to be 100 at catchments whose area is greater than 50,000 km2. It was set to be 30 at catchments whose area is less than 10,000 km2 and 60 otherwise.
path; (3) construct the pixel-response function (PRF) based on ti; (4) calculate CRF based on PRF with the way of linear averaging method; and (5) calculate the discharge
The improved routing method
at the outlet of the basin based on the result of the runoffgeneration model and CRF. The main advantage of NRF
The NRF routing method performed well in the Dongjiang
is that the routing accuracy is independent of cell size,
basin and the Willamette basin; however, the calibrated par-
because the CRF in NRF is not affected by the cell size
ameter (v45 ) in the Hydro1 k-driven NRF routing method
(Gong et al. ).
showed an unrealistic value (greater than 25 m/s) (Gong
For the wave velocity in grid, Beven & Kirkby ()
et al. , ). To improve the NRF, we combine the orig-
proposed that the overland flow velocity could be calculated
inal routing method with a wave function given by Sircar
based on the topography as follows:
et al. () and modified the equation as:
vi ¼ v45 tan (ci )
(3)
vi ¼ v45 ( tan (ci ))b
where vi is the wave velocity of the grid, and ci is the slope of the grid. v45 is the wave velocity in the grid whose slope is
45 . In the study of Gong et al. (), the authors calculate the wave velocity as follows: vi ¼ v45
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi tan (ci )
ti ¼
8 k P > > > <
i¼1 v45
k > P > > :
i¼1 v45
li (tan (ci ))b li (tan (c0 ))b
(6) ci > c0 (7) ci c0
where b is a parameter that reflects how sensitive the wave velocity is to slope. The value of the parameter b is related to (4)
the condition of the underlying surface of the catchment.
J. Li et al.
840
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
Compared with Equation (4), Equation (6) is more gen-
To measure the peak flow simulation performance, the
eral and more physically realistic. Because different basins
time lag and relative error of peak flow were also calculated.
have different underlying conditions, it cannot be known how sensitive the wave velocity is to slope. In this case, b varies in different basins. Equation (6) reflects the relationship between wave velocity and slope more flexibly and can more easily reflect the characteristics of different basins. Thus, Equation (5) is substituted with Equation (7) in the improved NRF routing method. There are two parameters in the improved NRF routing method (referred to as ‘after’ in the following) and one parameter in the original NRF routing method (referred to as ‘before’ in the following). Table 3 shows the routing parameter value sets for calibration. They were chosen based on computer capability limitations and the physical meaning of each parameter. There are 7 routing parameter value sets before and 42 sets after, composed of six kinds of parameter
Qre ¼
ΔT ¼ Tsim Tobs
warm-up period was 1 year for all catchments. The Nash–Sutcliffe (NSE) efficiency (Nash & Sutcliffe ) was used as the criterion for evaluating overall runoff simulation performance in this study.
the simulated maximum peak flow, Qobs max is the observed and Tobs is the day when Qobs max appears. ΔT is the time lag of the simulated peak flow relative to the observed peak flow. Qre and ΔT of all selected flood events in a catchment were averaged to reflect the overall simulation performance of the peak flow in that catchment as follows:
Qre ¼
l 1X jQre,j j l j¼1
(11)
ΔT ¼
l 1X jΔTj j l j¼1
(12)
where ΔT is the averaged time lag of the peak flow in the catchment, Qre is the averaged relative error of the peak flow in the catchment, and l is the number of flood events in the catchment. The parameters of the routing method were calibrated
Pn
NSE ¼ 1
(10)
maximum peak flow, Tsim is the day when Qsim max appears,
Evaluation criteria
ment was calibrated and validated in all catchments. The
(9)
where Qre is the relative error of the peak flow, Qsim max is
b and seven kinds of parameter v45 .
The model before and after the routing method improve-
Qsim max Qobs max Qobs max
2 i¼1 (Oi Si ) Pn 2 i¼1 (Oi Oi )
(8)
using the Monte Carlo algorithm (Barraquand & Latombe ), and the parameters of the runoff-generation model were calibrated using Covariance Matrix Adaption Evol-
where Oi is the observed runoff, Si is the simulated runoff,
ution (CMAES) (Hansen & Ros ). The Monte Carlo
and n is the length of the time step.
algorithm was used to search for the best routing parameter
Table 3
|
The parameters of the routing methods and their prior range for calibration
set. First, 300 runoff-generation parameter value sets were produced using LATIN-Hypercube sampling (Mckay et al.
Parameter
Explanation
Values
) according to the prior range of the parameters.
Before
v45 (m/s)
The wave velocity of a grid whose slope is 45
4, 5, 6, 7, 8, 9, 10
Table 3 shows the routing parameter sets. Second, for
After
b (–)
0.2, 0.25, 0.3, Power exponent reflecting 0.35, 0.4, how sensitive is the v to 0.45 slope v45 (m/s) The wave velocity of a grid 4, 5, 6, 7, 8, 9, whose slope is 45 10 The combination of all values yields 42 parameter sets
ameter value sets were used to drive the hydrological model,
each routing parameter set, all of the runoff-generation parand then the NSE was calculated based on the simulated runoff and the observed runoff. Third, the best routing parameters were selected based on the NSE. Then, the Covariance Matrix Adaption Evolution algorithm was used
841
J. Li et al.
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
to search for the best runoff-generation parameters under
period of the 15 catchments is 0.11, with a maximum of
the condition that the routing parameters are fixed.
0.44 in the calibration period and 0.43 in the validation period. From Figure 2, it can also be found that the improvement in NSE varies greatly in the 15 catchments of the
RESULTS AND DISCUSSION
Yangtze River basin after the improvement of the routing method.
Modeling performance assessment with the improved
In order to demonstrate the difference in the simulated and observed discharge hydrographs of the original and
NRF
improved routing methods, daily hydrographs of a represenBoth routing methods were taken as the routing module of
tative year 1968 for the Shimen catchment are plotted in
the WASMOD. To all catchments, the model was calibrated
Figure 3. It shows that with the original NRF routing
by using 7 years’ data series and validated by using another 4
method, the simulated flood peak is often later than the
years’ data series during the period of 1964–2008. Figure 2
observed flood peak. The improved NRF can better simulate
shows the performance of the daily runoff simulation
the flood peak. Tables 4 and 5 show the optimal parameters
before and after the improvement.
of the hydrological model with the original and the
Figure 2 shows that the model with the improved NRF
improved routing method.
routing method performs better than the model with the
In order to assess the flood simulation performance of
original NRF routing method for all of the study catchments
the model with the improved NRF routing method, the aver-
to varying extents. Taking the Baihe catchment and the
aged relative errors Qre and averaged time lags ΔT of all
Waizhou catchment as examples, the Baihe catchment has
flood events in each catchment were calculated using
an NSE of 0.87 in the calibration period and 0.74 in the vali-
Equations (11) and (12) and are shown in Figure 4. The
dation period after the improvement, while the NSE is 0.43
flood events whose peak flow was greater than three times
in the calibration period and 0.31 in the validation period
the average runoff of catchment were selected in this
before the improvement. The Waizhou catchment has an
study. It can be seen that the model with the improved
NSE of 0.82 in the calibration period and 0.78 in the vali-
NRF routing method provides better flood simulation per-
dation period after the improvement, while the NSE is
formance than the model with the original NRF routing
0.58 in the calibration period and 0.59 in the validation
method. After the improvement, the averaged time lags of
period before the improvement. The mean improvement in
the peak flow in almost all catchments are less than one
the NSE in both the calibration period and the validation
day, except for the Shigu and Waizhou catchments (Figure 4,
Figure 2
|
NSE in (a) the calibration period and (b) the validation period in all catchments by the NRF routing method before and after the improvement.
J. Li et al.
842
Figure 3
Table 4
|
|
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
The simulated and observed discharge hydrographs for the Shimen catchment in 1968 by the NRF routing method before and after the improvement.
The optimal parameter values of the hydrological model with the original routing method
Catchment
v45
b
a1
a2
a4
c1
c2
c4
c5
Pingshan
0.5
6
6.00
0.00
0.65
0.0075
0.00005
0.10
0.58
Shigu
0.5
10
6.00
0.00
0.99
0.0124
0.00008
0.27
0.50
Fushun
0.5
10
6.00
1.83
0.93
0.0092
0.00055
0.35
0.78
Gaochang
0.5
10
5.78
1.51
1.00
0.0064
0.00018
0.10
0.64
Beibei
0.5
10
0.05
5.00
0.81
0.0145
0.00017
0.12
0.59
Wulong
0.5
9
—
—
0.88
0.0087
0.00022
0.41
0.91
Baihe
0.5
10
—
—
0.71
0.1000
0.00008
0.15
0.69
Xiangtan
0.5
10
—
—
0.95
0.0015
0.00002
0.66
1.11
Taoyuan
0.5
10
—
—
0.94
0.0040
0.00007
1.00
1.14
Shimen
0.5
9
—
—
0.88
0.0879
0.00575
1.00
0.97
Waizhou
0.5
10
—
—
0.96
0.0088
0.00003
0.26
0.71
Lijiadu
0.5
10
—
—
0.96
0.0017
0.00001
1.00
1.10
Meigang
0.5
10
—
—
0.96
0.0026
0.00001
1.00
1.20
Shizhenjie
0.5
10
—
—
0.93
0.0010
0.00001
1.00
1.36
Dufengkeng
0.5
10
—
—
0.94
0.0089
0.00046
1.00
0.90
left panel). The averaged relative error of the peak flow in all
distributions histogram of time lags ΔT in each catchment
catchments decreases in nearly all of the catchments
with the original and improved NRF routing method. Fre-
(Figure 4, right panel). The amelioration is most evident in
quency is the number of flood events that ΔT in a certain
the Waizhou catchment, where Qre equals 0.32 with the
value is divided by the total number of flood events in this
original routing method and 0.17 with the improved routing
catchment. When the time lag ΔT is positive, it means that
method. The relative error of the peak flow decreases to the
the simulated flood peak is later than the actual one. On
minimum error range of 0.1 to 0.1 after the improvement
the contrary, when the time lag ΔT is negative, it means
in the routing method in all of the catchments except for
that the simulated flood peak is earlier than the actual
the Dufengkeng and Fushun catchments.
one. When ΔT equals zero, the hydrological model simulates
To evaluate the ability of simulating the occurrence of
the time of flood peak accurately. Figure 5 shows that with
flood peaks, the time lags ΔT in all selected flood events in
the original routing method, the frequency of positive time
each catchment were analyzed. Figure 5 plots the frequency
lag ΔT is larger than the frequency of negative time lag ΔT.
J. Li et al.
843
Table 5
|
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
The optimal parameter values of the hydrological model with the improved routing method
Catchment
v45
b
a1
a2
a4
c1
c2
c4
c5
Pingshan
0.2
5
6.00
0.00
0.47
0.0038
0.00015
0.11
0.58
Shigu
0.45
8
6.00
0.00
0.99
0.0125
0.00012
0.26
0.50
Fushun
0.45
10
0.99
2.32
0.12
0.0148
0.00224
0.14
0.66
Gaochang
0.2
5
2.88
6.00
1.00
0.0059
0.00028
0.91
0.64
Beibei
0.3
7
5.45
1.81
0.50
0.0120
0.00054
0.27
0.74
Wulong
0.45
8
—
—
0.88
0.0082
0.00026
0.40
0.90
Baihe
0.3
10
—
—
0.81
0.0090
0.00051
1.00
1.14
Xiangtan
0.25
4
—
—
0.22
0.0039
0.00012
0.21
0.79
Taoyuan
0.2
4
—
—
0.94
0.0040
0.00019
1.00
1.09
Shimen
0.25
8
—
—
0.91
0.0123
0.00080
1.00
1.19
Waizhou
0.35
8
—
—
0.97
0.0038
0.00007
0.91
0.89
Lijiadu
0.4
9
—
—
0.94
0.0044
0.00015
1.00
0.91
Meigang
0.4
9
—
—
0.94
0.0081
0.00043
1.00
0.96
Shizhenjie
0.25
4
—
—
0.95
0.0078
0.00085
1.00
0.89
Dufengkeng
0.45
9
—
—
0.76
0.0124
0.00104
0.24
0.78
Figure 4
|
The averaged time lag of peak flow (left panel) and relative error of peak flow (right panel) in all catchments by the NRF routing method before and after the improvement.
That indicates that the simulated peak flow is often delayed
model with the original routing method. Those phenomena
compared with the observed peak flow with the original
indicate that the improved routing methods can better serve
NRF routing method. There is a higher frequency of zero
the flood simulation.
time lag of peak flow (ΔT equals zero) in most of the catch-
These results indicate that the model with the improved
ments with the improved routing method than that with the
routing method performs better in both peak flow simu-
original routing method. Particularly, in the Gaochang,
lation and long-term runoff simulation than the model
Shizhengjie, and Meigang catchments, the frequency of
with the original routing method. Besides this, NRF’s advan-
zero time lag of peak flow is greater than 0.8 in the model
tage of being able to preserve the travel time information in
with the improved routing method and less than 0.4 in the
the PRF is maintained in the improved model.
844
Figure 5
J. Li et al.
|
|
The improvement study of the NRF flow routing method
The frequency distributions of time lags (day) for all of the catchments by the NRF routing method before and after the improvement.
Hydrology Research
|
51.5
|
2020
845
J. Li et al.
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
Analysis of the routing time calculated by the improved
by the original routing method and the improved routing
NRF
method. The runoff travel times were compared with the travel time provided by Zhang et al. () as a benchmark.
Figure 6(a) and 6(b) show the distributions of the travel time
For most catchments, the travel time between grids that
of all grids in all catchments that were calculated by the orig-
was calculated by the improved routing method was shorter
inal and the improved routing methods. Figure 6(d) plots the
than that calculated by the original routing method
mean grid travel time and mean slope of all of the catchments.
(Figure 6(a) and 6(b)). Besides this, the travel time between
The simulated travel time was evaluated based on two hydro-
the specific upstream hydrological station and the outlet
logical stations in each catchment: the outlet station and the
station was shorter and closer to the benchmark value for
specific upstream hydrological station that made the distance
all the catchments (Figure 6(c)). The main reason why the
between the two hydrological stations as large as possible.
travel time that was calculated by the improved routing
Figure 6(c) plots the runoff travel times between the upstream
method was longer than the benchmark value in some
hydrological station and the outlet station that were calculated
catchments is that the routing parameters v45 and b are
Figure 6
|
The evaluation of the travel times. (a) Box plots of the travel time in all catchments calculated by the model with the NRF routing method before the improvement and (b) after the improvement; (c) the averaged travel time in all of the catchments simulated by the model before and after the improvement compared with observations.
846
J. Li et al.
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
spatially constant over the whole catchment. However, rout-
the runoff travel times in these catchments, as calculated
ing refers to the process in which rainfall in a catchment
using the original routing method, were unrealistic and at
evolves along a certain path into discharge at the outlet.
least twice as long as the runoff travel times that were calcu-
The rainfall is first routed by a hillslope and then routed
lated using the improved routing model. On the other hand,
by the river before being transferred into discharge. Hill-
for the Dufengkeng, Fushun, and Wulong catchments, we
slope routing and river routing are two forms of routing
found no obvious change in the travel times that were calcu-
(Kirkby & Beven ). Usually, the runoff travels faster in
lated using the original and improved routing methods. The
river routing than in hillslope routing. However, the hill-
original and improved routing methods also provided a simi-
slope travel time and the river travel time are not
lar NSE in these catchments.
separated in both the original and improved NRF routing methods. The parameters v45 and b reflect the state of the
Analysis of the wave velocity calculated by the
whole basin; however, the benchmark value reflects the
improved NRF
travel time that is needed in the selected river section. Thus, the travel time in the river that is calculated by the
Figure 7 shows the theoretical curves of wave velocity versus
improved routing method will be a little longer than the
slope. Figure 8(a) and 8(c) plot the calibrated curves that
benchmark value. However, the travel time that was calculated in the Shigu catchment is much longer than the benchmark value, which is due to the Shigu catchment’s underlying surface. The shape factor of the Shigu catchment is the largest of all of the studied catchments, which means that the shape of the Shigu catchment is narrow and long and the water system’s development is immature. The above factors lead to much more hillslope routing and a longer travel time in the Shigu catchment. In addition, the snowmelt and groundwater are the main sources of runoff in the Shigu catchment. The snowmelt runoff moves more slowly down the hillslope than the rainfall runoff, so the optimal routing parameters that were calibrated using the observed runoff lead to a longer travel time in the Shigu catchment compared with the benchmark value, which only considers routing in the river. The travel time that was calculated using the original routing method produces a late peak flow and a smaller amount of flow (Figure 4). An unreasonable travel time is one of the main causes of a hydrological model’s poor performance in long-term runoff simulation. The improved routing method provides a more realistic travel time between the grids compared with the benchmark value. When analyzed in conjunction with Figures 2 and 6, it can be found that the NSE and the travel time in all catchments have a close relationship. From Figure 2, it can be seen that the NSE that was calculated by the original NRF is lower in some catchments, i.e., the Waizhou, Shimen, Beibei, and Baihe catchments. As shown in Figure 6(c),
Figure 7
|
The theoretical curve of velocity and slope: (a) wave velocity changes with slope with the parameter v45 before the improvement and (b) wave velocity changes with slope with the two parameters v45 and b after the improvement.
847
J. Li et al.
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
were calculated by the original routing method and the
larger NSE value. Thus, the calibrated value of v45 reaches
improved routing method (Table 3), respectively. From the
the upper limit of the prior range for calibration in 13
analysis in the section Analysis of the routing time calcu-
catchments, resulting in only three kinds of curves between
lated by the improved NRF, it is known that the improved
slope and calibrated wave velocity in 15 catchments. More-
routing method is able to achieve a more reasonable travel
over, in many basins, the optimal value of v45 in cali2 is
time. The reason for this is that the improved routing
larger than 10 m/s and has a maximum of 31 m/s; thus,
method has a more flexible wave velocity–slope curve
the grid wave velocity is too large to be realistic
(Figure 7) that can better reflect the real world.
(Figure 8(e)). Figure 8(b), 8(d) and 8(f) show the distri-
It can be seen from Figure 7(b) that the wave velocity of
butions of the wave velocity of all grids in all catchments
a grid is determined by the parameters b and v45 together. If
before the improvement with the strict prior range of par-
the parameter b remains constant, the wave velocity
ameter v45 for calibration; after the improvement with
increases as v45 increases. If the parameter v45 remains con-
the strict prior range of parameter v45 for calibration;
stant, the wave velocity increases as b decreases at grids with
before the improvement with the wider prior range of par-
a slope of less than 45 ; however, the wave velocity
ameter v45 for calibration. It also shows that the grid wave
increases as b increases at grids with a slope of greater
velocity is too large to be realistic before the improvement
than 45 . The effect that slope has on the wave velocity
with the wider prior range of parameter v45 for calibration.
varies across catchments due to the differences in the
The model can provide good runoff simulation perform-
hydrology of the underlying surface, which can be adjusted
ance
by the parameter b. The bigger b is, the more sensitive the
compensates for the problem of the routing time being
wave velocity is to slope. When b is equal to zero, the
too long due to the unreasonable distribution of the
because
the
excessively
large
value
of
v45
slope has no effect on wave velocity, which will be a con-
grids’ wave velocity. This indicates that the original routing
stant. A larger wave velocity results in a shorter runoff
method cannot adapt to the variation in the underlying sur-
travel time in a catchment.
face of the different basins. This is consistent with the
Each routing parameter value set refers to a certain
results from Gong et al. (), who found that the optimal
curve between slope and calibrated wave velocity. The orig-
value of v45 was physically unrealistic as it was determined
inal NRF only produces three kinds of curves for the 15
to be greater than 26 m/s with the original Hydro1 k-driven
catchments (Figure 8(a)). The curves of 13 catchments
NRF routing method.
coincide because they have the same optimal value of
The original NRF routing method uses a constant (0.5)
v45 . However, the improved NRF produces a different cali-
to reflect the influence that the slope has on the wave vel-
brated curve for each catchment (Figure 8(c)). Tables 6 and
ocity; however, it fails to adapt to the variation in the
7 present the optimal values of v45 that were calibrated by
underlying surface of the different basins. The improved
the original routing method with a strict range of 4–10 m/s
routing method makes it possible to fit the actual physical
(cali1), as in Table 3, and with a wider prior range of
relationship between the slope and wave velocity in a
4–31 m/s (cali2), respectively. Due to computer capability
more flexible way. The reasonable distribution of wave vel-
limitations, the increment of parameter v45 above 10 m/s
ocity leads directly to a more accurate simulation of the
is 3 m/s. Figure 8(e) plots the calibrated curves between
travel time, which in turn leads to a better simulation of
slope and calibrated wave velocity in cali2. In combination
the runoff.
with the optimal value of v45 shown in Tables 6 and 7, it
The improved NRF routing method belongs to hydrolo-
can be seen that the optimal value of v45 in cali1 is the
gical methods rather than hydraulics methods. There are
same as 10 m/s, reaching the upper limit of the prior
some other hydrological routing methods calculating the
range for nearly all of the catchments. When the prior
travel time of each grid and then routing the runoff to dis-
range for calibration is fixed to be 4–10 m/s, the larger
charge. But the parameter to calibrate and influence
the value of v45 , the shorter the travel time, which is
factors that have been taken into consideration are different
closer to the reality in most catchments. This results in a
in different methods.
848
Figure 8
J. Li et al.
|
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
Curves and box plots of calibrated wave velocity versus slope (a) and (b) before the improvement with the strict prior range of parameter v45 for calibration, (c) and (d) after the improvement with the strict prior range of parameter v45 for calibration, and (e) and (f) before the improvement with the wider prior range of parameter v45 for calibration.
For the routing method in Ducharne et al. (), only the
also the hydraulic radius, the roughness of the cell and the
slope between two grids will influence the velocity. For the
scale factor about the cell size will influence the wave velocity.
routing method in Guo et al. (), not only the slope, but
For the routing method in Bunster et al. (), it proposed a
J. Li et al.
849
Table 6
|
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
The values of the optimal routing parameter v45 with the prior range of 4–10 m/s
Catchment
Pingshan
Shigu
Fushun
Gaochang
Beibei
Optimal value (m/s)
6
10
10
10
10
NSE
0.88
0.78
0.80
0.81
0.61
Catchment
Wulong
Baihe
Xiangtan
Taoyuan
Shimen
Optimal value (m/s)
9
10
10
10
9
NSE
0.88
0.43
0.77
0.82
0.60
Catchment
Waizhou
Lijiadu
Meigang
Shizhenjie
Dufengkeng
Optimal value (m/s)
10
10
10
10
10
NSE
0.88
0.43
0.77
0.82
0.60
Table 7
|
The values of the optimal routing parameter v45 with a wider prior range of 4–31 m/s
Catchment
Pingshan
Shigu
Fushun
Gaochang
Beibei
Optimal value
6
13
13
22
19
NSE
0.88
0.82
0.80
0.89
0.87
Catchment
Wulong
Baihe
Xiangtan
Taoyuan
Shimen
Optimal value
9
31
16
13
31
NSE
0.88
0.83
0.85
0.85
0.79
Catchment
Waizhou
Lijiadu
Meigang
Shizhenjie
Dufengkeng
Optimal value
25
19
16
16
13
NSE
0.80
0.85
0.93
0.81
0.72
travel time formulation that accounts for the dynamics of the
Factors that influence the wave velocity calculation
upstream contributions, compared to the previous methods. In that method, the dynamics of the upstream contributions and
The wave velocity of a grid is affected not only by the grid’s
slope will influence the wave velocity. Du et al. () pro-
slope but also by many other factors, including the basin
posed a time variant routing method, the spatially distributed
shape factor, rainfall intensity, vegetation cover, branching
direct hydrograph travel time method (SDDH), in which
ratio, basin area, climate, and form of drainage. The factors
time variant runoff, grid length, the roughness of the cell,
that influence the travel time are so numerous that they may
and slope are used to calculate the wave velocity.
compensate for each other to some extent. It is difficult to
Some data such as hydraulic radius may not be available
calculate the wave velocity of each grid precisely. So, these
when applying the large-scale hydrological model. Thus, the
influencing factors were generalized to be reflected by the
routing method requiring too much input information may
parameter v45 , the parameter b, and slope in this study.
not preferable in a large-scale hydrological model. In the
The parameter b varies across basins due to differences in
improved NRF routing method, all the necessary infor-
basin characteristics, and it represents the underlying
mation can be taken from the DEM. Compared with the
surface.
routing methods that only take the slope into consideration,
The underlying surface influences the wave velocity and
the improved NRF can be adapted to more basins. Overall,
travel time in many ways. There is no strict and determinis-
the improved NRF routing method is simple and effective
tic functional relationship between basin characteristic and
for a large-scale hydrological model.
mean wave velocity in a large-scale hydrological model.
850
J. Li et al.
|
The improvement study of the NRF flow routing method
Hydrology Research
|
51.5
|
2020
Multivariate regression analysis (Alexopoulos ; Zhang
There are other factors that are not included in Equation
et al. ) is a mathematical analysis method that uses the
(13), and for R 2 it is not equal to 1. However, we have yet to
least squares method to model the relationship between a
find more accurate and detailed rules. The mean wave vel-
dependent variable and multiple independent variables. In
ocity that was predicted by Equation (13) was compared
this study, we used it to examine the relationship between
with the mean wave velocity that was calculated with the
the mean wave velocity of grids v (m/s) calculated with
improved routing method as shown in Figure 9, which
the improved routing method and several influencing fac-
shows an overall good result.
tors. The result equation will not be used to calculate the
In the future study, more data on the factors that influ-
wave velocity directly. But this method can identify basin
ence wave velocity, such as soil, vegetation cover, and the
characteristics that have a greater impact on the mean
branching ratio, should be inputted into the method in
wave velocity. Some variables that are related to wave vel-
order to examine the relationship between routing par-
ocity are chosen (Table 1): mean discharge at the outlet
ameters and basin characteristics. If appropriate, the
station Q (m3/s); area A (km2); shape factor F; mean slope
improved routing model has the potential to be applied to
S ( ); river length Lr (m); and range of elevation ΔH (m).
data-sparse areas by a parameter transformation based on
The area, shape factor, and river length are basin geometric
the distribution in a subzone of climate, soil, and vegetation,
factors which affect the path and sequence of runoff and
or even to calculate the travel time without parameters,
then influence the mean wave velocity. The slope and
which would enable this model to be used to simulate the
range of elevation influence the wave velocity, because
hydrological effect of climate and land-use change.
potential energy will be converted into kinetic energy. In the process of runoff routes, runoff will continually merge and diverge, and the speed between water will affect each
CONCLUSIONS
other. Thus, the mean discharge at the outlet station is In this paper, a flow routing algorithm was developed by
also taken as an independent variable. Based on data of the 15 studied catchments, multiple
combining an aggregation NRF with a velocity function.
linear regression analysis was done using the above factors as independent variables and mean wave velocity as the dependent variable. The value of R 2 was found to be equal to 0.73 and the significance level of the F-test was 0.051, which indicates that the mean wave velocity is associated (at the 90% level) with the above factors. The final regression equation is shown as Equation (13). The standardized regression coefficients of those variables are 0.23 (Q), 1.37 (A), 0.44 (F), 1.42 (S), 0.55 (Lr ), and 0.79 (ΔH). It can be inferred that the slope has the greatest positive effect on wave velocity and the area has the most negative effect.
v ¼ 0:0001884Q 0:00001691A þ 0:8231F þ 0:6517S 0:0003146ΔH þ 0:0004925Lr
(13)
where v (m/s) is the mean wave velocity of the grids; Q (m3/s) is the mean discharge at the outlet station; A (km2) is the area; F is the shape factor; S ( ) is the mean slope; Lr (m) is the river’s length; and ΔH (m) is the range of elevation.
Figure 9
|
A comparison between the predicted and calculated mean wave velocity of grids.
851
J. Li et al.
|
The improvement study of the NRF flow routing method
Hydrology Research
The improved routing NRF method uses a parameter b that
|
51.5
|
2020
REFERENCES
can reflect how sensitive the wave velocity is to a slope. The routing methods before and after the improvement were applied to 15 catchments in the Yangtze River using a daily WASMOD-M model based on Hydro1 k. The main conclusions can be summarized as follows: (1) The original NRF routing method was found to provide unsatisfactory runoff simulation performance in most of the studied catchments, with an unreasonable calibrated travel time and wave velocity of the grids, for it cannot adapt to the different underlying conditions in the different catchments. (2) After the improvement in the NRF routing method, the relationship between wave velocity and slope was found to be more flexible. It can simulate the complexity of the wave velocity distribution more effectively in all types of basins, resulting in a reasonable runoff travel time. (3) The runoff travel time that was calculated by the model with the improved NRF routing method was found to be shorter and more reasonable, which leads to better runoff simulation performance. (4) The model with the improved NRF routing method was found to yield better results with respect to long-term daily runoff time series and peak flow amounts than the model with the original NRF routing method. The improved model was found to perform better in catchments with different characteristics. The improved model also retains the best advantage of NRF; i.e., it records all of the travel time information from the PRF.
ACKNOWLEDGEMENT This study was partly supported by the National Natural Science Fund of China (grant no. 51539009) and the National Key Research and Development Program (grant no. 2017YFA0603702). The authors are grateful to the National Meteorological Information Center for providing the meteorology datasets and the Bureau of Hydrology of the
Changjiang
Water
Resources
Commission
providing the datasets for the Yangtze River basin.
for
Abou Rafee, S. A., Uvo, C. B., Martins, J. A., Domingues, L. M., Rudke, A. P., Fujita, T. & Freitas, E. D. Large-scale hydrological modelling of the Upper Paraná River Basin. Water 11 (5), 882. doi:10.3390/w11050882. Alexopoulos, E. C. Introduction to multivariate regression analysis. Hippokratia 14 (Suppl. 1), 23. doi:10.1111/j.13652362.2009.02222.x. Arora, V. K. & Boer, G. J. A variable velocity flow routing algorithm for GCMs. Journal of Geophysical Research: Atmospheres 104 (D24), 30965–30979. doi:10.1029/ 1999jd900905. Arora, V., Seglenieks, F., Kouwen, N. & Soulis, E. Scaling aspects of river flow routing. Hydrological Processes 15 (3), 461–477. doi:10.1002/hyp.161. Barraquand, J. & Latombe, J. C. A Monte-Carlo algorithm for path planning with many degrees of freedom. IEEE International Conference on Robotics and Automation. doi:10.1109/ROBOT.1990.126256. Beven, K. How far can we go in distributed hydrological modelling? Hydrology and Earth System Sciences 5 (1), 1–12. doi:10.5194/hess-5-1-2001. Beven, K. J. & Kirkby, M. J. A physically based, variable contributing area model of basin hydrology/Un modèle à base physique de zone d’appel variable de l’hydrologie du bassin versant. Hydrological Sciences Journal 24 (1), 43–69. doi:10.1080/02626667909491834. Bunster, T., Gironás, J. & Niemann, J. D. On the influence of upstream flow contributions on the basin response function for hydrograph prediction. Water Resources Research 55 (6), 4915–4935. doi:10.1029/2018WR024510. Chen, L., Chang, J., Wang, Y. & Zhu, Y. Assessing runoff sensitivities to precipitation and temperature changes under global climate-change scenarios. Hydrology Research 50 (1), 24–42. doi:10.2166/nh.2018.192. Du, J., Xie, H., Hu, Y., Xu, Y. & Xu, C.-Y. Development and testing of a new storm runoff routing approach based on time variant spatially distributed travel time method. Journal of Hydrology 369 (1–2), 44–54. doi:10.1016/j.jhydrol.2009. 02.033. Ducharne, A., Golaz, C., Leblois, E., Laval, K., Polcher, J., Ledoux, E. & de Marsily, G. Development of a high resolution runoff routing model, calibration and application to assess runoff from the LMD GCM. Journal of Hydrology 280 (1–4), 207–228. doi:10.1016/S0022-1694(03)00230-0. Fan, M., Mawuko, D. O., Shibata, H. & Ou, W. Spatial conservation areas for water yield hydrological ecosystem services with their economic values effects under climate change: a case study of Teshio watershed located in northernmost of Japan. Hydrology Research 50 (6), 1679–1709. doi:10.2166/nh.2019.009. Fekete, B. M., Vörösmarty, C. J. & Lammers, R. B. Scaling gridded river networks for macroscale hydrology:
852
J. Li et al.
|
The improvement study of the NRF flow routing method
development, analysis, and control of error. Water Resources Research 37 (7), 1955–1967. doi:10.1029/2001wr900024. Gong, L., Widen-Nilsson, E., Halldin, S. & Xu, C.-Y. Largescale runoff routing with an aggregated network-response function. Journal of Hydrology 368 (1–4), 237–250. doi:10. 1016/j.jhydrol.2009.02.007. Gong, L., Halldin, S. & Xu, C.-Y. Global-scale river routing – an efficient time-delay algorithm based on HydroSHEDS high-resolution hydrography. Hydrological Processes 25 (7), 1114–1128. doi:10.1002/hyp.7795. Guo, J., Xu, L. & Leung, L. R. A new multiscale flow network generation scheme for land surface models. Geophysical Research Letters 31 (23), 345–357. doi:10.1029/ 2004GL021381. Hansen, N. & Ros, R. Benchmarking a weighted negative covariance matrix update on the BBOB-2010 noisy testbed. In: Conference Companion on Genetic and Evolutionary Computation. doi:10.1145/1830761.1830788. Huang, P. C. & Lee, K. T. Efficient DEM-based overland flow routing using integrated recursive algorithms. Hydrological Processes 31 (5), 1007–1017. doi:10.1002/hyp.11080. Kharin, V., Flato, G., Zhang, X., Gillett, N., Zwiers, F. & Anderson, K. . Earth’s Future 6 (5), 704–715. doi:704715. 10.1002/2018EF000813. Kirkby, M. J. & Beven, K. Channel Network Hydrology. John Wiley & Sons, New York. Kizza, M., Rodhe, A., Xu, C.-Y. & Ntale, H. K. Modelling catchment inflows into Lake Victoria: uncertainties in rainfall–runoff modelling for the Nzoia River. Hydrological Sciences Journal 56 (7), 1210–1226. doi:10.1080/02626667. 2011.610323. Li, L., Xia, J., Xu, C.-Y. & Singh, V. P. Evaluation of the subjective factors of the GLUE method and comparison with the formal Bayesian method in uncertainty assessment of hydrological models. Journal of Hydrology 390 (3), 210–221. doi:10.1016/j.jhydrol.2010.06.044. Li, Z., Xu, Z. & Li, Z. Performance of WASMOD and SWAT on hydrological simulation in Yingluoxia watershed in northwest of China. Hydrological Processes 25 (13), 2001–2008. doi:10.1002/hyp.7944. Li, L., Xu, C.-Y. & Jain, S. Comparison of the global TRMM and WFD precipitation datasets in driving q large-scale hydrological model in Southern Africa. Hydrology Research 44 (5), 770–788. doi:10.2166/nh.2012.175. Li, H., Beldring, S. & Xu, C.-Y. Implementation and testing of routing algorithms in the distributed Hydrologiska Byråns Vattenbalansavdelning model for mountainous catchments. Hydrology Research 45 (3), 322–333. doi:10.2166/nh.2013.009. Li, L., Shen, M., Hou, Y., Xu, C.-Y., Lutz, A. F., Chen, J., Jain, S. K., Li, J. & Chen, H. Twenty-first-century glaciohydrological changes in the Himalayan headwater Beas River basin. Hydrology Earth System Sciences 23 (3), 1483–1503. doi:10.5194/hess-23-1483-2019. Ling, Z., Nan, Z., Xu, L., Yi, X., Hernández, F. & Li, L. Application of the MacCormack scheme to overland flow
Hydrology Research
|
51.5
|
2020
routing for high-spatial resolution distributed hydrological model. Journal of Hydrology 558, 421–431. doi:10.1016/ j.jhydrol.2018.01.048. Lu, W. & Qin, X. S. Integrated framework for assessing climate change impact on extreme rainfall and the urban drainage system. Hydrology Research 51 (1), 77–89. doi:10. 2166/nh.2019.233. Lu, G., Liu, J., Wu, Z., Hai, H., Xu, H. & Lin, Q. Development of a large-scale routing model with scale independent by considering the damping effect of sub-basins. Water Resources Management 29 (14), 5237–5253. doi:10.1007/ s11269-015-1115-7. Mckay, M. D., Beckkman, R. J. & Conover, W. Comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics 21 (2), 266–294. doi:10.2307/1271432. Miller, J. R., Russell, G. L. & Caliri, G. Continental-scale river flow in climate models. Journal of Climate 7 (6), 914–928. doi:10.1175/1520-0442(1994)007<0914:CSRFIC> 2.0.CO;2. Morisawa, M. Measurement of drainage-basin outline form. Journal of Geology 66 (5), 587–591. doi:10.1086/626538. Müller Schmied, H. & Döll, P. Assessment of the terrestrial water balance using the global water availability and use model WaterGAP – status and challenges. In: EGU General Assembly Conference. pp. 2921. Naden, P., Broadhurst, P., Tauveron, N. & Walker, A. River routing at the continental scale: use of globally-available data and an a priori method of parameter estimation. Hydrology and Earth System Sciences 3 (1), 109–123. Nash, J. E. & Sutcliffe, J. V. River flow forecasting through conceptual models part I – a discussion of principles. Journal of Hydrology 10 (3), 282–290. doi:10.1016/0022-1694(70) 90255-6. Ngongondo, C., Gong, L., Xu, C.-Y. & Alemaw, B. F. Flood frequency under changing climate in the upper Kafue River basin, southern Africa: a large scale hydrological model application. Stochastic Environmental Research and Risk Assessment 27 (8), 1883–1898. doi:10.1007/s00477-013-0724-z. Olivera, F., Famiglietti, J. & Asante, K. Global-scale flow routing using a source-to-sink algorithm. Water Resources Research 36 (8), 2197–2207. doi:10.1029/2000WR900113. Ragettli, S., Tong, X., Zhang, G., Wang, H., Zhang, P. & Stähli, M. Climate change impacts on summer flood frequencies in two mountainous catchments in China and Switzerland. Hydrology Research. In press. doi:10.2166/nh.2019.118. Sausen, R., Schubert, S. & Dümenil, L. A model of river runoff for use in coupled atmosphere-ocean models. Journal of Hydrology 155 (3–4), 337–352. doi:10.1016/0022-1694(94) 90177-5. Shepard, D. A two-dimensional interpolation function for irregularly-spaced data. In: Proceedings of the ACM National Conference. pp. 517–524. doi:10.1145/800186.810616. Sircar, J. K., Ragan, R. M., Engman, E. T. & Fink, R. A. A GIS based geomorphic approach for the digital computation of time-area curves. In: Civil Engineering Applications of
853
J. Li et al.
|
The improvement study of the NRF flow routing method
Remote Sensing in Water Resources Engineering (D. B. Stafford, ed.). ASCE, New York, pp. 28–296. Su, B. D., Jiang, T. & Jin, W. B. Recent trends in observed temperature and precipitation extremes in the Yangtze River basin, China. Theoretical and Applied Climatology 83 (1–4), 139–151. doi:10.1007/s00704-005-0139-y. Tian, Z., Jing, Q., Dai, T., Jiang, D. & Cao, W. Effects of genetic improvements on grain yield and agronomic traits of winter wheat in the Yangtze River Basin of China. Field Crops Research 124 (3), 417–425. doi:10.1016/j.fcr.2011.07.012. USGS HYDRO1 k Elevation Derivative Database. Available from: https://www.usgs.gov/centers/eros/ science/usgs-eros-archive-digital-elevation-hydro1 k?qtscience_center_objects=0#qt-science_center_objects (accessed 23 November). doi:10.5066/F77P8WN0. Wang, F., Ducharne, A., Cheruy, F., Lo, M. H. & Grandpeix, J. Y. Impact of a shallow groundwater table on the global water cycle in the IPSL land–atmosphere coupled model. Climate Dynamics 50 (9–10), 3502–3522. doi:10.1007/ s00382-017-3820-9. Wang, Y., Ni, J., Yue, Y., Li, J., Borthwick, A. G. L., Cai, X., Xue, A., Li, L. & Wang, G. Solving the mystery of vanishing rivers in China. National Science Review 6 (6), 1239–1246. doi:10.1093/nsr/nwz022. Weisstein, E. W. ‘Lambert Azimuthal Equal-Area Projection.’ From MathWorld – A Wolfram Web Resource. Available from: http://mathworld.wolfram.com/ LambertAzimuthalEqual-AreaProjection.html (accessed 23 November). Wen, Z., Xu, L. & Yang, S. A new multiscale routing framework and its evaluation for land surface modeling applications. Water Resources Research 48 (8), 8528. doi:10. 1029/2011WR011337. Widén-Nilsson, E., Halldin, S. & Xu, C.-Y. Global waterbalance modelling with WASMOD-M: parameter estimation
Hydrology Research
|
51.5
|
2020
and regionalisation. Journal of Hydrology 340 (1–2), 105–118. doi:10.1016/j.jhydrol.2007.04.002. Xu, C.-Y. WASMOD – the water and snow balance modeling system. In: Mathematical Models of Small Watershed Hydrology Applications (V. P. Singh & D. Frevert, eds). Water Resources Publication, Highlands Ranch, Colo, pp. 555–590. Xu, J., Yang, D., Yi, Y., Lei, Z., Chen, J. & Yang, W. Spatial and temporal variation of runoff in the Yangtze River basin during the past 40 years. Quaternary International 186 (1), 32–42. doi:10.1016/j.quaint.2007.10.014. Yang, X., Magnusson, J., Rizzi, J. & Xu, C.-Y. Runoff prediction in ungauged catchments in Norway: comparison of regionalization approaches. Hydrology Research 49 (2), 487–505. doi:10.2166/nh.2017.071. Yang, X., Magnusson, J. & Xu, C.-Y. Transferability of regionalization methods under changing climate. Journal of Hydrology 568, 67–81. doi:10.1016/j.jhydrol.2018.10.030. Yang, X., Magnusson, J., Huang, S., Beldring, S. & Xu, C.-Y. Dependence of regionalization methods on the complexity of hydrological models in multiple climatic regions. Journal of Hydrology 582, 124357. doi:10.1016/j.jhydrol.2019.124357. Yao, Z., Liu, Z., Huang, H., Liu, G. & Wu, S. Statistical estimation of the impacts of glaciers and climate change on river runoff in the headwaters of the Yangtze River. Quaternary International 336, 89–97. doi:10.1016/j.quaint. 2013.04.026. Zhang, Y., Min, Y., Zheng, J., Wang, G. & Niu, D. Compilation of Flood Forecasting Plan for the Yangtze River Basin. Hydrology Bureau of Yangtze River Water Resources Commission, Ministry of Water Resources, Wuhan (in Chinese). Zhang, G., Xie, H., Yao, T., Li, H. & Duan, S. Quantitative water resources assessment of Qinghai Lake basin using Snowmelt Runoff Model (SRM). Journal of Hydrology 519, 976–987. doi:10.1016/j.jhydrol.2014.08.022.
First received 25 November 2019; accepted in revised form 15 March 2020. Available online 23 June 2020
854
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Identification of regional water security issues in China, using a novel water security comprehensive evaluation model Jiping Yao, Guoqiang Wang
, Baolin Xue, Gang Xie and Yanbo Peng
ABSTRACT In order to solve regional water security issues, such as shortage of water resources, the aggravation of water pollution, the destruction of the ecological environment, etc., this study proposed the flood control security index, resource security index and ecological security index, respectively, according to the construction principle of human development index. Based on the above security indexes, a novel water security comprehensive evaluation model is established by combining the coupling coordination degree model and the state space model. The proposed model has the advantage of simple operation and fast data speed, which is convenient for water security evaluation in different periods and regions. Taking China as an example, the water security conditions were evaluated from 2007 to 2016 for 31 provincial-level administrative regions in China, including flood control security
Jiping Yao (corresponding author) Guoqiang Wang Baolin Xue Beijing Key Laboratory of Urban Hydrological Cycle and Sponge City Technology, College of Water Sciences, Beijing Normal University, Beijing 100875, China E-mail: wanggq@bnu.edu.cn Gang Xie Yanbo Peng Shandong Academy for Environmental Planning, Jinan 250000, China
index, resource security index, ecological security index and water security level of each region, and the specific problems of water security in each region were obtained. The evaluation results are consistent with the actual situation in each region, which provides the scientific basis for the local government authorities to formulate the corresponding regional water security policy. Key words
| ecological security index, flood control security index, regional water security, resources security index, water security evaluation model
INTRODUCTION Water is essential for maintaining the balance of life and the
et al. , a; Yang et al. ; Yao et al. c). How-
living environment. It is a basic natural resource and stra-
ever, in recent years, frequent floods, water shortage,
tegic resource to promote human economic development
pollution and water ecological damage have become serious
and social progress (Masseroni et al. ; Wang et al.
global challenges (Parry et al. ; Emam et al. ; Awan
b; Yao et al. a, b). Water security refers to the
et al. ; Han et al. ; Liu et al. a; Ledingham et al.
capacity of water resources with quantity and quality
). In 1972, the United Nations Conference on Environ-
guarantee required for human survival and development,
ment and Development predicted that a water crisis would
which can maintain the basin sustainability and human
take place following the oil crisis (Biswas ). In 2000,
and ecological environment health, and ensure people’s
the Hague and the World Water Week took ‘water security
life and property from water disasters (floods, landslides
in the 21st century’ as the theme of the world Ministerial
and droughts) (Ren et al. ; Fang et al. a; Wang
Level Conference (Falkenmark ). In 2009, United Nations Educational, Scientific and Cultural Organization
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying,
pointed out in its World Water Resources Development
adaptation and redistribution, provided the original work is properly cited
Report that the contradiction between supply and demand
(http://creativecommons.org/licenses/by/4.0/).
of water resources in human society is more prominent,
doi: 10.2166/nh.2020.014
855
J. Yao et al.
|
Identification of regional water security issues in China
Hydrology Research
|
51.5
|
2020
coupled with climate degradation and rapid population
global sustainable development challenges. Actually, the
growth, which leads to more serious water security problems
water security issues are caused by the comprehensive
(Russo et al. ; Jiang ; Kumar ; Li et al. ;
influence of society, economy, resources, environment
Fang et al. b). Therefore, it is necessary to study water
and ecology (Awan et al. ). However, at present, the
security and propose solutions to related issues from differ-
water security evaluation is only based on the character-
ent perspectives.
istics of the research area or the focus of the research
Many scholars study water security and related issues
problem, and puts forward the methods to solve the
from different perspectives. Harris & Kennedy ()
water security problem from different angles. The theory
pointed out that urban water supply should be integrated
and methods of water security evaluation are not compre-
into the actual urban development planning from the
hensive enough, and the universality of the proposed
perspective of water supply, and further explored the evol-
evaluation method is relatively low.
ution of urban water security. Rijsberman & van de Ven
In the 1990 Human Development Report, the United
() revealed the development of urban water security
Nations proposed the Human Development Index (HDI),
from a new perspective of water resources carrying capacity.
which comprehensively reflects the level of human develop-
Sullivan () proposed a water poverty index similar to the
ment among different countries and regions by using three
Consumer Price Index to reflect the impact of water short-
variables: human life index, education level index and
age on human beings. Falkenmark & Lundqvist ()
GDP index (Kawada et al. ). The index plays an impor-
paid attention to the water quantity problem, comprehen-
tant role in guiding the development strategies of developing
sively considering the shortage of natural water resources
countries. Therefore, this paper puts forward a flood control
and water quality shortage caused by water pollution, and
security index, resource security index and ecological secur-
provided a more objective and real basis for local relevant
ity index similar to HDI from flood control, resources and
departments to formulate water security protection policies.
ecology, and establishes an evaluation model that can com-
Ou et al. () used the established entropy weight-fuzzy
prehensively reflect the regional water security situation.
matter element model to evaluate the rural water safety situ-
The water security of the 31 provinces of China is analyzed
ation, which effectively reduced the influence of uncertainty
and evaluated by the proposed water security evaluation
and fuzziness in the evaluation process on the authenticity
model, and the distributions of flood control security
of water safety. Tian & Gang () used the pressure state
index, the resource security index and the ecological
response conceptual model to evaluate regional water
security index are respectively obtained. Additionally, the
security from the perspective of ecological security.
distributions of water security index which can comprehen-
Norman et al. () evaluated the water security situation
sively reflect the water security status of each region are
of a community in Canada with the method of water
obtained. All of the above indexes provide a scientific and
security index evaluation, and obtained good results. Con-
reliable basis for a comprehensive understanding of water
sidering the uncertainty of drought events, Dong & Xia
security problems in various regions of China and for the
() introduced the Dempster Shafer evidence theory
formulation of targeted strategies to solve water security
and evidence reasoning algorithm to assess the risk of
problems.
water security during drought. Gain et al. () put forward a framework that can quantitatively reflect the impact of human activities and natural water resources
MATERIALS AND METHODS
on water security by combining the factors that affect freshwater resources with the factors of social and econ-
Study area and data
omic development. Larson () took water security as the leading mode to study natural resource policy, and
According to the three key factors of water security, namely
comprehensively explored the situation of human water
flood control security, resource security and ecological
security by combining climate change with other pressing
security, the current situation and changes of water security
856
J. Yao et al.
|
Identification of regional water security issues in China
Hydrology Research
|
51.5
|
2020
in China from 2007 to 2016 are comprehensively evaluated.
index and ecological security index respectively. The three
In this paper, 31 administrative regions in mainland China
indexes are used to establish the water security evaluation
are taken as the research units of regional water security
model. The specific calculation methods are discussed below.
to study the temporal and spatial evolution of water security
The key points of flood control security are casualties
in China (Figure 1), and the key factors affecting the water
caused by flood, economic loss of water conservancy facili-
security status of China and each research unit are analyzed.
ties and disaster area loss caused by flood in the region.
Due to the lack of relevant data, Taiwan, Hong Kong and
Therefore, the flood control security index (FS) established
Macao were not included in the study. The dataset of this
includes three data sets: loss rate of disaster area Ra, regional
study came from the China Water Conservancy Bulletin,
flood disaster population rate Rc and the ratio of economic
China Water Development Statistical Yearbook, China
loss of water-saving facilities to direct economic loss Re,
Statistical
which reflects the regional social development water. The
Yearbook,
China
Environmental
Statistical
Yearbook and the China Flood and Drought Disaster
specific calculation formula is as follows:
Bulletin during the period 2007–2016 (Liu et al. a).
Methods Determination of security indexes of the regional water security evaluation model
8 Ra > > < Rc > > : Re
¼ ¼ ¼
Rad Ras Rcf Rcp Rew Red
(1)
where Rad denotes flood damage areas (km2), Ras denotes Based on the three key factors of flood control safety,
flood disaster areas (km2), Rcf denotes the number of
resource security and ecological security, which can compre-
flood-affected population (10,000 people), Rcp denotes the
hensively reflect the regional water security, this paper
number of population affected by disasters (10,000
establishes the flood control security index, resource security
people), Rew denotes economic loss of water-saving facilities
Figure 1
|
Overview of the study area.
857
J. Yao et al.
|
Identification of regional water security issues in China
Hydrology Research
|
51.5
|
2020
in floods (100,000,000 yuan), Red denotes direct economic
as the lower limit of per capita water resources in this
losses in floods (100,000,000 yuan). Since floods are
study (Liu et al. b).
driven by disasters caused by natural and human events,
The ecological security index (ES) focuses on the
the five-year moving average value is calculated based on
environmental conditions related to water, which reflects
the above data to reduce the impact of the annual flood con-
the security of water ecology through the ecological con-
trol security index. The regional flood control security index
ditions of rivers, lakes and reservoirs. The index includes
(FS) is defined as:
the ratio of the length of a river above Class III to total
8 1, Ri 0:01 > > P > > lg Ri > > < 0:2 0:4 × , P3 FS ¼ > lg Ri > > 0:3 0:3 × , > > 3 > : 0, Ri > 1
river length (RW) and the ratio between the number of non-eutrophication lakes to the total number of lakes 0:01 < Ri 0:1
(RE). The calculation formula for the ecological security (i ¼ a, c, e)
index (ES) is as follows:
0:01 < R1 1 (2)
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi RW 2 þ RE2 ES ¼ 2
(5)
The resource security index (RS) reflects the coordination between the per capita water resources, urban and rural development level and water supply capacity. The index shows the relative guarantee strength among regional
Construction of regional water security model
basic resources, economic development level and water
Based on the flood control security index, resource security
supply capacity. The resource security index (RS) includes
index and ecological security index, the water security index
per capita water resources, urbanization rate and per
is proposed to reflect regional water security. Due to the
capita water demand, and its calculation formula is as follows:
and ecology (Wilkinson & Bathurst ), this paper introduces the coupling coordination model (Cheng et al. )
8 > < 1, Wr 2W 1=2 RS ¼ 1 2 2 2 > 1 × [(1 W) þ (1 U) þ (1 C) ] , Wr < 2W : 3 (3) where W denotes the normalized water resource factor (see Equation (4)), U denotes the urbanization rate (%), and C is the per capita water storage capacity. 8 1, Wr > 2W > > < W ¼ lg Wr =lg 2W , > Wrmin > : Wrmin 0, Wr Wrmin
complementary relationship among flood control, resources
to characterize the interaction between flood control, resources and ecology, and the specific calculation formula is as follows: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 8 1=2 > > DFS↔RS ¼ [FS × RS=(FS þ RS)2 ] × (a × FS þ b × RS) > > < qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1=2 DRS↔ES ¼ [RS × ES=(RS þ ES)2 ] × (b × RS þ c × ES) > > q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi > > 1=2 : DFS↔ES ¼ [FS × ES=(FS þ ES)2 ] × (a × FS þ b × ES) (6)
Wrmin < Wr 2W
(4)
where DFS↔RS , DRS↔ES and DFS↔ES respectively represent the coupling coordination degree of the interaction between the flood control security and the resource security, the
where Wr denotes the per capita water resources volume, W 3
coupling coordination degree of the interaction between
is the global per capita water resources volume (6,123 m ),
the resource security and the ecological security, and the
and Wrmin is the lower limit of global per capita water
coupling coordination degree of the flood control security
resources volume. The per capita water resources of Israel
and the ecological security interaction. a, b and c are the
(a country with a serious water shortage), 97 m3, is taken
undetermined coefficients of the flood control security
J. Yao et al.
858
|
Identification of regional water security issues in China
Hydrology Research
index, resource security index and ecological security index 1 respectively, the value of which should be a ¼ b ¼ c ¼ 2 (Cheng et al. ). On the basis of the above-mentioned security interactions, the state space model (Li et al. ) is used to obtain a water security index (WSI) that can comprehen-
Table 2
|
|
51.5
|
2020
Average values of flood control security index in the study area from 2007 to 2016
Region
Ra
Rc
Re
FS
Beijing
0.5150
0.5031
0.2132
0.4258
Tianjin
0.7343
0.1358
0.1840
0.4737
Hebei
0.5656
0.0428
0.1871
0.5344
Shanxi
0.6055
0.0531
0.0718
0.5636
resource security and ecological security interaction on
Inner Mongolia
0.6150
0.2719
0.0954
0.4797
water security. Based on the above steps, a water security
Liaoning
0.6164
0.1722
0.2212
0.4629
evaluation model is established. The specific calculation for-
Jilin
0.5247
0.0649
0.2144
0.5137
mula is as follows:
Heilongjiang
0.6584
0.0154
0.0943
0.6020
Shanghai
0.3654
0.0321
0.0611
0.6145
Jiangsu
0.3190
0.0010
0.0613
0.7709
Zhejiang
0.4632
0.0084
0.1338
0.6285
Anhui
0.5147
0.0137
0.2187
0.5813
sively consider the impact of flood control security,
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi D2FS↔RS þ D2RS↔ES þ D2FS↔ES WSI ¼ 3
(7)
In order to scientifically and effectively show the advan-
Fujian
0.4454
0.0580
0.1926
0.5303
tages and disadvantages of water security in various regions,
Jiangxi
0.5247
0.0234
0.2297
0.5549
this paper classifies water security into five levels: unsafe,
Shandong
0.5712
0.0092
0.0783
0.6385
relatively unsafe, general security, relative security and
Henan
0.2643
0.0064
0.2314
0.6406
security according to the water security classification stan-
Hubei
0.3644
0.0454
0.1936
0.5494
dard (Ren et al. ) (see Table 1).
Hunan
0.5185
0.0356
0.2318
0.5369
RESULTS AND DISCUSSION
Guangdong
0.4352
0.0585
0.1526
0.5410
Guangxi
0.3543
0.0485
0.1351
0.5634
Hainan
0.5439
0.0399
0.0818
0.5750
Chongqing
0.4053
0.0854
0.1641
0.5246
Sichuan
0.5295
0.0612
0.1723
0.5253
Considering that the flood control security index, resource
Guizhou
0.4664
0.1050
0.1298
0.5197
security index and ecological security index are affected by
Yunnan
0.5745
0.1935
0.1744
0.4712
interannual changes, only relying on the evaluation results
Tibet
0.4157
0.2733
0.1922
0.4661
of a single year to determine the current situation of water
Shaanxi
0.6509
0.1643
0.1837
0.4707
security in each region of the study area will lead to a large
Gansu
0.7150
0.1307
0.1260
0.4929
deviation between the evaluation results and the actual situ-
Qinghai
0.5609
0.6281
0.2317
0.4088
ation. Consequently, based on the average values of the water
Ningxia
0.4540
0.2043
0.3703
0.4464
security evaluation indexes data for 2007–2016, the proposed
Xinjiang
0.6254
0.3694
0.3534
0.4088
water security evaluation model is used to obtain the flood control security index, the resource security index, the ecological security index and the water security index
water security status in 31 administrative regions of the
(Tables 2–4). The spatial distribution characteristics of
study area were analyzed (Figures 2–5).
Table 1
|
Characteristics of flood control security in the study
Water security evaluation standards
area Water security
Relatively
Generally
Relatively
level
Unsafe unsafe
safe
safe
Safe
Level threshold
[0,0.2) [0.2,0.4)
[0.4,0.6)
[0.6,0.8)
[0.8,1.0]
Table 2 shows the average value of flood control security index (FS) of each administrative region in the study area
J. Yao et al.
859
Table 3
|
|
Identification of regional water security issues in China
Hydrology Research
Average values of resources security index in the study area from 2007 to 2016
Table 4
|
|
51.5
|
2020
Average values of ecological security index in the study area from 2007 to 2016
Region
W
U
C
RS
Region
RW
RE
ES
Beijing
0.0869
0.8584
0.1194
0.2630
Beijing
0.8155
1.0000
0.9124
Tianjin
0.0523
0.8030
0.0834
0.2303
Tianjin
0.0916
0.0000
0.0648
Hebei
0.1883
0.4642
0.1379
0.2496
Hebei
0.4353
0.4219
0.4286
Shanxi
0.2657
0.4687
0.0943
0.2602
Shanxi
0.2689
0.0000
0.1901
Inner Mongolia
0.6706
0.5648
0.2053
0.4434
Inner Mongolia
0.7445
0.0000
0.5264
Liaoning
0.2861
0.6396
0.4130
0.4273
Liaoning
0.4391
1.0000
0.7723
Jilin
0.5264
0.5404
0.5955
0.5531
Jilin
0.6747
0.4290
0.5654
Heilongjiang
0.6373
0.5673
0.3549
0.5050
Heilongjiang
0.6690
0.4026
0.5521
Shanghai
0.2041
0.8885
0.0022
0.2603
Shanghai
0.4786
0.0000
0.3384
Jiangsu
0.4157
0.6121
0.0219
0.3051
Jiangsu
0.3497
0.1389
0.2661
Zhejiang
0.6721
0.6215
0.4005
0.5490
Zhejiang
0.7189
0.9088
0.8194
Anhui
0.5612
0.4551
0.2649
0.4141
Anhui
0.6167
0.5489
0.5838
Fujian
0.7374
0.6237
0.2529
0.4938
Fujian
0.7980
0.7488
0.7738
Jiangxi
0.7847
0.5178
0.3356
0.5100
Jiangxi
0.9369
1.0000
0.9690
Shandong
0.1185
0.5785
0.1037
0.2345
Shandong
0.4693
0.4599
0.4646
Henan
0.2341
0.4623
0.2264
0.2990
Henan
0.5095
0.6689
0.5946
Hubei
0.5916
0.5692
0.9986
0.6572
Hubei
0.8050
0.7883
0.7967
Hunan
0.6946
0.5069
0.3699
0.5056
Hunan
0.9821
0.0000
0.6945
Guangdong
0.6066
0.6307
0.2099
0.4476
Guangdong
0.7891
0.7387
0.7643
Guangxi
0.8142
0.4814
0.6873
0.6343
Guangxi
0.9585
1.0000
0.9795
Hainan
0.6451
0.5645
0.4140
0.5313
Hainan
0.9370
0.6689
0.8141
Chongqing
0.5656
0.6057
0.1991
0.4268
Chongqing
0.9928
0.0000
0.7020
Sichuan
0.6843
0.4768
0.2387
0.4364
Sichuan
0.8846
0.6645
0.7823
Guizhou
0.7236
0.4272
0.4199
0.5030
Guizhou
0.8142
1.0000
0.9118
Yunnan
0.7638
0.4638
0.7863
0.6399
Yunnan
0.8693
0.6594
0.7716
Tibet
1.0000
0.2796
0.4480
1.0000
Tibet
0.9948
1.0000
0.9974
Shaanxi
0.4558
0.4824
0.1298
0.3364
Shaanxi
0.6797
1.0000
0.8550
Gansu
0.3851
0.3781
0.1976
0.3148
Gansu
0.7485
1.0000
0.8832
Qinghai
0.9571
0.4614
0.9819
0.6879
Qinghai
0.9749
1.0000
0.9875
Ningxia
0.0542
0.5006
0.2218
0.2363
Ningxia
0.3671
0.0090
0.2597
Xinjiang
0.7689
0.4353
0.4179
0.5131
Xinjiang
0.9969
1.0000
0.9984
from 2007 to 2016, which is obtained from flood control
are more developed in society and economy, have more invest-
security evaluation factors (Ra, Rc, Re). Based on Table 2,
ment in infrastructure, and are more advanced in technology,
the spatial distribution of the average flood control security
so they have better performance in this regard (Liu et al.
index in the study area from 2007 to 2016 was obtained
b). On the contrary, the flood control security index of
(Figure 2). The flood control security index for Heilongjiang,
other provincial administrative areas is lower than 0.6,
Shanghai, Jiangsu, Zhejiang, Henan and Shandong exceeds
which indicates that the flood control security level in these
0.6, which indicates that the flood control security level of
areas is lower. In particular, the flood control security level
these areas is higher. This is mainly due to the fact that
of Xinjiang and Qinghai in the west of the study area is the
these areas are prone to more rainfall and flood events,
lowest at 0.4088. This low level is due to the rare rainfall,
860
Figure 2
J. Yao et al.
|
|
Identification of regional water security issues in China
Hydrology Research
|
51.5
|
2020
Spatial distribution of annual average flood control security index in the study area from 2007 to 2016.
rare flood and insufficient flood control infrastructure in wes-
Characteristics of resources security in the study area
tern China (especially the northwest) (Zhang et al. ). In this case, people are more vulnerable to floods and suffer
Resource security reflects the coordination of regional
more losses. These areas require a higher level of flood control
basic resources, economic development and water supply
alarms. Relevant government departments of the study area
capacity. Table 3 shows the average value of the resource
need to strengthen the unified management of water resources
security index (RS) for each administrative region in the
and flood control, and speed up the preparation or revision of
study area from 2007 to 2016, which was obtained from
emergency operation plans for flood control operation of
resource security evaluation factors (W, U, C ). Based on
important rivers. In addition, relevant departments need to
Table 3, the spatial distribution of the annual average
speed up the formulation of water distribution plans and
resource security index in the study area from 2007 to
water conservancy renovation plans for major rivers, improve
2016 was obtained (Figure 3). In all administrative regions
the unified dispatching management scheme, clarify the dis-
of the study area, the resource security index (RS) of
patching project and authority, and strengthen the flood risk
Tibet is equal to 1, which indicates that the level of resource
management.
security in the region is very high. This is mainly due to the
861
Figure 3
J. Yao et al.
|
|
Identification of regional water security issues in China
Hydrology Research
|
51.5
|
2020
Spatial distribution of annual average resource security index in the study area from 2007 to 2016.
relatively high per capita water resources in Tibet, which is
water resources are scarce (W < 0.12), which leads to the
twice as large as that in the rest of the world. In addition, the
lack of per capita water resources. Especially, the per capita
low urbanization rate, sparse population and relatively high
water resources of Tianjin are lower than that of the world
per capita water resources in this area make Tibet have a
minimum standard of water resources per capita (97 m3).
superior resource security effect (Liu et al. b). The resource
Additionally, the water supply capacity of Hebei, Shanxi,
security index (RS) of Qinghai, Hubei, Guangxi and Yunnan is
Shanghai, Jiangsu, Shandong, Shaanxi and Gansu is very
higher than 0.6, which indicates that the resource security
low (C < 0.14). This indicates that the above-mentioned
water in these areas is higher. The resource security indexes
administrative regions are facing great challenges in
of Beijing, Tianjin, Hebei, Shanxi, Shanghai, Jiangsu, Shan-
resource security, these regions need to reasonably adjust
dong, Henan, Shaanxi and Gansu are lower than 0.4, which
the industrial structure, optimize the industrial layout, and
indicates that the resource security level of these areas is rela-
improve the efficiency of water resource allocation accord-
tively low. This is mainly because the urbanization rate of
ing to the natural conditions and relevant development
Beijing and Tianjin is very high (U > 0.8), and the per capita
plans of water resources in each region.
862
Figure 4
J. Yao et al.
|
|
Identification of regional water security issues in China
Hydrology Research
|
51.5
|
2020
Spatial distribution of annual average ecological security index in the study area from 2007 to 2016.
Characteristics of ecological security in the study area
Xinjiang all exceed 0.8, which indicates that the ecological security of these areas is at a high level. According to the
Ecological security fully reflects the water quality and eco-
data in the China Water Development Statistical Yearbook
logical characteristics of rivers, lakes and reservoirs.
and China Water Resources Bulletin, the results of regional
Table 4 shows the average value of the ecological security
ecological security assessment during the period of 2007–
index (ES) for each administrative region in the study area
2016 show that the above-mentioned areas have better
from 2007 to 2016, which was obtained from ecological
environmental governance, and the proportion of rivers
security evaluation factors (RW, RE). Based on Table 4,
above class III and non-eutrophication lakes and reservoirs
the spatial distribution of the annual average ecological
that meet the requirements of RW and RE is relatively large.
security index in the study area from 2007 to 2016 was
However, the ecological security index related to water qual-
obtained (Figure 4). The results show that the ecological
ity of rivers, lakes and reservoirs in Tianjin and Shanxi is
security indexes of Beijing, Zhejiang, Jiangxi, Guangxi,
lower than 0.2. This indicates that the ecological security
Hainan, Guizhou, Shaanxi, Tibet, Gansu, Qinghai and
of Tianjin and Shanxi is at a low level, which may be
863
Figure 5
J. Yao et al.
|
|
Identification of regional water security issues in China
Hydrology Research
|
51.5
|
2020
Spatial distribution of annual average water security index in the study area from 2007 to 2016.
caused by the serious water pollution and eutrophication of
departments of environmental protection need to formulate
lakes and reservoirs in these two areas. The ecological secur-
effective measures to reduce the total amount of pollutants,
ity index of Shanghai, Jiangsu and Ningxia varies between
control the movement rate of pollutants, improve the effi-
0.2 and 0.4 and this indicates that the ecological security
ciency of sewage treatment, realize the comprehensive
of these areas is at the general level. In addition, the ecologi-
treatment of major rivers, and improve the ecological secur-
cal security index of the remaining areas is more than 0.5,
ity early warning system.
which indicates that the ecological security of these areas is at a relatively high level. In general, since the release of
Characteristics of water security in the study area
the Water Pollution Control Action Plan by the State Council in April 2015, the water ecological security situation in
Table 5 shows the average value of the water security index
the study area has been improved, and the water ecological
(WSI) for each administrative region in the study area from
environment protection has been strengthened (Liu et al.
2007 to 2016, which was obtained from FS, RS and ES.
b). In order to ensure the sustainable and healthy devel-
Based on Table 5, the spatial distribution of the annual aver-
opment of ecological security in the study area, the relevant
age water security index in the study area from 2007 to 2016
J. Yao et al.
864
Table 5
|
|
Identification of regional water security issues in China
Average values of water security index in the study area from 2007 to 2016
RS
ES
WSI
Security level
Hydrology Research
|
51.5
|
2020
the water security of these regions is in an unsafe state. This is mainly due to the low level of resource security and ecologi-
Region
FS
Beijing
0.4258 0.2630 0.9124 0.5337 Generally safe
be caused by the serious water shortage and low control rate
Tianjin
0.4737 0.2303 0.0648 0.2563 Relatively unsafe
of ecological water functional areas in these areas.
Hebei
0.5344 0.2496 0.4286 0.4042 Generally safe
Shanxi
0.5636 0.2602 0.1901 0.3380 Relatively unsafe
Inner Mongolia 0.4797 0.4434 0.5264 0.4832 Generally safe
cal security in Tianjin, Hebei, Shanxi and Ningxia, which may
To sum up, the northern part of the study area (Beijing, Tianjin, Hebei, Shanxi and Inner Mongolia) suffers from serious water shortage, and the water quality of lakes and
Liaoning
0.4629 0.4273 0.7723 0.5541 Generally safe
Jilin
0.5137 0.5531 0.5654 0.5441 Generally safe
Heilongjiang
0.6020 0.5050 0.5521 0.5530 Generally safe
Shanghai
0.6145 0.2603 0.3384 0.4044 Generally safe
Jiangsu
0.7709 0.3051 0.2661 0.4474 Generally safe
Zhejiang
0.6285 0.5490 0.8194 0.6656 Relatively safe
Anhui
0.5813 0.4141 0.5838 0.5264 Generally safe
Fujian
0.5303 0.4938 0.7738 0.5993 Relatively safe
Jiangxi
0.5549 0.5100 0.9690 0.6780 Relatively safe
Shandong
0.6385 0.2345 0.4646 0.4459 Generally safe
Jiangsu, Zhejiang, Anhui, Fujian, Shanxi and Shandong)
Henan
0.6406 0.2990 0.5946 0.5114 Generally safe
Hubei
0.5494 0.6572 0.7967 0.6678 Relatively safe
has strong flood control capacity and a high-water security
Hunan
0.5369 0.5056 0.6945 0.5790 Generally safe
security level of Shanghai and Jiangsu are relatively low.
Guangdong
0.5410 0.4476 0.7643 0.5843 Generally safe
The water security indexes of the central regions (Henan,
Guangxi
0.5634 0.6343 0.9795 0.7257 Relatively safe
Hubei and Hunan) in the study area are similar to that of
Hainan
0.5750 0.5313 0.8141 0.6402 Relatively safe
Chongqing
0.5246 0.4268 0.7020 0.5511 Generally safe
Sichuan
0.5253 0.4364 0.7823 0.5813 Generally safe
Guizhou
0.5197 0.5030 0.9118 0.6448 Relatively safe
Yunnan
0.4712 0.6399 0.7716 0.6276 Relatively safe
Tibet
0.4661 0.4622 0.9974 0.6419 Relatively safe
Shaanxi
0.4707 0.3364 0.8550 0.5540 Generally safe
Gansu
0.4929 0.3148 0.8832 0.5636 Generally safe
Qinghai
0.4088 0.6879 0.9875 0.6947 Relatively safe
Ningxia
0.4464 0.2363 0.2597 0.3141 Relatively unsafe
Xinjiang
0.4088 0.5131 0.9984 0.6401 Relatively safe
reservoirs is poor, which is reflected by the lower resource security index and ecological security index in these areas. The above problems are the main reasons for the unsafe water security in these areas. The northeast of the study area (Heilongjiang, Jilin and Liaoning) is relatively balanced in flood control security, resource security and ecological security, which leads to the overall high-water security index in these areas, and the water security is in a general state of security. The east part of the study area (Shanghai,
index. However, the resource security level and ecological
the eastern regions. Henan is the most populous area in the study area, which leads to the most serious problem of resource shortage. The south (Guangdong, Guangxi, Hainan) and southwest (Chongqing, Sichuan, Guizhou, Yunnan, Tibet) of the study area have a high water security index, and the lakes, rivers and reservoirs in these areas have good water quality. The water security situation in the northwest of the study area (Shaanxi, Ningxia, Gansu, Qinghai, Xinjiang) is generally good. However, Ningxia is facing relatively unsafe water resources and environmental conditions, which need more attention. The above conclusions are consistent with the published articles (Liu
was obtained (Figure 5). The results show that the water
et al. b; Wang et al. c; Zhao et al. ), which indi-
security index of Guangxi is the highest among all administra-
cates that the water security evaluation model proposed in
tive regions in the study area (WSI ¼ 0.7257), which indicates
this paper has scientific and reliable application value.
that the water security in this area is in a safe state. The water security indexes of Zhejiang, Jiangxi, Hubei, Hainan, Guizhou, Yunnan, Tibet, Qinghai and Xinjiang vary from 0.6 to
CONCLUSIONS
0.8, which indicates that these areas are in a relatively safe state, while the water security indexes of Tianjin, Hebei,
Based on flood control security, resource security and ecologi-
Shanxi and Ningxia are less than 0.4, which indicates that
cal security, which can comprehensively reflect the water
865
J. Yao et al.
|
Identification of regional water security issues in China
security situation, this paper establishes the corresponding
Hydrology Research
|
51.5
|
2020
REFERENCES
security indexes, and uses the coupling coordination model and state space model to organically combine these security indexes, and constructs a water security comprehensive evaluation model. The proposed model is used to study the flood control security, resource security and ecological security of each administrative region in 2007–2016 in China. The water security level of each administrative region in the study area is characterized according to the above-mentioned security and the water security issues between each region are analyzed and compared by using the water security level. The results show that, in terms of flood control security, except for the relative security of Heilongjiang, Shandong, Henan, Jiangsu, Shanghai, Zhejiang, the flood control security capacity of other administrative regions needs to be further improved. In terms of resource security, Beijing, Tianjin, Hebei, Shanxi, Shaanxi, Gansu and Ningxia have a relatively low level of resource security due to their low resource base conditions, these areas need to improve the efficiency of resource utilization and establish a sound resource security system. In addition, the uncoordinated level of resources, population and economic development in Shanghai, Jiangsu, Shandong and Henan also leads to the relatively low level of resource security, these areas need to establish appropriate resource optimization allocation programs to improve the coordination between social and economic development and resource utilization. In terms of ecological security, the level of ecological security in Tianjin, Shanxi, Shanghai, Jiangsu and Ningxia is relatively low, so it is urgent to strengthen the restoration and protection of ecological water environment in these areas. Compared with the traditional water security evaluation system, which needs multi-layer design, a complex calculation method and different evaluation indexes, the water safety evaluation model proposed in this paper has clear theory, simple structure, easy interpretation, strong robustness, and good promotion and application.
ACKNOWLEDGEMENTS This study was supported by the National Natural Science Foundation of China (Grant No. 51679006 and Grant No. 51879006).
Awan, U. K., Liaqat, U. W., Choi, M. & Ismaeel, A. A SWAT modeling approach to assess the impact of climate change on consumptive water use in Lower Chenab Canal area of Indus basin. Hydrol. Res. 47, 1025–1037. Biswas, A. K. Water crisis – current perceptions and future realities. Water Int. 24, 363–367. Cheng, K., Yao, J. & Ren, Y. Evaluation of the coordinated development of regional water resource systems based on a dynamic coupling coordination model. Water Sci. Technol. Water Supply 19, 565–573. Dong, Q. & Xia, L. Risk assessment of water security in Haihe River Basin during drought periods based on D-S evidence theory. Water Sci. Eng. 7, 119–132. Emam, A. R., Kappas, M. & Hosseini, S. Z. Assessing the impact of climate change on water resources, crop production and land degradation in a semi-arid river basin. Hydrol. Res. 46, 854–870. Falkenmark, M. Water security for the 21st century – Building bridges through dialogue. In: Proceedings of the 11th Stockholm Water Symposium, Stockholm, Sweden, 13–16 August 2001. Water Sci. Technol. 45, VII–VIII. Falkenmark, M. & Lundqvist, J. Towards water security: political determination and human adaptation crucial. Econ. Water Resour. 2, 489–503. Fang, Q., Wang, G., Liu, T., Xue, B.-L. & Yinglan, A. a Controls of carbon flux in a semi-arid grassland ecosystem experiencing wetland loss: vegetation patterns and environmental variables. Agric. For. Meteorol. 259, 196–210. Fang, Q., Wang, G., Xue, B., Liu, T. & Kiem, A. b How and to what extent does precipitation on multi-temporal scales and soil moisture at different depths determine carbon flux responses in a water-limited grassland ecosystem? Sci. Total Environ. 635, 1255–1266. Gain, A. K., Giupponi, C. & Wada, Y. Measuring global water security towards sustainable development goals. Environ. Res. Lett. 11, 124015. Han, D., Wang, G., Liu, T., Xue, B.-L., Kuczera, G. & Xu, X. Hydroclimatic response of evapotranspiration partitioning to prolonged droughts in semiarid grassland. J. Hydrol. 563, 766–777. Harris, J. M. & Kennedy, S. Carrying capacity in agriculture: global and regional issues. Ecol. Econ. 29, 443–461. Jiang, Y. China’s water security: current status, emerging challenges and future prospects. Environ. Sci. Policy 54, 106–125. Kawada, Y., Nakamura, Y. & Otani, S. An axiomatic foundation of the multiplicative human development index. Rev. Income Wealth 65, 771–784. Kumar, P. Hydrocomplexity: addressing water security and emergent environmental risks. Water Resour. Res. 51, 5827–5838. Larson, R. B. Water security. Northwest. Univ. Law Rev. 112, 139–200.
866
J. Yao et al.
|
Identification of regional water security issues in China
Ledingham, J., Archer, D., Lewis, E., Fowler, H. & Kilsby, C. Contrasting seasonality of storm rainfall and flood runoff in the UK and some implications for rainfall-runoff methods of flood estimation. Hydrol. Res. 50, 1309–1323. Li, J., Guan, X., Wang, G. & Jiang, H. Assessment on agricultural non-point source pollution in a drinking water source using remote sensing images. Chin. J. Ecol. 35, 3382–3392. Li, W., Yao, Y. & Shekhar, D. K. State-space model for transient behavior of membrane-based liquid desiccant dehumidifier. Int. J. Heat Mass Transfer 144, 118711. Liu, F., Song, X., Yang, L., Han, D., Zhang, Y., Ma, Y. & Bu, H. a Predicting the impact of heavy groundwater pumping on groundwater and ecological environment in the Subei Lake basin, Ordos energy base, Northwestern China. Hydrol. Res. 49, 1156–1171. Liu, H., Jia, Y., Niu, C., Gan, Y. & Xu, F. b Evaluation of regional water security in China based on dualistic water cycle theory. Water Policy 20, 510–529. Masseroni, D., Ercolani, G., Chiaradia, E. A. & Gandolfi, C. A procedure for designing natural water retention measures in new development areas under hydraulic-hydrologic invariance constraints. Hydrol. Res. 50, 1293–1308. Norman, E. S., Dunn, G., Bakker, K., Allen, D. M. & de Albuquerque, R. C. Water security assessment: integrating governance and freshwater indicators. Water Resour. Manage. 27, 535–551. Ou, B., Fu, S., Wang, Y. & Wang, L. The comprehensive evaluation of rural drinking water security in Yunnan Province. In: 2012 International Conference on Structural Computation And Geotechnical Mechanics (P. Yang & X. Jiang, eds). People’s Republic of China, Kunming, pp. 155–158. Parry, S., Hannaford, J., Lloyd-Hughes, B. & Prudhomme, C. Multi-year droughts in Europe: analysis of development and causes. Hydrol. Res. 43, 689–706. Ren, L., Li, Q. & Yuan, F. The hydrological cycle and water security in a changing environment in China. Hydrol. Res. 43, 1–2. Ren, Y., Yao, J., Xu, D. & Wang, J. A comprehensive evaluation of regional water safety systems based on a similarity cloud model. Water Sci. Technol. 76, 594–604. Rijsberman, M. A. & van de Ven, F. H. M. Different approaches to assessment of design and management of sustainable urban water systems. Environ. Impact Assess. Rev. 20, 333–345. Russo, T., Alfredo, K. & Fisher, J. Sustainable water management in urban, agricultural, and natural systems. Water 6, 3934–3956. Sullivan, C. Calculating a water poverty index. World Dev. 30, 1195–1210.
Hydrology Research
|
51.5
|
2020
Tian, J. & Gang, G. Research on regional ecological security assessment. In: 2012 International Conference on Future Energy, Environment And Materials (G. Yang, ed.). Pt B. Hong Kong, People’s Republic of China, pp. 1180–1186. Wang, G., Hu, X., Zhu, Y., Jiang, H. & Wang, H. Historical accumulation and ecological risk assessment of heavy metals in sediments of a drinking water lake. Environ. Sci. Pollut. Res. 25, 24,882–24,894. Wang, G., Li, J., Sun, W., Xue, B., Yinglan, A. & Liu, T. a Nonpoint source pollution risks in a drinking water protection zone based on remote sensing data embedded within a nutrient budget model. Water Res. 157, 238–246. Wang, P., Yao, J., Wang, G., Hao, F., Shrestha, S., Xue, B., Xie, G. & Peng, Y. b Exploring the application of artificial intelligence technology for identification of water pollution characteristics and tracing the source of water quality pollutants. Sci. Total Environ. 693, 133440. Wang, Q., Liu, Y.-y., Zhang, Y.-z., Tong, L.-j., Li, X., Li, J.-l. & Sun, Z. c Assessment of spatial agglomeration of agricultural drought disaster in China from 1978 to 2016. Sci. Rep. 9, 14393. Wilkinson, M. E. & Bathurst, J. C. A multi-scale nested experiment for understanding flood wave generation across four orders of magnitude of catchment area. Hydrol. Res. 49, 597–615. Yang, H., Wang, G., Wang, L. & Zheng, B. Impact of land use changes on water quality in headwaters of the Three Gorges Reservoir. Environ. Sci. Pollut. Res. 23, 11,448–11,460. Yao, J., Wang, G., Xue, B., Wang, P., Hao, F., Xie, G. & Peng, Y. a Assessment of lake eutrophication using a novel multidimensional similarity cloud model. J. Environ. Manage. 248, 109259. Yao, J., Wang, G., Xue, W., Yao, Z. & Xue, B. b Assessing the adaptability of water resources system in Shandong Province, China, using a novel comprehensive co-evolution model. Water Resour. Manage. 33, 657–675. Yao, J., Wang, P., Wang, G., Shrestha, S., Xue, B. & Sun, W. c Establishing a time series trend structure model to mine potential hydrological information from hydrometeorological time series data. Sci. Total Environ. 698, 134227. Zhang, Y., Wang, Y., Chen, Y., Liang, F. & Liu, H. Assessment of future flash flood inundations in coastal regions under climate change scenarios – a case study of Hadahe River basin in northeastern China. Sci. Total Environ. 693, 133550. Zhao, L., Gong, D., Zhao, W., Lin, L., Yang, W., Guo, W., Tang, X. & Li, Q. Spatial-temporal distribution characteristics and health risk assessment of heavy metals in surface water of the Three Gorges Reservoir, China. Sci. Total Environ. 704, 134883.
First received 19 January 2020; accepted in revised form 25 March 2020. Available online 11 May 2020
867
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Copula-based drought severity-area-frequency curve and its uncertainty, a case study of Heihe River basin, China Zhanling Li, Quanxi Shao, Qingyun Tian and Louie Zhang
ABSTRACT Copulas are appropriate tools in drought frequency analysis. However, uncertainties originating from copulas in such frequency analysis have not received significant consideration. This study aims to develop a drought severity-areal extent-frequency (SAF) curve with copula theory and to evaluate the uncertainties in the curve. Three uncertainty sources are considered: different copula functions, copula parameter estimations, and copula input data. A case study in Heihe River basin in China is used as an example to illustrate the proposed approach. Results show that: (1) the dependence structure of drought severity and areal extent can be modeled well by Gumbel; Clayton and Frank depart the most from Gumbel in estimating drought SAF curves; (2) both copula parameter estimation and copula input data contribute to the uncertainties of SAF curves; uncertainty ranges associated with copula input
Zhanling Li (corresponding author) Qingyun Tian School of Water Resources and Environment, MOE Key Laboratory of Groundwater Circulation and Environmental Evolution, China University of Geosciences, Beijing 100083, China E-mail: zhanling.li@cugb.edu.cn Quanxi Shao Louie Zhang CSIRO Data 61, Private Bag 5, Wembley, WA 6913, Australia
data present wider than those associated with parameter estimations; (3) with the conditional probability decreasing, the differences in the curves derived from different copulas are increasing, and uncertainty ranges of the curves caused by copula parameter estimation and copula input data are also increasing. These results highlight the importance of uncertainty analysis of copula application, given that most studies in hydrology and climatology use copulas for extreme analysis. Key words
| copula, drought, input data, parameter estimation, SPEI, uncertainty
INTRODUCTION Drought is a stochastic natural phenomenon that arises
Previous literature has been most concerned with the
from considerable deficiency in precipitation. With the
severity and the duration of drought due to their ability to
global temperature increasing, it becomes more common
characterize drought from the temporal scale (e.g., Lee &
and has great effects on terrestrial ecosystems (Mishra &
Kim ; Beguería et al., ; Rahmat et al. ; Tan
Singh ; Fang et al. a; Han et al. ). The frequency
et al. ; Afshar et al. ; Homdee et al. ; Fang
analysis of drought is important since it is capable of quan-
et al. b; Hameed et al. ). However, the spatial
tifying the drought characteristics (Rahmat et al. ).
extent of drought is also very important in drought assess-
Bivariate or even multivariate frequency analysis is more
ments. Recent researches have shown that the drought
informative than univariate analysis due to the complexity
area has been significantly increasing globally and regionally
and multivariate features of drought (Yu et al. ; Jiang
during the last several decades. The percentage of areas
et al. ; Rahmat et al. ; Amirataee et al. ).
under drought had increased by about 1.74% per decade from 1950 to 2008 globally (Dai ). The drought-affected
This is an Open Access article distributed under the terms of the Creative
area increased nearly 12-fold from 1950 to 2009 in China
Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying
(Wang et al. ). A significant increase in drought area
and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/ licenses/by-nc-nd/4.0/). doi: 10.2166/nh.2020.173
has been detected in the northwest of China, with an increasing rate of 3.96% per decade since the late 1990s
868
Z. Li et al.
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
(Yu et al. ). Therefore, taking both severity and areal
functions based on Bayesian framework; whereas the
extent into account is very necessary for better understand-
fixed drop-down menu options in the toolbox and the non-
ing drought from both a temporal and spatial scale, and it
editable display for the outputs restrict its wide application
would be valuable for drought-based decision-making.
to some extent.
Applying the concept of quantifying the frequency
This paper aims to develop a drought SAF curve with
analyses of both severity and areal extent based on the
copula theory and to evaluate the uncertainties in the
copula theory, the drought severity-area-frequency (SAF)
curve
curves can be derived. These curves are very useful in
parameter estimations, and copula input data. The following
drought planning and management since they can provide
issues will be addressed, namely: (i) to compare and select
the probabilities and return periods of the drought areal
the copula function with best performance in modeling
extent with different severities over a region (Reddy &
the dependence structure of drought severity and drought
Ganguli ; Amirataee et al. ). However, uncertainties
areal extent; (ii) to construct a drought SAF curve based
are usually involved in the estimation of drought SAF curves
on the conditional probability distribution in the study
due to the effects of different copula functions, marginal
area; and (iii) to quantify the uncertainties in drought SAF
distributions, input data, parameter estimations, etc.
curves from three uncertainty sources.
caused
by
different
copula functions,
copula
The former uncertainty sources, copula functions and marginal distributions, are of major concern in previous literatures. For example, Halwatura et al. () compared four marginal distribution functions and two types of
METHODS
copulas, and assessed the impacts of these uncertainties on the estimations of recurrence intervals (RIs) of drought
Drought index and its characteristics
events. They stated that no differences were found between the RIs derived from the different marginal distributions
A number of indices have been developed so far for drought
for mild drought events, and the effects of different copulas
quantification, monitoring, and analysis, such as PDSI
on RIs can nearly be negligible in their study. However,
(Palmer Drought Severity Index), SPI (Standardized
Zhao et al. () drew a slightly different conclusion by
Precipitation Index), and SPEI (Standardized Precipitation-
comparing six marginal distributions and three Archime-
Evapotranspiration Index), etc. Under the current climate
dean copulas in a case study of the Weihe River basin.
warming conditions, many studies show that, SPEI, devel-
They pointed out that the return period of drought varies
oped by Vicente-Serrano et al. (), could give a better
depending on the selected marginal distributions and
performance in drought assessments due to the fact that it
copula functions, and the differences become larger with
combines the multi-scalar character with the capacity of
the increase of return periods. Zhang et al. () evaluated
involvement of temperature effects on droughts (e.g., Tan
the influences of marginal distributions to the uncertainty of
et al. ; Homdee et al. ). Thus, SPEI was selected
joint distribution by employing a Bayesian framework, and
and 12-monthly scale of SPEI will be analyzed in this
the results showed that the stronger the tail of marginal
study. The temperature-based Thornthwaite method is
distribution, the greater the uncertainty of joint distribution,
used to calculate potential evapotranspiration (PET) as it
especially for extreme events. All the above studies made
only requires monthly mean temperature data and the latitu-
great contributions in enriching the uncertainty analysis
dinal coordinate of the location, which are readily available
in drought frequency curves. However, the uncertainties
at most meteorological stations. SPEI is computed with the
caused by copula parameter estimations and copula input
assumption that P-PET (precipitation minus potential eva-
data are rarely regarded. Sadegh et al. () presented a
potranspiration) series follow the Log-logistic distributions.
multivariate copula analysis toolbox which has the ability
Detailed information about SPEI and its calculation can
to quantify the uncertainties in drought frequency curves
be found in the studies of Vicente-Serrano et al. (),
associated with copula parameter estimations and copula
Beguería et al. (), and Hameed et al. ().
Z. Li et al.
869
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
Specifically, drought characteristics used here include drought severity (S) and drought areal extent (A). The sever-
(IDW)
spatial
interpolation
technology
|
is
51.5
|
2020
generally
employed to obtain the spatial distribution of SPEI values.
ity of drought events is defined as the average of SPEI values of all cells in the study area, which is located below the drought threshold (Amirataee et al. ). It is computed using the equation:
Suppose X and Y are two continuous random variables with marginal distribution functions CDFs u1 and u2 , then there
PN St ¼
Copula-based conditional distribution functions
i¼1
I[SPEIi,t SPEIthr ] × SPEIi,t for t ¼ 1, . . . T NCt
exists a copula C such that: (1)
where I[:] is the logical indicator function with the value 1 if its argument is true, and the value 0 otherwise. SPEIi,t is the
FXY (x, y) ¼ P(X x, Y y) ¼ C(u1 , u2 )
(3)
The conditional density functions are related to C(u1 , u2 ; θ) through:
value of SPEI in cell i and period t, SPEIthr is the threshold of SPEI. NCt represents the total number of drought cells in period t, N means the total number of cells in the study
C(u2 ju1 ; θ) ¼ P(U2 u2 jU1 ¼ u1 ) ¼
@ 2 C(u1 , u2 ; θ) @u1 @u2
(4)
area, T means the data record length. Positive SPEI values indicate wet periods, while negative values indicate dry
and when it is integrated with respect to u2 ,
periods compared with the normal conditions of the area (Vicente-Serrano et al. ). Light drought ( 0.5 to 0.99), moderate drought ( 1.0 to 1.49), severe drought ( 1.5 to 1.99), and extreme drought ( 2.0) are defined
uð2
C(u2 ju1 ; θ) ¼
tive and falls below a certain threshold value. A threshold
C(vju1 ; θ)dv ¼ 0
according to SPEI values (Vicente-Serrano et al. ). Drought events occur when the SPEI is continuously nega-
uð2
¼
@ @u1
0 uð2
0
@ 2 C(u1 , v; θ) dv @u1 @v
@C(u1 , v; θ) @C(u1 , u2 ; θ) dv ¼ @v @u1
(5)
value of 0.5 is selected here. In order to represent the extreme condition and to analyze the associated risk of droughts using the exceedance probability, the negative drought severity is transformed to positive values for fitting to the available distributions (Mishra & Singh ; Amirataee et al. ), thus, there is a negative sign in Equation (1). The areal extent of drought is defined as a specified percent of total area in period t in which the value of SPEI is below the specified threshold (Reddy & Ganguli ; Amirataee et al. ). It is computed using the following equation: PN At ¼
i¼1
I[SPEIi,t SPEIthr ] × ai for t ¼ 1, . . . T PN i¼1 ai
where (U1 , U2 ) ∼ C, and C is a bivariate copula distribution function with parameter θ, representing the bivariate dependence structure of variables X and Y. C(u2 ju1 ; θ) is the numeric vector of the conditional distribution function of the copula with parameter θ evaluated at u2 given u1 . Different copulas differ in describing the dependence structures. Comparisons of copulas could be necessary before bivariate drought frequency analysis. Ten copulas are selected and given in Table 1, including not only oneparameter Archimedean families such as Clayton, Gumbel, Frank, and Joe, but also two-parameter copulas, BB1
(2)
(Clayton–Gumbel), BB6 (Joe–Gumbel), BB7 (Joe–Clayton), and BB8 (Joe–Frank). The more flexible structures of twoparameter copulas allow for different non-zero lower
where ai is the area of cell i. Since SPEI is calculated orig-
and upper tail dependence coefficients (Brechmann &
inally at a limited number of sites where observations on
Schepsmeier ). Many of the above copulas are com-
climate variables are available, inverse distance weighting
monly used in hydrological and climatological applications
Z. Li et al.
870
Table 1
|
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
The cumulative density functions and parameter ranges of alternative copulas
No.
Copula family
C(u1 , u2 )
1
Normal
Φθ (Φ 1 (u1 ), Φ 1 (u2 ))
1 θ 1
2
Student-t
tθ, δ (t 1 δ (u1 ),
θ>0
3
Clayton
4
Gumbel
Parameter space
t 1 δ (u2 ))
θ>0
θ θ u θ 1 þ u2 1 1 exp [( lnu1 )θ þ ( lnu2 )θ ]θ 1
θ 1
5
Frank
1 (e θu1 1)(e θu2 1) ln 1 þ θ e θ 1
6
Joe
1 [(1 u1 )θ þ (1 u2 )θ (1 u1 )θ (1 u2 )θ ]θ
θ 1
7
BB1
1δ 1 θ θ θ δ 1 þ [(u δ 1 1) þ (u2 1) ]
δ > 0, θ 1
8
BB6
1θ 1 δ δ δ 1 1 exp {[ ln(1 (1 u1 )θ )] þ [ ln(1 (1 u2 )θ )] }
δ > 0, θ 1
9
BB7
1 1δ θ δ δ 1 1 [(1 (1 u1 )θ ) þ (1 (1 u2 )θ ) 1]
δ > 0, θ 1
10
BB8
θ≠0 1
0 " #1θ 1 1@ (1 (1 δu1 )θ )(1 (1 δu2 )θ ) A 1 1 δ 1 (1 δ)θ
δ > 0, θ 1
Note: For Normal, Φθ is the bivariate joint normal distribution with linear correlation coefficient θ, and Φ is standard normal marginal distribution. Formulas in this table are from Chen & Khashanah (2015).
(e.g., Xu et al. ; Zhang et al. , ; Zhao et al. ;
The AIC and BIC are defined as:
Ayantobo et al. ). Copula parameter estimation and goodness of fit
AIC ¼ 2D 2‘
(6)
BIC ¼ Dlnn 2‘
(7)
The copula parameter θ (or θ and δ) in Table 1 helps to measure the extent of relationship between variables
where D is the number of parameters in the model, ‘ is the
(Ayantobo et al. ). Its estimation from sample data is
log-likelihood value of the best parameter set, and n rep-
not straightforward. One of the commonly used approaches
resents the sample size. The derivations of both equations
is maximum likelihood estimation (MLE) (Brechmann &
appear to be similar. A lower AIC or BIC value associates
Schepsmeier ). Although the solution to MLE involves
with a better copula model. The AIC takes into account
the roots of a cubic equation, and cannot be written in a
both the complexity of the model and minimization of
simple form, the MLE can be easily found using numerical
model error residuals and provides a more robust measure
routines. The standard errors for MLE copula parameter
of quality of model predictions. It avoids the problem of
estimations are based on inversion of the Hessian matrix
over-parameterization by adding a penalty term based on
(Brechmann & Schepsmeier ).
the number of parameters (Sadegh et al. ; Sun et al.
The best copula is selected according to the Akaike and
). The BIC equation differs from the AIC equation
Bayesian information criteria (AIC and BIC, respectively).
with respect to its first term that depends on the sample
871
Z. Li et al.
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
size. The penalty term of BIC is larger than that of AIC,
2. Using the probability distribution type which has been
when the number of samples is too large, it can effectively
pre-specified according to the original severity and areal
prevent the model complexity from being too high due to
extent series, to fit the Sj* and Aj* for obtaining their opti-
the excessive model accuracy (Brechmann & Schepsmeier
mal parameter estimates and the corresponding marginal
).
distributions, CDF(Sj*) and CDF(Aj*). 3. Using CDF(Sj*), CDF(Aj*) and the best candidate copula pre-selected according to the original marginal distrij butions, to estimate the copula parameters θ^ (i ¼ 1, or
Uncertainty analysis
i
i ¼ 1, 2).
According to MLE, the optimal value of copula parameter
4. Calculating drought severities at given drought areal
can be derived, together with its standard errors which
extents according to the conditional probability func-
can be used to represent the parameter uncertainty ranges
tions, and then obtaining the drought curve SAFj*.
(without considering the uncertainty in marginal distributions in severity and areal extent). The effects of copula parameter estimations to drought SAF curves are quantified by random generating copula parameters based on the asymptotic normality first, then calculating the drought curves according to the new generating copula parameter values, and finally identifying the α=2th and (1 α)=2th
5. The above procedure is repeated m times to generate m bootstrap samples, m estimations of marginal distributions for severity and areal extent, m sets of copula j parameters θ^ , and m drought curves SAFj* ( j ¼ 1, 2, i
…, m). 6. From drought curves SAFj* ( j ¼ 1, …, m), identifying the α=2th and (1 α)=2th percentiles of drought curves
percentiles of drought curves which are defined as the
which is defined as the lower and upper limits of confi-
lower and upper limits of confidence intervals with given
dence intervals with given probabilities, and then
probabilities. These confidence intervals are the uncertainty
analyzing the contributions of copula input data to the
ranges of drought curves caused by copula parameter
uncertainty of drought curves.
estimations. To quantify the effects of copula input data uncertainty
The main steps of this study are as follows. The values of
to drought SAF curves, the non-parametric bootstrap
12-month SPEI at the stations are calculated first. Then, the
method, which has usually been used in quantifying the
drought severity and drought areal extent are determined
sample uncertainty in previous studies (e.g., Hu et al. ;
according to IDW spatial interpolation technique. The two
Vergni et al. ), is selected. The idea behind this
drought variables are fitted by a variety of theoretical
method is that since the distribution from which the
probability distributions. After that, copulas are used to
sample has been taken is unknown, the values in that
compare the performances of capturing the dependence
sample are the best guide to the true distribution. The theor-
structures of the two drought variables with goodness of
etical details of the bootstrap approach can be found in
fit tests. Drought SAF curves are then extracted with the
Efron (). The copula input data in this case consists of
best copula. Finally, the uncertainties in SAF curves
n sets of drought severity (S) and drought areal extent (A),
caused by different copulas, copula parameter estimations,
denoted as (S1, A1), (S2, A2), …, (Si, Ai), …, (Sn, An), in
and copula input data, are investigated.
which n means the data length. The Si and Ai series are fitted by theoretical probability distributions and the best distribution is selected based on goodness of fit tests. The
STUDY AREA AND DATA
specifics of bootstrap procedure are as follows: 1. In the jth bootstrap, resampling the original n set values
The Heihe River basin, which falls between 97.1 –102.0 E
of drought variables with replacement, and generating a
and 37.7 –42.7 N, is the second largest typical inland river
j*
bootstrapped sample of the same size n as ((S1 , (S2j*, A2j*), ……, (Sij*, Aij*), ……, (Snj*, Anj*).
A1j*),
basin in the northwest of China with a controlled catchment area of 143,000 km2 (Figure 1). It is one of the most drought-
Z. Li et al.
872
Figure 1
|
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
Locations of the meteorological stations in Heihe River basin.
prone river basins in China (Wang et al. ; Ayantobo
The locations and the detailed information of stations are
et al. ; Niu et al. ) with an annual precipitation of
shown in Figure 1 and Table 2.
approximately 150 mm across the basin, an annual average air temperature of 6–8 C, and annual potential evaporation of 1,000–2,000 mm. The runoff of the basin mainly comes
RESULTS AND ANALYSIS
from the upper reach, and consumed by large areas of irrigation in the middle reach and oasis ecosystem in the lower
Drought characteristics and their dependence
reach. Severe drought events would lead to inadequate
structures
water supply and thus affect the local agriculture and vulnerable oasis ecosystem. In this study, monthly precipitation
Drought severity and areal extent series over the study area
and temperature data spanning from 1960 to 2015 were col-
determined by Equations (1) and (2) are plotted in Figure 2.
lected from nine meteorological stations within the basin.
It is clear that, before 1997, the drought was mainly
Table 2
|
Climatic and geographical characteristics of stations used in the study during 1960–2015
No.
Station
Longitude
Latitude
Elevation (mm)
Annual precipitation (mm)
Annual mean temperature ( C)
1
Tuole
98.25
38.48
3,367
298
2.5
2
Yeniugou
99.36
38.26
3,320
418
2.8
3
Qilian
100.15
38.11
2,787
409
1.1
4
Shandan
101.05
38.48
1,765
201
6.6
5
Zhangye
100.17
39.05
1,483
128
7.5
6
Jiuquan
98.29
39.46
1,477
86
7.6
7
Gaotai
99.50
39.22
1,332
108
7.9
8
Dingxin
99.31
40.18
1,177
55
8.5
9
Ejinaqi
101.04
41.57
941
35
9.0
873
Z. Li et al.
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
previous studies (e.g., Loukas & Vasiliades ; Serinaldi et al. ; Reddy & Ganguli ; Afshar et al. ; Amirataee et al. ), ten univariate probability distributions
(Exponential,
Normal,
Lognormal,
Logistic,
Loglogistic, Gamma, Weibull, Log-Pearson III, Burr, and GEV distribution) are tested for fitting drought severity (S). Since the drought areal extent has definite ranges from 0 to 1, the bounded distributions, Beta, Power function, Pert, and Uniform, are tested as its candidate margins. Figure 2
|
Amirataee et al. () stated that Beta distribution Drought characteristics of Heihe River basin during the period of 1960–2015.
performed satisfactorily in fitting the drought area extent in their study. Kolmogorov–Smirnov (K-S), Anderson–
characterized by light drought conditions. However, several
Darling (A-D), and Chi-squared (C-S) are used to test the
moderate or even severe droughts were captured after 1997,
goodness of fit.
and furthermore, drought events with stronger severity and larger areal extent occurred more frequently than before.
Results are shown in Table 3. The critical values for the three tests are 0.061, 2.502, and 15.507 at 0.05 significance
It is necessary to evaluate the dependency between vari-
level, respectively. As shown, GEV is selected finally as
ables prior to the application of copula when analyzing the
the marginal distribution for drought severity, because it
bivariate joint probability distribution of such random vari-
not only passes the K-S, A-D, and C-S tests at 0.05 signifi-
ables (Amirataee et al. ). For this purpose, Pearson’s,
cance levels (all the statistic values are lower than the
Spearman’s, and Kendall’s correlation coefficients are
critical values for the three tests), but shows the best
applied to explore the dependency between drought severity
performance with the smallest statistic values. This
and drought areal extent. These correlation coefficients are
distribution was also proved to be suitable in fitting the
often used in measuring the monotone association. The
drought severity in many other studies (e.g., Lee & Kim
first belongs to the parametric method and sensitive to out-
; Reddy & Ganguli ). As to the variable of drought
liers, and the latter two belong to the non-parametric
areal extent, none of the distributions passes the tests.
methods and usually suggested for non-normally distributed
Then, we perform the fitting process based on its pseudo
data (Shong ).
observations (Ghoudi & Rémillard ). In light of the
The correlation coefficients between the two drought
references of Ghoudi & Rémillard () and Kojadinovic
variables are 0.678, 0.722, and 0.549, respectively, and all
et al. (), it is acceptable to replace the unknown mar-
the corresponding p-values are less than the significance
ginal by their empirical counterparts obtained from the
level of 0.05. It is concluded that there is a significant corre-
pseudo observations.
lation between the variables of drought severity and drought areal extent at 95% confidence level.
Goodness of fit of copula functions
Marginal distributions and copula goodness of fit
The results of modeling the joint probability distribution of the two drought variables by using ten alternative copula functions are summarized in Table 4, including the statistic
Determination of marginal distribution for drought variables
values of goodness of fit tests (AIC, BIC, and log-likelihood), the optimal parameter estimations, and their standard errors.
The first step in copula fitting consists of the identification of
Clearly, nearly all the copulas have close AIC, BIC, and log-
appropriate marginal distributions for the two drought vari-
likelihood values except for Frank which shows the greatest
ables. Since they are typically unknown, we first fit them by
departure from the others, indicating that most of the selected
using
copulas
several
theoretical
distributions.
According
to
have
similar
performance
in
capturing
the
Z. Li et al.
874
Table 3
|
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
Goodness of fit of theoretical probability distributions for drought severity (S) and drought areal extent (A)
Drought variable
No.
Probability distribution
Kolmogorov–Smirnov (K-S)
Anderson–Darling (A-D)
Chi-squared (C-S)
Drought severity (S)
1 2 3 4 5 6 7 8 9 10
Exponential Normal Lognormal Logistic Loglogistic Gamma Weibull Log-Pearson III Burr GEV
0.4475 0.0680 0.0434 0.0776 0.0580 0.0452 0.0841 0.0431 0.0481 0.0387
125.74 3.87 1.55 5.28 2.93 1.44 7.45 1.60 2.10 1.30
810.61 10.14 10.50 24.32 24.79 9.48 23.22 12.14 17.74 8.79
Drought areal extent (A)
1 2 3 4
Beta Power function Pert Uniform
0.1806 0.1266 0.1424 0.1767
35.46 38.91 37.16 16.43
101.81 112.88 188.62 48.12
Note: The minimum statistic values for K-S, A-D, and C-S tests are in bold italics.
Table 4
|
Goodness of fit of different copulas and their optimal parameter estimations
No.
Copula
Akaike information criteria (AIC)
Bayesian information criteria (BIC)
Log likelihood
Optimal parameter (standard error)
1
Normal
465.60
461.38
233.80
0.79 (0.006)
2
Student-t
465.43
460.00
234.72
0.79, 7.98 (0.019, 6.082)
3
Clayton
442.62
438.41
222.31
2.01 (0.415)
4
Gumbel
473.45
469.24
237.73
2.27 (0.114)
5
Frank
364.49
360.27
183.24
6.24 (1.637)
6
Joe
433.61
429.40
217.81
2.77 (0.190)
7
BB1
490.71
482.28
247.36
0.92, 1.60 (0.138, 0.096)
8
BB6
471.45
463.01
237.72
1.00, 2.27 (0.230, 0.350)
9
BB7
514.65
506.21
259.33
2.09, 1.82 (0.127, 0.151)
10
BB8
435.22
426.79
219.61
2.89, 0.99 (0.145, 0.004)
Note: The minimum AIC, BIC and log likelihood values are in bold italics.
dependence structures of drought severity and drought areal
of 5,000 samples are generated, and the scatter plot of gener-
extent in this study case. Comparatively, Gumbel, among the
ated samples, together with the observed datasets are shown
one-parameter copula functions, provides the best fit with the
in Figure 3. Different copula functions perform a little differ-
smallest AIC ( 473.45) and BIC ( 469.24) values. Among the
ently. The Normal, Student-t, and Frank are symmetric and
two-parameter copulas, BB7 performs the best. Student-t and
can be used to model symmetric correlation patterns of vari-
BB8 copulas even perform inferior to Gumbel according to
ables. Gumbel and Clayton are asymmetric and can model
the corresponding AIC and BIC values.
the asymmetrical tail correlation patterns. Gumbel exhibits
To better understand the performance of each copula in modeling the drought variables in this study, the generated
greater correlation in the upper tail, whereas Clayton exhibits greater correlation in the lower tail (Aas et al. ).
samples derived from the estimated theoretical copula by
From the scatterplot, the observed data in the two oppo-
Monte Carlo technique are compared with the observed
site corners appear clustering, seemingly displaying both
data for checking whether they have similar features. A set
upper- and lower-tail dependence, while the dependence
875
Figure 3
Z. Li et al.
|
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
Generated and observed CDF for drought severity and drought areal extent based on different copula functions. Gray dots present the Monte Carlo simulations from each copula and blue dots present the observed data. Please refer to the online version of this paper to see this figure in color: http://dx.doi.10.2166/nh.2020.173.
structures in the tails appear slightly asymmetric and a little
Drought SAF curves
skewed. Therefore, asymmetric copulas may be a good choice. Combined with the results of goodness of fit pre-
The drought SAF curve, developed by using conditional
sented in Table 3, Gumbel is then selected as the
probability function based on Gumbel copula, is plotted in
benchmark copula in this study. The two-parameter copulas,
Figure 4. The horizontal axis represents the percentage of
BB1, BB6, and BB7, are not selected herein since, on one
drought areal extent and the vertical axis represents the
hand, they show similar performance with Gumbel and,
drought severity. Only droughts with conditional probability
on the other hand, the higher model complexity (with
of less than 0.2 (corresponds to return period of more than
more parameters) would make a stimulation over condition-
five years) are considered in the analysis. It can be seen
ing of the model to some extent (Sadegh et al. ).
that the drought event with larger areal extent is generally
Actually, the optimality of copula functions could be dataset- and case-specific (Afshar et al. ). For example, Reddy & Ganguli () evaluated three copula families in modeling the joint dependence between the drought severity and drought areal extent in western Rajasthan of India, and found that the Gumbel–Hougaard copula best represents the drought properties in their study. Xu et al. () constructed a trivariate joint distribution based on the copula theory and concluded that the Joe and Gumbel copulas are more suitable to estimate the joint distribution of drought duration, affected area, and severity in southwest China. Amirataee et al. () regarded the Frank copula as the most appropriate in modeling the joint probability of drought severity and percent of area in the Lake Urmia basin, Iran. In our case, with considering the results of goodness of fit and the greater dependence in upper tails of the data, the one-parameter Gumbel is finally selected as the benchmark copula.
Figure 4
|
Drought severity-area-frequency (SAF) curves derived from Gumbel copula over the study area with six conditional probabilities (0.005, 0.01, 0.02, 0.05, 0.1, and 0.2).
876
Z. Li et al.
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
associated with higher severity. For example, with a con-
Clayton are 442.62 and 438.41 (Table 3), very close to
ditional probability of 0.1 (at the return period of ten
those for Gumbel ( 473.45 and 469.24). However, as
years), when the drought areal extent amounts to 20%, the
shown in Figure 3, the drought SAF curves derived from
average drought severity is 0.98, belonging to light drought;
these two copulas have a considerably larger difference,
while when the drought areal extent increases to 60% and
and Clayton gives larger possibilities of drought severity
80%, the severity increases to 1.17 and 1.30, resulting in
than Gumbel.
moderate droughts.
BB6 gives quite close drought severity estimations with
The drought SAF curve based on the conditional prob-
Gumbel no matter how much the conditional probability is.
ability can allow a water planner to assess the probability of
BB1, BB7, and Student-t copulas give similar curve shapes
a drought with a certain areal extent and severity occurring
with Gumbel, whereas they overestimate the drought severity
simultaneously. For example, for a drought areal extent reach-
overall. A similar conclusion can be drawn from Normal
ing 40%, the occurrence probability is 0.05 when the severity
when the drought areal extent is higher than 10%, overesti-
is 1.12, and 0.02 when the severity increases to 1.21. This
mating the drought severities comparatively. Joe and BB8
information cannot be obtained from the univariate fre-
overestimate the drought severity when the drought area is
quency analysis of droughts, in which only the probability
lower than about 10%, but underestimate the drought sever-
of univariate occurrence can be derived. The copula-based
ity when the drought areal extent becomes higher.
drought SAF curve can provide decision-makers with prob-
In summary, the drought SAF curves are influenced by
ability maps of drought severity and useful information on
copula choices, and the curves have different conditional
drought areal extent in the study area, yielding important
probabilities of occurrence depending on which copula is
information for water management purposes.
used. BB6 gives quite close estimations in drought severity to Gumbel. Normal, Student-t, BB1, BB7 overestimate the
Drought SAF curve uncertainties
drought severities, while Joe and BB8 underestimate the drought severities as a whole. Clayton and Frank copulas
Uncertainty caused by different copula functions
depart the most, and overestimate the drought severity much more. As the conditional probability decreases, the
In this subsection, whether different copulas lead to signifi-
differences in drought severity estimations from different
cantly different assessments of SAF curves is investigated.
copulas increase, although most of the differences are not
Figure 5 shows the conditional probability curves derived
large. It implies that the choice of copula should be
from different copula functions. Conditional probabilities of
implemented with caution since different copulas show
0.005, 0.01, 0.02, 0.05, 0.1, and 0.2 are compared. Gumbel
different performances in tail behaviors, especially when
copula is selected as the benchmark as stated above.
the focus is on extreme events (with small probabilities).
With a conditional probability of 0.2, most of the curves are close to each other, and only slight differences are found
Uncertainty caused by copula parameter estimations
between the benchmark and the alternatives, except for the two curves derived from Clayton and Frank. As the con-
After estimating the copula parameter(s) with the MLE
ditional probability decreases (conditional return period
method, 1,000 copula parameters are randomly generated
increases), these differences increase in general, being con-
from the Normal distribution with the mean and standard
sistent with the findings of Zhao et al. (), who pointed
error derived from MLE. The estimated optimal parameter
out that with an increase of the return period, the drought
value and standard error for Gumbel copula is 2.27 and
differences from different copula functions become larger.
0.114 (see Table 3). Such range of standard error in par-
Such differences are not very large in this study as a
ameter estimation is then translated to drought SAF
whole, except for those from Clayton and Frank.
curves, rendering the analysis of uncertainty. Combining
As to Clayton and Frank, large differences are detected.
with the conditional probability function and the generated
Interestingly, it is also noted that the AIC and BIC values for
copula parameters, a set of conditional probability isolines
Z. Li et al.
877
Figure 5
|
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
Drought severity-area-frequency (SAF) curves derived from different copula functions over the study area with six conditional probabilities (0.005, 0.01, 0.02, 0.05, 0.1, and 0.2).
are then derived, which will be used to analyze the SAF
SAF curves become wider with the conditional prob-
curve uncertainties caused by copula parameter estimations.
ability decreasing. This is because the lower the
Figure 6 presents the 95% CIs of conditional prob-
probability of the drought event, the less the number of
ability isolines associated with the uncertain parameter
the observed data, and the less information available
estimations of Gumbel. To avoid the curve overlapping,
for the analysis.
the six conditional probabilities are separated into two
Kendall’s tau method is also used to estimate the par-
plots. Obviously, the uncertainty in copula parameter
ameter in Gumbel, yielding an optimal parameter value of
estimation has a certain effect on the conditional
2.22 with a standard error of 0.130, which does not deviate
probability isolines. The uncertainty ranges of drought
much from the results of the MLE method. The corresponding
878
Figure 6
Z. Li et al.
|
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
Uncertainty ranges of drought severity-area-frequency (SAF) curve derived from copula parameter estimations over the study area with six conditional probabilities (0.005, 0.01, 0.02, 0.05, 0.1, and 0.2).
uncertainty ranges of conditional probability isolines are also quite close to the results from the MLE method. Uncertainty caused by copula input data Using bootstrap technique to generate 1,000 set of drought samples ((S1j*, A1j*), (S2j*, A2j*), …, (Sij*, Aij*), …, (S501j*, A501j*) (j ¼ 1, , 1,000)). To help visual comparisons of dependence structures among the original and the generated copula input data, the marginal distributions of the original samples and bootstrapped samples of drought severity and drought areal extent are plotted in Figure 7. The density of points in any area represents the density of the copula. It is observed that the bootstrapped samples well preserve
Figure 7
|
Dependence structures between drought severity and drought areal extent from 1,000 bootstrapped samples and the original samples.
the dependence structure between the drought severity and drought areal extent. Uncertainty bands in Figure 8 reveal the effects of copula input data uncertainty to the estimation of conditional drought severity. Obviously, with the conditional probability decreasing, the uncertainty bands are increasing, similar findings as those from the effects of copula parameter uncertainty. Table 5 presents the lower (2.5%
by copula input data present more widely. This reflects that the uncertainty of input data brings more uncertainty to drought SAF curves than that of copula parameter estimations. In other words, to improve the quality of copula input data can reduce the uncertainty of drought SAF curves more effectively.
percentile) and upper (97.5% percentile) limits and the ranges of estimated drought severity at a certain drought areal extent owing to both input data uncertainty and par-
CONCLUSIONS
ameter estimation uncertainty. The ranges resulting from both uncertainty sources grow remarkably as the con-
In this study, 12-month SPEI is calculated over the Heihe
ditional probability decreases. Compared with those
River basin in northwest China, the drought SAF curve is
caused by copula parameter, the uncertainty bands caused
developed, and then the uncertainties in the curve
Z. Li et al.
879
|
Figure 8
Table 5
|
|
Copula-based drought SAF curve and its uncertainty
Hydrology Research
|
51.5
|
2020
Uncertainty ranges of drought severity-area-frequency (SAF) curve derived from copula input data over the study area with six conditional probabilities (0.005, 0.01, 0.02, 0.05, 0.1, and 0.2).
Uncertainty bands of the estimated drought severity at given drought areal extent over the study area
Conditional probability Areal extent
0.005
0.01
0.02
0.05
0.1
0.2
20%-A
0.158 [1.088, 1.245] 0.133 [1.021, 1.155] 0.112 [0.964, 1.077] 0.088 [0.883, 0.971] 0.073 [0.820, 0.893] 0.059 [0.755, 0.814]
20%-B
0.108 [1.096, 1.204] 0.098 [1.030, 1.128] 0.083 [0.970, 1.053] 0.066 [0.888, 0.954] 0.050 [0.828, 0.879] 0.037 [0.763, 0.800]
40%-A
0.161 [1.117, 1.278] 0.140 [1.048, 1.188] 0.117 [0.991, 1.108] 0.088 [0.910, 0.999] 0.072 [0.848, 0.919] 0.058 [0.781, 0.839]
40%-B
0.114 [1.125, 1.240] 0.104 [1.060, 1.163] 0.087 [0.996, 1.083] 0.062 [0.917, 0.978] 0.049 [0.857, 0.906] 0.034 [0.791, 0.825]
60%-A
0.162 [1.134, 1.296] 0.141 [1.065, 1.206] 0.120 [1.006, 1.126] 0.090 [0.926, 1.016] 0.073 [0.863, 0.936] 0.058 [0.796, 0.854]
60%-B
0.109 [1.143, 1.252] 0.097 [1.079, 1.176] 0.081 [1.017, 1.098] 0.064 [0.933, 0.997] 0.043 [0.873, 0.915] 0.031 [0.807, 0.837]
80%-A
0.166 [1.148, 1.314] 0.146 [1.079, 1.224] 0.123 [1.019, 1.143] 0.093 [0.939, 1.032] 0.074 [0.877, 0.950] 0.059 [0.810, 0.868]
80%-B
0.119 [1.158, 1.277] 0.092 [1.093, 1.185] 0.082 [1.032, 1.114] 0.062 [0.949, 1.011] 0.042 [0.888, 0.930] 0.030 [0.822, 0.852]
Note: A and B mean the estimated drought severity uncertainties caused by copula input data (A) and copula parameter estimations (B); numbers in the brackets indicate the lower (2.5% percentile) and upper (97.5% percentile) limits, and those outside mean the ranges of 95% CIs.
associated with different copula functions, copula parameter
even presents quite close AIC and BIC values with
estimations, and copula input data are investigated.
Gumbel. The differences in drought SAF curves caused by
Results show that GEV is finally selected as the marginal
different copulas are increasing with the conditional prob-
distribution for drought severity, and pseudo observations
ability decreasing. As to the parameter estimations and
for drought areal extent. Gumbel shows the best perform-
input data, both of them result in the drought SAF curves
ance in capturing the dependence structure of drought
uncertainty. The uncertainty ranges caused by copula
severity and drought areal extent and thus is selected as
input data are larger than those caused by copula parameter
the benchmark copula. The estimations of drought SAF
estimations, and consequently, to improve copula input data
curves are influenced by copula choices. BB6 is capable of
quality is recommended for developing the reliability of such
giving close drought severity as Gumbel; Normal, Student-t,
frequency analysis. Results also show that the uncertainty
BB1, BB7 overestimate the drought severities; Joe and
bands in the drought SAF curves increase as the conditional
BB8 underestimate the drought severities when the drought
probability decreases. It indicates that larger uncertainties
areal extents become higher; Clayton and Frank produce the
exist in the frequency analysis of extreme values (with
most deviation in drought SAF curves although Clayton
small probability). This highlights the importance of
880
Z. Li et al.
|
Copula-based drought SAF curve and its uncertainty
uncertainty analysis in frequency study, given that most studies in hydrology and climatology focus on extreme values. Couple-based frequency analysis of drought with considering the uncertainty issues is capable of providing more beneficial probability information to water managers, and aiding in the development of mitigation strategies and disaster preparedness plans for different levels of drought.
ACKNOWLEDGEMENTS The first author is grateful for financial support from the China Scholarship Council and China University of Geosciences (Beijing), as well as the support provided by CSIRO during her visit in Australia. This study is supported by the Fundamental Research Funds for the Central Universities (No. 35832015028) and NSFC (No. 41101038). The authors also thank the anonymous referees for their comments and suggestions that have led to the quality of the paper being improved.
REFERENCES Aas, K., Czado, C., Frigessi, A. & Bakken, H. Pair-copula constructions of multiple dependence. Insurance: Mathematics and Economics 44 (2), 182–198. Afshar, M. H., Sorman, A. U. & Yilmaz, M. T. Conditional copula-based spatial-temporal drought characteristics analysis – a case study over Turkey. Water 8 (10), 426. Amirataee, B., Montaseri, M. & Rezaie, H. Regional analysis and derivation of copula-based drought severity-areafrequency curve in Lake Urmia basin, Iran. Journal of Environmental Management 206, 134–144. Ayantobo, O. O., Li, Y., Song, S., Javed, T. & Yao, N. Probabilitic modelling of drought events in China via 2dimensional joint copula. Journal of Hydrology 559, 373–391. Beguería, S., Vicente-Serrano, S. M., Reig, F. & Latorre, B. Standardized precipitation evapotranspiration index (SPEI) revisited: parameter fitting, evapotranspiration models, tools, datasets and drought monitoring. International Journal of Climatology 34 (10), 3001–3023. Brechmann, E. C. & Schepsmeier, U. Modeling dependence with C- and D-Vine copulas: the R package CDVine. Journal of Statistical Software 52 (3), 1–27. Chen, K. H. & Khashanah, K. Analysis of systemic risk: a dynamic Vine copula-based ARMA-EGARCH model. In: Transactions on Engineering Technologies, World Congress
Hydrology Research
|
51.5
|
2020
on Engineering and Computer Science (S.-I. Ao, H. K. Kim & M. A. Amouzegar, eds). Springer Nature, Singapore. Dai, A. Characteristics and trends in various forms of the palmer drought severity index during 1900–2008. Journal of Geophysical Research Atmospheres 16. https://doi.org/10. 1029/2010JD015541. Efron, B. Bootstrap methods: another look at the jackknife. Annals of Statistics 7, 1–26. Fang, Q. Q., Wang, G. Q., Xue, B. L., Liu, T. X. & Kiem, A. a How and to what extent does precipitation on multitemporal scales and soil moisture at different depths determine carbon flux responses in a water- limited grassland ecosystem? Science of the Total Environment 635, 1255–1266. Fang, Q. Q., Wang, G. Q., Liu, T. X., Xue, B. L. & Yinglan, A. b Controls of carbon flux in a semi-arid grassland ecosystem experiencing wetland loss: vegetation patterns and environmental variables. Agricultural and Forest Meteorology 259, 196–210. Ghoudi, K. & Rémillard, B. Empirical processes based on pseudo-observations II: the multivariate case. Asymptotic Methods in Stochastics 44, 381–406. Halwatura, D., Lechner, A. M., McIntyre, N. & Arnold, S. Uncertainties in estimating design droughts. 8th International Congress on Environmental Modelling and Software. International Environmental Modelling and Software Society (iEMSs), Toulouse, France. http://www. iemss.org/society/index.php/iemss-2016-proceedings. Hameed, M., Ahmadalipour, A. & Moradkhani, H. Apprehensive drought characteristics over Iraq: results of a multidecadal spatiotemporal assessment. Geosciences 8, 58. doi:10.3390/geosciences8020058. Han, D. M., Wang, G. Q., Liu, T. X., Xue, B. L., Kuczera, G. & Xu, X. Y. Hydroclimatic response of evapotranspiration partitioning to prolonged droughts in semiarid grassland. Journal of Hydrology 563, 766–777. Homdee, T., Pongput, K. & Kanae, S. A comparative performance analysis of three standardized climatic drought indices in the Chi River basin, Thailand. Agriculture and Natural Resources 50 (3), 211–219. Hu, Y., Liang, Z., Liu, Y., Wang, J., Yao, L. & Ning, Y. Uncertainty analysis of SPI calculation and drought assessment based on the application of Bootstrap. International Journal of Climatology 35 (8), 1847–1857. Jiang, C., Xiong, L., Xu, C. & Guo, S. Bivariate frequency analysis of nonstationary low-flow series based on the time-varying copula. Hydrological Processes 29 (6), 1521–1534. Kojadinovic, I., Yan, J., Leeuw, J. D. & Zeileis, A. Modeling multivariate distributions with continuous margins using the copula R package. Journal of Statistical Software 34 (9), 1346–1352. Lee, J. H. & Kim, C. J. Derivation of drought severity-durationfrequency curves using drought frequency analysis. Journal of Korea Water Resources Association 44 (11), 889–902.
881
Z. Li et al.
|
Copula-based drought SAF curve and its uncertainty
Loukas, A. & Vasiliades, L. Probabilistic analysis of drought spatiotemporal characteristics in Thessaly region, Greece. Natural Hazards & Earth System Sciences 4 (5/6), 719–731. Mishra, A. K. & Singh, V. P. Analysis of drought severity-areafrequency curves using a general circulation model and scenario uncertainty. Journal of Geophysical Research Atmospheres 114(D6). https://doi.org/10.1029/2008JD010986. Niu, J., Kang, S., Zhang, X. & Fu, J. Vulnerability analysis based on drought and vegetation dynamics. Ecological Indicators 105, 329–336. https://doi.org/10.1016/j.ecolind. 2017.10.048. Rahmat, S. N., Jayasuriya, N. & Bhuiyan, M. Development of drought severity-duration-frequency curves in Victoria, Australia. Australian Journal of Water Resources 19 (1), 31–42. Reddy, M. J. & Ganguli, P. Spatio-temporal analysis and derivation of copula-based intensity-area-frequency curves for droughts in western Rajasthan (India). Stochastic Environmental Research and Risk Assessment 27 (8), 1975–1989. Sadegh, M., Ragno, E. & Aghakouchak, A. Multivariate copula analysis toolbox (MvCAT): describing dependence and underlying uncertainty using a Bayesian framework. Water Resources Research 53, 5166–5183. Serinaldi, F., Bonaccorso, B., Cancelliere, A. & Grimaldi, S. Probabilistic characterization of drought properties through copulas. Physics & Chemistry of the Earth 34 (10–12), 596–605. Shong, N. Pearson’s Versus Spearman’s and Kendall’s Correlation Coefficients for Continuous Data. Master of Science Thesis, University of Pittsburg, Pittsburg, PA, USA. http://d-scholarship.pitt.edu/8056/1/Chokns_etd2010.pdf Sun, P., Zhang, Q., Yao, R., Singh, V. P. & Song, C. Low flow regimes of the Tarim River Basin, China: probabilistic behavior, causes and implications. Water 10 (4), 470. Tan, C., Yang, J. & Li, M. Temporal-spatial variation of drought indicated by SPI and SPEI in Ningxia Hui Autonomous Region, China. Atmosphere 6 (10), 1399–1421.
Hydrology Research
|
51.5
|
2020
Vergni, L., Lena, B. D., Todisco, F. & Mannocchi, F. Uncertainty in drought monitoring by the standardized precipitation index: the case study of the Abruzzo region (central Italy). Theoretical and Applied Climatology 128 (1), 13–26. Vicente-Serrano, S. M., Beguería, S. & López-Moreno, J. I. A multiscalar drought index sensitive to global warming: the standardized precipitation evapotranspiration index. Journal of Climate 23 (7), 1696–1718. Wang, X. J., Zhang, J. Y., Shahid, S., Elmahdi, A., He, R. M., Bao, Z. X. & Ali, M. Water resources management strategy for adaptation to droughts in China. Mitigation and Adaptation Strategies for Global Change 17 (8), 923–937. Wang, Y., Xu, Z., Zhang, B. & Li, Q. Monitoring the meteorological drought in the middle reaches of Heihe River basin based on TRMM precipitation data. 2016 IEEE International Geoscience and Remote Sensing Symposium, Beijing, China, pp. 4313–4316. doi: 10.1109/IGARSS.2016. 7730124. Xu, K., Yang, D., Xu, X. & Lei, H. Copula based drought frequency analysis considering the spatio-temporal variability in Southwest China. Journal of Hydrology 527, 630–640. Yu, M., Li, Q., Hayes, M. J., Svoboda, M. D. & Heim, R. R. Are droughts becoming more frequent or severe in China based on the standardized precipitation evapotranspiration index: 1951–2010? International Journal of Climatology 34 (3), 545–558. Zhang, Q., Xiao, M. & Singh, V. P. Uncertainty evaluation of copula analysis of hydrological droughts in the East River basin, China. Global and Planetary Change 129, 1–9. Zhang, D., Chen, P., Zhang, Q. & Li, X. Copula-based probability of concurrent hydrological drought in the Poyang lake-catchment-river system (China) from 1960–2013. Journal of Hydrology 553, 773–784. Zhao, P., Lü, H., Fu, G., Zhu, Y., Su, J. & Wang, J. Uncertainty of hydrological drought characteristics with copula functions and probability distributions: a case study of Weihe River, China. Water 9 (5), 334.
First received 29 November 2019; accepted in revised form 4 March 2020. Available online 6 May 2020
882
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Effects of antecedent soil water content on infiltration and erosion processes on loessial slopes under simulated rainfall Lan Ma, Junyou Li and Jingjing Liu
ABSTRACT Soil texture and antecedent soil water content (ASWC) are primary factors governing hillslope hydrological and erosion processes. We used simulated rainfall to investigate the runoff and erosion processes on sloped plots with three loessial soils and analyzed the effects of soil texture and ASWC on the hydrological processes. The results demonstrated that the average infiltration rate decreased with increasing clay content (i.e., Ansai (AS) loamy sand > Fuxian (FX) clay loam > Yangling (YL) clay). ASWC had little effect on infiltration processes for the YL clay but exerted a significant influence on
Lan Ma (corresponding author) Junyou Li Jingjing Liu Key Laboratory of State Administration of Forestry and Grassland on Soil and Water Conservation, School of Soil and Water Conservation, Beijing Forestry University, Beijing 100083, China E-mail: mlpcz@sina.com
infiltration for the FX and AS soils; this implies that infiltration models for loamy soils must consider the effects of ASWC. The Horton model was found to describe infiltration processes in these loessial soils better than the Kostiakov or Philip models. The YL clay yielded much less sediment than the FX and AS soils, and its sediment yield rate gradually decreased with the rainfall duration. There was a negative relationship between clay content and sediment yield under high ASWC, but no clear relation under low ASWC. These erosion differences derived from the splash erosion for the YL clay, and the depressions or rills occurred on the loamy soil plots. Key words
| antecedent soil water content, erosion, infiltration, simulated rainfall, soil texture
INTRODUCTION The Loess Plateau of China has an arid to semi-arid climate
important role in sustaining vegetation growth. In particular,
and has long experienced severe soil erosion due to its erod-
antecedent soil water content (ASWC) affects runoff gener-
ible soil texture, steep hillslopes, and frequent rain storms
ation and related soil erosion processes. Infiltration, runoff,
(Tang ). In fact, in some regions of steep terrain
and erosion processes within the Loess Plateau have been
within the plateau, soil loss rates reach up to 5,000–10,000
investigated previously to help conserve and use scarce
Mg km 2 a 1; such pronounced soil loss causes severe nutri-
soil and water resources across the region (Pan et al. ,
ent loss, induces widespread land degradation, and results in
; Wei et al. ; Mei et al. ).
frequent severe flooding along the Yellow River (Fu et al.
Soil infiltration processes under natural rainfall con-
). In arid and semi-arid areas, soil moisture plays an
ditions are complex, owing primarily to spatiotemporal variability in soil properties and rainfall (Geiger & Durnford
This is an Open Access article distributed under the terms of the Creative
; Dong et al. ). Typically, soil infiltration processes
Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying
are investigated by subjecting disturbed or undisturbed soil
and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/
columns to a stable hydraulic head or simulated rainfall;
licenses/by-nc-nd/4.0/).
the infiltrated water volume and the development of the
doi: 10.2166/nh.2020.013
883
L. Ma et al.
|
Effects of soil moisture on infiltration and erosion processes
Hydrology Research
|
51.5
|
2020
wetting front are commonly recorded (Youngs ). These
However, the infiltration of rainfall into soils is typically
soil column tests, particularly being subjected to ponding
greater for soils with more coarse particles than for those
hydraulic heads, only focus on a hydrological point, in con-
with finer particles under the same rainfall conditions.
trast to soil infiltration processes that occur on field
ASWC has an important impact on both rainfall–runoff
hillslopes during rainfall events (Boers et al. ).
and soil erosion processes, yet little information is available
Soil texture tends to become coarser from south to north on the Loess Plateau, while precipitation decreases from
regarding the effects of ASWC on the relationship between soil texture and erodibility.
600 to 200 mm a 1 (Tang ). Such differences in soil tex-
The primary objectives of the present study are as fol-
ture and precipitation have a pronounced effect on ASWC
lows: (1) to evaluate the effectiveness of three infiltration
and erosion processes on hillslopes (Lado et al. ).
models and (2) to investigate the differences in infiltration
Numerous well-established soil infiltration models have
and erosion processes between the three loessial soils. The
been reported previously, including the Horton, Philip,
results presented here will form a useful dataset for the
and Green–Ampt models. Both the selection of an appropri-
future assessment of soil erodibility and will help provide
ate model and the quantification of model parameters are
theoretical guidance for erosion control in the loess area.
necessary to describe hillslope rainfall–runoff processes. Previous studies have conducted soil column tests to elucidate the effects of soil properties or ASWC on infiltration
DATA AND METHODS
processes (Marcus et al. ; Makoto et al. ), and it is worth further comparing these findings of such previous
Experimental set-up
studies with infiltration processes on sloped plots under rainfall conditions.
Experiments were conducted in an indoor rainfall simulator
The Universal Soil Loss Equation (USLE) and the Water
with a side-sprinkle rainfall set-up. With this set-up, rainfall
Erosion Prediction Project (WEPP) model are two popular
intensity could be controlled by adjusting spray nozzle size
predictive models for soil erosion. Soil erodibility (K ) is an
and water pressure, and rainfall uniformity exceeded 85%.
important factor in the USLE and is typically assessed
The fall height of raindrops was approximately 16 m,
based on soil texture (Wischmeier & Mannering ).
which ensured kinetic energy similar to that of natural
Extensive field studies have been conducted by Simanton
rainfall.
et al. () and Elliot et al. () to develop methods of pre-
Six experimental plots were constructed, each with
dicting erodibility for cropland and rangeland soils based on
length, width, and depth of 2.00, 0.55, and 0.35 m, respect-
soil properties for use in the WEPP model. Nevertheless, as
ively; the plot size was selected considering the size of a
Laflen () suggested, the future expansion of databases
standard field runoff plot (5 × 20 m) and avoiding marginal
detailing soil properties will be essential for continued veri-
effects. The plots were constructed using steel, with small
fication of predictive models. In particular, establishing a
holes at the bottoms of the plots to allow soil water percola-
relationship between sediment yield and soil texture will
tion. The experimental slope was adjusted to 10 ; this is a
be important for the development and verification of such
common gradient for cultivated farmlands in the Loess Pla-
models.
teau of China. The tested soils were taken from Ansai
For a sloped plot, overland flow occurs when rainfall
County (AS, 36 520 N, 109 190 E), Fuxian County (FX,
intensity exceeds soil infiltration capacity. Soil erosion pro-
35 530 N, 108 360 E), and Yangling District (YL, 34 160 N,
cesses are controlled primarily by soil erodibility and
108 050 E), which are located along the north to south of
rainfall–runoff–erosion
raindrop
the Loess Plateau of China. The particle size distributions
dynamics,
including
splash and overland flow scour (Vermang et al. ).
of these soils are presented in Figure 1. The contribution
When soil detachment is controlled primarily by overland
by the weight of clay particles (<2 μm) to the total soil
flow, detachment rates tend to decrease with increasing
weight was 28%, 20%, and 12% for YL, FX, and AS soils,
(Wischmeier & Mannering ).
respectively. The YL, FX, and AS soils belong to the clay
soil clay content
884
L. Ma et al.
|
Effects of soil moisture on infiltration and erosion processes
Hydrology Research
|
51.5
|
2020
concentration was determined as the ratio of the total dry sediment mass collected to the total runoff volume, and the runoff coefficient was determined as the ratio of surface runoff to corresponding rainfall. A soil’s instantaneous infiltration rate ( fi) can be calculated according to the following formula (Pan et al. ):
fi ¼ I cos θ
Figure 1
|
10Ri St
(1)
where I is rainfall intensity (mm min 1), θ is slope gradient Particle size distributions for tested soils.
( ), t is the time interval required to collect the runoff sample (min), Ri is the runoff collected during the ith time
soil, clay loam soil, and loamy sand categories, respectively, according to the international soil texture classification system (Hillel ).
interval (mL), and S is the area of the plot (cm2). In empirical infiltration models, the Kostiakov equation is typically used to describe infiltration processes in exper-
The soils were gently crushed before being screened
imental soils (Kostiakov ):
with a 10 mm sieve. In each plot, the soil was packed into three layers with a thickness of 10 cm and a total depth of 30 cm and a soil density of 1.20 g cm 3. To avoid discontinu-
f(t) ¼ i0 t a
(2)
ities between neighboring layers in each plot, each soil layer was lightly raked before the next soil layer was packed. The
where f(t) is the instantaneous infiltration rate (mm min 1),
outlet of each plot was constructed like a shutter, with eight
i0 is the initial infiltration rate (mm min 1), t is the duration
steel sheets (0.55 × 0.02 m) attached to each plot at intervals
(min), and a is an empirical parameter associated with soil
of 2 cm. Rainfall I was simulated for a dry run with low
texture.
ASWC (approximately 0.13 m3 m 3). Rainfall II was applied 3 days after Rainfall I in the same soil plots for a wet run with high ASWC (0.3–0.4 m3 m 3).
Here, physically based models including the Horton and Philip models were selected to discuss infiltration processes for the three soils considered. The Horton equation can be written as follows (Horton ):
Data measurement and analysis Both rainfalls with the intensity of about 100 mm h 1 were
f(t) ¼ fc þ (f0 fc )e αt
(3)
administered for about 70 min, in line with typical storm characteristics for the Loess Plateau. Two replicates of
where f(t) is the instantaneous infiltration rate (mm min 1),
each treatment were subjected to the simulated rainfall sim-
fc is the ultimate infiltration capacity or stabilized infiltration
ultaneously. The time until runoff initiation was recorded for
rate (mm min 1), f0 is the initial infiltration rate (mm min 1),
each treatment; after runoff initiation, runoff and sediment
t is the duration (min), and α is the decay parameter. The
were collected for each test at 3-min intervals throughout
parameters fc and α depend primarily on soil properties
the rainfall event. When sediment had been deposited
and antecedent moisture. This equation assumes ponding
enough, the sediment was separated from water, oven
conditions for the soil throughout the infiltration process
dried for 24 h at 105 C, and weighed. Infiltration rates
due to the generation of overland runoff.
were calculated by subtracting the measured runoff rates from
the
rainfall
intensity.
The
average
sediment
Philip () suggested that soil infiltration rates decrease with time according to a power-law equation and
L. Ma et al.
885
|
Effects of soil moisture on infiltration and erosion processes
Hydrology Research
|
51.5
|
2020
order: YL clay (0.80) > FX clay loam (0.56) > AS loamy
can be expressed as follows:
sand (0.40). 1 f(t) ¼ A þ St 0:5 2
Under high ASWC (0.30–0.40 m3 m 3; Rainfall II), the
(4)
average infiltration rates and runoff coefficients for the three soils were 0.31–0.50 mm min 1 and 0.71–0.81, respectively.
1
where A is the constant infiltration rate (cm min ), t is
The AS loamy sand had a higher average and stabilized infiltra-
the duration (min), and S is the soil suction wetting rate
tion rate than the YL clay and FX clay loam, and there was no
(cm min 0.5), which can be obtained by fitting the cumulative
significant difference (p ¼ 0.05) between the YL clay and FX
infiltration amount and t 0.5 at the beginning of infiltration.
clay loam (Table 1). The average and stabilized runoff coeffi-
This equation assumes ponding infiltration into deep soil
cients for the three soil types decreased in the following
with uniform antecedent water content.
order: YL clay (0.81) > FX clay loam (0.78) > AS loamy sand
Analysis of variance (ANOVA) and paired t-tests were
(0.71). These results demonstrate that soil infiltration rates
used to analyze possible differences in soil infiltration and
increase as clay content decreases and indicate that the
sediment yield processes between the three soils and between
runoff coefficient decreases with decreasing clay content
high and low ASWC. Correlation and regression analyses
under similar rainfall conditions. This is in accordance with pre-
were also undertaken to investigate the relationship between
vious findings based on soil column tests (Makoto et al. ).
sediment yield and runoff. The above analyses were per-
Based on the comparison between Rainfall I and II,
formed using SPSS 19.0 (by SPSS Inc., an IBM Company).
ASWC had no significant effect (p ¼ 0.05) on the average infiltration rate and runoff coefficient for the YL clay. This behavior is similar to that reported by Benito et al. (),
RESULTS AND DISCUSSION
who demonstrated little difference in runoff or soil infiltration rates between dry and wet periods for clay soils. This
Runoff and infiltration
can be attributed to the low porosities typical of clay soils. In contrast, the infiltration rates for the AS loamy sand 3
3
Under low ASWC conditions (0.13 m m ; Rainfall I), the
and FX clay loam under low ASWC were both about two
average infiltration rates for the YL clay, FX clay loam,
times greater than those under high ASWC. This result
1
agrees with both Cerdà () and Jones (), who found
and AS loamy sand were 0.34, 0.66, and 0.97 mm min , respectively. Significant differences between the three soils
that field soils with sandy or loamy characteristics exhibit
were found at the p ¼ 0.05 level (Table 1). The runoff coeffi-
significantly greater infiltration rates during dry periods
cients for the different soil types decreased in the following
than during wet periods.
Table 1
|
Runoff, infiltration, and sediment yield characteristics for YL clay, FX clay loam, and AS loamy sand plots under low and high ASWC The stabilized valuea
The average value during the rainfall process Soil
3
ASWC (m3 m
)
IR b
RC
SYR
SC
IR
RC
SYR
SC
0.80 a
5.13 a
3.76 a
0.27 a
0.84 a
2.04 a
1.43 a
YL
0.13
0.34 a
FX
0.13
0.66 b
0.56 b
9.26 b
11.21 b
0.40 b
0.73 b
9.68 b
9.07 b
AS
0.13
0.97 c
0.40 c
2.37 c
3.77 a
0.80 c
0.56 c
1.88 a
2.11 a
YL
0.30
0.31 a
0.81 a
4.36 a
3.30 a
0.31 a
0.81 a
2.44 a
1.86 a
FX
0.35
0.34 a
0.78 a
32.72 b
27.36 b
0.25 a
0.86 a
12.75 b
9.70 b
AS
0.40
0.50 b
0.71 b
35.69 c
29.76 c
0.43 b
0.75 b
–
–
1
2
IR, infiltration rate (mm min ); RC, runoff coefficient; SYR, sediment yield rate (g m a The average of the last four observed values during the later phase of rainfall. b
1
3
min ); SC, sediment concentration (kg m ).
The same letter represents no significant differences at the p ¼ 0.05 level among the three soil using ANOVA.
L. Ma et al.
886
|
Effects of soil moisture on infiltration and erosion processes
Hydrology Research
|
51.5
|
2020
The stable infiltration rates and runoff coefficients
decreasing clay content. The regressed power equations
attained at the later stage of rainfall exhibited variation simi-
were converted into logarithmic lines, and the differences
lar to that observed for the average values presented in the
between the three soil types were further analyzed. Under
preceding paragraph. Significant differences in these par-
low ASWC, significant differences (p ¼ 0.05) in cumulative
ameters between low and high ASWC conditions were
infiltration processes were found between the three soils;
found for the FX clay loam and AS loamy sand, but not
under high ASWC, the AS loamy sand exhibited a signifi-
for the YL clay. These results provide further evidence that
cantly higher infiltration volume than both the FX clay
ASWC likely has little effect on clay soils but has a signifi-
loam and YL clay. Low ASWC generated a greater infiltra-
cant effect on loamy or sandy soils.
tion volume than high ASWC for both FX clay loam and
For the three soils considered, the cumulative infiltra-
AS loamy sand. In contrast, the cumulative infiltration
tion volume (F(t)) increased with rainfall duration; this
volume varied little between low and high ASWC for the
relationship can be described well by the power-law
YL clay (Figure 2 and Table 1), although the low ASWC cor-
equation F(t) ¼ atb (Figure 2 and Table 2). The exponent b
responded to a higher a-value in the regressed equations.
in the regressed equations was found to be in the ranges
This indicates that, for clay soils, ASWC may have a greater
0.52–0.77 and 0.74–0.83 for low and high ASWC, respect-
influence on initial infiltration processes than later period.
ively; moreover, values of b were found to increase with
Changes in the infiltration rate throughout the simulated rainfall event are shown in Figure 3 for all three soils and both low and high ASWC. Under low ASWC conditions (Rainfall I), overland sheet flow occurred 2.5, 5, and 7 min after the initiation of simulated rainfall for the YL clay, FX clay loam, and AS loamy sand, respectively. The infiltration rate for the YL clay decreased quickly to reach 0.3 mm min 1 after the initial 8.5 min and then stabilized; conversely, the infiltration rate decreased continuously throughout the rainfall duration for both the FX clay loam and AS loamy sand. However, under high ASWC conditions, the infiltration rates of all three soils decreased abruptly, with all soil types reaching stable values within 10 min. The Kostiakov, Horton, and Philip models were adopted
Figure 2
Table 2
|
|
Cumulative infiltration processes for YL clay, FX clay loam, and AS loamy sand
to fit the infiltration processes illustrated in Figure 3 using
plots under low (L) and high (H) ASWC.
the least square method (Table 3).
Best-fit equations for relationship between cumulative infiltration volume (F(t)), cumulative sediment yield (S(t)), and rainfall time (t) for soils considered under low and high ASWC F(t) 3
ASWC (m3 m
0.13
)
S(t)
Soil
Number
The best-fit equation
R2
The best-fit equation
R2
YL
23
F(t) ¼ 2.290t 0.517
0.974
S(t) ¼ 59.970t 0.430
0.949
0.668
0.998
S(t) ¼ 0.083t 2.145
0.992
0.998
S(t) ¼ 61.61ln(t) 105.9
0.967
0.760
0.978
S(t) ¼ 43.850t
0.995
0.998
S(t) ¼ 800.50ln(t) 1,202.0
0.971
0.995
S(t) ¼ 88.68e
0.965
0.13
FX
23
F(t) ¼ 2.719t
0.13
AS
22
F(t) ¼ 2.619t 0.767
0.30
YL
23
F(t) ¼ 0.741t
0.35
FX
23
F(t) ¼ 0.968t 0.743
23
F(t) ¼ 0.981t
0.40
AS
0.829
0.446
0.051t
L. Ma et al.
887
|
Effects of soil moisture on infiltration and erosion processes
Hydrology Research
|
51.5
|
2020
sand and YL clay were found to have greater f0-values than the FX clay loam. The greater f0-value obtained for the YL clay may be related to the abrupt decline in the infiltration rate during the initial phase of rainfall (Figure 3). The parameters A and S in the Philip model represent the stabilized and initial infiltration rates, respectively. Both A and S increased with increasing clay content. However, the Philip model generated a significantly lower stabilized and initial infiltration rate than the Horton model (i.e., A < fc and S < f0; Table 3). Under the simulated rainfall conditions, it was difficult to accurately observe the initial soil infiltration rate owing Figure 3
|
Infiltration rates vs. rainfall time for YL clay, FX clay loam, and AS loamy sand plots under low (L) and high (H) ASWC.
to the relatively low rainfall intensity. However, the Horton model produced stabilized infiltration rates ( fc) that were close to the observed values (Tables 1 and 3),
Based on the coefficient of determination (R 2), the
and higher R 2 values were obtained for the Horton model
Kostiakov model can describe infiltration processes better
than the Kostiakov and Philip models. These results suggest
for the FX clay loam and AS loamy sand than for the YL
that the Horton model is more suited than the Kostiakov or
clay. The i0-values in the Kostiakov model were found to
Philip models to describing the infiltration processes of loes-
1
for low
sial soils under the rainfall conditions considered here. Zhao
and high ASWC, respectively. Moreover, the FX clay loam
et al. () found that both the Horton and Philip models
and AS loamy sand exhibited significantly higher i0-values
can describe soil infiltration processes well for the loessial
than the YL clay. This finding supports the assertion that initial
soils of the Loess Plateau based on soil column tests.
infiltration rates decrease with increasing clay content.
Additionally, Genachte et al. () conducted experiments
vary in the ranges 0.77–2.93 and 0.62–1.0 mm min
The Horton model can give both initial ( f0) and stabil-
on the soil infiltration behaviors of Arenosol and Ferralsol
ized infiltration rates ( fc). For the three soils, the fitted
soils in tropical rain forests and suggested that the effective-
fc-values were found to decrease with increasing clay con-
ness of infiltration models depends partly on soil properties.
tent and were close to the observed values (Tables 1 and
In particular, Genachte et al. () found that the Philip
3). The fitted f0-values of the three soils were in the ranges
and Horton models could describe infiltration processes
1
under low (Rainfall I)
for the Arenosol better than the Kostiakov model, while
and high ASWC (Rainfall II), respectively. The AS loamy
all of the models considered (i.e., Philip, Kostiakov, and
1.50–2.51 and 1.44–1.75 mm min
Table 3
|
Fitted infiltration models for YL clay, FX clay loam, and AS loamy sand under low and high ASWC
Kostiakov model 3
Soil
ASWC (m3 m
YL
0.13
)
Horton model
Philip model
Number
f (t) ¼ i0 t a
R2
f (t) ¼ fc þ (f0 fc )e αt
23
f(t) ¼ 0.772t 0.29 0.46
0.44
f(t) ¼ 0.262 þ 2.254e 0.375t
0.91
f(t) ¼ 0.40 þ 1.096e
0.94
f(t) ¼ 0.675 þ 1.115e 0.051t
0.06t
FX
0.13
23
f(t) ¼ 2.836t
AS
0.13
22
f(t) ¼ 2.934t 0.34 0.21
0.37
f(t) ¼ 0.28 þ 1.471e
0.76
f(t) ¼ 0.293 þ 1.143e 0.289t
0.70
0.381t
YL
0.30
23
f(t) ¼ 0.622t
FX
0.35
23
f(t) ¼ 1.017t 0.35
23
0.21
AS
0.40
f(t) ¼ 0.998t
0.463t
f(t) ¼ 0.467 þ 1.261e
R2
1 f (t) ¼ A þ St 0:5 2
R2
0.94
f (t) ¼ 0.025 þ 1.387t 0.5
0.72
0.94
f (t) ¼ 0.069 þ 2.838t 0.5
0.94
0.96
f (t) ¼ 0.322 þ 3.210t 0.5
0.95
0.5
0.94
f (t) ¼ 0.11 þ 0.943t
0.91
f (t) ¼ 0.115 þ 1.026t 0.5
0.92
0.93
0.5
0.88
f (t) ¼ 0.295 þ 0.958t
0.77
888
L. Ma et al.
|
Effects of soil moisture on infiltration and erosion processes
Hydrology Research
|
51.5
|
2020
Horton models) could predict infiltration behaviors well for
behavior with an exponential distribution, with the cumulat-
the Ferralsol.
ive sediment yield accelerating during the later phase of the
The clay contents of the YL, FX, and AS soils were
rainfall. This behavior differs markedly from that of the
approximately 28%, 20%, and 13%, respectively (Figure 1),
other soil types (Figure 4). The YL clay exhibited similar
and their corresponding average and steady infiltration
sediment yield distributions for both high and low ASWC.
rates decreased in the following order: AS > FX > YL. This
In contrast, the obtained cumulative curves are markedly
provides further evidence that the soil infiltration rate
different under high and low ASWC for the FX and AS
decreases with increasing clay content. For both the FX
soils with high ASWC generating significantly greater sedi-
and AS soils, high ASWC generated a lower steady infiltra-
ment yields than low ASWC. These results indicate that
tion rate; in contrast, ASWC had little effect on the
the sediment yield may not be predicted effectively based
infiltration rate for the YL clay (Figure 3 and Table 2).
on rainfall duration under different ASWC conditions. Under low ASWC (Rainfall I), the FX clay loam
This indicates that, for soils with high clay content (i.e., >30%), ASWC has little effect on soil infiltration processes
plot
exhibited 2
an
average
sediment
yield
rate
of
1
during rainfall events. However, for loamy soils (i.e., clay
9.26 g m
content <20%), infiltration models should consider the
than the values obtained for the YL clay and AS loamy
effects of ASWC. These results may reflect the higher poros-
sand, respectively. Thus, the average sediment yield rate
ity of loamy soils relative to clay soils, allowing loamy soils
was lowest for the AS loamy sand, likely because this soil
to both hold more water and induce greater expansion of
exhibited the highest average infiltration rate (Table 1) and
soil volume.
the lowest average runoff rate. No significant difference
min ; this value is 1.8 and 3.9 times greater
(p ¼ 0.05) was found between the average sediment concenSediment yield
trations for the AS loamy sand and YL clay (Table 1). The higher sediment yield for the FX clay loam plot can be attrib-
Cumulative sediment yields increased with rainfall duration, exhibiting power-law, exponential, or logarithmic linear distributions (Figure 4 and Table 2). The differences found in the best-fit equations indicate that sediment yield processes (or rather, change processes in sediment yield with the rainfall duration) varied with soil texture and ASWC. Under high ASWC, the AS loamy sand exhibited sediment yield
uted to the occurrence and development of rills. In contrast, the difference in sediment yields between the AS loamy sand and YL clay can be attributed primarily to differences in soil infiltration processes. The AS loamy sand exhibited an average infiltration rate that was approximately three times greater than that of the YL clay. Therefore, the AS loamy sand plot can be considered to have had lower runoff erosion energy, even though its soil erodibility would have been higher than that of the YL clay (Wischmeier & Mannering ; Elliot et al. ). During the later phase of rainfall, the FX clay loam exhibited a stabilized sediment yield rate of 9.26 g m 2 min 1, which is almost five times that obtained for the YL clay and AS loamy sand (Figure 5). These results demonstrate that differences in the sediment yield rate between the three soil types are considerably greater during the later stage of rainfall than during the earlier stage. Under high ASWC (Rainfall II), FX clay loam and AS loamy sand exhibited similar sediment yield rates (32.7 and 35.7 g m 2 min 1, respectively) and sediment concen-
Figure 4
|
Cumulative sediment yield vs. rainfall duration for YL clay, FX clay loam, and AS loamy sand plots under low (L) and high (H) ASWC.
trations (27.36 and 29.76 kg m 3, respectively). These values are eight and nine times those of the YL clay,
889
L. Ma et al.
|
Effects of soil moisture on infiltration and erosion processes
Hydrology Research
|
51.5
|
2020
sediment yield rates under low ASWC. In contrast, initial sediment yield rates were lower than steady sediment yield rates for the FX soil; this could be attributed to the development of depressions under low ASWC for this soil type (Figure 5). Similar results have been reported previously for the YL and AS soils (Pan et al. ) and may reflect expansion and shrinkage of the surface soil layer and the greater raindrop splash during the initial rainfall stage. The FX clay loam and AS loamy sand plots generated significantly higher sediment yields under high ASWC (Rainfall II) than under low ASWC (Rainfall I). In contrast, no clear difference in sediment yield between low and high ASWC was observed for the YL clay plot (Table 1). This suggests that ASWC has a more significant effect on loamy or sandy soils than clay soils. Differences in sediment yield processes between low and high ASWC were considered using paired t-tests. No significant difference was found for the YL clay. However, high ASWC led to sediment yields that were 3.5 and 15 times higher than those under low ASWC for the AS loamy sand and FX clay loam, respectively. Benito et al. () found sediment concentration to be greater during wet periods than dry periods for all water-repellent soil sites considered. Here, the sediment yields obtained for the AS loamy Figure 5
|
(a) Sediment yield rate and (b) sediment concentration vs. rainfall duration for YL clay, FX clay loam, and AS loamy sand plots under low (L) and high (H) ASWC.
sand and FX clay loam under different ASWC support the findings of Benito et al. (), although the YL clay results do not. This highlights the importance of the effects of soil texture and ASWC on soil erosion.
respectively (4.36 g m 2 min 1 and 3.3 kg m 3) (Table 1).
Under low ASWC (Rainfall I), the sediment concen-
The greater sediment yields of the FX clay loam and AS
tration for the FX clay loam increased gradually, reaching
loamy sand are in accordance with their relatively low soil
approximately 17.5 kg m 3 after 35 min, and then stabilized
erodibility compared to the YL clay. Moreover, during the
at 10–15 kg m 3 (Figure 5). This behavior may be associated
later phase of rainfall, the sediment yield rates of the YL
with the development of rills. In contrast to the FX clay
clay and FX clay loam remained relatively stable, while
loam, both sediment yield rate and sediment concentration
the AS loamy sand plot exhibited significant increases in
for the YL clay and AS loamy sand plots increased rapidly,
sediment concentration and the sediment yield rate
peaking after 3 min, and then decreased slowly to reach
(Figure 5).
stabilized values during the later phase of rainfall (Figure 5).
In the present study, no clear relationship was observed
No rill development was observed in the YL clay and AS
between clay content and sediment yield under low ASWC,
loamy sand plots. This, combined with the occurrence of
although a negative relationship was observed under high
peak sediment yields during the initial phase of rainfall for
ASWC (Figures 4 and 5). This could be a result of different
these soils, supports the assertion that soil erosion in the
erosional forms, and depressions or rills tended to occur on
YL clay and AS loamy sand plots results primarily from rain-
the loamy soil plots with high ASWC. For the YL and AS
drop splash. Moreover, easily-eroded soils are typically
soils, initial sediment yield rates were greater than steady
transported during the initial phases of rainfall events,
890
L. Ma et al.
|
Effects of soil moisture on infiltration and erosion processes
Hydrology Research
|
51.5
|
2020
such that their supply is exhausted by the time later runoff phases occur. This phenomenon has been observed frequently on vegetated slopes or on relatively small plots of bare soil (Pan & Shangguan ). Under high ASWC (Rainfall II), notable differences in sediment yield processes were observed between the different soil types. The sediment yield rate and sediment concentration for the YL clay plot decreased gradually, while those for the AS loamy sand plot increased continuously. The decreasing trend observed for the YL clay plot can be attributed primarily to the dominance of raindrop splash erosion processes. Conversely, the continuous increase in erosion for the AS loamy sand plot can be attributed to soil collapse in the downslope area under high ASWC. Both the sediment
Figure 6
|
Relationship between the runoff rate and the sediment yield rate for YL clay, FX clay loam, and AS loamy sand plots under low (L) and high (H) ASWC.
yield rate and sediment concentration were found to increase linearly for the FX clay loam plot, peaking after 15 min,
infiltration rate initially decreased abruptly before remaining
before decreasing gradually to reach a constant value. This
relatively constant during the later phase of runoff (Figure 3).
behavior can be attributed primarily to the accelerated devel-
These different relationships under low and high ASWC indi-
opment of rills during the initial phase of rainfall and their
cate that ASWC has an important effect on the correlation
relative stability during later phases.
between the runoff rate and the sediment yield rate.
For the YL clay, no significant difference (p ¼ 0.05) in
When both low and high ASWC are considered
sediment yield processes was observed between low and
together, the results presented here indicate that the sedi-
high ASWC (i.e., between Rainfall I and II). In particular,
ment yield rate decreased linearly with the runoff rate for
the sediment concentration decreased rapidly from an initial
the YL clay; in contrast, higher runoff rates led to increased
3
after 12 min; then, it
sediment yield rate (as described by a power-law function)
decreased gradually to reach a stable value of 2.0 kg m 3
for both the AS loamy sand and FX clay loam (Figure 6).
during the later stage of rainfall. However, for the FX clay
In the present study, the plot surfaces were observed after
loam and AS loamy sand, high ASWC (Rainfall II) yielded
each rainfall. Scattered depressions were observed for the
significantly more sediment than low ASWC (Rainfall I)
AS and FX soils, while neither rills nor depressions were
(Figure 5). Although the AS loamy sand and FX clay loam
observed for the YL clay under either low or high ASWC.
yielded similar volumes of sediment overall, differences in
This suggests that raindrop splash dominates erosion
sediment processes between these two soil types were very
processes for the YL clay, while runoff scouring likely
significant (Figure 5).
dominates the AS and FX soils. For the YL clay, more soil
peak value of 23.6–3.5 kg m
particles were eroded by raindrop splash and delivered to Relationship between the sediment and runoff
the outlet at the beginning of the rainfall than later. Correspondingly, a negative relationship was observed between
Under low ASWC (Rainfall I), a significant positive relation-
the sediment yield rate and the runoff rate for the YL clay.
ship was observed between the sediment yield rate and the
Previous studies have observed a similar negative relation-
runoff rate for the FX clay loam, with a correlation coefficient
ship between erosion and the runoff rate for soils with
(R) of 0.76; conversely, a negative relationship (R ¼ 0.74) was
high clay content (Pan et al. ).
found for both the YL clay and AS loamy sand. However,
Under the same runoff rates, high ASWC led to greater
under high ASWC (Rainfall II), the relatively stable runoff
sediment yield rates for the FX clay loam and AS loamy
rate corresponded to a greater variation in the sediment
sand. This can be attributed primarily to the development
yield rate (Figure 6); this likely occurred because the soil
of rills and the collapse of the soil mass. However, for the
891
L. Ma et al.
|
Effects of soil moisture on infiltration and erosion processes
YL clay, low ASWC was found to generate greater sediment
Hydrology Research
Table 4
|
soil particles being more detachable under low ASWC.
51.5
|
2020
Best-fit equations between cumulative sediment yield (S) and cumulative runoff volume (R) for soils considered under low and high ASWC
yield rates than high ASWC during the initial phase of rainfall for the same runoff rates. This result can be attributed to
|
ASWCs (m3 m
3
)
The tested soils
The best-fit equations
Number
R2
The positive relationship between the runoff rate and
0.13
YL clay
S ¼ 80.85R 0.350
23
0.982
the sediment yield rate for the FX clay loam and AS
0.13
FX clay loam
S ¼ 13.84R 66.68
23
0.987
0.364
22
0.941
23
0.998
loamy sand plots supports a popular viewpoint that increas-
0.13
AS loamy sand
S ¼ 45.32R
ing the runoff rate tends to enhance the soil detachment rate
0.30
YL clay
S ¼ 45.05R 0.408
owing to increases in erosional force or runoff energy
0.35
FX clay loam
S ¼ 34.06ln(R) þ 36.2
23
0.952
(Laflen ). However, for the YL clay plot, increasing
0.40
AS loamy sand
S ¼ 95.27e (0.042R)
23
0.963
the runoff rate led to decreasing the sediment yield rate (Figure 6). These results suggest that inter-rill erosion dominated by raindrop splash (as occurred for the YL clay) may have little relation to overland runoff. For a rainfall depth of approximately 100 mm, the AS loamy sand and FX clay loam plots produced approximately 82 mm of runoff under high ASWC (Rainfall II). This is significantly greater than the runoff depths generated under low ASWC (Rainfall I), which were 44 and 60 mm for the AS loamy sand and FX clay loam, respectively. Under high ASWC, the FX clay loam and AS loamy sand plot yielded 2,159 and 2,355 g m 2 sediments, respectively; these values are 3 and 13 times, respectively, the corresponding values observed under low ASWC (Rainfall I; Figure 7). In contrast, runoff volume and sediment yield for the YL clay varied little between low and high ASWC. Regression analysis was undertaken to predict the relationship between cumulative sediment yield and cumulative runoff (Table 4). Generally, a linear or power-law equation
was found to effectively describe this relationship, except for the FX clay loam and AS loamy sand under high ASWC (Rainfall II). For the YL clay plot, and for the AS loamy sand plot under low ASWC only, the cumulative sediment yield increased with total runoff volume and the exponent of the regressed power equation was in the range of 0.35– 0.41. In this context, an exponent smaller than 1.0 indicates that the erosion rate would decrease gradually with rainfall duration. Such behavior is in line with that expected to occur when inter-rill erosion processes are dominated by raindrop splash. A linear relationship in this context indicates that the erosion rate remains relatively constant throughout the rainfall period; this is considered to be in line with the development of rill erosion within the FX clay loam plot under low ASWC (Table 4). Linear or power-law relationships between runoff and sediment load have been reported previously at the watershed scale (García-Ruiz et al. ; López-Tarazón et al. ; Tuset et al. ). However, the exponential equation obtained for the AS loamy sand plot under high ASWC indicates that the soil detachment rate accelerated gradually during the rainfall simulation; this behavior can be explained by soil collapse in the downslope area. In general, the relationship between runoff and sediment depends primarily on soil erosion processes, which in turn are controlled partly by soil texture and ASWC.
CONCLUSIONS The simulated rainfall experiments in sloped soil plots were Figure 7
|
Cumulative sediment yield vs. cumulative runoff for YL clay, FX clay loam, and AS loamy sand plots under low (L) and high (H) ASWC.
used to investigate the effects of soil texture and ASWC (0.13
and
0.35 m3 m 3)
on
infiltration
and
erosion
892
L. Ma et al.
|
Effects of soil moisture on infiltration and erosion processes
processes. The clay (<2 μm) contents of the YL, FX, and AS soils in this test corresponded to approximately 28%, 20%, and 12% by weight, respectively. Under the same rainfall conditions, average soil infiltration rates decreased with increasing clay content. ASWC had little effect on infiltration processes for the YL clay but had significant effects for the FX and AS soils. The Horton model was found to more effectively describe infiltration processes for these three loessial soils than the Kostiakov or Philip models. Under both low and high ASWC, the YL clay had a decreasing sediment yield rate with rainfall duration and generated much lower sediment yield than the FX and AS soils. The low sediment yield for the YL clay was mainly attributed to the dominance of raindrop splash erosion. The FX and AS soils had greater sediment yields under high ASWC than under low ASWC, which derived from the development of rills. There was no constant relationship between the sediment yield rate and the runoff rate, implying the importance of erosion form in predicting the soil erosion process on a hillslope.
ACKNOWLEDGEMENTS This study was supported by the National Natural Science Foundation of China (Grant No. 51779004). We would like to express great gratitude to the reviewers for their valuable suggestions made to improve the manuscript.
AUTHOR CONTRIBUTIONS L.M. conceived and designed the experiments; L.M., J.L., and J.L. performed the experiments and analyzed the data; L.M. wrote the paper.
REFERENCES Benito, E., Santiago, J. L., Blas, D. E. & Varela, M. E. Deforestation of water-repellent soils in Galicia (NW Spain): effects on surface runoff and erosion under simulated rainfall. Earth Surface Processes and Landforms 28, 145–155. Boers, T. M., Deurzen, V. F. J. M. P., Eppink, L. A. A. J. & Ruytenberg, R. E. Comparison of infiltration rates with
Hydrology Research
|
51.5
|
2020
an infiltrometer, a rainulator and a permeameter for erosion research in S E Nigeria. Soil Technology 5, 13–26. Cerdà, A. Seasonal variability of infiltration rates under contrasting slope conditions in southeast Spain. Geoderma 69, 217–232. Dong, H., Huang, R. Q. & Gao, Q. F. Rainfall infiltration performance and its relation to mesoscopic structural properties of a gravelly soil slope. Engineering Geology 230, 1–10. Elliot, W. J., Laflen, J. M. & Kohl, K. D. Effect of Soil Properties on Soil Erodibility. ASAE/CSAE, St. Joseph, MI, Paper No. 89-2150. Fu, B. J., Hu, C. X., Chen, L. D., Honnay, O. & Gulinck, H. Evaluating change in agricultural landscape pattern between 1980 and 2000 in the Loess hilly region of Ansai County, China. Agriculture, Ecosystems & Environment 114 (2–4), 387–396. García-Ruiz, J. M., Regüés-Muñoz, D., Alvera, B., Lana-Renault, N., Serrano-Muela, M. P., Nadal-Romero, E., NavasIzquierdo, A., Latrón, J., Bono, C. E. M. & Arnáez-Vadillo, J. Flood generation and sediment transport in experimental catchments affected by land use changes in the central Pyrenees. Journal of Hydrology 356, 245–260. Geiger, S. L. & Durnford, D. S. Infiltration in homogeneous sands and a mechanistic model of unstable flow. Soil Science Society of America Journal 64, 460–469. Genachte, G. V., Mallants, D., Ramos, J., Deckers, J. A. & Feyen, J. Estimating infiltration parameters from basic soil properties. Hydrological Processes 10, 687–701. Hillel, D. Introduction to Soil Physics. Academic Press, New York. Horton, R. E. An approach toward a physical interpretation of infiltration-capacity 1. Soil Science Society of America Journal 5 (C), 399–417. Jones, J. A. A. Global Hydrology: Processes, Resources and Environmental Management. Longman, Harlow. Kostiakov, A. N. On the dynamics of the coefficient of water percolation in soils and on the necessity for studying it from a dynamic point of view for purposes of amelioration. In: Transactions of 6th Committee International Society of Soil Science, Russia, Part A, pp. 17–21. Lado, M., Ben-Hur, M. & Shainberg, I. Soil wetting and texture effects on aggregate stability, seal formation, and erosion. Soil Science Society of America Journal 68, 1992–1999. Laflen, J. M. WEPP-erosion prediction technology. In: Soil Erosion and Dryland Farming (J. M. Laflen, J. L. Tian & C. H. Huang, eds). CRC Press, New York, pp. 557–566. López-Tarazón, J. A., Batalla, R. J., Vericat, D. & Balasch, J. C. Rainfall, runoff and sediment transport relations in a mesoscale mountainous catchment: the River Isabena (Ebro basin). Catena 82, 23–34. Makoto, H., Heinz, G. S. & Daiki, A. Removal of saline water due to road salt applications from columns of two types of sand by rainwater infiltration: laboratory experiments and model simulations. Water, Air, & Soil Pollution 230, 305.
893
L. Ma et al.
|
Effects of soil moisture on infiltration and erosion processes
Marcus, A. H., William, E. C., Richard, B. D., Greg, H., Shaun, L. & Kathrin, M. Effect of antecedent soil moisture on preferential flow in a texture-contrast soil. Journal of Hydrology 398, 191–201. Mei, X. M., Zhu, Q. K., Ma, L., Zhang, D., Wang, Y. & Hao, W. J. Effect of stand origin and slope position on infiltration pattern and preferential flow on a Loess hillslope. Land Degradation & Development 29, 1353–1365. Pan, C. Z. & Shangguan, Z. P. Runoff hydraulic characteristics and sediment generation in sloped grassplots under simulated rainfall conditions. Journal of Hydrology 331, 178–185. Pan, C. Z., Shangguan, Z. P. & Lei, T. W. Influences of grass and moss on runoff and sediment yield on sloped loess surfaces under simulated rainfall. Hydrological Processes 20, 3815–3824. Pan, C. Z., Ma, L. & Wainwright, J. Particle selectivity of sediment deposited over grass barriers and the effect of rainfall. Water Resources Research 52, 7963–7979. Philip, J. R. Theory of infiltration. In: Advances in HydroScience, Vol. 5 (T. H. Chow, ed.). Academic Press, New York, pp. 215–296. Simanton, J. R., Weltz, M. A., West, L. T. & Wingate, G. D. Rangeland Experiments for Water Erosion Prediction Project. ASAE/CSAE, St. Joseph, MI, Paper No. 87-2545.
Hydrology Research
|
51.5
|
2020
Tang, K. L. Soil and Water Conservation in China. Science Press, Beijing. Tuset, J., Vericat, D. & Batalla, R. J. Rainfall, runoff and sediment transport in a Mediterranean mountainous catchment. Science of the Total Environment 546, 114–132. Vermang, J., Demeyer, V., Cornelis, W. M. & Gabriels, D. Aggregate stability and erosion response to antecedent water content of a loess soil. Soil Science Society of America Journal 73 (3), 718–726. Wei, W., Jia, F. Y., Yang, L., Chen, L. D., Zhang, H. D. & Yu, Y. Effects of surficial condition and rainfall intensity on runoff in a loess Hilly area, China. Journal of Hydrology 513 (1), 115–126. Wischmeier, W. H. & Mannering, J. V. Relation of soil properties to its erodibility. Soil Science Society of America Journal 33 (1), 131–137. Youngs, E. G. Infiltration measurements – a review. Hydrological Processes 5, 309–320. Zhao, J. B., Zhang, Y., Chen, B. Q. & Dong, Z. B. Law of water infiltration of lower part of middle Pleistocene loess in Luochuan of Shaanxi. Acta Pedologica Sinica 46, 965–972.
First received 21 January 2020; accepted in revised form 10 March 2020. Available online 15 April 2020
894
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Assessment of hydrological drought based on nonstationary runoff data Xueli Sun, Zhanling Li and Qingyun Tian
ABSTRACT A nonstationary standardized runoff index (NSRI) is proposed by using the GAMLSS framework to assess the hydrological drought under nonstationary conditions. The definition of the NSRI is similar to that of SRI, but using a nonstationary Gamma distribution by incorporating meteorological variables and antecedent runoff as covariates to describe the characteristics of runoff series. The new drought index is then applied to the upper reach of the Heihe River basin. Four models are developed, in which one is stationary, and the other three are nonstationary with one, two and three covariates, respectively. Results show that, for the nonstationary runoff series, the nonstationary models are more robust and reliable than the stationary one. Among these models,
Xueli Sun Zhanling Li (corresponding author) Qingyun Tian MOE Key Laboratory of Groundwater Circulation and Environmental Evolution, China University of Geosciences (Beijing), Beijing 100083, China and School of Water Resources and Environment, China University of Geosciences (Beijing), Beijing 100083, China E-mail: zhanling.li@cugb.edu.cn
the model with two covariates performs the best. For the model with one covariate, the precipitation shows better in the fitting as a covariate in rainy seasons, and the antecedent runoff shows better in dry seasons. The NSRI identifies more drought events than SRI does, and the drought conditions in our case are mainly affected by precipitation. It is proved that the proposed new drought index is a more effective method for drought assessments under nonstationary conditions. Key words
| GAMLSS, hydrological drought, nonstationary, SRI
HIGHLIGHTS
• • • •
A nonstationary standardized runoff index is developed. Six alternative covariates and three kinds of nonstationary models are compared. The nonstationary model with two covariates performs the best. Hydrological drought conditions in this case is mainly affected by precipitation.
INTRODUCTION Drought is a kind of universal natural disaster which
the ecosystems more vulnerable (Zhang et al. ; Cheng
can bring serious ecological, environmental and social
; Fang et al. a; Han et al. a). Thus, drought
consequences. It may inhibit the growth of vegetation, accel-
has attracted wide attention from the researchers. The
erate the degradation of grassland and the disappearance of
accurate and timely drought assessment is crucial for
oasis, aggravate the reduction of rivers and lakes, and lead to
devising plans and carrying out the necessary mitigation measures especially under the background of climate
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.029
warming. Being a robust tool for drought assessment, many drought indices have been developed during the last decades
895
X. Sun et al.
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
(e.g., Nalbantis & Tsakiris ; Vicente-Serrano et al. ;
covariates herein. In addition, evapotranspiration is an
Bloomfield & Marchant ; Tabari et al. ). One of the
important process of water transfer in the hydrosphere
most widely used indices worldwide in the hydrological
and atmosphere, playing a vital role in the hydrological
drought assessment is the standardized runoff index (SRI;
cycle and the runoff processes (Han et al. b; Aa et al.
Luciano et al. ; Jiang et al. ; Zhou et al. ),
). The antecedent runoff (runoff in the previous
which is proposed by Shukla & Wood () based on the
month) contributes to the runoff changes at that month
theory of the standard precipitation index (SPI; McKee
especially in non-flood seasons (Ding et al. ). Conse-
et al. ). It remains the most broadly accepted index due
quently, the evapotranspiration and the antecedent runoff
to being simple to calculate, the limited data required, and
are also taken into account as alternative covariates. As
the different time scales of hydrological drought assessed.
the actual evapotranspiration data are not available
However, the stationarity assumption of the SRI is often
and not easy to calculate, potential evapotranspiration is
violated in the context of global warming and strong
used here. In summary, the six variables, precipitation,
human disturbance (Xiong & Guo ; Villarini et al.
temperature, relative humidity, wind speed, potential evapo-
; Liu et al. ). For example, Xiong & Guo ()
transpiration, and antecedent runoff, are incorporated in
found that the annual mean runoff series in the Yangtze
developing the NSRI.
River basin in China exhibited a clear decreasing trend
This study, taking the upper reach of the Heihe River
from 1882 to 2001. Villarini et al. () found that the
basin (the second largest inland river basin in northwest of
runoff series in the continental United States had significant
China) as the study area, first tests the stationarity of
change points during the 20th century. These mean that
runoff series over the study area and then develops an
many runoff series are not stationary any more, and the SRI
NSRI to assess hydrological droughts under nonstationary
with the assumption of the runoff data to be analyzed
conditions by using GAMLSS. GAMLSS is one of the tech-
keeping stationary is no longer suitable for fitting such non-
niques for simulating nonstationary time series and capable
stationarity data (Wang et al. b). Therefore, a modified
of providing a high degree of flexibility for solving nonsta-
SRI, namely, a nonstationary standardized runoff index
tionary modelling (e.g., Villarini et al. ; Zheng et al.
(NSRI), will be proposed in this study with incorporating
). It can describe the linear or nonlinear relationship
the nonstationary characteristic of runoff series.
between any statistical parameter of the random variable
NSRI is defined similarly with the SRI, but using a non-
sequence and the explanatory variable, and the description of
stationary Gamma distribution by incorporating covariates.
the random variable distribution function has a wide range of
Seeking appropriate covariates is the first, also one of the
types (Stasinopoulos & Rigby ). Many hydrologists
vital steps in constructing the nonstationary drought index
have applied this technique to analyze the nonstationary
(Vu & Mishra ). A single covariate is usually considered
hydrological sequences (Villarini et al. ; Zheng et al.
in previous literatures. For example, time or climatic vari-
). Six variables are considered as covariates in total and
ables are usually used to describe the precipitation
three nonstationary models are developed in terms of different
changes in developing the nonstationary meteorological
combinations of covariates. To verify the reliability of the
drought index (Russo et al. ; Li et al. ; Wang et al.
NSRI, it is finally compared with the traditional SRI according
a; Javad & Somayeh ). However, as known, the
to the historical droughts.
runoff changes are always affected by the recharge sources.
The outline of this paper is as follows. The study area,
The recharge sources for the upper reach of the Heihe River
data description, and the methodology of SRI and NSRI
basin include precipitation, glacial and snow melting, frozen
by using GAMLSS are covered in the section ‘Materials
soil thawing and groundwater (Ding et al. ; Wang et al.
and methods’. The modelling results and the choice of differ-
). These sources are affected by meteorological elements
ent combinations of covariates are discussed in the section
like temperature, relative humidity and wind speed. There-
‘Results and discussion’, followed by the conclusions in
fore, these meteorological elements are considered as
the section ‘Conclusion’.
896
X. Sun et al.
|
Assessment of hydrological drought based on nonstationary data
MATERIALS AND METHODS
Hydrology Research
|
51.5
|
2020
Monteith method (Allen et al. ), which is suggested by the Food and Agriculture Organization of the United
Materials
Nations. The locations of the stations are shown in Figure 1 and the detailed information are displayed in Table 1. These
Study area
datasets were obtained from the China Meteorological Data Network (http://data.cma.cn/) and the Hydrology
The Heihe River originates from Qilian County in the
Yearbook.
northeastern part of Qinghai Province. The river basin, between 97 370 E–102 060 E and 37 440 N–42 400 N, is located in the middle of Hexi Corridor in the northwest of China, with an area of 13,40,000 km2. The circulation
Methods
of the westerly belt in the middle and high latitudes and the polar air mass make the climate across the basin
Standardized runoff index
dry, strong wind and sunshine, large temperature differences, less precipitation and great evaporation, resulting in the scarcity of water resources in this basin (Gao & Zhao ). The basin is divided into three reaches, the upper reach, the middle reach and the lower reach. The annual precipitation for the upper reach is about 350 mm, with 70% occurring from June to October. The water resources over the basin are mainly generated from the upper reach and consumed in the middle and lower reaches. The water
Shukla & Wood () proposed the SRI, analogous to the SPI (McKee et al. ). Many studies have shown the efficacy of SRI in depicting the hydrological droughts (e.g., Luciano et al. ; Jiang et al. ). To obtain this index, fit the runoff series for a certain period by using Gamma distribution firstly and then transform it into a standard normal distribution through an equal probability transformation. The specific calculation process is as follows:
resources over the basin is 2.62 billion m3, supporting for
1. Let x(n) represents the runoff data at the nth month. Fifty-
1.5 million people and 384,000 ha of farmland irrigation
four years (1961–2014) of runoff data are available here,
(Ling ). The amounts and variations of water resources
thus n ranges from 1 to 648. For a given time scale of k
in the upper reach have great effects on agriculture, econ-
months (a time scale of 12-month is considered here),
omic development and ecological environment for the
the cumulative runoff xk (n) is calculated as:
middle and lower reaches. Thus, it is particularly necessary to carry on the researches about hydrological droughts in the upper reach.
xk (n) ¼
n X
(1)
x(i)
i¼n kþ1
Data 2. Fitting a two-parameter Gamma distribution, denoted as The monthly runoff data at three hydrological stations
xk (n) ∼ Gamma (μ, σ):
(Qilian, Zhamushike and Yingluoxia) and the meteorological data at two national meteorological stations (Qilian and Yeniugou) in the upper reach are used. The datasets span from 1961 to 2014. Runoff data are used to calculate the hydrological drought index, whereas meteorological
(xk (n))σ 2 1 exp[ (xk (n))=(σ 2 μ)] f(xk (n)jμ, σ) ¼ 2 1=σ Γ(1=σ 2 ) (σ 2 μ) 1
xk (n) > 0
1
(2)
data, including precipitation, temperature, relative humidity and wind speed, are used as covariates to establish the
where μ and σ are the location and scale parameters in
NSRI. The meteorological data are also used for calculating
Gamma, μ > 0 and σ > 0. Γ( ) is the mathematical
the potential evapotranspiration by employing the Penman–
Gamma function.
X. Sun et al.
897
Figure 1
Table 1
|
|
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
Locations of hydrological and meteorological stations in the upper reach of the Heihe River basin.
Hydrological and meteorological stations in the upper reach of the Heihe River basin
4. Converting cumulative probability to a standard normal distribution function:
Station name
Longitude and
DEM
Station type
(abbreviation)
latitude
(m)
Hydrological station
Qilian (QL)
100.25 E, 38.18 N 99.98 E, 38.23 N 100.18 E, 38.8 N
2,708
Zhamushike (ZMSK) Yingluoxia (YLX) Meteorological station
Qilian (QL) Yeniugou (YNG)
100.25 E, 38.18 N 99.61 E, 38.83 N
2,982 1,682 2,708 3,227
SRI ¼
8 > > > > > > > > > > > > > > > > < > > > > > > > > > > > > > > > > :
c0 þ c1 ω þ c2 ω2 1 þ d1 ω þ d2 ω2 þ d3 ω3 vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #ffi u " u 1 0 F(xk (n)) 0:5 ω ¼ tln F(xk (n))2
ω þ
c0 þ c1 ω þ c2 ω2 1 þ d 1 ω þ d 2 ω2 þ d 3 ω3 vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #ffi u " u 1 0:5 < F(xk (n)) 1 ω ¼ tln (1 F(xk (n))2 ω
(4)
3. The cumulative probability for a given time scale can be calculated as:
where c0 ¼ 2.515517, c1 ¼ 0.802853, c2 ¼ 0.010328, d1 ¼
ðx F(xk (n)) ¼ f(xk (n))dx 0
(3)
1.432788, d2 ¼ 0.189269, and d3 ¼ 0.001308 (Shukla & Wood ).
898
X. Sun et al.
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
Akaike information criterion (AIC) (Akaike ), together
Nonstationary standardized runoff index
with the worm plot, is used to test its goodness of fit. When modeling time series over 30 years, we should pay
Since the NSRI proposed here is normalized in the same
attention to the nonstationarity of the time series which is
way as the traditional SRI, the same drought-level standards
potentially caused by climate change (Russo et al. ).
for both indices are suggested (Guttman ). As shown in
The NSRI is then defined based on a nonstationary
Table 2, positive NSRI values indicate wet conditions, and
Gamma distribution with its location and scale parameters
negative values indicate dry conditions. The higher the
varying with covariates by using the GAMLSS framework
NSRI value, the wetter it is, and the lower the value, the
(Rigby & Stasinopoulos ; Stasinopoulos & Rigby ).
drier it is.
The GAMLSS framework was introduced by Rigby and Stasinopoulos (, ) and Akantziliotou et al. (), as a way to overcome some of the limitations associated with
RESULTS AND DISCUSSION
generalized linear models (GLMs) and generalized additive models (GAMs) (Nelder & Wedderburn ; Hastie & Tibshirani ). This framework is expanded to allow modeling not only the mean (or location) but other parameters of the distribution for a response variable as linear and/or nonlinear functions of explanatory variables and/ or random effects. In GAMLSS, the cumulative runoff xk (n) is modeled as xk (n) ∼ Gamma (μn , σ n ). It is assumed that the changes of runoff with covariates obey the following distribution parameters:
Stationary test Since it is the cumulative runoff series fitted by a distribution, the stationarity of cumulative runoff series is to be tested. The augmented Dickey–Fuller (ADF) test, an augmented version of the original Dickey–Fuller test (Dickey & Fuller ), is then performed to the 12-month cumulative runoff series from 1961 to 2014 to verify their stationarities. The computed Dickey–Fuller statistic is compared with the corresponding critical value at a certain
μn ¼ a0 þ
I X
ai zi (n)
(5)
i¼1
P-value. For example, when P ¼ 0.05, the critical value of the ADF test result is 3.5 (with the sample size of 50). If the computed Dickey–Fuller statistic is greater than this
σ n ¼ b0 þ
I X
critical value, it indicates that the null hypothesis of the bi zi (n)
(6)
i¼1
where μn and σ n are the location and scale parameters in
ADF test is accepted at the 0.05 significance level, that is, the sequence to be analyzed is nonstationary. In this study, we used the package of ‘tseries’ in r to do the ADF test which can output the P-values and the Dickey–Fuller
nonstationary Gamma distribution, a0 and b0 are constant terms, zi (i ¼ 1, 2,…, I ) is the covariate, and i is the number of covariates. Here, six variables are selected as alternative covariates, namely precipitation (P), temperature (T ), relative humidity (H ), wind speed (W ), potential evapotranspiration (E) and antecedent runoff (R). The RS algorithm is used to fit the nonstationary Gamma model in the GAMLSS framework (Rigby & Stasinopoulos , ). After fitting the nonstationary Gamma model, the remaining calculation process of NSRI is the same as SRI, converting the corresponding cumulative probability of nonstationary Gamma model to a standard normal distribution function. To prevent the model overfitting,
Table 2
|
Drought-level standards and the threshold values
Index value
Category
>2.00
Extreme wet
1.99 to 1.50
Very wet
1.49 to 1.00
Moderate wet
0.99 to 0.00
Near normal
0.00 to 0.99
Mild drought
1.00 to 1.49
Moderate drought
1.50 to 1.99
Severe drought
2.00
Extreme drought
X. Sun et al.
899
Table 3
|
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
Results of the ADF test for cumulative runoff data in the study area
QL
ZMSK
YLX
Month
P-value
Dickey–Fuller
P-value
Dickey–Fuller
P-value
Dickey–Fuller
1
0.02
3.84
0.26
2.78
0.07
3.34
2
0.02
3.84
0.26
2.77
0.07
3.35
3
0.02
3.81
0.26
2.77
0.07
3.34
4
0.03
3.75
0.26
2.79
0.07
3.35
5
0.03
3.70
0.28
2.73
0.11
3.16
6
0.02
3.95
0.30
2.68
0.17
3.01
7
0.02
3.86
0.28
2.74
0.21
2.90
8
< 0.01
4.23
0.24
2.83
0.22
2.89
9
< 0.01
4.16
0.24
2.84
0.10
3.17
10
< 0.01
4.15
0.27
2.75
0.09
3.26
11
0.01
4.14
0.27
2.76
0.08
3.28
12
0.01
4.14
0.27
2.76
0.08
3.29
statistic values directly. The P-value of this test in this pack-
stations, the results are opposite. Therefore, we conclude
age is obtained through interpolating from the critical value
that the cumulative runoff at these two stations can be con-
table of Banerjee et al. (). If the computed P-value is
sidered as nonstationary series. To model such series, the
larger than a given significance level, such as 0.05, the null
NSRI needs to be developed.
hypothesis is accepted. Otherwise, it is rejected. As shown in Table 3, the Dickey–Fuller statistics at the
Correlation test
QL station are less than 3.5, and the corresponding P-values are less than 0.05, which means that, at the 0.05
The selection of a suitable covariate is of great importance in
significance level, the cumulative runoff series at the QL
developing a NSRI. To test whether the runoff and the
station exhibit a stationary state. While at ZMSK and YLX
alternative variables over the study area are related to
Table 4
|
The correlation coefficients between runoff and the six variables in the study area
Pearson
Kendall
Spearman
Station
Variable
r
Sig.
r
Sig.
r
Sig.
ZMSK
Precipitation (P) Temperature (T ) Relative humidity (H ) Wind speed (W ) Potential evapotranspiration (E) Antecedent runoff (R)
0.892 0.786 0.759 0.138 0.654 0.706
0.000 0.000 0.000 0.000 0.000 0.000
0.667 0.728 0.611 0.054 0.538 0.615
0.000 0.000 0.000 0.038 0.000 0.000
0.862 0.911 0.807 0.079 0.761 0.825
0.000 0.000 0.000 0.042 0.000 0.000
YLX
Precipitation (P) Temperature (T ) Relative humidity (H ) Wind speed (W ) Potential evapotranspiration (E) Antecedent runoff (R)
0.918 0.788 0.813 0.041 0.654 0.718
0.000 0.000 0.000 0.294 0.000 0.000
0.693 0.727 0.641 0.056 0.532 0.618
0.000 0.000 0.000 0.033 0.000 0.000
0.879 0.911 0.839 0.094 0.759 0.830
0.000 0.000 0.000 0.016 0.000 0.000
900
X. Sun et al.
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
each other, a correlation test is carried out by using Spear-
Comparing the results of different numbers of covari-
man, Pearson and Kendall correlation tests. The test
ates, it can be found that the AIC values for M2 (with two
results are shown in Table 4, in which r indicates the corre-
covariates) show a little lower than those for M1 (with one
lation coefficient, sig. indicates the significance level. If sig.
covariate), also lower than those for M3 (with three covari-
<0.05, there is a significant correlation between the two
ates) (Table 5). Thus, M2 are selected for our analysis finally.
variables, and the higher the absolute value of r, the stronger
Specifically, M21 (precipitation and temperature as covari-
the correlation. Among the six variables, five of them
ates) and M22 (precipitation and relative humidity as
(precipitation, temperature, relative humidity, potential
covariates) are selected for ZMSK and YLX stations,
evapotranspiration and antecedent runoff) are strongly cor-
respectively, according to the lowest AIC values.
related with runoff with correlation coefficients greater than
To further evaluate the quality of fit, taking August
0.5, and wind speed shows weak correlations. Therefore,
runoff series at the YLX station as an example, the worm
except for wind speed, all variables are selected as covari-
plots of M0 and the best nonstationary M22 are provided
ates to develop the NSRI.
in Figure 2. The worm plot was first used by van Buuren
Nonstationary modeling with GAMLSS
residual point is within the acceptable area surrounded by
() to determine the model fit based on whether the the elliptic curve. According to Figure 2(a), it can be Goodness of fitting
observed that although most residual points are within the acceptable area, minority points are at 95% confidence
For better understating the interactions of different factors
level boundary and even outside the acceptable area,
and finding the main influencing factors of runoff variation
which means that M0 does not meet the reliability require-
in different seasons, four different types of models are devel-
ments quite well. While, in Figure 2(b), the M22 residual
oped with adding different variables to each nonstationary
points are all in the acceptable area, indicating that this
cases, Model 0 (M0), Model 1 (M1), Model 2 (M2) and
model satisfies the reliability and fits well. The performance
Model 3 (M3). M0 is the stationary case with constant
of the nonstationary model with meteorological elements as
Gamma parameters, and the remaining three are all non-
covariates was thus confirmed for August runoff series at the
stationary cases. M1, M2 and M3 take one, two and three
YLX station.
variables as covariates, respectively. Different combinations of covariates are shown in Table 5 for details, with 16 differ-
Percentile estimation
ent models developed in total. M11–M15 represent the nonstationary models with one covariate, M21–M27 rep-
Figure 3, also taking the cumulative runoff series in August at
resent the nonstationary models with two covariates and
the YLX station as an example, shows the five estimated per-
M31–M34 represent the nonstationary models with three
centile results (5th, 25th, 50th, 75th and 95th) from M0 and
covariates. The AIC values corresponding to each model
M22, as well as the observed runoff series. As shown in
are also displayed in the table. The model with the smallest
Figure 3(a), the quantile grayscale image is horizontally
AIC value is regarded as the best model.
linear, and the dynamic change of runoff series under climate
As shown, the stationary model, M0 has larger AIC
change conditions cannot be described well. Figure 3(b)
values, presenting the poor performance in the fitting.
reflects that the percentiles are constantly changing with
Comparatively, the nonstationary models, M1–M3, with
the observed data, and M22 with two meteorological factors
assuming that location and scale parameters in Gamma
as covariates shows better performance in describing the
change with the covariates, perform better in terms of the
nonlinear variation in the runoff data. It also implies that
lower AIC values. This is also largely supported by the
the influence of meteorological factors to runoff sequence
result of the stationary test. That is, the nonstationary time
cannot be neglected.
series show poor performance with stationary models and better performance with nonstationary models.
Overall, in the context of climate change, the nonstationary
Gamma
distribution
with
considering
X. Sun et al.
901
Table 5
|
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
Summary of AIC values for stationary (M0) and the nonstationary Gamma models (M1, M2, M3) for runoff data in the study area
Station
Model
ZMSK
M0 M1
M2
M3
YLX
M0 M1
M2
M3
Covariate
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
M0 M11 M12 M13 M14 M15 M21 M22 M23 M24 M25 M26 M27 M31 M32 M33 M34
Null P T H E R PT PH PE PR TH TE HE PTH PTE PHE THE
507 502 512 510 506 494 492 497 503 503 505 507 510 497 503 498 510
507 502 510 516 515 498 494 498 506 503 514 512 510 499 507 501 514
508 507 509 515 517 497 496 507 505 507 512 517 514 506 505 505 513
507 506 514 514 512 494 494 505 507 506 512 512 503 505 507 506 504
507 504 515 506 515 505 494 504 506 504 505 514 501 504 506 505 500
511 507 506 506 503 503 497 507 508 507 507 507 503 507 507 508 506
514 510 514 516 519 503 500 510 510 510 515 519 518 510 510 510 518
513 499 502 523 520 501 499 509 509 507 516 519 519 509 509 509 521
506 501 497 512 512 503 490 500 500 495 510 497 511 500 500 501 508
507 499 508 510 515 500 498 500 498 499 508 513 509 499 497 499 511
509 503 515 516 515 493 493 504 502 503 511 512 514 504 502 503 513
510 504 509 508 512 494 493 500 505 503 496 518 499 500 505 501 494
M0 M11 M12 M13 M14 M15 M21 M22 M23 M24 M25 M26 M27 M31 M32 M33 M34
Null P T H E R PT PH PE PR TH TE HE PTH PTE PHE THE
509 495 511 514 515 491 497 486 501 495 515 513 513 489 502 494 506
509 497 503 508 513 495 498 489 502 497 512 514 513 491 503 496 500
510 498 508 512 510 499 499 497 501 498 512 509 512 498 502 501 511
510 490 510 515 516 498 490 488 492 491 515 515 507 488 493 490 508
508 508 514 510 515 503 508 499 505 508 509 514 501 509 505 506 501
510 503 503 494 508 503 503 494 499 501 506 512 514 506 499 502 513
513 510 509 509 518 501 510 499 512 511 509 517 518 509 512 511 519
515 509 498 520 517 499 509 497 510 507 520 519 519 507 510 510 522
507 490 501 512 511 491 490 495 483 506 511 506 506 495 482 488 501
510 491 507 513 513 498 492 490 494 499 513 511 516 492 495 491 515
511 497 520 511 513 500 497 500 496 500 502 511 499 500 494 499 497
512 499 516 502 512 495 499 494 499 500 500 523 508 498 500 499 504
Note: P, precipitation; T, temperature; H, relative humidity; E, potential evapotranspiration; R, antecedent runoff. The smallest AIC value in each type of model is denoted in bold font.
Figure 2
|
Worm plots of M0 (a) and M22 (b) for August runoff series at the YLX station.
902
Figure 3
X. Sun et al.
|
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
Results of M0 (a) and M22 (b) for 12-month aggregated runoff series in August at the YLX station. The light grey region represents the area between the 0.05 and 0.95 quantiles, the dark grey region represents the area between the 0.25 and 0.75 quantiles, and the black line represents the median (0.5 quantile). The black dots are the observed values.
Figure 4
|
Twelve-month SRI and NSRI from 1961 to 2014 in the study area.
903
X. Sun et al.
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
meteorological factors as covariates tends to more robust in
several significant differences between SRI and NSRI esti-
describing the variation behavior exhibited in the runoff
mates can also be observed (Figure 4). For example, from
data in our study case.
August 1984 to May 1985, August 1999 to June 2000, and September 2010 to June 2011, SRI detects normal, while NSRI detects them as mild droughts.
Hydrological drought assessment Drought characteristics Estimation of the SRI and the NSRI The occurrence frequency of different drought grades is calFigure 4 shows the time series of drought index calculated
culated from both the SRI and the NSRI. As shown in
from the observed runoff data at three hydrological
Figure 5, compared with the results of the SRI, moderate
stations. For the QL station, only stationary SRI sequences
and severe droughts occur more frequently, while mild
are calculated from 1961 to 2014. This study defines a
and extreme droughts are less frequent according to the
drought event as a period of time when the index value
NSRI for the ZMSK station. For the YLX station, the fre-
is continuously negative and reaches a threshold of 1.0
quency of mild drought is higher, and the frequencies of
or lower. In this respect, the dry periods for the QL station
moderate and severe droughts are slightly lower in terms
were from July 1970 to September 1971, from November
of the NSRI.
1972 to August 1974, from June 1978 to June 1980, from
It also can be found that, in Figure 5, the two stations
September 1982 to May 1983, and from May 1992 to
present completely opposite results in the frequency of
August 1992. For ZMSK and YLX stations, both SRI and
drought occurrence, especially at mild and moderate
NSRI sequences are demonstrated in the figure. Overall,
grades. For the ZMSK station, the frequency of mild
the SRI and NSRI series at both stations provide close
drought decreased and that of moderate drought increased,
drought index in most cases, while also defining several
while for the YLX station, the results are the opposite. This
differences.
can be explained from the locations of the two stations. As
For the ZMSK station, the period of October 1990 to
displayed in Figure 1, the ZMSK station is located
January 1991 is detected as normal state from the SRI, but
upstream of the YLX station. The moderate drought in
moderate drought from the NSRI, the period of June to
the upstream (where the ZMSK station is located) may
July 2011 is detected as wet from the SRI, while drought
merely cause the weaker drought (mild drought) in the
from the NSRI. For the YLX station, both SRI and NSRI
downstream (where the YLX station is located), since
recorded extreme drought events from August 1973 to July
there are other tributaries flowing into the main stream
1974. The results obtained are consistent with historical
and then alleviating the drought grade in the downstream
records and related literature (Yang et al. ). However,
to some extent.
Figure 5
|
Drought occurrence frequency based on SRI and NSRI in the study area.
X. Sun et al.
904
Table 6
|
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
Drought characteristics from the SRI and the NSRI in the study area
SRI Station
Number
Peak
Longest duration (months)
Maximum severity
QL
11
2.92
25
59.83
ZMSK
12
2.29
21
26.69
YLX
10
2.63
22
43.73
NSRI Station
Number
Peak
Longest duration (months)
Maximum severity
ZMSK
18
2.42
22
33.62
YLX
19
2.47
22
37.66
The run-length theory can be used to extract drought
Discussion
charateristics, including the peak, duration and severity (Yevjevich ). Drought peak is defined as the minimum
Number of covariates
of the drought index during a drought event. Drought duration refers to the duration of drought from the start to the
Many meteorological factors show strong effects on
end of a drought event. Drought severity refers to the cumu-
droughts under the global warming in recent decades (e.g.,
lative sum of the differences between the drought index and
Núñez et al. ; Zarch et al. ), so many of these factors
the threshold during the drought (Yevjevich ; Gu et al.
are taken into account in the study of hydrological droughts.
). A threshold of 1.0 was set to identify the occurrence
Here, six variables are considered, and three nonstationary
of drought events in this study. Drought peak, duration and
models are developed. When only one variable is con-
severity were then extracted from the drought events accord-
sidered, M1, in which the variable of precipitation is
ing to the run-length theory. As shown in Table 6, the QL
taken as a covariate, shows the best for 9 of 24 series, and
station identifies 11 drought events, of which the drought
the series are concentrated basically in April, May, August,
peak reaches 2.92, the longest drought duration lasts for
September and October (Table 5). It indicates that precipi-
25 months and the maximum drought severity is 59.83.
tation is the main factor affecting runoff changes during
The ZMSK station identifies more drought events (18
these months. It is understandable since these months are
events) based on the NSRI, with higher drought peak
usually with more precipitation in the study area. M15,
( 2.42) and higher drought severity ( 33.62), compared
taking the variable of antecedent runoff as a covariate,
with the results of SRI (12 events, 2.29, 26.69). The
shows the best for 13 of 24 series, and most of them are con-
YLX station also identifies more drought events (19
centrated from November to about next April. This finding
events) based on the NSRI, while the drought peak
keeps consistent with the river recharge resources in differ-
( 2.47) and the maximum drought severity ( 37.66) are
ent seasons. In rainy seasons (e.g., summer), precipitation
lower than those from the SRI ( 2.63 and 43.73). The
contributes the runoff most, while in dry seasons (e.g.,
longest drought durations are quite close from both the
winter and spring), the river is mostly recharged by glacial
NSRI and the SRI for this station.
and snow melting water, permafrost melting water and
Table 7
|
The AIC values for nonstationary models with time as a covariate for runoff data in the study area
Station
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
ZMSK
497
497
498
497
496
497
505
498
490
499
494
496
YLX
495
496
498
497
495
495
501
498
485
492
494
495
905
X. Sun et al.
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
groundwater over the study area (Li et al. ; Wang et al.
M1, seven in M2 and four in M3). It is observed that in
). Pei et al. () and Wang et al. () concluded that
Figure 6(a) and 6(c), in the case of taking the variable of pre-
about 75% of runoff are recharged by precipitation in rainy
cipitation as the covariate, no matter what kinds of
seasons and more than 75% are recharged by snow melting
nonstationary models are (M21, M22, M23 and M24 with
in dry seasons in the Heihe River basin.
two covariates or M31, M32 and M33 with three covariates),
In the case of taking two variables as covariates (M2
the derived drought index are quite close with those from
models), all the corresponding AIC values show slightly
M11 (in which only the variable of precipitation is selected
lower than those for M1 models (Table 5). It indicates that
as the covariate). That is, as long as the combinations of pre-
two variables are better in describing the nonstationary
cipitation and other variables are taken as covariates for the
characteristics of the runoff data. In the case of taking
nonstationary model, the derived drought index curves are
three variables as covariates (M3 models), however, the
quite close. It indicates that the hydrological drought in
AIC values have gone up instead. This may be the result of
this study is mainly affected by precipitation, and the contri-
over-fitting caused by the over-complexity of the model.
bution of precipitation to hydrological droughts is great over
Generally speaking, when the number of covariates
the study area. This conclusion is also confirmed by Li et al.
increases, the complexity of the model increases, the likeli-
() and Wang et al. (, b). When other variables,
hood function also increases, it leads to AIC value
rather than precipitation, are considered as covariates
decreases. Whereas when the number is too large, the
(M12, M13, M14, M15, M25, M26, M27 and M34), the
growth rate of likelihood function slows down, which thus
derived drought index curves fluctuate a great deal, and
results in the increase of AIC (Akaike ). In addition,
large differences are found between them (Figure 6(b) and
as stated by Agilan & Umamahesh (), the increase of
6(d)). This indicates that the results of drought assessment
covariates would bring greater uncertainty in the final
differ greatly without considering precipitation. All these
assessment. Thus, M2, with two variables as covariates,
mean that, although other variables have certain effects to
are finally selected for our analysis. Note that many studies also included the time as a cov-
hydrological droughts, precipitation is still the major influence factor in hydrological drought in our case.
ariate in a nonstationary analysis (Russo et al. ; Wang et al. a; Javad & Somayeh ), although some studies
Performance of the SRI and the NSRI
have proved that time is not suitable in their cases due to its linear monotonic trends (Li et al. ). It is considered
According to Figure 4, the SRI and the NSRI at ZMSK and
as an additional covariate for comparison to discuss the
YLX stations display the close drought curves in most cases,
nonstationary runoff process in this study. As illustrated in
while there are still several differences. These differences are
Table 7, the corresponding AIC value is moderate, slightly
attempted to be explained by the meteorological variables.
higher than those from the selected models (M21 for the
According to Jiang et al. (), the hydrological drought
ZMSK station and M22 for the YLX station) and lower
in the upper reach of Heihe River basin lag behind the
than those from the stationary models (M0). Besides, since
meteorological drought by about 4 months. Therefore, we
using time as an explanatory variable has no physical mean-
deduce that the hydrological drought situation of a certain
ing, it can only describe the general trend of a hydrological
period is possibly related with the precipitation that slides
sequence over time (Vu & Mishra ). Thus, the time is
forward for 4 months.
not considered as a covariate in our nonstationary analysis.
For the ZMSK station, the period of October 1990 to January 1991 is detected as normal state from the SRI,
Effects of covariates to drought index
while moderate drought from the NSRI. It is found that the precipitation sliding forward for 4 months (from June
To discuss the effects of different covariates to the drought
to September in 1990) was 328.7 mm, lower than the
index, we compare the drought index calculated from
multi-year mean value of 333.7 mm for the same period.
three types and 16 nonstationary models (five models in
Less precipitation is more likely to cause drought, and
906
Figure 6
X. Sun et al.
|
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
Differences in the NSRI with different covariates in the study area (the models in the legend are the same as those in Table 5).
thus the conclusion drawn from the NSRI (drought) seems
March 2011 (6.4 mm) is less than the multi-year mean
to be more reliable. For the period of June to July 2011, it
value (11.0 mm), and the mean temperature (9.04 C)
is found that the observed precipitation from February to
shows higher than the multi-year mean value (8.28 C).
907
X. Sun et al.
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
Owing to the less precipitation and the higher temperature
of the SRI, but corresponds to mild drought (May 1964),
during this period, it is concluded that drought is more
normal state (June 1975) and moderate drought (December
likely to occur, and thus the detection from the NSRI
2004) in terms of the NSRI. A monthly runoff of 183 mm is
(drought) is more reasonable than that from the SRI (wet).
classified as normal state from the SRI, while classified as
For the YLX station, the drought conditions for periods
moderate wet (February 1982), normal (August 1999) and
of August 1984 to May 1985, August 1999 to June 2000 and
mild drought (September 2010) states from the NSRI.
September 2010 to June 2011 are different from the SRI and
In a changing environment, hydrological drought can no
the NSRI. It is found that the precipitation sliding forward
longer be evaluated solely on runoff. Zhang et al. ()
for 4 months in these three periods (336.8, 327.8 and
investigaed the impacts of both climate variations and
333.6 mm) are less than their multi-year mean value for
human activities to hydrological droughts in the middle
the same period (364.9, 358.3 and 345.5 mm), and the aver-
reach of Yangtze River in China. Wang et al. (b) incor-
age relative humidity of these three periods (54.50, 53.48
porated the climate-driven and human-induced variables
and 51.76%) are also lower than the multi-year mean
as covariates in the nonstationary model to indicate the non-
value (55.47, 54.27 and 52.76%). Therefore, the occurrence
stationary features of runoff series. In this study, the
of drought (drawn from the NSRI), rather than the normal
determination of NSRI depends not only on runoff data
state (drawn from SRI), is regarded as more reasonable.
but also on the key factors affecting runoff. Variables includ-
Another observation is that the same runoff value corre-
ing precipitation and temperature are considered as
sponds the same drought classification based on the SRI,
covariates in characterizing the nonstationary features in
which is because its calculation solely depends on runoff
the runoff data, and thus drought state detected by the
data (Shukla & Wood ), and the drought state detected
NSRI varies with the influence of covariates.
from the SRI keeps stable. However, this conclusion is not
We also calculated the NSRI with time as a covariate
applicable to the nonstationary model. The NSRI leads to
(denoted as the TSRI here) for comparison (Figure 7). It is
differences in drought classification with the same runoff
observed that TSRI differs a great deal with SRI especially
value. For example, a monthly runoff of 150 mm for the
for the first and last several years. This is because the location
YLX station corresponds to a mild drought state in terms
parameter is constant and close to its mean value in the SRI,
Figure 7
|
Twelve-month SRI and TSRI from 1961 to 2014 in the study area.
908
X. Sun et al.
|
Assessment of hydrological drought based on nonstationary data
Hydrology Research
|
51.5
|
2020
while it is time-dependent in the TSRI. This condition leads to
In general, the current study solves the problems of
more differences between TSRI and SRI for the early and last
hydrological drought assessment based on nonstationary
years of the record period and less differences for the middle
runoff data. It is proved that the proposed new drought
part of the period (Javad & Somayeh ). Furthermore, the
index, NSRI, is capable of providing more reliable results
linear monotonic trend of time also makes it not as appropri-
for drought assessment over the study area. Revelant find-
ate as other meteorological variables in characterizing the
ings are also very helpful in developing strategies for
nonstationary processes of runoff series (Li et al. ).
coping with local droughts and water resource risk manage-
These results support that, compared with the tra-
ment. It should be noted that, although the findings here are
ditional SRI, the new developed NSRI, with considering
in the context of NSRI at 12-month scale and limited
meteorological variables as covariates, is proved to be
stations, the framework and the methods in this study
more robust and more applicable in this drought assessment.
could be extended over further scales and regions. This paper mainly considers the relation of several meteorological elements to hydrological droughts. In fact, hydrological droughts are also induced by factors like the
CONCLUSIONS
underlying surface of the basin, vegetation coverage and human activities (such as reservoir opening and irrigation)
Appropriate drought indices are often important tools for
(Shukla & Wood ; Fang et al. b; Zhang et al.
regional drought assessments. In the context of changing
; Wang et al. a, b, c). The use of appropriate
environment, the traditional SRI which is defined based
quantitative factors to characterize human activities will be
on a stationary Gamma distribution is no longer suitable
considered in the future research for making the nonstation-
due to the nonstationary characteristics of runoff series
ary drought index more reliable under the complex
(Villarini et al. ; Russo et al. ). Thus, the NSRI is
conditions. Additionally, to identify the hydrological
developed in this study within the GAMLSS framework.
drought propagation and its key influencing factors is also
The proposed new drought index considers climate change
a hot topic and can be included in the future studies.
and incorporates six variables into calculation, accounting for the nonstationary features of runoff data. As a quantitative indicator, just like the SRI, the NSRI can assess drought
ACKNOWLEDGEMENTS
events at different time scales (e.g., 3, 6, 9 and 12 months) and allow comparison of climatic conditions at different
This study is supported by the Fundamental Research Funds
locations. The 12-month time-scale NSRI is applied to the
for the Central Universities (No. 39435832015028).
upper reach of the Heihe River basin. The conclusions can be summarized as follows. For the study area, (1) the fitting results of the nonstationary models
DATA AVAILABILITY STATEMENT
to those nonstationary runoff data (ZMSK and YLX stations) are much better than the stationary models according to AIC values; (2) for the nonstationary models with one covariate, precipitation shows better in the fitting as a covariate in rainy seasons, and antecedent runoff shows better in dry seasons, which is highly related with the recharge sources in
The meteorological dataset is available from the China Meteorological Data Network (http://data.cma.cn/). The hydrological dataset is available from the Hydrological Yearbook of the People’s Republic of China.
different seasons; the nonstationary models with two covariates perform better than those with one or three covariates;
REFERENCES
(3) the NSRI identifies more drought events than the SRI, and precipitation is the major factor affecting the hydrological drought conditions over the study area.
Aa, Y., Wang, G., Liu, T., Xue, B. & Kuczera, G. Spatial variation of correlations between vertical soil water and
909
X. Sun et al.
|
Assessment of hydrological drought based on nonstationary data
evapotranspiration and their controlling factors in a semiarid region. Journal of Hydrology 574, 53–63. Agilan, V. & Umamahesh, N. V. Covariate and parameter uncertainty in non-stationary rainfall IDF curve. International Journal of Climatology: A Journal of the Royal Meteorological Society 38 (1), 365–383. Akaike, H. A new look at the statistical model identification. IEEE Trans Autom Control 19 (6), 716–723. Akantziliotou, K., Rigby, R. A. & Stasinopoulos, D. M. The R implementation of generalized additive models for location, scale and shape. In: Statistical Modelling in Society: Proceedings of the 17th International Workshop on Statistical Modelling (M. Stasinopoulos & G. Touloumi, eds). Chania, pp. 75–83. Allen, R. G., Pereira, L. S., Raes, D. & Smith, M. Crop Evapotranspiration Guidelines for Computing Crop Water Requirements. Food and Agriculture Organization of the United Nations Rome, Italy. Banerjee, A., Dolado, J. J., Galbraith, J. W. & Hendry, D. F. Cointegration, Error Correction, and the Econometric Analysis of Non-Stationary Data. Oxford University Press, Oxford. Bloomfield, J. P. & Marchant, B. P. Analysis of groundwater drought building on the standardised precipitation index approach. Hydrology and Earth System Sciences 17 (12), 4769–4787. Cheng, M. Impact and Assessment of Drought Effects on Natural Ecosystem: A Case Study of Hei River Basin. Nanjing University of Information Science and Technology, Nanjing, China. Dickey, D. A. & Fuller, W. A. Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association 74 (366), 427. Ding, Y. J., Ye, B. S. & Liu, S. Y. Effect of climatic factors on streamflow in the alpine catchment of the qilian mountains. Acta Geographica Sinica 54 (05), 431–437. Fang, Q., Wang, G., Liu, T. & Xue, B. A. Y. a Controls of carbon flux in a semi-arid grassland ecosystem experiencing wetland loss: vegetation patterns and environmental variables. Agricultural and Forest Meteorology 259, 196–210. Fang, Q., Wang, G., Xue, B., Liu, T. & Kiem, A. b How and to what extent does precipitation on multi-temporal scales and soil moisture at different depths determine carbon flux responses in a water-limited grassland ecosystem? The Science Of The Total Environment 635, 1255–1266. Gao, X. C. & Zhao, L. The prediction and analysis of population carrying capacity in Heihe basin. Northwest Population Journal 31 (03), 120–123 þ 129. Gu, J. S., Li, Q. F., Niu, M. Y., Chen, Q. H., He, P. F., Zhou, Z. M., Zeng, T. S., Du, Y., Song, Y. & Han, X. Y. An analysis of meteorological drought characteristics in the Upper Reaches of Huaihe River based on multidimensional copula function. China Rural Water and Hydropower 61 (08), 83–87 þ 92. Guttman, N. B. Accepting the standardized precipitation index: a calculation algorithm. Journal of the American Water Resources Association 35 (2), 311.
Hydrology Research
|
51.5
|
2020
Han, D., Wang, G., Xue, B., Liu, T. A. Y. & Xu, X. a Evaluation of semiarid grassland degradation in North China from multiple perspectives. Ecological Engineering 112, 41–50. Han, D., Wang, G. & Liu, T. b Hydroclimatic response of evapotranspiration partitioning to prolonged droughts in semiarid grassland. Journal of Hydrology 563, 56. Hastie, T. J. & Tibshirani, R. J. Generalized Additive Models. Chapman and Hall, London. Javad, B. & Somayeh, H. A non-stationary reconnaissance drought index (NRDI) for drought monitoring in a changing climate. Water Resources Management 32 (8), 2611–2624. Jiang, Y. W., Zhang, X. F., Yang, L. X. & He, C. S. Analysis and comparison of spatial and temporal patterns of meteorological and hydrological drought indices in the Upper Reach of the Heihe River Watershed. Resources Science 36 (09), 1842–1851. Jiang, S. H., Wang, M. H., Ren, L. L., Xu, C. Y., Yuan, F., Liu, Y. & Shen, H. R. A framework for quantifying the impacts of climate change and human activities on hydrological drought in a semiarid basin of Northern China. Hydrological Processes 33 (7), 1075–1088. Li, L., Wang, Z. N. & Wang, Q. C. Influence of climatic change on flow over the Upper Reaches of Heihe River. Scientia Geographica Sinica 26 (01), 40–46. Li, J. Z., Wang, Y. X., Li, S. F. & Hu, R. A nonstationary standardized precipitation index incorporating climate indices as covariates. Journal of Geophysical Research 120 (23), 12082–12095. Ling, P. Q. The Coordinated Development Reseach of Population, Resource, Environment and Economy of Heihe River Basin. LanZhou University, Lanzhou, China. Liu, S. Y., Huang, S. Z., Xie, Y. Y., Wang, H., Leng, G., Huang, Q., Wei, X. & Wang, L. Identification of the non-stationarity of floods: changing patterns, causes, and implications. Water Resources Management 33 (3), 939–953. Luciano, T., Michele, L. & Ignacio, L. M. Investigation of scaling properties in monthly streamflow and Standardized Streamflow Index (SSI) time series in the Ebro basin (Spain). Physica A: Statistical Mechanics and its Applications 391 (4), 1662–1678. McKee, T. B., Doesken, N. J. & Kleist, J. The relationship of drought frequency and duration to time scales preprints. In 8th Conference on Applied Climatology. pp. 179–184. Nalbantis, I. & Tsakiris, G. Assessment of hydrological drought revisited. Water Resources Management 23 (5), 881–897. Nelder, J. A. & Wedderburn, R. W. M. Generalized linear models. Journal of the Royal Statistical Society A 135, 370–384. Núñez, J., Rivera, D., Oyarzún, R. & Arumí, J. L. On the use of standardized drought indices under decadal climate variability: critical assessment and drought policy implications. Journal of Hydrology 517, 458–470. Pei, H., Fang, S. F., Liu, Z. H. & Qin, Z. H. Design and application of distributed snowmelt-runoff model. Resources Science 30 (03), 454–459.
910
X. Sun et al.
|
Assessment of hydrological drought based on nonstationary data
Rigby, R. A. & Stasinopoulos, D. M. The GAMLSS project: a flexible approach to statistical modelling. In: New Trends in Statistical Modelling: Proceedings of the 16th International Workshop on Statistical Modelling (B. Klein & L. Korsholm, eds). Odense, pp. 249–256. Rigby, R. A. & Stasinopoulos, D. M. Generalized additive models for location, scale and shape. Applied Statistics 54, 507–554. Russo, S., Dosio, A., Sterl, A., Barbosa, P. & Vogt, J. Projection of occurrence of extreme dry-wet years and seasons in Europe with stationary and nonstationary standardized precipitation indices. Journal of Geophysical Research [Atmospheres] 118, 7628–7639. Shukla, S. & Wood, A. W. Use of a standardized runoff index for characterizing hydrologic drought. Geophysical Research Letters 35, L02405. Stasinopoulos, D. M. & Rigby, R. A. Generalized additive models for location scale and shape (GAMLSS) in R. Journal of Statistical Software 23 (7), 1–46. Tabari, H., Nikbakht, J. & Hosseinzadeh, T. P. Hydrological drought assessment in Northwestern Iran based on streamflow drought index (SDI). Water Resources Management 27 (1), 137–151. van Buuren, Y. Basic back: whether or not you’re among the 60 to 80% of us who experience severe and even debilitating back pain, you can have a healthy and pain-free back forever. Canadian Living 26 (4), 79. Vicente-Serrano, S. M., López-Moreno, J. I., Beguería, S., LorenzoLacruz, J., Azorin-Molina, C. & Morán-Tejeda, E. Accurate computation of a streamflowdrought index. Journal of Hydrologic Engineering 17 (2), 318–332. Villarini, G., Serinaldi, F., Smith, J. A. & Krajewski, W. F. On the stationarity of annual flood peaks in the continental United States during the 20th century. Water Resources Research 45 (8), W08417. Vu, T. M. & Mishra, A. K. Nonstationary frequency analysis of the recent extreme precipitation events in the United States. Journal of Hydrology 575, 999–1010. Wang, S. L., Liu, X. D., Jin, M., Zhang, X. L., Che, Z. X. & Wang, R. X. The impact of temperature and precipitation on the streamflow in the middle part of the Qilian Mountains, Northwestern China. Journal of Arid Land Resources and Environment 25 (01), 162–165. Wang, G., Liu, S. & Liu, T. a Modelling above-ground biomass based on vegetation indexes: a modified approach for biomass estimation in semi-arid grasslands. International Journal of Remote Sensing 40 (10), 3835–3854.
Hydrology Research
|
51.5
|
2020
Wang, Y. X., Duan, L. M., Liu, T. X., Li, J. Z. & Feng, P. b A non-stationary standardized streamflow index for hydrological drought using climate and human-induced indices as covariates. Science of The Total Environment 699, 134278. Wang, G., Li, J., Sun, W., Xue, B. A. Y. & Liu, T. c Non-point source pollution risks in a drinking water protection zone based on remote sensing data embedded within a nutrient budget model. Water Research 157, 238–246. Wang, Y. X., Li, J. Z., Feng, P. & Hu, R. a A time-dependent drought index for non-stationary precipitation series. Water Resources Management 29 (15), 5631–5647. Wang, Y. H., Yang, D. W., Lei, H. M. & Yang, H. B. b Impact of cryosphere hydrological processes on the river runoff in the upper reaches of Heihe River. Journal of Hydraulic Engineering 46 (09), 1064–1071. Xiong, L. H. & Guo, S. L. Trend test and change-point detection for the annual discharge series of the Yangtze River at the Yichang hydrological station. Hydrological Sciences Journal 49 (1), 99–112. Yang, L. X., Zeng, S. X., Jiang, Y. W., Gu, J. & He, C. S. Comparison of drought characteristics in the upstream of the Heihe River basin based on palmer drought severity index and standard precipitation index. Research of Soil and Water Conservation 24 (02), 132–136. Yevjevich, V. An Objective Approach to Definitions and Investigations of Continental Hydrologic Droughts, Vol. 6. Colorado State University, pp. 23–30. Zarch, M. A. A., Sivakumar, B. & Sharma, A. Droughts in a warming climate: a global assessment of standardized precipitation index (SPI) and reconnaissance drought index (RDI). Journal of Hydrology 526, 183–195. Zhang, X. Q., Sun, Y., Zheng, D. & Mao, W. Y. Regional response of temperature change in the arid regions of China to global warming. Arid Zone Research 27 (4), 592–599. Zhang, D., Zhang, Q., Qiu, J., Bai, P., Liang, K. & Li, X. Intensification of hydrological drought due to human activity in the middle reaches of the Yangtze River, China. Science of the Total Environment 637, 1432–1442. Zheng, J. T., Chen, F. L., Zhang, X. H., Long, A. H. & Liao, H. Analysis of design annual runoff of Manas River based on GAMLSS model. Climate Change Research 14 (03), 257–265. Zhou, H. H., Wang, Y. Q., Fang, G. H., Ye, Z. X. & Li, W. H. Application of standardized runoff index on hydrological drought characteristics identification in Aksu River. Journal of Water Resources and Water Engineering 30 (2), 6–11 þ 18.
First received 13 March 2020; accepted in revised form 10 August 2020. Available online 2 September 2020
911
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Spatiotemporal distributions and ecological risk assessment of pharmaceuticals and personal care products in groundwater in North China Jin Wu, Jingchao Liu, Zenghui Pan, Boxin Wang and Dasheng Zhang
ABSTRACT The contamination of surface water by pharmaceuticals and personal care products (PPCPs) has attracted widespread attention, but data regarding their impacts on groundwater (GW) are sparse. In river–GW interaction areas, rivers are likely an important source of PPCPs in aquifers, especially rivers impacted by sewage treatment plant effluent. Understanding the characterization, transport, and risk is valuable for the effective protection of vital aquatic ecosystem services, environmental health, and drinking water supplies. To attain this objective, statistics with spatial analysis and ecological risk were used to assess the effects of artificial recharge (AR) engineering on 16 PPCPs in aquifers in North China. The results indicated that 15 PPCPs were detected in unconfined and confined aquifers, with a few PPCPs being detected up to 1,000 ng/L. The most frequently detected
Jin Wu Jingchao Liu Zenghui Pan Boxin Wang Dasheng Zhang (corresponding author) Hebei Institute of Water Science, Shijiazhuang 050051, China E-mail: skyzhangdasheng@126.com Jin Wu Jingchao Liu College of Architecture and Civil Engineering, Beijing University of Technology, Beijing 100124, China
PPCPs were sulfisoxazole, sulfachloropyridazine, sulfamerazine, sulfamethazine, sulfamethoxazole, and ibuprofen. In addition, the spatial and seasonal variations in most PPCPs were significant. Furthermore, the maximum concentrations were compared to the predicted no-effect concentrations to evaluate the ecological risk, and four PPCPs were found to be of medium or high ecological risk. This study highlights that AR engineering has a significant ecological effect on GW. Key words
| ecological risk, groundwater, PPCPs, river–groundwater interaction, spatiotemporal distributions
INTRODUCTION China is one of the largest producers and consumers of
46% of the antibiotics were ultimately released into rivers
pharmaceuticals and personal care products (PPCPs) (Liu
through sewage effluent (Zhang et al. ). PPCPs have
& Wong ; Hanna et al. ). Due to widespread
been detected in surface water (SW), groundwater (GW),
consumption and inefficient treatment, using PPCPs has
sediment, soil, and vegetables around China, some of
put pressure on the environment, resulting in drinking
which are considered environmentally pseudopersistent
water resources in China being threatened (Thomas et al.
with continuous discharge (Kostich et al. ; Peng et al.
; Wang et al. a). Approximately 92,700 tons of anti-
; Wang et al. ; Zhang et al. ). Significant atten-
biotics were consumed in 2013 in China, and approximately
tion is being paid to the presence of PPCP residues in the environment due to their potential adverse effects on ecosys-
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying,
tems and human health. Specifically, studies have shown
adaptation and redistribution, provided the original work is properly cited
that mixtures of PPCPs exhibit greater effects than those of
(http://creativecommons.org/licenses/by/4.0/).
the individual compounds (Cleuvers ; Jiang et al. ;
doi: 10.2166/nh.2020.001
912
J. Wu et al.
|
Pharmaceuticals and personal care products in groundwater in North China
Hydrology Research
|
51.5
|
2020
Brandt et al. ). It is anticipated that environmental
the infiltration of sewage by artificial aquifer recharge and
legislation will be widened to cover a range of PPCPs.
bank filtration is expected to be the main source of PPCPs
However, the knowledge of their fate within water cycles
in unconfined aquifers (Teijon et al. ; Yang et al. ).
is still insufficient.
However, such infiltration processes are not expected to
GW maintains flows and levels in rivers and lakes, is
impact unconfined aquifers and deteriorate GW quality.
essential for the health of GW-dependent ecosystems, and
A few works have investigated the occurrence of PPCPs
is regarded as the most important source of drinking water
in confined aquifers, examining the selected PPCPs
in many parts of the world (Barnes et al. ; Lapworth
(diethyltoluamide, crotamiton, ethenzamide, propyphena-
et al. ). The riparian zone adjacent to rivers is often an
zone, carbamazepine, and caffeine) in confined aquifers
important area for residents in China, where the interactions
(up to 500 m in depth) in Tokyo due to leakage from decre-
between rivers and GW are active (Fang et al. ; Aa et al.
pit sewer networks (Kuroda et al. ). Generally, the GW
). The subsurface geochemistry in the riparian zone is
in unconfined aquifers is used as a drinking water source
changed greatly due to either the leakage of river water to
in North China and is considered a key factor for the
recharge the underlying GW or vice versa (Yang et al.
steady development of the national economy (Bo et al.
; Zhang et al. ). Increasing evidence shows that
; Li et al. ). Therefore, in developing management
rivers are one of the main sources of PPCPs in riverside
strategies to control GW pollution by PPCPs, the contami-
GW (Kuroda et al. ; Yang et al. a). Artificial recharge
nation and ecological risk of PPCPs in unconfined and
(AR) is considered a promising method to alleviate GW
confined aquifers both need to be assessed.
depletion by using reclaimed water, particularly in arid
Considering the issues raised above, the objectives of
areas (Grünheid et al. ; Li et al. ; Singh et al. ;
this study were to investigate the occurrence of PPCPs
Zheng et al. ). The PPCPs in rivers can enter GW
in different types of aquifers, including unconfined and con-
aquifers through the process of GW–SW exchange (i.e., via
fined aquifers, and to characterize the ecological risk for the
bank filtration or AR) (Buerge et al. ; Dougherty et al.
development of resistance to PPCPs in the GW as well as to
; Petrie et al. ). For example, Einsiedl et al. ()
optimize the management of GW recharging.
studied the transport of pharmaceuticals in karst GW systems affected by wastewater treatment. Díaz-Cruz & Barceló () reviewed PPCPs in different source waters
MATERIALS AND METHODS
used for artificial aquifer recharge purposes. In view of risk management and control, the ecological impact of AR
Study area
engineering on GW quality needs more attention since previous studies regarding ecological effects are sparse.
The studied river–GW interaction (RGI) area is located in
The ecological effect of PPCPs in GW, as well as their
North China (Figure 1). The area is characterized by hot,
relationship with environmental factors, is as yet poorly
humid summers and generally cold, windy, dry winters.
understood compared to those of SW. Previously, the detec-
Its annual temperature is ∼11.5 C, and the average precipi-
tion frequency and the ecological risk index have been
tation is ∼600 mm/a. AR engineering is using reclaimed
widely employed to describe the spatiotemporal character-
water which is produced by a membrane bioreactor with
istics of PPCPs in rivers and lakes (Sun et al. ; Yang
wastewater treatment plant effluent. Then the reclaimed
et al. b; Hanna et al. ; Bexfield et al. ). Since
water, about 38,000,000 m3/a, is introduced to the river.
more PPCPs are likely to have GW threshold values in the
Besides, some water containing treated and untreated
coming decades, more data related to occurrence and
municipal water from nearby factories and villages also
ecological effects are required. Moreover, studies seldom
flows into the river. The GW near the released reclaimed
discuss PPCPs in unconfined and confined aquifers simul-
water area is only recharged by reclaimed water in the wet
taneously recharged by reclaimed water. Among these
season. The mean concentrations of NH4þ, total nitrogen,
GW–SW exchange areas with sewage effluent discharge,
and total phosphorus in reclaimed water were 4.30, 10.64,
913
Figure 1
J. Wu et al.
|
|
Pharmaceuticals and personal care products in groundwater in North China
Hydrology Research
|
51.5
|
2020
GW sampling locations (a) and hydrogeological cross-sections of the study area (b).
and 0.78 mg/L, which indicate bad water quality. While
study area are widely distributed in plains and mountain
other water quality parameters ranged from medial to
valleys. The characteristics of the sedimentary layers are:
good quality according to the data analysis of 43 holes col-
north thin south thick and east thin west thick. The hydro-
lected in 2013 (Figure 1). The quaternary sediments in the
geological characteristics in the study area are that the
914
J. Wu et al.
|
Pharmaceuticals and personal care products in groundwater in North China
Hydrology Research
|
51.5
|
2020
depth of buried bedrock increases from north to south, the
control procedures were followed, usually including solvent
thickness of the quaternary strata is between 50 and
blank, procedure blank, and independent check standard.
300 m, and gradually increases from northeast to southwest.
Method detection limits (limits of detection, LODs) and
The lithology changes from coarse particles to fine particles,
quantification limits (limits of quantification, LOQs) were gen-
and the layers change from a single layer to multiple layers.
erally determined as the minimum detectable amount of an
The quaternary strata mainly include a sand pebble layer, a
analyte with a signal-to-noise ratio. Recoveries obtained by
sand gravel layer, and a silty clay layer.
spiking the analytes into GW ranged from 65 to 128%. The
Two subzones of the study area were identified, includ-
LOQs
were
0.2–6 ng/L
for
pharmaceuticals
in
GW.
ing the northern part (N zone) and southern part (S zone),
The concentrations of 16 PPCPs were determined for
based on the hydrogeological conditions. The N zone of
all GW samples. Sixteen PPCPs were classed and abbreviated
the study area was dominated by gravel and sand with
as
good permeability, while the S zone consisted of silty clay
sulfamonomethoxine (SMM), sulfathiazole (STZ), N4-acetyl-
with poor permeability. The lithology of the aquifer changed
sulfamethoxazol
follows:
enrofloxacin
(EFX),
(N4AcSMX),
erythromycin sulfisoxazole
(ETM), (SFS),
from sandy gravel to fine sand from north to south. The shal-
sulfachloropyridazine (SCP), sulfamerazine (SMR), sulfa-
low aquifer in the N zone is an unconfined aquifer (UA-N).
methazine (SMZ), sulfamethoxazole (SMX), trimethoprim
However, there are multiple layers of aquifers in the S
(TMP), caffeine (CAF), chloramphenicol (CAP), ibuprofen
zone with an unconfined aquifer (UA-S), first confined
(IBU), triclosan (TCS), and difloxacin (DIF). The main phys-
aquifer (FCA-S), and second confined aquifer (SCA-S). The
icochemical properties of the 16 target PPCPs are shown in
thicknesses of UA-N and UA-S ranged between 0–30 and
the Supplementary material, Table S1.
0–80 m, respectively. The thicknesses of FCA-S and SCA-S ranged between 30–50 and 50–80 m, respectively.
Leaching potential assessment
Sampling and analysis
Leaching potential assessment models were adopted to
GW samples were collected from 47 monitoring wells of riverside sections by the QED low-flow sampling equipment (Sample Pro™ sampling pump) in May 2016 (summer, wet season) and December 2016 (winter, dry season) (Figure 1). Twenty-four samples were collected in the N zone to represent the NS aquifer with the sampling depths at 50 m, and 20 samples were collected in the S zone to represent the UA-S, FCA-S, and SCA-S aquifers with the sampling depths at 30, 50, and 80 m, respectively. Detailed information about
the analysis procedure was
provided
elsewhere (Sui et al. ; Yang et al. b; Chen et al. ). In short, GW samples were pumped into 2 L glass bottles using a stainless-steel submersible pump. All samples were kept in precleaned containers at a cool temperature
assess the leaching potential of the selected PPCPs in the vadose zone. The model provides a quantitative value for representing the leaching potential. The model is described by Equation (1) by considering both the mobility and the persistence of chemicals. GUS ¼ log t1=2 (4 log KOC )
(1)
where GUS is the groundwater ubiquity score, KOC is the organic carbon partition coefficient, and t1/2 is the degradation half-life in the soil (days). The adjusted criteria were as follows: low leaching potential (GUS 1.8), moderate leaching potential (1.8 < GUS 2.8), and high leaching potential (2.8 GUS).
and then immediately transported to the laboratory for treatment. In the laboratory, the water samples were commonly
Ecological risk assessment
concentrated by preconditioned solid-phase extraction. The target antibiotics were subsequently analyzed using high/
The ecological risk assessment has been used to quantify the
ultra/ultrahigh-performance liquid chromatography–tandem
ecological effect exposed by environmental pollutants in
mass spectrometry. Appropriate quality assurance and quality
previous studies. In this study, risk quotients (RQs) were
915
J. Wu et al.
|
Pharmaceuticals and personal care products in groundwater in North China
Hydrology Research
|
51.5
|
2020
applied to assess the ecological risk of PPCPs in GW; a high
whole area, except for EFX. Six PPCPs need priority atten-
RQ suggests a high ecological risk and vice versa. The RQs
tion, including IBU, SMX, SMZ, SMR, SFS, and SCP,
for each PPCP in the GW sample were calculated with the
whose frequencies of detection are higher than 50% in
following equation:
each sampling season. This result was consistent with a previous study in which sulfonamides were the most extensively
RQ ¼
MEC PNEC
(2)
detected PPCPs in adjacent watersheds (Zou et al. ). However, sulfonamides in other countries, such as France, Germany, and Korea, had relatively lower concerns
where MEC is the measured concentration, and PNEC is the
regarding GW (Supplementary material, Table S3). This
predicted no-effect concentration. In this study, the chronic or
phenomenon can be explained by the different use patterns
acute toxicity data of the target antibiotics were collected from
of PPCPs between countries. In China, sulfonamides are
previous studies (Supplementary material, Table S2). The
extensively used in poultry and aquaculture because of
PNEC was obtained from the lowest no observed effect con-
their low price (Chen et al. ). The maximum concen-
centration (NOEC) with bold marks in Table S2. According
trations of each PPCP ranged from below the detectable
to the recommendations of the European technical guidance
limit to 1,186 ng/L, spanning several orders of magnitude.
document (TGD), NOECs were priority toxicity data (EC
Only TCS in FCA-S in the dry season had a peak con-
). Three levels of ecological risks were classified: between
centration of >1,000 ng/L. Large variations in PPCP
0.01 and 0.1 is low risk; between 0.1 and 1 is medium risk;
concentrations in this study may be attributed to variations
and larger than 1 is high risk.
in residence time and attenuation processes, such as dilution, adsorption to aquifer material, and degradation.
Statistical analysis
Therefore, the physicochemical properties of PPCPs also affect the migration and transformation in aquifers. By com-
Statistical tests were performed to identify major factors that
parison with other counties, the maximum concentration
likely affect the occurrence and contribution of PPCPs in
of PPCPs was at a relatively high level in a global context
GW. Most results for individual PPCPs were non-detections
(Table S3).
in this study, and the datasets do not conform to any distribution. Therefore, nonparametric statistical methods were
Temporal and spatial variations in PPCPs in GW
used to perform hypothesis testing. A paired samples t-test was conducted to compare the temporal differences
Temporal variations in PPCPs
between different seasons. Independent sample tests were conducted to compare the spatial differences between two
The type of PPCP detection frequency can be categorized
zones and among different aquifers. Statistical analyses
into three classes. The CAF is different from other PPCPs
were performed with SPSS Statistics V20.0 (SPSS, Inc.
with a higher detection frequency in the dry season
Quarry Bay, HK). The p-value used to indicate statistical sig-
(Table 1). For ETM, SMM, SCP, SMR, SMX, CAP, IBU,
nificance for all tests was 0.05.
and DIF, the detection frequencies in different aquifers were all higher in the wet season. For STZ, N4AcSMX, SFS, SMZ, TMP, and TCS, the highest detection frequencies
RESULTS AND DISCUSSION
in different aquifers were different. For instance, STZ had a high detection frequency in the dry season in the N zone
Overview of PPCPs in GW
and a high detection frequency in the wet season in the S zone. IBU was the most abundant PPCP in the GW in
The frequency of detection and the maximum concen-
both the dry and wet seasons for the whole study area, indi-
trations of the PPCPs in aquifers are summarized in
cating that GW pollution by IBU was not a sporadic but
Table 1. All the PPCPs were detected at least once in the
continuous event. Figure 2 shows the mean concentrations
916 J. Wu et al.
|
|
Maximum concentration (MC, ng/L) and frequency of detection (FD, %) of PPCPs in the dry and wet season in different aquifers of RGI
UA-N
UA-S
Dry
Wet
FCA-S
Dry
Wet
SCA-S
Dry
Wet
Dry
Wet
PPCPs
MC
FD
MC
FD
MC
FD
MC
FD
MC
FD
MC
FD
MC
FD
MC
FD
EFX
BLD
0.0
BLD
0.0
BLD
0.0
BLD
0.0
BLD
0.0
BLD
0.0
BLD
0.0
BLD
0.0
ETM
1.55
6.3
1.71
73.3
BLD
0.0
0.37
5.0
BLD
0.0
0.74
13.6
BLD
0.0
BLD
0.0
SMM
3.43
43.8
12.39
73.3
7.83
40.0
3.53
60.0
10.77
9.5
12.92
90.9
4.24
54.5
9.32
71.4
STZ
3.86
68.8
2.24
50.0
2.28
60.0
54.40
85.0
3.64
38.1
23.67
90.9
4.39
36.4
12.57
71.4
N4AcSMX
22.73
25.0
51.69
73.3
77.74
85.0
15.58
35.0
48.81
33.3
35.96
36.4
56.06
63.6
4.37
28.6
1.51
50.0
16.73
73.3
255.07
80.0
3.16
70.0
155.34
57.1
21.29
95.5
144.04
72.7
2.93
76.2
SCP
3.25
56.3
6.76
73.3
6.77
55.0
2.71
80.0
4.45
52.4
5.13
90.9
12.40
68.2
2.48
76.2
SMR
2.49
62.5
0.69
73.3
2.19
55.0
1.74
65.0
2.89
52.4
1.03
86.4
2.31
59.1
0.66
66.7
SMZ
0.49
56.3
0.83
73.3
56.47
95.0
19.50
95.0
12.57
76.2
9.49
86.4
8.58
95.5
33.44
81.0
SMX
19.29
100.0
11.01
100.0
54.19
75.0
8.12
65.0
15.36
95.2
7.45
81.8
20.56
77.3
15.08
71.4
TMP
1.03
81.3
3.23
100.0
0.54
95.0
0.23
45.0
0.86
90.5
0.64
50.0
4.89
100.0
2.01
52.4
CAF
782.62
56.3
461.87
26.7
249.29
5.0
BLD
0.0
398.51
19.0
BLD
0.0
235.52
4.5
BLD
0.0
CAP
0.55
43.8
0.16
66.7
9.66
60.0
1.07
95.0
9.23
47.6
1.01
95.5
8.37
22.7
0.48
85.7
IBU
24.50
100.0
42.99
100.0
30.92
90.0
34.35
100.0
64.21
95.2
23.19
95.5
190.14
77.3
7.91
90.5
TCS
18.05
25.0
109.90
26.7
5.19
5.0
BLD
0.0
1,186.38
19.0
BLD
0.0
10.81
4.5
BLD
0.0
DIF
BLD
0.0
22.05
73.3
BLD
0.0
BLD
0.0
BLD
0.0
5.59
13.6
BLD
0.0
BLD
0.0
Average FD
55.33
48.4
46.52
66.0
47.38
50.0
9.05
50.0
119.56
42.9
9.26
58.0
43.89
46.0
5.70
48.2
Hydrology Research
SFS
Pharmaceuticals and personal care products in groundwater in North China
Table 1
| 51.5
| 2020
J. Wu et al.
917
Figure 2
|
|
Pharmaceuticals and personal care products in groundwater in North China
Hydrology Research
|
51.5
|
2020
The mean concentration of the PPCPs in the unconfined aquifer (N and S zones) in different seasons. The concentration of EFX is not listed due to no available data. Bar graphs mean ± standard deviation (paired t-test, *P < 0.05; **P < 0.01).
for the PPCPs in the unconfined aquifer in different seasons.
Previous studies have shown that many factors influence
The mean concentrations of SMM, STZ, SCP, N4AcSMX,
the seasonal variations in PPCPs in aquatic environments,
TMP, ETM, IBU, and DIF in the N zone or the S zone
such as the consumption pattern, physicochemical proper-
were significantly higher in the wet season than those in
ties, and the flow conditions of rivers (Sun et al. ;
the dry season. For SFS, N4AcSMX, SMX, and CAP, the
Wang et al. b). Some PPCPs have seasonal uses, indicat-
mean concentrations were significantly lower in the wet
ing that their influent load will vary throughout the year. It is
season. Figure 3 shows the spatial distribution of the cumu-
anticipated that high consumption of some PPCPs in the wet
lative concentration of PPCPs in GW in the RGI area. From
season will lead to high occurrence and concentration in
the viewpoint of total concentration, PPCPs concentrations
rivers containing sewage. However, in the dry season,
in most sites were higher in the dry season than those in the
some factors will reduce the attenuation rate of PPCPs in
wet season. The higher occurrence of PPCPs in the dry
rivers, such as lower runoff, decreased river flow rate,
season (winter) was in accordance with previous studies
slower photolysis, thermal degradation, and biodegradation
(Sun et al. ).
in the cold season. As a result, the PPCPs in GW are directly
918
Figure 3
J. Wu et al.
|
|
Pharmaceuticals and personal care products in groundwater in North China
The distribution of PPCPs in GW in the RGI area.
Hydrology Research
|
51.5
|
2020
J. Wu et al.
919
|
Pharmaceuticals and personal care products in groundwater in North China
Hydrology Research
|
51.5
|
2020
and/or indirectly affected by the PPCPs in rivers. Therefore,
leaching potential group, EFX, SMM, and CAF were
the seasonal variation in PPCPs in aquifers in the RGI area
most likely to leach because of their high t1/2 and low
was affected by multiple factors simultaneously. Monitoring
KOC. TMP and IBU had moderate potential for leaching.
PPCPs at least twice a year (dry season/wet season) is essen-
TCS would be the compound most resistant to leaching. It
tial to obtain representative sample data to reflect the
is noted that due to relatively small chemical inputs, the tar-
dynamics of the seasonal change.
geted PPCPs are not expected to remain in the vadose zone for a long time. The different patterns of the 15 PPCPs in the
Spatial variations in PPCPs in different subzones In general, PPCP concentrations decreased from upstream (N zone) to downstream (S zone) (Figure 3). In the dry season, the mean concentration of CAF in the unconfined aquifer in the N zone was significantly higher than that in the S zone. In the wet season, the mean concentrations of SMM, TMP, IBU, CAF, TCS, N4AcSMX, ETM, DIF, and SMX were significantly higher in the N zone than those in the S zone. It is obvious that the dominant component in the N zone is CAF. In the S zone, SFS, IBU, and CAF were all abundant. CAF is known to be labile and has been suggested as a highly sensitive indicator of immediate contamination by sewage. The high CAF levels in the northern part were probably due to the high burden of sewage in the river water. The high concentrations of SFS and IBU in the southern part may be attributed to effluent from treated/untreated wastewater. As shown in Table 2, most of the targeted PPCPs appeared to have high leaching potential. Among the high
two subzones indicated that other factors would affect the occurrence in addition to the dilution of river flow due to precipitation. Dilution from sewage, degradation within the upstream sewer, rainfall, sampling mode, and pollution source characterization all contribute to this variability. Spatial variations in PPCPs in different aquifers Because information on confined aquifers in the northern part was not available, spatial variations in the PPCPs in different aquifers were analyzed in the southern part. As shown in Table 1, PPCPs in the dry season were more frequently detected in unconfined aquifers (50% for average frequency) than in the first confined aquifers (42% for average frequency in confined aquifers) and in the second confined aquifers (46% for average frequency). With regard to PPCPs in the wet season, the average frequency of detection in the second confined aquifers (48%) was lower than that in the unconfined aquifers (50%). For PPCPs in the first confined aquifers, the average frequency of detection (58%) was higher than that in the unconfined aquifers (50%). Figure 4 shows the mean concentrations of
Table 2
|
PPCPs in the confined aquifer in different seasons. Generally,
The logt1/2, logKoc, and GUS of PPCPs in soils
Name
logt1/2
logKOC
GUS
Leaching potential
EFX
2.56
1.174
7.22
High
ETM
2.56
2.754
3.19
High
SMM
1.88
1.678
4.35
High
STZ
1.88
2.207
3.36
High
SMR
1.88
2.076
3.61
High
SMZ
1.88
2.282
3.22
High
SMX
1.88
2.412
2.98
High
TMP
1.89
2.857
2.16
Middle
CAF
1.48
1
4.43
High
IBU
1.48
2.626
2.03
Middle
TCS
2,880
4.369
0.77
Low
DIF
2.56
2.743
3.21
High
PPCPs in confined aquifers have seldom been investigated, possibly because they were supposed to be protected from pollution by the vadose zone and the upper confining bed. However, it can be seen that the mean concentrations of some PPCPs in the two confined aquifers were relatively higher than those in the unconfined aquifer, suggesting insignificant attenuation effects by the upper confining bed. For many PPCPs, there may be multiple pathways into the GW, and difficulties in understanding these processes are compounded by the paucity of information compared to that for SW. The environmental behavior of PPCPs in aquifers depends not only on the specific hydrogeochemical conditions (Wang et al. b) but also on the physicochemical properties of the PPCPs in the GW system (Koh et al.
920
Figure 4
J. Wu et al.
|
|
Pharmaceuticals and personal care products in groundwater in North China
Hydrology Research
|
51.5
|
2020
The mean concentration of the PPCPs in aquifers (S zone) in different seasons. The concentration of EFX is not listed due to no available data. Mean contribution to each heavy metal concentration from each source.
). One explanation for the phenomenon of PPCPs in
defects in the well casing or via the backfilling sand
two confined GW samples is that those aquifers are con-
around the well casing, driven by the difference in the poten-
sidered to be recharged from the upper aquifers. Another
tial of aquifers. The GW was recharged by the river flows
reason involves contamination by unconfined GW via
deeper than 80 m in the area of the studied well.
J. Wu et al.
921
|
Pharmaceuticals and personal care products in groundwater in North China
Ecological risk of PPCPs
Hydrology Research
|
51.5
|
2020
are insufficient to reduce the ecological risk of all PPCPs to below the high or medium level in these areas. These pri-
The highest concentration and lowest PNEC were simul-
ority PPCPs should be given special attention to be strictly
taneously used as a worst-case scenario assuming ecological
regulated during GW recharging.
risk. Due to the lack of predictive toxicity data on the chronic effects of SMM N4AcSMX and DIF, their RQs were ignored. As shown in Table 3, eight PPCPs were detected in aquifers at
CONCLUSIONS
low-risk levels. The ecosystem risks of SFS, CAF, IBU, and TCS were found to be at least middle risk by using a RQ. Similar
With increasing artificial sewage recharge activities in the
to the concentration of PPCPs, the potential ecological risks
past several years, the detection frequencies and concen-
also exhibited spatial and seasonal variations to some extent.
tration levels of PPCPs in aquifers in North China were
Overall, the potential ecological risks posed by PPCPs to GW
generally higher than those reported in global GW. As a
could be a serious issue. The results of this study indicated
whole, IBU, SMX, SMZ, SMR, SFS, and SCP were widely
that PPCPs in GW in the RGI area also need to be monitored
detected in aquifers. The seasonal dynamics of some
for regulation and control by legislation due to their wide distri-
PPCPs showed statistically significant changes. Significant
bution and significant adverse ecological effects.
spatial variations in PPCPs in different subzones and in
From the perspective of ecological risk, SFS, CAF, IBU,
different aquifers were also found. However, the reasons
and TCS can be viewed as priority PPCPs because of middle
for the spatiotemporal variation were supposed to be com-
or high ecological risk. These results indicated that AR was a
plex and most likely caused by the combination of specific
source of PPCPs in GW, particularly in GW with short resi-
hydrogeochemical conditions and the physicochemical
dence times, and poses a threat to ecological systems. In
properties of the PPCPs in the GW system. Risk assessment
addition, another GW–SW exchange engineering method,
revealed that four antibiotics (DIF, IBU, CAF, and SFS)
river bank filtration, has been proved to be an important,
posed at least a medium risk in GW under the worst scen-
effective, and inexpensive technique for SW purification
arios. Finally, eight PPCPs (IBU, SMX, SMZ, SMR, SFS,
due to filtration, adsorption, and biodegradation (Munz
SCP, CAF, and TCS) were identified as priority control anti-
et al. ; Wang et al. c). However, these processes
biotics in the RGI area in North China.
Table 3
|
The ecological risk value of PPCPs in the dry and wet season in different aquifers of RGI
UA-N
UA-S
PPCPs
Dry
Wet
ETM
7.75 × 10 4
8.55 × 10 4
4
4
STZ
3.86 × 10
SFS
2.44 × 10 3
SCP
1.31 × 10
3
SMR SMZ
1.64 × 10
SMX TMP
3.55 × 10
CAF
78.26 × 10
CAP
9
7.4 × 10
IBU
2.45 × 10
TCS
FCA-S
Dry
Wet
Dry
1.85 × 10 4
NA 4
SCA-S
3
NA 4
2.24 × 10
2.28 × 10
5.44 × 10
3.64 × 10
2.7 × 10 4
4.1 × 10 1
5.1 × 10 3
2.5 × 10 1
3
3
3
3
2.72 × 10
2.73 × 10
1.09 × 10
1.8 × 10
3.66 × 10 3
1.02 × 10 3
3.22 × 10 3
2.56 × 10 3
6
6
4
5
Wet
Dry
Wet
3.7 × 10 4
NA
NA
2.37 × 10 3
4.39 × 10 4
1.25 × 10 3
3.4 × 10 2
2.3 × 10 1
4.73 × 10 3
3
3
1 × 10 3
2.07 × 10
5 × 10
4.25 × 10 3
1.52 × 10 3
3.4 × 10 3
5
5
3.16 × 10
2.86 × 10
1.11 × 10 4
2.77 × 10
1.88 × 10
6.5 × 10
2.05 × 10 3
1.17 × 10 3
5.77 × 10 3
8.63 × 10 4
1.63 × 10 3
7.93 × 10 4
2.19 × 10 3
1.61 × 10 3
3
2
3
4
3
3
2
9.02 × 10
2
4.19 × 10
9.7 × 10 4
5
1.12 × 10
1.85 × 10
8.09 × 10
2.96 × 10
46.18 × 10
24.93 × 10
NA
39.85 × 10
9
7
2.17 × 10
1.3 × 10
4.3 × 10
3.09 × 10
5.5 × 10
1
2
2.59 × 10
8
1.45 × 10
1.25 × 10
3.43 × 10
6.42 × 10
NA
5.93 × 10
7
2.19 × 10
1.69 × 10
6.92 × 10 3
NA
23.55 × 10
NA
8
1.37 × 10
1.13 × 10
6.44 × 10 9
2.32 × 10
19.01 × 10
7.9 × 10 1
NA
7
2
5.4 × 10
NA
922
J. Wu et al.
|
Pharmaceuticals and personal care products in groundwater in North China
Despite the uncertainties associated with sampling, geology, hydrology, and environment and parameters, the results obtained from this study are valuable. The environmental risk of PPCPs in GW recharged by river water has been questioned due to dilution, filtration, and degradation. However, the concentration and associated ecological risk indicate a significant threat by PPCPs to the RGI area, particularly to riverbank organisms. This work will help us gain insights into the contamination and monitoring of PPCPs in a GW–river interactive system and provide a basis to propose an appropriate mitigation strategy for PPCP management in similar watersheds around the world. Further studies should focus on the influencing factors of PPCPs during SW and GW exchange processes.
ACKNOWLEDGEMENTS This study was financially supported by the study on the watershed water management system of the Sino-German cooperation project and research on the key technology of water quality safety assurance in groundwater extraction reduction in Miyun district, Beijing. The authors would like to thank the editor and anonymous reviewers for their valuable comments that greatly improved the work. Author contributions as follows: J.W.: conceptualization, methodology, software, and writing the original draft; J.L.: data curation and visualization; Z.P.: investigation and writing – reviewing and editing; B.W.: investigation and resources; and D.Z.: supervision and funding acquisition.
SUPPLEMENTARY MATERIAL The Supplementary Material for this paper is available online at https://dx.doi.org/10.2166/nh.2020.001.
REFERENCES Aa, Y., Wang, G., Liu, T., Xue, B. & Kuczera, G. Spatial variation of correlations between vertical soil water and evapotranspiration and their controlling factors in a semiarid region. J. Hydrol. 574, 53–63. https://doi.org/10.1016/ j.jhydrol.2019.04.023.
Hydrology Research
|
51.5
|
2020
Barnes, K. K., Kolpin, D. W., Furlong, E. T., Zaugg, S. D., Meyer, M. T. & Barber, L. B. A national reconnaissance of pharmaceuticals and other organic wastewater contaminants in the United States – I) groundwater. Sci. Total Environ. 402 (2–3), 192–200. https://doi.org/10.1016/j.scitotenv.2008. 02.021. Bexfield, L. M., Toccalino, P. L., Belitz, K., Foreman, W. T. & Furlong, E. T. Hormones and pharmaceuticals in groundwater used as a source of drinking water across the United States. Environ. Sci. Technol. 53 (6), 2950–2960. https://doi.org/10.1021/acs.est.8b05592. Bo, Z., Mei, H., Yongsheng, Z., Xueyu, L., Xuelin, Z. & Jun, D. Distribution and risk assessment of fluoride in drinking water in the west plain region of Jilin province, China. Environ. Geochem. Health 25 (4), 421–431. https://xs.scihub. ltd/https://doi.org/10.1023/B:EGAH.0000004560.47697.91. Brandt, K. K., Amezquita, A., Backhaus, T., Boxall, A., Coors, A., Heberer, T. & Zhu, Y. G. Ecotoxicological assessment of antibiotics: a call for improved consideration of microorganisms. Environ. Int. 85, 189–205. https://doi.org/ 10.1016/j.envint.2015.09.013. Buerge, I. J., Buser, H. R., Kahle, M., Muller, M. D. & Poiger, T. Ubiquitous occurrence of the artificial sweetener acesulfame in the aquatic environment: an ideal chemical marker of domestic wastewater in groundwater. Environ. Sci. Technol. 43 (12), 4381–4385. https://doi.org/10.1021/es900126x. Chen, H., Jing, L., Teng, Y. & Wang, J. Characterization of antibiotics in a large-scale river system of China: occurrence pattern, spatiotemporal distribution and environmental risks. Sci. Total Environ. 618, 409–418. https://doi.org/10.1016/ j.scitotenv.2017.11.054. Cleuvers, M. Mixture toxicity of the anti-inflammatory drugs diclofenac, ibuprofen, naproxen, and acetylsalicylic acid. Ecotox. Environ. Safe 59 (3), 309–315. https://doi.org/10. 1016/S0147-6513(03)00141-6. Díaz-Cruz, M. S. & Barceló, D. Trace organic chemicals contamination in ground water recharge. Chemosphere 72 (3), 333–342. https://doi.org/10.1016/j.chemosphere.2008. 02.031. Dougherty, J. A., Swarzenski, P. W., Dinicola, R. S. & Reinhard, M. Occurrence of herbicides and pharmaceutical and personal care products in surface water and groundwater around Liberty Bay, Puget Sound, Washington. J. Environ. Qual. 39 (4), 1173–1180. https://dl.sciencesocieties.org/ publications/jeq/abstracts/39/4/1173. EC (European Commission) Technical Guidance Document in Support of Commission Directive 93//67/EEC on Risk Assessment for New Notified Substances and Commission Regulation (EC) No. 1488/94 on Risk Assessment for Existing Substances. Einsiedl, F., Radke, M. & Maloszewski, P. Occurrence and transport of pharmaceuticals in a karst groundwater system affected by domestic wastewater treatment plants. J. Contam. Hydrol. 117 (1–4), 26–36. https://doi.org/10.1016/j.jconhyd. 2010.05.008.
923
J. Wu et al.
|
Pharmaceuticals and personal care products in groundwater in North China
Fang, Q., Wang, G., Liu, T., Xue, B.-L. & Aa, Y. Controls of carbon flux in a semi-arid grassland ecosystem experiencing wetland loss: vegetation patterns and environmental variables. Agric. Forest. Meteorol. 259, 196–210. Grünheid, S., Amy, G. & Jekel, M. Removal of bulk dissolved organic carbon (DOC) and trace organic compounds by bank filtration and artificial recharge. Water Res. 39 (14), 3219–3228. https://doi.org/10.1016/j.watres.2005.05.030. Hanna, N., Sun, P., Sun, Q., Li, X., Yang, X., Ji, X., Zou, H., Ottoson, J., Nilsson, L. E., Berglund, B., Dyar, O. J., Tamhankar, A. J. & Stålsby Lundborg, C. Presence of antibiotic residues in various environmental compartments of Shandong province in eastern China: its potential for resistance development and ecological and human risk. Environ. Int. 114, 131–142. https://doi.org/10.1016/j.envint. 2018.02.003. Jiang, Y. H., Li, M. X., Guo, C. S., An, D., Xu, J., Zhang, Y. & Xi, B. D. Distribution and ecological risk of antibiotics in a typical effluent–receiving river (Wangyang River) in north China. Chemosphere 112, 267–274. https://doi.org/10.1016/ j.chemosphere.2014.04.075. Koh, Y. K., Chiu, T. Y., Boobis, A. R., Scrimshaw, M. D., Bagnall, J. P., Soares, A., Pollard, S., Cartmell, E. & Lester, J. N. Influence of operating parameters on the biodegradation of steroid estrogens and nonylphenolic compounds during biological wastewater treatment processes. Environ. Sci. Technol. 43, 6646–6654. https://doi.org/10.1021/es901612v. Kostich, M. S., Batt, A. L. & Lazorchak, J. M. Concentrations of prioritized pharmaceuticals in effluents from 50 large wastewater treatment plants in the US and implications for risk estimation. Environ. Pollut. 184, 354–359. https://doi. org/10.1016/j.envpol.2013.09.013. Kuroda, K., Murakami, M., Oguma, K., Muramatsu, Y., Takada, H. & Takizawa, S. Assessment of groundwater pollution in Tokyo using PPCPs as sewage markers. Environ. Sci. Technol. 46 (3), 1455–1464. https://doi.org/10.1021/es202059 g. Lapworth, D. J., Baran, N., Stuart, M. E. & Ward, R. S. Emerging organic contaminants in groundwater: a review of sources, fate and occurrence. Environ. Pollut. 163, 287–303. https://doi.org/10.1016/j.envpol.2011.12.034. Li, Z., Li, M., Liu, X., Ma, Y. & Wu, M. Identification of priority organic compounds in groundwater recharge of China. Sci. Total Environ. 493, 481–486. https://doi.org/10. 1016/j.scitotenv.2014.06.005. Li, J., Yang, Y., Huan, H., Li, M., Xi, B., Lv, N. & Yang, J. Method for screening prevention and control measures and technologies based on groundwater pollution intensity assessment. Sci. Total Environ. 551, 143–154. https://doi.org/ 10.1016/j.scitotenv.2015.12.152. Liu, J. L. & Wong, M. H. Pharmaceuticals and personal care products (PPCPs): a review on environmental contamination in China. Environ. int. 59, 208–224. https://doi.org/10.1016/ j.envint.2013.06.012. Munz, M., Oswald, S. E., Schäfferling, R. & Lensing, H. J. Temperature-dependent redox zonation, nitrate removal and
Hydrology Research
|
51.5
|
2020
attenuation of organic micropollutants during bank filtration. Water Res. https://doi.org/10.1016/j.watres.2019.06.041. Peng, X., Ou, W., Wang, C., Wang, Z., Huang, Q., Jin, J. & Tan, J. Occurrence and ecological potential of pharmaceuticals and personal care products in groundwater and reservoirs in the vicinity of municipal landfills in China. Sci. Total Environ. 490, 889–898. https://doi.org/10.1016/j.scitotenv. 2014.05.068. Petrie, B., Barden, R. & Kasprzyk-Hordern, B. A review on emerging contaminants in wastewaters and the environment: current knowledge, understudied areas and recommendations for future monitoring. Water Res. 72, 3–27. https://doi.org/10.1016/j.watres.2014.08.053. Singh, L. K., Jha, M. K. & Chowdary, V. M. Multi-criteria analysis and GIS modeling for identifying prospective water harvesting and artificial recharge sites for sustainable water supply. J. Clean Prod. 142, 1436–1456. https://doi.org/10. 1016/j.jclepro.2016.11.163. Sui, Q., Huang, J., Deng, S., Chen, W. & Yu, G. Seasonal variation in the occurrence and removal of pharmaceuticals and personal care products in different biological wastewater treatment processes. Environ. Sci. Technol. 45 (8), 3341–3348. https://doi.org/10.1021/es200248d. Sun, Q., Li, Y., Li, M., Ashfaq, M., Lv, M., Wang, H., Hu, A. & Yu, C. PPCPs in Jiulong River estuary (China): spatiotemporal distributions, fate, and their use as chemical markers of wastewater. Chemosphere 150, 596–604. https:// doi.org/10.1016/j.chemosphere.2016.02.036. Teijon, G., Candela, L., Tamoh, K., Molina-Díaz, A. & Fernández-Alba, A. R. Occurrence of emerging contaminants, priority substances (2008/105/CE) and heavy metals in treated wastewater and groundwater at Depurbaix facility (Barcelona, Spain). Sci. Total Environ. 408 (17), 3584–3595. https://doi.org/10.1016/j.scitotenv. 2010.04.041. Thomas, R., Gough, R. & Freeman, C. Linear alkylbenzene sulfonate (LAS) removal in constructed wetlands: the role of plants in the treatment of a typical pharmaceutical and personal care product. Ecol. Eng. 106, 415–422. https://doi. org/10.1016/j.ecoleng.2017.06.015. Wang, Z., Zhang, X. H., Huang, Y. & Wang, H. Comprehensive evaluation of pharmaceuticals and personal care products (PPCPs) in typical highly urbanized regions across China. Environ. Pollut. 204, 223–232. https://doi.org/ 10.1016/j.envpol.2015.04.021. Wang, G., Li, J., Sun, W., Xue, B., Aa, Y. & Liu, T. a Non-point source pollution risks in a drinking water protection zone based on remote sensing data embedded within a nutrient budget model. Water Res. 157, 238–246. https://doi.org/10. 1016/j.watres.2019.03.070. Wang, P., Rene, E. R., Yan, Y., Ma, W. & Xiang, Y. b Spatiotemporal evolvement and factors influencing natural and synthetic EDCs and the microbial community at different groundwater depths in the Chaobai watershed: a long-term field study on a river receiving reclaimed water.
924
J. Wu et al.
|
Pharmaceuticals and personal care products in groundwater in North China
J. Environ. Manage. 246, 647–657. https://doi.org/10.1016/j. jenvman.2019.05.156. Wang, P., Yao, J., Wang, G., Hao, F., Shrestha, S., Xue, B., Xie, G. & Peng, Y. c Exploring the application of artificial intelligence technology for dentification of water pollution characteristics and tracing the source of water quality pollutants. Sci. Total Environ. 693, 133440. Yang, Z., Cai, Y. & Mitsch, W. J. Ecological and hydrological responses to changing environmental conditions in China’s river basins. Ecol. Eng. 76, 1–6. https://doi.org/10.1016/ j.ecoleng.2014.12.007. Yang, L., Cheng, Q., Lin, L., Wang, X., Chen, B., Luan, T. & Tam, N. F. Partitions and vertical profiles of 9 endocrine disrupting chemicals in an estuarine environment: effect of tide, particle size and salinity. Environ. Pollut. 211, 58–66. https://doi.org/10.1016/j.envpol.2015.12.034. Yang, Y., Ok, Y. S., Kim, K. H., Kwon, E. E. & Tsang, Y. F. a Occurrences and removal of pharmaceuticals and personal care products (PPCPs) in drinking water and water/sewage treatment plants: a review. Sci. Total Environ. 596, 303–320. https://doi.org/10.1016/j.scitotenv.2017.04.102. Yang, L., He, J. T., Su, S. H., Cui, Y. F., Huang, D. L. & Wang, G. C. b Occurrence, distribution, and attenuation of pharmaceuticals and personal care products in the riverside groundwater of the Beiyun River of Beijing, China. Environ.
Hydrology Research
|
51.5
|
2020
Sci. Pollut. Res. 24 (18), 15838–15851. https://xs.scihub.ltd/ https://doi.org/10.1007/s11356-017-8999-0. Zhang, Q.-Q., Ying, G.-G., Pan, C.-G., Liu, Y.-S. & Zhao, J.-L. Comprehensive evaluation of antibiotics emission and fate in the river basins of China: source analysis, multimedia modeling, and linkage to bacterial resistance. Environ. Sci. Technol. 49, 6772–6782. https://doi.org/10.1021/acs.est.5b00729. Zhang, Y., Lv, T., Carvalho, P. N., Zhang, L., Arias, C. A., Chen, Z. & Brix, H. Ibuprofen and iohexol removal in saturated constructed wetland mesocosms. Ecol. Eng. 98, 394–402. https://doi.org/10.1016/j.ecoleng.2016.05.077. Zhang, P., Cai, Y., Yang, W., Yi, Y., Yang, Z. & Fu, Q. Multiple spatio-temporal patterns of vegetation coverage and its relationship with climatic factors in a large dam-reservoirriver system. Ecol. Eng. 138, 188–199. https://doi.org/10. 1016/j.ecoleng.2019.07.016. Zheng, G., Cao, J. R., Cheng, X. S., Ha, D. & Wang, F. J. Experimental study on the artificial recharge of semiconfined aquifers involved in deep excavation engineering. J. Hydrol. 557, 868–877. https://doi.org/10.1016/j.jhydrol.2018.01.020. Zou, S. C., Xu, W. H., Zhang, R. J., Tang, J. H., Chen, Y. J. & Zhang, G. Occurrence and distribution of antibiotics in coastal water of the Bohai Bay, China: impacts of river discharge and aquaculture activities. Environ. Pollut. 159, 2913–2920. https://doi.org/10.1016/j.envpol.2011.04.037.
First received 1 January 2020; accepted in revised form 15 March 2020. Available online 20 April 2020
925
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Impacts of bias nonstationarity of climate model outputs on hydrological simulations Yu Hui, Yuni Xu, Jie Chen, Chong-Yu Xu and Hua Chen
ABSTRACT Bias correction methods are based on the assumption of bias stationarity of climate model outputs. However, this assumption may not be valid, because of the natural climate variability. This study investigates the impacts of bias nonstationarity of climate models simulated precipitation and temperature on hydrological climate change impact studies. The bias nonstationarity is determined as the range of difference in bias over multiple historical periods with no anthropogenic climate change for four different time windows. The role of bias nonstationarity in future climate change is assessed using the signal-to-noise ratio as a criterion. The results show that biases of climate models simulated monthly and annual precipitation and temperature vary with time, especially for short time windows. The bias nonstationarity of precipitation plays a great role in future precipitation change, while the role of temperature bias is not important. The bias nonstationarity of climate model outputs is amplified when driving a hydrological model for hydrological simulations. The increase in the length of time window can mitigate the impacts of bias nonstationarity for streamflow projections.
Yu Hui Jie Chen (corresponding author) Hua Chen State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University, Wuhan, China E-mail: jiechen@whu.edu.cn Yu Hui Changjiang Institute of Survey, Planning, Design and Research, Wuhan, China Yuni Xu Bureau of Hydrology, ChangJiang Water Resources Commission, Wuhan, China
Thus, a long time period is suggested to be used to calibrate a bias correction method for hydrological climate change impact studies to reduce the influence of natural climate variability. Key words
| bias nonstationarity, climate change signal, climate model outputs, hydrology, natural climate variability
Chong-Yu Xu Department of Geosciences, University of Oslo, Oslo, Norway
HIGHLIGHTS
• • •
The biases of GCM precipitation and temperature vary with time, due to natural climate variability. The bias nonstationarity of precipitation plays a great role in future precipitation change, while the role of temperature bias is not important. The bias nonstationarity of precipitation and temperature has great considerable impacts on future streamflow changes.
INTRODUCTION The assessment of climate change impacts on the hydrologi-
Marhaento et al. ; Shen et al. ; Lu & Qin ;
cal cycle has been widely investigated during recent years
Ragettli et al. ). Global climate models (GCMs) can pro-
(Graham et al. ; Li et al. ; Chen et al. a;
vide climate variables (e.g. precipitation and temperature) for the future period used as inputs of hydrological models
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.254
for hydrological impact studies. However, the coarse resolution of GCMs does not meet the need of high resolution for the hydrological models (Maraun et al. ). In parallel,
926
Y. Hui et al.
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
GCMs are imperfect representations of reality with systema-
of climate models simulated precipitation are nonstationary
tic biases found between climate model simulations and
even for two close historical periods, while temperature
observations. To resolve these problems, a number of down-
biases are relatively stationary. This study attributed the
scaling methods have been developed during the last two
bias nonstationarity of precipitation to natural climate varia-
decades. Especially, bias correction becomes a standard pro-
bility (multi-decadal climate variability). A similar study was
cedure when using regional climate model (RCM) outputs
also carried out by Wang et al. (), who tested bias non-
for hydrological impact studies (van Pelt et al. ; Teutsch-
stationarity of precipitation in the eastern United States
bein & Seibert ; Olsson et al. ; Zhuan et al. ).
determined by a skill score, which compares the errors of
During recent years, bias correction methods are also com-
a downscaling method over the validation period with the
monly used for GCM outputs. Since the resolution of
errors of observations between calibration and validation
climate model outputs is lower than that of observations,
periods. The results show that precipitation biases are non-
bias correction methods also act as downscaling methods.
stationary at most of the stations, especially for the annual
The usually used bias correction methods range from
extreme precipitation. Taking into account multi-decadal cli-
simple mean-based scaling to sophisticated distribution-
mate variability, Nahar et al. () investigated the bias
based mapping and multivariate or/and multisite correction.
stationarity of six GCMs simulated monthly and seasonal
The commonly used bias correction methods are usually
precipitation and temperature over multi-decadal time
based on an assumption that biases of climate models out-
periods from 1900 to 1999 in Australia. One hundred
puts are stationary. In other words, these bias correction
samples of the observations and GCM simulations gener-
methods assume climate model outputs present the same
ated using bootstrapping was used to calculate uncertainty
biases in magnitude and direction between historical and
in biases with a 95% confidence, which represents a possible
future periods. However, this assumption may not be
range of bias if bias stationarity exists. When the actual bias
valid, as pointed out in a few recent studies (e.g. Buser
goes beyond this uncertainty, the bias was assumed to be
et al. ; Ehret et al. ; Gutierrez et al. ; Maurer
nonstationary. This study showed that biases of precipitation
et al. ; Velázquez et al. ; Dixon et al. ). For
and temperature are not stationary for some regions in Aus-
example, Christensen et al. () evaluated bias nonstatio-
tralia (e.g. east coast of Australia), because of the natural
narity on simulated monthly precipitation and temperature
climate variability.
from an ensemble of 13 RCMs over Europe. They found
For a historical period with no anthropogenic forcing,
that biases of simulated precipitation and temperature vary
the bias nonstationarity is attributed to natural climate varia-
as a second-order function of observations, suggesting that
bility. Natural climate variability refers to variations in the
bias nonstationarity exists in different climatic regimes.
mean state and other statistics of the climate, due to natural
Maraun () verified the bias stationary assumption in
internal processes or natural external forcing. It involves a
RCMs for European seasonal mean temperature and pre-
wide range of time scales, from one day to the next, as
cipitation sums using a pseudo-reality approach, which
well as from one year or multi-decade to the next. For the
considers one climate model as pseudo observation to com-
application of a bias correction method in climate change
pare with other climate model simulations. The results
impact studies, the decadal and multiple decadal time
showed that biases are relatively stable in general, but bias
scales are widely and implicitly used. At the decadal or
nonstationarity was identified in some regions where
multi-decadal time scales, the natural climate variability
changes in potentially relevant physical variables are signifi-
includes some natural modes of decadal or multi-decadal cli-
cant. This study was conducted in the climate model world,
mate variability, such as El Niño/Southern Oscillation
the transferability to real world needs to be further investi-
(ENSO), Atlantic Multidecadal Oscillation (AMO), Pacific
gated. To test the bias nonstationarity in real-world
Decadal Oscillation (PDO), and Interdecadal Pacific Oscil-
climate, Chen et al. () compared the biases between cli-
lation (IPO). The phase of these modes has been found to
mate model simulations and corresponding observations
be linked with changes in precipitation and temperature.
over two historical periods. The results showed that biases
For example, the PDO warm (cold) phase periods link to
927
Y. Hui et al.
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
the decrease (increase) in precipitation in the majority of
propagation of bias nonstationarity in hydrological impact
China (Ouyang et al. ). Precipitation decreases during
studies. The first objective is achieved by separating a long
IPO warm phase periods, while it increases during IPO
historical climate series (100 years) to decadal and multi-
cool phase periods in east Australia (Power et al. ).
decadal time periods to calculate the difference in bias
However, since the unpredictability and complexity of natu-
among different periods. Precipitation and temperature,
ral dynamic processes make it difficult for climate models to
which are primary variables for hydrological simulations,
accurately capture the characteristics, GCMs from Coupled
are investigated. The second objective is achieved by com-
Model Intercomparison Project phase 5 (CMIP5) have been
paring the range of biases among multiple periods to
found to poorly represent the phase in natural modes of
climate change signals between future and historical
decadal or multi-decadal climate variability (Polade et al.
periods. The last objective is achieved by analyzing stream-
; Ruiz-Barradas et al. ; Bellenger et al. ;
flow time series simulated by a hydrological model using
Fuentes-Franco et al. ). This poor representation for
the above two approaches.
GCMs could lead to biases between simulations and obser-
The rest of the paper is organized as follows: the follow-
vations when sampling in a finite decadal or multi-decadal
ing section introduces the study area and hydro-climate data
time, and further lead to changes in bias among different
followed by the methodology. The main results are then pre-
decadal or multi-decadal periods (Nahar et al. ).
sented, and the discussion and conclusions are presented in
In climate change impact studies, natural climate varia-
the final section.
bility used to be estimated based on multi-decadal climate model simulations in the absence of anthropogenic-induced climate change (Hulme et al. ; Arnell ; Chen et al.
STUDY AREA AND DATA
b). In these methods, the differences among several separated multi-decadal periods obtained from a long time
Study area
series of unforced simulations were supposed to represent the multi-decadal natural climate variability. Similarly, to
The case study was conducted at the Hanjiang River basin
evaluate bias nonstationarity of climate model outputs, the
(Figure 1), located in south-central China, which is the lar-
estimate of bias nonstationarity is depicted as the range of
gest tributary of the Yangtze River basin. The Hanjiang
the biases among multi-decadal historical periods, while
River flows through Shaanxi and Hubei Provinces with a
the timescale of each period may have an effect on the
drainage area of 159,000 km2 and a length of 1,567 km.
results of estimation. The different temporal scales of
The watershed above the Danjiangkou Reservoir, with a
sampling can lead to changes in the magnitude and direc-
sub-basin area of 95,200 km2 and a length of 918 km, was
tion of bias, due to different phases between GCM outputs
used in this study. The Danjiangkou reservoir is the water
and observations. Therefore, the impacts of different
source of the Middle Route of the South-to-North Water
temporal scales of natural climate variability on bias
Transfer Project in China. The watershed has a subtropical
nonstationarity of climate model outputs need to be investi-
monsoon climate. The mean annual precipitation is about
gated, especially for their propagation in hydrology.
840 mm, of which 70–80% of the total amount falls in the
Even though few studies (Maraun ; Chen et al. ;
wet season from May to September. The average maximum
Velázquez et al. ; Nahar et al. ) have investigated the
and minimum temperatures are 24–29 and 0–3 C, respect-
bias nonstationarity of climate model outputs, the role of
ively. The daily mean discharge of the Hanjiang River is
bias nonstationarity in future climate change was not
approximately 1,150 m3/s.
investigated, especially in hydrological impact studies. Accordingly, the objectives of this study are to investigate:
Data
(1) the bias nonstationarity of climate model outputs in the context of natural climate variability, (2) the role of
This study used both observed and GCM-simulated precipi-
bias nonstationarity in future climate change, and (3) the
tation and temperature for the Hanjiang watershed. The
928
Figure 1
Y. Hui et al.
|
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
Location map of the Hanjiang River basin. The upstream part above the Danjiangkou reservoir was used in this study.
inflow to the Danjiangkou reservoir calculated using a water
21st century. The other greenhouse gas emission scenarios
mass balance method was also used for hydrological model
(RCP8.5) are also presented and discussed in the Discussion
calibration and validation. The streamflow time series cover-
section. Since this study only used monthly and annual data,
ing the 1961–2000 period was provided by the Bureau of
the daily precipitation and temperature were summed (for
Hydrology of the Changjiang Water Resources Commission.
precipitation) or averaged (for temperature) to monthly and
The observed precipitation and temperature were
annual values. In order to run a lumped hydrological
obtained from the gridded Climatic Research Unit (CRU)
model, all gridded precipitation and temperature within and
Time-series (TS) data, produced by CRU at the University
surrounding the Hanjiang watershed were averaged to a
of East Anglia, UK. This dataset provides monthly gridded
single time series using the Thiessen polygon method for
data with a high resolution of 0.5 × 0.5 . The latest version
both observations and GCM simulations.
CRU TS version 4.01 (Harris et al. ) covering the period 1901–2000 was used in this study. The GCM-simulated daily precipitation and maximum
METHODOLOGY
and minimum temperatures were extracted from the database of CMIP5. In order to capture the uncertainty related to cli-
A climate time series for the past–present period may contain
mate models, 17 GCMs were used in this study (Table 1).
both natural climate variability and anthropogenic climate
The time period covers from 1901 to 2100. All GCMs’
change signal. In order to investigate the impacts of natural
simulations during the 1901–2005 period were driven by
climate variability on bias correcting climate model outputs,
historical climate forcing (including natural solar, volcanic
the climate change signal needs to be first removed. A
variations and anthropogenic radiative forcing), while those
simple linear detrending method of Zhuan et al. () was
during the 2006–2100 period were generated under the
used to remove the anthropogenic climate change signal for
representative concentration pathway (RCP) 4.5. RCP4.5
observed and GCM-simulated monthly precipitation and
corresponds to the medium anthropogenic forcing for the
temperature for the 1901–2000 period. This detrending
Y. Hui et al.
929
Table 1
|
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
Basic information of the used 17 CMIP5 models
Resolution (Longitude × ID
Model name
Modeling center
Institution
Latitude)
1
ACCESS1.0
CSIRO-BOM
1.875 × 1.25
2
ACCESS1.3
Commonwealth Scientific and Industrial Research Organization (CSIRO) and Bureau of Meteorology (BOM), Australia
3
BCC-CSM1.1(m)
BCC
Beijing Climate Center, China Meteorological Administration
1.125 × 1.25
4
CNRM-CM5
CNRM-CERFACS
Centre National de Recherches Météorologiques/Centre Européen de Recherche et Formation Avancée en Calcul Scientifique
1.4 × 1.4
5
CSIRO-Mk3.6.0
CSIRO-QCCCE
Commonwealth Scientific and Industrial Research Organization in collaboration with Queensland Climate Change Centre of Excellence
1.9 × 1.9
6
CanESM2
CCCMA
Canadian Centre for Climate Modelling and Analysis
2.8 × 2.8
7
GFDL-CM3
NOAA GFDL
NOAA Geophysical Fluid Dynamics Laboratory
2.5 × 2.0
8
GFDL-ESM2G
2.5 × 2.0
9
GFDL-ESM2M
2.5 × 2.0
10
INM-CM4
INM
Institute for Numerical Mathematics
2.0 × 1.5
11
IPSL-CM5A-LR
IPSL
L’Institut Pierre-Simon Laplace
3.75 × 1.9
12
IPSL-CM5A-MR
2.5 × 1.25
13
IPSL-CM5B-LR
3.75 × 1.9
14
MIROC-ESM-CHEM
15
MIROC-ESM
16
MIROC5
17
MRI-CGCM3
1.875 × 1.25
Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research Institute (The University of Tokyo), and National Institute for Environmental Studies
2.8 × 2.8
MIROC
Atmosphere and Ocean Research Institute (The University of Tokyo), National Institute for Environmental Studies, and Japan Agency for Marine-Earth Science and Technology
1.4 × 1.4
MRI
Meteorological Research Institute
1.1 × 1.1
MIROC
2.8 × 2.8
method first detects the trend and breakpoint using the
variability at the temporal scale, the 100-year precipitation
Mann–Kendall test (Mann ; Kendall ). If a trend
and temperature time series are divided into several inde-
exists, it is removed for the period after the breakpoint
pendent decadal and multi-decadal periods. The time
using a linear method. To preserve the seasonality, the
window consists of 10, 20, 33 and 50 years for the 100-
detrending method was applied for each month, respectively.
year period. For example, when the time window is 20 years, the 100-year period is divided into five 20-year periods
Calculation of difference in bias
(1901–1920, 1921–1940, 1941–1960, 1961–1980 and 1981– 2000). The first period of each time window (e.g. 1901–
After detrending, the variation of observed and GCM-simu-
1910 for 10-year window) is used as the baseline period.
lated precipitation and temperature within the 1901–2000
For each decadal or multi-decadal period, the biases of
period is considered to be only attributed to the natural
GCM-simulated monthly and annual precipitation (BP)
climate variability. Since natural climate variability is
and temperature (BT) relative to observations are calculated
inherently complex and manifests itself over various tem-
using Equations (1) and (2), respectively:
poral and spatial scales, this study only investigates its impacts on bias correcting climate model outputs over dec-
BPi ¼ (Pmod,i Pobs,i )=Pobs,i
(1)
BTi ¼ Tmod,i Tobs,i
(2)
adal and multi-decadal temporal and watershed spatial scales. In order to investigate the impacts of natural climate
Y. Hui et al.
930
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
where i indicates the number of decadal or multi-decadal
where DBmax and DBmin indicate the maximum and mini-
historical periods, Pmod and Pobs indicate the mean values
mum values of differences in bias between historical
of simulated and observed precipitation over the period,
periods and the baseline period, respectively. The outliers
and Tmod and Tobs indicate the mean values of simulated
were removed when calculating the RB to ensure the rea-
and observed temperature.
sonability of estimation. The outliers are detected if they
The differences in bias of precipitation (DBP) and tempera-
are larger than Q75 þ 1.5 × (Q75 Q25) or smaller than
ture (DBT) between the baseline period and all other historical
Q25 1.5 × (Q75 Q25), where Q75 and Q25 indicate the
periods are then calculated for each GCM. The use of the differ-
25th and 75th percentiles, respectively. It is worth noting
ence in bias is beneficial to compare with the results from
that when the time window length is 50 years, the RB is
different GCMs. The differences in bias represent the possible
just equal to the absolute value of difference in bias between
changes of GCM bias due to natural climate variability at dec-
the 1951–2000 period and the 1901–1950 baseline period.
adal and multi-decadal scales. The range of difference in biases
Meanwhile, the CCS of each future period is calculated
is used to judge the bias stationarity of climate model outputs.
relative to the historical baseline period. The 100-year
The impacts of natural climate variability on bias nonstationar-
period from 2001 to 2100 is also divided into several deca-
ity is investigated by using multiple time windows. The main
dal or multi-decadal periods as the same time window as
steps are illustrated in Figure 2 for a 20-year time window.
the historical period. By doing this, the SNR can be calculated for each time scale. The climate change signal for precipitation (CCSP) and temperature (CCST) was calcu-
Calculation of climate change signals relative to
lated using Equations (4) and (5):
differences in bias To identify the role of bias nonstationarity in future climate change, the signal-to-noise ratio (SNR) is used as a criterion
CCSPj ¼ (Pmod,j Pmod,b )=Pmod,b
(4)
CCSTj ¼ Tmod,j Tmod,b
(5)
to quantify the extent of bias nonstationarity relative to future climate change. The SNR is defined as the ratio of cli-
where j indicates the number of decadal or multi-decadal
mate change signal (CCS) of future periods and the range of
periods from 2001 to 2100, and subscript b indicates the
bias (RB) among multiple historical periods.
baseline period (e.g. 1901–1910, 1901–1920, 1901–1933 or
RB is calculated based on the differences in bias among multiple periods using Equation (3): RB ¼
1901–1950). These precipitation and temperature time series without detrending were used to retain climate change tendency.
max (jDBmax j, jDBmin j) if DBmax DBmin > 0 if DBmax DBmin 0 jDBmax DBmin j
(3)
Then, SNR is calculated as the absolute ratio of CCS and RB for both monthly and annual precipitation and temperature. If the CCS is relatively smaller than the RB, e.g. the SNR being smaller than 1, the RB is considered to have large impacts on future climate changes and their hydrological impacts and vice versa.
Hydrological modeling The hydrological simulations were carried out using a Figure 2
|
Illustration of the calculation of the differences in bias between four multiple
lumped hydrological model named as the Two-Parameter
time periods and the baseline period. Precipitation at the 20-year time window
Monthly Water Balance Model (Xiong & Guo ). It is
is used as an example. BP indicates bias between GCM-simulated and observed precipitation, and DBP indicates the difference in bias between
a simple and monthly lumped rainfall-runoff model with
future 20-year period and the 1901–1920 baseline period.
only two physical parameters. The first parameter is c,
Y. Hui et al.
931
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
which represents a coefficient to take account of the effect of
periods for the Hanjiang watershed, indicating a good
the time scale change. The second parameter is SC, which
performance.
represents the field capacity of a watershed. The monthly
Using the calibrated hydrological model, the streamflow time series are generated for each time window using the
runoff (Q) was simulated using Equations (6) and (7):
detrended observed and GCM-simulated precipitation and Et ¼ c × EPt × tanh (Pt =EPt )
(6)
Qt ¼ (St 1 þ Pt Et ) × tanh [(St 1 þ Pt Et )=SC]
(7)
temperature for the historical period. Meanwhile, the streamflow time series are also generated for each time window of the future periods. The propagation of precipitation and temperature bias nonstationarity in hydrology was investigated using the same methods as presented above under
where t indicates the number of months, E and EP indicate
‘Calculation of difference in bias’ and ‘Calculation of climate
the monthly actual and potential evapotranspiration,
change signals relative to differences in bias’.
respectively, P indicates the monthly areal precipitation, and S indicates the water content in the soil. The monthly potential
evapotranspiration
was
calculated
by
the
RESULTS
Thornthwaite method (Thornthwaite ; Xu & Singh ), using the monthly mean temperature. This model
Bias nonstationarity of precipitation and temperature
has been tested and proved to be highly efficient in several watersheds with the monsoon-rainfall-dominated climate
Figure 3 presents the difference in bias of mean annual pre-
in the south of China (Xiong & Guo ; Guo et al. ;
cipitation between each historical period and the baseline
Chen et al. ; Bai et al. ).
period for four different time window lengths: 10, 20, 33
The observed monthly streamflow was used to model calibration
(1981–2000).
resent differences in bias of all historical periods relative to
Model calibration was done automatically using shuffled
the baseline period for all 17 GCMs, and the thick line is
complex evolution-University of Arizona (SCE-UA) algor-
their median value. The x axis represents the midpoints of
ithm (Duan et al. ). The optimal combination of
the time window. The results show that the difference in
parameters was chosen on the basis of the Nash–Sutcliffe
bias of mean annual precipitation varies with time over his-
efficiency (NSE) coefficient (Nash & Sutcliffe ). The
torical periods, indicating that the bias of precipitation is
values of NSE are 0.79 for both calibration and validation
not stationary. The difference in bias reflects the impacts of
Figure 3
|
(1961–1980)
and
validation
and 50 years. For each time window length, the dash lines rep-
Difference in bias of mean annual precipitation between each historical period and the baseline period for 10-, 20-, 33- and 50-year windows. The dashed lines are for 17 GCMs and the thick line is their median value.
932
Y. Hui et al.
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
natural climate variability on bias nonstationarity, because
particular, the magnitude and uncertainty of differences in
the trend of precipitation change was removed. The results
bias are larger for short than long time windows. For
in Figure 3 show that the natural climate variability has
example, the median value of difference in temperature
great impacts on bias nonstationarity for mean annual pre-
bias changes between –0.7 and –0.05 C for the 10-year
cipitation, especially for the short time window. However,
window, between –0.5 and –0.05 C for the 20-year
this problem becomes less important as the time window
window, between –0.28 and –0.06 C for the 33-year
gets longer, as indicated by the fact that the difference in
window, and it is –0.13 C for the 50-year window. The nega-
bias decreases with the extension of the time window. For
tive values of the difference in bias indicate the bias over the
example, in terms of the median value over all GCMs, the
first baseline period is larger than that over the following
difference in bias changes between –11.5 and 3.8% for the
several historical periods. The large uncertainty related to
10-year window, between 0.5 and 5.5% for the 20-year
climate models is observed, especially for the short time
window, between –2.7 and –0.4% for the 33-year window,
window. Since the temperature time series were detrended
and it is –4.4% for the 50-year window. This indicates that
before calculating the difference in bias, the variation
the decadal variability has more impacts on bias nonstatio-
resulted from natural climate variability. Even though the
narity than multi-decadal variability. Not only the median
bias varies with time due to natural climate variability, its
of difference in bias but also the uncertainty (the difference
impacts on future climate change need to be investigated
between maximum and minimum values over the whole
by comparing future climate change signals.
period) related to GCMs decreases with the longer time window. The uncertainty of difference in bias is 56.4% for
Impacts of bias nonstationarity on future climate
the 10-year window, 27% for the 20-year window, 17.5% for
changes
the 33-year window, and 15.8% for the 50-year window. Therefore, using a long reference period to filter out low fre-
In order to investigate the role of bias nonstationarity in
quency modes of variability may lessen the impacts of
future climate change, the SNR of mean annual precipi-
natural climate variability on bias correction methods. How-
tation and temperature is calculated for all four time
ever, if just looking at a single climate model, the difference in
windows. Figure 5 presents the SNR of mean annual precipi-
bias can reach 10%, even for the 50-year time window.
tation and temperature over multiple time periods from
Figure 4 presents the difference in bias of mean annual
2001 to 2100 for 10-, 20-, 33- and 50-year windows. Each
temperature between each historical period and the baseline
boxplot depicts the distribution of SNR from an ensemble
period for four different time windows. The results show
of 17 GCMs. The larger SNR values suggest that the bias
that the differences in temperature bias vary with time. In
range is relatively less important compared with the climate
Figure 4
|
Difference in bias of mean annual temperature between each historical period and the baseline period for 10-, 20-, 33- and 50-year windows. The dashed lines are for 17 GCMs and the thick line is their median value.
933
Figure 5
Y. Hui et al.
|
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
Signal-to-noise ratio (SNR) of (a) mean annual precipitation and (b) temperature from 2001 to 2100 for 10-, 20-, 33- and 50-year windows. The black dashed line indicates the value of SNR is equal to 1.
change signal and vice versa. The unity of SNR means that
length of time window. However, the median values of
the range of bias is equal to the climate change signal. In
SNR are still less than one, even for the 33- and 50-year
this case, the calculated climate change signal may just be
time window. The above results indicate that great attention
the difference in bias between two periods. Figure 5(a)
should be paid to natural climate variability when investi-
shows that SNR values of precipitation are consistently
gating the change of precipitation, in the face of the
less than one for the 10-year window of all future periods.
increase in the greenhouse gas emission scenario.
In this case, it is hard to judge the existence of climate
However, different results are observed for annual mean
change signal. The calculated climate change signal may
temperature. Even though the difference in bias varies with
be just a part of the range of bias caused by the natural cli-
time, the anthropogenic climate change is significantly
mate variability, especially taking into account the fact
greater than the bias nonstationarity for annual mean temp-
that the SNR values are mostly constant over time. With
erature, especially for the future periods (Figure 5(b)). In
the increase in the length of time window, the bias nonsta-
other words, even though the bias nonstationarity can
tionarity of mean annual precipitation becomes less
affect the detection of the real anthropogenic climate
important, as indicated by some GCMs presenting SNR
change of annual mean temperature, the influence is limited.
values being greater than one. In particular, the uncertainty
Comparing to future climate change signal, the temperature
related to GCMs becomes larger with the increase in the
bias can be considered as stationary. However, for the first
934
Y. Hui et al.
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
and second decades, the natural climate variability still plays
January to December. The median value of the SNR was cal-
an important role in anthropogenic climate change or the
culated over 17 GCMs for each period. Generally, the bias of
climate change signal may not exist, as indicated by the rela-
monthly precipitation is nonstationary, while the tempera-
tively small SNR. The SNR values increase with time, which
ture is stationary compared to future climate change. In
is caused by the increase in the anthropogenic climate
other words, the impacts of bias nonstationarity on future
change, as the range of bias is supposed to be constant.
climate changes are large for monthly precipitation, but
Meanwhile, the dispersion of boxplot is gradually amplified
it is limited for monthly temperature. Similar to annual
with time, resulting from the increase in the GCM uncer-
precipitation
tainty in the far future. Additionally, the magnitude and
nonstationarity on future climate changes lessen for the
dispersion of the SNR become larger with the extension of
longer time window, as indicated by the increase in the
time window. This implies that the impacts of bias nonstatio-
SNR. However, the monthly variability is observed for
narity on future climate changes are important for the long
these impacts. In terms of the monthly precipitation
time window.
(Figure 6(a)), the SNR in the wet season (May–September)
and
temperature,
the
impacts
of
bias
The SNR of precipitation and temperature is also calcu-
is larger than in that the dry season (October–April),
lated at monthly scale. Figure 6 shows the portrait diagram
especially for the mid and far future, while the SNR is still
of the median value of the SNR for mean monthly precipi-
smaller than one for 10-, 20- and 33-year windows. When
tation and temperature over multiple time periods from
the time window increases to 50 years, the SNR in the wet
2001 to 2100 with 10-, 20-, 33- and 50-year windows,
season becomes larger than 1. This indicates that the
respectively. The y-axis represents the 12 months from
impacts of bias nonstationarity on future climate changes
Figure 6
|
Medians of signal-to-noise ratio (SNR) of (a) mean monthly precipitation and (b) temperature from 2001 to 2100 for 10-, 20-, 33- and 50-year windows.
935
Y. Hui et al.
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
are less important in the wet season than in the dry season.
hydrological simulations. The results show that the impacts
Therefore, the impacts of bias nonstationarity in the dry
of natural climate variability are amplified from climate
season need to be paid more attention than in the wet
world to hydrological world. For example, the difference
season and annual scale. In terms of the mean monthly
in bias for all GCMs varies between –30 and 20% for
temperature (Figure 6(b)), the SNR from May to November
mean annual precipitation, while that varies between –60
is larger than other months for the mid and far future, and
and 60% for mean annual streamflow for the 10-year time
the maximum values of the SNR present between May and
window. This is because of the non-linear process from
August over each decadal or multi-decadal period. In
climate to hydrology. Similar to precipitation and tempera-
addition, the comparison of the SNR between annual and
ture, the extent of the impacts of natural climate variability
monthly temperature shows that the SNR at the monthly
gradually mitigates as the time window gets longer. For
scale is smaller than that at the annual scale, especially for
example, the median value of difference in streamflow
short time windows, while for the 50-year window, the
bias changes between –12.3 and 25.6% for the 10-year
SNR between June and October is comparable to or even
window, between 8.5 and 20.2% for the 20-year window,
larger than that at the annual scale. In the near future
between –0.9 and 5.9% for the 33-year window, and it is
(2001–2030), the SNR in each month is comparable to or
–6.2% for the 50-year window. The uncertainty of difference
smaller than one for the short time window, because of
in streamflow bias related to GCMs also reduces as the
the weak climate change signal.
extension of the time window. The uncertainty is 121.6% for the 10-year window, 58.2% for the 20-year window,
Propagation of bias nonstationarity in hydrology
35% for the 33-year window and 27.1% for the 50-year window.
The propagation of bias nonstationarity of climate models
The impacts of bias nonstationarity of climate model
simulated precipitation and temperature in hydrology is
outputs on future streamflow changes are also evaluated
investigated by running a hydrological model over the Han-
using the SNR as a metric. Figure 8 shows the SNR of
jiang watershed. Figure 7 presents the difference in bias of
mean annual streamflow over multiple periods from 2001
mean annual streamflow between each historical period
to 2100 for 10-, 20-, 33- and 50-year windows. In the
and the baseline period for 10-, 20-, 33- and 50-year time
future, the magnitude of the streamflow change between
windows. The difference in bias between GCM-driven and
the future and baseline period is much smaller than the
observation-driven streamflow also varies with time, which
range of bias, as indicated by the small SNR values,
reflects the impacts of natural climate variability on
especially for the 10- and 20-year time windows. This
Figure 7
|
Difference in bias of mean annual streamflow between each historical period and the baseline period for 10-, 20-, 33- and 50-year windows. The dashed lines are for 17 GCMs and the thick line is their median value.
936
Figure 8
Y. Hui et al.
|
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
Signal-to-noise ratio (SNR) of mean annual streamflow for 10-, 20-, 33- and 50-year windows. The black dashed line indicates the value of SNR is equal to 1.
indicates that the natural climate variability plays an impor-
50-year time window, the median values of SNR are smaller
tant role in future streamflow changes, especially for the
than one, indicating that the natural climate variability is
short time window. In other words, previous studies that
still more important than the anthropogenic climate
attribute the streamflow change to anthropogenic climate
change for future streamflow simulations. These results of
change may not be correct when taking into account the
streamflow are similar to those of mean annual precipi-
impacts of climate change. Similar to precipitation and
tation, but the SNR of the former is smaller than the latter,
temperature, the importance of natural variability reduces
due to the larger range of bias and lower streamflow change.
with the extension of the time window, as indicated by the
Figure 9 presents the median of SNR of mean monthly
increase in the SNR. For example, the median values of
streamflow across 17 GCMs for 10-, 20-, 33- and 50-year
the SNR change between 0.07 and 0.21 for the 10-year
windows. The results show that the bias nonstationarity
window, between 0.09 and 0.3 for the 20-year window,
has large impacts on the change of future monthly stream-
between 0.38 and 0.85 for the 33-year window, and between
flow, even though the impacts mitigate with the extension
0.61 and 1.01 for the 50-year window. The uncertainty
of the time window. For example, the values of SNR are
related to GCMs becomes greater for the longer time
consistently smaller than one for 12 months and all
window. With the exception of the second period of the
future periods in 10- and 20-year windows. This is also
Figure 9
|
Medians of signal-to-noise ratio (SNR) of mean monthly streamflow for 10-, 20-, 33- and 50-year windows.
937
Y. Hui et al.
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
the case for most months and future periods in the 33-year
be enough for calibrating a bias correction method for
window. For the 50-year window, the values of SNR
impact studies.
increase to larger than 1 in the wet season, indicating that the anthropogenic climate change may be more impor-
Uncertainty of baseline observations
tant than the bias nonstationarity resulting from natural climate variability for the wet season. In addition, the
In this study, the CRU dataset was used as the baseline to calcu-
impact of bias nonstationarity on future climate change is
late the bias of GCM-simulated precipitation and temperature.
amplified in hydrological simulations, as indicated by the
The CRU data is gridded data obtained by interpolating gauged
fact that the SNR of streamflow is smaller than that of pre-
precipitation and temperature to regular grids. Due to limited
cipitation and temperature.
gauges used for interpolation, the CRU dataset may also be biased. In order to investigate the uncertainty related to the baseline dataset, another dataset produced by the University
DISCUSSION AND CONCLUSION
of Delaware (version 4.01, available at www.esrl.noaa.gov/ psd/data/gridded/data.UDel_AirT_Precip.html), with a spatial
Results summary
resolution of 0.5 from 1901 to 2000, the same as the CRU dataset, is also used to estimate the difference in bias between each
This study first investigated the bias nonstationarity of GCM-
historical period and the baseline period. Figure 10 shows the
simulated precipitation and temperature in the context of
difference in bias for mean annual precipitation and tempera-
natural climate variability for multiple time windows. The
ture obtained from CRU and the University of Delaware
role of bias nonstationarity in future climate changes was
(USA) dataset for 10-, 20-, 33- and 50-year windows. Similarly,
estimated by comparing the range of bias in the historical
the difference in bias varies with time for all time windows.
periods to climate change signal in the future periods.
Meanwhile, the range of difference in bias decreases with the
Finally, the propagation of bias nonstationarity in hydrologi-
extension of the time window. Generally, the range of differ-
cal simulations is investigated by running a hydrological
ence in bias is similar for two datasets, even though the
model. The main findings are as follows: (1) the biases of
magnitude is not exactly the same.
GCM precipitation and temperature vary with time, due to natural climate variability, especially for the short time win-
Uncertainty of greenhouse gas emission scenarios
dows; (2) precipitation bias nonstationarity plays an important role in future precipitation changes at annual
The role of bias nonstationarity on future climate changes may
and monthly scales, while the importance of temperature
rely on greenhouse gas emission scenarios, as different scen-
bias nonstationarity in future temperature change is not sig-
arios predict different future climate change signal. All the
nificant; (3) the bias nonstationarity of climate model
above results are based on the scenario of RCP4.5. In order
outputs is amplified when driving a hydrological model for
to investigate the uncertainty related to the greenhouse gas
hydrological simulations. In other words, the bias nonstatio-
emission scenario, the climate change signal of mean annual
narity of precipitation and temperature has considerable
precipitation and temperature predicted by RCP4.5 and
impacts on future streamflow changes for this specific water-
RCP8.5 are compared (Figure 11). The results showed that cli-
shed. This implies that the usually predicted hydrological
mate change signals predicted by RCP4.5 and RCP8.5 are
change in the future may be just the result of bias nonstatio-
comparable for precipitation, while the former is much smaller
narity due to natural climate variability. In addition, the
than the latter for temperature in terms of both median value
increase in the length of time window mitigates the impacts
and uncertainty related to GCMs for all time windows. Thus,
of bias nonstationarity on streamflow projections. Thus, it
the use of different greenhouse gas emission scenarios would
suggests using a long period (e.g. 50 years) for calibrating a
not change the conclusion that the role of precipitation bias
bias correction method in hydrological climate change
nonstationarity is important in future precipitation change.
impact studies. The climatology of 20 or 30 years may not
However, the impacts of temperature bias nonstationarity
938
Figure 10
Y. Hui et al.
|
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
The difference in bias of (a) mean annual precipitation, (b) temperature obtained from CRU (blue) and the University of Delaware dataset (red) for 10-, 20-, 33- and 50-year windows. The dash lines are for 17 GCMs, and the thick line is their median value. Please refer to the online version of this paper to see this figure in color: http://dx.doi.10. 2166/nh.2020.254.
may depend on emission scenarios. For a high emission scen-
deviation of bias over historical periods. The median value is
ario (e.g. RCP8.5), the bias nonstationarity will become even
obtained from all 17 GCMs. The results show that the range
less important, while for a low emission scenario (e.g.
of bias is much larger than the standard deviation of bias for
RCP4.5), it may become important for near future periods.
both precipitation and temperature, especially over the short time window. When using the standard deviation of bias
Methods to estimate natural climate variability
instead of the range of bias, the lower values of ‘noise’ can lead to larger SNR, indicating the impacts of bias nonstationar-
For the historical period, the difference in bias between two
ity would be less important. However, the impacts of
periods is caused by natural climate variability, because the
precipitation bias are still great for most GCMs in the near
anthropogenic climate change is pre-removed. Thus, to investi-
and mid future, as indicated by the SNR being smaller than 1.
gate the role of bias nonstationarity in future climate change, the range of bias is defined as ‘noise’, i.e. the difference between
Future work
maximum and minimum values of biases across multiple periods with outliers deleted. An alternative approach using
This study only investigated the bias nonstationarity of precipi-
the standard deviation of bias over multiple periods can also
tation and temperature for the historical period. Thus, the bias
be used to estimate natural climate variability. Table 2 presents
nonstationarity only resulted from natural climate variability.
a comparison of the median of range of bias and standard
However, for a future period, bias nonstationarity resulted
Y. Hui et al.
939
Figure 11
|
|
Impacts of bias nonstationarity on hydrology
Hydrology Research
|
51.5
|
2020
The climate change signal of (a) mean annual precipitation and (b) temperature under RCP4.5 (blue) and RCP8.5 (red) for 10-, 20-, 33- and 50-year windows. The dashed lines are for 17 GCMs, and the thick line is their median value. Please refer to the online version of this paper to see this figure in color: http://dx.doi.10.2166/nh.2020.254.
Table 2
|
A comparison of the median of range of bias and standard deviation of bias over historical periods for 10-, 20-, and 33-year windows
Precipitation (%)
Temperature ( C)
Standard
Standard
the atmospheric physics, due to the limitation of inadequate knowledge of climate system, such as cloud and convective precipitation mechanisms, surface albedo feedback and land–atmosphere interactions. Thus different models simulate somewhat different responses to the same external forcing.
Range
deviation
Range
deviation
10-year window
25.63
6.35
0.89
0.28
This may lead to different biases between different climate
20-year window
12.31
5.29
0.57
0.22
models and observations. As the attribution of both natural
33-year window
6.13
3.41
0.38
0.19
climate variability and climate model sensitivity, the perform-
50-year window
4.67
3.31
0.16
0.11
ance of a bias correction method may be more deteriorated in the future climate change than historical period. This will
from both natural climate variability and climate model sensi-
transfer to hydrological simulations, implying that hydrologi-
tivity (Chen et al. ; Velázquez et al. ; Hui et al. ).
cal simulations forced by bias-corrected precipitation and
The climate model sensitivity is more significant than natural
temperature may perform even worse when taking into
climate variability, especially for precipitation bias nonstatio-
account the impacts of climate model sensitivity. The impacts
narity (Hui et al. ). Climate models were proposed using
of climate model sensitivity on bias nonstationarity can be
different structures and parameterization schemes to represent
investigated in future studies.
940
Y. Hui et al.
|
Impacts of bias nonstationarity on hydrology
Due to the limited length of observed data, 100 years data were divided into 10, five, three and two nonoverlapping periods respectively corresponding to four different window lengths (10, 20, 33 and 50 years). For each time window, the biases over these finite periods were used to represent the range of biases (RB) caused by natural climate variability. However, small sample size, e.g. two non-overlapping periods for the 50-year window, may lead to large uncertainty of the estimated RB. To investigate the impacts of sample size on the estimation of bias variability, the use of multiple member GCMs may be a solution (Chen & Brissette ). In addition, this study only used the TwoParameter Monthly Water Balance Model to simulate monthly runoff. The hydrological model uncertainty was not considered. All these can be avenues for future studies.
ACKNOWLEDGEMENTS This work was partially supported by the National Natural Science Foundation of China (Grant No. 51779176 and 51539009) and the National Key R&D Program of China (2019YFC0408903). The authors would like to acknowledge the contribution of the World Climate Research Program Working Group on Coupled Modelling, which is responsible for CMIP. We wish to thank the climate modeling groups (listed in Table 1 of this paper) for producing and making available their model output.
DATA AVAILABILITY STATEMENT Data cannot be made publicly available; readers should contact the corresponding author for details.
REFERENCES Arnell, N. W. Relative effects of multi-decadal climatic variability and changes in the mean and variability of climate due to global warming: future streamflows in Britain. J. Hydrol. 270 (3–4), 195–213. Bai, P., Liu, X. M., Liang, K. & Liu, C. M. Comparison of performance of twelve monthly water balance models in different climatic catchments of China. J. Hydrol. 529, 1030–1040.
Hydrology Research
|
51.5
|
2020
Bellenger, H., Guilyardi, E., Leloup, J., Lengaigne, M. & Vialard, J. ENSO representation in climate models: from CMIP3 to CMIP5. Clim. Dyn. 42 (7–8), 1999–2018. Buser, C. M., Kunsch, H. R., Luthi, D., Wild, M. & Schar, C. Bayesian multi-model projection of climate: bias assumptions and interannual variability. Clim. Dyn. 33, 849–868. Chen, J. & Brissette, F. P. Reliability of climate model multimember ensembles in estimating internal precipitation and temperature variability at the multi-decadal scale. Int. J. Climatol. 39, 843–856. Chen, H., Guo, S. L., Xu, C.-Y. & Singh, V. P. Historical temporal trends of hydro-climatic variables and runoff response to climate variability and their relevance in water resource management in the Hanjiang basin. J. Hydrol. 344 (3–4), 171–184. Chen, J., Brissette, F. P. & Lucas-Picher, P. Assessing the limits of bias-correcting climate model outputs for climate change impact studies. J. Geophys. Res. Atmos. 120 (3), 1123–1136. Chen, J., Brissette, F. P. & Lucas-Picher, P. a Transferability of optimally-selected climate models in the quantification of climate change impacts on hydrology. Clim. Dyn. 47, 3359–3372. Chen, J., St-Denis, B. G., Brissette, F. P. & Lucas-Picher, P. b Using natural variability as a baseline to evaluate the performance of bias correction methods in hydrological climate change impact studies. J. Hydrometeorol. 17, 2155–2174. Christensen, J. H., Boberg, F., Christensen, O. B. & Lucas-Picher, P. On the need for bias correction of regional climate change projections of temperature and precipitation. Geophys. Res. Lett. 35, L20709. Dixon, K. W., Lanzante, J. R., Nath, M. J., Hayhoe, K., Stoner, A., Radhakrishnan, A., Balaji, V. & Gaitan, C. F. Evaluating the stationarity assumption in statistically downscaled climate projections: is past performance an indicator of future results? Clim. Change 135, 395–408. Duan, Q., Sorooshian, S. & Gupta, V. K. Effective and efficient global optimization for conceptual rain-runoff models. Water Resour. Res. 28 (4), 1015–1031. Ehret, U., Zehe, E., Wulfmeyer, V., Warrach-Sagi, K. & Liebert, J. HESS opinions ‘Should we apply bias correction to global and regional climate model data?’. Hydrol. Earth Syst. Sci. 16, 3391–3404. Fuentes-Franco, R., Giorgi, F., Coppola, E. & Kucharski, F. The role of ENSO and PDO in variability of winter precipitation over North America from twenty first century CMIP5 projections. Clim. Dyn. 46 (9–10), 3259–3277. Graham, L. P., Andreasson, J. & Carlsson, B. Assessing climate change impacts on hydrology from an ensemble of regional climate models, model scales and linking methods – a case study on the Lule River basin. Clim. Change 81, 293–307. Guo, S. L., Wang, J. X., Xiong, L. H., Ying, A. W. & Li, D. F. A macro-scale and semi-distributed monthly water balance model to predict climate change impacts in China. J. Hydrol. 268 (1–4), 1–15. Gutierrez, J. M., San-Martin, D., Brands, S., Manzanas, R. & Herrera, S. Reassessing statistical downscaling
941
Y. Hui et al.
|
Impacts of bias nonstationarity on hydrology
techniques for their robust application under climate change conditions. J. Clim. 26 (1), 171–188. Harris, C. R. U., & Jones, I. C. & D, P. Climatic Research Unit (CRU) Time-Series (TS) Version 4.01 of High-Resolution Gridded Data of Month-by-Month Variation in Climate (Jan. 1901–Dec. 2016). Centre for Environmental Data Analysis. Oxon, UK. Hui, Y., Chen, J., Xu, C.-Y., Xiong, L. H. & Chen, H. Bias nonstationarity of global climate model outputs: the role of internal climate variability and climate model sensitivity. Int. J. Climatol. 39 (4), 2278–2294. Hulme, M., Barrow, E. M., Arnell, N. W., Harrison, P. A., Johns, T. C. & Downing, T. E. Relative impacts of human-induced climate change and natural climate variability. Nature 397 (6721), 688–691. Kendall, M. G. Rank Correlation Methods. Griffin, London, UK. Li, L., Diallo, I., Xu, C.-Y. & Stordal, F. Hydrological projections under climate change in the near future by RegCM4 in Southern Africa using a large-scale hydrological model. J. Hydrol. 528, 1–16. Lu, W. & Qin, X. S. Integrated framework for assessing climate change impact on extreme rainfall and the urban drainage system. Hydrol. Res. 51, 77–89. Mann, H. B. Nonparametric tests against trend. Econometrica 13, 245–259. Maraun, D. Nonstationarities of regional climate model biases in European seasonal mean temperature and precipitation sums. Geophys. Res. Lett. 39, L06706. Maraun, D., Wetterhall, F., Ireson, A. M., Chandler, R. E., Kendon, E. J., Widmann, M., Brienen, S., Rust, H. W., Sauter, T., Themeßl, M., Venema, V. K. C., Chun, K. P., Goodess, C. M., Jones, R. G., Onof, C., Vrac, M. & Thiele-Eich, I. Precipitation downscaling under climate change: recent developments to bridge the gap between dynamical models and the end user. Rev. Geophys. 48 (3), RG3003. Marhaento, H., Booij, M. J. & Hoekstra, A. Y. Attribution of changes in stream flow to land use change and climate change in a mesoscale tropical catchment in Java, Indonesia. Hydrol. Res. 48 (3–4), 1143–1155. Maurer, E. P., Das, T. & Cayan, D. R. Errors in climate model daily precipitation and temperature output: time invariance and implications for bias correction. Hydrol. Earth Syst. Sci. 17 (6), 2147–2159. Nahar, J., Johnson, F. & Sharma, A. Assessing the extent of non-stationary biases in GCMs. J. Hydrol. 549, 148–162. Nash, J. E. & Sutcliffe, J. V. River flow forecasting through conceptual models. J. Hydrol. 10, 282–290. Olsson, T., Jakkila, J., Veijalainen, N., Backman, L., Kaurola, J. & Vehvilainen, B. Impacts of climate change on temperature, precipitation and hydrology in Finland-studies using bias corrected Regional Climate Model data. Hydrol. Earth Syst. Sci. 19 (7), 3217–3238. Ouyang, R., Liu, W., Fu, G., Liu, C., Hu, L. & Wang, H. Linkages between ENSO/PDO signals and precipitation,
Hydrology Research
|
51.5
|
2020
streamflow in China during the last 100 years. Hydrol. Earth Syst. Sci. 18 (9), 3651–3661. Polade, S. D., Gershunov, A., Cayan, D. R., Dettinger, M. D. & Pierce, D. W. Natural climate variability and teleconnections to precipitation over the Pacific-North American region in CMIP3 and CMIP5 models. Geophys. Res. Lett. 40 (10), 2296–2301. Power, S., Casey, T., Folland, C., Colman, A. & Mehta, V. Inter-decadal modulation of the impact of ENSO on Australia. Clim. Dyn. 15, 319–324. Ragettli, S., Tong, X., Zhang, G., Wang, H., Zhang, P. & Stähli, M. Climate change impacts on summer flood frequencies in two mountainous catchments in China and Switzerland. Hydrol. Res. https://doi.org/10.2166/nh.2019.118. Ruiz-Barradas, A., Nigam, S. & Kavvada, A. The Atlantic Multidecadal Oscillation in twentieth century climate simulations: uneven progress from CMIP3 to CMIP5. Clim. Dyn. 41, 3301–3315. Shen, M., Chen, J., Zhuan, M., Chen, H., Xu, C.-Y. & Xiong, L. Estimating uncertainty and its temporal variation related to global climate models in quantifying climate change impacts on hydrology. J. Hydrol. 556, 10–24. Teutschbein, C. & Seibert, J. Bias correction of regional climate model simulations for hydrological climate-change impact studies: review and evaluation of different methods. J. Hydrol 456–457, 12–29. Thornthwaite, C. W. An approach toward a rational classification of climate. Geograph. Rev. 38 (1), 55–94. Van Pelt, S. C., Kabat, P., Ter Maat, H. W., Van den Hurk, B. J. J. M. & Weerts, A. H. Discharge simulations performed with a hydrological model using bias corrected regional climate model input. Hydrol. Earth Syst. Sci. 13 (12), 2387–2397. Velázquez, J. A., Troin, M., Caya, D. & Brissette, F. Evaluating the time-invariance hypothesis of climate model bias correction: implications for hydrological impact studies. J. Hydrometeorol. 16 (5), 2013–2026. Wang, Y., Sivandran, G. & Bielicki, J. M. The stationarity of two statistical downscaling methods for precipitation under different choices of cross-validation periods. Int. J. Climatol. 38 (S1), e330–e348. Xiong, L. H. & Guo, S. L. A two-parameter monthly water balance model and its application. J. Hydrol. 216 (1–2), 111–123. Xu, C.-Y. & Singh, V. P. Evaluation and generalization of temperature-based methods for calculating evaporation. Hydrol. Process. 15 (2), 305–319. Zhuan, M. J., Chen, J., Shen, M. X., Xu, C.-Y., Chen, H. & Xiong, L. H. Timing of human-induced climate change emergence from internal climate variability for hydrological impact studies. Hydrol. Res. 49 (2), 421–437. Zhuan, M. J., Chen, J., Xu, C.-Y., Zhao, C., Xiong, L. H. & Liu, P. A method for investigating the relative importance of three components in overall uncertainty of climate projections. Int. J. Climatol. 39, 1853–1871.
First received 2 May 2020; accepted in revised form 25 August 2020. Available online 7 October 2020
942
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Drought prediction models driven by meteorological and remote sensing data in Guanzhong Area, China Jianzhu Li, Siyao Zhang, Lingmei Huang, Ting Zhang and Ping Feng
ABSTRACT Drought is an important factor that limits economic and social development due to its frequent occurrence and profound influence. Therefore, it is of great significance to make accurate predictions of drought for early warning and disaster alleviation. In this paper, SPEI-1 was confirmed to classify drought grades in the Guanzhong Area, and the autoregressive integrated moving average (ARIMA), random forest (RF) and support vector machine (SVM) model were established. Meteorological data and remote sensing data were used to derive the prediction models. The results showed the following. (1) The SVM model performed the best when the models were developed using meteorological data, remote sensing data and a combination of meteorological and remote sensing data, but the model’s corresponding kernel functions are different and include linear, polynomial and Gaussian radial basis kernel functions, respectively. (2) The RF model driven by the remote sensing data and the SVM model driven by the combined meteorological and remote sensing data were found to perform better than the model driven by the corresponding other data in the Guanzhong Area. It is difficult to accurately measure drought with the single meteorological data. Only by considering the combined factors can we more accurately monitor and predict drought. This study can provide an important scientific basis for regional drought warnings and predictions. Key words
| drought prediction, integrated autoregressive moving average model, random forest model, standardized precipitation evaporation index, support vector machine model
HIGHLIGHTS
• •
SPEI-1 was used to analyze the temporal distribution characteristics of drought and the main driving factors in Guanzhong Area, China. Drought grades were selected as the dependent variable, and the meteorological, geographical and vegetative factors were selected as the independent variables to establish an autoregressive integrated moving average (ARIMA) model, random forest (RF) model and support
• •
vector machine model. Meteorological data and remote sensing data were used as independent variables to derive prediction models, respectively. Comparing the models driven by remote sensing data only and the combination of meteorological and remote sensing data, the RF model driven by the remote sensing data and the SVM model driven by the combined meteorological and remote sensing data were found to
•
perform better than the model driven by the corresponding other data in Guanzhong Area. This study can provide an important scientific basis for regional drought warning and prediction.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.184
Jianzhu Li Siyao Zhang Ting Zhang (corresponding author) Ping Feng State Key Laboratory of Hydraulic Engineering Simulation and Safety, Tianjin University, Tianjin, China E-mail: zhangting_hydro@tju.edu.cn Lingmei Huang Faculty of Water Resources and Hydroelectric Engineering, Xi’an University of Technology, Xi’an, China
943
J. Li et al.
|
Drought prediction by meteorological and remote sensing data
Hydrology Research
|
51.5
|
2020
INTRODUCTION As a severe natural disaster, drought not only affects
shortages in crops and has shown better performances in
economic development but also influences water resources,
capturing spatial and temporal characteristics (Yuan ).
agriculture, ecology and environments (Mazdiyasni &
Based on SPI and PDSI concepts, Vicente-Serrano et al.
Aghakouchak ; Malik et al. ; Zhang et al. a,
() proposed a standardized precipitation evaporation
b; Guo et al. ). Due to the uncertainty in when
index (SPEI) (Yuan & Zhou ; Vicente-Serrano et al.
droughts begin and end, it is very difficult to predict drought.
), which can express the dry and wet conditions of the
Therefore, drought has become one of the critical factors
land surface on multiple scales with the potential evapor-
limiting the sustainable development of the economy and
ation included. Zhuang et al. () and Wang & Chen
society in many areas (Yuan & Zhou ; Zhang et al.
() investigated the application of the SPEI in China
a, b; Dai et al. ). As it increases in severity,
and found that this index has good applicability in China
drought has a profound impact on biogeochemical pro-
and can accurately represent the occurrence of drought.
cesses in terrestrial ecosystems (Fang et al. ). Drought
Hernandez & Annette () applied SPI and SPEI in con-
also affects the water absorption of vegetation by affecting
junction with precipitation and temperature projections
soil water content and increases the sensitivity of vegetation
from two general circulation models at six major urban cen-
to water energy, which plays an important role in terrestrial
ters of south Texas spanning five climatic zones. Both the
water, energy and carbon cycles (Fang et al. ; Yinglan
models predicted a progressively increasing aridity in the
et al. a, b). A warming climate is expected to perturb
region throughout the 21st century. Li et al. () investi-
the hydrological cycle, resulting in changes in both the fre-
gated the spatiotemporal characteristics of drought in the
quency and duration of drought (Han et al. ). To
Weihe River Basin by employing the SPEI index. Vega
alleviate drought effects, it is of great significance to
et al. () investigated hydrological patterns in the Brazi-
strengthen the study of regional drought characteristics
lian rainforest through a 9-month SPEI series and
and to make more accurate drought predictions for early
determined the Hurst exponents from detrended time
warnings and disaster mitigation (Miao ).
series of days with precipitation and accumulated monthly
The drought index is a primary factor in drought predic-
rainfall. The researchers found that the Hurst exponent cor-
tion, and many indices were presented in the previous
related positively with the monthly mean rainfall. The SPEI
studies for different drought types. The standardized precipi-
not only considers temperature and precipitation but also
tation index (abbreviated as SPI, see McKee et al. ;
evapotranspiration (Lu ), and it has been confirmed to
Bonaccorso et al. ) and the Palmer drought severity
be applicable and better than other drought indices in the
index (abbreviated as PDSI, refer to Palmer ) are two
Guanzhong Area (Xu ). Therefore, the SPEI was
common indicators used to characterize regional drought.
selected as the drought index to evaluate the drought
The SPI was proposed by McKee et al. () and has
events in the Guanzhong Area in this paper.
been widely used as an indicator of drought. This index
The traditional drought index is usually based on hydro-
only requires long-term (generally more than 30 years)
meteorological data measured at stations, and the spatial
precipitation data (Che & Li ; Li et al. ), but it
resolution of the index does not necessarily meet the
cannot reflect the seasonal distribution characteristics of
requirements of drought monitoring in large-scale regions.
precipitation (Dong & Xie ). The PDSI considers the
Meteorological satellites, which are widely used in drought
balance of water resources, including precipitation and
remote sensing monitoring (Di et al. ; Sahoo et al.
evaporation processes, additional runoff values, soil moist-
), can acquire multitemporal, multispectral, continuous
ure content and other conditions (Zargar et al. ), and
and complete data (Kogan ; Yilmaz et al. ). At pre-
the PDSI has been extensively used to monitor long-term
sent, the most commonly used and better-performing remote
drought (Liu et al. ). This index also considers water
sensing drought index is the normalized vegetation index
944
J. Li et al.
|
Drought prediction by meteorological and remote sensing data
Hydrology Research
|
51.5
|
2020
(NDVI) (Farrar et al. ; Bannari et al. ; Sun ),
The main aims of this study are: (1) to analyze the tem-
which can effectively monitor the dynamic changes in veg-
poral distribution characteristics of drought and the main
etation cover (Kogan ; Chen et al. ). Ichii et al.
driving factors in the Guanzhong Area, China; (2) to build
() suggested that there was a relative difference in the
an ARIMA model, RF model and SVM model, and predict
correlation between NDVI and natural meteorological fac-
drought grades by using gauge-based and remote sensing
tors at different geographical latitudes and that NDVI was
monitoring data, respectively; and (3) to select the best pre-
significantly correlated with temperature as well as regional
diction method and the corresponding predictors.
precipitation. Yan et al. () studied the relationship between NDVI and climate data in coastal areas of the Jiangsu Province and confirmed that both temperature and
STUDY AREA AND DATA
precipitation showed a significantly positive correlation with NDVI data in the region.
Study area
To reduce the losses caused by drought, drought prediction is needed on the basis of real-time drought monitoring.
The Guanzhong Area is located in the middle of the
Commonly used prediction methods include time series
Shaanxi Province, between the Loess Plateau and the
analysis, artificial neural networks (Moody & Darken
Qinling Mountains, China. The geographical coordinates
), support vector machines (SVMs) (Ahmad et al. ;
are 33 340 N–35 520 N, 106 180 E–110 380 E, with an area of
Weng ) and random forests (RFs) (Breiman ;
36,000 km2. The administrative divisions in the Guanzhong
Gislason et al. ; Cong ). The autoregressive inte-
Area include the cities of Xi’an, Xianyang, Weinan, Baoji
grated moving average (ARIMA) (Box & Jenkins ;
and Tongchuan (Wei et al. ), and this area is an impor-
Shumway & Stoffer ) model proposed by Box & Jenkins
tant area connecting the eastern and western parts and
() is a common model used in time series analysis (Han
the northern and southern parts of China (Qiao ).
et al. ). Yurekli et al. () used the ARIMA model to
Additionally, this area is a key construction area of the
simulate the 5-year monthly runoff observation data of the
Belt and Road Initiative.
Kelkit River (Bai et al. ). Zhang et al. () used the
The topography and geomorphology of the Guanzhong
ARIMA model to predict the drought of the northern
Area are complex with elevations between 270 and
Haihe River in China, indicating that the ARIMA model
2,439 m. The altitudes are high in the western, northern
has a good prediction accuracy. Machine learning has
and southern regions and low in the eastern and middle
been widely used to predict drought. For example, Fan
areas (Figure 1). The topography descends in steps from
et al. () established a drought prediction model in
the mountainous area to the center of the basin and in
autumn in the Zhejiang Province based on the SVM
turn includes the piedmont alluvial plain, the loess plateau
method with a radial basis kernel function and a cross-vali-
and the terrace of the river valley (Qiao ).
dation approach. The optimal model parameters were then
The Guanzhong Area is located in the transitional zone
determined, and they concluded that the developed model
between arid and humid regions (Li et al. ), which
has high prediction accuracy (Fan et al. ). Wu et al.
belongs to a continental monsoon climate with cold winters,
() employed the RF model to analyze the drought
hot summers and distinct seasonal features. During the same
grades of 21 representative stations from 1962 to 2012 in
period of heat and rain, drought is more likely to occur in
the Huaihe River Basin. The researchers found that the over-
the Guanzhong Area. The average annual precipitation is
all average prediction accuracy is higher than the weather
500–700 mm, with precipitation concentrated mostly in
system’s weather prediction accuracy, suggesting acceptable
summer and autumn and little precipitation occurs in
prediction results. ARIMA, RF and SVM have been used in
winter (Qiao ).
drought prediction, but the performance of these three
The special topography results in climatic conditions of
methods needs to be further explored to improve the accu-
high temperature and low rainfall. The area is mostly plain
racy of drought prediction in the Guanzhong Area.
formed by loess sedimentation and river alluvial sediments
945
Figure 1
J. Li et al.
|
|
Drought prediction by meteorological and remote sensing data
Hydrology Research
|
51.5
|
2020
Location of the Guanzhong Area and the meteorological stations.
1976–1980
and
1994–1997
(Gao
).
with soft soil, poor water retention capacity, and therefore,
1971–1972,
the area is prone to drought. In terms of agriculture,
Droughts in the Guanzhong Area occurred mostly in
approximately 70% of the land on the Guanzhong Plain
spring and summer, and continuous drought events fre-
is covered by cropland, which mainly includes grain
quently occurred, lasting from spring to summer and from
crops, fruit woodlands and vegetables. More than 50% of
summer to autumn (Wu ).
the croplands in this area are rainfed. There are many irrigated croplands in the western and middle parts, while
Data
most of the croplands in the east are rainfed (Zhou et al. ). Due to the dense population and developed agricul-
The data used for drought prediction in the Guanzhong
ture, the increasing industrial and agricultural water
Area include gauge-based meteorological data and remote
supply led to a more severe drought situation. Thus, the
sensing data.
Guanzhong Area is known by the saying, ‘9 drought years out of 10 years’ (Yu ).
The gauge-based meteorological data include the daily wind speed, precipitation, air temperature, air pressure, sun-
According to the ‘Record of Natural Disasters in the
shine hours and relative humidity of 11 meteorological
Shaanxi Province’, from the 2nd century to 1949 AD,
stations in the Guanzhong Area, covering the period from
there were more than 600 drought events in the Shaanxi
January 1960 to December 2016. The data were downloaded
Province, and 326 in the Guanzhong Area, accounting
from the Meteorological Data Sharing Service System of the
for 54% of the total drought records (Meteorological
China Meteorological Administration (http://data.cma.cn/).
Station of Shaanxi Meteorological Bureau ). There
The remote sensing data are the land surface tempera-
were 22 large-scale drought events in the Guanzhong
ture of the Guanzhong Area and the NDVI from July 2002
Area from 1949 to present, especially the continuous
to June 2011. These data were downloaded from the Inter-
severe droughts that occurred in 1959–1961, 1966–1967,
national Scientific Data Mirror website of the Chinese
J. Li et al.
946
|
Drought prediction by meteorological and remote sensing data
Academy of Sciences Computer Network Information Center
(http://www.gscloud.cn).
The
digital
elevation
model (DEM) was also obtained from this website. Detailed information on the remote sensing data used in this study is shown in Table 1.
METHODS
Hydrology Research
Table 2
|
|
51.5
|
2020
The classification of drought based on SPEI
SPEI
Drought grades
Number
( 0.5,0)
no drought
1
( 1, 0.5]
light drought
2
( 1.5, 1]
moderate drought
3
( 2, 1.5]
heavy drought
4
∞, 2]
severe drought
5
Calculation of the SPEI-1 series
Predictive models
The SPEI index is calculated as follows:
ARIMA model
(1) Calculate the monthly water surplus and deficiency Di, Di ¼ Pi – PETi, with i indicating the month, Pi indicating
The ARIMA model is one of the most commonly used
the monthly precipitation (mm) and PETi the monthly
models in time series analyses (Zhao et al. ). The data
potential evaporation (mm), which is calculated using
series are formed by the prediction index with time regarded
the Morton-type Penman formula (Mei ).
as a random sequence. The dependence relation of this
(2) Calculate the probability distribution of the monthly
random series reflects the continuity of the original data in
water surplus and deficit Di by using the log–logistic
time, which has both the influence of external factors and
probability distribution function. (3) Standardize the log–logistic probability distribution function F(x) for monthly accumulated water surplus and deficit Di (Li et al. ).
its own change laws (Wu ). The ARIMA modeling steps are as follows: (1) Determine if the time series is stationary. The time series can be differentiated to obtain a stationary series if it is a
The severity of drought can be graded according to the value of SPEI, as shown in Table 2 (Zhang et al. ).
non-stationary series. The difference operators can be explained as each observation minus the previous one,
According to the multi time scale characteristics of the
in which d represents the difference times. The autore-
SPEI, the SPEI on different time scales can reflect changes
gressive moving average (ARMA) model can be
in humidity and dryness in different periods (Yang et al.
described as ARMA (p,q), and the ARIMA model can
). SPEI-1 refers to the SPEI index on a monthly time
be described as ARIMA (p,d,q). When the time series
scale. The trend of SPEI values on different time scales is
is stationary, d ¼ 0, and the ARIMA model becomes
consistent overall, but the trend of humidity and dryness reflected by SPEI values on short time scales is more specific. Therefore, SPEI-1 is selected in this paper.
the ARMA model (Ghashghaie & Nozari ). (2) Determine p and q (p represents the lag order of the autoregressive processes, and q represents the lag order of the moving average processes (Ghashghaie &
Table 1
|
Nozari )). The autocorrelation function (ACF) Detailed information on the remote sensing data
graph and the partial ACF (PACF) graph are drawn, Spatial
Temporal
and the ARMA(p,q) models are judged based on the tail-
Production name
resolution
resolution
ing and truncation of the ACF and PACF graph (Wu
China 1 km land temperature monthly synthetic products (MYDLT1M)
1 km
month
).
China 500 m NDVI synthetic products (MYDND1M)
500 m
month
30 m resolution DEM data (GDEMV2)
30 m
–
(3) Fit the model and perform the normality test, autocorrelation test and white noise test on the residuals. (4) Use the ARIMA model to inspect and predict the monitoring series. For the seasonal non-stationary time series,
947
J. Li et al.
|
Drought prediction by meteorological and remote sensing data
Hydrology Research
|
51.5
|
2020
seasonal differences are also needed to obtain the
The SVM was originally proposed for the problem of
stationary time series and form the ARIMA (p,d,q)
binary pattern classification in the case of linear separability.
(P, D,Q)S model, where D represents the order of seaso-
Given a set of observation samples S ¼ {(x1, y1), (x2, y2)…}
nal differences; P and Q represent the order of seasonal
⊂X × { 1, 1}, X⊂Rn is called the input space or the input
autoregression and the order of seasonal moving
characteristic space, and yi∈{ 1,1} is the sample class tag.
averages, respectively; and S represents the length of
The purpose of classification is to find a classification hyper-
the seasonal period.
plane that completely separates the two classes. Let G ¼ {ω x þ b ¼ 0| ω ∈ Rn, x∈X, b∈R} be all hyperplane sets
RF model
that can completely and correctly classify S. In all hyperplanes, the maximum interval classifier is looking for an
The RF model uses the bootstrap resampling method to
optimal hyperplane that satisfies the two types of classifi-
extract multiple samples from the original samples, builds
cation intervals (the sum of the sample-to-hyperplane
the decision tree for each bootstrap sample, and then com-
distance closest to each other from the hyperplane). The
bines the predictions of multiple decision trees to obtain
linear inseparable problem of the input space can be trans-
the final prediction results by voting (Breiman ; Fang
formed into a linear separable problem by using the
et al. ). Specifically, the RF model is a combined classification model composed of many decision tree classification models (h(X,θi), i ¼ 1, 2, …, k), and the parameter set (θi) is an independent and identically distributed random vector. Given the independent variable X, each decision tree classi-
appropriate kernel functions to operate in two classes. Common kernel functions are linear kernel functions (Altman ), polynomial kernel functions, Gaussian radial basis kernel functions and sigmoid kernel functions (Wang ; Burges ).
fication model selects the optimal classification results by one-vote voting rights (Wang ). The modeling steps for the RF model are as follows: (1) Randomly extract M samples from the fitting data set T, and then fit the extracted data set. (2) For each instance, generate a decision tree, and at each node of the tree: a. randomly extract a subset of m variables from the p valid overall features; b. choose the best variables and the best partition from the set of m variables; and c. continue until the tree is fully generated (Wang ). (3) Perform RF predictions using all trees.
RESULTS AND DISCUSSION Calculation of the SPEI-1 series The SPEI-1 series of 11 meteorological stations in the Guanzhong Area are obtained based on monthly meteorological data. Here, we show only the results of Baoji Station in Figure 2. The SPEI-1 series of the Guanzhong Area was calculated according to the SPEI-1 series at the 11 stations, as shown in Figure 3. The Guanzhong Area is prone to drought disasters because of the frequent alternation of dry and wet conditions. In 1962, 1963, 1967, 1969, 1976, 1979, 1994– 2002 and 2007, the Guanzhong Area suffered severe
SVM model
drought events as shown in Figure 3 based on the drought classification in Table 2.
The SVM model seeks the best compromise between the
The occurrence frequency and percentage of drought at
complexity of the model (learning-intensive reading of
all grades in 12 months at Baoji Station are acquired and
specific training samples) and the ability to learn (the ability
shown in Table 3. Moderate drought mainly occurs in
to identify any samples without error) based on limited
autumn, while the occurrence of heavy drought is concen-
sample information for better promotion ability (Wang
trated in spring and summer. In addition, Baoji Station
).
experienced no severe drought during 1960–2016.
J. Li et al.
948
|
Drought prediction by meteorological and remote sensing data
Figure 2
|
SPEI-1 series of the Baoji Station.
Figure 3
|
SPEI-1 series of the Guanzhong Area.
Table 3
|
Hydrology Research
|
51.5
|
2020
The number and proportion of droughts at all grades in 12 months at Baoji Station
Moderate drought
Heavy drought
Severe drought
12.26%
7
9.21%
2
4.76%
0
–
6
5.66%
5
6.58%
4
9.52%
0
–
11
10.38%
5
6.58%
4
9.52%
0
–
7.71%
12
11.32%
6
7.89%
3
7.14%
0
–
8.81%
6
5.66%
5
6.58%
5
11.90%
0
–
39
8.59%
7
6.60%
6
7.89%
4
9.52%
0
–
37
8.15%
10
9.43%
5
6.58%
5
11.90%
0
–
8
40
8.81%
6
5.66%
7
9.21%
4
9.52%
0
–
9
41
9.03%
5
4.72%
6
7.89%
4
9.52%
0
–
10
37
8.15%
10
9.43%
8
10.53%
2
4.76%
0
–
11
36
7.93%
9
8.49%
10
13.16%
2
4.76%
0
–
12
37
8.15%
11
10.38%
6
7.89%
3
7.14%
0
–
Month
No drought
Light drought
1
34
7.49%
13
2
41
9.03%
3
37
8.15%
4
35
5
40
6 7
J. Li et al.
949
|
Drought prediction by meteorological and remote sensing data
In the ‘History of Natural Disasters in the Shaanxi Province’ and ‘Report on China’s Disaster Situation 1949–1995’
Hydrology Research
|
51.5
|
2020
in the Guanzhong Area, with a predicted length of 12 months.
(Liu ), the Guanzhong Area suffered flood disasters in the mid-1960s and early 1980s, and among them, 1964,
Drought prediction by the ARIMA model
1983 and 1984 were serious flooding years. There were drought events in the early 1960s, mid-late 1970s and
The SPEI-1 series of the 11 meteorological stations in the
1990s, among which, the continuous drought events that
Guanzhong Area were used for ARIMA modeling and pre-
occurred in 1959–1961, 1966–1967, 1971–1972, 1976–
diction. The ARIMA model was fitted using the SPEI-1
1980, 1994–1997 and 1999–2002 were severe. Drought
series from January 1960 to December 2015 and predicted
and floods occurred alternately from the mid-1980s to the
drought grades from January to December 2016 with a
early 1990s, and the climate was more humid in the 21st
length of 12 months. The ARIMA model parameters corre-
century (Cai et al. ; Lei et al. ). The wet and dry con-
sponding to each station are shown in Table 5, and the
ditions in Figure 3 are consistent with previous studies.
fitting and prediction results of Baoji Station are shown in
Therefore, the SPEI-1 can reasonably reflect the occurrence
Figure 4. The ARIMA model parameters at different stations
of drought in the Guanzhong Area.
have regional heterogeneity, depending on the natural con-
The number and proportion of droughts at all grades were calculated according to the SPEI-1 series at the 11
ditions, such as the underlying surface, geographical location and climatic characteristics.
stations, as shown in Table 4. The proportion of droughts
According to Figure 4(a), the SPEI-1 fitting series of the
at 11 stations is nearly 35%, mainly light and moderate
ARIMA model at each station is mostly consistent with the
droughts, and severe droughts hardly occur. In addition,
measuring series, which shows that the ARIMA model is
the number of droughts at all grades decreases with the
basically in line with reality and that the fitting performs
severity of drought.
well. The prediction SPEI-1 series shown in Figure 4(b) fluctuates in the range of 1 to 1, suggesting that the Guan-
Drought prediction by different models
zhong Area was in a no drought or light drought state in 2016. There is no sudden increase or decrease in SPEI-1
The three models previously mentioned in the ‘Predictive
values in 2016, and the change is relatively small. Among
models’ section were used to predict the drought grades
them, the prediction SPEI declined in 2016. The SPEI-1
Table 4
|
The number and proportion of droughts at all grades
Station
No drought
Light drought
Moderate drought
Heavy drought
Severe drought
Baoji
454
66.57%
108
15.84%
77
11.29%
43
6.30%
0
0.00%
Fengxiang Luochuan
451
66.13%
106
15.54%
90
13.20%
35
5.13%
0
0.00%
440
64.42%
122
17.86%
89
13.03%
30
4.39%
2
0.29%
Tongchuan
459
67.30%
106
15.54%
77
11.29%
38
5.57%
2
0.29%
Wugong
447
65.54%
108
15.84%
93
13.64%
33
4.84%
1
0.15%
Yaoxian
458
67.06%
110
16.11%
73
10.69%
41
6.00%
1
0.15%
Changwu
456
66.86%
102
14.96%
87
12.76%
34
4.99%
3
0.44%
Pucheng
457
67.01%
105
15.40%
84
12.32%
33
4.84%
3
0.44%
Longxian
461
67.60%
108
15.84%
87
12.76%
23
3.37%
3
0.44%
Yongshou
456
66.86%
99
14.52%
92
13.49%
34
4.99%
1
0.15%
Qindu
452
66.08%
118
17.25%
81
11.84%
26
3.80%
7
1.02%
average
454
66.49%
108
15.88%
85
12.39%
34
4.93%
2
0.31%
J. Li et al.
950
Table 5
|
|
Drought prediction by meteorological and remote sensing data
Hydrology Research
|
51.5
|
2020
The ARIMA model parameters corresponding to each station
Station
ARIMA(p,d,q)(P,D,Q)S
Station
ARIMA(p,d,q)(P,D,Q)S
Baoji
ARIMA(5,4,3)(1,0,1)12
Fengxiang
ARIMA(5,4,3)(1,0,2)12
Luochuan
ARIMA(3,1,2)(0,1,0)12
Tongchuan
ARIMA(2,2,3)(0,1,0)12
Wugong
ARIMA(3,1,2)(0,1,0)
12
Yaoxian
ARIMA(1,1,2)(0,1,0)12
Changwu
ARIMA(0,1,2)(1,1,0)12
Pucheng
ARIMA(1,2,2)(0,1,0)12
Longxian
ARIMA(1,1,2)(0,1,1)
12
Yongshou
ARIMA(4,3,3)(2,1,0)12
Qindu
ARIMA(1,1,2)(0,1,0)12
Guanzhong Area
ARIMA(0,1,2)(0,1,0)12
Figure 4
|
Fitting and prediction results of the ARIMA model at Baoji Station. (a) Fitting results of the ARIMA model in Baoji Station. (b) Prediction results of the ARIMA model in Baoji Station.
was lower after August 2016 when the drought was serious.
The fitting and prediction of SPEI values were graded
The SPEI-1 was at a higher level before August 2016, that is,
according to Table 2, and the qualified rate of the drought
the drought degree was lighter.
grade was calculated. The qualified fitting and prediction
J. Li et al.
951
|
Drought prediction by meteorological and remote sensing data
Hydrology Research
|
51.5
|
2020
rates of the ARIMA model at each station are shown in
temperature, average humidity, average wind speed, sunshine
Table 6. The drought prediction grades of the ARIMA
hours, rainfall and potential evaporation of the 11 meteorolo-
model can represent the actual situation to some extent,
gical stations in the Guanzhong Area were selected as the
and the average qualified fitting and prediction rates are
independent variables. The data from January 1960 to
0.5337 and 0.6061, respectively. Except for the qualified pre-
December 2015 were used as the fitting data, and the data
diction rate of the Longxian station, which is 0.7500 and is
from January 2016 to December 2016 were used as the pre-
significantly higher than the average, the qualified rates of
diction data with a predicted length of 12 months.
other stations fluctuate around the average. The case
The qualified fitting and prediction rates corresponding
where the qualified fitting rate is lower than the qualified
to the 11 stations of the RF model constructed by the
prediction rate may be due to the uneven length of the fitting
meteorological data were calculated and are shown in
and prediction data sets. The fitting data were used from Jan-
Table 7. The average qualified fitting and prediction rates
uary 1960 to December 2015 (56 years), and the prediction
are 0.7623 and 0.6515, respectively. The qualified prediction
data were used from January to December 2016 (1 year), so
rates are significantly different from the average qualified
that the internal variance in the fitting process was much
prediction rates at some stations, such as Baoji, Tongchuan,
larger than the prediction process, that is, the fitting process
Wugong and Yongshou with prediction accuracies of
had a large error with respect to the prediction process.
0.5000, 0.7500, 0.8333 and 0.5000, respectively. The qualified fitting and prediction rates of the other stations fluctuate around the average. Due to the noise in the fitting
Drought prediction by the RF model
data, the overfitting phenomenon is obvious in the model. RF model driven by meteorological data and accuracy assessment. SPEI-1 grade (no drought: 1, light drought: 2,
RF model driven by remote sensing data and accuracy
moderate drought: 3, heavy drought: 4, special drought: 5)
assessment. After classifying the SPEI-1 grades of 11 stations
was selected as the dependent variable. The monthly mean
in the Guanzhong Area, kriging spatial interpolation was per-
values of daily maximum temperature, daily minimum
formed to obtain the SPEI-1 spatial distribution. The DEM,
Table 6
|
Qualified fitting and prediction rate of the ARIMA model
Station
Fitting
Prediction
Station
Fitting
Prediction
Baoji
0.5417
0.5000
Changwu
0.5372
0.5833
Fengxiang
0.5461
0.6667
Pucheng
0.4940
0.6667
Luochuan
0.4717
0.5000
Longxian
0.6741
0.7500
Tongchuan
0.5134
0.6667
Yongshou
0.5655
0.5833
Wugong
0.4985
0.5833
Qindu
0.5149
0.5833
Yaoxian
0.5134
0.5833
Average
0.5337
0.6061
Table 7
|
Drought qualified fitting and prediction rates of the RF models in 11 stations (meteorological data)
Station
Fitting
Prediction
Station
Fitting
Prediction
Baoji
0.7887
0.5000
Changwu
0.7708
0.5833
Fengxiang
0.7723
0.5833
Pucheng
0.7827
0.8333
Luochuan
0.7426
0.5833
Longxian
0.7455
0.6667
Tongchuan
0.7753
0.7500
Yongshou
0.7307
0.5000
Wugong
0.7857
0.8333
Qindu
0.7321
0.6667
Yaoxian
0.7589
0.6667
Average
0.7623
0.6515
J. Li et al.
952
|
Drought prediction by meteorological and remote sensing data
Hydrology Research
|
51.5
|
2020
slope, slope direction, rainfall, NDVI, daytime land tempera-
values of daily maximum temperature, daily minimum temp-
ture (LTD) and nighttime land temperature (LTN) were used
erature, average humidity, average wind speed, sunshine
as independent variables, and SPEI-1 grades were used as the
hours, rainfall and potential evaporation obtained from 11
dependent variables. The RF model was constructed with
meteorological stations in the Guanzhong Area were used
the data from July 2002 to June 2010 as the fitting data and
as the independent variables, as well as DEM, slope, and
the data from July 2010 to June 2011 as the prediction data.
slope direction, rainfall, NDVI, LTD and LTN obtained
The average qualified fitting and prediction rates of the RF
from remote sensing products. SPEI-1 grades were used as
model constructed by remote sensing data in the Guanzhong
the dependent variable. The data from July 2002 to June
Area are 0.6746 and 0.6836, respectively. The independent
2010 were used for fitting the model, and the data from
variables corresponding to the 11 stations were extracted,
July 2010 to June 2011 were used for prediction. The quali-
and then, the qualified fitting and prediction rates were calcu-
fied fitting and prediction rates corresponding to the 11 stations are shown in Table 9. The average qualified fitting
lated as shown in Table 8. As seen from Table 6, the qualified fitting and prediction
and prediction rates are 0.6316 and 0.5388, respectively.
rates of the 11 stations are 0.6913 and 0.5616, respectively.
Except for the qualified prediction rate of Longxian station
Except for the qualified fitting rate of Luochuan and the qua-
(0.6667), which is significantly higher than the average qua-
lified prediction rate of Qindu, the qualified rates of other
lified prediction rate, the qualified rates of the other stations
stations fluctuate around the average qualified rates. The
fluctuate around the average.
model works well for the 11 stations, but the RF model built by using remote sensing data also presents a significant
Drought prediction by the SVM model
overfitting problem due to data noise. SVM model driven by meteorological data and accuracy RF model driven by combined meteorological and remote
assessment. By using the same fitting data and prediction
sensing data and accuracy assessment. The monthly mean
data as in the ‘RF model driven by meteorological data
Table 8
|
Drought qualified fitting and prediction rates of the RF models at 11 stations (remote sensing data)
Station
Fitting
Prediction
Station
Fitting
Prediction
Baoji
0.6979
0.6563
Changwu
0.6458
0.4792
Fengxiang
0.6667
0.5625
Pucheng
0.7396
0.5313
Luochuan
0.5833
0.4896
Longxian
0.6979
0.6250
Tongchuan
0.7188
0.5000
Yongshou
0.6771
0.4896
Wugong
0.7500
0.6042
Qindu
0.7813
0.6771
Yaoxian
0.6458
0.5625
Average
0.6913
0.5616
Fitting
Prediction
Table 9
|
Drought qualified fitting and prediction rates of the RF models in 11 stations (meteorological and remote sensing data)
Station
Fitting
Prediction
Station
Baoji
0.6667
0.6042
Changwu
0.5625
0.4583
Fengxiang
0.5938
0.4688
Pucheng
0.6667
0.5729
Luochuan
0.5625
0.5208
Longxian
0.6875
0.6667
Tongchuan
0.6042
0.4479
Yongshou
0.6354
0.4896
Wugong
0.6875
0.5938
Qindu
0.6875
0.6250
Yaoxian
0.5938
0.4792
Average
0.6316
0.5388
953
J. Li et al.
|
Drought prediction by meteorological and remote sensing data
and accuracy assessment’ section, four common kernel func-
Hydrology Research
Table 11
|
|
51.5
|
2020
The qualified rates of four common kernel functions (remote sensing data)
tions (linear kernel function, polynomial kernel function, Kernel
Linear kernel
Polynomial kernel
Gaussian radial basis kernel
Sigmoid kernel
function) were used to construct the SVM model as well
functions
function
function
function
function
Gaussian radial basis kernel function and sigmoid kernel as fit and predict the drought grades. The average qualified
Fitting
0.7247
0.7795
0.7523
0.6680
fitting and prediction rates at the 11 stations are shown in
Prediction
0.7269
0.7188
0.6548
0.7016
Table 10.
Difference
0.0022
0.0607
0.0975
0.0336
As shown in Table 10, the polynomial kernel function and the linear kernel function perform best in the fitting and prediction, respectively. The Gaussian radial basis
function in the prediction ranks second, and it differs from
kernel function ranks second for both the qualified fitting
the first-ranked linear kernel function by only 0.0081. In
and prediction rates. The qualified fitting and prediction
addition, the differences in the qualified fitting and predic-
rates of the sigmoid kernel function differ by 0.0364,
tion rates of the four kernel functions are less than 0.1,
which is the smallest and the most stable among the four
which shows that they are relatively stable. Therefore, the
kernel functions, but the qualified rates of fitting and predic-
polynomial kernel function is more suitable for drought pre-
tion are the lowest. The qualified rate of linear kernel
diction with remote sensing data driving the SVM model.
function fitting is 0.7532, which is only 0.0437 smaller than the highest Gaussian radial basis kernel function, and the qualified rate is the highest in the prediction process.
SVM model driven by combined meteorological and remote
The difference between qualified fitting and prediction
sensing data and accuracy assessment. Using the same fit-
rates is 0.0498, which is only larger than the sigmoid
ting data and prediction data as in the ‘RF model driven
kernel function (ranks first). Therefore, the linear kernel
by combined meteorological and remote sensing data and
function is considered to be more suitable for drought moni-
accuracy assessment’ section, four kernel functions were used to construct the SVM model as well as fit and predict
toring in the Guanzhong Area.
the drought grades. The qualified fitting and prediction SVM model driven by remote sensing data and accuracy
rates of the four kernel functions are shown in Table 12.
assessment. Using the same fitting data and the prediction
The Gaussian radial basis kernel function and the linear
data shown in the ‘RF model driven by remote sensing
kernel function perform best in the fitting and prediction
data and accuracy assessment’ section, four kernel functions
process, respectively. The Gaussian radial basis kernel func-
were used to construct the SVM model as well as fit and pre-
tion ranks second with a qualified prediction rate of 0.7652
dict the drought grades. The qualified fitting and prediction
and only differs from the linear kernel function (ranks first)
rates of the four kernel functions are shown in Table 11.
by 0.0075. That is, the Gaussian radial basis kernel function
The polynomial kernel function and the linear kernel
performs well in both fitting and prediction. Therefore, the
function perform best in the fitting and prediction process,
Gaussian radial basis kernel function is more suitable for
respectively. The qualified rate of the polynomial kernel
drought prediction with the SVM model driven by combined meteorological and remote sensing data.
Table 10
|
The average qualified rates of four common kernel functions (meteorological data)
Table 12
|
The qualified rates of four common kernel functions (meteorological and remote sensing data)
Linear
Polynomial
Gaussian radial
Sigmoid
Kernel
kernel
kernel
basis kernel
kernel
functions
function
function
function
function
Fitting
0.7532
0.7969
0.7676
0.6273
Prediction
0.8030
0.6970
0.7121
Difference
0.0498
0.0999
0.0555
Linear
Polynomial
Gaussian radial
Sigmoid
Kernel functions
kernel function
kernel function
basis kernel function
kernel function
0.5909
Fitting
0.8532
0.7727
0.9441
0.6799
0.0364
Prediction
0.7727
0.7348
0.7652
0.7273
954
J. Li et al.
|
Drought prediction by meteorological and remote sensing data
Comparison of the models
Hydrology Research
|
51.5
|
2020
The qualified prediction rates of the Gaussian radial basis kernel function rank second, the differences from the linear
Comparison of different models driven by meteorological data
radial basis kernel function which ranks first are small, and the performance is excellent. Therefore, the SVM model (Gaussian radial basis kernel function) driven by combined
Comparing the qualified fitting and prediction rates of the
meteorological and remote sensing data is more suitable for
ARIMA model, the RF model and the SVM model driven
drought prediction in the Guanzhong Area.
by meteorological data, it is found that among the three
According to the ‘Comparison of different models
models, the SVM model performs best in drought predic-
driven by meteorological data, Comparison of the models
tion, where the polynomial kernel function and linear
driven by remote sensing data, and Comparison of
kernel function perform best in the fitting and prediction
models driven by combined meteorological and remote sen-
process, respectively. The qualified fitting rate of the linear
sing data’ sections, the SVM model is superior to the other
kernel function is 0.7532, which is only 0.0437 smaller
two models for the following reasons. (1) The SVM model
than the highest Gaussian radial basis kernel function, and
performs better than the RF model when dealing with unba-
the qualified rate in the prediction is the highest. The differ-
lanced data. Drought grades in the Guanzhong Area are
ence between qualified fitting and prediction rates is 0.0498,
unbalanced, with light and moderate drought as the main
which is relatively stable. Therefore, it is considered that the
types, and heavy and severe droughts are rare. The samples
SVM model (linear kernel function) is more suitable for
of drought levels are unbalanced. (2) The ARIMA model is
drought monitoring in the Guanzhong Area.
better in terms of linear prediction and less effective in terms of nonlinear prediction. The SVM model can also
Comparison of the models driven by remote sensing data
solve nonlinear problems through the application of kernel functions.
Comparing the qualified rates of the RF model and SVM model based on remote sensing data, the SVM model per-
Comparison of the model-driven data
forms better in the prediction of drought, in which the polynomial kernel function and linear kernel function per-
For the RF model, except for Longxian station, the qualified
form best in the fitting and prediction, respectively. The
fitting and prediction rates show that the performance of
polynomial kernel function ranks second in the qualified
remote sensing data in the RF model performed better than
prediction rate and only differs from the linear kernel func-
the combined meteorological and remote sensing data. The
tion (ranks first) by 0.0081, and the difference between the
use of meteorological data will reduce the qualified fitting
qualified fitting and prediction rates is only 0.0607, which
and prediction rates of the RF model. However, the qualified
is relatively stable. Therefore, the SVM model (polynomial
fitting and prediction rates of remote sensing data at the 11
kernel function) driven by remote sensing data is more suit-
stations are only slightly higher than those of the combined
able for drought monitoring in the Guanzhong Area.
meteorological and remote sensing data. Therefore, the RF model driven by the remote sensing data performed well
Comparison of models driven by combined meteorological and remote sensing data
for drought monitoring in the Guanzhong Area.
Comparing the qualified rates of the RF model and SVM
other three kernel functions based on the remote sensing
model based on combined meteorological and remote sen-
data are lower than those driven by the combined meteoro-
sing data, the SVM model performs better than the RF
logical and remote sensing data. The difference in the
model in the prediction of drought, in which the Gaussian
qualified rates of the four kernel function models based on
In the SVM model, except for the polynomial kernel function, the qualified fitting and prediction rates of the
radial basis kernel function and linear kernel function per-
different data was calculated, which is shown in Table 13.
form best in the fitting and prediction process, respectively.
The differences between the fitting and prediction of the
955
Table 13
J. Li et al.
|
|
Drought prediction by meteorological and remote sensing data
Difference in the qualified rates of the SVM model for different data (absolute values)
Hydrology Research
|
51.5
|
2020
terms of drought prediction using meteorological data. The SVM model (polynomial kernel function) is
Linear
Polynomial
Gaussian radial
Sigmoid
Kernel
kernel
kernel
basis kernel
kernel
functions
function
function
function
function
Fitting
0.1285
0.0068
0.1918
0.0119
Prediction
0.0458
0.0160
0.1104
0.0257
superior if the models are driven by remote sensing data, while the SVM model (Gaussian radial basis kernel function) outperforms the other models if the predictors are a combination of meteorological and remote sensing data. (3) When using the remote sensing data and the combi-
polynomial and sigmoid kernel functions are mostly less
nation of meteorological and remote sensing data to
than 0.05, respectively, but large for the other kernel func-
build the RF model, the use of meteorological data has
tions. The use of meteorological data has different effects
a small reduction effect on the qualified fitting and pre-
on the qualified fitting and prediction rates for different
diction rate; the use of meteorological data has a
kernel functions and has little influence on the polynomial
different effect on qualified fitting and prediction rates
and sigmoid kernel functions. The use of meteorological
of different kernel functions in the SVM model. In con-
data significantly improves the qualified fitting and predic-
clusion, the RF model driven by the remote sensing data
tion rate of the Gaussian radial basis kernel function and
and the SVM model driven by the combined meteorolo-
the qualified fitting rate of the linear kernel function. The
gical and remote sensing data performed better than the
use of remote sensing data significantly improved the quali-
model driven by the corresponding other data in the
fied fitting and prediction rates (except for the polynomial
Guanzhong Area.
kernel function for the fitting and linear kernel functions
Different types and lengths of meteorological data and
for prediction). It is difficult to accurately measure the pro-
remote sensing data used in this paper may affect the fitting
cess of drought using a source of meteorological data. Only
and prediction accuracy of meteorological and remote sen-
by considering the factors of precipitation, temperature, veg-
sing data models, which should be discussed in the future.
etation growth and so on, can drought be monitored and predicted more accurately. Therefore, the SVM model driven by the combined meteorological and remote sensing data is better for drought monitoring in the Guanzhong Area.
ACKNOWLEDGEMENT This work is supported by National Natural Science
CONCLUSIONS Based on the calculation of the SPEI series of 11 meteorological stations in the Guanzhong Area, the applicability of the SPEI to the drought characterization in the Guanzhong Area was analyzed. Three different models were used to predict the drought grades, and the main conclusions are as follows: (1) The identified drought events based on SPEI-1 were in line with the recorded droughts, suggesting that the SPEI-1 data can reasonably reflect the occurrence of drought events in the Guanzhong Area. (2) The SVM model (linear kernel function) performs the best, while the ARIMA model performs the worst in
Foundation of China (No. 51479130).
REFERENCES Ahmad, S., Kalra, A. & Stephen, H. Estimating soil moisture using remote sensing data: a machine learning approach. Advances in Water Resources 33 (1), 69–80. Altman, N. S. An introduction to kernel and nearest-neighbor non-parametric regression. American Statistician 46 (3), 175–185. Bai, Z., Zhang, L., Wang, J. & Huang, Y. Research on forecast of meteorological drought in Yunnan based on ARIMA Model. Yangtze River 46 (15), 6–9. Bannari, A., Morin, D., Bonn, F. & Huete, A. R. A review of vegetation indices. Remote Sensing Reviews 13 (1–2), 95–120.
956
J. Li et al.
|
Drought prediction by meteorological and remote sensing data
Bonaccorso, B., Bordi, I., Cancelliere, A., Rossi, G. & Sutera, A. Spatial variability of drought: an analysis of the SPI in Sicily. Water Resources Management 17, 273–296. Box, G. E. & Jenkins, G. M. Time Series Analysis Forecasting and Control. Holden-Day, San Francisco. Breiman, L. Random forests. Machine Learning 45 (1), 5–32. Burges, C. J. A tutorial on support vector machines for pattern recognition. Data Mining and Knowledge Discovery 2, 127–167. Cai, X., Ye, D., Li, Q., Zhang, C. & Wang, N. Temporal and spatial variation characteristics of drought in Shaanxi based on CI index. Agricultural Research in the Arid Areas 31 (5), 1–8. Che, S. & Li, C. Time and space characteristics of drought in Shijiazhuang based on standardized precipitation index. Meteorological Science and Technology 38 (1), 66–70. Chen, T., De, J. R. A. M., Liu, Y., Van, D. W. G. R. & Dolman, A. J. Using satellite based soil moisture to quantify the water driven variability in NDVI: a case study over mainland Australia. Ecological Informatics 5 (5), 400–409. Cong, L. Analysis of Drought Evolution Characteristics and Drought Prediction in Semi-Arid Areas. Master’s Thesis, Shenyang Agricultural University, Shenyang. Dai, M., Huang, S. Z., Huang, Q., Leng, G. Y., Guo, Y., Wang, L. & Zheng, X. D. Assessing agricultural drought risk and its dynamic evolution characteristics. Agricultural Water Management 231, 106003. Di, L. P., Rundquist, D. C. & Han, L. H. Modeling relationships between NDVI and precipitation during vegetative growth cycles. International Journal of Remote Sensing 15 (10), 2121–2136. Dong, Q. & Xie, P. Progress in hydrological drought research. Journal of China Hydrology 34 (4), 1–7. Fan, G., Zhang, Y., Liu, M. & Mao, Y. Research on drought prediction based on support vector machine. Chinese Journal of Agrometeorology 32 (3), 475–478. Fang, K., Wu, J., Zhu, J. & Xie, B. A review of random forest methods. Statistics & Information Forum 26 (3), 32–38. Fang, Q., Wang, G., Xue, B., Liu, T. & Kiem, A. How and to what extent does precipitation on multi-temporal scales and soil moisture at different depths determine carbon flux responses in a water-limited grassland ecosystem? Science of the Total Environment 635, 1255–1266. Fang, Q., Wang, G., Liu, T., Xue, B., Sun, W. & Shrestha, S. Unraveling the sensitivity and nonlinear response of water use efficiency to the water-energy balance and underlying surface condition in a semiarid basin. Science of the Total Environment 699, 134405. Farrar, T. J., Nicholson, S. E. & Lare, A. R. The influence of soil type on the relationships between NDVI, rainfall, and soil moisture in semiarid Botswana.I. NDVI response to soil moisture. Remote Sensing of Environment 50 (2), 121–133. Gao, P. Precipitation Forecast in Guanzhong Basin and Its Application in Drought Research. Master’s Thesis, Chang’an University, Xi’an.
Hydrology Research
|
51.5
|
2020
Ghashghaie, M. & Nozari, H. Effect of dam construction on Lake Urmia: time series analysis of water level via ARIMA. Journal of Agricultural Science and Technology 20 (S), 1541–1553. Gislason, P. O., Benediktsson, J. A. & Sveinsson, J. R. Random forests for land cover classification. Pattern Recognition Letters 27 (4), 294–300. Guo, Y., Huang, S. Z., Huang, Q., Leng, G. Y., Fang, W., Wang, L. & Wang, H. Propagation thresholds of meteorological drought for triggering hydrological drought at various levels. Science of the Total Environment 712, 136502. Han, P., Wang, P. X., Tian, M., Zhang, Y., Liu, J. & Zhu, D. Application of the ARIMA models in drought forecasting using the standardized precipitation index. In: International Conference on Computer and Computing Technologies in Agriculture. Springer, Berlin, Heidelberg, pp. 352–358. Han, D., Wang, G., Liu, T., Xue, B. L., Kuczera, G. & Xu, X. Hydroclimatic response of evapotranspiration partitioning to prolonged droughts in semiarid grassland. Journal of Hydrology 563, 766–777. Hernandez, E. & Annette, U. V. Standardized precipitation evaporation index (SPEI)-based drought assessment in semiarid south Texas. Environmental Earth Sciences 71 (6), 2491–2501. Ichii, K., Hashimoto, H., Nemani, R. & White, M. Modeling the interannual variability and trends in gross and net primary productivity of tropical forests from 1982 to 1999. Global and Planetary Change 48 (4), 286. Kogan, F. N. Remote sensing of weather impacts on vegetation in nonhomogeneous areas. International Journal of Remote Sensing 11 (8), 1405–1420. Kogan, F. N. Droughts of the late 1980s in the United States as derived from NOAA polar-orbiting satellite data. Bulletin of the American Meteorological Society 76 (5), 655–668. Lei, X., Li, Q., Wang, J., Li, H. & Li, H. The drought and flood evolution law in the Guanzhong Area and the analyses of the present century’s drought and flood characteristics. Journal of Catastrophology 31 (3), 101–109. Li, J., Mo, S., Shen, B., Si, H. & Wang, Y. Analysis of drought characteristics in Weihe River Basin based on SPEI. Journal of Xi’an University of Technology 32 (1), 70–76. Liu, X. Natural Disasters and Preventive Measures in the Northwest Region Since the Founding of the People’s Republic (1994–2000). Master’s Thesis, Tianjin University of Commerce, Tianjin. Liu, W., An, S., Liu, G. & Guo, A. Further correction of Palmer’s drought pattern. Journal of Applied Meteorological Science 15 (2), 207–215. Lu, J. Temporal and spatial variation characteristics of drought in Yungui area from 1960 to 2014 based on SPEI and run-length theory. Journal of Zhejiang University (Science Edition) 45 (3), 363–372. Malik, A., Kumar, A. & Singh, R. P. Application of heuristic approaches for prediction of hydrological drought using multi-scalar streamflow drought index. Water Resources Management 33 (11), 3985–4006.
957
J. Li et al.
|
Drought prediction by meteorological and remote sensing data
Mazdiyasni, O. & Aghakouchak, A. Substantial increase in concurrent droughts and heatwaves in the United States. Proceedings of the National Academy of Sciences 112 (37), 11484–11489. Mckee, T. B., Doesken, N. J. & Kleist, J. The relationship of drought frequency and duration to time scales. In: Proceedings of the 8th Conference on Applied Climatology, January 17–22 Anaheim, pp. 179–184. Mei, X. Temporal and Spatial Distribution of Actual Evapotranspiration in Guanzhong Agricultural Region of Shaanxi Province. Master’s Thesis, Xi’an University of Technology, Xi’an. Meteorological Station of Shaanxi Meteorological Bureau Record of Natural Disasters in the Shaanxi Province. Shaanxi Meteorological Bureau, Xi’an. Miao, Z. Analysis and Prediction of Meteorological Drought Characteristics in Baojixia Irrigation District. Master’s Thesis, Northwest A&F University, Yangling. Moody, J. & Darken, C. J. Fast learning in networks of locally tuned processing units. Neural Computation 1 (2), 281–294. Palmer, W. C. Meteorological Drought. Research Paper No.45 US Department of Commerce Weather Bureau, Washington, DC. Qiao, L. Study on Drought Early Warning and Emergency Water Source Allocation in Guanzhong Area. Doctoral Dissertation, Chang’an University, Xi’an. Sahoo, A. K., Sheffield, J., Pan, M. & Wood, E. F. Evaluation of the tropical rainfall measuring mission multi-satellite precipitation analysis (TMPA) for assessment of large-scale meteorological drought. Remote Sensing of Environment 159, 181–193. Shumway, R. H. & Stoffer, D. S. An approach to time series smoothing and forecasting using the EM algorithm. Journal of Time Series Analysis 3 (4), 253–264. Sun, Q. Temporal and Spatial Variation Characteristics of Drought in China From 1981 to 2015 Based on VCI Index. Master’s Thesis, Jiangsu Normal University, Nanjing. Vega, H. M., Lima, J. R. & Cerniak, S. N. SPEI and Hurst analysis of precipitation in the Amazonian Area of Brazil. Revista Brasileira de Meteorologia 34 (2), 325–334. Vicente-Serrano, S. M., Beguer, A. S. & Lpez-Moreno, J. I. A multi-scalar drought index sensitive to global warming: the standardized precipitation evapo-transpiration index. Journal of Climate 23 (7), 1696–1718. Wang, Z. Big Data Mining and Application. Tsinghua University Press, Beijing. Wang, L. & Chen, W. Applicability analysis of standardized precipitation evapotranspiration index in drought monitoring in China. Plateau Meteorology 33 (2), 423–431. Wei, Z., Chen, S. & Huang, Y. Temporal and spatial characteristics of potential evapotranspiration in Shaanxi from 1981 to 2000 and its response to climate factors. Scientia Geographica Sinica 35 (8), 1033–1041. Weng, Q. Remote sensing of impervious surfaces in the urban areas: requirements, methods, and trends. Remoting Sensing of Environment 117 (2), 34–49.
Hydrology Research
|
51.5
|
2020
Wu, J. SPSS Statistical Analysis Starts From Scratch. Tsinghua University Press, Beijing. Wu, H. Study on the Characteristics of Water Resources Change and Drought Vulnerability in Guanzhong Plain. Doctoral Dissertation, Chang’an University, Xi’an. Wu, J., Chen, Y. & Yu, S. Drought prediction based on random forest model. China Rural Water and Hydropower 11, 17–22. Xu, H. Analysis of the Characteristics of Drought Time and Space Evolution and Vulnerability Assessment in Shaanxi Province. Doctoral Dissertation, Chang’an University, Xi’an. Yan, S., Lu, Q., Zhang, J., Zhang, Z. & Bai, S. Evolution characteristics of NDVI in coastal areas of Jiangsu Province and its response to regional climate change. Journal of Nanjing Forestry University (Natural Sciences Edition) 36 (01), 43–47. Yang, S. Y., Meng, D., Li, X. J. & Wu, X. L. Multi scale response of vegetation change to SPEI meteorological drought index in North China from 2001 to 2014. Journal of Ecology 38 (3), 1028–1039. Yilmaz, M. T., Hunt, E. R. & Jackson, T. J. Remote sensing of vegetation water content from equivalent water thickness using satellite imagery. Remote Sensing of Environment 112 (5), 2514–2522. Yinglan, A., Wang, G., Liu, T., Shrestha, S., Xue, B. L. & Tan, Z. a Vertical variations of soil water and its controlling factors based on the structural equation model in a semi-arid grassland. Science of the Total Environment 691, 1016–1026. Yinglan, A., Wang, G., Liu, T., Xue, B. L. & Kuczera, G. b Spatial variation of correlations between vertical soil water and evapotranspiration and their controlling factors in a semi-arid region. Journal of Hydrology 574, 53–63. Yu, Y. Comparison of Drought Monitoring and Drought Time and Space Characteristics in Guanzhong Region Based on CWSI and TVDI. Master’s Thesis, Shaanxi Normal University, Xi’an. Yuan, X. Study on Establishment of Hydrological Drought Index and Calculation Method of Drought Frequency. Master’s Thesis, Taiyuan University of Technology, Taiyuan. Yuan, W. & Zhou, G. Theoretical analysis and research prospects of drought index. Advance in Earth Sciences 19 (6), 982–991. Yurekli, K., Kurune, A. & Ozturk, F. Application of linear stochastic models to monthly flow data of Kelkit steam. Ecological Modeling 183, 67–75. Zargar, A., Sadiq, R., Naser, B. & Khan, F. I. A review of drought indices. Environmental Reviews 19 (17), 333–349. Zhang, Y., Hao, Z., Wang, Y., Li, M. & Chen, E. The relationship between multi-scale drought characteristics and climate index of Taiyuan based on SPEI and SPI index. Ecology and Environmental Sciences 23 (9), 1418–1424. Zhang, B., Wang, S. & Wang, Y. a Copula-based convectionpermitting projections of future changes in multivariate
958
J. Li et al.
|
Drought prediction by meteorological and remote sensing data
drought characteristics. Journal of Geophysical Research– Atmospheres 124 (14), 7460–7483. Zhang, X., Chen, N., Sheng, H., Ip, C., Yang, L. & Chen, Y. b Urban drought challenge to 2030 sustainable development goals. Science of the Total Environment 693, 133536. Zhang, Y., Li, W., Chen, Q., Pu, X. & Xiang, L. Multi-models for SPI drought forecasting in the north of Haihe River Basin, China. Stochastic Environmental Research and Risk Assessment 31 (10), 1–11. Zhao, X., Zhu, N. & Huang, L. Time series modeling algorithm and empirical analysis based on ARIMA model.
Hydrology Research
|
51.5
|
2020
Journal of Guilin University of Electronic Technology 32 (5), 410–415. Zhou, X. J., Peng, X. W., Tansey, K., Zhang, S. Y., Li, H. M. & Wang, L. Developing a fused vegetation temperature condition index for drought monitoring at field scales using Sentinel-2 and MODIS imagery. Computers and Electronics in Agriculture 168, 105144. Zhuang, S., Zuo, H., Ren, P., Xiong, G. & Li, B. Application of standardized precipitation evaporation index in China. Climatic and Environmental Research 18 (5), 617–625.
First received 24 December 2019; accepted in revised form 29 March 2020. Available online 23 June 2020
959
© 2020 The Authors Hydrology Research
|
51.5
|
2020
The influences of sponge city construction on spring discharge in Jinan city of China Kangning Sun, Litang Hu and Xiaomeng Liu
ABSTRACT In recent years, intense human activities have threatened to dry up the well-known karst springs in Jinan, China. Sponge city construction program was one of the measures aiming to improve the recharge to groundwater and also protect sources of spring discharge. An influence study of sponge city construction on groundwater is necessary while not fully evaluated. In this paper, a threedimensional numerical groundwater flow model was developed to address this issue. Model calibration showed that the simulated groundwater level successfully reproduced the observed results. Then, 12 scenarios were established and predicted according to different precipitation conditions and the achieved degrees of sponge city construction. The results indicated that the sponge city construction was conducive to the rise of regional groundwater level after 20 years. However, the groundwater level around the spring groups would only increase by an average of 0.22 m, and the annual spring discharge would increase by approximately 9.00 million m3 after 20 years. Results revealed that the extent of spring discharge recovery was not evident in a short time frame. The proper positioning of sponge city construction was suggested to be considered further to
Kangning Sun Litang Hu (corresponding author) College of Water Sciences, Beijing Normal University, Beijing 100875, China; Beijing Key Laboratory of Urban Hydrological Cycle and Sponge City Technology, Beijing 100875, China; and Engineering Research Center of Groundwater Pollution Control and Remediation of Ministry of Education, Beijing 100875, China E-mail: litanghu@bnu.edu.cn Xiaomeng Liu The Government of Beishicao Town Shunyi District, Beijing 101300, China
balance the protection of springs with the issue of waterlogging. Key words
| fractured-karst aquifer, Jinan springs, numerical simulation, sponge city
INTRODUCTION Karst areas account for approximately 7–12% of the total
southern and northern regions of China, including the
land area of the Earth; however, these areas provide more
Yunnan, Guizhou, Sichuan, Chongqing, Guangxi, Hunan,
than 25% of the drinking water for the world’s population
and Hubei regions, which comprise a total area of approxi-
(Ford & Williams ; Wang et al. ). Owing to its
mately 1.76 × 106 km2 (Lu ).
rich reserves, large water yield, extensive distribution, and
Jinan, the capital city of Shandong Province in China, is
good water quality, karst water has become an especially
famous for its numerous karst springs, including four big
important groundwater resource for the world (Bakalowicz
spring groups called Baotu, Black Tiger, Five Dragon, and
). Besides, karst springs are also the sources for local
Pearl springs. However, sustained and rapid economic
rivers and lakes which are important for local ecosystems.
development that occurred in the city from 1975 to 2003
Additionally, the unique topography of karst areas is a sig-
caused a sharp increase in groundwater exploitation, result-
nificant source of tourism for the development of local
ing in continuous groundwater level decline and the cut off
economies. Many karst catchment areas are present in
of springs. For example, the longest continuous cutoff period
This is an Open Access article distributed under the terms of the Creative
). The Chinese government closely monitored the
of Baotu spring was 926 days from 1999 to 2001 (Qian et al. Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited
spring cutoff problem and conducted research on the restor-
(http://creativecommons.org/licenses/by/4.0/).
ation of the groundwater level and the protection of local
doi: 10.2166/nh.2020.008
960
K. Sun et al.
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
springs. Furthermore, various measures were taken to allevi-
() and Wang et al. (), aiming at estimation of the
ate severe situations. For example, the utilization of
causes of groundwater level decline and the drying up of
groundwater was partially replaced by surface water (such
springs. These investigations and studies provide good refer-
as water from the Yellow and Yangtze Rivers), many pump-
ences for our further analysis. To our knowledge, the
ing wells were shut down in spring catchment areas,
influence of sponge city construction on groundwater flow
artificial groundwater recharge increased, and the infiltra-
has not been fully examined, especially the short-term and
tion
the
long-term influences of the project on groundwater level
implementation of these measures, the springs have been
and spring discharge. The purpose of this study was to
flowing since 2003; however, due to low water levels and
develop a numerical model to evaluate the influence of
flow, the springs are still at risk.
sponge city construction on the regional groundwater level
of
surface
water
also
increased.
Since
The sponge city program was announced by the Chinese
and spring discharge in consideration of the implementation
government in 2013 to address the problem of urban water-
of sponge city construction in Jinan. First, we established a
logging. This approach emphasized the use of natural
three-dimensional (3D) numerical model of groundwater
systems (e.g., changing the conditions of the land surface
flow. Then, the model was calibrated based on the esti-
for precipitation infiltration, control, and storage). Orig-
mation of the parameters, and the factors controlling the
inally, the term ‘sponge city’ meant a city could function
groundwater discharge were analyzed. Finally, scenarios
like a sponge, demonstrating great resilience to environ-
portraying different construction and precipitation con-
mental changes and natural disasters. This approach was
ditions
were
prepared
to
predict
changes
in
the
similar to the best management practices (Ice ) and
groundwater level and spring discharge. This study will pro-
low impact development system of the United States (Xu
vide
et al. ), the water-sensitive urban design of Australia
construction and the protection of springs in Jinan and simi-
(Ahammed ), and the sustainable urban drainage
lar areas.
theoretical
reference
for
further
sponge
city
system of the UK (Ellis & Lundy ). Today, the concept of a sponge city has evolved into a modern stormwater management approach to help solve drainage issues, fully utilize
STUDY AREA
land resources, and promote sustainable development. From 2015 to 2016, two pilot batches of sponge cities (30 cities in
Background
total) were supported by the Chinese Central Government, and the total investment was about 42.3 billion RMB ( Jia
The study area (Figure 1) is located in Jinan, the capital city
et al. ). Jinan was among the first batch of pilot sponge
of Shandong Province, with an area of approximately
cities in China in 2015, the purposes of which were to
1,500 km2 and coordinates of approximately 36 280 –
increase precipitation infiltration, raise the groundwater
36 460 N, 116 400 –117 140 E. Its eastern and western borders
level, and to influence the springs.
are the Dongwu and Mashan Faults, respectively. The
Since the 1950s, many field investigations and studies
southern boundary is the groundwater watershed, and the
regarding aquifer properties, origin of springs, groundwater
northern boundary the zone of contact metamorphism and
quantity and quality have been carried out in Jinan using
magmatism. The elevation gradually decreases from south
borehole logging, long-term observation of groundwater
to north. The northern area is a piedmont alluvial plain
level and spring flow, tracer test and numerical simulation.
and an intermountain plain with an elevation of approxi-
The earliest two-dimensional groundwater flow model with
mately 20 m–50 m. The urban district is mainly distributed
one-year calibration was established in 1989 to explore the
in this area. To the south is a continuously low and hilly
balance of groundwater supply and spring protection, and
area with an elevation just over 200 m. Further south is a
then over five groundwater flow models were progressively
steep lower-middle mountainous area with a maximum
constructed (Wang et al. ). Two typical three-dimen-
elevation of 800 m. The study area has a temperate continen-
sional groundwater models were developed by Qian et al.
tal climate. The annual average amount of precipitation
961
Figure 1
K. Sun et al.
|
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
Location and geology map of study area.
there is 670.47 mm, while the annual average amount of
(P) strata are buried and exposed only slightly in the north-
evaporation is approximately 1,475.6 mm. Moreover, pre-
west. The Quaternary (Q) strata is well exposed in the
cipitation is mainly concentrated during the months of
northern area. Several large faults are distributed through-
June to September, which accounts for 75% of the total
out the study area in northwest and northeast directions.
annual precipitation. The four main rivers in the study
The Dongwu Fault is impermeable in most areas, demon-
area are the Yellow, Xiaoqing, Dashahe, and Yufu Rivers.
strating weak permeability only in the northeastern part of the spring area; the Mashan Fault is impermeable in the
Geology and hydrogeology
southern mountainous area and permeable in the northern area. The Qianfoshan Fault is in the central part of the
As shown in Figure 1, the Archean Taishan Group (Art),
Jinan spring area and is impermeable in the south and per-
consisting of metamorphic rocks, is distributed in the
meable in the north. The remaining small faults are all
southern area and is the basement of the study area. The
permeable.
Cambrian (∈) strata is characterized by interbeds of lime-
Pore water in a Quaternary aquifer and fractured-karst
stone and shale, which are well exposed from south to
groundwater in an Ordovician aquifer are the main types
north. The Ordovician (O) strata is composed of thick-
of groundwater in the study area, which account for more
bedded limestone and mainly distributed in the middle of
than 80% of the total groundwater discharge, especially
the study area. Most of the Carboniferous (C) and Permian
the karst aquifer, which is the major recharge resource for
962
K. Sun et al.
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
the springs. Precipitation recharges the groundwater mainly
exploitation was transferred from urban areas to suburban
through surface fissures that have developed in the meta-
areas, and the exploitation decreased suddenly from 30 ×
morphic rock and limestone in the southern and central
104 m3/d to 12 × 104 m3/d in the urban areas. However,
areas, which are the main recharge sources for groundwater
the groundwater level and spring discharge were still
(Figure 2). Surface water and irrigation infiltration are also
reduced, and the springs continued to dry up. The last
important components for groundwater recharge. Recently,
stage was the period that began post-2001, after which
artificial recharge has become another important recharge
more effective measures were taken by the Jinan govern-
source. The karst groundwater in the study area is mainly
ment to protect the springs. For example, water from the
discharged through groundwater exploitation, springs, and
Yellow and Yangtze Rivers and the Wohushan and Yuqin-
recharge of the Quaternary pore water, while the pore
ghu reservoirs was utilized, and the groundwater was
water is mainly discharged through evaporation, springs,
gradually replaced by these surface waters in the water
and groundwater exploitation.
supply system. At the same time, artificial recharge of the groundwater system was conducted. The implementation
Groundwater utilization
of these measures has ensured continuous flowing of the springs since 2003. Furthermore, the groundwater level
Groundwater utilization can be classified into four stages (Figure 3). During the first stage (pre-1965), groundwater exploitation was limited, and precipitation was the main factor affecting the groundwater regime, and so the highest
began to rise, and spring discharge gradually increased. However, the springs still face the risk of drying up and are in a low water level state, especially from April to June each year.
spring discharge reached 51 × 104 m3/d in 1962. The second stage was the period from 1965 to 1980. Groundwater exploitation gradually increased from 10 × 104 m3/d
METHODS AND DATA
in 1965 to approximately 30 × 104 m3/d in 1980 in the urban district. The total groundwater exploitation in the
Simulation method
spring area reached 60 × 104 m3/d in 1972 and then increased to 70 × 104 m3/d by 1980. Due to the overexploita-
Equivalent porous media method
tion of groundwater, the springs have begun to dry up during each dry season since 1972, and the number of dried-up days
Flow in a fractured-karst aquifer is usually a non-Darcy flow.
increased correspondingly. The third stage was the period
Flow simulation in a fractured-karst medium is challenging.
from 1981 to 2000. For spring protection, groundwater
According to the condition of whether physical processes
Figure 2
|
Schematic diagram of genesis of Jinan karst spring (Wang et al. 2014).
K. Sun et al.
963
Figure 3
|
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
Changes in precipitation, spring discharge, and groundwater exploitation in urban district; average groundwater level and number of dried-up days of Baotu spring from 1958 to 2010.
are considered, these models can roughly be classified into
spatial difference, and it only measures the structural
two categories (Hartmann et al. ; Meng et al. ).
relationship between inputs and outputs. Some examples
The first category is a physically based model. This type of
of a lumped model are the commonly used artificial neural
model usually applies a finite differential method or a
network model (Meng et al. ), linear or nonlinear reser-
finite element method to discretize a fractured-karst aquifer
voir model (Padilla & Pulido-Bosch ), wavelet analysis
into several small grids, where different hydrogeological par-
(Labat et al. ), and regression analysis (Felton & Currens
ameters and boundary, recharge, and discharge conditions
). Lumped models cannot reflect the hydraulic par-
are given. Owing to the high heterogeneity and anisotropy
ameters of an aquifer, the flow direction of groundwater,
of karst aquifer media, it is difficult to accurately character-
or the velocity of groundwater flow.
ize the hydrogeological parameters of an aquifer; thus,
The geology structure in the study area is complex, and
aquifers are often generalized. Many simplification methods
the aquifer contains karst, fissures, and porous aquifers with
have been proposed, such as the equivalent porous media
different pore sizes, which are difficult to generalize with
model (Scanlon et al. ; Dragoni et al. ), double
one another. In this study, a convenient and efficient equiv-
media model (Robineau et al. ), triple media model
alent medium simulation method was chosen, which is
(Chen & Hu ), discrete fractured network model
widely used at present in physically based models. In par-
(Dverstorp et al. ), and discrete conduit network
ticular, when the main aquifer media are dissolution pores
model (Ghasemizadeh et al. ). The second category is
rather than caves and channels, this kind of model can be
a lumped model; this type of model takes a karst ground-
used to simulate water balances and trends of the regional
water system as a unified entity without considering
groundwater flow (Scanlon et al. ). Therefore, this
964
K. Sun et al.
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
method could be employed in our study area. Likewise, it
permeable. The Qianfoshan Fault was represented as a
has been successfully applied in most of the karst aquifer
weakly permeable boundary in the southern section and a
systems in northern China (Kang et al. ).
permeable boundary in the northern part. The bottom boundary of the model was set as the bottom of the upper
Simulation program
Quaternary overburden and the lower Cambrian, and the total thickness of the model was approximately 500–
This study chose a polygon grid finite difference groundwater
600 m. There were three vertical model layers. The first
modeling system (named PGMS) to establish a 3D ground-
layer mainly consisted of Quaternary, Cambrian, and Ordo-
water flow numerical model. The PGMS was based on a
vician strata, and the thickness of this layer was
polygonal grid, which was more convenient, flexible, and
approximately
accessible and had more advantages than a rectangular
included upper Cambrian strata with a thickness of approxi-
grid. Moreover, the PGMS could deal well with problems
mately 58–100 m. The third layer mainly consisted of
associated with karst groundwater and karst springs, and it
Middle to Lower Cambrian strata, and the thickness was
was applied in the Heihe River Basin (Hu et al. ) with
approximately 170–470 m. Among them, the first and third
good results. In addition, the authors have the source code
layers were major aquifers. The model was represented as
of the PGMS, which could be further improved based on
a heterogeneous anisotropy with a 3D spatial structure
the need for other numerical simulations.
and transient groundwater flow.
Uncertainty
analysis
software
called
185–380 m.
The
second
layer
mainly
Uncertainty
Quantification Python Laboratory (UQ-PyL) (Wang et al.
Data preparation
) was chosen for parameter optimization. UQ-PyL coupled with the PGMS through ‘control files’ (‘template’
The main data sources concerning groundwater level and
and ‘driver’ files). The template file was a parameter control
spring discharge are listed in Table 1 and include topogra-
file that wrote the parameters in a separate file, and the
phy, precipitation, hydrogeology, observation wells, spring
model file generated by the PGMS needed this file to read
discharge, and many research reports (Wang et al. ).
the parameters automatically. The driver file contained all
Sources and sinks were processed in this study. The main
the adjustable parameters. The model parameter settings
recharge sources included were infiltration from precipi-
could be changed in this file, and the results of the model
tation, rivers, and artificial recharge, and the return flow of
operation could also be processed.
irrigation water. The main rivers studied were the Yufu,
Conceptual model and data preparation
Table 1
Conceptual model
|
Lists of data sources used in the model
Data items
Descriptions
Topography
Spatial resolution of 16 m and over the study area, and resolution with 0.5 m on the northeast of spring area
Precipitation
Daily precipitation data from 2014 to 2016 at 9 stations, and annual precipitation data from 1956 to 2017 at 20 stations
Hydrogeology
Hydrogeological maps and borehole descriptions from previous reports (Wang et al. ), 10 hydrogeological cross sections
Observation wells
Annual groundwater level data from 1956 to 2017 at 100 wells and monthly water level data from 2014 to 2016 at 17 wells
Spring discharge
Annual springs discharge from 1958 to 2010 and monthly data from 2014 to 2017
The Jinan spring catchment was selected as the model area. The boundary conditions were set according to previous studies (Wang et al. ). The northern boundary of the model area was regarded as a no-flow boundary, but a permeable boundary was present where the Yellow River flows. The southern boundary was generalized as an impermeable boundary. The eastern boundary was represented as a no-flow boundary, except the northeastern section of the spring area, which was a weakly permeable boundary. The western boundary was represented as a noflow boundary, except the northern plains area, which was
965
K. Sun et al.
|
Influences of sponge city construction on spring discharge
Beidasha, Yellow, and Xiaoqing Rivers. Artificial ground-
Hydrology Research
|
51.5
|
2020
Model discretization and calibration
water recharge was primarily from two strong seepage zones (reaches of the Yufu and Xingji Rivers). The area of irrigation was determined by remote sensing images.
Model discretization and zonation of hydrogeological parameters
Groundwater evaporation was calculated from a model based on the depth of groundwater, soil types, and given
The study area was divided into three layers, and each layer
limits (Hu et al. ). Groundwater exploitation occurred
was divided into 6,488 auxiliary triangles for a total of
in local regions. Spring discharge was calculated based on
19,464 auxiliary triangles (Figure 4). The period of model
the groundwater level and the thickness and permeability
calibration was from January 2014 to December 2016. The
of underlying media, which included Baotu, Black Tiger,
time step was set as one month; therefore, there were a
Five Dragon, and Pearl springs. The critical water level
total of 36 time steps. The trial-and-error method was used
elevation of the spring groups was set according to the
to calibrate the model against observed data to achieve the
actual outflow elevation. The outcrops of Baotu, Black
smallest possible objection function, which was defined as
Tiger, Five Dragon, and Pearl springs were 27.01 m,
the sum of the squares of the differences between the
27.30 m, 26.20 m, and 26.77 m, respectively.
observed and calculated results, including both the water
Figure 4
|
Three-dimensional (3D) network model sketch of research area.
K. Sun et al.
966
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
level and spring discharge. Then, UQ-PyL software was used
field and spring discharge (Serial Zones 16 and 31,
to optimize the parameters.
Table 2). The zonation of the three aquifers is presented
Many aquifer parameters, including horizontal (Kxx and Kyy) and vertical (Kzz) hydraulic conductivities,
in Figure 5, and the optimized parameters are listed in Table 2.
specific yield (Sy), and specific storage (Ss), were involved due to the complexity of the aquifer system (Kzz ¼ ½ Kxx, Kyy ¼ Kxx). Based on the obtained borehole data, the hydrogeological
parameters
Comparison of observed and simulated groundwater level data
of the three simulation
layers were subdivided into 71 zones. The hydrogeological
Calibration targets were calculated as the root mean
parameters were initially assigned, and then they were
square of the difference between the observed and simu-
adjusted and fixed after the model was calibrated. Consid-
lated groundwater levels (Δh). There was a total of 17
ering the changes of the hydrogeological parameters
observation wells (13 in the plains area and four in the
caused by sponge city construction, the first layer of the
mountainous area), and the observation data were col-
sponge city pilot area was subdivided separately to
lected from 2014 to 2016. The results (Table 3)
better simulate the changes in the groundwater flow
indicated that the number of absolute errors between
Table 2
|
Lists of calibrated and optimized hydrogeological parameters
Serial zone
Kxx (m/d)
Ss (10
1
0.02
2
0.02
3
4
1
Sy
Serial zone
Kxx (m/d)
Ss (10
0.40
0.10
25
40.00
0.60
0.10
26
20.00
0.05
0.60
0.20
27
4
0.03
0.40
0.20
5
0.06
1.00
0.20
6
0.05
0.60
7
0.04
0.60
8
0.02
9
0.05
10
4
1
4
1
Sy
Serial zone
Kxx (m/d)
Ss (10
7.00
0.05
49
10.00
0.10
0.01
5.00
0.05
50
5.00
0.10
0.01
80.00
5.00
0.05
51
0.10
0.20
0.05
28
70.00
7.00
0.05
52
0.05
0.50
0.10
29
30.00
7.00
0.05
53
0.10
0.20
0.15
0.25
30
1.00
1.00
0.01
54
0.10
0.10
0.15
0.20
31
10.00
7.00
0.01
55
0.10
0.20
0.10
0.60
0.001
32
50.00
0.50
0.005
56
0.05
1.00
0.10
0.60
0.20
33
5.00
1.00
0.05
57
0.01
1.00
0.05
0.05
0.60
0.20
34
40.00
0.10
0.001
58
0.01
0.50
0.10
11
0.01
1.00
0.05
35
30.00
0.10
0.005
59
5.00
2.00
0.01
12
0.008
1.00
0.05
36
0.10
1.00
0.01
60
5.00
2.00
0.05
13
50.00
7.00
0.05
37
0.10
1.00
0.05
61
0.10
0.50
0.01
14
30.00
7.00
0.01
38
1.00
0.50
0.10
62
0.10
0.20
0.01
15
30.00
4.00
0.10
39
50.00
5.00
0.10
63
0.10
1.00
0.01
16
30.00
4.00
0.005
40
1.00
1.00
0.10
64
0.10
2.00
0.05
m
)
m
)
m
)
Sy
17
10.00
4.00
0.20
41
10.00
5.00
0.10
65
0.10
2.00
0.05
18
40.00
7.00
0.05
42
10.00
5.00
0.10
66
40.00
1.00
0.01
19
30.00
7.00
0.04
43
3.00
2.00
0.10
67
40.00
1.00
0.01
20
30.00
7.00
0.10
44
10.00
5.00
0.10
68
40.00
1.00
0.10
21
5.00
1.00
0.20
45
3.00
1.00
0.10
69
40.00
1.00
0.05
22
4.00
1.00
0.10
46
1.00
1.00
0.10
70
0.10
0.50
0.01
23
1.00
1.00
0.10
47
10.00
0.50
0.10
71
0.10
0.20
0.10
24
5.00
7.00
0.05
48
10.00
0.50
0.05
K. Sun et al.
967
Figure 5
Table 3
|
|
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
Parameter zonation of three layers.
Statistics of absolute errors and percentages for 17 observation wells
Absolute errors (m)
Δh 0.5
0.5<Δh 1.0
1.0<Δh 3.0
Δh>3.0
Total
All observation holes
Frequency of occurrence Percentage (%)
146 27.29
107 20.00
170 31.78
112 20.93
535 100
Observation holes in plain area
Frequency of occurrence Percentage (%)
131 33.08
98 24.75
135 34.09
32 8.08
396 100
Observation holes in mountain area
Frequency of occurrence Percentage (%)
15 10.79
9 6.48
35 25.18
80 57.54
139 100
the simulated and observed values were less than 0.5 m,
Groundwater budget analysis
1.0 m, and 3.0 m, which accounted for 27%, 47%, and 79%, respectively. The relative errors within 5%, 10%,
There were four springs in total, and observation data were
and 20% accounted for 70%, 85%, and 94%, respectively.
collected from 2014 to 2016. The observed average annual
The goodness-of-fit in the plains area was higher than
spring discharge for Pearl, Five Dragon, Baotu, and Black
that of the mountainous area. If the observation holes
Tiger springs was 0.06, 0.12, 0.18, and 0.16 × 108 m3, respect-
in the plains area were solely analyzed, the absolute
ively. The simulated average annual spring discharge for
errors were less than 0.5 m, 1.0 m, and 3.0 m, accounting
Pearl, Five Dragon, Baotu, and Black Tiger springs was
for 33%, 58%, and 92%, respectively; the relative errors
0.07, 0.14, 0.22, and 0.17 × 108 m3, respectively. For the
within 5%, 10%, and 20% accounted for 77%, 92%,
four springs, the observed and simulated annual total dis-
and 95%, respectively. The fitting accuracy of the obser-
charges were 0.53 and 0.60 × 108 m3, respectively, and the
vation holes in the mountainous area was slightly poor,
absolute error of discharge was approximately 0.07 ×
but the relative errors of the simulated value that were
108 m3 with a relative error (the ratio of the absolute differ-
less than 20% also accounted for 91%, and more than
ence of observed and simulated discharge to the observed
half of them were less than 5%. Figure 6 shows the
results) of approximately 13%. Meanwhile, for the obser-
comparison
observed
vation wells around the four springs (Figure 6(e)), the
groundwater levels (corresponding with observation
simulated water level was consistent with the observed
wells a, b, c, d, e, and f in Figure 1). The results indicated
results.
between
the
simulated
and
that the prediction could adequately reflect the actual groundwater level.
The simulated groundwater budget in the study area from 2014 to 2016 was obtained. The main groundwater
968
Figure 6
K. Sun et al.
|
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
Comparison of observed and simulated groundwater levels: (a), (b), (c), (d), (e), and (f) represent observation wells a, b, c, d, e, and f from Figure 1.
recharge sources included infiltration from precipitation and
and 24% of the total recharge, respectively. From 2014 to
rivers (natural river infiltration and artificial recharge), and
2016, the total annual value of groundwater recharge was
the average recharge values were 3.24 and 0.77 × 108 m3
3.39, 3.93, and 5.21 × 108 m3, respectively, and the total
(accounting for 78% and 19% of the total recharge), respect-
annual value of groundwater discharge was 2.43, 2.49, and
ively. The main groundwater discharge sources included
2.56 × 108 m3, respectively. The changes in groundwater sto-
groundwater exploitation and springs, and the average dis-
rage from 2014 to 2016 were 0.96, 1.44, and 2.65 × 108 m3,
8
3
charge was 1.41 and 0.61 × 10 m , accounting for 57%
respectively, and the groundwater system was always in a
969
K. Sun et al.
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
positive balance. The increase in groundwater storage
and when combined with our own research purposes, this
should have a beneficial impact on the recovery of spring
area was divided into two zones. Zone a was the develop-
discharge.
ment
and
construction
area
with
a
sponge
city
construction target of runoff coefficient and annual runoff control rate of 0.7 and 75%, respectively. Zone b was the
MODEL APPLICATION AND RESULTS Construction of Jinan sponge city
mountainous area with a sponge city construction target of runoff coefficient and annual runoff control rate of 0.4 and 85%, respectively. The total area of the pilot area was 39 km2, including 22 km2 in Zone a and 17 km2 in Zone b.
The sponge city pilot was located in the southeast region of the study area with mountains to the south; the pilot was
Scenarios set
approximately 1.8 km north of the springs (Figure 7). The elevation of the pilot area ranged from 23 m to 460 m. In
The most direct effect of sponge city construction on
this area, the mountains and plains were merged, and the
groundwater was an increase in the infiltration capacity of
plains area (with its low permeability) was primarily utilized
the underlying surface; namely, the degree of sponge city
for development and construction. According to the con-
construction could be represented by a precipitation infiltra-
struction and implementation plans for Jinan sponge city,
tion coefficient. Twelve scenarios (S) were set, as shown in
the planning targets were to be set by each planning area,
Table 4. Scenarios S0–S3 corresponded to the infiltration
Figure 7
|
Sponge city pilot area.
K. Sun et al.
970
Table 4
|
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
Scenarios set for model prediction
Infiltration coefficient scenarios S0
S1
S2
S3
Zone a
Zone b
Zone a
Zone b
Zone a
Zone b
Zone a
Zone b
Precipitation scenarios
0.15
0.30
0.19
0.34
0.23
0.42
0.30
0.60
Dry year
D-S0
D-S1
D-S2
D-S3
Normal year
N-S0
N-S1
N-S2
N-S3
Wet year
W-S0
W-S1
W-S2
W-S3
conditions caused by different degrees of sponge city con-
The increment of the groundwater level under different
struction. Among them, S0 was the scenario when sponge
precipitation conditions in the pilot area was analyzed when
city construction had not yet begun, while S3 was the scen-
sponge city construction was completed according to the
ario when sponge city had been completed according to the
initial target. In successive ‘dry’ years, the areas where the
initial planning target.
groundwater level rose more than 2 m, 4 m, 6 m, and 8 m
Precipitation conditions were simulated according to
accounted for approximately 58%, 25%, 9%, and 3% of the
dry (D), normal (N), and wet (W) years (obtained by the
total area, respectively. In successive ‘normal’ years, the
P-III curve), based on the multi-year precipitation series.
areas where the groundwater level rose more than 2 m,
The amount of precipitation (P) in the wet year (P ¼ 25%),
4 m, 6 m, and 8 m accounted for approximately 71%, 40%,
normal year (P ¼ 50%), and dry year (P ¼ 75%) was
14%, and 7% of the total area, respectively. In successive
782.45 mm, 648.73 mm, and 534.87 mm, respectively. The
‘wet’ years, the areas where the groundwater level rose
precipitation conditions (D, N, or W) corresponded with
more than 2 m, 4 m, 6 m, and 8 m accounted for approxi-
the benchmark scenarios (S), and were represented by
mately 86%, 50%, 22%, and 11% of the total area,
D-S0, N-S0, and W-S0, respectively. The prediction time
respectively. No matter what the precipitation conditions,
frame was 20 years, and the time step was one month.
the groundwater level in most of the pilot area could rise more than 2 m after 20 years. Sponge city construction
Results analysis Regional water level distribution Under different precipitation conditions, the groundwater
generally raised the groundwater level of the pilot area and its surroundings, and the higher the degree of sponge city construction and the more precipitation there was, the larger the increment, and the wider the range of increment.
level changed in scenarios S1, S2, and S3, in contrast to that of scenario S0 (Figure 8). No matter what the scenario
Change in groundwater level around springs
was, it was demonstrated that the most rapid and largest increase of groundwater level was in the mountainous
We took observation well e as an example to demonstrate
area, and it gradually expanded to the surrounding areas
the influence of sponge city construction on the ground-
and formed a high value area of the regional groundwater
water level near the spring groups. Under different
level. When the infiltration conditions were the same, the
precipitation conditions, the increment of the groundwater
groundwater level increased gradually with an increase in
level from scenarios S1, S2, and S3 was compared to that
precipitation. When the precipitation conditions were the
of scenario S0 (Figure 9), and 240 data were collected for
same, we observed that the greater the infiltration capacity
each scenario. Compared with scenario D-S0, where the
of the underlying surface, the greater the increment of the
maximum increment of the groundwater level was 0.09,
groundwater level.
0.13, and 0.27 m, the average increment of the groundwater
971
Figure 8
K. Sun et al.
|
|
Influences of sponge city construction on spring discharge
Distribution of increased groundwater level for various scenarios in pilot area surroundings after 20 years.
Hydrology Research
|
51.5
|
2020
972
Figure 9
K. Sun et al.
|
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
Statistics for increment of groundwater level in various scenarios (S) in observation well e.
level was 0.03, 0.06, and 0.13 m in scenarios D-S1, D-S2,
under various precipitation conditions. It can be seen that
and D-S3, respectively. Compared with scenario N-S0,
when the infiltration conditions were the same, the spring
where the maximum increment of the groundwater level
discharge increased the most in wet years. When the precipi-
was 0.15, 0.23, and 0.40 m, the average increment of the
tation remained the same, the larger the infiltration capacity
groundwater level was 0.03, 0.08, and 0.18 m in scenarios
was, the more the spring discharge increased; 20 years later,
N-S1, N-S2, and N-S3, respectively. Compared with scenario
the sponge city construction could increase the annual
W-S0, where the maximum increment of the groundwater
spring discharge by 0.0672, 0.0766, and 0.0875 × 108 m3 in
level was 0.16, 0.23, and 0.46 m, the average increment of
a dry year, normal year, and wet year, respectively. The
the groundwater level was 0.04, 0.10, and 0.22 m in scen-
observed data indicated that the discharge of Pearl spring
arios W-S1, W-S2, and W-S3, respectively. The simulation
in 2014 was 0.058 × 108 m3. It could be seen that under
results indicated that when precipitation was sufficient, the
the condition of continuous wet years, sponge city construc-
groundwater level near the spring groups would increase
tion could increase the annual spring discharge there by
by 0.46 m, and the average water level would increase by
approximately 1.5 times the Pearl spring discharge after 20
0.22 m compared to the results that would occur in the
years.
absence of the sponge city after 20 years. The higher the construction level of the sponge city and the more precipitation there is, the more the groundwater level near the spring groups should rise.
DISCUSSION Based on our analysis, we found that sponge city construc-
Change in spring discharge
tion, due to an increased precipitation infiltration capacity of the underlying surface, could affect the precipitation to
Figure 10 shows the increment of spring discharge in scen-
recharge the local groundwater, which would be beneficial
arios S1, S2, and S3 compared with that of scenario S0
for the increment of the groundwater level in and around
973
Figure 10
K. Sun et al.
|
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
Increment of spring discharge over prediction time frame.
the sponge city area. After 20 years, in all scenarios, the
and Ordovician limestones were exposed over a large area
groundwater level should have increased. The higher the
(Figure 2). The pilot test area of sponge city construction
degree of sponge city construction, the greater the amount
was almost located in the local groundwater system of the
of precipitation and the more obvious the uplift effect
Jinan watershed and partially belonged to the recharge
would be. However, the effect of sponge city construction
area. Therefore, when the infiltration coefficient of the
with regard to increasing the groundwater level and spring
groundwater in the sponge city pilot area increases, it
discharge near the spring groups was not proportional.
could affect the local groundwater flow within a certain
The average increment of the groundwater level near the
range, which has a limited effect on spring restoration (the
spring groups should rise by about 0.22 m, the maximum
simulated maximum annual increment percentage was
increment of the groundwater level near the spring groups
approximately 5%). The construction of the sponge city
could rise by 0.46 m, and the annual spring discharge
was a project that necessitated both huge quantities in con-
8
3
should increase by approximately 0.09 × 10 m in continu-
struction and monetary funds. The construction of the Jinan
ous wet years after 20 years. In 20 continuous wet years,
sponge city cost approximately 7.831 billion yuan, and it
the proportion of annual increment gradually changes
involved the reconstruction of a building area, garden and
from 0.40% to 5.00% when compared with the simulated
green space, roadways, an urban water system, among
annual spring discharge for the four springs in the scenario
other aspects. To increase groundwater recharge to the
S0. As time goes on, the influence of sponge city construc-
greatest extent possible, so as to protect the springs by
tion on spring discharge is increasingly prominent.
means of sponge city construction, it would be necessary
Therefore, an immediate response by the increment of
to expand the pilot area; however, this method could require
spring discharge would not be observed in a short time
further investments. It could be more effective if the pilot
frame.
area were moved to the southern part of the Jinan water-
In the Jinan spring area, the main groundwater recharge
shed. Owing to the short distance between the recharge
source was from the southern region where the Cambrian
and discharge areas of the watershed and the large
974
K. Sun et al.
|
Influences of sponge city construction on spring discharge
Hydrology Research
|
51.5
|
2020
topographic slope when rainstorms occur, the runoff would
0.22 m, on average, and the maximum annual spring dis-
not be able to be intercepted in time, which would result in
charge would increase by approximately 0.09 × 108 m3.
urban waterlogging; therefore, how to balance the protec-
The increased maximum annual spring discharge after 20
tion of the springs with the issue of waterlogging will be a
years could account for approximately 5% of the simulated
question worthy of further consideration.
average annual spring discharge. It was not possible for water managers to observe an immediate response of the groundwater level and spring discharge in a short time
CONCLUSIONS
frame after the implementation of the project. To increase spring discharge after the implementation of
As the first batch of pilot sponge cities in China in 2015, one
sponge city construction in Jinan, one method would be to
of the important purposes of sponge city construction in
further expand the pilot area, which would incur further
Jinan was to increase the precipitation infiltration and pro-
expenses; the other method, to increase spring discharge,
tect the springs. The effect of sponge city construction on
would be to move the pilot area to the southern region,
the groundwater system in Jinan was discussed in detail in
which would probably enhance the recharge sources of
this study. Groundwater types vary from the karst aquifer
the groundwater system. Jinan has a waterlogging issue
in the southern region to the porous media in the northern
due to its short distance from the recharge zone to the dis-
region of the study area. The groundwater mainly received
charge zone. The combined influence of waterlogging and
infiltration recharge from precipitation in the southern
sponge city construction on the groundwater level was pre-
area. Four well-known major springs were located in the
sent and not discussed in this article; however, the proper
northern part of the study area, and the sponge city pilot
positioning of a sponge city pilot area could depend on prop-
area in Jinan was located in the north part. A finite-differ-
erly balancing the protection of springs with the issue of
ence groundwater flow model in the Jinan spring area was
waterlogging.
developed to evaluate the influences of the project on the groundwater system. The equivalent media method was used to characterize the karst and fissure aquifers. Model calibration was conducted based on the observed ground-
ACKNOWLEDGEMENTS
water level and spring discharge data, using a trial-anderror method and optimized software (UQ-PyL). After
This work was supported by the National Natural Science
model calibration, the model was applied to evaluate the
Foundation
of
China
(Grant
Nos.
41877173
and
changes in the groundwater level and spring discharge
41831283)
and
the
National
Key
Research
and
after sponge city construction near the pilot area.
Development Program of China (2018YFC0407900). The
The results from the 12 scenarios showed that the larger
authors would also like to acknowledge the financial
the infiltration capacity of the sponge city pilot area, the
support
more the groundwater level and spring discharge rose.
Hydrological Process Modeling of Jinan ‘Sponge City’ and
Additionally, the more precipitation that fell, the more the
Beijing Advanced Innovation Program for Land Surface
uplift of the groundwater level occurred. Sponge city con-
Science.
struction
was
beneficial
to
the
increment
of
of
the
project
named
Water
Cycle
the
groundwater level in the pilot area. After 20 years, even under the condition of dry years, the groundwater level in most areas of the pilot could rise more than 2 m. However, the effect of the sponge city construction on the protection of the springs was not very obvious. After 20 years, under the condition of continuous wet years, the groundwater level around the spring groups would only increase by
REFERENCES Ahammed, F. A review of water-sensitive urban design technologies and practices for sustainable stormwater management. Sustainable Water Resources Management 3 (3), 1–14. https://doi.org/10.1007/s40899-017-0093-8.
and
975
K. Sun et al.
|
Influences of sponge city construction on spring discharge
Bakalowicz, M. Karst groundwater: a challenge for new resources. Hydrogeology Journal 13 (1), 148–160. https://doi. org/10.1007/s10040-004-0402-9. Chen, C. X. & Hu, L. T. A review of the seepage-pipe coupling model and its application. Hydrogeology & Engineering Geology 35 (3), 70–75 (In Chinese). Dragoni, W., Mottola, A. & Cambi, C. Modeling the effects of pumping wells in spring management: the case of Scirca spring (central Apennines, Italy). Journal of Hydrology 493, 115–123. https://doi.org/10.1016/j.jhydrol.2013.03.032. Dverstorp, B., Andersson, J. & Nordqvist, W. Discrete fracture network interpretation of field tracer migration in sparsely fractured rock. Water Resources Research 28 (9), 2327–2343. https://doi.org/10.1029/92WR01182. Ellis, J. B. & Lundy, L. Implementing sustainable drainage systems for urban surface water management within the regulatory framework in England and Wales. Journal of Environmental Management 183, 630–636. https://doi.org/ 10.1016/j.jenvman.2016.09.022. Felton, G. K. & Currens, J. C. Peak flow rate and recessioncurve characteristics of a karst spring in the inner bluegrass, central Kentucky. Journal of Hydrology 162 (1–2), 99–118. https://doi.org/10.1016/0022-1694(94)90006-X. Ford, D. C. & Williams, P. W. Karst Hydrogeology and Geomorphology. Wiley, Chichester, UK. Ghasemizadeh, R., Hellweger, F., Butscher, C., Padilla, I., Vesper, D., Field, M. & Alshawabkeh, A. Review: groundwater flow and transport modeling of karst aquifers, with particular reference to the North Coast Limestone aquifer system of Puerto Rico. Hydrogeology Journal 20 (8), 1441–1461. https://doi.org/10.1007/s10040-012-0897-4. Hartmann, A., Goldscheider, N., Wagener, T., Lange, J. & Weiler, M. Karst water resources in a changing world: review of hydrological modeling approaches. Reviews of Geophysics 52 (3), 218–242. https://doi.org/10.1002/2013RG000443. Hu, L. T., Chen, C. X., Jiao, J. J. & Wang, Z. J. Simulated groundwater interaction with rivers and springs in the Heihe river basin. Hydrological Processes 21 (20), 2794–2806. https://doi.org/10.1002/hyp.6497. Ice, G. History of innovative best management practice development and its role in addressing water quality limited waterbodies. Journal of Environmental Engineering 130 (6), 684–689. https://doi.org/10.1061/(ASCE)0733-9372(2004) 130:6(684). Jia, H. F., Yu, S. L. & Qin, H. P. Low impact development and sponge city construction for urban stormwater management. Frontiers of Environmental Science & Engineering 11 (4), 20. https://doi.org/10.1007/s11783-017-0989-4. Kang, F. X., Jin, M. G. & Qin, P. R. Sustainable yield of a karst aquifer system: a case study of Jinan springs in Northern
Hydrology Research
|
51.5
|
2020
China. Hydrogeology Journal 19 (4), 851–863. https://doi.org/ 10.1007/s10040-011-0725-2. Labat, D., Ababou, R. & Mangin, A. Introduction of wavelet analyses to rainfall/runoffs relationship for a karstic basin: the case of Licq-Atherey karstic system (France). Groundwater 39 (4), 605–615. https://doi.org/10.1111/ j.1745-6584.2001.tb02348.x. Lu, Y. R. Karst water resources and geo-ecology in typical regions of China. Environmental Geology 51 (5), 695–699. https://doi.org/10.1007/s00254-006-0381-3. Meng, X. M., Yin, M. S., Ning, L. B., Liu, D. F. & Xue, X. W. A threshold artificial neural network model for improving runoff prediction in a karst watershed. Environmental Earth Sciences 74 (6), 1–10. https://doi.org/10.1007/s12665-0154562-9. Padilla, A. & Pulido-Bosch, A. Simple procedure to simulate karstic aquifers. Hydrological Processes 22 (12), 1876–1884. https://doi.org/10.1002/hyp.6772. Qian, J. Z., Zhan, H. B., Wu, Y. F., Li, F. L. & Wang, J. Q. Fractured-karst spring-flow protections: a case study in Jinan, China. Hydrogeology Journal 14 (7), 1192–1205. https://doi. org/10.1007/s10040-006-0061-0. Robineau, T., Tognelli, A., Goblet, P., Renard, F. & Schaper, L. A double medium approach to simulate groundwater level variations in a fissured karst aquifer. Journal of Hydrology 565, 861–875. https://doi.org/10.1016/j.jhydrol. 2018.09.002. Scanlon, B. R., Mace, R. E., Barrett, M. E. & Smith, B. Can we simulate regional groundwater flow in a karst system using equivalent porous media models? Case study, Barton Springs Edwards aquifer, USA. Journal of Hydrology 276 (1–4), 137–158. https://doi.org/10.1016/S0022-1694(03)00064-7. Wang, Q. B., Yan, J. S., Duan, X. M. & Gao, Z. D. Study on Groundwater Circulation and Sustainable Development in Jinan Spring Area of Karst in North China. Jinan Publishing House, Jinan, China (In Chinese). Wang, C., Duan, Q. Y., Tong, C. H., Di, Z. H. & Gong, W. A GUI platform for uncertainty quantification of complex dynamical models. Environmental Modelling & Software 76, 1–12. https://doi.org/10.1016/j.envsoft.2015.11.004. Wang, G., Liu, S., Liu, T., Fu, Z., Yu, J. & Xue, B. Modelling above-ground biomass based on vegetation indexes: a modified approach for biomass estimation in semi-arid grasslands. International Journal of Remote Sensing 40 (10), 3835–3854. Xu, C. Q., Jia, M. Y., Xu, M., Long, Y. & Jia, H. F. Progress on environmental and economic evaluation of low-impact development type of best management practices through a life cycle perspective. Journal of Cleaner Production 213, 1103–1114. https://doi.org/10.1016/j.jclepro.2018.12.272.
First received 8 January 2020; accepted in revised form 6 April 2020. Available online 20 May 2020
976
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Quantification of climate change and land cover/use transition impacts on runoff variations in the upper Hailar Basin, NE China Yuhui Yan, Baolin Xue, Yinglan A, Wenchao Sun and Hanwen Zhang
ABSTRACT Quantification of runoff change is vital for water resources management, especially in arid or semiarid areas. This study used the Soil and Water Assessment Tool (SWAT) distributed hydrological model to simulate runoff in the upper reaches of the Hailar Basin (NE China) and to analyze quantitatively the impacts of climate change and land-use change on runoff by setting different scenarios. Two periods, i.e., the reference period (before 1988) and the interference period (after 1988), were identified based on long-term runoff datasets. In comparison with the reference period, the contribution rates of both climate change and land-use change to runoff change in the Hailar Basin during the interference period were 83.58% and 16.42%, respectively. The simulation analysis of climate change scenarios with differential precipitation and temperature changes suggested that runoff changes are correlated positively with precipitation change and that the impact of precipitation change on runoff is stronger than that of temperature. Under different economic
Yuhui Yan Baolin Xue (corresponding author) Yinglan A Wenchao Sun Hanwen Zhang College of Water Sciences, Beijing Normal University, Beijing 100875, China E-mail: xuebl@bnu.edu.cn Baolin Xue Yinglan A Wenchao Sun Beijing Key Laboratory of Urban Hydrological Cycle and Sponge City Technology, Beijing, China
development scenarios adopted, land use was predicted to have a considerable impact on runoff. The expansion of forests within the basin might induce decreased runoff owing to enhanced evapotranspiration. Key words
| climate change, Hailar Basin, land cover/use change, runoff, SWAT model
INTRODUCTION Water resources form the foundation of sustainable socio-
Piao et al. ). The water cycle in many watersheds has
economic development (Barnett et al. ). However,
been affected considerably, and runoff in major watersheds
owing to continued global extreme climatic events and the
has shown a rapid decrease (Bao et al. ; Wang et al.
negative influence of human activities, China is facing
a). Human activities affect the hydrological cycle and
water resource problems that are becoming increasingly
the formation process of water resources by changing the
severe and that represent an important constraining factor
mode of land use to varying degrees (Wheater & Evans
on China’s future sustainable socioeconomic development
; Sterling et al. ), which can result in a series of pro-
(Barnett et al. ; Piao et al. ; Voeroesmarty et al.
blems such as wetland shrinkage and groundwater funneling
). In recent decades, under the influence of global cli-
(Kaushal et al. ; Sterling et al. ; Li et al. ). There-
mate change, the climate in northern China has shown an
fore, it is of considerable scientific and practical importance
obvious trend of warming and drying (Fang et al. ;
to investigate the impact of climate change and land-use change on the hydrological cycle to resolve water resources
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.022
problems in a changing environment. Runoff is a vital link in the hydrological cycle that is also important in relation to the allocation of water resources
977
Y. Yan et al.
|
Climate and land-use impacts on runoff variation
Hydrology Research
|
51.5
|
2020
within a basin (Milly et al. ). Changes in runoff directly
The impact of climate change and land-use change on
affect life and production activities in a basin (Piao et al.
runoff varies spatially, and therefore, research is usually con-
). Therefore, it is of considerable importance to under-
ducted in specific watersheds (A et al. a). Many studies
take quantitative research on the impact of climate and
have assessed the impact of climate change and land-use
land-use changes on runoff. Methods commonly used for
change on runoff in various watersheds in China. Research
the quantitative analysis of the impact of environmental
in an agricultural catchment of a tributary of the Jinghe
changes on runoff can be divided into three categories: com-
River on the Loess Plateau highlighted that climate variabil-
parative basin tests, statistical analysis methods, and
ity had greater influence than land-use change on the
hydrological model simulations (Mishra & Singh ).
surface hydrological processes during 1981–2000 (Li et al.
The comparative watershed method is used for the manual
). Results of a study in the Dongjiang River Basin indi-
evaluation of the human impact on runoff by changing the
cated that climate variability and human activities each
natural geographical conditions (one or more watershed
accounted for approximately 50% of the runoff change in
characteristics) of the test watershed. However, this
the low-flow period (Zhou et al. ). Analysis in three
approach is usually considered best for studying the effect
main tributary subcatchments in the Yellow River Basin
of climate change in small watersheds but it is difficult to
illustrated that climate change accounts for only 8% of the
find two similar medium- or large-sized watersheds, and
total decline in mean annual runoff, whereas human activi-
even the same watershed might undergo notable changes
ties account for 92% (Li et al. a). However, few studies
at different stages (Lorup et al. ). Statistical analysis
have conducted relevant research on the Hailar Basin.
can be used to analyze the trends of the change of hydrome-
The Hailar Basin is a typical northern inland river basin
teorological data, but it cannot consider the spatial
in NE China. In recent years, it has encountered increas-
heterogeneity of watersheds or the mechanisms via which
ingly serious water shortage problems (Chi et al. ;
land-use change and climate change might affect runoff
Wang et al. b). Therefore, it is of considerable urgency
change in watersheds (Gampe et al. ; Ishida et al.
that research be conducted to alleviate the water shortage
). Therefore, comprehensive physically based tools are
problems in this basin (Wang et al. a). This study used
needed to obtain as much information as possible from
the SWAT model to simulate runoff in the Hailar Basin,
limited existing data (Li et al. ). Hydrological models
and the impact of climate change and land-use change on
provide a framework for conceptualizing and studying
runoff was analyzed quantitatively. The main objectives of
relationships (Wang & Xu ). By linking model par-
this study were as follows: (1) to determine abrupt change
ameters
surface
points in runoff according to runoff time series, (2) to ana-
features, hydrological models can establish relationships
lyze quantitatively the contribution rates of climate change
among climate, human activities, and runoff (Leavesley
and land-use change to runoff, and (3) to investigate runoff
; Jothityangkoon et al. ). For example, Cuo et al.
response under different climate change and land-use
() used the variable infiltration capacity hydrological
change scenarios. The ultimate aim was to provide support
model to analyze quantitatively the impact of climate
for strategic planning and allocation of water resources in
change and land-use change on runoff in the upper reaches
the Hailar Basin.
directly
with
physically
observable
of the Yellow River Basin (China). Mango et al. () used the Soil and Water Assessment Tool (SWAT) distributed hydrological model to analyze quantitatively the impact of
STUDY AREA AND DATA
climate change and land-use change on runoff in the Mala River Basin (Kenya). In hydrological model simulations,
Study area
the mechanism of runoff formation is consistent with or without environmental change and no detailed data of
The Hailar Basin (47 380 –50 160 N, 117 430 –122 20 E) is
human activities are needed (Legesse et al. ). Therefore,
located in the northeast of the Inner Mongolia Autonomous
this study chose the hydrological model simulation method.
Region to the southwest of the city of Hulunbeier (Figure 1).
978
Figure 1
Y. Yan et al.
|
|
Climate and land-use impacts on runoff variation
Hydrology Research
|
51.5
|
2020
Map showing location of the upper Hailar River Basin.
It has a temperate continental monsoon climate, and it
period is derived primarily from snowmelt and precipitation.
2
covers an area of 54,500 km . The basin is located at the
The summer flood season usually occurs during June–
junction of the western slopes of the Daxing’anling moun-
October, which is the period with the most concentrated
tains and the northeastern edge of the Inner Mongolia
precipitation (Xue et al. b; A et al. b).
High Plain (Wang et al. b). It is a fan-shaped basin that has large topographic fluctuation (Han et al. a).
Data sources
The general trend of elevation (range: 536–1,706 m) is from low in the west to high in the east. The upper reaches
The data used in this study were divided into two parts. Part
of the Hailar Basin constitute the main area of the basin
1 data were used to analyze the water resources situation of
(Han et al. b). There are two flood seasons annually: a
the upper reaches of the Hailar Basin and to determine the
spring flood season and a summer flood season. The
relationship between runoff and climate factors. Part 2 were
spring flood season that usually occurs during March–May
the input data required by the SWAT model. The input data
reaches its peak in May (Fang et al. b). Runoff in this
of the SWAT model also included two main parts: spatial
979
Y. Yan et al.
|
Climate and land-use impacts on runoff variation
Hydrology Research
|
51.5
|
2020
data and attribute data. Spatial data mainly included digital
to soil parameters of the United States Geological Survey
elevation model data (90 × 90 m), land use/cover data
standard. The soil distribution data in the Hailar Basin are
(1,000 × 1,000 m), soil distribution data, and spatial distri-
shown in Figure 2(b). This study used daily meteorological
bution data of the hydrological stations and meteorological
and runoff data from 1980 to 2012. Meteorological data
stations. Attribute data mainly included soil type data,
included daily precipitation, maximum and minimum temp-
meteorological data, and hydrological data. Index tables of
erature, humidity, and wind speed.
land-use type and soil type were established in the modeling process to ensure the required data in the model corresponded to each database. Digital elevation model data: first, geometric correction
METHODOLOGY
was performed, and then clipping and projection transformation were undertaken using ArcGIS. Land use/cover data:
Mutation analysis
first, the original data were downloaded from the Resources and Environment Science Data Center of the Chinese
Mann–Kendall mutation test
Academy of Sciences, and reclassification and projection transformation performed using ArcGIS. The data were
The Mann–Kendall (M-K) test is a widely used non-
divided into six categories: forest land, grassland, water
parametric test method recommended by the World
area, urban land, unused land, and cultivated land, as
Meteorological Organization. In recent years, many studies
shown in Figure 2(a). Soil type data and soil attribute data:
have adopted the M-K method to analyze the changes of
these data were obtained from the Harmonized World Soil
trends in time series of precipitation, runoff, temperature,
Database, which contains a large number of soil parameters.
and water quality (Hamed & Rao ; Yue et al. ).
The data are presented in a gridded format using the
In this study, the M-K catastrophe test was used to ana-
WGS1984 coordinate system. The soil classification system
lyze the catastrophe of the runoff time series. The M-K
adopted is mainly FAO-90. As the data provided in the data-
mutation test defines statistical variables by constructing
base are international standards, the SPAW software was
order columns. Assuming time series X ¼ {x1 , x2 , . . . , xn },
required to convert the data from international standards
ri represents the cumulative number of the ith sample, and
Figure 2
|
(a) Distribution of land use/cover types and (b) distribution of soil types.
Y. Yan et al.
980
|
Climate and land-use impacts on runoff variation
Hydrology Research
k X
ri ¼
1 0
|
2020
For a time series, the principle of the sliding t-test is to ri
(k ¼ 2, 3, 4, . . . , n)
(1)
i¼1
51.5
Sliding t-test technique
xi is greater than xj (1 j i), we have
sk ¼
|
extract two subsequences of the main time series and then to test whether there is a significant difference between
xi > xj (j ¼ 2, 3, . . . , i) other
the average values of those two subsequences (Machiwal (2)
& Jha ). If there is a significant difference, the sequence is considered to have mutation (Zhao et al. ). Assuming
Under the assumption that the time series is random and
time series X ¼ {x1 , x2 , . . . , xn }. Then, by taking a certain time in the series as a reference point and taking
independent, we have the following:
subsequence sk E(sk ) UFk ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Var(sk )
and
subsequence
X2 ¼ {x1 , x2 , . . . , xn } forward and backward, respectively, (k ¼ 1, 2, . . . , n)
(3)
based on the reference point, statistic t can be obtained based on the two subsequences as follows:
When the elements x1 , x2 , . . . , xn are independent of t¼
each other and continuously and uniformly distributed: E(sk ) ¼ k(k 1)=4
(4)
Var(sk ) ¼ k(k 1)(2k þ 5)=72
(5)
UFi is a standard normal distribution. Given significance level α, if |UFi| > U, there is an obvious trend change in the sequence, and the critical value of UF and UB is ±1.96. Arranging time series X in the reverse order and then performing the calculation according to the above equation,
2 1 x x rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 1 s þ n1 n2
(7)
P 1 1 ¼ (1=n1 ) ni¼1 2 ¼ (1=n2 ) In the above equation, x xi , x q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi Pn2 (n1 s21 þ n2 s22 )=(n1 þ n2 2). i¼1 xi , and s ¼ Then, statistic t obeys the t distribution with n1 þ n2 2 degrees of freedom. The specific steps of the sliding t-test for mutation are described in the following. 1. On the basis of the determination of the reference point, the lengths of subsequences n1 and n2 are
we have
X1 ¼ {x1 , x2 , . . . , xn }
determined. Normally, the lengths of the two subseUBk ¼ UFk 0 k ¼ nþ1 k
(k ¼ 1, 2, . . . , n)
(6)
quences are taken as the same, i.e., n1 ¼ n2. Then, subsequences n1 and n2 are taken forward and backward, respectively.
By analyzing the statistical sequences UFk and UBk, the
2. By sliding the reference point backward in turn, taking
trend change of sequence X can be analyzed further, which
out the corresponding subsequences, and calculating
allows the mutation time to be defined and the mutation
the
region to be identified. If the value of UFk is >0, it indicates
statistics can be obtained.
corresponding
statistics,
the
n (n1 þ n2) þ 1
that the sequence shows an upward trend and vice versa.
3. The obtained statistics are arranged in sequence to obtain
When the value exceeds the critical line, it indicates that
the statistics sequence, the significance level @ is selected,
the upward or downward trend is significant. If the UFk
and the corresponding standard statistics t@ are obtained
and UBk curves have intersection points and the intersection
from a lookup table. The statistics sequence is drawn into
points are between the critical straight lines, then the time
a polyline graph. If jti j > t@ , the datum point can be
corresponding to the intersection points is the time when
adjudged the abrupt change point of the sequence
the abrupt change is considered to start (Hamed ).
(Dittmer ; Zhao et al. b).
981
Y. Yan et al.
|
Climate and land-use impacts on runoff variation
Hydrology Research
SWAT model
|
51.5
|
2020
reaches of the Hailar Basin based on three indices: the decision coefficient R 2, Nash–Sutcliffe coefficient Ens, and
The SWAT model is a distributed hydrological model devel-
relative error RE (Bhatta et al. ). The expressions of
oped by the Agricultural Research Center of the United
the three indices are as follows:
States Department of Agriculture (Arnold et al. ). The SWAT model is used widely for the simulation of hydrological cycle processes because its strong physical basis gives it
Pn
2 sim (Qobs Qobs Qsim ave )(Qi ave ) i 2 Pn 2 obs sim Qobs Qsim ave ) ave ) i¼1 (Qi i¼1 (Qi
the ability to simultaneously integrate the effects of topography, soil, land use, and weather (Yang et al. ). The model can simulate runoff change under different climate
"P Ens ¼ 1
change conditions, land-use conditions, soil conditions, and watershed management conditions; thus, the SWAT model was selected for the runoff simulations conducted
i¼1
R2 ¼ P n
Pn RE ¼
n obs i¼1 (Qi Pn obs i¼1 (Qi
2
Qsim i )
(9)
#
2
Qobs ave )
(Qobs Qsim ) Pn i obs i i¼1 Qi
i¼1
(10)
(11)
in this study (Moriasi et al. ). The model is calculated is the measured runoff value, In the above equations, Qobs i
according to the following water balance equation:
Qsim i SWt ¼ SW0 þ
t X
(Rday Qsurf Ea Wseep Qgw )
(8)
i¼1
where SW0 is the initial soil water content on the ith day, Rday is the precipitation on the ith day, Qsurf is the surface runoff on the ith day, Ea is the soil evaporation and plant transpiration on the ith day, Wseep is the seepage flow on the ith day, and Qgw is the amount of groundwater on the ith day.
is the simulated runoff value, Qobs ave is the measured aver-
age runoff value, and Qsim ave is the simulated average runoff value. The R 2 values (range: 0–1) represent the fitting degree between the simulated and measured values. The Nash–Sutcliffe coefficient (range: 0–1) is used to assess the quality of the hydrological model simulation results. Only when Ens > 0.5 can the simulation results be accepted. A value of Ens close to 1 indicates that the model simulation quality is good and its credibility is high. If the value of the relative error RE is controlled to within 25%, the simulation results of the model can be considered within an acceptable
Model calibration and validation
range (Moriasi et al. ; Asl-Rousta et al. ).
In this study, the simulation of daily runoff was performed
Quantification of climate and land-use contributions to runoff change
first. Then, the SWAT-CUP software was used to calibrate and verify the model. The data used included land-use data in 1980 and 2000 and meteorological data from 1980
Based on the SWAT model, this study used scenario analysis
to 2012. Based on runoff data recorded at the Bahou
to separate the influences of various factors on runoff, i.e.,
Station, the Latin Hypercube One factor At a Time
assuming that climatic factors or human activity factors
(LH-OAT) method was used to analyze the sensitivity of
remain constant and that another factor changes accord-
the parameters, and those parameters with strong sensitivity
ingly, the contribution rate of this factor to runoff can be
were selected as adjustment parameters (van Griensven
analyzed quantitatively.
et al. ; Arnold et al. ). The SUFI-2 algorithm was
This study used the M-K mutation test and sliding t-test
used to determine the optimal ranges and values of the par-
to determine the time point of runoff mutation in the water-
ameters through iterative calculation, and the optimal
shed. The year before the time point of abrupt change was
values of the parameters were introduced into the model
defined as a period of natural runoff, i.e., runoff was
through internal adjustment (Anoh et al. ). In this
assumed undisturbed by climate change and land-use
study, the effective evaluation method was used to evaluate
change, whereas runoff after the time point of abrupt
the simulation accuracy of the SWAT model in the upper
change was considered disturbed by both climate change
Y. Yan et al.
982
Table 1
|
|
Climate and land-use impacts on runoff variation
Hydrology Research
|
51.5
|
2020
based on the following scenarios: current precipitation
Model simulation scenario settings
unchanged, 10% and 20% increase in precipitation, 10%
Scenario
Land use/cover data
Meteorological data
1
1980s
1980–1988
unchanged, 1 C and 2 C increase in temperature, and
2
1980s
1989–2012
1 C and 2 C decrease in temperature.
3
2000s
1980–1988
4
2000s
1989–2012
and 20% decrease in precipitation, current temperature
To explore the influence of land use on runoff, this study used the cellular automata–Markov (CA–Markov) model to simulate land-use development scenarios based on existing land-use datasets of the Hailar Basin. Land use in the
and land-use change. To quantify the contributions of land-
upper reaches of the Hailar Basin was simulated for 2030
use change and climate change to runoff change, four scen-
and 2050, and the associated runoff was investigated accord-
arios were established (Table 1). A comparison of the simulated runoff under scenario 2 with that under scenario
ing to historical development trends and existing problems within the Hailar Basin.
1 allowed the analysis of the impact of climate change on runoff. Similarly, a comparison of the simulated runoff under scenario 3 with that under scenario 1 allowed the
Land use/land cover prediction
analysis of the impact of land-use change on runoff. A comparison of the runoff simulated under scenario 4 with that
This study used the CA–Markov model to predict future land
under scenario 1 allowed the analysis of the impact of
use/land cover. The CA–Markov model is a powerful tool
both climate change and land-use change on runoff. Thus,
that combines the advantages of the Markov chain and cel-
the contribution rates of climate change and land-use
lular automata.
change to runoff change were obtained.
The CA–Markov model describes land-use change from
In this study, Q1, Q2, Q3, and Q4 were used to represent
one period to another, which allows predictions of the
the average annual runoff simulated under scenarios 1, 2, 3,
future trend of land-use change. The following formula can
and 4, respectively. Thus, Q2 Q1 represents the impact of
be used to predict land use:
climate change on runoff, Q3 Q1 represents the impact of land use/cover change (land-use change) on runoff, and
Stþi ¼ Pij St
(14)
Q4 Q1 represents the total change of runoff within the basin. In this paper, we define the following:
where St and Stþi are the states of the land-use structure at time t and t þ i, respectively and Pij is the state-transition
Q2 Q1 ωcl ¼ Q4 Q1
(12)
Q3 Q1 Q4 Q1
(13)
ωh ¼
matrix. A cellular automaton represents a gridded dynamics model with strong spatial computing capability. The CA– Markov model can be expressed as follows: Stþi ¼ f(St , N)
(15)
Scenario setting and model analysis where S is a finite discrete state set element, N is a cellular To investigate the impact of climate change on runoff in the
neighborhood, t and t þ i are different moments, and f is
upper reaches of the Hailar Basin, the change values of pre-
the cell transformation rule in local space.
cipitation and temperature were determined according to
The model can simulate the temporal and spatial evol-
the possible range of future climate change. In simulating
utions of complex systems, and it has been used widely in
runoff in the upper reaches of the Hailar Basin, 25 different
many studies (Mitsova et al. ; Sang et al. ). The
climate change combination programs were considered
detailed parameters and steps of land-use prediction using
Y. Yan et al.
983
|
Climate and land-use impacts on runoff variation
Hydrology Research
|
51.5
|
2020
the CA–Markov model are as follows. (1) Data format con-
upward or downward trend is considered significant. The
version and reclassification are performed to obtain fixed
range beyond the critical line is determined as the time
land-use types. (2) A state-transition probability matrix and
zone in which the abrupt change occurs. If there is an inter-
a transition area matrix are obtained using the CA–
section point between the UFk and UBk curves and that
Markov module. (3) A transitional fitness image set is estab-
intersection point lies between the critical lines, the time
lished. (4) The CA filter and number of cycles are
corresponding to the intersection point is the time at
determined. (5) The accuracy of the predicted image is eval-
which the abrupt change starts. As shown in Figure 4(a),
uated according to the actual image. In this study, a map of
the UFk value after 1995 is <0, which indicates that the
land use in 2000 was used as the basic image, and an assem-
runoff sequence at the Bahou station began to decline con-
bly transition probability matrix of land-use maps in 2005
tinuously after 1995. In 2003, the UFk value was lower
and 2015 was used to predict the land-use situation in
than 1.96, indicating that runoff had a significant decreas-
2030 and 2050.
ing trend based on the significance level test of 0.05. As shown in Figure 4(a), during 1980–1995, the UFk and UBk curves had intersection points between the critical lines at
RESULTS
around 1980, 1982, 1986, 1988, and 1992. Therefore, the sliding t-test method was used to determine the mutation year in this study.
Runoff trend analysis
In this study, the sliding t-test method was also used to The linear analysis and 5-year moving average curve of
analyze the Bahou Station runoff time series. To avoid vari-
runoff during 1980–2012 at the Bahou Station at the outlet
able point drift caused by different subsequence lengths, the
of the upper reaches of the Hailar Basin are shown in
sliding length was changed repeatedly. Finally, it was deter-
Figure 3. In the past 30 years, annual runoff at the Bahou
mined that the length of the two sliding sequences should be
Station has fluctuated but generally diminished with a rate
4, i.e., n1 ¼ n2 ¼ 4, and 33 statistics relating to the Bahou
that has become more obvious since 1990. Annual runoff
Station formed the corresponding statistical sequence. If
at the Bahou Station has ranged from a maximum value of
the significance level @ ¼ 0:05 were selected, the statistic
8
69.82 × 10 m 8
3
3
in 1984 to a minimum value of 10.28 × 8
3
10 m in 2007 with a trend of 0.24 × 10 m /a.
t–t(6) and the critical value were found to be t0.05(6) ¼ 1.943, and each statistical sequence and critical value are
In this study, the M-K method and sliding t-test method
plotted on the graph for the mutation analysis. The analysis
were used to test for catastrophe in the runoff sequence. UFk
of Figure 4(b) reveals that for the annual runoff series, the t
values >0 indicate that the sequence shows an upward trend
statistics of the Bahou Station exceed the critical values in
and vice versa. When the values exceed the critical line, the
two places: 1987 and 1988, both of which are possible abrupt years. As the M-K mutation test identified 1988 as the intersection point of the UFk and UBk curves, it can be considered that mutation occurred in the annual runoff at the Bahou Station in 1988. As water resources are affected by various climate change factors, e.g., rising temperature and extreme weather, runoff of the major rivers in northern China currently shows an overall downward trend. Therefore, the identified sudden change is probably attributable to climate change because temperature rise leads to increased evapotranspiration. Based on these two methods, the mutation point was set to 1988, and the study period was divided into the reference period (before 1988) and the inter-
Figure 3
|
Results of the trend analysis of runoff at the Bahou Station.
ference period (after 1988).
Y. Yan et al.
984
Figure 4
|
|
Climate and land-use impacts on runoff variation
Hydrology Research
|
51.5
|
2020
Results of (a) the M-K mutation test and (b) the moving t-test.
|
Model calibration and validation
Table 2
Parameter sensitivity analysis and model calibration are clo-
Parameter
sely linked, and both are essential processes for runoff
name
simulation using hydrological models. The SWAT-CUP soft-
GW_DELAY Groundwater delay factor
1
(30,450) 45.75
ALPHA_BF
Base-flow recession constant
2
(0,1)
SMFMN
Snowmelt coefficient on 3 December 21
( 10,10) 2.42
CANMX
Maximum canopy storage
(0,100)
76.25
SOL_AWC
Available water capacity 5 of the soil layer
(0,1)
0.15
SMFMX
Snowmelt coefficient on 6 June 21
( 10,10) 6.75
GWQMN
Threshold depth of water in the shallow aquifer required for return flow to occur
Parameter sensitivity analysis
The most
ware can be used to perform sensitivity analysis and calibration on relevant parameters for a runoff simulation. Combined with the actual situation of the basin, the LHOAT method provided in SWAT-CUP was used in this study to select parameters related to runoff simulation in the northern region for sensitivity analysis and calibration, as shown in Table 2. Through
the
sensitivity analysis
of
the
SWAT
model, 11 parameters with high sensitivity were selected to calibrate and verify the model (Table 2). In this study, the preheating period of the model was taken as 1988–1989, 1990–2006 was taken as the calibration period, and 2007–2012 was taken as the validation period. The SWAT-CUP software was used to calibrate the model parameters, and the monthly runoff at the Bahou Station was used to calibrate and verify the
Rank Ranges
4
7
(0,2)
value
0.90
0.96
CN2
Curve number
8
(20,100) 46
SMTMP
Snow melting
9
( 10,10) 2.75
ESCO
Soil evaporation compensation coefficient
10
(0.01,1)
SFTMP
Snow temperature
11
( 10,10) 3.63
model. In the calibration period, the R 2, Ens, and RE values of
suitable Variable name
0.46
the simulated monthly runoff at the Bahou Station were 0.75, 0.71, and 20.8%, respectively, while the corresponding values in the validation period were 0.74, 0.70, and
Contribution of land use and climate change to runoff
20.3%, respectively, proving the suitability of the SWAT
variation
model for runoff simulation in the upper reaches of the Hailar Basin. The simulation results are shown in
According to the abrupt change years, the study period was
Figures 5 and 6.
divided into the reference period (before 1988) and
985
Y. Yan et al.
|
Climate and land-use impacts on runoff variation
Hydrology Research
Figure 5
|
Correlation between simulated and measured values of monthly streamflow at the Bahou Station: (a) calibration period and (b) validation period.
Figure 6
|
Measured and simulated monthly streamflow graphs (Bahou Station) for the calibration period and the validation period.
interference period (after 1988). Using the meteorological
Table 3
data and land-use data of different periods and by establishing different scenarios, the contribution rates of land-use change and climate change to runoff in the upper reaches of the Hailar Basin were calculated. As shown in Tables 3 and 4, the average annual runoff simulated under scenarios 1–4 was 125.43, 104.05, 121.23, and 99.85 m3/s, respect-
|
|
51.5
|
2020
Comprehensive scenario simulation results
Influence of land-use
Influence of climate
Scenario runoff (m3/s)
change on runoff (m3/s)
change on runoff (m3/s)
1
125.43
–
–
2
104.05
–
21.38
3
121.23
4.2
–
4
99.85
4.2
21.38
Annul average
ively. The analysis of scenarios 1 and 2 suggests that under the influence of climate change, annual average runoff has
suggests that under the influence of land-use change,
decreased by 21.38 m3/s. The analysis of scenarios 1 and 3
annual average runoff has decreased by 4.2 m3/s. The
Y. Yan et al.
986
Table 4
|
|
Climate and land-use impacts on runoff variation
Hydrology Research
|
51.5
|
2020
Simulated response of land-use change and climate change
Time
Simulated annul average runoff (m3/s)
Varied runoff (m3/s)
Influence of land-use change on runoff (m3/s)
Influence of climate change on runoff(m3/s)
Reference period
1980–1988
125.43
–
–
–
Interference period
1989–2012
99.85
25.58
4.2(16.42%)
21.38(83.58%)
Period
analysis of scenarios 1 and 4 suggests that under the joint
Table 6
|
Relative variation of mean annual runoff for different scenarios (%)
influence of climate change and land-use change, annual average runoff has decreased by 25.58 m3/s. Specifically, Q2 Q1 ¼ 21.38 m3/s,
Q3 Q1 ¼ 4.2 m3/s,
Q4 Q1 ¼
Variations in precipitation Variations in temperature
ΔP ¼ 20%
ΔP ¼ 10%
ΔP ¼ 0
ΔP ¼ 10%
ΔP ¼ 20%
25.58 m3/s, ωcl ¼ 83:58%, and ωh ¼ 16:42%. The contri-
ΔT ¼ þ2
19.82
7.15
7.25
16.86
34.70
bution rate of climate change (land-use change) to runoff
ΔT ¼ þ1
23.59
10.72
6.50
13.96
30.86
change in the basin is 83.58% (16.42%); consequently, it
ΔT ¼ 0
26.30
13.34
0
11.40
27.59
can be concluded that climate change has a greater impact
ΔT ¼ 1
28.47
16.08
7.20
9.45
25.20
than land-use change on runoff change in the basin.
ΔT ¼ 2
29.79
18.38
12.46
7.92
23.14
Scenario simulation of climate change and land
greater the increase in runoff. Values when ΔP ¼ 0 indicate
use/cover change
the response of runoff to temperature change when precipitation is constant. It can be seen that runoff decreases as
Climate change scenarios
temperature increases, and the more the temperature increases,
As shown in Table 5, under the condition of constant basin temperature, surface runoff will increase with the increase of precipitation. Under the condition of unchanged precipitation, surface runoff will change with the change of temperature. In all temperature-drop scenarios, surface runoff shows an increasing trend, whereas in all temperature-rise scenarios, surface runoff shows a decreasing trend. As shown in Table 6, values when ΔT ¼ 0 represent the
the more runoff decreases. Values when neither ΔT nor ΔP are zero indicate the runoff response effect when both temperature and precipitation change. It can be seen that when temperature decreases and precipitation increases, the increase in runoff is greatest. When temperature increases and precipitation decreases, a decrease in runoff is greatest. When precipitation increases by more than 10%, runoff increases markedly, indicating that precipitation has a more significant impact on runoff.
response of runoff to changed precipitation at constant temperature. It can be seen that the increase of precipitation
Land-use change scenarios
increases runoff, and the more precipitation increases, the According to the trend of land-use change in the Hailar Table 5
|
Relative variation of mean annual runoff for different scenarios (m3/s)
with current economic development and ecological protection in this region, this study simulated land-use scenarios
Variations in precipitation Variations in temperature P(1 þ 20%) P(1 þ 10%) P
Basin, and with reference to the problems encountered
P(1–10%) P(1–20%)
T þ 2 C
24.11
8.70
8.82 20.51 42.20
T þ 1 C
28.69
13.04
7.91 16.98 37.57
T
31.99
16.23
0.00
13.87 33.56
T 1 C
34.63
19.56
8.76
11.49 30.65
T 2 C
36.23
22.36
15.16 9.63
28.15
in 2030 and 2050 using the CA–Markov model and analyzed the impact of future land-use change on watershed runoff. In simulating and predicting future land-use situations, three development models were established: natural development, ecological protection, and economic development. The distribution maps of land use in 2030 and 2050 under these three economic development models are shown in Figure 7.
Y. Yan et al.
987
Figure 7
|
|
Climate and land-use impacts on runoff variation
Hydrology Research
|
51.5
|
2020
Distribution of land-use types under different economic development scenarios: (a) natural growth scenario in 2030, (b) natural growth scenario in 2050, (c) ecological protection scenario in 2030, (d) ecological protection scenario in 2050, (e) economic development scenario in 2030, and (f) economic development scenario in 2050.
Y. Yan et al.
988
Table 7
|
|
Climate and land-use impacts on runoff variation
Hydrology Research
|
51.5
|
2020
Proportion of land-use types in different scenarios (%)
Natural growth Land-use type
Ecological protection
Economic development
2030 (%)
2050 (%)
2030 (%)
2050 (%)
2030 (%)
2050 (%)
Cultivated land
7.902
7.983
7.728
7.658
7.937
8.228
Forest
34.294
34.501
34.294
35.966
34.190
34.094
Meadow
50.456
50.027
50.560
50.108
50.454
49.864
Water
0.720
0.895
0.721
0.895
0.686
0.895
Constructed land
0.639
0.779
0.743
0.860
0.744
1.104
Unused land
5.989
5.815
5.954
4.513
5.989
5.815
It can be seen from Table 7 that future areas of various
the proportion of the grassland area will still decrease in
land-use types will increase or decrease to varying degrees.
2030 and 2050; however, in comparison with the natural
In the case of natural growth, from 2030 to 2050, the pro-
growth scenario, both an increase in the proportion of the
portions of cultivated land, forest land, water areas, and
forest area and a decrease in runoff will be greater. Under
construction land will increase, while the proportions of grass-
the condition of economic development, the proportion of
land and unused land will decrease. In the case of ecological
the forest area will increase slightly in 2030 and 2050, the
protection, from 2030 to 2050, the proportions of forest land,
proportion of the grassland area will decrease greatly, and
water areas, and construction land will increase, while the pro-
the proportion of cultivated land will increase markedly;
portions of cultivated land, grassland, and unused land will
however, runoff in the basin will still show a decreasing
decrease. In the context of economic development, from
trend. This shows that the increase/decrease of the forest
2030 to 2050, the proportions of cultivated land, water areas,
land area has a reasonably direct impact on the increase/
and construction land will increase, while the proportions of
decrease of runoff. Forest land can improve both the
forest land, grassland, and unused land will decrease.
regional climate and the soil environment while reducing
The runoff simulation results under the different land-
the surface temperature and direct evaporation of water to
use scenarios are shown in Table 8. Under the natural
a certain extent, which can play a role in conserving water
growth scenario, ecological protection scenario, and econ-
resources and intercepting runoff (Xue et al. b). In
omic development scenario, runoff shows a decreasing
future planning, the proportion of ecological land (e.g.,
trend, but the degree of runoff reduction differs between
forest land and grassland) should be adjusted appropriately
the scenarios. Under the condition of natural growth, the
to alleviate the trend of decreasing runoff within the basin.
proportion of the forest area will increase, the proportion of the grassland area will decrease, and runoff will decrease in 2030 and 2050. Under the condition of ecological protec-
DISCUSSION
tion, the proportion of the forest area will still increase and According to the derived results, runoff in the Hailar Basin has shown a downward trend over the previous 30 years, Table 8
2030
2050
|
Change of runoff under different land-use scenarios
Land-use change scenarios
Rate of change in runoff (%)
Natural growth scenario Ecological protection scenario Economic development scenario
0.515 0.704 0.371
Natural growth scenario Ecological protection scenario Economic development scenario
0.942 1.408 0.152
with a more obvious downward trend since around 1990. A decrease in runoff in the basin is attributable to both decreased rainfall and changes of the underlying surface conditions (Xue et al. a; Zhao et al. a), e.g., increase of forest land and grassland areas, increase of forest land coverage rate, greater retention of surface runoff, and increase of surface evaporation in the basin (Wang et al. ). According to the results of the abrupt change test,
989
Y. Yan et al.
|
Climate and land-use impacts on runoff variation
Hydrology Research
|
51.5
|
2020
annual runoff in the basin experienced an abrupt change in
). Precipitation has a direct effect on runoff change,
1988. According to the UF curve, runoff has shown a decreas-
whereas the effect of temperature on runoff change is indir-
ing trend since 1988. Moreover, the UF curve exceeds the
ect (Fu et al. ; Li et al. b). The simulation scenarios
0.05 critical line after 2003, indicating that the trend of
of land-use change showed that in 2030 and 2050, under the
decrease is very significant. The main reason might be that
three different scenarios of natural development, ecological
climate change has caused rainfall in the basin to change
development, and economic development, runoff in the
abruptly (Fang et al. a). However, since China
basin will show a decreasing trend to differing degrees. The
implemented the policy of returning farmland to forest in
main reason might be an increase in the proportion of the
the 1990s, the rate of coverage of forest land in the basin
forest land area and a decrease in the proportion of the grass-
has increased, which will have increased interception by veg-
land area (Xue et al. a). An increase in the forest land
etation and delayed runoff (Wang et al. ; Xue et al. a).
area has the functions of conserving water and fixing soil,
Based on the determined catastrophe years, this study
which have a certain effect on intercepting runoff (Wang
considered annual runoff data during 1980–2012 as the
et al. ; Xu et al. ). Moreover, an increase in the
research object and divided the research period into the
forest land area leads to increased evapotranspiration,
reference period (before 1988) and the interference period
which can improve both the regional microclimate and the
(after 1988). By setting different scenarios, it was established
soil environment to a certain extent, reducing watershed
that the contribution rate of climate change (land-use
runoff (Wang et al. ; El Kateb et al. ). Compared
change) to runoff change in the basin was 83.58%
with land-use change, climate change has a more significant
(16.42%). The main land-use types in the upper reaches of
impact on runoff change in the Hailar Basin. To optimize
the Hailar Basin are grassland and forest land, and there
water resources allocation in the basin, the layout of land
has been little change in the underlying surface for many
use must be optimized to cope with the challenges brought
years. Moreover, the area is located in the grassland area
by climate change (Fang et al. ). It will be important to
of northern China, where the population density is low
consider changing the proportion of ecological land (e.g.,
and there is little impact from human activities (Xue et al.
forest land and grassland) to regulate and control runoff in
b; Wang et al. ). Consequently, given the global
the basin (Xue et al. b; Zhang et al. ).
increase in temperature and regional decrease of precipi-
Certain problems encountered in this research should
tation, climate change can be considered the main reason
be resolved in future work. First, this study used only
for runoff change in the Hailar Basin.
runoff data from 1980 to 2012 for the mutation analysis,
In order to further explore the runoff variations under
and the time series was incomplete. The next step would
climate change and land-use change, this paper sets up
be to collect a dataset that is more complete for specifying
different future scenarios for research based on the results
the mutation year. Second, the predictions of land use in
of runoff variation attribution analysis. In this study, differ-
2030 and 2050 were based on the CA–Markov model and
ent scenarios were established to explore the changes of
relied mainly on the transformation matrix derived using
watershed runoff under the effects of both climate change
land-use maps from 2005 to 2015. However, the implemen-
and land-use change. The climate change scenarios
tation of land-use management strategies by the Chinese
showed that runoff in the basin will increase with the
government might change in the future, which would have
increase of precipitation. For every 10% increase in precipi-
a certain impact on the land use/coverage scenarios pre-
tation, the runoff will increase by 13.34% on average. Runoff
dicted in this study for 2030 and 2050. Third, different
in the basin will decrease with an increase in air tempera-
combinations of temperature and rainfall were used to
ture. For every 1 C increase in air temperature, runoff will
drive the SWAT model. The climate change scenario was
decrease by 6.5% on average. Overall, the influence on
reasonably simple and did not start from actual conditions.
runoff of precipitation change is greater than that of temp-
In the next step, the climate forecast could be corrected
erature change. The main reason might be that runoff in
based on actual local climate conditions to provide the
this basin is replenished primarily by precipitation (A et al.
improved simulation of future runoff.
990
Y. Yan et al.
|
Climate and land-use impacts on runoff variation
Hydrology Research
CONCLUSIONS
|
51.5
|
2020
will be greater than the amount of runoff derived from snowmelt, which will lead to an overall reduction in
This study employed the SWAT distributed hydrological
runoff. A significant positive correlation was found
model to assess the hydrological response of the Hailar
between the change in precipitation and runoff
Basin to climate change and land-use change, and to analyze
change. Moreover, it was found that when precipitation
its impact quantitatively. First, the runoff change trend at the
increases
Bahou Station on the upstream reaches of the Hailar Basin
(decreases)
and
temperature
decreases
(increases), an increase (decrease) in runoff is greatest.
over the previous 30 years was analyzed, and the year of
(5) The projected impact of land use on runoff under differ-
abrupt change in runoff was determined. The SWAT
ent development scenarios was different. Runoff under
model was used to simulate runoff, and the contribution
the three development modes of natural development,
rates of climate change and land-use change to surface
ecological protection, and economic development
runoff were separated. Based on the principle of the water
showed a decreasing trend. In comparison with the
balance, the contributions of these two factors to runoff
natural development scenario, the reduction in runoff
were calculated. Using combinations of different climate
under ecological protection will be greater. Under the
and land-use scenarios, the natural runoff under the influ-
economic development scenario, runoff will still show
ence
a decreasing trend. It was established that forest land
of
climate
change
and
land-use
change
was
simulated. The derived results are presented below:
has a certain effect on decreasing runoff flow, i.e., an increase in the forest area will lead to decreased surface
(1) Annual runoff at the Bahou Station generally showed a
runoff. In future planning, the proportion of ecological
downward trend that exhibited diminishing volatility.
land (e.g., forest land and grassland) should be adjusted
The long-term runoff sequence changed abruptly in
appropriately to alleviate the trend of decreasing runoff
1988. This study divided the runoff sequence into the
within the basin.
reference period (before 1988) and the interference period (after 1988). (2) The SWAT model demonstrated satisfactory applica-
ACKNOWLEDGEMENTS
bility to simulating runoff at the Bahou Station. The R 2, Ens, and RE values of the monthly runoff simulated
This study was supported by the National Natural Science
in the calibration period were 0.75, 0.71, and 20.8%,
Foundation of China (Grant No. 31670451) and the
respectively, while the corresponding values in the vali-
Fundamental Research Funds for the Central Universities
dation period were 0.74, 0.70, and 20.3%, respectively,
(No. 2017NT18). We thank James Buxton MSc from
proving that the model was suitable both for the calcu-
Liwen Bianji, Edanz Group China (www.liwenbianji.cn./
lation of the runoff contribution rate and for the
ac), for editing the English text of this manuscript.
simulation of runoff under different scenarios. (3) Compared with land-use change, climate change was found to have a more significant impact on runoff
COMPETING FINANCIAL INTERESTS
change in the river basin. The contribution rate of climate change (land-use change) to runoff change in
The authors declare no competing financial interest.
the river basin was 83.58% (16.42%). (4) According to different combinations of temperature and rainfall change, temperature decrease will lead to evaporation and snowmelt runoff decrease, which will result in increased surface runoff in the watershed. A rise in temperature will increase both snowmelt runoff and evaporation; however, the amount of evaporation
REFERENCES A, Y., Wang, G., Sun, W., Xue, B. & Kiem, A. Stratification response of soil water content during rainfall events under different rainfall patterns. Hydrological Processes 32, 3128–3139.
991
Y. Yan et al.
|
Climate and land-use impacts on runoff variation
A, Y., Wang, G., Liu, T., Shrestha, S., Xue, B. & Tan, Z. a Vertical variations of soil water and its controlling factors based on the structural equation model in a semi-arid grassland. Science of the Total Environment 691, 1016–1026. A, Y., Wang, G., Liu, T., Xue, B. & Kuczera, G. b Spatial variation of correlations between vertical soil water and evapotranspiration and their controlling factors in a semiarid region. Journal of Hydrology 574, 53–63. Anoh, K. A., Koua, T. J. J., Kouame, K. J., Jourda, J. P. & Laurent, F. Modelling water flow in a complex watershed in humid a tropical area using SWAT: a case study of Taabo watershed in Ivory Coast. International Journal of River Basin Management 16, 157–167. Arnold, J. G., Srinivasan, R., Muttiah, R. S. & Williams, J. R. Large area hydrologic modeling and assessment – part 1: model development. Journal of the American Water Resources Association 34, 73–89. Arnold, J. G., Moriasi, D. N., Gassman, P. W., Abbaspour, K. C., White, M. J., Srinivasan, R., Santhi, C., Harmel, R. D., van Griensven, A., Van Liew, M. W., Kannan, N. & Jha, M. K. SWAT: model use, calibration, and validation. Transactions of the ASABE 55, 1491–1508. Asl-Rousta, B., Mousavi, S. J., Ehtiat, M. & Ahmadi, M. SWAT-based hydrological modelling using model selection criteria. Water Resources Management 32, 2181–2197. Bao, Z., Zhang, J., Wang, G., Fu, G., He, R., Yan, X., Jin, J., Liu, Y. & Zhang, A. Attribution for decreasing streamflow of the Haihe River basin, northern China: climate variability or human activities? Journal of Hydrology 460, 117–129. Barnett, T. P., Adam, J. C. & Lettenmaier, D. P. Potential impacts of a warming climate on water availability in snowdominated regions. Nature 438, 303–309. Bhatta, B., Shrestha, S., Shrestha, P. K. & Talchabhadel, R. Evaluation and application of a SWAT model to assess the climate change impact on the hydrology of the Himalayan River Basin. Catena 181, 104082. Chi, D., Wang, H., Li, X., Liu, H. & Li, X. Estimation of the ecological water requirement for natural vegetation in the Ergune River basin in Northeastern China from 2001 to 2014. Ecological Indicators 92, 141–150. Cuo, L., Zhang, Y., Gao, Y., Hao, Z. & Cairang, L. The impacts of climate change and land cover/use transition on the hydrology in the upper Yellow River Basin, China. Journal of Hydrology 502, 37–52. Dittmer, K. Changing streamflow on Columbia basin tribal lands – climate change and salmon. Climatic Change 120, 627–641. El Kateb, H., Zhang, H., Zhang, P. & Mosandl, R. Soil erosion and surface runoff on different vegetation covers and slope gradients: a field experiment in Southern Shaanxi Province, China. Catena 105, 1–10. Fang, J. Y., Chen, A. P., Peng, C. H., Zhao, S. Q. & Ci, L. Changes in forest biomass carbon storage in China between 1949 and 1998. Science 292, 2320–2322.
Hydrology Research
|
51.5
|
2020
Fang, Q., Wang, G., Liu, T., Xue, B.-L. & A, Y. a Controls of carbon flux in a semi-arid grassland ecosystem experiencing wetland loss: vegetation patterns and environmental variables. Agricultural and Forest Meteorology 259, 196–210. Fang, Q., Wang, G., Xue, B., Liu, T. & Kiem, A. b How and to what extent does precipitation on multi-temporal scales and soil moisture at different depths determine carbon flux responses in a water-limited grassland ecosystem? Science of the Total Environment 635, 1255–1266. Fang, Q., Wang, G., Liu, T., Xue, B., Sun, W. & Shrestha, S. Unraveling the sensitivity and nonlinear response of water use efficiency to the water-energy balance and underlying surface condition in a semiarid basin. Science of the Total Environment 699, 134405. Fu, Z., Jiang, H., Wang, G., A, Y., Xue, B. & Wang, H. Effects of soil properties on plant community structure in a semi-arid grassland. Chinese Journal of Ecology 37, 823–830. Gampe, D., Nikulin, G. & Ludwig, R. Using an ensemble of regional climate models to assess climate change impacts on water scarcity in European river basins. Science of the Total Environment 573, 1503–1518. Hamed, K. H. Trend detection in hydrologic data: the MannKendall trend test under the scaling hypothesis. Journal of Hydrology 349, 350–363. Hamed, K. H. & Rao, A. R. A modified Mann-Kendall trend test for autocorrelated data. Journal of Hydrology 204, 182–196. Han, D., Wang, G., Liu, T., Xue, B.-L., Kuczera, G. & Xu, X. a Hydroclimatic response of evapotranspiration partitioning to prolonged droughts in semiarid grassland. Journal of Hydrology 563, 766–777. Han, D., Wang, G., Xue, B., Liu, T., A, Y. & Xu, X. b Evaluation of semiarid grassland degradation in North China from multiple perspectives. Ecological Engineering 112, 41–50. Ishida, K., Ercan, A., Trinh, T., Kavvas, M. L., Ohara, N., Carr, K. & Anderson, M. L. Analysis of future climate change impacts on snow distribution over mountainous watersheds in Northern California by means of a physically-based snow distribution model. Science of the Total Environment 645, 1065–1082. Jothityangkoon, C., Sivapalan, M. & Farmer, D. L. Process controls of water balance variability in a large semi-arid catchment: downward approach to hydrological model development. Journal of Hydrology 254, 174–198. Kaushal, S. S., Groffman, P. M., Band, L. E., Elliott, E. M., Shields, C. A. & Kendall, C. Tracking nonpoint source nitrogen pollution in human-impacted watersheds. Environmental Science & Technology 45, 8225–8232. Leavesley, G. H. Modeling the effects of climate change on water resources – a review. Climatic Change 28, 159–177. Legesse, D., Vallet-Coulomb, C. & Gasse, F. Hydrological response of a catchment to climate and land use changes in Tropical Africa: case study South Central Ethiopia. Journal of Hydrology 275, 67–85.
992
Y. Yan et al.
|
Climate and land-use impacts on runoff variation
Li, Z., Liu, W.-z., Zhang, X.-c. & Zheng, F.-l. Impacts of land use change and climate variability on hydrology in an agricultural catchment on the Loess Plateau of China. Journal of Hydrology 377, 35–42. Li, X., Cheng, G., Ge, Y., Li, H., Han, F., Hu, X., Tian, W., Tian, Y., Pan, X., Nian, Y., Zhang, Y., Ran, Y., Zheng, Y., Gao, B., Yang, D., Zheng, C., Wang, X., Liu, S. & Cai, X. Hydrological cycle in the Heihe River Basin and its implication for water resource management in endorheic basins. Journal of Geophysical Research – Atmospheres 123, 890–914. Li, H., Shi, C., Zhang, Y., Ning, T., Sun, P., Liu, X., Ma, X., Liu, W. & Collins, A. L. a Using the Budyko hypothesis for detecting and attributing changes in runoff to climate and vegetation change in the soft sandstone area of the middle Yellow River basin, China. Science of the Total Environment 703, 135588. Li, Q. , Wang, G., Wang, H., Shrestha, S., Xue, B-L., Sun, W. & Yu, J. b Macrozoobenthos variations in shallow connected lakes under the influence of intense hydrologic pulse changes. Journal of Hydrology 584, 124755, http://doi.org/ 10.1016/j.jhydrol.2020.124755. Lorup, J. K., Refsgaard, J. C. & Mazvimavi, D. Assessing the effect of land use change on catchment runoff by combined use of statistical tests and hydrological modelling: case studies from Zimbabwe. Journal of Hydrology 205, 147–163. Machiwal, D. & Jha, M. K. Comparative evaluation of statistical tests for time series analysis: application to hydrological time series. Hydrological Sciences Journal 53, 353–366. Mango, L. M., Melesse, A. M., McClain, M. E., Gann, D. & Setegn, S. G. Land use and climate change impacts on the hydrology of the upper Mara River Basin, Kenya: results of a modeling study to support better resource management. Hydrology and Earth System Sciences 15, 2245–2258. Milly, P. C. D., Dunne, K. A. & Vecchia, A. V. Global pattern of trends in streamflow and water availability in a changing climate. Nature 438, 347–350. Mishra, A. K. & Singh, V. P. A review of drought concepts. Journal of Hydrology 391, 204–216. Mitsova, D., Shuster, W. & Wang, X. A cellular automata model of land cover change to integrate urban growth with open space conservation. Landscape and Urban Planning 99, 141–153. Moriasi, D. N., Arnold, J. G., Van Liew, M. W., Bingner, R. L., Harmel, R. D. & Veith, T. L. Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Transactions of the ASABE 50, 885–900. Piao, S., Fang, J., Ciais, P., Peylin, P., Huang, Y., Sitch, S. & Wang, T. The carbon balance of terrestrial ecosystems in China. Nature 458, 1009–1013. Piao, S., Ciais, P., Huang, Y., Shen, Z., Peng, S., Li, J., Zhou, L., Liu, H., Ma, Y., Ding, Y., Friedlingstein, P., Liu, C., Tan, K., Yu, Y., Zhang, T. & Fang, J. The impacts of climate
Hydrology Research
|
51.5
|
2020
change on water resources and agriculture in China. Nature 467, 43–51. Sang, L., Zhang, C., Yang, J., Zhu, D. & Yun, W. Simulation of land use spatial pattern of towns and villages based on CAMarkov model. Mathematical and Computer Modelling 54, 938–943. Sterling, S. M., Ducharne, A. & Polcher, J. The impact of global land-cover change on the terrestrial water cycle. Nature Climate Change 3, 385–390. van Griensven, A., Meixner, T., Grunwald, S., Bishop, T., Diluzio, A. & Srinivasan, R. A global sensitivity analysis tool for the parameters of multi-variable catchment models. Journal of Hydrology 324, 10–23. Voeroesmarty, C. J., McIntyre, P. B., Gessner, M. O., Dudgeon, D., Prusevich, A., Green, P., Glidden, S., Bunn, S. E., Sullivan, C. A., Liermann, C. R. & Davies, P. M. Global threats to human water security and river biodiversity. Nature 467, 555–561. Wang, G. & Xu, Z. Assessment on the function of reservoirs for flood control during typhoon seasons based on a distributed hydrological model. Hydrological Processes 25, 2506–2517. Wang, G., Song, J., Xue, B.-L., Xu, X. & Otsuki, K. Land use and land cover change of Hulun Lake Nature Reserve in Inner Mongolia, China: a modeling analysis. Journal of the Faculty of Agriculture Kyushu University 57, 219–225. Wang, G., Yang, H., Wang, L., Xu, Z. & Xue, B. Using the SWAT model to assess impacts of land use changes on runoff generation in headwaters. Hydrological Processes 28, 1032–1042. Wang, Z., Lai, C., Chen, X., Yang, B., Zhao, S. & Bai, X. Flood hazard risk assessment model based on random forest. Journal of Hydrology 527, 1130–1141. Wang, Z., Deng, X., Song, W., Li, Z. & Chen, J. What is the main cause of grassland degradation? A case study of grassland ecosystem service in the middle-south Inner Mongolia. Catena 150, 100–107. Wang, G., Hu, X., Zhu, Y., Jiang, H. & Wang, H. a Historical accumulation and ecological risk assessment of heavy metals in sediments of a drinking water lake. Environmental Science and Pollution Research 25, 24882–24894. Wang, W., Fang, Q., Wang, G., Li, R., Xue, B. & Wang, H. b Simulation of CO2 source, sink, and flux temporal and spatial distributions in Hulun Buir grassland. Acta Ecologica Sinica 38, 7288–7299. Wang, G., Li, J., Sun, W., Xue, B., A, Y. & Liu, T. a Non-point source pollution risks in a drinking water protection zone based on remote sensing data embedded within a nutrient budget model. Water Research 157, 238–246. Wang, G., Liu, S., Liu, T., Fu, Z., Yu, J. & Xue, B. b Modelling above-ground biomass based on vegetation indexes: a modified approach for biomass estimation in semi-arid
993
Y. Yan et al.
|
Climate and land-use impacts on runoff variation
grasslands. International Journal of Remote Sensing 40, 3835–3854. Wheater, H. & Evans, E. Land use, water management and future flood risk. Land Use Policy 26, S251–S264. Xu, X., Yang, D., Yang, H. & Lei, H. Attribution analysis based on the Budyko hypothesis for detecting the dominant cause of runoff decline in Haihe basin. Journal of Hydrology 510, 530–540. Xue, B.-L., Wang, L., Li, X., Yang, K., Chen, D. & Sun, L. a Evaluation of evapotranspiration estimates for two river basins on the Tibetan Plateau by a water balance method. Journal of Hydrology 492, 290–297. Xue, B.-L., Wang, L., Yang, K., Tian, L., Qin, J., Chen, Y., Zhao, L., Ma, Y., Koike, T., Hu, Z. & Li, X. b Modeling the land surface water and energy cycles of a mesoscale watershed in the central Tibetan Plateau during summer with a distributed hydrological model. Journal of Geophysical Research – Atmospheres 118, 8857–8868. Xue, B.-L., Guo, Q., Otto, A., Xiao, J., Tao, S. & Li, L. a Global patterns, trends, and drivers of water use efficiency from 2000 to 2013. Ecosphere 6, 174. Xue, B.-L., Li, Z., Yin, X.-A., Zhang, T., Iida, S. i., Otsuki, K., Ohta, T. & Guo, Q. b Canopy conductance in a two-storey Siberian boreal larch forest, Russia. Hydrological Processes 29, 1017–1026. Xue, B.-L., Guo, Q., Hu, T., Wang, G., Wang, Y., Tao, S., Su, Y., Liu, J. & Zhao, X. a Evaluation of modeled global vegetation carbon dynamics: analysis based on global carbon flux and above-ground biomass data. Ecological Modelling 355, 84–96. Xue, B.-L., Guo, Q., Hu, T., Xiao, J., Yang, Y., Wang, G., Tao, S., Su, Y., Liu, J. & Zhao, X. b Global patterns of woody residence
Hydrology Research
|
51.5
|
2020
time and its influence on model simulation of aboveground biomass. Global Biogeochemical Cycles 31, 821–835. Yang, J., Reichert, P., Abbaspour, K. C., Xia, J. & Yang, H. Comparing uncertainty analysis techniques for a SWAT application to the Chaohe Basin in China. Journal of Hydrology 358, 1–23. Yue, S., Pilon, P. & Cavadias, G. Power of the Mann-Kendall and Spearman’s rho tests for detecting monotonic trends in hydrological series. Journal of Hydrology 259, 254–271. Zhang, M., Liu, N., Harper, R., Li, Q., Liu, K., Wei, X., Ning, D., Hou, Y. & Liu, S. A global review on hydrological responses to forest change across multiple spatial scales: importance of scale, climate, forest type and hydrological regime. Journal of Hydrology 546, 44–59. Zhao, F. F., Xu, Z. X., Huang, J. X. & Li, J. Y. Monotonic trend and abrupt changes for major climate variables in the headwater catchment of the Yellow River basin. Hydrological Processes 22, 4587–4599. Zhao, G., Tian, P., Mu, X., Jiao, J., Wang, F. & Gao, P. a Quantifying the impact of climate variability and human activities on streamflow in the middle reaches of the Yellow River basin, China. Journal of Hydrology 519, 387–398. Zhao, Y., Zou, X., Zhang, J., Cao, L., Xu, X., Zhang, K. & Chen, Y. b Spatio-temporal variation of reference evapotranspiration and aridity index in the Loess Plateau Region of China, during 1961–2012. Quaternary International 349, 196–206. Zhou, Y., Lai, C., Wang, Z., Chen, X., Zeng, Z., Chen, J. & Bai, X. Quantitative evaluation of the impact of climate change and human activity on runoff change in the Dongjiang River Basin, China. Water 10, 511.
First received 22 February 2020; accepted in revised form 3 April 2020. Available online 13 May 2020
994
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Analysis for spatial-temporal matching pattern between water and land resources in Central Asia Ying Zhang, Zhengxiao Yan, Jinxi Song, Anlei Wei, Haotian Sun and Dandong Cheng
ABSTRACT Central Asia, the pioneering place of the ‘Belt and Road’, is under the threat of prominent water issues. Based on the Gini coefficient model and the matching index, the amount of the total renewable water resources and the cultivated land area were introduced to evaluate the matching pattern between the water and land resources in Central Asia. The water problem of Kazakhstan, being the most prominent, shows low water resources per unit area with the highest reclamation rate. The matching degree for the upstream countries of the Aral Sea (Kyrgyzstan, Tajikistan) was better than those of the downstream countries (Turkmenistan, Uzbekistan, Kazakhstan). The Gini coefficient in Central Asia was 0.32, smaller than that of the global average value (0.59). The overall water available for use and the matching cultivated land resources was reasonable. Large differences exist in the matching degree in water distribution and utilization among Central Asian countries. The matching index of water and land resources in Central Asia was 1.25, similar to the matching degree estimated from the Gini coefficient model. Moreover, rational measures are suggested to alleviate the issue of water and land resources matching in Central Asia. Key words
Ying Zhang Zhengxiao Yan Jinxi Song (corresponding author) Anlei Wei Haotian Sun Dandong Cheng Shaanxi Key Laboratory of Earth Surface System and Environmental Carrying Capacity, College of Urban and Environmental Sciences, Northwest University, Xi’an 710127, China E-mail: jinxisong@nwu.edu.cn Jinxi Song State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau, Institute of Soil and Water Conservation, Chinese Academy of Sciences and Ministry of Water Resources, Yangling 712100, China
| ArcGIS, spatial and temporal pattern, water and land resources matching, water distribution and utilization
HIGHLIGHTS
• • • • •
The Gini coefficient model is to evaluate the overall level of matching degree. The matching index model is introduced for evaluating the spatial matching pattern. The matching situation for the upstream countries of the Aral Sea was better than those of the downstream countries. Two models perform similarly in matching pattern. The matching index of water and land resources in Central Asia was 1.25.
INTRODUCTION Central Asia, located in the hinterland of the Eurasian con-
world (Unger-Shayesteh et al. ). The economic benefits
tinent, lacks water resources (Gafurov et al. ). It is one of
of water use in Central Asia are lower than those in other
the regions threatened by serious water problems in the
areas of Asia (Lee & Jung ). Although the total amount of water resources is seemingly large in this
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying,
region, in fact, only 24.4% is available for use by humans
adaptation and redistribution, provided the original work is properly cited
(Yang et al. ). On top of that, multiple transboundary
(http://creativecommons.org/licenses/by/4.0/).
rivers exist in Central Asia (Zhupankhan et al. ).
doi: 10.2166/nh.2020.177
995
Y. Zhang et al.
|
Matching pattern between water and land resources in Central Asia
Hydrology Research
|
51.5
|
2020
However, the mismatched spatial distributions of water and
conflicts. The correct evaluation of the formation and con-
land resources, along with the intense human activities (e.g.,
sumption of water resources and its spatial matching with
overexploitation of water resources), has ultimately led to
land resources is critical to alleviating water problems
the serious water crisis in Central Asia’s river basins. This
(Yang et al. ). Most of the current state-of-the-art studies
is the main reason for ongoing water conflicts in the region’s
have focused on the water problems in Central Asia from the
transboundary rivers and the ecological disaster of the Aral
aspects of the state of water resources, the water cycle pro-
Sea (Chen et al. ).
cesses, the transboundary river management, and so on
Kyrgyzstan and Tajikistan, located in the upstream of
(Yang et al. ). Although previous studies exist for quan-
the Aral Sea Basin, have the most abundant water
titative analysis, they mainly focus on the impact of
resources and are called the ‘water towers’ of Central
climate change on the water cycle and the water environ-
Asian countries. However, due to insufficient water conser-
ment of the basin (Zhu et al. ). However, there is a
vancy engineering measures, water resources have flowed
lack of study on the matching patterns between water and
into downstream countries, so the amount of available
land resources. Therefore, a quantitative study on spatial dis-
water resources in upstream countries is limited. Mean-
tribution difference and the matching pattern change of
while, the downstream countries of the Aral Sea Basin
water and land resources is the basis of a spatial optimal
(Kazakhstan, Turkmenistan, and Uzbekistan) enjoy more
allocation of regional water resources, offering important
available water resources, which flow from upstream
guidance for efficient utilization of water and land resources
countries, than their actual domestic water storage (Deng
in Central Asia.
et al. ). Owing to the limited water storage, the down-
Spatial-temporal dynamics of water and land resources
stream countries are also water-deficient, especially the
matching have attracted many researchers’ attention. Most
oil-rich Turkmenistan, which has the saying that ‘water is
researchers have studied the overall matching of water and
more expensive than oil’.
land resources in the region on the macro-scale of time
In the former Soviet Union era, the upstream countries
and space. Liu et al. () constructed the matching analy-
focused on the construction of water conservancy facilities,
sis model of agricultural water and land resources using the
providing downstream countries with water resources and
total water resources as a parameter and evaluated the
power resources such as farmland irrigation water, while
matching pattern of agricultural water and land resources
downstream countries provided more cultivated land and
in Northeast China. Liu et al. () established the Gini
agricultural products for the upstream (Jalilov et al. ).
coefficient model for agricultural water and land resources
Since the independence of Central Asian countries, a host
allocation and the model for measuring the matching
of international projects has been launched to achieve sus-
index of the broadly defined agricultural water resources
tainable water management, but only a few have been
and land resources (MIBAWRLR) to assess the matching
implemented (Abdullaev et al. ). Uncompensated
degree of the agricultural water and land of the Jiansanjiang
water use patterns for downstream countries have caused
Administration and 15 farms in Heilongjiang Province,
dissatisfaction in upstream countries. The issue of ‘energy-
China. Based on the water and land resources and agricul-
water-irrigation’ has been plaguing relations between Cen-
tural output value, Huang et al. () used the data
tral Asian countries (Bernauer & Siegfried ). Water
envelopment analysis (DEA) model to analyze the situation
and land resources, as the basic materials for production
of water and land resources in Sichuan Province, China.
and life, are the basis for ensuring national food security
Only a few researchers analyzed the equilibrium state of
(Zhang et al. a). For the five Central Asian countries
supply and demand for agricultural land and water
dominated by irrigated agriculture, the shortage of water
resources at the micro-scale. Zhang et al. (b) con-
and land resources has delayed the process of agricultural
structed a water and land resources matching index
modernization (Riquelme & Ramos ). Generally speak-
calculation model in micro-scale and studied the temporal
ing, the mismatching between the formation area and
and spatial dynamics of crop water and land resource
consumption area of water resources will lead to water
matching in the Naoli River Basin, China.
996
Y. Zhang et al.
|
Matching pattern between water and land resources in Central Asia
The matching of water and land resources can be calcu-
Hydrology Research
|
51.5
|
2020
MATERIALS AND METHODS
lated from three aspects: calculating the matching index of water and land resources with the total water resources
Description of study area
volume per hectare as the index; evaluating the overall matching situation of water and land resource in the
Central Asia refers to the eastern part of the Caspian Sea,
region by the Gini coefficient model, based on DEA to
south of Western Siberia, north of Afghanistan and Central
analyze the matching characteristics of water and land
Asia region to the west of Xinjiang, including Kazakhstan,
resources. Although these models are widely used to study
Uzbekistan,
the dynamics of regional water and land resources match-
(56 340 E–87 190 E, 35 080 N–55 260 N). With a total area of
ing, each model has its shortcomings. For the DEA model,
nearly 4.00 × 108 hm2 (excluding coastal waters) and a total
the water and land resources are used as the input indicators
population of about 7.08 × 107 in 2017, Central Asia is one
and the agricultural output as the output index to study the
of the most sparsely populated areas in the world today.
matching degree of water and land resources (Huang et al.
The total renewable water resources in this area are 2.28 ×
). However, multiple factors have been found to affect
1011 m3 and the cultivated area is 3.84 × 107 hm2. The rivers
agricultural output value: the amount of agricultural
in Central Asia are mostly inland rivers and lakes, mainly
equipment, labor force, water and land resources, etc.
including the Amu Darya, the Syr River, the Chu River, the
Therefore, the results of the DEA model are not satisfactory
Aral Sea, and so on (Figure 1). Its terrain is high in the east
when only water resources and land resources are taken as
and low in the west, dominated by plains and hills. The
research variables. For the Gini coefficient model, different
total area of drylands is 5.17 × 106 km2 (above 80% of the
division degrees of units in the region can yield different
global total temperate desert area) in Central Asia, belonging
results. Meanwhile, it can only reflect the overall matching
to the temperate desert climate and grassland continental cli-
situation of water and land resources in the study area and
mate (Li et al. ). The vegetation is mainly grassland and
the relative matching of each unit in the region. For the
desert. In most areas, the climate is dry with strong evapor-
matching index model, although it reveals the overall match-
ation (Yao et al. ). The annual precipitation is below
ing status of water and land resources in the region, the
300 mm, and the annual precipitation in the desert near the
Kyrgyzstan,
Tajikistan,
and
Turkmenistan
spatial and temporal distribution and dynamic character-
Aral Sea and Turkmenistan is only 75–100 mm. The territory
istics of water and land resources matching at the micro-
area of the five Central Asian countries varies greatly, with
scale are still unclear. At the same time, the model does
Kazakhstan’s territory being vast, twice the total area of the
not take into account the differences in the use of water
other four countries. However, although Uzbekistan has a
resources between different types of cultivated land use.
total area of only one-sixth of Kazakhstan, its population is
However, this model can make up for the errors caused by
nearly 40% greater than that of Kazakhstan.
the division of the unit based on the Gini coefficient, reflect the relative space–time ratio of water and land resources matching. To achieve comprehensive results, we use the
Data sources
Gini coefficient model and the matching index model to analyze the matching situation of water and land resources
Since the data of water and land resources in Central Asia
in Central Asia.
are not continuous, this paper uses the data of 1992, 1997,
Due to the complexity of transboundary water resources in
2002, 2007, 2012, and 2016 from the Food and Agriculture
Central Asia, this paper takes the five countries of Central Asia
Organization of the United Nations (FAO a). The culti-
as the basic unit, then uses total renewable water resource and
vated area is used as the parameter of cultivated land
total water withdrawal as the water resource parameters, and
resources, including arable land and permanent crops. For
the cultivated area as the cultivated land resource parameter
water resources, the amount of agricultural water, total
to study the spatial–temporal matching relationship of water
water withdrawal, and renewable water resources are
and land resources in five Central Asian countries.
used. The renewable water resource includes the amount
997
Figure 1
Y. Zhang et al.
|
|
Matching pattern between water and land resources in Central Asia
Hydrology Research
|
51.5
|
2020
Research site map of Central Asia.
of surface water and groundwater available in the country.
volume per hectare. It is a quantitative relationship that
The land cover types are obtained from the global land
reflects the spatial and temporal distribution of water
cover data product with a resolution of 500 m (MCD12Q1
resources and cultivated land resources available for agri-
data set) of the Land Process Distributed Active Archive
cultural production in a specific region. It focuses on
Center (LPDAAC ). The area of cropland, grassland,
regional water resources and cultivated land, and coordi-
and forest land are collected by the FAO (FAO b).
nation and suitability of regional water resources and cultivated land resources in space–time distribution (Liu
Theory of water and land resource matching
et al. ). The higher the matching degree, the better the fundamental conditions of agricultural production are.
Matching index model of water and land resources
2. Calculating model: The matching level of water resources and cultivated areas with five countries in Central Asia
1. Basic concept: The matching index of water and land resources is expressed by the total water resources
are selected as the basic units. The agricultural water withdrawal
is
determined
by
the
proportion
of
998
Y. Zhang et al.
|
Matching pattern between water and land resources in Central Asia
Hydrology Research
|
51.5
|
2020
agricultural water used to the whole water consumption
Smaller G value indicates relatively even distribution,
(agricultural water, industrial water, and municipal
while larger G value indicates relatively uneven distribution.
water) in Central Asia (Liu et al. ). The calculation formula is as follows:
Because the spatial distribution of water resources is diversified, the Gini coefficient is also applicable in the study of the patterns of water and land resource matching.
Rwl i ¼ Wi αi =Li ,
(i ¼ 1, 2, 3, 4, 5)
(1)
According to the spatial distribution characteristics of water and land resources, the Gini curve construction
is the matching index of water and land where Rwl i resources of the ith country in Central Asia; Wi is the
model for the matching problem of water and land resources in Central Asia is as follows (Liu et al. ):
total renewable water resource of the ith country, 108 m3; Li is the area of the cultivated area of the ith country, 104 hm2; αi is the proportion of agricultural water in the ith country to the total amount of social and economic water. The matching index of water and land resources in Central Asia reflects the mean value of the matching index of water and land resources among countries in Central Asia. The calculation formula is as follows:
1. Take the five countries in Central Asia as the basic calculation units, and calculate the total water resources volume per hectare at each country, then according to the size of this index, the five Central Asian countries are ranked in ascending order. 2. Calculate the percentage of total water resources and cultivated land resources in each unit. 3. The cumulative proportion of water and land resources in each country is calculated following step 1.
Rpj ¼
n X
Rwl i =n
(2)
i¼1
4. Draw XY scatter plots, the X coordinate being the cumulative proportion of the cultivated area of each unit, and the Y coordinate is the cumulative proportion
is the matching index of water and land
of the water resources parameters of each unit, then
resources in area j; Rwl i is the matching index of water
the Lorenz curve of water and land resources is
where
Rpj
and land resources of the ith country in area j; and n is the number of i countries in area j.
constructed. 5. Obtain the area of the figure between the Lorentz curve and the slash with an angle of 45 , and an integral
Gini coefficient model
between 0 and 1. Then, the Gini coefficient is twice the figure area.
The Gini coefficient was proposed by Italian economist Corrado Gini in 1921 when studying income inequality. It has been recognized and applied in many fields. The Gini
If the G value is between 0 and 1, and the smaller the G
coefficient (G), also known as the Lorentz coefficient, is
value is, the higher the degree of matching. When G ¼ 0, the
determined by calculating the area of the Lorentz curve
Lorenz curve completely coincides with the 45 line, and
graph using MATLAB software. First, the Lorentz curve
the water and soil resources are completely matched. On
needs to be fitted. Assuming that the area between the
the contrary, when G ¼ 1, the matching degree of water
actual distribution curve of income and the absolute equal
and land resources is extremely poor, that is, the cultivated
distribution curve is A, and the area between the actual dis-
land resources in this area are very rich, and the water
tribution curve of income and the coordinate axis is B, then
resources are in serious shortage.
the Gini coefficient is calculated as follows:
The internationally accepted model for dividing the Gini coefficient is as follows. For example, G ∈ (0, 0.2) indicates
G¼
A AþB
(3)
that the degree of matching between water resources and land resources is high (Table 1).
Y. Zhang et al.
999
Table 1
|
|
Matching pattern between water and land resources in Central Asia
Hydrology Research
|
51.5
|
2020
higher than domestic water resources. Kyrgyzstan and Taji-
Gini coefficient division standard (Dorfman 1979)
G
Matching results
(0,0.2)
High
[0.2,0.3)
Consistent
[0.3,0.4)
Reasonable
[0.4,0.5)
Inconsistent
[0.5,1)
Poor
kistan are net exit water resources countries, of which, Kyrgyzstan’s net exit water volume accounts for 76% of domestic water resources. The climate in Central Asia is highly arid with annual precipitation generally below 300 mm. Due to topography and climate, upstream Kyrgyzstan and Tajikistan are the main water sources in Central Asia, with an exit water volume of 76 billion m3, but the main water users are Uzbekistan and Turkmenistan.
RESULTS The structure and spatial distribution of water and land
Composition and utilization of land resources
resources in Central Asia The total area of Central Asia is nearly 4.00 × 108 hm2. Composition and utilization of water resources
Regarding land types, grassland is the largest proportion (279.5 × 104 km2) and bare areas the second (67.7 ×
The total freshwater in the five Central Asian countries is 3
above 1 trillion m , but the real available water resources 3
104 km2). Nearly half of the areas show a natural landscape of desert and semi-desert, and the vegetation coverage in the
are only 206 billion m , of which, surface water is 187 billion
north and southeast is higher (Figure 2). The highest pro-
m3 and the non-repetitive groundwater is 19 billion m3
portions of cropland, forest land, and grassland are in
(Table 2). Kazakhstan has the largest water resources,
Uzbekistan (11.21%), Turkmenistan (8.78%), and Kazakh-
accounting for 36.69% of Central Asia. Turkmenistan
stan (69.44%), respectively (Figure 3). Owing to the
has the least (0.68%). Agricultural water consumption
disintegration of the former Soviet Union, large-scale culti-
accounted for the largest proportion of social and economic
vation was abandoned in Central Asia. Subsquently, with
water use, with the proportion being 66%, 91%, 93%, 90%,
the independence of Central Asian countries, the social
and 94% in Kazakhstan, Tajikistan, Kyrgyzstan, Uzbekistan,
economy gradually recovered, and so did the cultivated
and Turkmenistan, respectively (FAO a). The five Cen-
areas. Therefore, the cultivated area and crop yield in Cen-
tral Asian countries have numerous transboundary rivers
tral Asia showed a trend of rapid decline first and then a
with large stream flow. Notably, Kazakhstan has a net
slow rise (Kienzler et al. ).
entry water volume of 34.2 billion m3, followed by Uzbeki-
The production potential of land resources in Central
stan. The net entry water volume of Turkmenistan is much
Asia is huge, but it is constrained by water resources.
Table 2
|
Water resources assessment in the five Central Asia countries (FAO 2016a; United Nations 1998) Average precipitation
Overlap between Surface
Groundwater/
surfaces water and
Water resources/ Entry-exit water
Available water
Water resources
Country in depth/mm
water/108 m3
108 m3
groundwater/108 m3
108 m3
volume/108 m3
resources/108 m3
per capita/m3
KAZ
693
161
100
754
342
1,096
4,142
250
KGZ
533
441
136
112
465
259
206
7,692
TJK
691
638
60
30
668
508
160
7,488
TKM
161
10
4
0
14
233
247
243
UZB
206
95
88
20
163
341
504
511
1,877
449
262
2,064
149
1,117
4,015.2
Sum
Y. Zhang et al.
1000
Figure 2
|
|
Matching pattern between water and land resources in Central Asia
Hydrology Research
|
51.5
|
2020
The types of land cover in Central Asia in 2015.
Meanwhile, a series of ecological and environmental pro-
(Figure 4(b)). Moreover, the spatial distribution of water
blems in the process of land use seriously restrict the
resources and cultivated land resources in Central Asia is
sustainable use of land resources in Central Asia. Effective
inconsistent (Figure 4). Kazakhstan is the most prominent
approaches to these problems are the key to maintaining
country, whose water resources per unit area is the lowest,
sustainable development in Central Asia.
but the reclamation rate is the highest in the region. This spatial dislocation of water and land resources in Central
Spatial distribution of water and land resources
Asia severely limits the increase of regional grain production and sustainable use of resources, which in return, exacer-
The water resources per unit area can reflect the richness
bates the contradiction of water problems in Central Asia.
and shortage of regional water resources, and the reclamation rate can reflect the degree of reclamation of
Matching pattern of water and land resources in Central
regional cultivated land resources. Five countries in Central
Asia
Asia are taken as the basic regional units, then the natural breakpoint method of GIS is used to divide the Central Asian region into five grades according to the water
Matching degree of water and land resources based on the Gini coefficient
resources per unit area and the reclamation rate, producing the spatial distribution pattern of water and land resources
The total renewable water resource is taken as the water
(Figure 4).
resource parameter, and the cultivated land area is used as
The spatial distribution of water resources in Central
the land resource parameter. The Gini coefficient of Central
Asia is not uniform and is characterized by a larger
Asia in 1992, 1997, 2002, 2007, 2012, and 2016 is shown in
amount of water resources in Kazakhstan and a lower
Figure 5.
amount of water resources in Kyrgyzstan (Figure 4(a)). The
The Gini coefficient showed little variation in 1992,
distribution of cultivated land in Central Asia is uneven
1997, 2002, 2007, 2012, and 2016, and the matching
Y. Zhang et al.
1001
|
Matching pattern between water and land resources in Central Asia
Hydrology Research
|
51.5
|
2020
Spatial matching pattern of water and land resources based on the matching index model Since potential errors may exist in the unit division of the Gini coefficient, the matching index model can be used to study the matching situation of water and land resources. Based on the data of total renewable water resource, cultivated area, and proportion of agricultural water in five Central Asian countries in 1992, 1997, 2002, 2007, 2012, and 2016, the water and land resources matching model is Figure 3
|
Proportion of land types in Central Asia in 2016.
used to obtain the matching index of water and land resources in Central Asian countries, as shown in Table 3. To visually indicate the difference in the matching situ-
situation of water and land resources in Central Asia is
ation of water and land resources, the matching index of
relatively reasonable (Figure 5(a)–5(f)). The average Gini
water and land resources in 2016 is divided into five cat-
coefficient of Central Asia in 1992–2016 was 0.32. Accord-
egories according to the Natural Breaks proposed by
ing to the internationally recognized Gini coefficient
Jenks: Poor matching (0.001< R 0.242), Moderate match-
division standard, the degree of matching between water
ing (0.242 < R 0.922), Relatively good matching (0.922<
resources and land resources is reasonable. Therefore, the
R 1.164), Good matching (1.164 < R 1.610), and Excel-
total renewable water resource is not a decisive factor for
lent matching (R > 1.610). Moreover, the other years are
water problems in Central Asia.
analyzed based on this criterion to form a contrast. Finally,
Due to data availability, the total water withdrawal was used as the water resource parameter and the cultivated area
Figures 7 and 8 illustrate the spatial matching pattern of water and land resources in Central Asia.
as the land resource parameter. Then, the Gini coefficient in
Combined with the multi-year average of Table 3, the
Central Asia in 2008 and 2014 were calculated, as shown in
matching index of Kazakhstan is the smallest and the match-
Figure 6.
ing index of Tajikistan is the largest, the latter being six times
The average Gini coefficient was 0.60 (Figure 6). The total
that of the former. By taking the five countries as a whole
water withdrawal indicates that the matching situation of
and calculating the multi-year average matching index of
water and land resources in Central Asia is extremely poor.
water and land resources from 1992 to 2016, the matching
Figure 4
|
Spatial distribution of water and land resources in five Central Asian countries in 2016.
Y. Zhang et al.
1002
|
Matching pattern between water and land resources in Central Asia
Figure 5
|
Lorentz curve of water and land resource matching in Central Asia.
Figure 6
|
Gini coefficient for Central Asia.
Hydrology Research
|
51.5
|
2020
Y. Zhang et al.
1003
Table 3
|
|
Matching pattern between water and land resources in Central Asia
Hydrology Research
|
51.5
|
2020
Matching index of water and land resources in five Central Asian countries from 1992 to 2016
Country
1992
1997
2002
2007
2012
2016
Mean
Kazakhstan
0.251
0.269
0.286
0.256
0.236
0.242
0.257
Kyrgyzstan
1.601
1.557
1.569
1.624
1.625
1.610
1.598
Tajikistan
2.054
2.198
2.279
2.284
2.281
2.310
2.234
Turkmenistan
1.565
1.349
1.138
1.109
1.164
1.164
1.248
Uzbekistan
0.943
0.949
0.916
0.938
0.926
0.922
0.932
Mean
1.283
1.264
1.238
1.242
1.246
1.250
1.254
index of water and land resources in Central Asia is 1.25
space and time, in combination with excessive and
(Table 3), which is Good matching. The spatial matching
often uncontrolled withdrawal for irrigation, creates
pattern of water and land resources in Central Asia varies
water scarcity, especially in the southern area.
across different periods. The matching situation of each
2. The average Gini coefficient in Central Asia is 0.32, and
country is as follows: Kazakhstan changed from Poor
the matching situation of water and land resources is rela-
matching in 2007 to Moderate matching in 2012 (Figure 7(d)
tively reasonable. It shows that Central Asia is regarded
and 7(e)); Uzbekistan fluctuated between Relatively good
as a whole and its water and land resources are balanced
matching and Moderate matching; Turkmenistan has
to a certain extent. The total renewable water resource is
changed from Good matching in 1997 to Relatively good
not a decisive factor contributing to the water problems
matching in 2002 (Figure 7(b) and 7(c)); Kyrgyzstan
in Central Asia. The average Gini coefficient obtained
remained as Good matching; Tajikistan was Excellent
from the total water withdrawal as a parameter of water
matching from 1992 to 2016 (Figure 7(a)–7(f)). Meanwhile,
resources in Central Asia was 0.60, and the matching situ-
differences exist among the matching index of water and
ation of water and land resources is Poor matching. This
land resources in different countries in the same period.
is in stark contrast to the results obtained from the total
The overall situation of Turkmenistan, Kyrgyzstan, and Taji-
renewable water resource as a water resource parameter
kistan is better; the upstream countries (Kyrgyzstan,
(from Reasonable matching to Poor matching). The water
Tajikistan) are better than the downstream countries (Turk-
and land resources of Central Asia as a whole are in bal-
menistan, Uzbekistan, Kazakhstan) (Figure 8).
ance, and the contradiction of water issues mainly lies in the distribution and utilization of water resources among countries.
DISCUSSION
3. The matching index of water and land resources is a quantitative indicator for the equilibrium situation and
Analysis of spatial distribution and matching pattern
satisfaction degree of water resources and cultivated
results
land resources in a certain area. The matching index of water and land resources in Central Asia is 1.25,
1. A clear dislocation of water resources and cultivated land
which is Good matching. However, the matching situ-
resources has been identified in Central Asia. Kazakhstan
ation varies greatly among different countries. The
is the most prominent, exhibiting low water resource per
matching situation of water and land resources in the
unit area with the highest reclamation rate in the region.
upstream countries (Kyrgyzstan, Tajikistan) is better
The shortage of water resources has become the main
than that in the downstream countries (Turkmenistan,
factor restricting agricultural development in the region.
Uzbekistan, Kazakhstan). This is because, after the dis-
Kazakhstan is not water-scarce in terms of total water
integration of the Soviet Union, the mode of free
supply per capita. However, the uneven distribution in
water supply to the downstream countries caused
Y. Zhang et al.
1004
Figure 7
|
|
Matching pattern between water and land resources in Central Asia
Hydrology Research
|
51.5
|
2020
Matching index of water and land resources in five Central Asian countries from 1992 to 2016.
dissatisfaction in the upstream countries. The upstream
to provide agricultural products for the upstream
countries to provide water and electricity for the
countries and other management systems have not
downstream countries and the downstream countries
been adopted.
Y. Zhang et al.
1005
Figure 8
|
|
Matching pattern between water and land resources in Central Asia
Hydrology Research
|
51.5
|
2020
Variation trend categories of resource matching in five Central Asian.
Comparison of the two models
the results of the two models also show consistency. They all show the renewable water resources is not a key factor
Two models are used in the matching of water and land
affecting water issues in Central Asia (Table 4).
resources in Central Asia. Although great discrepancies in the matching pattern are identified among the five countries, the results of the overall level of matching between water
Suggestions on water resource management
and land resources in Central Asia are reasonable. Two models both show that the matching of water and land
The Central Asian states share many common problems and
resources are balanced to a certain extent in Central Asia.
unaddressed tasks in the management of water resources
For the Gini coefficient model, the average of the matching
(Zhupankhan et al. ). Based on the current situation of
situation is reasonable, similar to the matching degree esti-
water and land resources matching in Central Asia, the fol-
mated from the matching index (Good matching). In terms
lowing suggestions are put forward.
of the renewable water resource, the two models are
1. Improve the water resources management and allocation
almost, in the matching of water and land resources in Cen-
system in Central Asia and formulate a scientific plan for
tral Asia, reasonable. At the same time, from the perspective
the relative matching of water and land resources. The
of each year, combining their respective evaluation criteria,
matching of water and land resources in Central Asian
Y. Zhang et al.
1006
Table 4
|
|
Matching pattern between water and land resources in Central Asia
Hydrology Research
|
51.5
|
2020
Comparison of the results of the two models
1992
1997
2002
2007
2012
2016
Gini coefficient
0.24
0.33
0.30
0.30
0.31
0.31
Matching results
Consistent
Reasonable
Reasonable
Reasonable
Reasonable
Reasonable
Matching index
1.28
1.26
1.24
1.24
1.25
1.25
Matching results
Good matching
Good matching
Good matching
Good matching
Good matching
Good matching
countries has fluctuated with time, indicating that the
cooperation with the international community, introduce
water resources management and distribution system
advanced water resources management programs, draw
needs to be improved, and it is impossible to ensure the
on national policies with outstanding effects on water
supply of resources among countries. The present situ-
control, and ultimately achieve social-economic-ecologi-
ation of water resources and future supply capacity
cal sustainable development in Central Asia.
should be combined to formulate reasonable social development strategies (Zhang et al. ). 2. Work out reasonable measures for constructing water
Limitations of this study
conservancy projects. The future for Central Asian development lies within the intertwined use of transboundary
1. The total amount of water and land resources will change
water resources. Upstream countries use water for hydro-
due to industrialization, urbanization, water conservancy
power while downstream countries use the water for
construction, and so on. Therefore, future studies should
agriculture and industry. Thus, water use also needs to
consider the dynamic changes in the matching of water
integrate transboundary sectors. Upstream countries pro-
and land resources in the border areas of Central Asia.
duce energy that downstream countries can use for
At the same time, the matching of water and land
agriculture and exploiting mineral resources (Zhupan-
resources should be refined to evaluate the micro-states
khan et al. ). Meanwhile, it is necessary to properly
of crop water demand and land use in each state,
handle the contradiction of water problems and formu-
which also needs further research.
late reasonable water distribution, water supply, and
2. This paper analyzes the matching of water and land
compensation water agreements. The water contradiction
resources in Central Asia on the time and space scales.
in Central Asia is not the total renewable water resources,
However, due to the fact that appropriate data were not
but the total water withdrawal used by various countries.
available, it is impossible to estimate historical data and
The upstream countries are rich in water resources, but
make scientific predictions for future data. Further
the matching degree of water and land resources lags
research can improve this to make a more comprehen-
behind the downstream countries. Therefore, the five
sive analysis.
Central Asian countries should consider the relevant compensation factors when formulating the transbound-
CONCLUSIONS
ary water resources allocation plan. 3. Strengthen cooperation with the international community, which requires cooperation among the countries of
This paper takes the five Central Asian countries as the
Central Asia that goes far beyond the current situation.
research unit, and uses the total renewable water resource
Major reforms are necessary and external pressures
and total water withdrawal as the water resources par-
from neighboring Russia and China are likely required
ameters, and constructs the model of the Gini coefficient
to make this happen (Howard & Howard ). There-
and the matching index between water and land resources.
fore,
Then, reasonable suggestions are put forward according to
Central
Asia
needs
to
actively
carry
out
1007
Y. Zhang et al.
|
Matching pattern between water and land resources in Central Asia
the current situation of water problems in Central Asia. The matching situation of water and land resources in Central Asia is generally better. The average Gini coefficient is 0.32, and the water and land resources matching index is 1.25. The main contradiction of the Central Asian water problem is the distribution and utilization among countries. Therefore, relevant departments should improve the water resources management and distribution system in Central Asia, formulate reasonable water distribution, water supply, and compensation water agreements to achieve social-economic-ecological sustainable development in Central Asia.
ACKNOWLEDGEMENTS The research is supported by the Strategic Priority Research Program
of
the
Chinese
Academy
of
Sciences
(XDA2004030201-2), the National Key R&D Program on Monitoring, Early Warning, and Prevention of Major Natural Disasters (Grant No. 2017YFC1502506).
REFERENCES Abdullaev, I., Kazbekov, J., Manthritilake, H. & Jumaboev, K. Participatory water management at the main canal: a case from South Ferghana canal in Uzbekistan. Agricultural Water Management 96 (2), 317–329. Bernauer, T. & Siegfried, T. Climate change and international water conflict in Central Asia. Journal of Peace Research 49 (1), 227–239. Chen, Y., Li, Z., Fang, G. & Li, W. Large hydrological processes changes in the transboundary rivers of Central Asia. Journal of Geophysical Research: Atmospheres 123 (10), 5059–5069. Deng, M. J., Long, A. H., Zhang, Y., Li, X. & Lei, Y. Assessment of water resources development and utilization in the five Central Asia countries. Advances in Earth Science 25 (12), 1347–1356. Dorfman, R. A formula for the Gini coefficient. The Review of Economics and Statistics 61 (1), 146–149. FAO a AQUASTAT Main Database. Food and Agriculture Organization of the United Nations (FAO), Rome, Italy. Available from: http://www.fao.org/nr/water/aquastat/data/ query/index.html?lang=en (accessed 5 December 2019). FAO b AQUASTAT Main Database. Food and Agriculture Organization of the United Nations (FAO), Rome, Italy. Available from: http://www.fao.org/faostat/zh/#data/EL (accessed 5 December 2019).
Hydrology Research
|
51.5
|
2020
Gafurov, A., Kriegel, D., Vorogushyn, S. & Merz, B. Evaluation of remotely sensed snow cover product in Central Asia. Hydrology Research 44 (3), 506–522. Howard, K. W. & Howard, K. K. The new ‘Silk road economic belt’ as a threat to the sustainable management of Central Asia’s transboundary water resources. Environmental Earth Sciences 75 (11), 976. Huang, K. W., Yuan, P. & Liu, G. Research on Water and Soil Resources Matching in Sichuan Province Based on DEA. China Rural Water and Hydropower, China. Jalilov, S. M., Keskinen, M., Varis, O., Amer, S. & Ward, F. A. Managing the water–energy–food nexus: gains and losses from new water development in Amu Darya River Basin. Journal of Hydrology 539, 648–661. Kienzler, K. M., Lamers, J. P. A., McDonald, A., Mirzabaev, A., Ibragimov, N., Egamberdiev, O., Ruzibaev, E. & Akramkhanov, A. Conservation agriculture in Central Asia – what do we know and where do we go from here? Field Crops Research 132, 95–105. Lee, S. O. & Jung, Y. Efficiency of water use and its implications for a water-food nexus in the Aral Sea Basin. Agricultural Water Management 207, 80–90. Li, Z., Chen, Y., Fang, G. & Li, Y. Multivariate assessment and attribution of droughts in Central Asia. Scientific Reports 7 (1), 1316. Liu, Y. S., Gan, H. & Zhang, F. G. Analysis of the matching patterns of land and water resources in Northeast China. Acta Geographica Sinica 61 (8), 847–854. Liu, D., Liu, C., Fu, Q., Li, M., Faiz, M. A., Khan, M. I., Li, T. X. & Cui, S. Construction and application of a refined index for measuring the regional matching characteristics between water and land resources. Ecological Indicators 91, 203–211. LPDAAC The Global Land Cover Data Product. Land Process Distributed Active Archive Center (LPDAAC). Available from: https://ladsweb.modaps.eosdis.nasa.gov/ search/ (accessed 5 December 2019). Riquelme, F. J. M. & Ramos, A. B. Land and water use management in vine growing by using geographic information systems in Castilla-La Mancha, Spain. Agricultural Water Management 77 (1–3), 82–95. Unger-Shayesteh, K., Vorogushyn, S., Merz, B. & Frede, H. G. Water in Central Asia-perspectives under global change. Global and Planetary Change 110 (Part A), 1–152. United Nations Strengthening Cooperation for Rational and Efficient use of Water and Energy Resources in Central Asia. Available from: http://www.unece.org/speca/welcome.html (accessed 5 December 2019). Yang, S. T., Yu, X. Y., Ding, J. L., Zhang, F., Wang, F. & Ma, Y. A review of water issues research in Central Asia. Acta Geographica Sinica 72 (1), 79–93. Yang, Z., Song, J., Cheng, D., Xia, J., Li, Q. & Ahamad, M. I. Comprehensive evaluation and scenario simulation for the water resources carrying capacity in Xi’an city, China. Journal of Environmental Management 230, 221–233.
1008
Y. Zhang et al.
|
Matching pattern between water and land resources in Central Asia
Yao, J., Hu, W., Chen, Y., Huo, W., Zhao, Y., Mao, W. & Yang, Q. Hydro-climatic changes and their impacts on vegetation in Xinjiang, Central Asia. Science of the Total Environment 660, 724–732. Zhang, Y., Yu, J., Wang, P. & Fu, G. Vegetation responses to integrated water management in the Ejina basin, northwest China. Hydrological Processes 25 (22), 3448–3461. Zhang, J., Chen, Y., Li, Z., Song, J., Fang, G., Li, Y. & Zhang, Q. a Study on the utilization efficiency of land and water resources in the Aral Sea Basin, Central Asia. Sustainable Cities and Society 51, 101693. Zhang, Y., Lei, G. P., Zhang, H. Q. & Li, J. b Spatiotemporal dynamics of land and water resources matching of cultivated
Hydrology Research
|
51.5
|
2020
land use based on micro scale in Naoli River Basin. Transactions of the Chinese Society of Agricultural Engineering 35 (8), 185–194. Zhu, B., Yu, J., Rioual, P., Gao, Y., Zhang, Y. & Xiong, H. Climate effects on recharge and evolution of natural water resources in middle-latitude watersheds under arid climate. In: Environmental Management of River Basin Ecosystems (M. Ramkumar, K. Kumaraswamy & R. Mohanraj, eds). Springer, Cham, Switzerland, pp. 91–109. Zhupankhan, A., Tussupova, K. & Berndtsson, R. Water in Kazakhstan, a key in Central Asian water management. Hydrological Sciences Journal 63 (5), 752–762.
First received 12 December 2019; accepted in revised form 24 April 2020. Available online 22 June 2020
1009
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Different runoff patterns determined by stable isotopes and multi-time runoff responses to precipitation in a seasonal frost area: a case study in the Songhua River basin, northeast China Jie Li, Wei Dai, Yang Sun, Yihui Li, Guoqiang Wang and Yuanzheng Zhai
ABSTRACT Runoff patterns are crucial to determine the hydrological response to climate change, especially in a seasonal frost area. In this study, multi-time runoff responses to meteoric precipitation for the period from July 2014 to June 2016 and the period from 1955 to 2010 were obtained to identify different runoff patterns in the Songhua River basin, northeast China, based on six stations. Two distinctly different runoff responses are exhibited: a periodic one in response to precipitation in the Nen River and a constant one in the Second Songhua River under different scales. Stable isotopes in the plain with diverse characteristics also supported these runoff patterns. What is more, gradual runoff relatively less sensitive to precipitation in the Second Songhua Rive was attributed to upstream dam constructions. Furthermore, the Second Songhua River contributes more water to the main stream
Jie Li Guoqiang Wang Yuanzheng Zhai (corresponding author) Engineering Research Center of Groundwater Pollution Control and Remediation, Ministry of Education, College of Water Sciences, Beijing Normal University, Beijing 100875, China E-mail: diszyz@163.com Wei Dai China Ordnance Industry Survey and Geotechnical Institute Co., Ltd, Beijing 100053, China
during January to March at the seasonal scale and in the 2000s at the annual scale, with low precipitation during those periods. This study could have implications for water management in the Songhua River basin. Key words
| base flow, daily runoff, precipitation, Songhua River basin, stable isotopes
Yang Sun The Bureau of Hydrology (Information Center), Songliao Water Resources Commission, Changchun 130021, China Yihui Li ShanDong Harbor Construction Group Co., Ltd, Rizhao 276800, China
INTRODUCTION Climate change has been widely reported in recent years and
Precipitation is an important driver for the global hydrological
is expected to continue (IPCC ; Han et al. ). It is very
cycle (Dai et al. ; Sivapalan et al. ; Jimenez Cisneros
likely that precipitation has increased over the mid-latitude
et al. ), carbon cycle (Chapin et al. ; Fang et al. ),
land areas of the northern hemisphere since 1901. Meanwhile,
and so on, and has attracted the attention of hydrologists
increases in the number of days with precipitation and inten-
and decision-makers.
sity of heavy precipitation with some seasonal and/or
Precipitation has an important impact on runoff, and
regional variation have been observed over many regions,
runoff patterns with diverse runoff responses to precipitation
even where there had been a reduction in annual total precipi-
are vital to assess the impacts of precipitation change on the
tation (Goswami et al. ; Donat et al. ; Liu et al. ).
hydrological cycle (Vogel et al. ; Berghuijs et al. ). Since precipitation elasticity to runoff was first defined by
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.183
Schaake (), many studies have been conducted to identify the response of runoff to precipitation based on long-term hydrometeorological data (Sankarasubramanian et al. ;
1010
J. Li et al.
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
Chiew ; Novotny & Stefan ). In this case, a quick
ridge of the hills is the northwestern watershed and the ridge
response in which the maximum runoff peak coincided with
of the Wanda Mountain is the southeastern watershed of the
the rainfall peak was observed (Changnon & Kunkel ;
Songhua River basin (Figure 1). Within the river basin, the
Onda et al. ), whereas in other cases, the variability of
highest elevation is 2,691 m at the Baitou Peak of the Chang-
runoff was reduced due to check dams (Batalla et al. ;
bai Mountain and the lowest elevation is about 57 m at the
Abbasi et al. ), indicating less sensitivity to precipitation.
river outlet. The Songnen and Sanjiang Plains, with an
Unraveling runoff patterns and their spatial-temporal vari-
elevation of 57–128 m, are situated in the central and eastern
ations is critical for the prediction and management of
parts, respectively. The areas of mountain, hill, and plain
surface water resources (Birsan et al. ; Wu et al. ).
within the basin account for 42.7%, 29.1%, and 27.4% of the
Runoff patterns in the frost regions are sensitive in response
total drainage area, respectively.
to climate change (Woo & Winter ; Kong & Pang ),
The Songhua River has a total channel length of 2,309 km
which is affected by seasonal evolution of soil freezing and
with an average channel gradient of about 0.00042. The Nen
thawing. The Songhua River basin, located in the far northeast
River in the north and the Second Songhua River in the
of China, is a typical seasonal frost area. There is a clear decreas-
south are the two sources for the Songhua River (Figure 1).
ing trend in annual precipitation found from 1960 to 2009 in the
The Nen River, originating from the northern part of the Dax-
Songhua River basin (Li et al. ). Meanwhile, a declining
inganling (Greater Khingan) Mountain, has a channel length
trend at 0.81 mm season 1 a 1 was also observed for the
of 1,370 km and a channel gradient of 0.00066, while the
summer series (Liang et al. ). Consequently, the stream
Second Songhua River with a channel length of 958 km and
flow in the main stream of the Songhua River basin shows a
a channel gradient of 0.00196 originates from the Tianchi
decreasing trend during 1955–2004 with dry years in the
Lake located in the central part of the Changbai Mountain.
1970s and after 2000 (Miao et al. ). Although there are sev-
The two rivers join the Songhua River at Sanchahe and flow
eral studies focused on the runoff changes at certain
northeastward before joining the Heilongjiang River at the
hydrological gauge stations, tributaries, and even the whole
outlet of the river basin (Figure 1).
basin of the Songhua River (e.g., Xu & Ma ; Feng et al.
The Songhua River basin has a mean annual tempera-
; Meng & Mo ; Pan & Tang ; Wang et al. ),
ture of 3–5 C with significant monthly temperature
the runoff patterns and their response to precipitation in the
difference. Mean temperatures are below 20 C in January
tributaries and their contributions to the main stream in the sea-
and 25 C in July. The mean annual precipitation in the river
sonal scale are rarely involved.
basin during the period 1955–2010 was 525.6 mm, more
The aims of this study are: (1) to display different runoff
than 75% of which occurred during the rainy season from
patterns of the Songhua River with precipitation–runoff
June to September (Liu et al. ). In comparison, precipi-
relationships in different time scales (daily, monthly, and
tation is very low from December to February and
annual) and isotopic evidence; and (2) to quantify the contri-
represents only 5% of the total precipitation. Furthermore,
butions of the two sources to the runoff of the main stream.
spatial distribution of precipitation across the river basin is
This work will provide important insights into future sus-
heterogeneous. The mean annual precipitation in the south-
tainable water resource management and planning in the
eastern part of the river basin is about 700–800 mm, while
study area.
that in the western part is only 400 mm.
STUDY AREA
DATA AND METHODS
This study was conducted in the Songhua River catchments in northeast China (119 520 –132 310 E, 41 420 –51 380 N). It has a
Data descriptions
drainage area of 5.57 × 105 km2 with the ridges of the Daxinganling Mountain and the Changbai Mountain as its
Data used in this study included daily runoff at Jiamusi, Yilan,
northwestern and southeastern boundaries, respectively. The
Tonghe, Harbin, Dalai, and Fuyu stations from 1 July 2014 to
J. Li et al.
1011
Figure 1
|
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
Location map of the Songhua River basin showing the six hydrological gauging stations.
30 June 2016 (Figure 1), which were obtained from the
for Dalai and Fuyu stations during 1956–2010 were taken
Songliao Water Resources Commission (http://www.slwr.
from Wang et al. () and Yu (). Isotopic data were
gov.cn). There are two hydrological stations located in the
obtained from Zhang et al. () and Yang ().
tributaries, which are Fuyu station in the Second Songhua
Data for the isotopic composition of 61 surface water
River, and Dalai station in the Nen River. Harbin, Tonghe,
samples were taken from previously published studies
Yilan, and Jiamusi stations are in the main stream from the
(Yang ; Zhang et al. ). These samples were collected
upstream to the lower reach (Figure 1). The lowermost
in both the headwater area and the central plain along the
station, Jiamusi station, controls a drainage area of 5.28 ×
Nen River and the Second Songhua River.
105 km2, accounting for 94.8% of the total drainage area. Daily precipitation amount data for the stations during the period of 2014–2016 are available from the National Meteor-
METHODS
ological Information Center of the China Meteorological
The double-mass curve plotted between precipitation and
Administration (http://data.cma.cn/). Annual runoff data
runoff is commonly used to reveal the streamflow change.
1012
J. Li et al.
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
The basic idea of double mass analysis is the gradual sum-
(Figure 2). When temperatures are low, the river freezes
mation of two variables according to the same time span,
over, and snow has not melted. The lowest runoff
with one variable as abscissa and the other as ordinate. If
recorded during a year is generally considered as the
the interaction between these two variables is stable, all
base flow. Our data show that the base flow increases
the corresponding points should lie in a straight line. The
from upstream to downstream, with a value of 472 m3/s
cumulative anomaly was used to identify the turning year
at Jiamusi, 423 m3/s at Yilan, 380 m3/s at Tonghe, and
in runoff changes during the period 1955–2010. The
358 m3/s at Harbin. The variation coefficient for the
anomaly is defined as annual value minus average value.
runoff during January to March was 0.14, 0.16, 0.10,
To calculate the contributions of total runoff from the
and 0.13 for Jiamusi, Yilan, Tonghe, and Harbin stations,
two different sources, a two-component mixing model was
respectively, and showed a steady trend during the dry
used. The two-component separation of a runoff hydrograph
season. Runoff began to increase after March, which
into runoff from Nen River and runoff from the Second
can be attributed to snow melt in the basin and the thaw-
Songhua River can be described as:
ing of the river network as temperatures rise, and is referred to as spring flood. Owing to the increase in pre-
qN=S
QN ¼ QN þ QSS
(1)
cipitation (Figure 2), runoff remains high in the rainy season. There are two clear peaks in river runoff during the rainy season from July to September, which are
where QN and QSS are runoff from the Nen River and the
observed at all stations (Figure 2).
Second Songhua River, respectively, and qN=S is the pro-
Runoff at the main stream (Figure 2(a)–2(d)) started to
portion of the Nen River runoff in the total runoff from
increase in April, in accordance with the increase in precipi-
the Nen River and the Second Songhua River.
tation. There were two or three peaks in runoff observed
The variation coefficient of runoff was employed to analyze fluctuation of daily or annual runoff based on: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u N u1 X σ 2 Cv ¼ , σ ¼ t (Q(t) Q) N Q t¼1
during the rainy season. Dalai station in the Nen River showed a similar trend to what was observed at the stations in the main stream (Figure 2(e)). The maximum runoff value was 2,840 m3/s on 29 June 2015. The lowest
(2)
where Q(t) is the daily or annual runoff for day/year (t), and Q is the average of Q(t). When the value of Cv approaches zero, the runoff tends to be constant throughout the period.
values are observed in the winter season with low precipitation, indicating the base flow. Runoff characteristics for Fuyu station in the Second Songhua River were strikingly different, with only gradual variations throughout the year (Figure 2(f)). The only obvious increases were observed in the months of July 2015 and June 2016. Therefore, the runoff mechanism must be different for the two tributaries.
RESULTS Monthly precipitation-runoff characteristics Daily precipitation-runoff characteristics To further understand the intra-annual variation in runoff, There are four stations in the main stream of the Songhua
monthly mean runoff was calculated at six stations
River. Figure 2 shows the temporal variations in daily
(Figure 3). Similar to the daily runoff variations, except
runoff at these four stations. Seasonal changes in runoff
for Fuyu station in the Second Songhua River, all gauge
are similar at all stations with high values in the rainy
stations showed a synchronous tendency with an annual
season. The lowest runoff was observed during January
peak flow in the rainy season. On the contrary, although
to March at all stations, which is in good agreement
the intra-annual distribution of precipitation is similar to
with the lowest precipitation recorded during this period
that of other stations, considerable changes in intra-annual
J. Li et al.
1013
Figure 2
|
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
Daily variations of runoff at Jiamusi, Yilan, Tonghe, Harbin, Dalai, and Fuyu stations in the the Songhua River (2014–2016). The dotted lines and the bars present runoff in left axis and precipitation in right axis, respectively.
J. Li et al.
1014
Figure 3
|
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
Relationships between monthly runoff and precipitation. The curves and the bars present runoff in left axis and precipitation in right axis, respectively.
variability of streamflow were not observed in Fuyu station.
ranges 478–3,054 and 877–3,826 m3/s, respectively. The
Therefore, the mainstream stations of the main stream
streamflow series at the two stations demonstrate similar
showed similar intra-annual distributions of discharge
fluctuations. Both stations experienced a high extreme
to Dalai station of the Nen River, indicating a greater
flow in 1960 and 1998.
flow contribution from Nen River than the Upper Songhua River.
The annual runoff during the period 1956–2010, measured at the two tributaries’ stations, is also presented in Figure 4. Average runoff at Dalai station was 651 m3/s
Long-term annual runoff changes
with a range of 146 and 1,971 m3/s and a variation coefficient of 0.56. The highest annual runoff was approximately
The annual streamflow series from the Harbin, Jiamusi,
13.5 times higher than the lowest value observed. While
Dalai, and Fuyu gauging stations are presented in Figure 4.
annual runoff decreased before 1980, it showed an increas-
Annual streamflow discharge velocities at the Harbin
ing trend in the 1980s. The annual runoff in the 2000s was
and Jiamusi stations were mostly concentrated in the
significantly lower than that in the 1990s. Similar to the
J. Li et al.
1015
Figure 4
|
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
Change of mean annual runoff during 1955–2010 at Harbin, Jiamusi, Dalai, and Fuyu hydrological stations.
daily/monthly runoff variations (Figures 2 and 3), there is a
DISCUSSION
similar annual changing trend at Dalai station compared to Jiamusi and Harbin stations in the main stream. Fuyu station
The runoff responses to monthly precipitation
had an average runoff of 465 m3/s with a range of 170 to 888 m3/s. The variation coefficient was 0.35, indicating a
To ascertain the controlling factors of runoff change, the
more steady flow when compared to Dalai station, which
monthly
is in accordance with the daily runoff data presented in
monthly runoff are presented in Figure 5. There is a
Figures 2(f) and 3(f).
good relationship between cumulative precipitation and
Figure 5
|
cumulative
precipitation
Double-mass curve for cumulative monthly precipitation and cumulative monthly runoff at six stations in the Songhua River.
and
cumulative
J. Li et al.
1016
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
cumulative monthly runoff for all the stations. Especially
points observed in April. This indicates that snow melt
in the rainy season, streamflow in the study area corre-
water contributes a lot to the runoff increase in the
sponded well with precipitation, which indicates that
spring season.
the precipitation is the dominant factor affecting stream-
To further confirm the effect of precipitation on the
flow. Meanwhile, due to low precipitation in the dry
runoff change in the stations, the monthly precipitation
season (from November to April in the next year), most
amount and monthly runoff are plotted in Figure 6. At
points are concentrated in Figure 5. However, they are
Fuyu station in the Second Songhua River, the monthly
mostly located above the fitted line with the furthest
runoff is constant when the monthly precipitation amount
Figure 6
|
Relationship between the monthly precipitation amount and monthly runoff in the Songhua River.
J. Li et al.
1017
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
is lower than 100 mm (Figure 3(f)). When the monthly pre-
(Figure 8), indicating the controlling effect of precipitation
cipitation amount is higher than 100 mm, the increasing
on runoff changes at these stations in the annual scale.
monthly runoff is observed responsively. In contrast, the
However, the cumulative anomaly of annual runoff at
obvious increase of monthly runoff is observed when the
Fuyu station in the Second Songhua River markedly differed
monthly precipitation amount exceeds 20 mm at other
from that of other stations both in the Nen River and the
stations. This means that precipitation plays a more impor-
main stream (Figure 9(a)), which displayed a gentle trend.
tant role in the runoff change of the Nen River and the
There is an obvious jump at Dalai, Harbin, and Jiamusi
main stream. Therefore, although the spatial distribution of
stations during the year of 1998 (Figure 9(a)). The same
monthly precipitation in one year displays similar periodical
‘turning year’ was also identified in the studies of Wang
change characteristics at these stations (Figure 3), the runoff
et al. () and Song et al. ().
variations performed in two diverse ways. The different performances were also obvious in the cumulative anomaly of
Isotopic evidences of runoff mechanisms
monthly runoff in Figure 7. The stations in the Nen River (Dalai) and in the main stream (Jiamusi, Harbin, Tonghe,
The stable isotope compositions of oxygen and hydrogen,
and Yilan) displayed a similar curve, with an increasing
expressed as δ 18O and δ 2H values, respectively, are com-
trend from April to October and a decreasing trend from
monly used to identify water sources (Clark & Fritz
October to April in the next year. This is in good agreement
). The δ 18O and δ 2H values in the mountain area
with the intra-annual variations of precipitation amount.
ranged from 14.9 to 12.0‰ and 106.1 to 84.4‰,
Meanwhile, a steady trend was observed at Fuyu station
respectively (Figure 10). The mean values for the Nen
in the Second Songhua River, which indicates that the influ-
River in the mountain area were 13.5‰ and 94.0‰,
ence of precipitation on runoff is relatively weak compared
respectively. The Second Songhua River had similar
to that of other stations.
δ 18O and δ 2H values of 13.5‰ and 97.0‰, respectively. These values are more negative than the mean
The runoff responses to annual precipitation
value of summer precipitation at GNIP Qiqihaer station ( 9.6‰ for δ 18O), indicating recharge from higher
There is an obvious relationship between annual precipi-
elevation. The mean elevation of the catchments of the
tation and runoff at Dalai, Harbin, and Jiamusi stations
rivers was estimated to be 1,100 m asl above the elevation
Figure 7
|
River change trends of the cumulative anomaly of monthly runoff in the Songhua River.
J. Li et al.
1018
Figure 8
|
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
Relationship between annual precipitation amount and annual runoff (1955–2004) at (a) Jiamusi, (b) Harbin, and (c) Dalai stations.
of Qiqihaer (147 m asl) using the isotopic altitude effect
strengthening the evidence of recharge from mountainous
of 0.25‰/100 m (Clark & Fritz ). They are in the
areas both in the Second Songhua River basin and the
range of the Changbai or Daxinganling Mountains,
Nen River basin.
J. Li et al.
1019
Figure 9
|
Figure 10
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
Change trends of (a) cumulative yearly runoff and (b) cumulative anomaly of annual runoff.
| δ 18O–δ 2H relationships for rivers in the Songhua River basin. The summer amount-weighted mean isotopic composition for precipitation events is shown as a plus sign. NR and SSR represent the Nen River and the Second Songhua River, respectively.
Although similar isotopic values were observed in the
intersection of the GMWL with the river water line (RWL,
headwater area, the two rivers displayed different isotopic
dotted line in Figure 10 of δ 2H ¼ 4.3δ 18O 28.5) of the
characteristics in the central plain. The samples taken in
Second Songhua River has δ 18O and δ 2H values of
2
the Second Songhua River plot on a line of δ H ¼ 18
2
10.4‰ and 73.2‰, respectively, similar to those of
5.1δ O 21.2 (R ¼ 0.97) indicate obvious evaporation
summer precipitation. The isotopic data are therefore in
effects. The samples located on the global meteoric water
good agreement with the constant runoff observed in the
2
18
line (GMWL, black line in Figure 8 of δ H ¼ 8δ O þ 10)
Second Songhua River (Figures 2(f), 3(f) and 4), which indi-
are those sampled from the downstream river. The
cates that the runoff in the Second Songhua River is mostly
1020
J. Li et al.
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
composed of uniform water. In the downstream, it appears
are relatively fewer, in agreement with the accordant
to be affected by significant surface water evaporation. For
runoff response to precipitation.
the Nen River, samples in the plain also plot below the GMWL, indicating the effect of evaporation. However, these samples are dispersive (hollow squares in Figure 10)
Contributions of tributaries to the main stream
with no obvious correlation. Considering that the majority of samples plot near the isotopic values of summer precipi-
If the total flow of the Songhua River is composed of water
tation, we conclude that summer precipitation plays an
from the two tributary rivers, the proportion originating
important role in the runoff of the Nen River. This is in
from the Nen River during each day can be calculated
accordance with the high runoff flow during the rainy
(Figure 11). During the period from January to March,
season, shown in Figures 2(e), 3(e), and 4.
these values are lower than 0.5, which indicates that
The insensitive runoff responses to precipitation and iso-
during this period, the Second Songhua River contributed
topic characteristics in the Second Songhua River may be
more water to the total runoff in the main stream. During
attributed to its upstream water conservancy facilities. A
the rest of the year, the Nen River appears to play a more
large number of dams in the Second Songhua River have
important role in providing water to the mean stream
been constructed since 1953, including the Baishan, Hon-
(Figure 11). We assumed that the base flow is equal to the
gshi, and Fengman Dams, which can attenuate flood
flow of the catchment during the lowest discharge period,
waves (Jia ). On the contrary, dams in the Nen River
observed from January until March, the months with the
Figure 11
|
Temporal variations in the contribution of the two tributary rivers to the Songhua River on the daily scale. The gray columns show the proportion of runoff in the main stream originating from Nen River (q(N/S) in the y-axis). The parts between the dashed lines indicate periods during which the main stream receives more water from the Second Songhua River than from the Nen River.
Figure 12
|
Temporal variations of the proportion of runoff in the main stream originating from Nen River on the annual scale (q(N/S) in the y-axis). The mean annual runoff is from Dalai and Fuyu stations, located in the two tributaries of the Songhua River (1956–2010). The gray areas indicate the period during which the main stream received more water from the Second Songhua River.
1021
J. Li et al.
|
Runoff patterns determined by multi-time runoff responses to precipitation
Hydrology Research
|
51.5
|
2020
lowest precipitation. In other words, the base flow of the
runoff patterns should be considered in the future water
Second Songhua River is higher than that of the Nen River.
resource management.
The annual proportion of runoff from the Nen River is shown in Figure 12. The data showed that the runoff in the main stream was dominated by water originating from
ACKNOWLEDGEMENTS
the Nen River before 1999, while the Second Songhua River contributed more in the 2000s (gray area in Figure 12).
This work is supported by the National Natural Science
In fact, this was supported by the fact that the precipitation
Foundation of China (Grants 41602276 and 41877174)
decreased in the 2000s. On the other hand, due to the higher
and the Fundamental Research Funds for the Central
contributions from the Second Songhua River in the 2000s,
Universities. Many thanks to Dr Fengping Li from Jilin
the cumulative anomaly of annual runoff at Harbin station is
University for her insightful advice and great help. We are
less than that of Dalai station (Figure 9(b).
grateful to Fei Wang for writing codes to process data.
These trends are in good agreement with the results from daily and monthly runoff calculations discussed in the previous section. In conclusion, our data show that when precipitation is low, either during the dry season (January to March), or during dry years (the 2000s), more runoff in the main stream was received from the Second Songhua River. Therefore, the impact of climate change, especially the precipitation change on streamflow, must be evaluated under different spatial and temporal scales, to obtain more information for water resources management.
CONCLUSIONS Using observed records of daily runoff at six gauging stations along the Songhua River during July 2014 to June 2016 and annual runoff (1956–2010) at four stations, we analyzed the runoff responses to precipitation change, and quantified the relative contribution of runoff from tributaries. Two different runoff patterns were determined: a periodic one in response to precipitation in the main stream and the Nen River; and a constant one in the Second Songhua River under different scales. The isotopic values in samples from the downstream of the Nen River plot near the isotopic values of summer precipitation, further indicating the influence of precipitation on runoff changes in this area. The steady runoff in the Second Songhua River with less sensitivity to precipitation may be due to upstream dams. A two-component mixing model showed that the Second Songhua River contributes more water to the main stream during January to March on the daily scale, and in the 2000s on the annual scale, both of which correspond to low precipitation periods. The different
REFERENCES Abbasi, N. A., Xu, X., Lucas-Borja, M. E., Dang, W. & Liu, B. The use of check dams in watershed management projects: examples from around the world. Science of the Total Environment 676, 683–691. Batalla, R. J., Gomez, C. M. & Kondolf, M. Reservoir-induced hydrological changes in the Ebro River basin (NE Spain). Journal of Hydrology 290, 117–136. Berghuijs, W. R., Larsen, J. R., van Emmerik, T. H. M. & Woods, R. A. A global assessment of runoff sensitivity to changes in precipitation, potential evaporation, and other factors. Water Resources Research 53, 8475–8486. Birsan, M., Molnar, P., Burlando, P. & Pfaundler, M. Stream flow trends in Switzerland. Journal of Hydrology 314, 312–329. Changnon, S. A. & Kunkel, K. E. Climate-related fluctuations in midwestern floods during 1921–1985. Journal of Water Resources Planning and Management 121 (4), 326–334. Chapin, F. S., Matson, P. A. & Vitousek, P. M. Principles of Terrestrial Ecosystem Ecology. Springer, New York, USA. Chiew, F. H. S. Estimation of rainfall elasticity of streamflow in Australia. Hydrological Sciences Journal 51, 613–625. Clark, I. D. & Fritz, P. Environmental Isotopes in Hydrogeology. Lewis, New York, USA, p. 328. Dai, A., Qian, T. T., Trenberth, K. E. & Milliman, J. D. Changes in continental freshwater discharge from 1948 to 2004. Journal of Climate 22, 2773–2792. Donat, M. G., Alexander, L. V., Yang, H., Durre, I., Vose, R., Dunn, R. J. H., Willett, K. M., Aguilar, E., Brunet, M., Caesar, J., Hewitson, B., Jack, C., Klein Tank, A. M. G., Kruger, A. C., Marengo, J., Peterson, T. C., Renom, M., Oria Rojas, C., Rusticucci, M., Salinger, J., Elrayah, A. S., Sekele, S. S., Srivastava, A. K., Trewin, B., Villarroel, C., Vincent, L. A., Zhai, P., Zhang, X. & Kitching, S. Updated analyses of temperature and precipitation extreme indices since the beginning of the twentieth century: the HadEX2 data set. Journal of Geophysical Research-Atmosphere 118, 2098–2118.
1022
J. Li et al.
|
Runoff patterns determined by multi-time runoff responses to precipitation
Fang, Q., Wang, G., Xue, B., Liu, T. & Kiem, A. How and to what extent does precipitation on multi-temporal scales and soil moisture at different depths determine carbon flux responses in a water-limited grassland ecosystem? The Science of the Total Environment 635, 1255–1266. Feng, X., Zhang, G. & Yin, X. Hydrological responses to climate change in Nenjiang River basin, northeastern China. Water Resources Management 25 (2), 677–689. Goswami, B. N., Venugopal, V., Sengupta, D., Madhusoodanan, M. S. & Xavier, P. K. Increasing trend of extreme rain events over India in a warming environment. Science 314 (5804), 1442–1445. Han, D., Wang, G., Liu, T., Xue, B., Kuczera, G. & Xu, X. Hydroclimatic response of evapotranspiration partitioning to prolonged droughts in semiarid grassland. Journal of Hydrology 563, 766–777. IPCC Summary for policymakers. In: Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor & H. L. Miller, eds). Cambridge University Press, Cambridge, UK and New York, USA. Jia, F. The evaluation research on runoff effect of Fengman Hydropower Station. Water Conservancy Science and Technology and Economy 24 (2), 66–71 (in Chinese). Jimenez Cisneros, B. E., Oki, T., Arnell, N. W., Benito, G., Cogley, J. G., Doll, P. & Mwakalila, S. S. Freshwater resources. In: Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (A. Revi, D. Satterthwaite, F. Aragon-Durand & J. CorfeeMorlot, eds). Cambridge University Press, Cambridge, UK, pp. 229–269. Kong, Y. & Pang, Z. Evaluating the sensitivity of glacier rivers to climate change based on hydrograph separation of discharge. Journal of Hydrology 434–435, 121–129. Li, F., Zhang, G. & Xu, Y. Spatiotemporal variability of climate and streamflow in the Songhua River Basin, Northeast China. Journal of Hydrology 514, 53–64. Liang, L., Li, L. & Liu, Q. Precipitation variability in Northeast China from 1961 to 2008. Journal of Hydrology 404, 67–76. Liu, Z., Xia, Z., Yu, L. & Wang, J. Temporal and spatial variation of characteristics of precipitation in Songhua River Basin during 1958–2009. Journal of Natural Resources 27 (6), 990–1000 (in Chinese). Liu, R., Liu, S., Cicerone, R. J., Shiu, C., Li, J., Wang, J. & Zhang, Y. Trends of extreme precipitation in eastern China and their possible causes. Advances in Atmospheric Sciences 32 (8), 1027–1037. Meng, D. & Mo, X. Assessing the effect of climate change on mean annual runoff in the Songhua River basin, China. Hydrological Processes 26 (7), 1050–1061.
Hydrology Research
|
51.5
|
2020
Miao, C., Yang, L., Liu, B., Gao, Y. & Li, S. Streamflow changes and its influencing factors in the mainstream of the Songhua River basin, Northeast China over the past 50 years. Environmental Earth Sciences 63, 489–499. Novotny, E. V. & Stefan, H. G. Stream flow in Minnesota: indicator of climate change. Journal of Hydrology 334 (3–4), 319–333. Onda, Y., Tsujimura, M., Fujihara, J. & Ito, J. Runoff generation mechanisms in high-relief mountainous watersheds with different underlying geology. Journal of Hydrology 331, 659–673. Pan, J. & Tang, L. Distributed hydrological simulation and runoff variation analysis for the upper basin of Songhua River. Journal Hydroelectric Engineering 32 (5), 58–63 (in Chinese). Sankarasubramanian, A., Vogel, R. M. & Limbrunner, J. F. Climate elasticity of streamflow in the United States. Water Resources Research 37 (6), 1771–1781. Schaake, J. C. From climate to flow. In: Climate Change and U.S. Water Resources (P. E. Waggoner, ed.). John Wiley, New York, USA, pp. 177–206. Sivapalan, M., Savenije, H. H. G. & Bloschl, G. Sociohydrology: A new science of people and water. Hydrological Processes 26, 1270–1276. Song, X., Mu, X., Gao, P., Wang, F. & Wang, S. Trends of runoff variation from 1900 to 2005 at Harbin Station of Songhua River. Journal of Natural Resources 24 (10), 1803–1809 (in Chinese). Vogel, R. M., Wilson, I. & Daly, C. Regional regression models of annual streamflow for the United States. Journal of Irrigation and Drainage Engineering 125 (3), 148–157. Wang, S., Wang, Y., Ran, L. & Su, T. Climatic and anthropogenic impacts on runoff changes in the Songhua River basin over the last 56 years (1955-2010), Northeastern China. Catena 127, 258–269. Woo, M. & Winter, T. C. The role of permafrost and seasonal frost in the hydrology of northern wetlands in North America. Journal of Hydrology 141, 5–31. Wu, C. S., Yang, S. L. & Lei, Y. P. Quantifying the anthropogenic and climatic impacts on water discharge and sediment load in the Pearl River (Zhujiang), China (1954– 2009). Journal of Hydrology 452–453, 190–204. Xu, J. & Ma, Y. Response of the hydrological regime of the Yellow River to the changing monsoon intensity and human activity. Hydrological Sciences Journal l54 (1), 90–100. Yang, X. Estimation of Groundwater Recharge and Renewal Rate Based on Environmental Isotopes in Songnen Plain. China University of Geosciences, Beijing, China. Yu, T. Research on the Impact of Climate Change on the Annual Runoff of Typical Watershed in North China. North China Electric Power University, Beijing, China. Zhang, B., Song, X., Zhang, Y., Han, D., Yang, L. & Tang, C. Relationship between surface water and groundwater in the second Songhua River basin. Advances in Water Science 25 (3), 336–347 (in Chinese).
First received 21 December 2019; accepted in revised form 24 April 2020. Available online 10 July 2020
1023
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Water balance changes in response to climate change in the upper Hailar River Basin, China Junfang Liu, Baolin Xue, Yinglan A, Wenchao Sun and Qingchun Guo
ABSTRACT Projected climate change will have a profound effect on the hydrological balance of river basins globally. Studying water balance modification under changing climate conditions is significant for future river basin management, especially in certain arid and semiarid areas. In this study, we evaluated water balance changes (1981–2011) in the upper Hailar River Basin on the Mongolian Plateau. To evaluate the hydrological resilience of the basin to climate change, we calculated two Budyko metrics, i.e. dynamic deviation (d ) and elasticity (e). The absolute magnitude of d reflects the ability of a basin to resist the influence of climate change and maintain its stable ecological function, whereas parameter e is used to assess whether a basin is hydrologically elastic. Results revealed modification of the hydrological balance during the study period has manifested as a decreasing
Junfang Liu Baolin Xue (corresponding author) Yinglan A Wenchao Sun College of Water Sciences, Beijing Normal University, Beijing 100875, China E-mail: xuebl@bnu.edu.cn Baolin Xue Wenchao Sun Beijing Key Laboratory of Urban Hydrological Cycle and Sponge City Technology, Beijing, China
trend of runoff and runoff-precipitation ratio. Correspondingly, basin-averaged evapotranspiration has also shown a decreasing trend, attributable mainly to precipitation. Furthermore, the calculated elasticity (e ¼ 8.03) suggests the basin has high hydrological resilience, which indicates the basin ecosystem may maintain its hydrological function to a certain extent under a changing climate.
Qingchun Guo School of Environment and Planning, Liaocheng University, Liaocheng, 252000, China
The results of this study could assist water resource management in the study area and the prediction of ecosystem response to future climate change. Key words
| Budyko curve, climate change, Hailar River Basin, resilience, resistance, water balance
HIGHLIGHTS
•
Water balance changes (1981–2011) in the upper Hailar River Basin on the Mongolian Plateau
•
The hydrological balance during the study period has manifested as a trend of decrease of runoff
•
were investigated.
and a decreased runoff–precipitation ratio. The calculated elasticity (e ¼ 8.03) suggests the basin has high hydrological resilience, which indicates the basin ecosystem might maintain its hydrological function to a certain extent under a changing climate.
INTRODUCTION The arid and semiarid areas that cover >50% of China’s
Tibetan Plateau and in northern China. Grassland is the
national mainland territory are distributed primarily on the
main vegetation type in such regions that are usually subjected to water limitations and intense effects associated
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying,
with climate warming (Fang et al. a; Han et al. a;
adaptation and redistribution, provided the original work is properly cited
A et al. b). Therefore, arid and semiarid areas are ecolo-
(http://creativecommons.org/licenses/by/4.0/).
gically vulnerable and sensitive to climate extremes (Miao
doi: 10.2166/nh.2020.032
1024
J. Liu et al.
|
Climate change influences on water balance changes
Hydrology Research
|
51.5
|
2020
et al. ; Fang et al. b). Moreover, vegetation transpira-
absorb change induced by external factors and retain its eco-
tion
accounts for a large proportion of the total
logical function (Creed et al. ). In recent years, this
evapotranspiration in arid and semiarid regions and thus it
concept has been applied in hydrological sciences (Trenbath
plays an essential role in the regional hydrological balance
; Gerten et al. ). The concept of hydrological
(Han et al. a; Wang et al. a; A et al. a).
resilience is described as the capability of a basin to
Research has shown that drought in arid and semiarid
maintain stability in multiple hydrological equilibrium
areas is projected to become intensified in the future,
states. Creed et al. () found that climate warming was
which could trigger considerable change in the processes
projected to change forest runoff, so he calculated the resili-
of the hydrological balance and affect regional water
ence and resistance of 12 watersheds across North America
resources (Dai ; Wang et al. a). Therefore, it is of
and concluded that the forest type is the dominant factor
critical importance to study the modification of the water
affecting the elasticity of a specific watershed. Helman
balance under the effects of climate change in arid and
et al. () calculated these two metrics of forests in the
semiarid regions to ensure sustainable water resources
Eastern Mediterranean and found that a drier climate may
management.
induce higher resilience compared with a more humid cli-
Projected global climate change in the 21st century
mate. Therefore, these two metrics have been applied and
could have a considerable effect on the hydrological balance
to some extent can reflect the characteristics of the river
of many river basins, especially in certain arid and semiarid
basin following the climate change. In this study, we used
regions, in terms of important variables such as precipi-
the Budyko theoretical curve to describe the relationship
tation, runoff, and evaporation (Cuo et al. ; Zhang
between basin resilience and climate change (Shen et al.
et al. ; Wang et al. b). Earlier studies have investi-
). The Budyko curve, which comprises a dryness index
gated changes of the hydrological balance due to climate
(DI ¼ PET/P) and an evaporative index (EI ¼ AET/P),
change in many river basins of northern China. In the
describes the relationship between potential evaporation
middle section of the Yellow River Basin, both streamflow
and actual evaporation (Troch et al. ). The Budyko
and precipitation have exhibited downward trends and
curve defines two basin states with evaporation being lim-
evaporation has presented an upward trend during the pre-
ited by either energy supply or water supply, which is
vious 60 years (He et al. ; Bao et al. ). Cuo et al.
determined by the calculated value of the DI (Figure 1). A
() analyzed observed streamflow changes in the upper
value of DI <1 indicates an energy-limited basin, whereas
Yellow River Basin using a modified VIC model and results
a value of DI >1 indicates a water-limited basin. Based on
showed that streamflow has decreased during recent decades. In the Kuye River Basin in Northwest China, Yang & Yang () found annual runoff has declined significantly during the past 60 years. Generally, streamflow has tended to diminish and evaporation has tended to increase in many basins of northern China owing to climate change (Wang & Hejazi ; Wang et al. , b; Shen et al. ). However, the extent to which these basins may be resistant to climatic perturbations should be explored further with regard to the prediction of basin responses to future climate change (Xue et al. ). In hydrology, the concepts of resistance and resilience, which are taken from the field of ecology, are two metrics used to quantify basin response to climate change (Zhang
Figure 1
|
The Budyko theoretical curve. The horizontal coordinate is the dryness index (DI ¼ PET/P) and the vertical coordinate is the evaporative index (EI ¼ AET/P).
et al. ; Williams et al. ). In ecological studies, a resi-
The solid line represents two different basin states, i.e. the energy and water limitations to the EI, and the dashed line represents the theoretical Budyko
lient ecosystem is defined as one that has the ability to
curve.
1025
J. Liu et al.
|
Climate change influences on water balance changes
the Budyko curve, we calculated two metrics, i.e. dynamic
Hydrology Research
|
51.5
|
2020
MATERIALS AND METHOD
deviation (d ) and elasticity (e), which can be used to quantify the resilience and resistance of a basin to the effects of
Study area
climate change (Creed et al. ; Helman et al. ). Dynamic deviation, which is described as the vertical depar-
The study area comprised the upper reaches of the Hailar
ture of the EI from the corresponding value calculated using
River Basin that drains a watershed of 43,345 km2 (Fang
the theoretical Budyko curve, represents the resistance of a
et al. ). This area is located in northeastern Inner Mon-
basin in terms of the runoff change caused by climate
golia, China (47 350 –50 120 N, 118 450 –122 400 E), and it
change. A positive (negative) value of the dynamic deviation
usually experiences extreme weather caused by the influ-
indicates that runoff generated is smaller (greater) than the
ence of the East Asian summer monsoon, which means it
value estimated using the Budyko theoretical equations,
is highly sensitive to climate change (Figure 2). The river
and its absolute magnitude reflects the extent of the runoff
originates in the Greater Khingan Mountains and finally
change relative to the inherent runoff calculated by the
joins the Erguna River (Duan et al. ). The length of
Budyko theoretical curve. Smaller values of the dynamic
the main river is 708.5 km. The study area has a temperate
deviation indicate higher basin resistance. Elasticity is the
continental monsoon climate, i.e. short cool summers and
second metric that can be used to reflect the hydrological
long cold winters (Fang et al. ). Mean annual precipi-
resilience of a basin, which represents the extent to which
tation and mean annual temperature in this river basin are
a watershed can hold this partitioning pattern after climate
347.6 mm and –1.2 C, respectively, according to observed
perturbations. A basin with high elasticity means runoff pre-
meteorological data during 1979–2018 (www.tpedatabase.
dictions within the basin responds highly consistently with
cn). The study area is 510–1,622 m above sea level and its
the Budyko curve, i.e. when a change in the DI results in
topography is predominantly mountains, hills, and wetlands.
a change in the EI, the ecological system moves along the Budyko curve. This study investigated the upstream area of the upper Hailar River Basin, which is situated in northeastern
Standardized precipitation and evapotranspiration index
Inner Mongolia, China. The upper Hailar River Basin is a primary tributary of the Erguna River and it is the
In this study, the standardized precipitation and evapotran-
main water source for the local industry and agriculture.
spiration index (SPEI) was selected as the drought
Moreover, its location belongs to the ecologically vulner-
monitoring index. This index considers the statistical distri-
able area. So it is meaningful to detect the water
bution of precipitation and the potential evapotranspiration
balance changes following the climate change and quan-
at the same time and it can reflect the regional drought
tify its hydrological resilience and resistance to climate
more comprehensively. According to Abbasi et al. (),
change for future water resource management. In this
the SPEI can be calculated as described in the following.
study, we employed the Mann–Kendall test to analyze
First, the water-year potential evapotranspiration (ET0)
the changing hydrological balance of the study basin
is calculated using the radiation-based formulation of Priest-
and adopted two metrics to determine the basin’s hydrolo-
ley and Taylor (Dewes et al. ), as shown below:
gical resilience, which is used as a supplement. The objective was to investigate the following: (1) the changes in hydroclimatic variables/climate in the study basin over the previous 30 years, (2) the basin response to the
PET ¼ 1:26
Δ (RN G) Δþγ
(1)
where RN is net radiation, Δ is the gradient of saturated
changes in climate and water balances variables, and (3)
vapor pressure, G is soil heat flux, and γ is the psychrometric
the resilience of the basin to climate change. The findings
constant. The unit of the variable PET is mm/d. Second, the
of this study will assist water resources management in
difference between monthly precipitation and evapotran-
the basin.
spiration is calculated as Di ¼ Pi PETi , where i is the
J. Liu et al.
1026
Figure 2
|
|
Climate change influences on water balance changes
Hydrology Research
|
51.5
|
2020
Map of the study area and its position in China.
month counter. Third, the accumulation sequence of water
according to the following formulas:
profit and loss over different timescales is established using Equation (2), where k is the timescale (here, k ¼ 12): β¼
Dkn ¼
k 1 X
(Pn i PETn i ), n k
(2)
i¼0
2w1 w0 (6w1 w0 6w2 )
(w0 2w1 )β α¼ 1 1 Γ 1þ Γ 1 β β
(4)
(5)
Fourth, the Dkn data series should be fitted and normalized to calculate the SPEI. Vicente-Serrano et al. () showed that the log-logistic density function is the best fitting function to fit the Dkn data series through contrasting
1 1 Γ 1 γ ¼ w0 αΓ 1 þ β β
the different types of parameter function. The expression of the log-logistic probability density function including three parameters is as follows:
f(x) ¼
x γ β 2 β x γ β 1 1þ α α α
ws ¼
n 1X l 0:35 s 1 Xl , n i¼1 n
(6)
(7)
where ws is the probability weighted moment, s is taken as 0, (3)
1, or 2, l is the sequence of accumulated water deficit X in ascending order (X1 X2,…, Xn), and Γ(β) is the gamma function. Through the three-parameter log-logistic prob-
where α, β, and γ are the parameters of scale, shape, and
ability distribution function, the cumulative probability on
beginning, respectively. The linear moment method is
a given timescale can be calculated using Equation (7)
adopted to estimate the fitting parameters of this function
(Polong et al. ). Then, the SPEI is calculated using
J. Liu et al.
1027
|
Climate change influences on water balance changes
Hydrology Research
Equation (9): "
|
51.5
|
2020
where P, ET, and ET0 are mean annual precipitation, mean
annual actual evaporation, and potential evapotranspira-
β # 1
F(x) ¼ 1 þ
α x γ
SPEI ¼ W
C0 þ C1 W þ C2 W 2 1 þ d1 W þ d2 W 2 þ d3 W 3
(8)
tion, respectively. Parameter w is a constant determined by the characteristics of the watershed, e.g. vegetation type
(9)
where C0 ¼ 2:515517, C1 ¼ 0.802853, C2 ¼ 0.010328, d1 ¼ 1.432788, d2 ¼ 0.189269, and d3 ¼ 0.001308 (Gao et al. ). Finally, W is calculated using Equation (11): P ¼ 1 F(x) ( pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 ln (P) P 0:5 W ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 ln (P 1) P > 0:5
(10) (11)
Having obtained the SPEI, a criterion was required to determine the occurrence of drought (Table 1). In this study, if the value of the SPEI was less than –0.5, we considered a drought phenomenon happened; if the value of the SPEI was positive, we considered the drought period over (Begueria et al. ).
and soil type (Qiu et al. ). In this study, we used the adjusted equation assuming w ¼ 2 to describe the theoretical ET0 relationship between the dryness index DI ¼ and the P ET evaporative index EI ¼ . P The parameters of deviation (d) and elasticity (e) were calculated for the studied basin to represent the potential departure from the Budyko theoretical curve of the DI and EI points (Creed et al. ). Deviation is described as the vertical departure of the EI from the corresponding B value calculated from the theoretical Budyko curve, which is composed of two parts: static and dynamic deviation. Static deviation is calculated as the mean annual EI minus its theoretical B value obtained from the mean annual DI, which is the inherent deviation with average normal climate conditions (s ¼ (EIAVG B), Figure 3). Dynamic deviation
Calculation of Budyko metrics Based on the Budyko hypothesis, the annual water balance can be described using the function of water (precipitation) and energy (potential evaporation). Among the various forms of equations for Budyko curves, we selected the following (Zhang et al. ): ET ¼ P
ET0 P ET0 ET0 1 þ 1þw P P
Table 1
|
1þw
(12)
Classification of the standardized precipitation evapotranspiration index Figure 3
Drought categories
SPEI values
Extreme drought
–2.0
Severe drought
–2.0 to –1.0
Moderate drought
–1.0 to –0.5
Normal
–0.5 to 0.5
Moderate wet
0.5–1.0
Severe wet
1.0–2.0
Extreme wet
2.0
|
Graphical representation of three Budyko metrics: static deviation (s), dynamic deviation (d), and elasticity (e). Static deviation (s) is the inherent deviation with average normal climate conditions, which is calculated as the average EI minus the B value calculated from the Budyko theoretical curve according to the mean DI under normal climate conditions (s ¼ (EIN BN )). Dynamic deviation (d) is the departure of the mean annual EI from the B value with climate change after considering the inherent static deviation, which is the additional deviation induced by climate change [d ¼ (EId Bd ) s)]. Its absolute magnitude reflects the hydrologic resistance of the basin to climate change. Elasticity (e) is the ratio of the range of the DI to the range of the residual EI, which is the magnitude of the EI departure from its corresponding B value according to the Budyko theoretical cure for the entire period [e ¼ (DImax DImin )=(EIR, max EIR, min)]. High elasticity indicates a basin has hydrological resilience.
J. Liu et al.
1028
|
Climate change influences on water balance changes
Hydrology Research
|
51.5
|
2020
(d ) is the departure of the mean annual EI from the B value
where statistic S is the cumulative number of values at time i
with climate change after considering the inherent static
larger than at time j. Under the assumption of random inde-
deviation [d ¼ (EID BD ) s]. Thus, dynamic deviation is
pendence of the time series, the statistic UFk can be defined
the additional deviation induced by climate change and its
by the following formula:
absolute magnitude reflects the hydrologic resistance of the basin to climate change. A value of d close to zero indicates the basin has high hydrological resistance (Helman
Si E(Si ) UFi ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Var(Si )
i ¼ 1, 2, , n
(16)
et al. ). Elasticity (e) is defined as the ratio of the range of the DI (DImax DImin ) to the range of the EIR
In Equation (16), UF1 ¼ 0, and E(Sk) and Var(Sk)
(EIR, max EIR, min), which is the magnitude of the depar-
represent the expected value and the variance, respectively,
ture of the EI from its corresponding B value according to
of the cumulative value Sk , which can be obtained as
the Budyko theoretical cure for the entire period of climate
follows:
change
[e ¼ (DImax DImin )=(EIR, max EIR, min)].
The
points representing the DI and the EI with low departures from the Budyko curve tend to have large values of e, indi-
E(Si ) ¼
i(i 1) 4
cating that the basin has high elasticity, and vice versa. In addition, we used a threshold of e ¼ 1 to distinguish elastic
Var(Si ) ¼
i(i 1)(2i 5) 72
(17)
(18)
and inelastic basins. The actual evapotranspiration (ET0) was calculated based on the water balance (Xue et al. , ):
UFi is a standard normal distribution, which is a sequence of statistics calculated from the order of time
ET0 ¼ P Q 4S
(13)
where P is precipitation, R is streamflow, E is evapotran-
series X. Given significance level α, a condition of jUFi j > Uα indicates an obvious trend change in the sequence. Using the inverse time series, we calculated the
spiration, and ΔS is the change of water storage volume
UFi again using the above calculation process, where
(Bao et al. ). We considered water storage negligible,
UBi ¼ UFi and i ¼ n, n 1, , 1 for the same test
assuming a steady state for the study period (1980–2011).
method as described above.
The Mann–Kendall test
Data sources
The nonparametric Mann–Kendall test can be used to analyze change trends and breakpoints of hydrological time series (Sung et al. ). It can accommodate many types of samples because it does not need the samples to follow any particular distribution and it is rarely disturbed by abnormal data (Liang et al. ). For an assumed data series X (x1 , x2 , xn ), n is the length of the data series. First, the cumulative statistic S should be calculated as follows: Sk ¼
k X
ri ¼
basin outlet, which operated during 1981–2011, because we only needed annual streamflow data of the entire river basin. The daily flow was summed to an annual amount, multiplied by the corresponding length of time, and then divided by the basin area to obtain the annual streamflow. Gridded precipitation and temperature data (0.025 × 0.025 ) measured during 1981–2011 were collected from GHCN-Monthly datasets (www.ncdc.noaa.gov/ghcnm/v3.
r1
k ¼ 2, 3, n
(14)
i¼1
We used daily Q data from the hydrological station at the
1 xi > xj 0 xi xj
php) (Zhao et al. ). Net radiation data from 1981 to 2011 with 500-m spatial resolution and 30-d composite tem-
j ¼ 1, 2, i 1
(15)
poral resolution were downloaded from the Global Land Data Assimilation System.
J. Liu et al.
1029
|
Climate change influences on water balance changes
RESULTS
Hydrology Research
|
51.5
|
2020
of annual mean precipitation before 2000 was much smaller than after 2001. According to the P value calculated from
Trends of climatic variables in the Hailar River Basin
the linear regression (P ¼ –0.48), the trend of decrease of annual
mean
precipitation
was
insignificant
at
the
This study found that the Hailar River Basin has become
α ¼ 0:05 level (Figure 4(a)). Moreover, the results of the
warmer and drier in recent decades, consistent with the
Mann–Kendall test revealed an oscillation in mean annual
results of Han et al. (b). During the study period,
precipitation before 1998 and after 1998, indicating that
annual mean precipitation was 429.56 mm, with the highest
the trend of decrease of mean annual precipitation became
(lowest) annual mean precipitation of 621.41 (248.31) mm
slightly more evident (Figure 4(b)). Therefore, although the
in 1998 (2007). The time series of areal precipitation in
trend of decrease of mean annual precipitation was mild,
the Hailar Basin showed a decreasing trend during the
the scale of the trend intensified.
studied 30 years (Figure 4(a)). The multiyear annual mean
During 1981–2011, average annual temperature in the
precipitation was 483.83, 443.56, and 367.52 mm in 1981–
study area was –1.81 C, with the lowest (highest) average
1990, 1991–2000, and 2001–2011, respectively. The multi-
annual temperature of –3.53 C (–0.19 C) in 1984 (2007).
year mean annual precipitation during 1991–2000 was
There has been a trend of increase in average annual temp-
40.27 mm less than in 1981–1990 but 76.04 mm higher
erature over the 30-year study period (Figure 4(c)). The
than in 2001–2011, which indicates that the rate of decrease
multiyear annual mean temperature was –2.4, –1.56 and
Figure 4
|
Change trend of (a) annual mean precipitation and (b) annual mean temperature in the Hailar River Basin during 1981–2011. Mann–Kendall test results for (c) precipitation and (d) temperature (red line: calculated UF value, blue line: calculated UB value). Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/ nh.2020.032.
J. Liu et al.
1030
|
Climate change influences on water balance changes
Hydrology Research
|
51.5
|
2020
–1.51 C in 1981–1990, 1991–2000, and 2001–2011, respect-
the SPEI in the Hailar River Basin has risen and fallen
ively. The multiyear mean annual temperature during 1991–
during the 30-year study period. During 1981–1991 (except
2000 was 0.84 C higher than in 1981–1990 but 0.04 C
1987), the UF value was positive, indicating that the value
lower than in 2001–2011, which indicates that the rate of
of the SPEI showed an overall trend of rise during this
increase in temperature has decreased slightly in compari-
period. During 1992–2011 (except 1993 and 1998), the UF
son with earlier years. Statistically, the P value (P ¼
value was <0 and it exceeded the confidence interval of
0.4835) calculated from the linear regression indicates that
0.05 after 2007, showing that the SPEI decreased gradually
this trend of increase was not significant and that its process
and that the change trend of the SPEI was significant after
of change was uncertain (Figure 4(c)). In addition, according
2007. Within the confidence interval, the UF and UB
to the results of the Mann–Kendall test, the trend of increase
curves intersect in 1998, indicating that 1998 was the break-
in annual mean temperature was significant after 1993 at the
point of the SPEI and that drought has intensified in the
α ¼ 0:05 level. Generally, despite the uncertain trend of
Hailar River Basin since 1998 (Zhai et al. ).
change of temperature, it was clearer than that of precipitation (Figure 4(d)).
Water balance change in the Hailar River Basin
We also analyzed the change of humid/dry climate in the Hailar River Basin based on the SPEI. It was found
The water balance of the Hailar River Basin has changed
that drought intensified during the study period, especially
during the study period (Figure 6). During 1981–2011,
during 1998–2011 (Figure 5). The variations of the
annual mean runoff at the outlet hydrological station has
12-month SPEI and the Mann–Kendall test results of the
decreased at the rate of 1.46 mm/year (Figure 6(a)). The
SPEI for the Hailar River Basin during 1981–2011 are
change trend of annual mean runoff was similar to that of
shown in Figure 5. The general trend of decrease indicates
precipitation (Figure 4(a)). During 1981–1990, 1991–2000,
that this area has experienced increasing occurrence and
and 2001–2011, annual mean runoff was 102.26, 86.88,
intensification of drought. The maximum (minimum) SPEI
and 52.51 mm, respectively. Annual mean runoff during
of 2.76 (–2.69) was in 2007 (1998), which was the wettest
1991–2000 was 15.38 mm lower than in 1981–1990 but
(driest) year in the study period. The UF curve shows that
34.37 mm higher than in 2001–2011. Therefore, the rate of annual mean runoff decrease was increased during the study period. As shown in Figure 6(a), 1998 represents the breakpoint of the annual mean runoff series. Prior to 1998, repeated oscillation occurred in the time series of annual mean runoff, whereas annual mean runoff showed a more obvious decreasing trend after 1998. Throughout the entire data series, the decreasing trend of annual runoff was insignificant at the α ¼ 0:05 level, according to the obtained P value (P ¼ –0.53). In addition, annual mean evaporation has decreased generally during the 30-year study period with a change trend similar to but more evident than that of annual mean runoff (Figure 6(b)). However, the rate of decrease of annual mean evaporation has been smaller than the change trend of precipitation. The proportions of precipitation allocated to runoff and evaporation have also changed during the study period, i.e. a
Figure 5
|
Variation of the calculated annual SPEI and the Mann–Kendall test results of the SPEI in the upper Hailar River Basin during 1981–2011 (red line: calculated UF value, blue line: calculated UB value). Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/nh.2020.032.
smaller proportion of precipitation has generated runoff, while a greater proportion has been evaporated. As shown in Figure 7, there has been a general trend of decrease in
J. Liu et al.
1031
|
Climate change influences on water balance changes
Hydrology Research
Figure 6
|
Change trends of hydrological variables in the upper Hailar River Basin in the studied 30 years: (a) runoff and (b) evaporation.
Figure 7
|
Change trends of (a) ET/P and (b) R/P during 1981–2011.
|
51.5
|
2020
the runoff–precipitation ratio and a trend of increase in
during 1981–1990, 1991–2000, and 2001–2011, respectively.
the evaporation–precipitation ratio. The multiyear mean
It is evident that there was no change in the relationship
runoff–precipitation ratio was 0.206, 0.194, and 0.143
between precipitation and runoff during 1981–2000, i.e.
J. Liu et al.
1032
|
Climate change influences on water balance changes
Hydrology Research
|
51.5
|
2020
the runoff–precipitation ratio during 1991–2000 was 0.012
runoff predicted using the Budyko curve, although it was
higher than during 1981–1990, which indicates that the
very close to the theoretical runoff.
same percentage of runoff was generated at this time. How-
Dynamic deviation (d ) describes the vertical departure
ever, there was a significant decrease in the runoff–
of points (DI, EI) in the five-year dry period from the
precipitation ratio after 2000, i.e. the decrease of 0.051
Budyko curve when considering the static deviation. Its
during 2001–2011 relative to 1991–2000 indicates less pre-
absolute magnitude reflects the hydrological resistance of
cipitation was converted into runoff. The change trend of
the basin to climate change, according to runoff change, in
the runoff–precipitation ratio was consistent with the trend
comparison with the runoff estimated using the Budyko
of annual mean runoff.
curve. In the Hailar River Basin, the calculated value of d was –0.0183, which is less than zero and indicates higher
Resistance and resilience of the watershed Based on the calculated SPEI, we calculated five-year moving averages for the period 1981–2011. Finally, we selected 1981–1985 as the wet period with maximum average values of the SPEI and 2004–2008 as the dry period with minimum average values of the SPEI (Figure 8). According to the DI and EI in the normal climate period, s was calculated using the equation described in the above. An obtained value of s < 0 indicated that runoff generated in the five-year wet period was larger than the predicted runoff based on the Budyko curve, and vice versa. Obtained points distributed near the Budyko theoretical curve (|s| < 0.05) indicated the pre-warming runoff fitted the predicted values using the Budyko curve, which reflected the inherent characteristics of the basin. The value of s calculated for the Hailar River Basin was
runoff than estimated according to the Budyko curve. Moreover, the absolute value of d was close to zero, which shows the extent of runoff change has been small under the effects of climate change, reflecting high hydrological resistance of the basin to climate change. The hydrological resilience of a basin is reflected by its elasticity (e). In a basin with high elasticity, runoff change under the effects of climate change is consistent with the Budyko curve. For example, as can be seen from Figure 8, when change in the DI results in change of the EI along the Budyko curve, the basin has hydrological resilience (Trenbath ). Conversely, change in the DI that leads to change in the EI that deviates from the Budyko curve reflects a basin without hydrological resilience. The value of elasticity we calculated was 8.03, which is much greater than 1 and reflects the high hydrological resilience of the Hailar River Basin.
0.035, i.e. a positive value whose absolute value is within the scope of 0.05. It reflects that pre-warming runoff produced under a normal climate was smaller than the
DISCUSSION During 1981–2011 in the Hailar River Basin, there was a trend of decrease in annual mean precipitation and a trend of increase in annual mean temperature (Figure 4). As the calculated results showed, the rate of decrease of precipitation and the rate of increase of temperature both increased gradually to some degree, indicating that the climate of the upper Hailar River Basin during the studied 30-year period has changed and that this change might become intensified in the future (Figure 4). Moreover, the observed decreased precipitation and increased temperature
Figure 8
|
Distribution of 10 points chosen from the 30-year study period according to the SPEI in the Budyko theoretical framework.
are projected to lead to intensification of drought, consistent with the results of Wang et al. (). This type of
1033
J. Liu et al.
|
Climate change influences on water balance changes
Hydrology Research
|
51.5
|
2020
phenomenon is consistent with that observed in the region
(wet) period (Xue et al. ; Han et al. a; Sun et al.
of the Hulunbuir grassland during 1960–2017 (Wang et al.
). This flexibility of water use efficiency in different
b). Under the background of climate change, the water balance in the upper Hailar River Basin has changed. Annual
periods increased the basin’s ability to cope with climate change and further enhanced the hydrological resilience of the basin to climate change (Ponce-Campos et al. ).
mean runoff has shown a decreasing trend during the study period (Figure 6(b)), consistent with the simulated results (Duan et al. ). Precipitation is the origin of
CONCLUSIONS
runoff generation and the decrease of precipitation induced the reduction of runoff. In addition, a greater proportion of
In the Hailar River Basin, the major features of climate
precipitation is now evaporated and less transformed into
change during the 30-year study period (1981–2011) are rep-
runoff (Figure 7(a) and 7(b)). However, annual mean evap-
resented by increased temperature, decreased precipitation,
oration also shown a decreasing trend due to the
and intensified drought. We used the Mann–Kendall test to
substantial decrease of precipitation. Owing to the increase
investigate the water balance changes in this basin based on
of temperature, a greater proportion of precipitation is now
certain hydrological variables, e.g. precipitation and runoff.
evaporated (Han et al. a). In addition, increased veg-
Based on hydrological data obtained during the study
etation coverage has led to a greater uptake of water and
period, it was determined that the water balance change
enhanced evapotranspiration (Zhao et al. ; Fang et al.
has been manifest as trends of decrease of runoff and evapor-
a; Bao et al. ). Thus, the proportion of precipitation
ation. In addition, we used two Budyko metrics to quantify
converted to runoff has been reduced and the amount of
the resistance and resilience of runoff in this basin to the
runoff has declined.
effects of climate change. The results showed the basin has
In our study, the calculated values of d and e were – 0.0183 and 8.03, respectively. The values of these two
high hydrological resilience and resistance and that it could retain its ecological function in a changing climate.
metrics reflect the high level of hydrological resistance and resilience of the Hailar River Basin, confirming that the predicted runoff change has high consistency with the Budyko curve. The obtained value of e > 1 indicates that the change
ACKNOWLEDGEMENTS
in the DI during the entire study period was greater than the
This study was supported by the National Natural Science
change in the EI (Creed et al. ). Several of the factors
Foundation
of
China
(Grant
31670451)
and
the
that contribute to high elasticity were analyzed and among
Fundamental Research Funds for the Central Universities
the dominant influencing hydrological factors were those
(No. 2017NT18). We thank James Buxton MSc from
that contribute to certain changes in ET0. For example, sea-
Liwen Bianji, Edanz Group China (www.liwenbianji.cn./
sonal precipitation has an effect on the evaporative indices
ac), for editing the English text of this manuscript.
based on the Budyko framework (Gentine et al. ; Williams et al. ). Annual precipitation in the Hailar River Basin (30–350 mm) is concentrated primarily in summer (Fang et al. a). Given the high temperature and large quantity of precipitation in summer, potential evaporation in this water-limited area is maximized and the range
DATA AVAILABILITY STATEMENT All relevant data are included in the paper or its Supplementary Information.
of the DI is increased. Additionally, Klein et al. () argued that some basins can adjust their own ecophysiologi-
REFERENCES
cal factors to cope with drought induced by climate change, such as WUE (Xue et al. ; Yao et al. ). Therefore, water use efficiency in the basin is large (small) in the dry
A, Y., Wang, G., Liu, T., Xue, B. & Kuczera, G. a Spatial variation of correlations between vertical soil water and
1034
J. Liu et al.
|
Climate change influences on water balance changes
evapotranspiration and their controlling factors in a semiarid region. J. Hydrol. 574, 53–63. A, Y., Wang, G. Q., Liu, T. X., Shrestha, S., Xue, B. L. & Tan, Z. X. b Vertical variations of soil water and its controlling factors based on the structural equation model in a semi-arid grassland. Sci. Total Environ. 691, 1016–1026. Abbasi, A., Khalili, K., Behmanesh, J. & Shirzad, A. Drought monitoring and prediction using SPEI index and gene expression programming model in the west of Urmia Lake. Theor. Appl. Climatol. 138, 553–567. Bao, Z., Zhang, J., Wang, G., Chen, Q., Guan, T., Yan, X., Liu, C., Liu, J. & Wang, J. The impact of climate variability and land use/cover change on the water balance in the Middle Yellow River Basin, China. J. Hydrol. 577, 123942. https:// doi.org/10.1016/j.jhydrol.2019.123942. Begueria, S., Vicente-Serrano, S. M., Reig, F. & Latorre, B. Standardized precipitation evapotranspiration index (SPEI) revisited: parameter fitting, evapotranspiration models, tools, datasets and drought monitoring. Int. J. Climatol. 34, 3001–3023. Creed, I. F., Spargo, A. T., Jones, J. A., Buttle, J. M., Adams, M. B., Beall, F. D., Booth, E. G., Campbell, J. L., Clow, D., Elder, K., Green, M. B., Grimm, N. B., Miniat, C., Ramlal, P., Saha, A., Sebestyen, S., Spittlehouse, D., Sterling, S., Williams, M. W., Winkler, R. & Yao, H. Changing forest water yields in response to climate warming: results from long-term experimental watershed sites across North America. Global Change Biol. 20, 3191–3208. Cuo, L., Zhang, Y., Gao, Y., Hao, Z. & Cairang, L. The impacts of climate change and land cover/use transition on the hydrology in the upper Yellow River Basin, China. J. Hydrol. 502, 37–52. Dai, A. Drought under global warming: a review. Clim. Change 2, 45–65. Dewes, C. F., Rangwala, I., Barsugli, J. J., Hobbins, M. T. & Kumar, S. Drought risk assessment under climate change is sensitive to methodological choices for the estimation of evaporative demand. PLoS One 12, e0174045. Duan, L., Liu, T., Wang, X., Luo, Y. & Wu, L. Development of a regional regression model for estimating annual runoff in the Hailar River Basin of China. J. Water Resour. Prot. 2, 934–943. Fang, Q., Wang, G., Liu, T., Xue, B.-L. & Aa, Y. a Controls of carbon flux in a semi-arid grassland ecosystem experiencing wetland loss: vegetation patterns and environmental variables. Agric. For. Meteorol. 259, 196–210. Fang, Q., Wang, G., Xue, B., Liu, T. & Kiem, A. b How and to what extent does precipitation on multi-temporal scales and soil moisture at different depths determine carbon flux responses in a water-limited grassland ecosystem? Sci. Total Environ. 635, 1255–1266. Fang, Q., Wang, G., Liu, T., Xue, B., Sun, W. & Shrestha, S. Unraveling the sensitivity and nonlinear response of water use efficiency to the water–energy balance and underlying surface condition in a semiarid basin. Sci. Total Environ. 699, 134405. https://doi.org/10.1016/j.scitotenv.2019.134405.
Hydrology Research
|
51.5
|
2020
Gao, X., Zhao, Q., Zhao, X., Wu, P., Pan, W., Gao, X. & Sun, M. Temporal and spatial evolution of the standardized precipitation evapotranspiration index (SPEI) in the Loess Plateau under climate change from 2001 to 2050. Sci. Total Environ. 595, 191–200. Gentine, P., D’Odorico, P., Lintner, B. R., Sivandran, G. & Salvucci, G. Interdependence of climate, soil, and vegetation as constrained by the Budyko curve. Geophysical Research Letters 39, L19404. http://doi.org/10.1029/ 2012gl053492. Gerten, D., Lucht, W., Schaphoff, S., Cramer, W., Hickler, T. & Wagner, W. Hydrologic resilience of the terrestrial biosphere. Geophys. Res. Lett. 32, L21408. https://doi.org/10. 1029/2005GL024247. Han, D., Wang, G., Liu, T., Xue, B.-L., Kuczera, G. & Xu, X. a Hydroclimatic response of evapotranspiration partitioning to prolonged droughts in semiarid grassland. J. Hydrol. 563, 766–777. Han, D., Wang, G., Xue, B., Liu, T., Y, A. & Xu, X. b Evaluation of semiarid grassland degradation in North China from multiple perspectives. Ecol. Eng. 112, 41–50. He, B., Miao, C. Y. & Shi, W. Trend, abrupt change, and periodicity of streamflow in the mainstream of Yellow River. Environ. Monit. Assess. 185, 6187–6199. Helman, D., Lensky, I. M., Yakir, D. & Osem, Y. Forests growing under dry conditions have higher hydrological resilience to drought than do more humid forests. Glob. Chang. Biol. 23, 2801–2817. Klein, T., Di Matteo, G., Rotenberg, E., Cohen, S. & Yakir, D. Differential ecophysiological response of a major Mediterranean pine species across a climatic gradient. Tree Physiol. 33, 26–36. Liang, L. Q., Li, L. J., Liu, C. M. & Cuo, L. Climate change in the Tibetan Plateau Three Rivers Source Region: 1960–2009. Int. J. Climatol. 33, 2900–2916. Miao, L., Jiang, C., Xue, B., Liu, Q., He, B., Nath, R. & Cui, X. Vegetation dynamics and factor analysis in arid and semi-arid Inner Mongolia. Environ. Earth Sci. 73, 2343–2352. Polong, F., Chen, H., Sun, S. & Ongoma, V. Temporal and spatial evolution of the standard precipitation evapotranspiration index (SPEI) in the Tana River Basin, Kenya. Theor. Appl. Climatol. 138, 777–792. Ponce-Campos, G. E., Moran, M. S., Huete, A., Zhang, Y., Bresloff, C., Huxman, T. E., Eamus, D., Bosch, D. D., Buda, A. R., Gunter, S. A., Scalley, T. H., Kitchen, S. G., McClaran, M. P., McNab, W. H., Montoya, D. S., Morgan, J. A., Peters, D. P. C., Sadler, E. J., Seyfried, M. S. & Starks, P. J. Ecosystem resilience despite large-scale altered hydroclimatic conditions. Nature 494, 349–352. Qiu, H., Niu, J. & Phanikumar, M. S. Quantifying the space– time variability of water balance components in an agricultural basin using a process-based hydrologic model and the Budyko framework. Sci. Total Environ. 676, 176–189.
1035
J. Liu et al.
|
Climate change influences on water balance changes
Shen, Q., Cong, Z. & Lei, H. Evaluating the impact of climate and underlying surface change on runoff within the Budyko framework: a study across 224 catchments in China. J. Hydrol. 554, 251–262. Sun, W., Jin, Y., Yu, J., Wang, G., Xue, B., Zhao, Y., Fu, Y. & Shrestha, S. Integrating satellite observations and human water use data to estimate changes in key components of terrestrial water storage in a semi-arid region of North China. Sci. Total Environ. 698, 134171. https://doi. org/10.1016/j.scitotenv.2019.134171. Sung, J. H., Chung, E.-S., Kim, Y. & Lee, B.-R. Meteorological hazard assessment based on trends and abrupt changes in rainfall characteristics on the Korean peninsula. Theor. Appl. Climatol. 127, 305–326. Trenbath, B. R. Multispecies cropping systems in India – predictions of their productivity, stability, resilience and ecological sustainability. Agrofor. Syst. 45, 81–107. Troch, P. A., Carrillo, G., Sivapalan, M., Wagener, T. & Sawicz, K. Climate-vegetation-soil interactions and long-term hydrologic partitioning: signatures of catchment coevolution. Hydrol. Earth Syst. Sci. 17, 2209–2217. Vicente-Serrano, S. M., Beguería, S. & López-Moreno, J. I. A multiscalar drought index sensitive to global warming: the standardized precipitation evapotranspiration index. J. Clim. 23, 1696–1718. Wang, D. & Hejazi, M. Quantifying the relative contribution of the climate and direct human impacts on mean annual streamflow in the contiguous United States. Water Resour. Res. 47, W00J12. https://doi.org/10.1029/2010WR010283. Wang, G., Yang, H., Wang, L., Xu, Z. & Xue, B. Using the SWAT model to assess impacts of land use changes on runoff generation in headwaters. Hydrol. Process. 28, 1032–1042. Wang, G., Hu, X., Zhu, Y., Jiang, H. & Wang, H. a Historical accumulation and ecological risk assessment of heavy metals in sediments of a drinking water lake. Environ. Sci. Pollut. Res. Int. 25 (24), 882–24,894. Wang, G., Liu, S., Liu, T., Fu, Z., Yu, J. & Xue, B. b Modelling above-ground biomass based on vegetation indexes: a modified approach for biomass estimation in semi-arid grasslands. Int. J. Remote Sens. 40, 3835–3854. Wang, G., Li, J., Sun, W., Xue, B., Y, A. & Liu, T. a Non-point source pollution risks in a drinking water protection zone based on remote sensing data embedded within a nutrient budget model. Water Res. 157, 238–246. Wang, P. Z., Yao, J. P., Wang, G. Q., Hao, F. H., Shrestha, S., Xue, B. L., Xie, G. & Peng, Y. B. b Exploring the application of artificial intelligence technology for identification of water pollution characteristics and tracing the source of water quality pollutants. Sci. Total Environ. 693. UNSP 133440, https://doi.org/10.1016/j.scitotenv.2019.07.246. Williams, C. A., Reichstein, M., Buchmann, N., Baldocchi, D., Beer, C., Schwalm, C., Wohlfahrt, G., Hasler, N., Bernhofer, C., Foken, T., Papale, D., Schymanski, S. & Schaefer, K.
Hydrology Research
|
51.5
|
2020
Climate and vegetation controls on the surface water balance: synthesis of evapotranspiration measured across a global network of flux towers. Water Resour. Res. 48, W06523. https://doi.org/10.1029/2011WR011586. Xue, B.-L., Komatsu, H., Kumagai, T. O., Kotani, A., Otsuki, K. & Ohta, T. Interannual variation of evapotranspiration in an eastern Siberian larch forest. Hydrol. Process. 26, 2360–2368. Xue, B.-L., Wang, L., Li, X., Yang, K., Chen, D. & Sun, L. Evaluation of evapotranspiration estimates for two river basins on the Tibetan Plateau by a water balance method. J. Hydrol. 492, 290–297. Xue, B.-L., Guo, Q., Otto, A., Xiao, J., Tao, S. & Li, L. Global patterns, trends, and drivers of water use efficiency from 2000 to 2013. Ecosphere 6. UNSP 174, http://dx.doi.org/10.1890/ ES14-00416.1. Xue, B.-L., Guo, Q., Gong, Y., Hu, T., Liu, J. & Ohta, T. The influence of meteorology and phenology on net ecosystem exchange in an eastern Siberian boreal larch forest. J. Plant Ecol. 9, 520–530. Xue, B.-L., Guo, Q., Hu, T., Xiao, J., Yang, Y., Wang, G., Tao, S., Su, Y., Liu, J. & Zhao, X. Global patterns of woody residence time and its influence on model simulation of aboveground biomass. Glob. Biogeochem. Cycles 31, 821–835. Yang, H. & Yang, D. Derivation of climate elasticity of runoff to assess the effects of climate change on annual runoff. Water Resour. Res. 47. UNSP 174, https://doi.org/10.1029/ 2010WR009287. Yao, J. P., Wang, P. Z., Wang, G. Q., Shrestha, S. G., Xue, B. L. & Sun, W. C. Establishing a time series trend structure model to mine potential hydrological information from hydrometeorological time series data. Sci. Total Environ. 698, 134227. https://doi.org/10.1016/j.scitotenv.2019.134227. Zhai, J. Q., Liu, B., Hartmann, H., Su, B. D., Jiang, T. & Fraedrich, K. Dryness/wetness variations in ten large river basins of China during the first 50 years of the 21st century. Quat. Int. 226, 101–111. Zhang, L., Dawes, W. R. & Walker, G. R. Response of mean annual evapotranspiration to vegetation changes at catchment scale. Water Resour. Res. 37, 701–708. Zhang, Y., Pena-Arancibia, J. L., McVicar, T. R., Chiew, F. H., Vaze, J., Liu, C., Lu, X., Zheng, H., Wang, Y., Liu, Y. Y., Miralles, D. G. & Pan, M. Multi-decadal trends in global terrestrial evapotranspiration and its components. Sci. Rep. 6, 19124. https://doi.org/10.1038/srep19124. Zhao, G. J., Tian, P., Mu, X. M., Jiao, J. Y., Wang, F. & Gao, P. Quantifying the impact of climate variability and human activities on streamflow in the middle reaches of the Yellow River basin, China. J. Hydrol. 519, 387–398. Zhao, H., Huang, W., Xie, T., Wu, X., Xie, Y., Feng, S. & Chen, F. Optimization and evaluation of a monthly air temperature and precipitation gridded dataset with a 0.025 spatial resolution in China during 1951–2011. Theor. Appl. Climatol. 138, 491–507.
First received 9 March 2020; accepted in revised form 15 May 2020. Available online 7 July 2020
1036
© 2020 The Authors Hydrology Research
|
51.5
|
2020
The spatial pattern of periphytic algae communities and its corresponding mechanism to environmental variables in the Weihe River Basin, China Yixin Liu, Jiaxu Fu, Dandong Cheng, Qidong Lin, Ping Su, Xinxin Wang and Haotian Sun
ABSTRACT Periphytic algae is a useful indicator of aquatic ecological conditions. We investigated the periphytic algae on natural substrate and the environmental variables at 44 sites on three river systems in the Weihe River Basin (WRB). A total of 84 species are identified, representing 37 genera. The most common genera were Navicula, Oscillatoria, Nitzschia, Scenedesmus, Cymbell, and Fragilaria. One-way analysis of variance (ANOVA) indicated significant differences among the three river systems in environmental variables (p < 0.05). Non-metric multidimensional scaling (NMDS) analyses also showed differences in periphytic algae communities in the three river systems (p < 0.05) and identified different dominant species in each river system. Canonical correspondence analysis (CCA) and Monte Carlo permutation tests revealed that nutrient concentration, WT, and altitude were the most important variables affecting the structure and distribution of periphytic algae communities. Chemical variables were the most accounted for environmental variables (12.5%), while physical variable and geographical factors (5.8% in total) play a relevant minor role. Our results demonstrate that Navicula pupula, Navicula radiosq, Nitzschia palea, and Nitzschia denticula, exhibiting wide ecological amplitude, are tolerant of high concentrations of nutrient pollution. Variation of periphytic algae communities in WRB is due to the combination of anthropogenic and natural factors including agricultural and domestic wastes water inputting, land use patterns, geology, climatic changes, and river hydrology. Key words
| CCA, environmental variables, nutrients, periphytic algae, Weihe River Basin
HIGHLIGHTS
• • • • •
Structure and distribution of periphytic algae communities were investigated in the Weihe River Basin. The gradient of organic pollution detected in environmental variables was the main gradient, and the periphytic algae responded to this gradient. The types of substrates in river impact periphytic algae abundance, causing higher periphytic algae abundance in cobbles substrate. The variation of periphytic algae communities is attributed to the combination of anthropogenic factors and natural factors. The results support the periphytic algae as ecological indicators in the Weihe River Basin.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/ licenses/by-nc-nd/4.0/) doi: 10.2166/nh.2020.031
Yixin Liu Jiaxu Fu Dandong Cheng (corresponding author) Qidong Lin Ping Su Xinxin Wang Haotian Sun Shaanxi Key Laboratory of Earth Surface System and Environmental Carrying Capacity, College of Urban and Environmental Sciences, Northwest University, Xi’an 710127, China E-mail: chengdandong@hotmail.com
1037
Y. Liu et al.
|
Periphytic algal communities and responses to environmental variables
Hydrology Research
|
51.5
|
2020
INTRODUCTION Periphytic algae are found in bodies of water ranging from
With the reform and opening-up policy in the 1980s,
small ponds to the global ocean. They are a major contribu-
China began a period of rapid urbanization. The proportion
tor to primary productivity in the river, playing a vital role in
of the urban population increased from 17.9% in 1978 to
aquatic ecosystems (Kolmakov et al. ; Pandey et al.
52.6% in 2012, while urban built-up areas increased by
). Periphytic algae have been widely associated with
78.5% (Bai et al. ). At the same time, rapid urbanization
different geographical regions with specific environmental
has brought a series of complex water environmental pro-
conditions (Soininen ; Kovács et al. ; Chen et al.
blems. It greatly affects the hydrological system by channel
a). Composition and relative abundance of the periphy-
morphology, water quality, and aquatic biota (White &
tic algae community vary with their ecological affinity and
Greer ). The Weihe River Basin (WRB) is the main
preferences (Potapova & Charles ; Tornés et al. ).
grain-producing area, and is an important industrial and com-
Their structure and distribution are reflected in spatially
mercial center in northwest China. The Weihe River is a major
and temporally based geomorphological conditions, climate,
source of domestic water, industrial water, and irrigation in the
hydrology, and water physicochemical properties (Pan et al.
central plain. There are 76 major cities with a total population
; Bere et al. ). Based on observations in the natural
of 22 million in the basin (Chang et al. ). Due to the
river and experiments, studies have shown that the algae
increase in population, industry, and farmland area, the WRB
will not be nutrient-limited when the nutrients concen-
also faces serious water environmental problems. Presently,
tration is P > 30 μg/L and N > 1 mg/L in the river
most studies have focused on the water quality and quantity
ecosystem (Westlake ). Water temperature affects algal
of the WRB (Jiake et al. ; Wei et al. ; Song et al. ).
photosynthetic metabolism through its control of enzyme
Research on the effects of urbanization on the aquatic biota
reaction rates (Stevenson et al. ). Periphytic algae abun-
in WRB has typically focused on fish communities and macro-
dance reaches a peak at a certain level with suitable bed
invertebrates (Wu et al. ; Su et al. ), with few studies
light levels (Yang & Flower ). Hydrodynamic conditions
available on algae communities. Studies on periphytic algae
also influence periphytic algae biomass, and the suitable
communities and their relationship with environmental vari-
velocity and river depth are 0.9–1.1 m/s and 0.40–0.48 m,
ables have scarcely been carried out in the WRB.
respectively (Wang et al. a). Periphytic algae are
This study focuses on the abundance, species structure,
effective environmental indicators due to their wide distri-
and distribution of periphytic algae communities in the WRB.
to
The specific objectives are to investigate the relationships
environmental changes (Stevenson et al. ). They
between periphytic algae communities and environmental vari-
respond rapidly to changes in environmental variables,
ables, to detect which ecological factors explain most of the
reflecting the overall ecological quality and the effects of
variation. Our results may provide a valuable baseline for
different stressors (Bona et al. ; Vasiljević et al. ).
future water quality assessments in the WRB in China.
bution
range,
numerous
species,
and
sensitivity
Individual algae species often show a clear preference for specific substrates and habitats (Winter & Duthie ). Substantial differences exist in species composition and
MATERIALS AND METHODS
abundance of periphytic algae communities from the same sites but different substrates such as sand, rock surface, submerged, or emergent macrophytes (Fisher & Dunbar ;
Study area
Bere & Tundisi ). Therefore, understanding the relation-
The study was carried out in the WRB (Figure 1), which is
ship between environmental factors and the distribution of
located in the northwest of China. The WRB (east longitude
periphytic algae communities is important for developing
103 50 –110 400 , north latitude 33 400 –37 250 ) has a drainage
algae-based water quality indices.
area of 1.35 × 105 km2, with annual mean temperatures
Y. Liu et al.
1038
Figure 1
|
|
Periphytic algal communities and responses to environmental variables
Hydrology Research
|
51.5
|
2020
The locations of the sampling sites in the Weihe River Basin.
between 7.8 and 13.5 C and the annual mean rainfall ran-
systems has a different number of sampling sites: 15 in
ging from 558 to 750 mm which increases from north to
WRS (W1–W15), 17 in JRS (J1–J17) and 12 in BRS (B1–
south (Wang et al. b). The climate in the WRB is charac-
B12) (Figure 1). For the cobble-type sedimentary rivers,
terized by the temperate continental monsoon. Therefore,
within the range of 100 m upstream and downstream of
precipitation and runoff exhibited strong inter-annual vari-
the sampling point, nine stones were selected (surface area
ations and similar intra-annual variations. The runoff from
of the stones <200 cm2), and scraped with a toothbrush to
July to October is approximately 65% of the mean annual
a 7.07 cm2 area (Kelly et al. ). For the silt-type sedimen-
runoff (Chang et al. ). Also, most of the WRB is covered
tary rivers, the surface layer of silt on the sampled stones
by highly erodible loess and soil erosion is serious because
were scraped with an area of 63.62 cm2 using an inverted
of sparse vegetation, uneven rainfall distribution, and high
Petri dish along the riverbed (Carpenter & Waite ).
intensity of heavy rain (Song et al. ). There are three
The samples were preserved with 4% formaldehyde and
different water systems in the WRB, including the Weihe
2% Lugol’s iodine solution. An aliquot of the periphytic
River System (WRS), the Jinghe River System (JRS), the Bei-
algae suspension was cleaned by using hydrochloric acid
luoRiver System (BRS), and their tributaries, as well as other
and hydrogen peroxide. The acid-cleaned samples were
small independent streams (Figure 1).
mounted on microscope slides applying the high refraction mountant Naphrax. Species were identified and a minimum
Sampling and analysis
of 400 valves were counted per slide at 1,000× magnification using a microscope (Olympus BX51, Olympus, Tokyo,
Periphyton samples were collected from 44 sites in the three
Japan). Algal cells identification was carried out according
water systems of WBR in June 2017, each of the water
to the standard guides of Krammer & Lange-Bertalot
1039
Y. Liu et al.
|
Periphytic algal communities and responses to environmental variables
Hydrology Research
|
51.5
|
2020
(, , a, b), Zhu & Chen (), and Hu & Wei
structure. Both NMDS and ANOSIM analyses were per-
(). The relative abundance of each observed taxa was
formed using PRIMER version 5 (Clarke & Warwick ). A total of 18 rivers geographical, hydrological, physical
calculated. Basic physical and chemical water parameters were
and chemical variables were considered for data analysis,
measured directly at the sampling sites during sampling.
including Altit, WT, width, depth, SD, velocity, flux, pH,
The water temperature (WT), pH, dissolved oxygen (DO),
DO, EC, TDS, NO3-N, NH4-N, Kþ, Mg2þ, Cl , TN and
electrical conductivity (EC), and total dissolved solids
TP. All of the environmental variables, except pH, were
(TDS) were measured in situ using a portable water quality
transformed as log10 (x þ 1) before analysis. One-way
meter (HACH HQ40d). River velocity was acquired by
analysis of variance (ANOVA) was used to evaluate
using a portable flow meter (MGG/KL-DCB); river width
the significance of the differences among the WRS, the
data was obtained with the help of Trupulse 200; river
JRS and the BRS sites based on the physical and chemical
depth was measured by using a terrain probe; Secchi
variables of the transformed water, completed by IBM
depth (SD) was measured using a Secchi disc, except for
SPSS Statistics for Windows (25). Principal component
some of the restored rivers that were transparent to the
analysis (PCA) was applied to explore the main environ-
bottom; latitude longitude and altitude (Altit) information
mental
was obtained with a Global Positioning System (GPS
CANOCO 5.0 (Ter Braak & Smilauer ).
gradients
among
the
sampling
sites,
using
Etrex 201X). Two parallel water samples were collected
Multivariate analyses are mathematical tools that detect
with a water sampler and poured into a 1,000 mL bottle
the relationship between periphytic algae and environ-
at each sampling point. The water sample was fixed with
mental variables, indicating the main variables and
acid, stored in a 4 C incubator and transported back to
revealing the similarities among algae samples. For analysis
the laboratory to measure the concentrations of nitrate
of their relation, 28 species were retained with rare species
(NO3-N, mg·L–1), ammonium (NH4-N, mg·L–1), total nitro-
(<1%) removed. The removal of the species with a relative
–1
–1
gen (TN, mg L ), total phosphorus (TP, mg·L ) following
abundance of less than 1% can minimize the influence of
the Chinese Government standard for Water and Waste-
rare species in the analysis. Algae abundance data were
þ
water Monitoring and Analysis (). Major ions (K ,
transported by log10 (x þ 1) to stabilize variances and give
Mg2þ, Cl ) were measured using the Dionex Ion Chrom-
more weight to the larger species often found at low relative
atography System (ICS; ICS-1000, Dionex, Sunnyvale,
abundance in periphytic algae communities, which are
CA, USA).
important for defining assemblage (Tison et al. ). Detrended correspondence analysis (DCA) was conducted to detect the gradient length of the algae abundance data.
Data analysis
Gradient lengths were then used to select the appropriate model (linear or unimodal model) for the constrained
The counts of each algae taxon were expressed as relative
ordinations. In this study, the longest gradient of four
abundance before the analysis. The structural properties of
ordination axes was 3.8, indicating that a linear or unimodal
the periphytic algae community at each site were used to
model could be applied to the ordination analysis (Ter
0
characterize by the Shannon–Weiner index (H ) and species
Braak & Verdonschot ). Hence, the unimodal ordina-
richness (S) (Bellinger et al. ; Wang et al. ). These
tion technique of canonical correspondence analysis
indicators are commonly used in the biological assessment
(CCA) was used to assess the relationships between environ-
of water quality (Spellerberg & Fedor ). Non-metric
mental variables and periphytic algae communities from
multidimensional scaling (NMDS) analyses were performed
different sites. A Monte Carlo permutation test (999 permu-
to visualize the periphytic algae structure distribution
tations, p 0.05) was used to reduce the environmental
characteristics of the WRB. Furthermore, the analysis of
variables to those correlating significantly with the first
similarity (ANOSIM, Bray-Curtis distance measure, 999 per-
two CCA axes. Both DCA and CCA were performed using
mutations) was used to test the difference in the community
CANOCO 5.0 (Ter Braak & Smilauer ).
Y. Liu et al.
1040
|
Periphytic algal communities and responses to environmental variables
Hydrology Research
|
51.5
|
2020
RESULTS Environmental gradients In this study, considerable fluctuations were identified in hydrological characteristics, water physical and chemical properties among sampling sites (Table 1). One-way ANOVA indicated that among all the environmental variables, width, pH, TDS, NO3-N, Mg2þ, and Cl– were significantly different, p < 0.05 (Table 1). The average width and the concentration of NO3-N in the BRS sites were lower than those in the WRS and the JRS (p < 0.05), while the concentrations of TDS and Cl– in the BRS sites were higher than other sites (p < 0.05). The average pH and the concentration of Mg2þ in the BRS sites was higher
Figure 2
|
Analysis of the principal components (PCA) of sampling sites based on environmental variables. WRS, JRS, and BRS are indicated by different symbols.
than in WRS and JRS (p < 0.001). Principal component analysis (PCA) accounted for 63.17% of the total variability in the environmental data in the first two axes (Figure 2).
TDS (Figure 2). The second axis explained 21.44% of the
The first axis of PCA explained 41.73% of the variance, indi-
variance, mainly correlating with pH, flux, and river width
–
þ
cating that variables were primarily related to Cl , K , and Table 1
|
(Figure 2).
Mean value (range in parentheses) from three water system sites in July 2017
WRS
JRS
BRS
Abbreviation
n ¼ 15
n ¼ 17
n ¼ 12
P
Altit
913 (340–2,452)
1,049 (410–1,950)
821 (340–1,360)
0.162
Water temperature ( C)
WT
22 (6.5–28.5)
21.5 (9.0–25.5)
26 (20.3–29.5)
0.088
River width (m)
Width
73.99 (12.3–210)
40.21 (16.9–105)
33.13 (2.5–133)
0.033*
River depth (m)
Depth
0.47 (0.2–1)
0.47 (0.25–0.65)
0.42 (0.25–0.5)
0.671
Water velocity (m/s)
Velocity
0.54 (0.29–1.02)
0.55 (0.1–1.29)
0.52 (0.07–1.27)
0.901
Environment variables
Altitude (m)
3
Flux (m /s)
Flux
16.53 (2.87–70.40)
8.71 (1.93–19.53)
6.11 (0.47–16.46)
0.088
Secchi depth (m)
SD
0.16 (0.07–0.52)
0.10 (0–0.35)
0.08 (0–0.3)
0.134
Dissolved oxygen (mg/L)
DO
8.45 (5.55–10.03)
8.69 (7.24–11.92)
9.36 (3.98–13.93)
0.79
pH
pH
9.00 (8.16–9.59)
9.31 (8.01–10.05)
9.63 (9.05–10.14)
0.001**
Electrical conductivity (μs/cm)
EC
837.07 (391–1,618)
789.02 (132.4–1,893)
1,274.33 (336–2,155)
0.075
Total dissolved solids (mg/L)
TDS
372.75 (125–845)
682.45 (89.6–2,115)
893.17 (312–1,944)
0.003*
Total nitrogen (mg/L)
TN
5.55 (1.09–8.21)
4.81 (1.84–9.76)
5.86 (2.78–10.05)
0.423
Total phosphorus (mg/L)
TP
0.11 (0.03–0.25)
0.20 (0.03–1.95)
0.13 (0.02–0.47)
0.808
Nitrate (mg/L)
NO3-N
2.31 (0.19–4.19)
1.21 (0.04–2.50)
1.65 (0.47–3.59)
0.018*
Ammonium (mg/L)
NH4-N
0.34 (0.09–0.63)
0.31 (0.10–0.82)
0.29 (0.01–0.74)
0.687
Potassium (mg/L)
Kþ
5.07 (0.49–14.68)
5.12 (0.50–11.87)
5.21 (3.82–7.25)
0.683
2þ
Magnesium (mg/L)
Mg
21.28 (7.64–59.54)
52.05 (4.87–131.69)
70.40 (20.6–135.72)
0.000**
Chloride (mg/L)
Cl-
145.36 (1.02–380.23)
265.71 (0.84–824.23)
486.26 (160.89–960.55)
0.008*
The value of n is the number of water samples in the sites. Variables significantly different among the river systems with p < 0.05 and p < 0.001 are marked (based on the ANOVA). *p < 0.05; **p < 0.001.
Y. Liu et al.
1041
|
Periphytic algal communities and responses to environmental variables
Hydrology Research
|
51.5
|
2020
Spatial distribution of periphytic algae community
vulgare, Oscillatoria tenuis and Phormidium tenue. In the
structure
BRS sites, Navicula, Scenedesmus and Oscillatori were dominant species, but the most abundant species different
A total of 84 periphytic algae taxa species were identified,
from WRS and JRS were Navicula gracilis, Navicula
distributing in 37 genera. Only 28 species and 14 genera
pupula and Scenedesmus quadricauda.
were found to reach a relative abundance of more than
In the analyzed set of samples, the density of periphytic
1% (Figure 3). The most common genera were Navicula
algae ranged from 525 to 5.9 × 105 ind/cm2, species richness
(46.5% of all counted), Oscillatoria (5.8%), Nitzschia
from 2 to 29, and the Shannon–Weiner index from 0.58 to
(4.8%), Scenedesmus (4.0%), Cymbell (3.6%), Fragilaria
4.43 (Figure 4). The periphytic algae density showed a low
(3.3%), Diatoma (3.2%), Phormidium (3.1%), Synedra
level in the whole WRB. The average of periphytic algae
(2.8%), Cyclotella (2.0%), Srauroneis (1.9%), Cocooneis
density was 7.35 × 104 ind/cm2 while the highest density
(1.3%), Achnanthes (1.2%) and Melosira (1.0%). These 14
and the lowest density appeared at sites J1 and J11, respect-
genera accounted for 84.5% of all the values counted
ively (Figure 4(a)). The average species richness of the
(Figure 3). The dominant genus of each site differed in rela-
periphytic algae communities in WRS, JRS, and BRS were
tive abundance. In the WRS sites, Navicula, Fragilaria,
13, 11, and 14, respectively, with sites J2 and J9 displaying
Scenedesmus, and Melosira occupied a higher relative abun-
the highest and lowest species richness (Figure 4(b)). The
dance (20% in at least one sample). Navicula radiosq,
average Shannon–Weiner indexes in WRS, JRS, and BRS
Navicula palcentula, Fragilaria intermedia, Scenedesmus
were 2.90, 2.45, and 2.99, respectively. The peak and
acutiformis and Melosira granulata were the most dominant
valley values of the Shannon–Weiner index appeared at
species (a relative abundance of 10% in at least one sample).
sites J2 and J15 (Figure 4(c)). According to the results of
In the JRS sites, Navicula, Cymbella, Diatoma, Oscillatoria
NMDS and ANOSIM (r ¼ 0.291, p ¼ 0.001), the periphytic
and Phormidium occupied a higher relative abundance. The
algae community structures were significantly different
most abundant species were Cymbella ventricosa, Diatoma
among WRS, JRS, and BRS (Figure 5).
Figure 3
|
Percentage composition of 14 genera (relative abundance of each genus >1%) at the sampling sites.
Y. Liu et al.
1042
Figure 4
|
|
Periphytic algal communities and responses to environmental variables
Hydrology Research
|
51.5
|
2020
Structure and distribution of periphytic algae community in the Weihe River Basin: (a) algae density, (b) species richness, (c) Shannon–Weiner index.
Mg2þ, Cl–, TN) that significantly contributed to the algae species assemblages (Figure 6). The eigenvalues of the first two CCA axes (λ1 ¼ 0.316, λ2 ¼ 0.133) explained 49.1 and 20.7% of the variation in species and environmental variables, respectively (Table 2). The species-environment correlations were high for both of the ordination axes (r1 ¼ 0.833, r2 ¼ 0.842). Axis 1 was positively correlated with TDS (r ¼ 0.74), EC (r ¼ 0.48), WT (r ¼ 0.41), TN (r ¼ 0.35), Kþ (r ¼ 0.41), Mg2þ (r ¼ 0.73), Cl– (r ¼ 0.69) and negatively with Altit (r ¼ –0.23). Axis 2 was positively correlated with Altit (r ¼ 0.61) and negatively with EC (r ¼ 0.49) and WT (r ¼ –0.4). Figure 5
|
Nonparametric multidimensional scaling (NMDS) ordination of sampling sites.
The environmental variables play important roles in the distribution of periphytic algae communities. Frequently
Relationship between periphytic algae communities
found species such as Cocconeis placentula, Navicula
and environment variables
radiosq, Navicula pupula, Stauroneis ancepse, Nitzschia denticule, Nitzschia palea and Synedra acus exhibited
CCA analyzed 28 species and 18 environmental variables
wide tolerance to environmental variables (Figure 6(a)).
at 44 sites (Supplementary material, Table A.1). The
These species were positively associated with Altit, WT,
Monte Carlo unrestricted permutation test (p < 0.05) ident-
EC, and TN. Most sites in WRS and JRS were distributed
ified eight environmental variables (Altit, WT, TDS, EC, Kþ,
in the second and the third quadrant, where the samples
Y. Liu et al.
1043
Figure 6
|
|
Periphytic algal communities and responses to environmental variables
Hydrology Research
|
51.5
|
2020
Canonical correspondence analysis (CCA) showed the relationship of periphytic algae species (a) and sampling sites (b) with significant environmental variables. Legend: sp1: Achnanthes dispar, sp2: Cocconeis placentula, sp3: Cyclotella meneghiniana, sp4: Cymbella ventricosa, sp5: Diatoma vulgare, sp6: Fragilaria intermedia, sp7: Melosira granulate, sp8: Navicula cari, sp9: Navicula cari var.angusta, sp10: Navicula dicephala, sp11: Navicula exigua, sp12: Navicula gracilis, sp13: Navicula graciloides, sp14: Navicula minuscula, sp15: Navicula palcentula, sp16: Navicula pupula, sp17: Navicula radiosq, sp18: Navicula simplex, sp19: Nitzschia denticule, sp20: Nitzschia palea, sp21: Oscillatoria tenuis, sp22: Phormidium tenue, sp23: Scenedesmus acutiformis, sp24: Scenedesmus bijugayus, sp25: Scenedesmus quadricauda, sp26: Stauroneis anceps, sp27: Synedra acus, sp28: Synedra ulna.
Table 2
|
Summary of the CCA of most dominant periphytic algae species composition in
environmental stressors, owing to urbanization and industri-
44 samples concerning the eight environmental variables
alization taking place in the entire watershed, contributing to
Axes
1
2
3
4
Eigenvalues
0.316
0.133
0.055
0.045
Species-environment correlations
0.833
0.842
0.725
0.665
Cumulative percentage variance of species data
15.9
22.7
25.4
27.7
Species–environment relation
49.1
69.8
78.3
85.3
nutrients and organic pollution continuously (Song et al. ). Wastewater from industrial and agricultural activities, as well as an urban settlement, is the main source that contributes a great number of solid organic pollution and thermal pollution (Milovanovic ; Luo et al. ; N’guessan et al. ). Algae are an essential component in maintaining the health of aquatic ecosystems (Kelly et al.
were closely related to Altit (Figure 6(b)). The parameter
). In general, periphytic algae are regarded as a good indi-
was highly positively associated with Navicula simplex,
cator of water quality. Many studies have demonstrated that
Scenedesmus acutiformis, Scenedesmus bijugayus and
the composition and structure of the periphytic algae commu-
Scenedesmus quadricauda. However, the sites in the BRS
nity are affected by various environmental variables
were mainly dispersed in the fourth quadrant. Samples
(Chessman et al. ; Soininen et al. ; Urrea & Sabater
were associated with high WT and EC (Figure 6(b)).
; Panahy Mirzahasanlou et al. ). In this study, we compared the species composition and distribution of periphytic algae communities among three
DISCUSSION
river systems. Though each river system was represented and dominated by different periphytic algae taxa, we found
Rivers, an important part of the ecosystem, not only have eco-
that the genera Navicula existed at all samples and species
logical functions but also provide various services for people.
N. radiosq, N. pupula and N. simplex were found in almost
However, river ecosystems are increasingly impacted by
every site. From the whole WRB, the species diversity and
1044
Y. Liu et al.
|
Periphytic algal communities and responses to environmental variables
Hydrology Research
|
51.5
|
2020
abundance of periphytic algae changed significantly. Mean-
the concentration of various particles, pollutants and ions
while, we found that algae abundance was higher at the
(Quilbé et al. ). EC is an index that can reflect the total
sites of JRS (J1–J5), part sites of WRS (W11, W13) and BRS
ionic concentration in natural water. In this research, EC
(B4, B5, B10), where the forest coverage is high and the sub-
became an important determinant to distinguish the distri-
strate of the river is dominated by boulders and cobbles
bution of the periphytic algae communities in the WRB.
(Supplementary material, Table A.2). The impact of habitat
We found that the average values of EC concentration in
quality on the periphytic algae communities is more signifi-
downstream sites was 1,110 μm/cm, higher than 759 μm/cm
cant, especially the type of substrate (Eloranta & Andersson
measured in upstream sites. TDS and nutrient levels play
; Bere & Tundisi ). Soil and water loss in the WRB
an important role in benthic periphytic algae community
is serious, and a large amount of sediment is dumped into
structure. Most of the abundant and common species were
the main river, causing soft-sediment substrate at many sites.
characteristic of eutrophic ecosystems (Van Dam et al.
Periphytic algae abundance in the soft-sediment substrate is
). In the WRB, high TN concentration from anthropo-
lower than the stone substrate (Townsend & Gell ).
genic sources are likely to cause genera Naviacula (46.5%)
Periphytic algae communities were associated with
and Nitzschia (4.8%), which have wide ecological amplitude
three sets of environmental variables in the WRB. One set
and pollution tolerance, dominating in samples (Goma et al.
is a geographic variable, and only one variable is Altit. The
; Delgado & Pardo ; Chen et al. ).
other two sets are physical and chemical variables of the
WT can affect the concentration of dissolved oxygen in
WRB. The physical variable is mainly WT. Chemical vari-
water and the respiratory rate of aquatic organisms, so it
ables consist of TDS, TN, and nutrient ion concentration.
also is a major factor affecting the periphytic algae community
For the geographical and physical variables, the WT chan-
(Chen et al. b). Moreover, because that altitude drastically
ged significantly due to the large east–west span of the
drops from upstream to downstream in the WRB, WT also
WRB and drastic variation in altitude. The similarity of per-
changes evidently and becomes a significant factor affecting
iphytic algae community is related to the altitude distance,
the periphytic algae. Thermal pollution is equally a reason
and the altitude gradient affects the biodiversity by affecting
for WT rising. Thermal energy absorbed and stored by
the local environmental factors (Teittinen et al. ).
urban impervious underlying surface and rainwater runoff
Chemical variables are significantly correlated with periphy-
and point discharge from wastewater treatment plants
tic algae data indicating that they are a key factor for the
increase WT (Van Buren et al. ; Kinouchi et al. ).
distribution of periphytic algae communities (Potapova & Charles ; Soininen et al. ; Tan et al. ).
Recently, the impact of catchment-scale variables on river ecosystems has attracted researchers’ attention. A widely used
The three sets of environmental variables provided infor-
approach is to establish relationships between community pat-
mation on how they affected the periphytic algae community
terns and environmental variables (Liu et al. ). In this
structure (Leira & Sabater ; Blettler et al. ). The CCA
study, Altit, WT and nutrition concentration were retained in
indicated that the chemical variables (12.5%) were the most
the CCA as significantly affecting algal distribution environ-
accounted for in the contribution to the periphytic algae com-
mental factors. The impact of increased nutritional levels
munity structure in the WRB, while physical variables and
caused by human activities is significantly greater than natural
geographical factors (5.8% in total) played a relevant minor
factors. Our work has important implications for river bio-
role. The periphytic algae composition in the river system
monitoring and management in the study area, especially in
downstream sites was greatly influenced by high EC, Mg2þ,
other subtropical regions throughout China.
–
Cl and TN concentration. The high concentration of these variables may be due to the domestic sewage and industrial wastewater draining to the river (Ma et al. ). The other
CONCLUSIONS
main reason is due to soil erosion which not only carries sediment into the river, increasing water turbidity, but also
There are complex variables which affect the structure and
dissolves soluble nutrients and ions in the soil, increasing
distribution of periphytic algae communities, including
1045
Y. Liu et al.
|
Periphytic algal communities and responses to environmental variables
hydrological, geographical and physiochemical factors. Our results revealed that altitude, EC, TN, WT and major ions were identified as the main variables with a significant influence on the structure and distribution of periphytic algae communities. Three river systems were investigated, each of them with different dominant species, and corresponding to the different variables. Multivariate analyses are a good tool for interpreting species data but a large part of variation remained unexplained in most instances. This uncertainty may be related to other variables which can explain why variation and species have broad tolerance to the variables (Passy ; Centis et al. ; Porter-Goff et al. ; Tolkkinen et al. ). Consequently, the main determinants of variation of periphytic algae communities may result from a combination of the change in land-use patterns by man, natural phenomena including geology and climatic changes and river hydrology.
ACKNOWLEDGEMENTS This study was supported by the National Natural Science Foundation
of
China
(Grant
Nos.
51679200
and
51379175), Science and Technology Project of Shaanxi Provincial
Water
Resources
Department
(Grant
No.2018slkj-12).
DATA AVAILABILITY STATEMENT All relevant data are included in the paper or its Supplementary Information.
REFERENCES Bai, X. M., Shi, P. J. & Liu, Y. S. Realizing China’s urban dream. Nature 509 (7499), 158–160. Bellinger, B. J., Cocquyt, C. & O’Reilly, C. M. Benthic diatoms as indicators of eutrophication in tropical streams. Hydrobiologia 573, 75–87. Bere, T. & Tundisi, J. G. The effects of substrate type on diatom-based multivariate water quality assessment in a tropical river (Monjolinho), São Carlos, SP, Brazil. Water Air Soil Pollut. 216 (1–4), 391–409.
Hydrology Research
|
51.5
|
2020
Bere, T., Mangadze, T. & Mwedzi, T. Variation partitioning of diatom species data matrices: understanding the influence of multiple factors on benthic diatom communities in tropical streams. Sci. Total Environ. 566, 1604–1613. Blettler, M. C., Oberholster, P. J., Madlala, T., Eberle, E. G., Amsler, M. L., De Klerk, A. R., Truter, J. C., Marchese, M. R., Latosinski, F. G. & Szupiany, R. Habitat characteristics, hydrology and anthropogenic pollution as important factors for distribution of biota in the middle Paraná River, Argentina. Ecohydrol. Hydrobiol. 19 (2), 296–306. Bona, F., Falasco, E., Fassina, S., Griselli, B. & Badino, G. Characterization of diatom assemblages in mid-altitude streams of NW Italy. Hydrobiologia 583, 265–274. Carpenter, K. D. & Waite, I. R. Relations of habitat-specific algal assemblages to land use and water chemistry in the Willamette Basin, Oregon. Environ. Monit. Assess. 64 (1), 247–257. Centis, B., Tolotti, M. & Salmaso, N. Structure of the diatom community of the River Adige (North-Eastern Italy) along a hydrological gradient. Hydrobiologia 639 (1), 37–42. Chang, J., Wang, Y., Istanbulluoglu, E., Bai, T., Huang, Q., Yang, D. & Huang, S. Impact of climate change and human activities on runoff in the Weihe River Basin, China. Quat. Int. 380, 169–179. Chen, X., Bu, Z., Stevenson, M. A., Cao, Y., Zeng, L. & Qin, B. a Variations in diatom communities at genus and species levels in peatlands (central China) linked to microhabitats and environmental factors. Sci. Total Environ. 568, 137–146. Chen, X., Zhou, W., Pickett, S. T., Li, W., Han, L. & Ren, Y. b Diatoms are better indicators of urban stream conditions: a case study in Beijing, China. Ecol. Indic. 60, 265–274. Chen, S., Zhang, W., Zhang, J., Jeppesen, E., Liu, Z., Kociolek, J. P., Xu, X. & Wang, L. Local habitat heterogeneity determines the differences in benthic diatom metacommunities between different urban river types. Sci. Total Environ. 669 (1), 711–720. Chessman, B., Growns, I., Currey, J. & Plunkett-Cole, N. Predicting diatom communities at the genus level for the rapid biological assessment of rivers. Freshwater Biol. 41 (2), 317–331. Clarke, K. R. & Warwick, R. M. Changes in marine communities: an approach to statistical analysis and interpretation. Mt. Sinai J. Med. 40 (5), 689–692. Delgado, C. & Pardo, I. Comparison of benthic diatoms from Mediterranean and Atlantic Spanish streams: community changes in relation to environmental factors. Aquat. Bot. 120, 304–314. Eloranta, P. & Andersson, K. Diatom indices in water quality monitoring of some South-Finnish rivers. SIL Proc. 1922– 2010 26, 1213–1215. Fisher, J. & Dunbar, M. Towards a representative periphytic diatom sample. Hydrol. Earth Syst. Sci. Discuss. 11 (1), 399–407. Goma, J., Rimet, F., Cambra, J., Hoffmann, L. & Ector, L. Diatom communities and water quality assessment in
1046
Y. Liu et al.
|
Periphytic algal communities and responses to environmental variables
Mountain Rivers of the upper Segre basin (La Cerdanya, Oriental Pyrenees). Hydrobiologia 551, 209–225. Hu, H. J. & Wei, Y. X. The Freshwater Algae of China: Systematics, Taxonomy and Ecology. Science Press, Beijing, China. Jiake, L., Huaien, L., Bing, S. & Yajiao, L. Effect of non-point source pollution on water quality of the Weihe River. Int. J. Sediment Res. 26 (1), 50–61. Kelly, M., Cazaubon, A., Coring, E., Dell’Uomo, A., Ector, L., Goldsmith, B., Guasch, H., Hürlimann, J., Jarlman, A. & Kawecka, B. Recommendations for the routine sampling of diatoms for water quality assessments in Europe. J. Appl. Phycol. 10 (2), 215–224. Kinouchi, T., Yagi, H. & Miyamoto, M. Increase in stream temperature related to anthropogenic heat input from urban wastewater. J. Hydrol. 335 (1–2), 78–88. Kolmakov, V., Anishchenko, O., Ivanova, E., Gladyshev, M. & Sushchik, N. Estimation of periphytic microalgae gross primary production with DCMU-fluorescence method in Yenisei River (Siberia, Russia). J. Appl. Phycol. 20 (3), 289–297. Kovács, C., Kahlert, M. & Padisák, J. Benthic diatom communities along pH and TP gradients in Hungarian and Swedish streams. J. Appl. Phycol. 18 (2), 105–117. Krammer, K. & Lange-Bertalot, H. Bacillariophyceae 1 Teil: Naviculaceae. Süsswasserflora von Mitteleuropa. Gustav Fischer Verlag, Jena. Krammer, K. & Lange-Bertalot, H. Bacillariophyceae 2. Teil: Bacillariaceae, Epthemiaceae, Surirellaceae. Süsswasserflora von Mitteleuropa. Gustav Fischer Verlag, Stuttgart. Krammer, K. & Lange-Bertalot, H. a Bacillariophyceae 3. Teil: Centrales, Fragilariaceae, Eunotiaceae. Süsswasserflora von Mitteleuropa. Gustav Fischer Verlag, Stuttgart. Krammer, K. & Lange-Bertalot, H. b Bacillariophyceae 4. Teil: Achnanthaceae, Kritische Erg?nzungen zu Navicula (Lineolatae) und Gomphonema. Süsswasserflora von Mitteleuropa. Gustav Fischer Verlag, Stuttgart. Leira, M. & Sabater, S. Diatom assemblages distribution in Catalan rivers, NE Spain, in relation to chemical and physiographical factors. Water Res. 39 (1), 73–82. Liu, S., Xie, G., Wang, L., Cottenie, K., Liu, D. & Wang, B. Different roles of environmental variables and spatial factors in structuring stream benthic diatom and macroinvertebrate in Yangtze River Delta, China. Ecol Indic 61, 602–611. Luo, H. B., Luo, L., Huang, G., Liu, P., Li, J. X., Hu, S., Wang, F. X., Xu, R. & Huang, X. X. Total pollution effect of urban surface runoff. J. Environ. Sci. 21 (9), 1186–1193. Ma, J., Ding, Z., Wei, G., Zhao, H. & Huang, T. Sources of water pollution and evolution of water quality in the Wuwei basin of Shiyang river, Northwest China. J. Environ. Manage. 90 (2), 1168–1177. Milovanovic, M. Water quality assessment and determination of pollution sources along the Axios/Vardar River, Southeastern Europe. Desalination 213 (1–3), 159–173. N’guessan, Y. M., Probst, J.-L., Bur, T. & Probst, A. Trace elements in stream bed sediments from agricultural
Hydrology Research
|
51.5
|
2020
catchments (Gascogne region, SW France): where do they come from? Sci. Total Environ. 407 (8), 2939–2952. Pan, Y., Herlihy, A., Kaufmann, P., Wigington, J., Van Sickle, J. & Moser, T. Linkages among land-use, water quality, physical habitat conditions and lotic diatom assemblages: a multi-spatial scale assessment. Hydrobiologia 515 (1–3), 59–73. Panahy Mirzahasanlou, J., Ramezanpour, Z., Nejadsattari, T., Imanpour Namin, J. & Asri, Y. Temporal and spatial distribution of diatom assemblages and their relationship with environmental factors in Balikhli River (NW Iran). Int. J. Ecohydrol. Hydrobiol. 20 (1), 102–111. Pandey, L. K., Bergey, E. A., Lyu, J., Park, J., Choi, S., Lee, H., Depuydt, S., Oh, Y.-T., Lee, S.-M. & Han, T. The use of diatoms in ecotoxicology and bioassessment: insights, advances and challenges. Water Res. 118, 39–58. Passy, S. I. Diatom ecological guilds display distinct and predictable behavior along nutrient and disturbance gradients in running waters. Aquat. Bot. 86 (2), 171–178. Porter-Goff, E. R., Frost, P. C. & Xenopoulos, M. A. Changes in riverine benthic diatom community structure along a chloride gradient. Ecol. Indic. 32, 97–106. Potapova, M. G. & Charles, D. F. Benthic diatoms in USA rivers: distributions along spatial and environmental gradients. J. Biogeogr. 29 (2), 167–187. Potapova, M. & Charles, D. F. Distribution of benthic diatoms in US rivers in relation to conductivity and ionic composition. Freshwater Biol. 48 (8), 1311–1328. Quilbé, R., Rousseau, A. N., Duchemin, M., Poulin, A., Gangbazo, G. & Villeneuve, J.-P. Selecting a calculation method to estimate sediment and nutrient loads in streams: application to the Beaurivage River (Québec, Canada). J. Hydrol. 326 (1–4), 295–310. Soininen, J. Responses of epilithic diatom communities to environmental gradients in some Finnish rivers. Int. Rev. Hydrobiol. 87 (1), 11–24. Soininen, J., Paavola, R. & Muotka, T. Benthic diatom communities in boreal streams: community structure in relation to environmental and spatial gradients. Ecography 27 (3), 330–342. Song, J., Xu, Z., Hui, Y., Li, H. & Li, Q. Instream flow requirements for sediment transport in the lower Weihe River. Hydrol. Processes 24 (24), 3547–3557. Song, J., Yang, X., Zhang, J., Long, Y., Zhang, Y. & Zhang, T. Assessing the variability of heavy metal concentrations in liquid-solid two-phase and related environmental risks in the Weihe river of shaanxi province, China. Int. J. Environ. Res. Public Health 12 (7), 8243–8262. Spellerberg, I. F. & Fedor, P. J. A tribute to Claude Shannon (1916–2001) and a plea for more rigorous use of species richness, species diversity and the ‘Shannon–Wiener’Index. Global Ecol. Biogeogr. 12 (3), 177–179. Stevenson, R. J., Bothwell, M. L., Lowe, R. L. & Thorp, J. H. Algal Ecology: Freshwater Benthic Ecosystem. Academic Press, USA.
1047
Y. Liu et al.
|
Periphytic algal communities and responses to environmental variables
Stevenson, R. J., Pan, Y., Manoylov, K. M., Parker, C. A., Larsen, D. P. & Herlihy, A. T. Development of diatom indicators of ecological conditions for streams of the western US. J. N. Am. Benthol. Soc. 27 (4), 1000–1016. Su, P., Wang, X., Lin, Q., Peng, J., Song, J., Fu, J., Wang, S., Cheng, D., Bai, H. & Li, Q. Variability in macroinvertebrate community structure and its response to ecological factors of the Weihe River Basin, China. Ecol. Eng. 140, 105595. Tan, X., Ma, P., Xia, X. & Zhang, Q. Spatial pattern of benthic diatoms and water quality assessment using diatom indices in a subtropical river, China. Clean Soil Air Water 42 (1), 20–28. Teittinen, A., Kallajoki, L., Meier, S., Stigzelius, T. & Soininen, J. The roles of elevation and local environmental factors as drivers of diatom diversity in subarctic streams. Freshwater Biol. 61 (9), 1509–1521. Ter Braak, C. J. & Šmilauer, P. Canoco Reference Manual and User’s Guide: Software for Ordination, Version 5.0. Microcomputer Power Press, Ithaca, USA. Ter Braak, C. J. & Verdonschot, P. F. Canonical correspondence analysis and related multivariate methods in aquatic ecology. Aquat. Sci. 57, 255–289. Tison, J., Park, Y.-S., Coste, M., Wasson, J., Ector, L., Rimet, F. & Delmas, F. Typology of diatom communities and the influence of hydro-ecoregions: a study on the French hydrosystem scale. Water Res. 39 (14), 3177–3188. Tolkkinen, M., Mykrä, H., Virtanen, R., Tolkkinen, M., Kauppila, T., Paasivirta, L. & Muotka, T. Land use impacts on stream community composition and concordance along a natural stress gradient. Ecol. Indic. 62, 14–21. Tornés, E., Cambra, J., Gomà, J., Leira, M., Ortiz, R. & Sabater, S. Indicator taxa of benthic diatom communities: a case study in Mediterranean streams. Int. J. Limnol. 43 (1), 1–11. Townsend, S. A. & Gell, P. A. The role of substrate type on benthic diatom assemblages in the Daly and Roper Rivers of the Australian wet/dry tropics. Hydrobiologia 548, 101–115. Urrea, G. & Sabater, S. Epilithic diatom assemblages and their relationship to environmental characteristics in an agricultural watershed (Guadiana River, SW Spain). Ecol. Indic. 9 (4), 693–703. Van Buren, M., Watt, W. E., Marsalek, J. & Anderson, B. Thermal enhancement of stormwater runoff by paved surfaces. Water Res. 34 (4), 1359–1371. Van Dam, H., Mertens, A. & Sinkeldam, J. A coded checklist and ecological indicator values of freshwater diatoms from the Netherlands. Netherland J. Aquat. Ecol. 28 (1), 117–133.
Hydrology Research
|
51.5
|
2020
Vasiljević, B., Simić, S. B., Paunović, M., Zuliani, T., Krizmanić, J., Marković, V. & Tomović, J. Contribution to the improvement of diatom-based assessments of the ecological status of large rivers – the Sava River case study. Sci. Total Environ. 605, 874–883. Wang, J., Meier, S., Soininen, J., Casamayor, E., Pan, F., Tang, X., Yang, X., Zhang, Y., Wu, Q., Zhou, J. & Shen, J. Regional and global elevational patterns of microbial species richness and evenness. Ecography 40 (3), 393–402. Wang, H., Li, Y., Li, J., An, R., Zhang, L. & Chen, M. a Influences of hydrodynamic conditions on the biomass of benthic diatoms in a natural stream. Ecol. Indic. 92, 51–60. Wang, W., Song, J., Zhang, G., Liu, Q., Guo, W., Tang, B., Cheng, D. & Zhang, Y. b The influence of hyporheic upwelling fluxes on inorganic nitrogen concentrations in the pore water of the Weihe River. Ecol. Eng. 112, 105–115. Wei, F. Water and Waste Water Monitoring and Analysis Methods, 4th edn. Water and Waste Water Monitoring and Analysis Method Committee, China Environmental Science Press, Beijing (in Chinese). Wei, S., Yang, H., Song, J., Abbaspour, K. C. & Xu, Z. System dynamics simulation model for assessing socio-economic impacts of different levels of environmental flow allocation in the Weihe River Basin, China. Eur. J. Oper. Res. 221 (1), 248–262. Westlake Temporal Changes in Aquatic Macrophytes and Their Environment. Dynamique de Populations et Qualite de l’Eau, pp. 109–138. White, M. D. & Greer, K. A. The effects of watershed urbanization on the stream hydrology and riparian vegetation of Los Penasquitos Creek, California. Landsc. Urban Plan. 74 (2), 125–138. Winter, J. G. & Duthie, H. C. Stream epilithic, epipelic and epiphytic diatoms: habitat fidelity and use in biomonitoring. Aquat. Ecol. 34 (4), 345–353. Wu, W., Xu, Z., Yin, X. & Zuo, D. Assessment of ecosystem health based on fish assemblages in the Wei River basin, China. Environ. Monit. Assess. 186 (6), 3701–3716. Yang, H. & Flower, R. J. Effects of light and substrate on the benthic diatoms in an oligotrophic lake: a comparison between natural and artificial substrates. J. Phycol. 48 (5), 1166–1177. Zhu, H. Z. & Chen, J. Y. Bacillariophyta of the Xizang Plateau. Science Press, Beijing, China.
First received 15 March 2020; accepted in revised form 12 June 2020. Available online 14 August 2020
1048
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Spatiotemporal variation and tendency analysis on rainfall erosivity in the Loess Plateau of China Yongsheng Cui, Chengzhong Pan, Chunlei Liu, Mingjie Luo and Yahui Guo
ABSTRACT Rainfall erosivity is an important factor to be considered when predicting soil erosion. Precipitation data for 1971–2010 from 39 stations located in the Loess Plateau of China were collected to calculate the spatiotemporal variability of rainfall erosivity, and the long-term tendency of the erosivity was predicted using data from the HadGEM2-ES model. Statistical analyses were done using Mann– Kendall statistic tests and ordinary Kriging interpolation. The results showed that the annual mean rainfall erosivity in the Loess Plateau decreased from 1,286.02 MJ mm hm 2 h 1 a 1 in 1971–1990 to 1,201.46 MJ mm hm 2 h 1 a 1 in 1991–2010 and mainly occurred in July to August. The rainfall erosivity decreased from the southeast to the northwest of the Loess Plateau and was closely related to the annual precipitation amount. However, the effect of annual precipitation on rainfall erosivity weakened under climate change: the annual precipitation increased and the rainfall erosivity decreased. Climate change, however, had little influence on the spatial variation in rainfall erosivity in the Loess Plateau. The results obtained can facilitate the prediction of spatial and temporal variations in soil erosion in the Loess Plateau. Key words
| Loess Plateau of China, rainfall erosivity, Representative Concentration Pathway (RCP) scenarios, spatiotemporal variation, tendency prediction
HIGHLIGHTS
• • •
Both precipitation and rainfall erosivity showed an insignificant decreasing trend for 1971–2010. The climate model predicts an increasing precipitation but decreasing rainfall erosivity. South and southeast of the Loess Plateau are areas susceptible to rainfall erosion under climate change.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.030
Yongsheng Cui Chengzhong Pan (corresponding author) Chunlei Liu Mingjie Luo Yahui Guo Key Laboratory of Water Sediment Sciences, College of Water Sciences, Beijing Normal University, Beijing 100875, China E-mail: pancz@bnu.edu.cn
1049
Y. Cui et al.
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
GRAPHICAL ABSTRACT
INTRODUCTION Soil erosion is one of the most important ecological problems
of rainfall erosivity in Brazil. Some other researchers have con-
in the world. Nearly a third of arable land has been exposed to
ducted spatiotemporal analyses of rainfall erosivity in different
7
2
soil erosion, which has increased at a rate of 1.0 × 10 hm per
regions (Angulo-Martínez & Beguería ; Panagos et al.
year (Pimentel et al. ). Soil erosion removes fertile soil and
; Xie et al. ). A high-resolution distribution map of rain-
seriously threatens agricultural production and food security.
fall erosivity has been established using global data, and the
The eroded soil usually enters rivers, causing siltation and
global rainfall erosivity was estimated to be about 2,190.0
water eutrophication, and thus endangers water safety and
MJ mm hm 2 h 1 (Panagos et al. ). Due to the difficulty
affects normal production and life. Predicting and protecting
in obtaining rainfall data with a high temporal resolution,
against soil erosion are of great significance to the construction
Vrieling et al. () used satellite data and combined rainfall
of an ecological civilization and to the sustainable develop-
intensity and rainfall erosivity to establish a method for fore-
ment
Consequently,
casting rainfall erosivity, thereby demonstrating a potential
mechanisms of soil erosion and measures to control it have
of
the
economy
and
society.
tool for soil erosion prediction in data-poor areas. Previous
been hot research topics (Lal ; Pimentel ).
studies have enhanced our understanding of spatiotemporal
Rainfall is the driving force and prerequisite for soil ero-
variations in rainfall erosivity, but these studies were con-
sion (Zhang et al. ). The Universal Soil Loss Equation
strained by the limited availability of daily rainfall data and
(USLE) devised by Wischmeier & Smith () and the
by the length of rainfall data. Studying available long-term rain-
Revised Universal Soil Loss Equation (RUSLE) proposed by
fall data and estimating rainfall erosivity could provide
Renard & Freimund () and Renard et al. () for the
information on the potential trends of soil erosion, especially
United States are widely used to predict soil erosion (Anees
in areas that are lacking data.
et al. ). The rainfall erosivity (R-factor) in these models indi-
In the Loess Plateau of China, intensive rainfall and
cates the ability of rainfall to cause soil erosion and is
sparse vegetation can lead to relatively serious soil erosion.
considered the best factor for studying the response of soil ero-
The soil erosion modulus can reach 1.0 × 107 kg km 2 a 1.
sion to rainfall changes (Nearing et al. ). Meusburger et al.
A large amount of sediment enters the Yellow River and
() collected rainfall data for 1989–2010 and used the pro-
causes many problems, including river channel siltation and
duct of the total kinetic energy of rainfall (E) and the 30-min
the deterioration of aquatic ecosystems (Liu & Liu ).
maximum rainfall intensity (I30) as the measurement index of
Many researchers have analyzed the spatial and temporal dis-
rainfall erosivity to study the temporal and spatial variation
tribution of rainfall erosion on the Loess Plateau in different
of rainfall erosivity in Switzerland. Da Silva () used rain-
periods (Xin et al. ; Abd Elbasit et al. ; Yang & Lu
fall data in conjunction with geographic information system
). Fu et al. () calculated all the RUSLE parameters
(GIS) for spatial interpolation to map the spatial distribution
using GIS based on regional data from the Yanhe watershed
1050
Y. Cui et al.
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
and concluded that the annual soil loss was 0.5–2.0 ×
6.35 × 105 km2. Most areas of the plateau have semi-humid
107 kg km 2. Sun et al. () analyzed the effects of topogra-
or semi-arid climates, and the climates in different areas are
phy and land-use patterns on soil erosion in the Loess Plateau
quite different. The annual mean temperature is 3.6–14.3 C,
and suggested that ‘Grain-to-Green Program’ is effective for
and the annual evaporation is 1,400–2,000 mm. The main
preventing soil erosion. Xu () carried out a quantitative
crops grown are wheat, corn, soybeans, and sorghum. The
analysis of the correlations between vegetation coverage
study area is characterized by complex landforms and deep
(Cf ), annual rainfall erosivity (Re), and annual precipitation
soil layers (mainly loose and easily eroded dark loessial soil
(Pm). Xu’s results indicated that Re will increase rapidly
and loessal soil). The annual precipitation ranges from 200
when Pm > 300 mm, and that when Pm > 530 mm, the rate
to 750 mm, with large inter-annual and seasonal variations.
at which Re increases with Pm becomes higher. Current
There are frequent storms, which are the main driving force
research generally focuses on specific watersheds where
of soil erosion. The annual mean soil erosion modulus of
daily rainfall data are easily available. Consequently, there
the Loess Plateau is 0.5–1.0 × 107 kg km 2 a 1, and the thick-
are relatively few papers addressing the long-term spatiotem-
ness of the soil layer lost every year is about 1 cm.
poral variation of rainfall erosivity in the Loess Plateau. More analysis is needed for the region as a whole based on available
Data
long-term daily rainfall data. The results of the analysis could form the basis for soil loss prediction in the region. Moreover,
We used data from the 39 representative meteorological
the impact of climate change on soil erosion needs to be
stations located in the Loess Plateau. The daily rainfall
understood (Almagro et al. ). Few studies, however,
data from 1971 to 2010 were selected for investigating the
have been carried out that have made predictions and ana-
spatiotemporal variability of precipitation. The location of
lyses of variations in rainfall erosivity under projected
each meteorological station is shown in Figure 1: SRTM
climate change in the Loess Plateau of China. Consequently,
(Shuttle Radar Topography Mission) data from http://srtm.
a climate change perspective and decision basis cannot be
csi.cgiar.org/srtmdata/, with spatial resolution approxi-
provided for decision makers.
mately 30 meters on the line of the equator. Daily rainfall
The overall aim of this study was to address these knowl-
data were obtained from the China Meteorological Data
edge gaps. Precipitation data from 1971 to 2010 were
Service Center (http://data.cma.cn/). The Linfen meteorolo-
collected from 39 representative stations in the Loess Plateau
gical station does not have data for 2000, and the Yongji
of China to analyze the spatiotemporal variability of rainfall
station does not have data for 1981–1990.
erosivity. The main objectives of this study are:
In addition to the data from the above meteorological
1. to determine the spatial and temporal variability of
stations, data were also obtained using a GCMs (Hadley Center Global Environment Model version 2 (HadGEM2-
annual rainfall erosivity in the Loess Plateau; 2. to predict the change tendency of future rainfall erosivity using a typical global circulation model (GCM).
ES)) forced by two Representative Concentration Pathway (RCP) scenarios – RCP4.5 and RCP8.5 (Chou et al. ; Yan et al. ). The HadGEM2-ES is an earth system category GCM that was developed by the Hadley Center, and the resolution is about 1.875 for longitude and 1.25 for latitude (Collins et al. ). The RCP scenarios were developed
MATERIALS AND METHODS
by the research community by considering emissions, conStudy area
centrations, and land-use trajectories and are labeled according to the expected values of global radiative forcing
0
0
The Loess Plateau is located between 100 52 –114 31 E and
in 2100. RCP4.5 is considered an intermediate scenario
33 370 –41 250 N, including most or part of Shanxi, Shaanxi,
that assumes emissions reduction during the 21st century
Gansu, Qinghai, and Henan Provinces and the Inner
by employing clean technologies and stringent climate pol-
Mongolia and Ningxia Regions. The total area is about
icies. This scenario predicts a global forcing radiation of
Y. Cui et al.
1051
Figure 1
|
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
Location map of meteorological stations and sample spots on the Loess Plateau.
∼4.5W m2 and concentrations of ∼650 ppm CO2-eq at
months. In this study, half-monthly rainfall erosivity esti-
stabilization after 2100. In contrast, RCP8.5 is a kind of
mates were obtained using the model proposed by Zhang
business-as-usual scenario characterized by no implemen-
et al. (). The model has been shown to be suitable for
tation of climate policies, low rates of technology
the prediction of rainfall erosivity for the Loess Plateau
development, high energy intensities, and a reliance on
(Wu et al. ). Rainfall erosivity was calculated as follows:
fossil fuels that leads to higher emissions than RCP4.5 over time. Global forcing radiation and atmospheric concentrations of CO2-eq are projected to reach >8.5 W m
2
and
>1,370 ppm in 2100, respectively (Almagro et al. ). Daily precipitations for 2020–2100 are available for the RCP scenarios. The boundaries of the Loess Plateau were taken to 33.75 –41.25 N and 101.25 –114.375 E. Daily rain-
Ri ¼ α
k X
(Pj )β
(1)
j¼1
where Ri is the rainfall erosivity index in the ith half-month period (MJ·mm·hm 2·h 1); k is the number of days in the half-month period; Pj is the erosive rainfall (mm) on the
fall data for 48 sample spots were obtained (Figure 1) from
jth day in the half-month period. Daily precipitation is
https://esgf-data.dkrz.de/search/cmip5-dkrz/.
required to be greater than or equal to 12 mm, otherwise calculated as 0, and the threshold of 12 mm is consistent with the Chinese standard for erosive rainfall. The parameters α, β need to be determined for the model and are calculated
METHODS
from the following formulae:
Calculation of the rainfall erosivity values
α ¼ 21:586β 7:1891
The USLE and the RUSLE models use an algorithm and
β ¼ 0:8363 þ
daily precipitation to determine rainfall erosivity in half-
18:177 24:455 þ Pd12 Py12
(2) (3)
Y. Cui et al.
1052
|
Spatiotemporal variation of rainfall erosivity
where Pd12 is the average daily precipitation (mm) with a
Hydrology Research
|
51.5
|
2020
RESULTS
daily rainfall of 12 mm or more; Py12 represents the annual mean precipitation (mm) with a daily rainfall of 12 mm or
Annual variation in precipitation and erosive rainfall
more. During the calculation, the half-month period is defined as the 15th day of each month: the 1st to 15th day
Data from the Linfen and Yongji meteorological stations
of each month is one half month, and the rest of the
were excluded from the analysis because data from some
month is calculated as another half month. The whole
years were missing. Data of the remaining 37 meteorological
year is divided into 24 half-month periods, and the monthly,
stations were averaged to obtain the precipitation variation
annual rainfall erosivity, and annual mean rainfall erosivity
for the Loess Plateau from 1971 to 1990 and from 1991 to
are obtained.
2010 (Figure 2(a) and 2(b)). The mean precipitation from
In the present study, the average rainfall erosivity refers
1971 to 1990 and from 1991 to 2010 was 425.26 and
to the average of data from different meteorological stations
399.87 mm, respectively. The annual precipitation ranges
over a given time, while the annual mean rainfall erosivity
were 205.39 mm (1971–1990) and 251.99 mm (1991–
refers to the average of data over a period of time for a
2010), and the coefficients of variation were 0.15 and 0.14,
specific location.
respectively. In general, years with a mean precipitation of >500 mm in the Loess Plateau from 1971 to 2010 accounted for only 7.5% of the years, and years with less than 400 mm accounted for 45.0% (1971–1990 accounted for 15.0%). The
DATA PROCESSING AND ANALYSIS METHODS
annual precipitation showed an overall decreasing trend from 1971 to 2010.
MATLAB R2018b (The MathWorks, Inc., USA) and
The mean erosive rainfall on the Loess Plateau from
ORIGIN 2020 (OriginLab Corporation, USA) were used
1971 to 1990 and from 1991 to 2010 was 242.59 and
for statistical analysis and mapping of data. The ordinary
229.49 mm, respectively, which accounted for about 57.0%
Kriging interpolation method and the geostatistical modules
of the mean precipitation, and showed a decreasing trend
of ArcGIS 10.2 (Esri, Inc., USA) were used for the spatial
from 1971 to 2010. There is a strong linear regression
interpolation of precipitation and rainfall erosivity. In
relationship between mean erosive rainfall and precipitation
addition, the non-parametric Mann–Kendall trend analysis
(R 2 ¼ 0.93, P < 0.05). The results of the Mann–Kendall test
method was used to test the significance of changes in
showed that the erosive rainfall trend from 1971 to 2010
meteorological elements (Hamed & Ramachandra Rao
has increased from 1973 to 1981 and decreased from 1982
; Ahmad et al. ).
to 2010 (except for 1985 and 1996). Both trends are
Figure 2
|
Annual variation of erosive rainfall and precipitation on the Loess Plateau for (a) 1971–1990 and (b) 1991–2010.
Y. Cui et al.
1053
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
insignificant and changeable, which indicates that erosive
erosivity and precipitation was not strong. Taking 1989
rainfall is not only affected by the precipitation amount
and 1990 as an example, the rainfall erosivity was basically
but also reflects the annual changes in precipitation.
the same, but the difference in precipitation was 17.5%. This shows that rainfall erosivity is also affected by rainfall distribution and other factors. The Mann–Kendall tests showed
Temporal variation of rainfall erosivity
that, in general, the inter-annual changes in rainfall erosivity
The intra-annual distribution of rainfall erosivity for the years from 1971 to 2010 is shown in Figure 3(a) and 3(b). The rainfall erosivity during the year formed a single-peak distribution, first increasing and then decreasing. The annual precipitation distribution and the rainfall erosivity
increased from 1975 to 1981 and decreased from 1982 to 2010. This is similar to the trends for erosive rainfall, but, as shown in Figure 4(a) and 4(b), the annual decrease rate in rainfall erosivity during 1991–2010 was slightly slower than that during 1971–1990, and eventually stabilized.
followed a similar pattern: in January and February, the rainfall erosivity was almost zero, it then gradually increased,
Spatial distribution of rainfall erosivity
reaching a peak in July and August, and then gradually decreased to zero in November. The total rainfall erosivity
As shown in Figure 5, the annual mean rainfall erosivity
in July to August accounted for 61.5% (1971–1990) and
decreased from southeast to northwest. For the area as a
59.4% (1991–2010) of the year’s rainfall erosivity, indicating
whole, the annual mean rainfall erosivity was 1,286.02 and
that summer rainfall, especially in July and August, was the
1,201.46 MJ mm hm 2 h 1 a 1 in 1971–1990 and 1991–
main cause of soil erosion. In 1991–2010, the precipitation
2010, respectively. The maximum values obtained (for the
decreased slightly from that of 1971 to 1990, but the maxi-
Wutai Mountain and Anze station in the province of
mum still occurred in July and August, and was about
Shanxi) were 2,433.14 MJ mm hm 2 h 1 a 1 (1971–1990)
2
for both series.
and 2,144.56 MJ mm hm 2 h 1 a 1 (1991–2010). The mini-
The average rainfall erosivity for 1971–1990 and 1991–
mum values obtained for the two periods (for the Jingyuan
200.0 MJ mm hm
1
h
2010 was 1,278.68 and 1,200.60 MJ mm hm 2 h 1, respect-
station,
Gansu 2
1
province)
were
277.57
and
298.94
1
ively, and the coefficients of variation were 0.24 and 0.19,
MJ mm hm
respectively. Rainfall erosivity was essentially stable at
fall erosivity from 1991 to 2010 was slightly lower than
around
1,200 MJ mm hm 2 h 1
h
a , respectively. The annual mean rain-
2010
that from 1971 to 1990 and showed a very high decrease
(Figure 4(b)). Overall, rainfall erosivity fluctuated with the
in the northern part of Shanxi Province. The areas with
increases and decreases in annual precipitation, but there
the greatest rainfall erosivity were, however, concentrated
were some years where the correlation between rainfall
in the southeast of the Loess Plateau. Consequently, most
Figure 3
|
from
2004
to
Relation between half-monthly rainfall erosivity and monthly precipitation in the Loess Plateau for (a) 1971–1990 and (b) 1991–2010.
Y. Cui et al.
1054
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
Figure 4
|
Annual variation in rainfall erosivity for (a) 1971–1990 and (b) 1991–2010.
Figure 5
|
Spatial distribution of rainfall erosivity on the Loess Plateau for (a) 1971–1990 and (b) 1991–2010. The spatial variation of annual mean rainfall erosivity from 2020 to 2100 under RCPs is shown in Supplementary material, Appendix A.
of Shanxi Province, central and southern Shaanxi were sus-
Provinces, and below 300.0 mm a 1 for the Ningxia and
ceptible to rainfall erosion from 1971 to 2010. In these
Inner Mongolia Regions, while the annual mean precipi-
areas, soil conservation measures should be strengthened
tation value distribution from 1971 to 1990 was relatively
to reduce soil erosion.
discrete in Gansu Province, which was about 170.60– 598.97 mm a 1. The annual mean precipitation from 1991
Tendency analysis of rainfall erosivity
to 2010 was 143.45–585.32 mm a 1, which was 5.5% lower than that from 1971 to 1990. The province with the
Precipitation data for representative areas of the Loess Pla-
highest mean precipitation changed from Shanxi Province
teau are shown in Table 1. (The data were obtained using
in 1971–1990 to Shaanxi Province in 1991–2010, while
ordinary Kriging interpolation. Because there are very few
Ningxia still had the lowest. The standard deviation (STD)
representative meteorological stations that could be selected
of the average precipitation in Shanxi Province decreased
in Henan and Qinghai Provinces, and the areas of these Pro-
by 32.8% from 1971–1990 to 1991–2010. The reduction in
vinces are relatively small, these Provinces are not included
the other regions was, however, much smaller.
in Tables 1 and 2). The annual mean precipitation from 1971 1
Under the RCP4.5 scenario, the annual mean precipi505.12 ±
to 1990 was 398.50 ± 114.80 mm a . The annual mean pre-
tation
cipitation was about 500.0 mm a 1 for Shanxi and Shaanxi
130.02 mm a 1, which is 26.8 and 34.1% higher than the
from
2020
to
2100
would
be
Y. Cui et al.
1055
Table 1
|
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
1
Annual precipitation statistics for different provinces in the Loess Plateau (mm a
|
51.5
|
2020
)
Inner Mongolia
Gansu
Ningxia
Shanxi
Shaanxi
Observed (1971–1990)
Max. Min. Mean STD
501.44 137.53 297.82 92.78
598.97 170.60 436.66 129.79
493.70 170.24 256.47 90.63
546.51 443.33 504.33 94.69
608.61 279.45 497.25 91.18
Observed (1991–2010)
Max. Min. Mean STD
430.08 143.45 296.46 86.95
549.67 186.27 402.31 113.46
459.83 171.36 249.50 85.19
532.82 403.91 463.91 63.67
585.32 298.40 471.32 82.82
RCP4.5 (2020–2100)
Max. Min. Mean STD
518.10 206.40 341.16 67.42
1,050.60 245.03 546.09 105.58
585.14 296.13 420.12 –
826.60 229.48 680.54 89.13
1,048.18 400.28 537.73 108.90
RCP8.5 (2020–2100)
Max. Min. Mean STD
519.07 219.93 350.56 65.47
1,012.28 263.41 542.25 96.70
585.44 316.37 428.91 –
861.09 238.75 699.95 101.37
1,012.40 411.85 538.33 105.46
Table 2
|
2
Statistics on rainfall erosivity for different provinces in the Loess Plateau (MJ mm hm
h
1
a
1
)
Inner Mongolia
Gansu
Ningxia
Shanxi
Shaanxi
Observed (1971–1990)
Max. Min. Mean STD
1,584.97 274.67 901.20 364.44
1,807.94 251.49 1,066.96 515.89
1,290.51 316.21 491.31 208.05
2,405.43 831.70 1,681.49 469.92
2,054.79 524.64 1,557.78 342.67
Observed (1991–2010)
Max. Min. Mean STD
1,311.57 300.76 830.76 335.34
1,678.56 281.53 984.16 455.61
1,318.07 305.11 573.06 257.84
2,090.51 877.03 1,512.01 360.49
2,026.55 787.54 1,511.63 316.75
RCP4.5 (2020–2100)
Max. Min. Mean STD
967.26 161.32 447.29 164.89
1,513.95 194.80 560.28 77.91
776.07 262.86 459.26 –
2,573.76 263.82 1,466.75 309.46
2,104.44 515.64 914.64 222.85
RCP8.5 (2020–2100)
Max. Min. Mean STD
1,031.66 135.53 476.01 143.54
1,515.07 235.42 629.24 102.97
920.24 349.35 542.92 –
3,494.06 236.67 1,705.36 429.95
2,184.66 571.43 988.91 279.86
means for 1971–1990 and 1991–2010, respectively. The
219.93–1,012.40 mm a 1, which was an increase of about
maximum
reach
35.9% compared with 1991–2010. The regional variations
1,050.60 mm a 1. The spatial distribution of precipitation
precipitation
in
RCP4.5
would
in annual mean precipitation were essentially the same as
similar to the spatial distribution for 1971–1990: Shanxi
those in the RCP4.5 scenario, with a maximum of
Province had the largest annual mean precipitation of
1,000.0 mm a 1 (Gansu and Shaanxi) and a minimum of
680.54 mm a 1, while the provinces with the lowest precipi-
only 219.93 mm a 1 (Inner Mongolia).
tation changed from Ningxia (1971–2010) to Inner 1
From historical data (Table 2), the annual mean rainfall
Mongolia, with a precipitation of only 341.16 mm a . The
erosivity in the Loess Plateau from 1971 to 1990 was calcu-
range of precipitation under the RCP8.5 scenario was
lated as 251.49–2,405.43 MJ mm hm 2 h 1 a 1, with an
1056
Y. Cui et al.
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
average of 1,137.75 ± 487.57 MJ mm hm 2 h 1 a 1. Shanxi
from 1971–1980 to 1981–1990 was 604.89 to 394.74
Province had the highest rainfall erosivity, which reached
MJ mm hm 2 h 1 a 1. In the southern Gansu Province, the
2
1
2,405.43 MJ mm hm
1
from 1971 to 1990, followed
central part of Shaanxi Province, and the northern Shanxi
by Shaanxi Province with 1,557.78 MJ mm hm 2 h 1 a 1.
h
a
Province, rainfall erosivity from 1981 to 1990 was less
The rainfall erosivity in Ningxia was low, and the STD was
than that in 1971–1980, while the rainfall erosivity in Ning-
also low. The STD for Gansu was the largest; this indicates a
xia and some parts of Gansu Province and the southern part
large spatial difference, which is similar to the results obtained
of Shanxi Province increased the most. The spatial variation
from the analysis of data from meteorological stations for
of rainfall erosivity from 1991 to 2010 was not the same as
1991–2010. The annual mean rainfall erosivity from 1991 to
that from 1971 to 1990. After the 2000s, rainfall erosivity
2
2010 was 1,082.32 ± 418.69 MJ mm hm
1
1
a . Compared
increased significantly in southern Shaanxi Province and
with 1971–1990, except for Ningxia, which showed an erosiv-
the central part of Shanxi Province, while the other areas
h
ity increase of 16.6% for 1991–2010, other provinces and
showed an overall decreasing trend. For example, rainfall
regions all showed decreases. For example, erosivity in
erosivity in Qinghai Province and Inner Mongolia decreased
Shanxi and Shaanxi decreased by 10.1 and 3.0%, respectively,
by about 429.88 MJ mm hm 2 h 1 a 1.
which represents small spatial differences.
Under the RCP4.5 scenario, the variation of rainfall ero-
The calculated rainfall erosivity in future climate scen-
sivity from 2020–2060 to 2061–2100 was 66.89 to
arios was lower than that obtained from the historical
595.59 MJ mm hm 2 h 1 a 1. The areas with increased rain-
data. Under the RCP4.5 and RCP8.5 scenarios, the rainfall
fall erosivity are concentrated in the southern Loess Plateau,
erosivity in 2020–2100 would be 769.64 ± 433.32 and
particularly in southern Shanxi and in the central and
868.49 ± 507.97 MJ mm hm 2 h 1 a 1, which is a 28.9 and
southern part of Shaanxi Province, while the rainfall erosiv-
19.8% decrease, respectively, compared with 1991–2010.
ity in the other regions in 2061–2100 is slightly less than that
The areas with the highest rainfall erosivity are still in
in 2020–2060. In contrast, under the RCP8.5 scenario, the
Shanxi Province in the RCP scenarios, while the lowest
rainfall erosivity increased greatly and was concentrated in
areas change from being in Ningxia (according to the his-
the southeast of the Loess Plateau (southern Shanxi and
torical data) to Inner Mongolia. These results are similar
northern Henan Provinces). In the other regions, rainfall
to the results for the distribution of precipitation. The mini-
erosivity in RCP8.5 also increased, and the overall increase
mum values for erosivity from 2020 to 2100 would be
was around 99.19–2,009.03 MJ mm hm 2 h 1 a 1. This indi-
2
161.32 MJ mm hm
1
1
135.53
cates that the rainfall erosivity increased during the second
MJ mm hm 2 h 1 a 1 (RCP8.5). In the RCP8.5 scenario,
half of the 21st century under the RCP8.5 scenario, and
the rainfall erosivity ranges from 135.53 to 3,494.06
hence, the degree of soil erosion in the southeast of the
MJ mm hm 2 h 1 a 1 and is 12.8% higher than the RCP4.5
Loess Plateau would also increase.
h
a
(RCP4.5)
and
values. The value distribution under RCP8.5 is, however, more discrete than under RCP4.5. The maximum erosivity in Shanxi Province increases by 35.8% from RCP4.5 to
DISCUSSIONS
RCP8.5, while the increases in the other provinces and regions are small. This arises because heavy rainfall
Temporal changes in rainfall erosivity
becomes more concentrated in Shanxi Province in the RCP8.5 scenario.
The mean precipitation on the Loess Plateau was
The rainfall erosivity in different periods was divided
412.57 mm from 1971 to 2010 and showed an insignificant
into two parts according to the length of the series, after
decreasing trend, with an annual decrease of about
the annual mean rainfall erosivity was interpolated by
0.88 mm. These results are similar to those in the study by
ordinary Kriging in ArcGIS in every part, and the rainfall
Wu et al. (). The mean precipitation in 2020–2100, how-
erosivity variation figure (Figure 6) was obtained using
ever, increased by 30.3% (RCP4.5) and 32.1% (RCP8.5)
raster calculation. The variation in the rainfall erosivity
compared with that in 1971–2010. The spatial distribution
Y. Cui et al.
1057
Figure 6
|
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
Variation of annual mean rainfall erosivity for (a) 1981–1990 minus 1971–1980, (b) 2001–2010 minus 1991–2000, and (c and d) 2061–2100 minus 2020–2060.
of future precipitation indicates that precipitation will
mean rainfall erosivity from 1991 to 2010 decreased by
increase in the southeast of the study area and decrease in
5.0%. Similarly, the rainfall erosivity from 2020 to 2100 was
the northwest. Consequently, Shanxi Province may have
28.9% (RCP4.5) and 19.8% (RCP8.5) less than that in 1991–
an increased risk of soil erosion.
2010. Almagro et al. () also obtained similar results for
Precipitation distribution causes annual rainfall erosiv-
Brazil. The similarity in results may have been caused by
ity to appear as a unimodal distribution (Figure 3). The
the rainfall in the future climate scenarios being relatively dis-
maximum half-month rainfall erosivity between 1971–1990
persed, with few short-duration heavy rainfall events. An
2
and 1991–2010 was about 200.0 MJ mm hm
1
and
alternative explanation is that the HadGEM2-ES model out-
occurs mainly in July to August, which indicates that
puts are based on meteorological factors and ignore the
h
heavy rainfall during the summer is the main cause of soil
effect of topography on rainfall intensity (Djebou et al.
erosion in the Loess Plateau. The comprehensive utilization
). Consequently, the model outputs may lead to underes-
of
managements),
timations of the erosivity of future rainfall. For further
vegetation measures (e.g. afforestation), and agricultural
research, comprehensive analysis should be combined with
measures (e.g. no-tillage) are the key elements of ecological
factors such as topography and land-use data. However,
environmental development that are required to prevent soil
despite the aforementioned limitations, the results of the pre-
erosion in the Loess Plateau.
sent study can still be used as a reference for the study of soil
engineering
measures
(e.g.
slope
In terms of inter-annual changes, the annual mean rain-
erosion under climate change in the future.
1,139.75 ±
To improve the analysis of the relationship between
487.57 MJ mm hm 2 h 1 a 1, with an insignificant decreas-
rainfall erosivity and precipitation over various time scales,
ing rate of 1.83 MJ mm hm 2 h 1 a 1 (Figure 4). When
the present study made use of the Standardized Precipi-
compared with the erosivity from 1971 to 1990, the annual
tation Index (SPI) and used the SPI PROGRAM to
fall
erosivity
from
1971
to
1990
was
Y. Cui et al.
1058
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
calculate the SPI value (obtained from https://drought.unl.
April and reached a maximum in May, while in normal and
edu/droughtmonitoring/SPI/SPIProgram.aspx).
time
humid years, the maxima were often reached in August and
scale of the historical data was 36 and 60 months, and 60
September, and then gradually decreased and reached zero
and 96 months for the estimated data. The results are
in November. The rainfall erosivity from July to September
shown in Supplementary material, Appendix B. Based on
in drought years accounted for about 50.0% for the whole
the SPI value and the classification criteria of McKee et al.
year, while in normal and humid years, this period accounted
(), cases were selected for a moderate drought year
for 75.7–90.0% of the year’s total.
The
( 1.49 SPI 1.00), a normal year ( 0.99 SPI 0.99),
A discussion of the data obtained from the RCP scen-
and a moderate humid year (1.00 SPI 1.49). The selected
arios follows. When compared with historical data, it can
hydrological years and SPI values are shown in Table 3.
be seen that precipitation derived from the RCP scenarios
It can be seen from Figure 7 that the monthly rainfall
is more dispersed and is mainly concentrated in April to
erosivity has a good corresponding relationship with precipi-
September, with the peak generally occurring in July and
tation. However, the precipitation amount is not the
August (Figure 7(c) and 7(d)). Although the precipitation
determining condition for the variation in rainfall erosivity.
increased from April to June, most of the precipitation
For example, the precipitation in June 2005 decreased by
events were of long-duration and low-rainfall intensity,
11.1% compared with May 2005, but the rainfall erosivity
with daily precipitation less than 12 mm. This precipitation,
increased by 74.9%. The increase was mainly the result of
therefore, had little effect on the distribution of annual rain-
the greater rainfall intensity in summer. The threshold of
fall erosivity. The long-duration, low-intensity rainfall events
erosive rainfall adopted in this paper was a daily rainfall
were the results of climate change and other factors. Precipi-
of P 12 mm. Consequently, a lot of the precipitation was
tation from April to June accounted for about 30.0% of the
excluded and could not be used as erosive rainfall when cal-
annual precipitation on the whole, while rainfall erosivity
culating rainfall erosivity. The use of the erosive rainfall
during this period only accounted for about 25.0% of the
threshold also leads to deviations between rainfall erosivity
annual rainfall erosivity. Except for the heavy rainfall erosiv-
and precipitation (Xie et al. ).
ity that occurred in July in humid years, the monthly rainfall
The percentage of precipitation from April to June in
erosivity values for different hydrological years were essen-
moderate drought years was relatively large and accounted
tially below 500.0 MJ mm hm 2 h 1. In a similar fashion to
for 34.0% of the annual precipitation (1971–1990), while it
the results from the historical data, 83.3% of the months
only accounted for 20.0% of the annual precipitation in
had rainfall erosivity below 200.0 MJ mm hm 2 h 1 in the
normal and humid years. In moderate drought years, the pre-
RCP4.5 and RCP8.5 scenarios.
cipitation from July to August accounted for about 25.3% of the annual precipitation, while in normal and moderate
Spatial variation of rainfall erosivity and its response to
humid years, it reached more than 35.0% (1991–2010). The
precipitation
historical data show that the monthly distribution of rainfall erosivity during different hydrological years is essentially
In the 1980s, rainfall erosivity decreased in southern Gansu,
the same. In drought years, rainfall erosivity increased from
central Shaanxi, and northern Shanxi and increased slightly
Table 3
|
SPI values and hydrological years for the different data series
Data series
Moderate drought years
Normal years
Moderate humid years
1971–1990s
1983 (SPI36 ¼ 1.06)
1976 (SPI36 ¼ 0.50)
1978 (SPI36 ¼ 1.72)
1991–2010s
2002 (SPI36 ¼ 1.01)
2008 (SPI36 ¼ 0.08)
2005 (SPI36 ¼ 1.15)
2020–2100s (RCP4.5)
2045 (SPI60 ¼ 1.13)
2052 (SPI60 ¼ 0.38)
2057 (SPI60 ¼ 1.12)
2020–2100s (RCP8.5)
2048 (SPI60 ¼ 1.09)
2062 (SPI60 ¼ 0.44)
2071 (SPI60 ¼ 1.02)
Y. Cui et al.
1059
Figure 7
|
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
Monthly rainfall erosivity and precipitation variations for (a) 1971–1990, (b) 1991–2010, (c) 2020–2100 (RCP4.5), and (d) 2020–2100 (RCP8.5).
in Ningxia and some parts of Gansu and southern Shanxi.
and RCP8.5 scenarios, the rainfall erosivity in 2020–2100
After 2000, it increased in central and southern Shaanxi
decreased by 28.9 and 19.8%, respectively, compared with
and central Shanxi but decreased in Inner Mongolia. How-
that in 1991–2010, respectively, but the range of variations
ever, on the whole, the spatial distribution of rainfall
is larger (Table 2). The maximum annual mean rainfall ero-
erosivity in the Loess Plateau shows a decreasing trend
sivity reaches 3,494.06 MJ mm hm 2 h 1 a 1, while the
from the southeast to the northwest. Most parts of Shanxi
minimum is only 135.53 MJ mm hm 2 h 1 a 1.
Province, and the central and southern parts of Shaanxi
To analyze the causes of increased precipitation and
remained the areas most threatened by rainfall erosion in
reduced rainfall erosivity in future climate modes and to
the Loess Plateau, and thus requires ecological management
further study the quantitative relationship between rainfall
interventions to prevent soil erosion.
erosivity and precipitation, the present study compared
In the future climate scenarios, Shanxi continues to be
annual rainfall erosivity and annual precipitation. (Missing
the province with the greatest rainfall erosivity, while the
data were recorded as 0, and the historical data comprised
area with the lowest rainfall erosivity changes Ningxia (as
39 × 40 ¼ 1,560 pairs; 26 sample spots in the Loess Plateau
determined by historical data) to Inner Mongolia. This rep-
were selected for the predicted data, and 26 × 81 ¼ 2,106
resents a relatively regular distribution of highs in the
pairs of data were generated in the RCP4.5 and RCP8.5 scen-
southeast and lows in the northwest. Areas with increased
arios, respectively.) The fitting results are shown in Figure 8.
rainfall erosion after 2060 are also concentrated in the
In the historical data, the annual precipitation was
south (RCP4.5) and southeast (RCP8.5). In the RCP4.5
mainly concentrated between 200 and 600 mm. Initially,
Y. Cui et al.
1060
Figure 8
|
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
Relationship between annual rainfall erosivity and precipitation for (a) 1971–2010, (b) 2020–2100 (RCP4.5), and (c) 2020–2100 (RCP8.5).
rainfall erosivity increased at a small rate. The rate of
MJ mm hm 2 h 1. In contrast to the RCP4.5 scenario, the
increase then gradually increased before gradually decreas-
annual rainfall erosivity in the RCP8.5 scenario was
ing. The annual rainfall erosivity was generally below
mainly concentrated below 3,000 MJ mm hm 2 h 1. The
2
1
h . The logistic function provided a
logistic function fits the RCP4.5 data well (R 2 ¼ 0.49, P <
good fitting effect with the historical data (R 2 ¼ 0.77, P <
0.05), but the function does not converge with the RCP8.5
0.01). The formula is as follows:
data. The Allometric1 function in ORIGIN 2020 was used
2,500 MJ mm hm
to do the fitting, and a good fitting effect was also obtained Ry ¼ 7112:09
7037:33
(4)
1 þ (Py =880:61)2:25
for the RCP8.5 data (R 2 ¼ 0.55 (P < 0.05), Figure 8(c)). For the same rainfall amount, when P < 442.51 mm, the rainfall erosivity of the RCP4.5 scenario is greater than that of
2
1
where Ry is the annual rainfall erosivity (MJ·mm·hm ·h ),
RCP8.5, and vice versa. Under the future climate modes,
and Py is the annual precipitation (mm). When the annual
the annual rainfall erosivity gradually increases and
precipitation was 200–600 mm, the annual rainfall erosivity
becomes more discrete with an increase in precipitation,
was about 316.68–2,163.19 MJ mm hm 2 h 1.
which indicates that the rainfall data generated by the
When compared with the historical data, it was evident
HadGEM2-ES model are complex. However, as Figure 8
that the annual precipitation in the RCP4.5 scenario was
shows, when the annual precipitation is less than
mainly concentrated in the 300–900 mm range, and the
1,000.0 mm, the 95% confidence band is narrow, which
annual rainfall erosivity was generally below 2,500
indicates that two functions are applicable for the prediction
1061
Y. Cui et al.
|
Spatiotemporal variation of rainfall erosivity
Hydrology Research
|
51.5
|
2020
of rainfall erosivity using data from HadGEM2-ES. It is
and 41771305) and the National Key Research and
clear, therefore, that the detailed changes of future rainfall
Development Program of China (2017YFC0504702). We
erosivity require further quantitative research.
thank Paul Seward, PhD, from Liwen Bianji, Edanz Group China (www.liwenbianji.cn/ac), for editing the English text of a draft of this manuscript.
CONCLUSIONS Rainfall data from 1971 to 2010 for 39 typical meteorologi-
DATA AVAILABILITY STATEMENT
cal stations in the Loess Plateau were collected to calculate the spatiotemporal variation of rainfall erosivity. Addition-
All relevant data are included in the paper or its Supplemen-
ally, future trends in rainfall erosivity were predicted using
tary Information.
the RCP4.5 and RCP8.5 scenarios. The main conclusions are as follows:
REFERENCES
1. From 1971 to 2010, the mean rainfall erosivity on the Loess Plateau was 1,239.64 MJ mm hm 2 h 1. Overall, annual rainfall erosivity exhibited a slightly decreasing trend. On a monthly basis, rainfall erosivity appeared as a unimodal distribution throughout the year and was mainly concentrated in July to August. There were some differences, however, in the distribution during different hydrological years. 2. The change in annual mean rainfall erosivity was different for different time series, while the spatial distribution of rainfall erosivity showed an overall decrease from the southeast to the northwest. The province with the highest rainfall erosivity was Shanxi, while Ningxia had the lowest. Southern Shanxi, and central and southern Shaanxi were areas susceptible to rainfall erosion in the Loess Plateau. 3. Rainfall erosivity obtained from future climate scenarios is lower than that obtained from historical data. While the precipitation amount increased in future scenarios, the relationship between precipitation and rainfall erosivity was relatively discrete. Future rainfall erosivity presents a relatively regular spatial distribution pattern: high in the southeast and low in the northwest. Areas with intense rainfall erosion on the Loess Plateau were concentrated in the south (RCP4.5) and southeast (RCP8.5).
ACKNOWLEDGEMENTS This research was financially supported by the National Natural Science Foundation of China (Grants 41530858
Abd Elbasit, M. A. M., Huang, J., Ojha, C., Yasuda, H. & Adam, E. O. Spatiotemporal changes of rainfall erosivity in Loess Plateau, China. ISRN Soil Science 2013, 1–8. http://dx. doi.org/10.1155/2013/256352. Ahmad, I., Tang, D., Wang, T., Wang, M. & Wagan, B. Precipitation trends over time using Mann-Kendall and Spearman’s Rho tests in Swat River Basin, Pakistan. Advances in Meteorology 2015, 1–15. http://dx.doi.org/10. 1155/2015/431860. Almagro, A., Oliveira, P. T. S., Nearing, M. A. & Hagemann, S. Projected climate change impacts in rainfall erosivity over Brazil. Scientific Reports 7, 8130. https://doi.org/10. 1038/s41598-017-08298-y. Anees, M. T., Abdullah, K., Nawawi, M. N. M., Norulaini, N. A. N., Piah, A. R. M., Fatehah, O., Syakir, M. I., Zakaria, N. A. & Omar, A. K. M. Development of daily rainfall erosivity model for Kelantan State, Peninsular Malaysia. Hydrology Research 5 (49), 1434–1451. https://doi.org/10.2166/nh.2017.020. Angulo-Martínez, M. & Beguería, S. Estimating rainfall erosivity from daily precipitation records: a comparison among methods using data from the Ebro Basin (NE Spain). Journal of Hydrology 379 (1–2), 111–121. https://doi.org/10. 1016/j.jhydrol.2009.09.051. Chou, S. C., Lyra, A., Mourão, C., Dereczynski, C., Pilotto, I., Gomes, J., Bustamante, J., Tavares, P., Silva, A., Rodrigues, D., Campos, D., Chagas, D., Sueiro, G., Siqueira, G. & Marengo, J. Assessment of climate change over South America under RCP 4.5 and 8.5 downscaling scenarios. American Journal of Climate Change 03 (05), 512–527. http://dx.doi.org/10.4236/ajcc.2014.35043. Collins, W. J., Bellouin, N., Doutriaux-Boucher, M., Gedney, N., Halloran, P., Hinton, T., Hughes, J., Jones, C. D., Joshi, M., Liddicoat, S., Martin, G., Connor, F., Rae, J., Senior, C., Sitch, S., Totterdell, I., Wiltshire, A. & Woodward, S. Development and evaluation of an earth-system model – HadGEM2. Geoscientific Model Development 4 (4), 1051–1075. https://doi.org/10.5194/gmd-4-1051-2011.
1062
Y. Cui et al.
|
Spatiotemporal variation of rainfall erosivity
Da Silva, A. M. Rainfall erosivity map for Brazil. Catena 57 (3), 251–259. https://doi.org/10.1016/j.catena.2003.11.006. Djebou, D. C. S., Singh, V. P. & Frauenfeld, O. W. Analysis of watershed topography effects on summer precipitation variability in the Southwestern United States. Journal of Hydrology 511, 838–849. https://doi.org/10.1016/j.jhydrol. 2014.02.045. Fu, B. J., Zhao, W. W., Chen, L. D., Zhang, Q. J., Lü, Y. H., Gulinck, H. & Poesen, J. Assessment of soil erosion at large watershed scale using RUSLE and GIS: a case study in the Loess Plateau of China. Land Degradation & Development 16 (1), 73–85. https://doi.org/10.1002/ldr.646. Hamed, K. H. & Ramachandra Rao, A. A modified MannKendall trend test for autocorrelated data. Journal of Hydrology 201 (4), 182–196. https://doi.org/10.1016/S00221694(97)00125-X. Lal, R. Soil erosion research methods. Geographical Review 4. https://doi.org/10.2307/215130. Liu, L. & Liu, X. H. Sensitivity analysis of soil erosion in the Northern Loess Plateau. Procedia Environmental Sciences 2, 134–148. https://doi.org/10.1016/j.proenv.2010.10.017. McKee, T. B., Doesken, N. J. & Kleist, J. The relationship of drought frequency and duration to time scales. In: Paper Presented at the 8th Conference on Applied Climatology, Anaheim, CA. Meusburger, K., Steel, A., Panagos, P., Montanarella, L. & Alewell, C. Spatial and temporal variability of rainfall erosivity factor for Switzerland. Hydrology and Earth System Sciences 16 (1), 167–177. https://doi.org/10.5194/hess-16167-2012. Nearing, M. A., Pruski, F. F. & O’Neal, M. R. Expected climate change impacts on soil erosion rates: a review. Journal of Soil and Water Conservation 59 (1), 43–50. Panagos, P., Ballabio, C., Borrelli, P., Meusburger, K., Klik, A., Rousseva, S., Tadić, M. P., Michaelides, S., Hrabalíková, M., Olsen, P., Aalto, J., Lakatos, M., Rymszewicz, A., Dumitrescu, A., Beguería, S. & Alewell, C. Rainfall erosivity in Europe. Science of the Total Environment 511, 801–814. https://doi.org/10.1016/j.scitotenv.2015.01.008. Panagos, P., Borrelli, P., Meusburger, K., Yu, B., Klik, A., Jae, L. K., Yang, J. E., Ni, J., Miao, C., Chattopadhyay, N., Sadeghi, S. H., Hazbavi, Z., Zabihi, M., Larionov, G. A., Krasnov, S. F., Gorobets, A. V., Levi, Y., Erpul, G., Birkel, C., Hoyos, N., Naipal, V., Oliveira, P., Bonilla, C. A., Meddi, M., Nel, W., Al, D. H., Boni, M., Diodato, N., Van Oost, K., Nearing, M. & Ballabio, C. Global rainfall erosivity assessment based on high-temporal resolution rainfall records. Scientific Reports 7 (1), 4175. https://doi.org/10.1038/s41598-017-04282-8. Pimentel, D. Soil erosion: a food and environmental threat. Environment Development and Sustainability 8 (1), 119–137. https://doi.org/10.1007/s10668-005-1262-8. Pimentel, D., Harvey, C., Resosudarmo, P., Sinclair, K., Kurz, D., McNair, M., Crist, S., Shpritz, L., Fitton, L., Saffouri, R. &
Hydrology Research
|
51.5
|
2020
Blair, R. Environmental and economic costs of soil erosion and conservation benefits. Science 267 (5201), 1117–1123. Available from: https://www.jstor.org/stable/ 2886079. Renard, K. G. & Freimund, J. R. Using monthly precipitation data to estimate the R-factor in the revised USLE. Journal of Hydrology 157 (1–4), 287–306. https://doi.org/10.1016/00221694(94)90110-4. Renard, K. G., Foster, G. R., Weesies, G., McCool, D. & Yoder, D. Predicting Soil Erosion by Water: A Guide to Conservation Planning with the Revised Universal Soil Loss Equation (RUSLE). US Government Printing Office, Washington, DC. Sun, W., Shao, Q., Liu, J. & Zhai, J. Assessing the effects of land use and topography on soil erosion on the Loess Plateau in China. Catena 121, 151–163. https://doi.org/10.1016/j. catena.2014.05.009. Vrieling, A., Sterk, G. & de Jong, S. M. Satellite-based estimation of rainfall erosivity for Africa. Journal of Hydrology 395 (3–4), 235–241. https://doi.org/10.1016/j. jhydrol.2010.10.035. Wischmeier, W. H. & Smith, D. D. Predicting Rainfall Erosion Losses – A Guide to Conservation Planning. USDA, Science and Education Administration, Hyattsville, MD, USA. Wu, L., Liu, X. & Ma, X. Spatiotemporal distribution of rainfall erosivity in the Yanhe River Watershed of Hilly and Gully Region, Chinese Loess Plateau. Environmental Earth Sciences 75 (4). https://doi.org/10.1007/s12665-015-5136-6. Xie, Y., Yin, S., Liu, B., Nearing, M. A. & Zhao, Y. Models for estimating daily rainfall erosivity in China. Journal of Hydrology 535, 547–558. https://doi.org/10.1016/j.jhydrol. 2016.02.020. Xin, Z., Yu, X., Li, Q. & Lu, X. X. Spatiotemporal variation in rainfall erosivity on the Chinese Loess Plateau during the period 1956–2008. Regional Environmental Change 11 (1), 149–159. https://doi.org/10.1007/s10113-010-0127-3. Xu, J. Precipitation–vegetation coupling and its influence on erosion on the Loess Plateau, China. Catena 64 (1), 103–116. https://doi.org/10.1016/j.catena.2005.07.004. Yan, D., Werners, S. E., Ludwig, F. & Huang, H. Q. Hydrological response to climate change: the Pearl River, China under different RCP scenarios. Journal of Hydrology: Regional Studies 4, 228–245. https://doi.org/10.1016/j.ejrh. 2015.06.006. Yang, F. & Lu, C. Spatiotemporal variation and trends in rainfall erosivity in China’s dryland region during 1961–2012. Catena 133, 362–372. https://doi.org/10.1016/j.catena.2015. 06.005. Zhang, W. B., Xie, Y. & Liu, B. Y. Rainfall erosivity estimation using daily rainfall amounts. Scientia Geographica Sinica 22 (06), 705–711 (in Chinese with English abstract). http://dx.doi.org/10.3969/j.issn.1000-0690.2002.06.012.
First received 24 March 2020; accepted in revised form 3 July 2020. Available online 31 July 2020
1063
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Potential impact of water transfer policy implementation on lake eutrophication on the Shandong Peninsula: a difference-in-differences approach Jia He, Jiping Yao, Aihua Li, Zhongxin Tan, Gang Xie, Huijian Shi, Xuan Zhang, Wenchao Sun and Peng Du
ABSTRACT Traditional research on lake eutrophication has failed to consider the effect of the South-to-North Water Transfer Project (SNWTP) policy; thus, the difference-in-differences (DID) model, which is usually applied to economic factors, was innovatively introduced to evaluate the effect of such policies on lake eutrophication. Nansi Lake and Dongping Lake in the Shandong Peninsula were selected as the experimental group, and Daming Lake and Mata Lake were selected as the control group. The eutrophication indices of the experimental group and the control group were calculated by the measured chlorophyll-a, total phosphorus, total nitrogen, water transparency and chemical oxygen demand data and used as the explanatory variables of the DID model. Nine environmental and socio-economic factors, such as dissolved oxygen and rural population, were selected as the control variables of the DID model to analyze the impact of the SNWTP policy on lake eutrophication. A joint consideration of environmental and socio-economic factors showed that the eutrophication degree of the experimental lakes deteriorated by 7.10% compared with the control under the
Jia He Jiping Yao Zhongxin Tan (corresponding author) Xuan Zhang Wenchao Sun Peng Du Beijing Key Laboratory of Urban Hydrological Cycle and Sponge City Technology, College of Water Sciences, Beijing Normal University, Beijing 100875, China E-mail: tanzhongxin@bnu.edu.cn Aihua Li Gang Xie Huijian Shi Shandong Academy for Environmental Planning, Jinan 250000, China
influence of the implemented policy. Dissolved oxygen is the main factor affecting the eutrophication of the Shandong Peninsula. This study verifies that the DID model has the potential for use in quantitative analyses of the effect of the SNWTP policy on lake eutrophication. Key words
| difference-in-differences approach, lake eutrophication, parallel trends, policy impact, South-to-North Water Transfer Project
HIGHLIGHTS
• • •
Using DID model to estimate the impact of policy on lake eutrophication is reasonable. Eutrophication deteriorated by 7.10% under the effect of policy and control variables. DO is the key control variables influencing the eutrophication of the Shandong Peninsula.
INTRODUCTION China’s South-to-North Water Transfer Project (SNWTP) is
improving the ecological environment of the water demand
of great strategic significance for alleviating water shortages,
area and promoting sustainable economic and social development. China has invested approximately $20 billion and
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.047
resettled >300,000 people to construct the pipeline, and the SNWTP has become the largest and most expensive interbasin water transfer mega-project in the world
1064
J. He et al.
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
Hydrology Research
|
51.5
|
2020
(Resources ; Li et al. ). The final water volume trans-
nutrients has detrimental effects in the receiving water
ferred is expected to reach 44.8 billion m3/year by 2050 when
system, such as algal blooms and decreased dissolved
the
fully
oxygen (DO) (Barrow ; Zeng et al. ). Although a
implemented (Zhuang et al. ). The first phase of the east-
water transfer project can improve the water quality, it
ern route of the SNWTP (SNWTP-ER) was successfully
might also increase algal growth, and nutrient-rich water
completed and has been operational since late 2015 (Li
pumped from upstream causes cyanobacteria blooms in
eastern,
central
and
western
et al. ). A total of more than 3 × 10
routes
10
are
3
m of water has
the receiving reservoir (Davies et al. ; Zhang et al.
been transferred from the lower Yangtze River in Yangzhou
). Previous studies on the SNWTP-ER have not deter-
city, Jiangsu Province, and crossed the Huai River to the
mined whether it could control algal blooms or contribute
water shortage areas in the Yellow River basin (Guo et al.
to eutrophication in the water shortage period (Wu et al.
a, b). When completed, the SNWTP-ER will consist
). This research generally examines the influence of
of 1,156 km of canals and 54 pumping stations designed to
the SNWTP on environmental factors (total nitrogen and
lift water up to 65 m over the Yellow River (Wang et al.
total phosphorus) and hydrological factors (water quantity
). The pumped water will be diverted from the south
flow rate) directly based on water quality status and monitor-
to the north, primarily through the existing Grand Canal,
ing data (Guo et al. b; Rogers et al. ; Xu et al. ).
and impound in a chain of natural lakes, as regulating
However, to the best of our knowledge, few studies have
reservoirs, namely Gao-Bao-Shaobo Lake (GBSL), Hongze
considered the impact of SNWTP policy implementation
Lake (HZL), Luoma Lake (LML), Nansi Lake (NSL) and
hidden behind the monitoring data. In other words, the
Dongping Lake (DPL) (Wang et al. ; Zhang ; Wu
possibility of eutrophication in the receiving reservoirs with-
et al. ; Guo et al. , a).
out
policy
implementation
should
be
comparatively
Any interbasin water transfer project causes complex
analyzed when considering the impact of the SNWTP
physical, chemical, hydrological and biological changes to
policy. If only the effects of potential changes of the receiv-
the receiving system (Zeng et al. ; Yang et al. ; Yao
ing reservoirs on eutrophication are compared in parallel
et al. a, b, ; Yinglan et al. a). The water and
before and after SNWTP policy implementation, then the
sediment quality of the SNWTP-ER is of particular concern
impact of policy implementation itself could be largely
due to its large contribution to the total volume of transferred
ignored and the socio-economic and environmental effects
water, and it is also one of the most potentially polluted routes
of SNWTP operation policy can be investigated (Rogers
given its proximity to urban and industrial activities (Wu et al.
et al. ). This study aims to establish an approach for
). With the vigorous development of the manufacturing
determining the influence of water transfer policy on receiv-
industry in the Yangtze River Delta and the Bohai Rim,
ing reservoir eutrophication. Nansi Lake and Dongping
large amounts of untreated industrial wastewater are directly
Lake were selected as the objects of the transferred water
discharged to the lakes and rivers along the eastern route, and
policy, and Daming Lake and Mata Lake (without water
agricultural production has led to augmented fertilizer appli-
transferred) were treated as controls. Nansi Lake and
cation, which has substantially increased the loadings of
Dongping Lake are the largest freshwater shallow lakes in
nutrients (including nitrogen and phosphorus) and organic
the Shandong Peninsula and play a vital role in absorbing
matter into river streams, thereby deteriorating water quality
pollutants and storing water (Zhuang et al. ). The differ-
(Li et al. ; Hou et al. ; Gao et al. ; Sheng &
ence-in-differences (DID) model, a common policy impact
Webber ; Kuo et al. ; Yinglan et al. b).
model used in economics, was introduced to establish a
Eutrophication has been recognized as the primary
comprehensive modeling approach for evaluating the
water quality issue for most of the lake ecosystems in the
impacts of SNWTP policy implementation on eutrophica-
world (Smith & Schindler ; Wang et al. ), and nitro-
tion in Nansi and Dongping Lakes and analyzing the key
gen and phosphorus are the primary reasons for algal
factors and processes driving the underlying mechanisms.
blooms caused by excess nutrients (Diersing ; Fang
The objectives of the study are to: (1) unravel the spatial
et al. ). Transferred water with a high content of
and temporal distribution of lake eutrophication in Nansi
J. He et al.
1065
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
Hydrology Research
|
51.5
|
2020
Lake, Dongping Lake, Daming Lake and Mata Lake; (2) use
from south to north and 5–25 km wide from east to west,
the SNWTP policy as a single factor or combine environ-
has an area of 1,266 km2 and a total storage volume of
mental and socio-economic factors as a composite factor
6.37 × 1010 m3 and is the first largest freshwater lake in the
to establish the DID model for evaluating variation trends
Shandong Peninsula along the eastern route. Nansi Lake
in eutrophication on water transfer lakes; (3) apply parallel
consists of Nanyang Lake, Dushan Lake, Zhaoyang Lake
trend and robustness analyses to validate the feasibility of
and Weishan Lake without physical boundaries defining
the DID model; and (4) identify the primary driving factors
each lake, and it was divided into the upper and lower sec-
that affect eutrophication and supply policy recommen-
tions by the Erji Dam Pumping Station Hinge Project in
dations to develop management strategies for water quality
1960. The upper lake lies on the northern side of the Erji
safety and pollution control.
Dam, and the lower lake is located on the southern side. Five state-controlled monitoring sections were located in Nansi Lake, namely the Qianbaikou (S1) and Nanyang
MATERIALS AND METHODS
(S2) sections in the upper lake, the Erjiba (S3) section near the Erji Dam and the Dajuan (S4) and Daodong (S5)
Study area
sections in the lower lake (Figure 1). Dongping Lake (35 300 –36 200 N, 116 000 –116 300 E) is
0
0
0
0
Nansi Lake (34 27 –35 20 N, 116 34 –117 21 E) is located
located in Dongping County, west of Shandong Peninsula,
in Weishan County, southwest Shandong Peninsula, China
China (Figure 1). The lake has an annual mean depth of
(Figure 1). The lake, which is approximately 126 km long
2–4 m and a total storage volume of 4 × 109 m3 and covers
Figure 1
|
Location of Nansi Lake, Dongping Lake, Mata Lake and Daming Lake and the SNWTP-ER route in the Shandong Peninsula.
1066
J. He et al.
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
Hydrology Research
|
51.5
|
2020
627 km2, and it is the second-largest freshwater lake in the
implementation effect of projects or public policies. The
Shandong Peninsula along the eastern route. The Hubei
notable superiority of DID estimation is derived from its
(S6), Huxin (S7) and Hunan (S8) state-controlled monitor-
simplicity as well as its potential to circumvent many of
ing sections are distributed in the north, middle and south
the endogenous problems that typically arise when making
of Dongping Lake, respectively. Relying on its special geo-
comparisons between heterogeneous individuals (Meyer
graphical position and function along the SNWDP-ER,
; Bertrand et al. ). In the assessment, the policy
Dongping Lake serves as the final reservoir and an essential
experimental group and the control group generally do
flood control project in the lower reaches of the Yellow
not have complete randomness in sample allocation. The
River (Yao et al. a, b).
experiment involving a nonrandom allocation policy exper-
Since the implementation of the SNWDP-ER policy in
imental group and control group is known as a natural trial,
2015, Nansi Lake and Dongping Lake have served as the
and its important feature is that systematic differences might
water-supplying lakes and impounded reservoirs; however,
occur between the experimental group and the control
maintaining good water quality while meeting water
group prior to the implementation of the experiment. If
demands remains a great challenge (Grant et al. ). In
the initial difference is ignored and only a horizontal com-
recent years, Nansi Lake is surrounded by the dense indus-
parison between the experimental group and the control
trial and population zones, which has resulted in a large
group is performed after implementing the experiment, the
amount of industrial and domestic sewage discharged into
estimated experimental effect is likely to be biased due to
Nansi Lake each year, thus exacerbating water pollution
the mixed effect of the initial difference. The DID model
and eutrophication (Li ; Yao et al. a, b).
was first introduced in 1985 to solve this problem
Severe contamination and mining, chemical, electric
(Ashenfelter & Card ), and since then, increasing
power, manufacturing and many other industries within
attention has been focused on applying the model.
the main tributary river basin are the main factors responsible for the load of pollutants in Dongping Lake (Guo
Setting and verification of the DID model in lake
et al. ; Wang et al. ; Xu et al. ).
eutrophication research
Daming Lake is located in Jinan City and Mata Lake is located in Zibo City, and both lakes were identified as key
Since the route of the SNWTP-ER was opened to water
nature reserves of Shandong Peninsula, China (Figure 1).
supply, its impact on lake eutrophication in Shandong Pro-
The Lixiating (S9) and Mata Lake (S10) state-controlled
vince might be due to the policy effect of the project and
monitoring sections are distributed in Daming Lake and
the time effect of the time trend changes in lake water eutro-
Mata Lake, respectively. The SNWTP-ER policy has not
phication. The question of how to analyze the policy effect
been implemented on Daming Lake and Mata Lake, and
for correctly evaluating the impact of the route of the
the spatial location and characteristics are adjacent to
SNWTP-ER on lake eutrophication in Shandong Province
Nansi Lake and Dongping Lake. Reducing the difference
is highly important. The DID model can effectively analyze
between the experimental groups and the control groups is
and objectively evaluate the policy effect. In summary, we
expected to improve the accuracy of the DID model. In
apply the DID model to research the change of lake eutro-
this study, considering the comprehensiveness, integrity
phication in Shandong Province before and after the route
and availability of the data, among the 13 lakes and reser-
of the SNWTP-ER.
voirs in the Shandong Peninsula, Daming Lake and Mata
Using trophic level indices (TLIs) as the explained variable, ‘treated’ ¼ 1 indicates the water inflow lake of the
Lake were treated as the control group.
section in the line of the SNWTP-ER and ‘treated’ ¼ 0 indiBasic principle of the DID model
cates that the lake where the section is located has no water diversion of the SNWTP-ER line. Time is used to
DID estimation has become an increasingly popular method
express the ‘time’. The value of the year when the
of
SNWTP-ER is open to water and the following years is 1;
estimating
the
econometric
evaluation
of
the
1067
J. He et al.
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
otherwise, the value is 0. We use ‘did’ to represent the implementation effect of the SNWTP-ER, i.e., the intersection of ‘treated’ and ‘time’. Additionally, xit is the control variable, which includes the time fixed effect and regional fixed effect, and θit represents the constant term and disturbance term. The basic measurement model is shown as follows: Yit ¼ β 0 þ β 1 didit þ β 2 treated þ β 3 time þ αxit þ θit
Table 1
Hydrology Research
|
|
51.5
|
2020
Standard values of the TLI
Grades
Meaning
TLI
TN
TP
COD
Level I
Oligotrophic
<30
0.2
0.02
2
Level II
Mesotrophic
30–50
0.5
0.1
4
Level III
Light eutropher
50–60
1.0
0.2
6
Level IV
Middle eutropher
60–70
1.5
0.3
10
Level V
Hyper eutropher
>70
2.0
0.4
15
(1)
where β 0 represents the common initial eutrophication
Lake, Dongping Lake, Daming Lake and Mata Lake were
mean value of all sections before the SNWTP-ER, β 1
calculated using chlorophyll-a (Chl-a), total phosphorus
represents the policy effect of the SNWTP-ER after control-
(TP), total nitrogen (TN), water transparency (SD) and
ling the initial eutrophication difference and common trend,
chemical oxygen demand (CODMN), which were computed
β 2 represents the difference in lake initial eutrophication
from
between the experimental group and the control group
material, Table S1. Water samples from the monitoring
before and after the SNWTP-ER, β 3 represents the trend of
stations in Shandong Province were collected from January
eutrophication of the experimental group and the control
2009 to December 2017 and analyzed for Chl-a, TP, TN, SD
group before and after the SNWTP-ER and α is the coeffi-
and CODMN based on the methods outlined by the Chinese
cient of the control variable xit .
national water environmental protection standard in China
Parallel trend and robustness analyses were used to validate the DID models. The parallel trend assumption
Equations
(2)–(7)
presented
in
Supplementary
(GB11914-89, GB11893-89, GB11894-89 and SL88-1994) (Wei ; Huo et al. ).
indicates that the experimental group and the control group follow parallel paths over time. This assumption allows the DID to account for unobserved variables, which are assumed to remain fixed over time (Dimick & Ryan ; Zhou et al. ). Replacement of the interpreted variable was selected as the robustness analysis. If the regression results are consistent with the original results and the core variables are significant, the estimation can be considered robust.
TLI
X
¼
m X
Wj × TLI (j)
(2)
j¼1
where Wj is the correlative weight for the TLI of j, TLI (j) is P the TLI of j, TLI ð Þ is the comprehensive eutrophication index and the Wj values of Chl-a, TP, TN and CODMN are 0.2663, 0.1879, 0.1790 and 0.1834, respectively. The units of Chl-a, TP, TN, SD and CODMN are mg/m3, mg/L, mg/L, m and mg/L, respectively.
Selection of interpreted and control variables for the DID model in lake eutrophication research
TLI (Chl-a) ¼ 10 × [2:5 þ 1:086 × ln (Chl-a)]
(3)
TLI (TP) ¼ 10 × [9:436 þ 1:624 × ln (TP)]
(4)
TLI (TN) ¼ 10 × [5:453 þ 1:694 × ln (TN)]
(5)
TLI (SD) ¼ 10 × [5:118 þ 1:940 × ln (SD)]
(6)
TLI (CODMN ) ¼ 10 × [0:109 þ 2:661 × ln (CODMN )]
(7)
The eutrophication index treated as an interpreted variable was estimated by standard values. Due to the varied topographies, environmental background, industrial layout and human activities, the assessment methods for lake eutrophication are diverse for different regions because the comprehensive TLI is widely applied to evaluate the conditions of water quality (Liu et al. ). The standard
Environmental factors and socio-economic factors were
values of the TLIs are given in Table 1. The TLIs of Nansi
selected as the control variables, and nine control variable
1068
J. He et al.
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
Hydrology Research
|
51.5
|
2020
factors in total are used in this paper. The experimental
surface water reservoir section in Shandong Province from
group included five sections of Nansi Lake, and three sec-
January 2009 to December 2017 and the Shandong Statisti-
tions of Dongping Lake and Daming Lake and Mata Lake
cal Yearbook from 2010 to 2018.
each contained a section as a control. A total of 1,080 water samples from each section were collected monthly from January 2009 to December 2017. Four environmental
RESULTS AND DISCUSSION
factors, including water temperature (Tw) and DO, were measured with a portable multiparameter water quality ana-
Spatial and temporal distribution of lake eutrophication
lyzer (YSI Professional Plus, Yellow Springs, Ohio, USA).
in the Shandong Peninsula
The N:P ratios were calculated by the TN and TP values. Hours of sunshine (HS) were collected from the ‘Statistical
To analyze the spatial and temporal variation trend of eutro-
Yearbook of Shandong Province’ from 2009 to 2017 (details
phication in Nansi Lake, Dongping Lake, Daming Lake and
are listed in Supplementary material, Table S2; Lu ).
Mata Lake affected by the implementation of the SNWTP-
Five socio-economic factors, namely gross domestic pro-
ER policy, the TLIs of each section in four lakes from
duct (GDP), gross output value of agriculture (GOVA), gross
2009 to 2017 were calculated using the CODMN, TP, TN,
output value of animal husbandry (GOVAH), rural popu-
Chl-a and SD (Figure 2). The spatial and temporal variation
lation (RP) and number of inbound tourists (NIT) received
trends of three state-controlled sections in Dongping Lake
by each city, were also collected from the ‘Statistical Year-
are generally unanimous, and they all indicated significant
book of Shandong Province’ from 2009 to 2017 (Lu ).
improvements in 2010, from the middle eutropher to meso-
Detailed data are listed in Supplementary material,
trophic levels. However, after the official water transfer
Table S2. GDP refers to the final result of production activi-
began in 2015, an obvious slight upward trend is suggested
ties of all permanent units in the region in a certain period,
throughout Dongping Lake as shown in the lower left of
which is the sum of the added value of each industry and
Figure 2. The water quality of Hubei and Hunan sections
reflects the relationship between the economic development
increases to light eutropher, and the center of the lake
level of the region and lake eutrophication. GOVA refers to
remains at the mesotrophic level. Certain differences
the total scale and total results of agricultural production in
occur in the spatial and temporal variation trends of five
the region in a certain period and reflects the relationship
state-controlled sections in Nansi Lake, although significant
between agricultural nonpoint source pollution and lake
improvements were not observed after the implementation
eutrophication. GOVAH refers to the total scale and results
of the water transfer policy in 2015. The eutrophication
of animal husbandry production in a certain period in the
degree was more severe in the northern portion than the
region and reflects the relationship between livestock
southern portion of Nansi Lake. An improvement of the
waste and lake eutrophication. RP refers to the total popu-
eutrophication degree at the Qianbaikou and Nanyang
lation except for the total permanent population living in
sections occurred in 2014, although it subsequently deterio-
the urban area and reflects the relationship between
rated to middle eutropher at the Nanyang section. The
undischarged domestic garbage and lake nutrition. NIT
analysis of the TLI values did not provide obvious results
refers to the number of inbound international tourists and
for the effect of policy implementation in the Erjiba,
domestic tourists and reflects the relationship between
Dajuan and Daodong sections based on the maintenance
household garbage and lake eutrophication.
of the mesotrophic level from 2013.
Data sources of the DID model in the study area
Lake decreased from hypereutropher to mesotrophic in a
The variable data of the DID model applied to lake eutrophi-
in 2015. The lowest TLI values of Daming Lake appeared in
cation research for each monitoring section in the study area
2012 and 2013, and they slightly increased to the light eutro-
are taken from the water quality monitoring data of the
pher level from 2014.
In the control lakes, the eutrophication level of Mata year-by-year manner and significant improvements occurred
J. He et al.
1069
Figure 2
|
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
Hydrology Research
|
51.5
|
2020
Spatial and temporal variation of TLIs in 10 state-controlled sections from 2009 to 2017.
Quantitative and qualitative analysis of the impact of
the P-value passing the 5% level test, and the fitting degree
SNWTP-ER policy on lake eutrophication
of the relationship has slightly improved. The β1 value of
An analysis of the temporal and spatial distribution shows
of the SNWTP-ER policy, the eutrophication degree of the
that the eutrophication degree could be affected by the
experimental lakes deteriorates by 6.20% compared with
SNWTP-ER policy. However, it is impossible to quantitat-
that of the control lakes.
Model 2 indicates that when considering only the influence
ively and qualitatively analyze the impact of the policy.
Models 3 and 4 added the environmental control vari-
Therefore, the DID model was introduced to estimate the
ables shown in Table 2, such as Tw, DO, HS and N:P
potential influence of the SNWTP-ER policy on the lake
ratio. There are no fixed time and regional effects in
eutrophication degree in the Shandong Peninsula.
Model 3, and the significance of the P-value is only 10%.
The results are shown in Table 2 and the Model 1 and
After fixing the time and regional effects in Model 4, the
Model 2 regressions considered only the effect of the
relationship between the TLIs and policy effects and
SNWTP-ER policy without control variables. Model 1
environmental factors is significant, with R 2 ¼ 0.70, and
does not include fixed time and regional effects, and the
the policy effect is significantly improved at the 1% level.
results showed that the relationship is rather poor, with
The quantitative and qualitative analysis results of Model 4
R 2 ¼ 0.30 and a P-value at only 10% significance. Model 2
show that the β1 value is positive at 5.89, which indicates
includes fixed time and regional effects, and the SNWTP-
that the eutrophication degree of the experimental lakes is
ER policy effect of lake eutrophication is significant, with
significantly increased by 5.89% compared with the control
1070 J. He et al.
|
|
DID regression estimations for TLIs from different effect factors
Type
Policy effect
Regression
Model 1 Model 2 Model 3 Model 4
Model 5 Model 6
Model 7
Model 8
Model 10
Model 11
β1
6.20*
6.20**
4.33*
5.89***
9.76***
10.75***
5.48**
10.46*** 7.68**
7.02***
7.10***
Time effect
No
Control
No
Control
No
Control
Control
No
Control
Control
Control
Regional effect No
Control
No
Control
No
Control
Control
No
Control
Control
Control
0.51
0.64
0.70
0.60
0.58
0.48
0.82
0.82
0.80
0.80
R
2
0.33
Environmental effect
Socio-economic effect
Comprehensive effect
Model 9
Coefficient of control variable Tw
NA
NA
NA
1.76***
NA
NA
NA
NA
1.67***
1.67***
1.56***
DO
NA
NA
NA
2.50*** NA
NA
NA
NA
1.37**
2.02***
2.10***
HS
NA
NA
NA
0.01**
NA
NA
NA
NA
0.01
NA
NA
N:P ratio
NA
NA
NA
0.08*** NA
NA
NA
NA
0.11***
0.10***
0.10***
03
GDP
NA
NA
NA
NA
NA
0.01
NA
NA
2.37 × 10
NA
NA
GOVA
NA
NA
NA
NA
NA
6.13 × 10 06**
5.62 × 10 06***
NA
6.46 × 10 06***
6.66 × 10 06***
6.53 × 10 06***
GOVAH
NA
NA
NA
NA
NA
5.66 × 10 06*
5.68 × 10 06**
NA
3.10 × 10 06
NA
NA
RP
NA
NA
NA
NA
NA
0.02
NA
NA
0.03**
0.05***
0.05***
NIT
NA
NA
NA
NA
NA
0.51
NA
NA
0.57*
0.31
NA
Using DID model to identify the impact of SNWTP policy on lake eutrophication
Table 2
Note: *, ** and *** indicate significant at the level of 10, 5 and 1%, respectively.
Hydrology Research
| 51.5
| 2020
1071
J. He et al.
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
Hydrology Research
|
51.5
|
2020
lakes after the SNWTP-ER policy was implemented. The
In summary, although Model 2, Model 4 and Model 7
value of β1 in Model 4 is lower than that in Model 2,
evaluate the impact of policy from different perspectives,
which suggests that the estimated eutrophication degree
Model 11 represents a more reasonable model for
decreased by 0.31% under the combined influence of
estimating the influence of the water transfer policy on
policy and environmental effect factors compared with
lake eutrophication.
that when only the policy effect is considered. Tw and DO
As shown in Table 2, the TLIs of Model 11 are signifi-
are the principal influence factors with regression coeffi-
cantly negatively correlated with the DO and N:P ratio
cients of 1.76 and 2.50, respectively.
and positively correlated with the Tw, GOVA and RP. The
Models 5–7 add GDP, GOVA, GOVAH, RP and NIT as
concentration of DO in the lake water could reflect the bio-
socio-economic control variables. Without the fixed time and
logical processes of algae through photosynthesis and
regional effects, the significance of Model 5 is at the 1% level
respiration (Zeng et al. ). The eutrophication degree
2
and the relationship coefficient R is 0.60. However, Model 7
usually decreases with an increase in the concentration of
includes fixed time and regional effects, and the significance
DO. According to the law of Redfield, the N:P ratio of
2
of the policy effect decreased to the 5% level, with R below
algal cells is in the range of 16:1, and the eutrophication
0.5. This result could occur because the selected control vari-
condition of water indicates that nitrogen and phosphorus
ables are not comprehensive without the fixed time and
reach the appropriate proportion, which leads to the out-
regional effects. The value of β1 in Model 7 is similar to
break of lake eutrophication (Geider & La Roche ).
that of Model 4, which indicates that the degree to which
The low concentration of phosphorus in the lakes of
eutrophication is affected by socio-economic factors and
Shandong Province is the key to inhibiting water eutrophica-
environmental factors does not significantly differ. In con-
tion, and the TLIs might decrease with an increase in the
trast, as shown in Table 2, the regression coefficients of the
N:P ratio. Tw is another essential environmental influence
socio-economic factors are notably low, suggesting that the
factor that affects the eutrophication degree. For every
impact of socio-economic variables on eutrophication might
1 C increase in annual temperature, the annual algal bio-
be lower than that of environmental variables.
mass is expected to increase by 0.145 times (Ye et al. ).
Models 8–11 combine the environment effect factors
In addition to the direct impact of environmental factors
and socio-economic factors as comprehensive control vari-
on lake eutrophication, the indirect impact of social and
ables. Model 11 includes the fixed time and regional
economic factors on lake water quality is also increasingly
effects, and the SNWTP-ER policy effect of lake eutrophica-
significant. GOVA can reflect the relationship between agri-
tion is significant with a P-value reaching the 1% level, and
cultural nonpoint source pollution and lake nutrients. An
the fitting degree of the relationship is obviously improved
increase in the GOVA number increases the risk of nitrogen
with R 2 ¼ 0.80. The best estimation accuracy among the
and phosphorous fertilizers and pesticides entering the lake
11 models is that of Model 11, which includes five compre-
along with surface runoff. An increase of RP directly leads to
hensive factors. Compared with the accuracy of the other 10
domestic garbage pollution in the water body, resulting in
models developed in this paper, the estimation accuracy
eutrophication of the lake.
from Model 11 is reasonable because this model considered the combined effect rather than only the environment effect
Parallel trend and robustness analyses of the DID model
or socio-economic influence. The β1 value of 7.10 indicates that under the combined influence of the SNWTP-ER
Parallel trend analysis is a necessary prerequisite for policy
policy with all of the effect factors, the eutrophication
effect evaluations using the DID model. If the parallel
degree of the experimental lakes deteriorates by 7.10% com-
trends between the experimental group and the control
pared with the control lakes, which is 0.90% higher than
group are not significantly different, the probability of bias
that when considering only the influence of policy. Tw and
in the empirical results might be reduced. In other words,
DO are the main influence factors, with regression coeffi-
in this study, prior to the implementation of the SNWTP
cients of 1.56 and 2.10, respectively.
policy, the variation trend of the TLI values of the
J. He et al.
1072
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
experimental group (Nansi Lake and Dongping Lake) are the same as those of the control group (Mata Lake and Daming Lake) or are different but have not changed significantly over time. The above two cases could indicate that the eutrophication degrees of the experimental group and the control group before the implementation of the policy have the same trend.
Table 3
Hydrology Research
|
2020
Regression
β1
Policy effect
Model 12 Model 13
0.20*** 0.20***
Environmental effect
Model 14 Model 15
0.18*** 0.19***
Socio-economic effect
Model 16 Model 17 Model 18
0.17** 0.19*** 0.20**
Comprehensive effect
Model 19 Model 20 Model 21
0.20*** 0.18*** 0.17***
allel trends between the experimental group and the control group before 2015 (left of the red line). After 2015, a signifi-
|
Type
tion trends for all outcomes as shown in Figure 3(a). The
cant difference appears between the two groups, with a
51.5
Robustness analysis of DID estimations for TP as interpreted variables
Statistical evidence indicated comparable preintervenvariation trend of the mean TLI values did not reject the par-
|
Note: ** and *** indicate significant at the level of 5 and 1%, respectively.
decreasing trend of eutrophication degree in the control group and an increasing trend in the experimental group,
Comparison of the assessment ability of the DID model
which meets the preconditions of evaluating the policy
with unconsidered policy effect models on lake
impact using the DID model. Moreover, Figure 3(b) illus-
eutrophication
trates that the coefficient fluctuates approximately 0 before the policy is implemented and is significantly positive 1
To further assess the evaluation ability of the DID models
year after the policy is implemented. This result indicates
for lake eutrophication established in this study, a Pearson
that the parallel trends assumption outlined above can be
correlation analysis and multiple linear regression were
evaluated using a regression model in this study and that
applied to evaluate eutrophication without considering the
the policy effect appears after the SNWTP policy is
policy effect for comparative analysis. First, the Pearson cor-
implemented, with apparent changes over time.
relation analysis was established between the nine control
To verify the robustness of the DID model, the inter-
variables and TLI values. The results listed in Table 4
preted variables could be redefined for the DID regression.
show that the TLIs are significantly correlated with DO,
TP was replaced in the interpreted variables, and the
HS, GOVA, GOVAH and NIT, with the strongest corre-
policy cut-off point was still 2015. The robustness analysis
lation between GOVAH and NIT, which is not consistent
results listed in Table 3 suggest that the core variables
with the conclusion of the DID model. Moreover, HS,
passed the significance test with a P-value of 5 or 1%,
GOVAH, GOVA and NIT were negatively correlated with
which is consistent with the regression results in Table 2.
the TLIs, indicating that when the Pearson’s analysis did
Figure 3
|
Parallel trends of the eutrophication degree between the experimental group and the control group. Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/nh.2020.047.
J. He et al.
1073
Table 4
|
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
Hydrology Research
|
51.5
|
2020
Pearson correlations among Tw, DO, HS, N:P ratio, GDP, GOVA, GOVAH, RP, NIT and TLI Tw
DO
HS
N:P ratio
GDP
GOVA
GOVAH
RP
NIT
TLI
Tw
1.000
DO
0.005
1.000
HS
0.024
0.030
1.000
N:P ratio
0.001***
0.196
0.005
1.000
GDP
0.212***
0.011
0.008
0.305***
1.000
GOVA
0.276***
0.011
0.017
0.041***
0.135***
1.000
GOVAH
0.197
0.000***
0.023***
0.056***
0.093***
0.846***
RP
0.013
0.075**
0.093**
0.220
0.198***
0.201***
0.183
1.000
NIT
0.029
0.071***
0.049
0.000***
0.012*
0.134**
0.318***
0.009
1.000
TLI
0.012
0.181***
0.067**
0.007
0.003
0.220***
0.375***
0.017
0.409***
1.000
1.000
Note: *, ** and *** indicate significant at the level of 10, 5 and 1%, respectively.
not consider the influence of policy, a certain amount of
Policy recommendations
deviation occurred in the analysis results. A multiple linear regression is established using five control variables selected
Based on the above quantitative and qualitative analysis
in Model 11 as shown below.
results of the impact of the SNWTP-ER policy on lake eutrophication, the main influencing factors are selected from Model 11, and we propose the following three suggestions.
Y ¼ 63:0 þ 1:41 × Tw 2:89 × DO þ 0:0049 × N:P
First, while developing the local agricultural economy,
0:000004 × GOVA þ 0:0058 RP
(8)
additional attention should be focused on agricultural nonpoint
source
pollution
discharge
for
control
and
governance (Wang et al. , a, b). In agricultural The results showed that the relationship between the
production, especially in crop production, fertilizers and
TLIs and five control variables was significant, with P ¼
pesticides are heavily used every year, which might cause
2
0.00 < 0.05; however, the correlation coefficient R ¼ 0.51 2
a large amount of nitrogen and phosphorus to enter the
was not better than that of the DID Model 11, with R ¼
lake via the circulation process and surface runoff. In
0.80. In addition, the N:P ratio and RP did not pass the
other words, an increase in agricultural output value might
t-test of significance at P-values greater than 0.05, although
enhance the discharge of agricultural pollutants, meaning
DO is still treated as a dominant factor affecting water qual-
that the eutrophication degree of lakes might continuously
ity, which is consistent with the DID model and the Pearson
deteriorate. One suggestion might be to adjust the planting
correlation analysis. Moreover, the DID model could be
varieties of crops and reasonably control the amount of pes-
used as an alternative method to evaluate and quantify the
ticides and fertilizers used in the lake area. Another
impact of SNWTP policy on lake water quality in Shandong
suggestion is to establish a wetland isolation zone surround-
Province. The advantage of the DID model is that it could
ing the lake to decrease the risk of nitrogen and phosphorus
reveal the extent of policy impacts hidden in the experimen-
entering the lake by way of nutrient substances consumed by
tal monitoring data. In addition, the DID model offers a
the growth of aquatic plants. Additionally, the development
diversified perspective that includes policy, environment
of the economy is expected to result in advanced production
and socio-economic diversification to analyze the impact
technology and management experience that can alleviate
of the SNWTP on lake water quality, thus making the
the pollution caused by agricultural production to a certain
entire assessment and analysis more comprehensive than
extent and improve the adverse impact of agricultural econ-
conventional models.
omic development on lake eutrophication.
1074
J. He et al.
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
Hydrology Research
|
51.5
|
2020
Second, the shortage of rural sewage treatment must be
correlation analysis and multiple linear regression result
addressed. Production and domestic garbage generated by
showed that the DID models are more suitable for analyzing
an increasing population are expected to have a serious
lake eutrophication in Shandong Province than convention-
impact on the aquatic ecology of the lake if it is not handled
al models. The DID models identified the environmental
properly and directly enters the water body. It is suggested
factors that should receive additional attention to prevent
that human activities should be separated from ecological
eutrophication of Nansi Lake and Dongping Lake. It is
nature reserves to a certain extent. The capacity of domestic
suggested that attention should be focused on controlling
sewage treatment should be enhanced, and upgrades and
and treating nonpoint agricultural source pollution emis-
reconstruction should be performed in sewage treatment
sions to compensate for the shortage of rural sewage
plants (Scheren et al. ). In areas where flush toilets
treatment, increasing the circulation of lakes and restoring
are mainly used, latrine improvement in the countryside
the natural hydrological fluctuation rhythm of a portion of
and sewage treatment should be integrated. In areas where
the lakes.
traditional dry latrines and waterless latrines are mainly used, fecal pollution-free treatment and resource utilization should be conducted and construction space should be
ACKNOWLEDGEMENTS
reserved for later sewage treatment. Third, the circulation of lakes should be improved, and
The project was financially supported by the National Key
the natural hydrological fluctuation rhythms should be
R&D Program of China (Grant No. 2016YFC0401308),
restored in a portion of the lakes. The diversion of the
the Chinese National Special Science and Technology
SNWTP-ER might be contrary to the natural flow of the
Program of Water Pollution Control and Treatment (Grant
lake, which changes the natural flow direction of the water
No. 2017ZX07302004), the Fundamental Research Funds
body to a certain extent and might cause disturbances in
for the Central Universities (Grant No. 2019NTST19) and
the ecosystem. From the perspective of improvement
the National Natural Science Foundation of China (Grant
measures, optimizing lake fishery management, improving
Nos. 51679006, 51879006 and 41603113) and the 111
the effective interception capacity of nutrients in the lake-
Project (B18006).
side and implementing lake ecological restoration projects (Zhu et al. ) are the keys to controlling lake eutrophication and improving the regional lake ecological quality in
DATA AVAILABILITY STATEMENT
the future. Data cannot be made publicly available; readers should con-
CONCLUSIONS The DID model is feasible for the quantitative and qualitat-
tact the corresponding author for details.
REFERENCES
ive detection of the impact of SNWTP-ER policy on lake eutrophication in the Shandong Peninsula. Our results demonstrated that Model 11, which included the influence of policy and considered a composite of environmental and socio-economic factors, acquired the optimal fitting degree with R 2 ¼ 0.80, and in this model, the eutrophication degree of the transferred lake deteriorated by 7.10% compared with that of the control lake without water transfer. DO was the main factor influencing the eutrophication of the Shandong Peninsula. A comparison of the Pearson
Ashenfelter, O. & Card, D. Using the longitudinal structure of earnings to estimate the effect of training programs. Rev. Econ. Stat. 67, 648–660. Barrow, C. Long-distance water transfer – a Chinese case study and international experiences – Biswas, AK, Dakang, Z, Nickum, JE, Changming, L. Third World Q. 6, 237–238. Bertrand, M., Duflo, E. & Mullainathan, S. How much should we trust differences-in-differences estimates? Q. J. Econ. 119, 249–275. Davies, B. R., Thoms, M. & Meador, M. An assessment of the ecological impacts of inter-basin water transfers, and their
1075
J. He et al.
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
threats to river basin integrity and conservation. Aquat. Conserv. 2, 325–349. Diersing, N. Phytoplankton Blooms: The Basics. Florida Keys National Marine Sanctuary, National Oceanic and Atmospheric Administration, Washington, DC. Dimick, J. B. & Ryan, A. M. Methods for evaluating changes in health care policy: the difference-in-differences approach. JAMA 312, 2401–2402. Fang, Q., Wang, G., Liu, T., Xue, B., Sun, W. & Shrestha, S. Unraveling the sensitivity and nonlinear response of water use efficiency to the water-energy balance and underlying surface condition in a semiarid basin. Sci. Total Environ. 699, 134405. Gao, X., Zhuang, W., Chen, C.-T.A. & Zhang, Y. Sediment quality of the SW coastal Laizhou Bay, Bohai Sea, China: a comprehensive assessment based on the analysis of heavy metals. PLoS ONE 10, 1–27. Geider, R. J. & La Roche, J. Redfield revisited: variability of C:N:P in marine microalgae and its biochemical basis. Eur. J. Phycol. 37, 1–17. Grant, E. H. C., Lynch, H. J., Muneepeerakul, R., Arunachalam, M., Rodriguez-Iturbe, I. & Fagan, W. F. Interbasin water transfer, riverine connectivity, and spatial controls on fish biodiversity. PLoS ONE 7, 1–7. Guo, C., Chen, Y., Li, W., Xie, S., Lek, S. & Li, Z. Food web structure and ecosystem properties of the largest impounded lake along the eastern route of China’s south-to-north water diversion project. Ecol. Inform. 43, 174–184. Guo, C., Chen, Y., Liu, H., Lu, Y., Qu, X., Yuan, H., Lek, S. & Xie, S. a Modelling fish communities in relation to water quality in the impounded lakes of China’s south-to-north water diversion project. Ecol. Model. 397, 25–35. Guo, C., Chen, Y., Xia, W., Qu, X., Yuan, H., Xie, S. & Lin, L.-S. b Eutrophication and heavy metal pollution patterns in the water supplying lakes of China’s south-to-north water diversion project. Sci. Total Environ. 711, 134543. Hou, Q., Yang, Z., Ji, J., Yu, T., Chen, G., Li, J., Xia, X., Zhang, M. & Yuan, X. Annual net input fluxes of heavy metals of the agro-ecosystem in the Yangtze River Delta, China. J. Geochem. Explor. 139, 68–84. Huo, S., Ma, C., Xi, B., Su, J., Zan, F., Ji, D. & He, Z. Establishing eutrophication assessment standards for four lake regions, China. J. Environ. Sci (China) 25 (10), 2014–2022. Kuo, Y., Liu, W., Zhao, E., Li, R. & Muñoz-Carpena, R. Water quality variability in the middle and down streams of Han River under the influence of the middle route of south-north water diversion project, China. J. Hydrol. 569, 218–229. Li, W. Research on Water Quality of Nansi Lake and the Inflowing Rivers. Tianjin University, Tianjin. Li, X., Liu, L., Wang, Y., Luo, G., Chen, X., Yang, X., Gao, B. & He, X. Integrated assessment of heavy metal contamination in sediments from a coastal industrial basin, NE China. PLoS ONE 7, e39690. https://doi.org/10.1371/ journal.pone.0039690.
Hydrology Research
|
51.5
|
2020
Li, Q., Peng, Y., Wang, G., Wang, H., Xue, B. & Hu, X. A combined method for estimating continuous runoff by parameter transfer and drainage area ratio method in ungauged catchments. Water 11, 1104. Li, Q., Wang, G., Wang, H., Shrestha, S., Xue, B., Sun, W. & Yu, J. Macrozoobenthos variations in shallow connected lakes under the influence of intense hydrologic pulse changes. J. Hydrol. 584, 124755. Liu, X., Zhang, G., Sun, G., Wu, Y. & Chen, Y. Assessment of lake water quality and eutrophication risk in an agricultural irrigation area: a case study of the Chagan Lake in Northeast China. Water 11, 2380. Lu, W. Editorial Board and Staff. Shandong Statistical Yearbook–2018. China Statistics Press, Beijing. Meyer, B. D. Natural and quasi-experiments in economics. J. Bus. Econ. Stat. 13, 151–161. Ministry of Water Resources South-North Water Transfer Project Masterplan (Summary). Ministry of Water Resources, Beijing. Rogers, S., Chen, D., Jiang, H., Rutherfurd, I., Wang, M., Webber, M., Crow-Miller, B., Barnett, J., Finlayson, B., Jiang, M., Shi, C. & Zhang, W. An integrated assessment of China’s south–north water transfer project. Geogr. Res. 58, 49–63. Scheren, P., Zanting, H. A. & Lemmens, A. M. C. Estimation of water pollution sources in Lake Victoria, East Africa: application and elaboration of the rapid assessment methodology. J. Environ. Manage. 58, 235–248. Sheng, J. & Webber, M. Incentive-compatible payments for watershed services along the eastern route of China’s southnorth water transfer project. Ecosyst. Serv. 25, 213–226. Smith, V. H. & Schindler, D. W. Eutrophication science: where do we go from here? Trends Ecol. Evol. 24, 201–207. Wang, C., Wang, Y. & Wang, P. Water quality modeling and pollution control for the eastern route of South-to-North Water Transfer Project in China. J. Hydrodyn. 18, 253–261. Wang, G., Fang, Q., Teng, Y. & Yu, J. Determination of the factors governing soil erodibility using hyperspectral visible and near-infrared reflectance spectroscopy. Int. J. Appl. Earth Obs. Geoinf. 53, 48–63. Wang, G., Hu, X., Zhu, Y., Jiang, H. & Wang, H. Historical accumulation and ecological risk assessment of heavy metals in sediments of a drinking water lake. Environ. Sci. Pollut. Res. 25, 24882–24894. Wang, G., Li, J., Sun, W., Xue, B., Yinglan, A. & Liu, T. a Nonpoint source pollution risks in a drinking water protection zone based on remote sensing data embedded within a nutrient budget model. Water Res. 157, 238–246. Wang, P., Yao, J., Wang, G., Hao, F., Shrestha, S., Xue, B., Xie, G. & Peng, Y. b Exploring the application of artificial intelligence technology for identification of water pollution characteristics and tracing the source of water quality pollutants. Sci. Total Environ. 693, 133440. Wang, H., Yan, H., Zhou, F., Li, B., Zhuang, W. & Shen, Y. Changes in nutrient transport from the Yangtze River to the East China Sea linked to the Three-Gorges Dam and water transfer project. Environ. Pollut. 256, 113376.
1076
J. He et al.
|
Using DID model to identify the impact of SNWTP policy on lake eutrophication
Wei, F. Methods of Monitoring and Analysis of Water and Wastewater. China Environmental Science, Beijing. Wu, B., Wang, G., Jiang, H., Wang, J. & Liu, C. Impact of revised thermal stability on pollutant transport time in a deep reservoir. J. Hydrol. 535, 671–687. Wu, B., Wang, G., Wang, Z., Liu, C. & Ma, J. Integrated hydrologic and hydrodynamic modeling to assess water exchange in a data-scarce reservoir. J. Hydrol. 555, 15–30. Wu, Y., Dai, R., Xu, Y., Han, J. & Li, P. Statistical assessment of water quality issues in Hongze Lake, China, related to the operation of a water diversion project. Sustainability 10, 1885. Xu, Z., Cai, X., Yin, X., Su, M., Wu, Y. & Yang, Z. Is water shortage risk decreased at the expense of deteriorating water quality in a large water supply reservoir? Water Res. 165, 114984. Xu, T., Yang, T. & Xiong, M. Time scales of external loading and spatial heterogeneity in nutrients-chlorophyll a response: implication on eutrophication control in a large shallow lake. Ecol. Eng. 142, 105636. Yang, H., Wang, G., Wang, L. & Zheng, B. Impact of land use changes on water quality in headwaters of the Three Gorges Reservoir. Environ. Sci. Pollut. Res. 23, 11448–11460. Yao, J., Wang, G., Xue, B., Wang, P., Hao, F., Xie, G. & Peng, Y. a Assessment of lake eutrophication using a novel multidimensional similarity cloud model. J. Environ. Manage. 248, 109259. Yao, J., Wang, G., Xue, W., Yao, Z. & Xue, B. b Assessing the adaptability of water resources system in Shandong Province, China, using a novel comprehensive co-evolution model. Water Resour. Manage. 33, 657–675. Yao, J., Wang, P., Wang, G., Shrestha, S., Xue, B. & Sun, W. Establishing a time series trend structure model to mine potential hydrological information from hydrometeorological time series data. Sci. Total Environ. 698, 134227.
Hydrology Research
|
51.5
|
2020
Ye, C., Shen, Z. M., Zhang, T., Fan, M. H., Lei, Y. M. & Zhang, J. D. Long-term joint effect of nutrients and temperature increase on algal growth in Lake Taihu, China. J. Environ. Sci (China) 23, 222–227. Yinglan, A., Wang, G., Liu, T., Shrestha, S., Xue, B. & Tan, Z. a Vertical variations of soil water and its controlling factors based on the structural equation model in a semi-arid grassland. Sci. Total Environ. 691, 1016–1026. Yinglan, A., Wang, G., Liu, T., Xue, B. & Kuczera, G. b Spatial variation of correlations between vertical soil water and evapotranspiration and their controlling factors in a semiarid region. J. Hydrol. 574, 53–63. Zeng, Q., Qin, L. & Li, X. The potential impact of an interbasin water transfer project on nutrients (nitrogen and phosphorous) and chlorophyll a of the receiving water system. Sci. Total Environ. 536, 675–686. Zhang, Q. The South-to-North Water Transfer Project of China: environmental implications and monitoring strategy. J. Am. Water Resour. Assoc. 45, 1238–1247. Zhang, J., Ni, W., Luo, Y., Stevenson, R. & Qi, J. Response of freshwater algae to water quality in Qinshan Lake within Taihu Watershed, China. Phys. Chem. Earth 36, 360–365. Zhou, H., Taber, C., Arcona, S. & Li, Y. Difference-indifferences method in comparative effectiveness research: utility with unbalanced groups. Appl. Health Econ. Health Policy 14, 419–429. Zhu, G. X. H., Zhu, M., Zou, W., Guo, C. J. P. D. W., Zhou, Y., Zhang, Y. & Qin, B. Changing characteristics and driving factors of trophic state of lakes in the middle and lower reaches of Yangtze River in the past 30 years. J. Lake Sci. 31, 1510–1524. Zhuang, W., Ying, S. C., Frie, A. L., Wang, Q., Song, J., Liu, Y., Chen, Q. & Lai, X. Distribution, pollution status, and source apportionment of trace metals in lake sediments under the influence of the South-to-North Water Transfer Project, China. Sci. Total Environ. 671, 108–118.
First received 16 April 2020; accepted in revised form 3 July 2020. Available online 31 July 2020
1077
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Succession of phytoplankton in a shallow lake under the alternating influence of runoff and reverse water transfer Qing Li, Guoqiang Wang
, Zhongxin Tan and Hongqi Wang
ABSTRACT Both runoff and water diversion can interfere with the physical and chemical environment of a lake and affect aquatic organisms. In this study, previously obtained data were used to analyze the phytoplankton community, water quality, water level, and temperature in Dongping Lake (DPH) before, during, and after the water diversion caused by the South-to-North Water Transfer Project. The results showed that the total density and diversity index of phytoplankton decreased in the water transfer period, and was related to low temperature. Temperature also affected the recovery of phytoplankton community structure when the water transfer period ended. In a water transfer
Qing Li (corresponding author) Guoqiang Wang Zhongxin Tan Hongqi Wang Beijing Key Laboratory of Urban Hydrological Cycle and Sponge City Technology, College of Water Sciences, Beijing Normal University, Beijing 100875, China E-mail: wanggq@bnu.edu.cn
cycle, changes in dominant genera were more drastic than that of a whole phytoplankton community, and dominant genera were sensitive to total phosphorus (TP) and total nitrogen (TN) changes. Water transfer alleviated the deterioration of water quality in DPH, but water transfer process increased the risk of water pollution. Runoff from Dawen River carried TN, TP, and chemical oxygen demand (COD) into DPH in the rainy season, which indirectly affected phytoplankton, while it also carried phytoplankton directly into DPH. Overall, these findings provide a clear understanding of the impact of water transfer projects on ecology in shallow lakes. Key words
| Dongping Lake, phytoplankton community, reverse water transfer, runoff, shallow lake, the South-to-North Water Transfer Project
HIGHLIGHTS
• • • •
Dominant genera change greater than genera in a whole phytoplankton community. Reverse water transfer and runoff affect phytoplankton community alternately. There is a risk of water quality deterioration during the water transfer process. Runoff transfers water, pollutants and phytoplankton to the lakes.
INTRODUCTION The influence of hydrology on the water ecosystem in shal-
change the original hydrologic conditions in lakes and dis-
low lakes is obvious. Both natural runoff and water
rupt natural, stable patterns. This affects the habitats of
diversion projects affect lakes by changing hydrology and
aquatic organisms and controls their community structure
water quality. In particular, the water transfer projects
and distribution. When hydrology changes, a series of water physical and chemical properties such as water
This is an Open Access article distributed under the terms of the Creative
level, flow velocity, and water quality are changed (Tuttle
Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying
et al. ; Liu et al. ). When hydrology is the driving
and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/
factor, the specific responses of the aquatic communities
licenses/by-nc-nd/4.0/).
in lakes are diverse and complex.
doi: 10.2166/nh.2020.163
1078
Q. Li et al.
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
|
51.5
|
2020
Most natural runoff is caused by rainfall. In areas signifi-
introduced by water transfer projects and water pollutants
cantly affected by the monsoon climate, natural runoff is
in water conveyance channels affect the water quality in
concentrated in the rainy season. During this period,
receiving lakes (Zhuang ; Liu et al. ; Zhuang et al.
runoff is characterized by a big flow amount and high flow
). Unlike natural runoff, man-made water transfer pro-
velocity that results in flooding or seasonal pulses (Rodger
jects have changed the natural pattern and broken the
et al. ; Galir Balkić et al. ; Han et al. ). However,
dynamic and stable influence of natural runoff on lakes.
rainfall-runoff during dry seasons is characterized by low
Although the effects of water diversion are complex, the
flow velocity and a small water amount. Natural runoff
reverse water diversion exacerbates this complexity. Accord-
showed the same pattern in every hydrological year in
ingly, it is necessary to study the ecology in shallow lakes
areas controlled by a temperate continental climate. For a
under the influence of reverse water transfer projects; how-
long time, runoff has increased the water amount and
ever, few studies have investigated reverse water transfer.
brought pollution (Wang et al. a, b), which con-
As the obviously seasonal succession, phytoplankton
stantly affects the water ecosystem in lakes (Silva et al.
community change rapidly in a short period with changes
; Sun et al. ). Runoff affects the phytoplankton suc-
in habitat environments. Accordingly, phytoplankton com-
cession by changing temperature and nutrients (Cao et al.
munity changes are commonly used to reflect the influence
). Increased runoff in the rainy season has also intro-
of external disturbances (Sharov ; Snit’ko & Snit’ko
duced pollutants, which favor the massive development of
; Toporowska et al. ; Weng et al. ). Community
Cyanobacteria in lakes (Rao et al. ; Silva et al. ;
structure, diversity, total density, and biomass are all
Sun et al. ) and also reduce the abundance of diatoms
important indicators of habitat environment changes
(Da Silva et al. ) or phytoplankton biomass (Cobbaert
(Deng et al. ; Özkan et al. ; Anneville et al. ).
et al. ). Flood pulses in rainy seasons have also been
In the shallow lakes, the water depth is small and the
shown to influence community structure by affecting food
water stratification not obvious, characteristics at different
and nutrient levels (Galir Balkić et al. ). In addition to
depths tend to be consistent, and the exchange of sedi-
the recognition that rainfall runoff increases nutrients in
ments and water is more frequent than in deep lakes
lakes, runoff also has the potential to dilute pollutants
(Scheffer ; Ongun Sevindik et al. ). Phytoplankton
(Cobbaert et al. ; Ho & Michalak ). These factors
in shallow lakes is more responsive to outside influences,
complicate the effects of runoff in lakes.
such as changes in temperature, hydrological conditions,
Increasingly, water transfer projects are being designed
and water quality. The effect of water quality on phyto-
to solve the problem of uneven distribution of water
plankton in lakes is obvious, total nitrogen (TN), total
resources. One of the main functions of these projects is to
phosphorus (TP), and ammonia nitrogen (NH4-N) are clo-
introduce exogenous water to a receiving lake in dry sea-
sely related to the phytoplankton in most lakes (Borics
sons. In areas in which natural precipitation is extremely
et al. ; Zhu et al. ; Tang et al. ). Temperature
scarce, water diversion has helped alleviate the decline of
also affects the phytoplankton succession, the increase in
water levels and deterioration of water quality, then affect-
temperature contributes to the increase in total phyto-
ing algal blooms in lakes (Amano et al. ; Li et al. ;
plankton biomass (Markensten et al. ). Also, Rao
Huang et al. ; Yao et al. ). In the dry season, water
et al. () thought that the effect of temperature on phyto-
diversion increases the water amount, raises the water
plankton was highly species-specific and temperature also
level, and changes the habitat environment, leading to a
modulated the effects of nutrients on phytoplankton. In
chain reaction in the community structure of aquatic organ-
addition, water level fluctuation also has an obvious influ-
isms (Guo et al. ). Under the influence of water transfer
ence on phytoplankton succession (Kivrak ; Tian
projects, the dominant species of the phytoplankton com-
et al. ; Qian et al. ; Rao et al. ). However,
munity has changed because of the changing water quality
under the influence of runoff and water transfer, the
(Amano et al. ) and hydrodynamic conditions in lakes
effect of temperature, hydrological conditions and water
(Li et al. ). What is more, both exogenous water
quality on phytoplankton should be rethought.
1079
Q. Li et al.
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
|
51.5
|
2020
Rainfall in northern China is mostly concentrated in
an important flood detention area in the lower reaches of
summer and autumn, while the region is cold and dry
the Yellow River, but is also the last regulating and storing
during spring and winter. Dongping Lake (DPH) is a typical
lake of the eastern route of the SNWTP. Accordingly, the
shallow lake located in northern China that mainly receives
protection of the ecosystem in the DPH is very important.
river runoff from the Dawen River (DWR) in the rainy
The surface area of DPH is about 125 km2 and the lake
season. In the dry season before the South-to-North Water
basin is narrow in the north and wide in the south (Figure 1).
Transfer Project (SNWTP), the water level in DPH fre-
DPH is a shallow lake with an annual average water depth
quently dropped because of reduced inflow, which
of about 2 m. The eastern part of the lake is mainly hilly,
seriously threatens the lake ecosystem (He et al. ).
while there are dams to the south and west. Few rivers
Implementation of the SNWTP alleviated the water loss
flow into DPH. The DWR flows from east to west into the
and eutrophication in DPH (Hu et al. ). However, the
lake, and then flows into the Yellow River. DPH mainly
SNWTP also introduced the possibility of adverse impacts
depends on surface runoff and lake surface precipitation
on the environment (Liu et al. ), and even caused bio-
supply. The region in which DPH is located is characterized
logical invasion (Qin et al. ). To date, the SNWTP has
by a cold and dry winter and hot and rainy summer, with
mainly transferred water into DPH in the dry season. There-
more than 70% of the annual precipitation falling during
fore, after implementation of the SNWTP, DPH mainly
the flood season. Therefore, the main inflow of the lake is
received runoff in the rainy season and transferred water
from the DWR in the rainy season. The eastern route of
in the dry season. The alternations of runoff and transferred
the SNWTP was completed in 2013. Liuchang River
water have complicated the effects on the ecology in DPH.
serves as the route from south to north to bring water
Accordingly, water transfer projects participate in the
from the Nansihu Lake (NSH) to DPH via Denglou station
effect of runoff on the aquatic organisms in lakes. What is
(DL). The water transfer period of the SNWTP is generally
more, it is important to understand how runoff and water
concentrated in winter and spring, when there is less rainfall
transfer under the continuous influence of the SNWTP
and low water depth in DPH. The water transfer period in
affect the ecology in DPH within a water transfer cycle,
DPH in 2018 was from December 2018 to June 2019,
but this has not yet been investigated. Therefore, we (1)
during which the inflow was mainly composed of the
investigated and analyzed the phytoplankton and water
inflow of the DWR and the transferred water from the
quality in DPH and the DWR before, during, and after the
upper lakes of NSH (UNSH) via the Liuchang River. There-
water transfer period and (2) analyzed the main factors
fore, the inflow during the water transfer period and non-
influencing phytoplankton and the impacts of runoff
water transfer period is different.
inflow
and
diversion
inflow
on
the
phytoplankton
community in DPH. Additionally, (3) the phytoplankton
Data description and sample collection
community structural changes and dominant taxa replacement of phytoplankton under the alternating influence of
Water diversion in DPH always occurs in the dry season, so
runoff and water diversion were discussed.
one year is considered as a water transfer cycle, including before the water transfer period (S1) (August 2018), during the water transfer period (S2) (March 2019), and after the
MATERIALS AND METHODS
water transfer period (S3) (July 2019). Water quality samples and phytoplankton samples were collected at four sites in
Study area
DPH and one site in the DWR. Water quality samples were collected at one site in the UNSH.
DPH, located in Shandong Province, China, is a major fresh-
Quantitative phytoplankton samples were collected
water lake with diverse aquatic life. This lake has many
using a 2.5 L water sampler from a depth of 0.5 m, after
functions including flood control, storage, irrigation, water
which, samples were fixed with Lugol’s solution and 40%
supply, and preservation of biodiversity. DPH is not only
formalin. Qualitative phytoplankton samples were collected
Q. Li et al.
1080
Figure 1
|
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
|
51.5
|
2020
The location of Dongping Lake (DPH). The arrows indicate the directions of the water flow, and the dotted arrow indicates the direction of the water transfer.
using a 25# floating net and then were kept at 4 C. All
laboratory, total nitrogen (TN), total phosphorus (TP),
samples were collected twice at each sampling site. After
ammonia nitrogen (NH4-N), and chemical oxygen demand
the samples were sent to the laboratory, the phytoplankton
(COD) were measured. Data describing the transferred
was identified and quantified. Upon arrival in the
water and water level were obtained from the local
Q. Li et al.
1081
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
|
51.5
|
2020
government, while rainfall and temperature data were
phytoplankton phylum. The calculation and visualization
downloaded from the China Meteorological Data Service
of RDA results are based on the R package ‘vegan’. Niche breadth models were used to analyze the change
Center.
in phytoplankton community with the environment. Shannon–Wiener index (Equation (3)) is often used to calculate
Materials and methods
niche breadth (Heino & Tolonen ; Li et al. ), and its significance is different to Equation (1):
Calculation of phytoplankton community diversity We calculated the diversity to evaluate changes in phyto-
H0i ¼
) to evaluate the diversity of the phytoplankton community:
(Pij log2 Pij )
(3)
j¼1
plankton communities during different periods. Here, we used the Shannon–Wiener index (Equation (1)) (Shannon
R X
where H0i is the niche breadth index of the i-th genus; R is the total number of sampling sites; Pij is the ratio of the numbers of the i-th species at the j-th sampling site to the total number of the i-th genus.
H¼
N X
(Pi log2 Pi )
(1) Identification of dominant taxa
i¼1
where H is the Shannon–Wiener index; Pi is the density
Dominant genera in the phytoplankton were screened to
ratio of the i-th genus; and N is the total number of the
analyze community changes. The dominant genera were
genera.
determined according to the dominance. The genus in the
We used the Jaccard similarity coefficient (Equation (2)) (Jaccard ) to obtain the community similarity:
top 25% of dominance was selected as the dominant genus. The formula used to calculate dominance was as follows:
j C¼ aþb j
(2)
PM Yi ¼
j¼1
N
nij
× fi
(4)
where C is the Jaccard similarity coefficient of the community; j is the number of the common genera in different
where Yi is the dominance of the i-th genus; nij is the density
communities; a and b are the number of genera in particular
of the i-th genus at the j-th site; M is the site number; N is the
communities.
total density of phytoplankton; and fi is the occurrence frequency of the i-th genus.
Analysis of key driving factors Redundancy analysis (RDA) was selected to evaluate the
RESULTS
relationship between environmental factors and phytoplankton communities. RDA is an ordination method that
Variations in the phytoplankton community in a water
combines regression analysis with principle component
transfer cycle
analysis, which is widely used to identify influencing factors. Detrended correspondence analysis (DCA) is required before RDA can be applied. If the first axis value of the gra-
Changes in phytoplankton community structure during different periods
dient length is less than 3.0, the RDA method is appropriate. Here, environmental factors include TN, TP, NH4-N,
In the whole water transfer cycle, eight phyla and 59
COD, and temperature in water. The biomes includes
genera, including Cyanophyta, Bacillariophyta, Chrysophyta,
Q. Li et al.
1082
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
Table 1
|
|
51.5
|
2020
The composition difference of the phytoplankton in a water transfer cycle
Scenarios
Genus number
Similarity coefficient
Dominant genus number
Similarity coefficient
S1 ∩ S2 ∩ S3
9
0.11
0
0
S1 ∩ S2
10
0.20
0
0
S2 ∩ S3
13
0.33
1
0.04
S1 ∩ S3
24
0.50
3
0.14
DWR ∩ S2
9
0.39
DWR ∩ S3
18
0.45
similarities and differences revealed that a total of nine genera appeared in three periods at the same time with a similarity coefficient of 0.11. Moreover, the community similarity coefficient was 0.20 in S1 and S2, 0.33 in S2 and S3, and 0.50 in S1 and S3. Hence, the community similarity between two non-water transfer periods was higher than that between the water transfer period and the non-water Figure 2
|
Variations in phytoplankton community structure in a water transfer cycle. S1, S2, and S3 refer to before, during, and after the water transfer period,
transfer period.
respectively.
Pyrrophyta, Cryptophyta, Euglenophyta, Chlorophyta, and
Variations in phytoplankton diversity during different periods
Xanthophyta, were identified in DPH (Figure 2). Community structure in S1, S2, and S3 showed that Chlorophyta,
To further understand the changes in different phytoplank-
Cyanophyta, and Bacillariophyta were the dominant phyla,
ton taxa in a water transfer cycle, we analyzed the
while Cryptophyta and Euglenophyta were not identified
changes in qualitative structure of phytoplankton. As
during the water transfer period and Xanthophyta was not
shown in Figure 3, the Shannon–Wiener index of phyto-
identified during two non-water transfer periods (S1 and
plankton was significantly different between the water
S3). Total density of phytoplankton decreased rapidly
transfer period and the non-water transfer period. During
during the water transfer period, and then recovered slowly after the water transfer period. Cyanophyta showed the same pattern as the total density. Cyanophyta dominated the phytoplankton community before the water transfer period, decreased during the water transfer period, and recovered slowly after the water transfer period. The proportion of Chlorophyta increased rapidly during the water transfer period, then decreased after the water transfer period. The proportion of Chrysophyta also increased during the water transfer period, while the proportion of Bacillariophyta was stable in different periods and did not change significantly, indicating that this phylum had good adaptability to changing environments. Additionally, 40, 21, and 32 genera of phytoplankton were identified during the three periods, respectively (Table 1). Analysis of the
Figure 3
|
The diversity index of the phytoplankton in a water transfer cycle. The white boxes are Shannon–Wiener indexes, the grey boxes are richness values.
Q. Li et al.
1083
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
|
51.5
|
2020
the water transfer period, the Shannon–Wiener index was
with the highest dominance in S1 was Phormidium. Three
relatively low, and the richness value was also low; this indi-
dominant genus in S1 and S2 were the same, Cyclotella,
cated that low phytoplankton diversity was probably due to
Dactylococcopsis, and Oscillatoria. Additionally, the domi-
low richness. Moreover, the Shannon–Wiener index differed
nance value of Oscillatoria in S1 and S3 was different, and
greatly among sites during the water transfer period. Specifi-
it presented the smallest dominance value in S3. No
cally, the Shannon–Wiener index in the central area (D04)
repeat dominant genus was found in S1 and S2.
(1.61) and the northern part of DPH (D01) (1.41) was smaller than at other sites, while the Shannon–Wiener index
Environmental factors affecting phytoplankton
near the entrance of the DWR (D05) (2.74) and at the
communities
southern part of DPH (D07) (2.95) was larger than at other sites. During two non-transfer periods, the Shannon–
Variations in environmental and hydrological factors
Wiener index of genus was relatively high and consistent at each site. Overall, the density of the phytoplankton com-
The water level in DPH fluctuated throughout the water
munity in the study area was relatively uniform before and
transfer cycle (Figure 5), being generally low in S1 and
after the water transfer period. After the water transfer
increasing in S2. In 2018, Shandong Province experienced
period, phytoplankton diversity gradually recovered, but
the heaviest rainfall in years. During that year, rainfall in
was not the same as the original state, which might also
the rainy season was much heavier than that in the historical
be related to the insufficient sampling interval between S2
period (570 mm), and the water storage volume in DPH was
and S3.
63% higher than usual. Therefore, the rainfall runoff replenishment of DPH resulted in the overall elevation of the
Dominant taxa during different periods
water level in the rainy season. The highest water level in S2 came as a result of the heavy rainfall. DPH was dry in
Ten, five, and nine dominant genera were screened in S1,
the dry season. The rainfall in post-flood season of 2018
S2, and S3, respectively. Based on the distribution of the
and pre-flood season of 2019 was lower than the post-
dominant genera in the three different periods, the domi-
flood (110 mm) and pre-flood (257 mm) seasons in the
nance of each genus was not significantly different, but
same historical period, and the water level dropped accord-
none of the dominant genera were the same (Figure 4).
ingly (Figure 6). At this time, the transferred water flowed
There was only one genus that was dominant in both S2
into DPH, which alleviated the water level decline. Water
and S3, namely, Chlorella, which was the dominant genus
demand around DPH changes with the seasons. The drop
with the highest dominance value in S2 and S3. The genus
of water level after March was due to an increase in agricultural activities; accordingly, the transferred water amount is adjusted to keep the water level in DPH stable. Water quality is one of the major factors affecting aquatic communities. In a whole water transfer cycle, water quality differed in different periods, and different water quality indicators also showed different changes. After a water transfer cycle, TP and NH4-N in DPH decreased, while COD and TN increased. Additionally, TN rose rapidly in S2 and fell in S3, which returned to close to the original level. COD decreased slightly in S2, but showed a higher value in S3. These changes indicated that the water quality fluctuated within a water transfer cycle, but that it returned to close to its original level when the
Figure 4
|
The dominant genera in different periods. The codes are the same as Table 2.
water transfer was completed. However, the water quality
Q. Li et al.
1084
Figure 5
|
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
|
51.5
|
2020
The dynamics of water level and temperature in DPH during a water transfer cycle.
did not return to the original level after water diversion
niche breadth index was largest on the COD gradient and
when it was not in the rainy season. It is possible that the
smallest on the TP gradient, indicating that the change of
effects of precipitation on the improvement of water quality
COD in DPH had no significant impact on the dominant
in the lake were not reflected yet in the collected samples
genera, while the change in TP had a greater impact on
during the third sampling.
the
Key factors driving phytoplankton distribution
mental gradient, indicating that they were well adapted to
dominant
genera.
Phormidium,
Merismopedia,
Chlorella, and Cyclotella ranked high on each environthe changing environment. Merismopedia and Cyclotella To better understand the environmental impact on phyto-
were present throughout the water transfer cycle, even
plankton, it is necessary to identify the main impact
though they were not dominant genera in every period.
factors. According to the RDA results (Figure 7), water qual-
Chlorella appeared in S2 and S3, while Phormidium
ity had great influence on phytoplankton community in
appeared in S1 and S3.
phylum level. Especially, the COD and water temperature (T water) in DPH greatly affected the phytoplankton. COD was related to the Bacillaroiphyta. T water was positively
DISCUSSION
correlated with both Cryptophyta and Pyrrophyta, but negatively correlated with Xanthophyta. TN was positively
Water transfer and river inflow jointly affect water
correlated with Xanthophyta, but negatively correlated
quality and water level during the water transfer period
with Cyanophyta, while NH4-N was closely correlated with Cyanophyta. To further analyze the dominant genera
Water transfer is always implemented in the dry season,
changes along environmental gradient, niche breadth
when there is little rainfall and limited runoff flow into
values were calculated. As shown in Table 2, the mean
DPH. DPH mainly accepts transferred water and little
Q. Li et al.
1085
Table 2
|
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
|
51.5
|
2020
Niche breadth index of the dominant genera along environmental gradients
Code
Dominant genera
TP
COD
TN
NH4-N
RNP
Mean
Rank
1
Phormidium
1.4
1.05
1.05
1.87
1.04
1.282
1
2
Merismopedia
1.22
0.97
1.05
1.58
0.87
1.138
3
3
Scenedesmus
0.86
0.54
0.55
1
0.53
0.696
5
4
Westella
0.41
0.44
0.42
0.77
0.46
0.5
8
5
Cyclotella
0.6
1.24
0.8
0.83
0.69
0.832
4
6
Dactylococcopsis
0.51
0.73
0.43
0.61
0.48
0.552
6
7
Oscillatoria
0.49
0.79
0.45
0.51
0.39
0.526
7
8
Cryptomonas
0.38
0.24
0.18
0.45
0.07
0.264
14
9
Melosira
0.2
0.17
0.16
0.15
0.1
0.156
20
10
Fragilaria
0.3
0.15
0.22
0.27
0.07
0.202
18
11
Dinobryon
0.35
0.31
0.98
0.3
0.33
0.454
9
12
Tribonema
0.15
0.19
0.94
0.17
0.3
0.35
11
13
Ankistrodesmus
0.08
0.14
0.7
0.13
0.21
0.252
15
14
Synedra
0.06
0.1
0.43
0.09
0.51
0.238
17
15
Chlorella
0.45
1.63
1.48
0.7
1.62
1.176
2
16
Micryocystis
0.09
0.91
0.03
0.23
0.47
0.346
12
17
Chroococcus
0.06
0.65
0.03
0.14
0.34
0.244
16
18
Eudorina
0.06
0.52
0.01
0.13
0.26
0.196
19
19
Nitzschia
0.1
0.89
0.07
0.21
0.5
0.354
10
20
Pandorina
0.03
0.77
0.05
0.18
0.48
0.302
13
Mean
0.39
0.621
0.501
0.516
0.486
Rank
5
1
3
2
4
The table shows the sensitivity of each dominant genus to environmental indicators, and the bold values represent the average value. The larger the value was (Mean), the higher the ranking was (Rank), indicating that the species was less sensitive to the corresponding environmental indicators. This provided important support for the discussion of the relationship between phytoplankton and water quality indicators in this paper. Bold is for better presentation of important data.
inflows in the dry season. According to the section ‘Environ-
TN content in DWR. Although NH4-N content in DWR
mental factors affecting phytoplankton communities’, the
was higher than in DPH, it was not high enough to affect
phytoplankton community structure was closely related to
the NH4-N in DPH considering the DWR inflow amount.
water quality factors, and the influence of water diversion
More importantly, a large amount of transferred water
on water quality in DPH was obvious. Compared with the
with low NH4-N content also reduced the influence of
results before the water transfer period, TN increased,
NH4-N in DWR inflow on DPH. COD and TP showed
NH4-N decreased during the water transfer period. Based
little change during the water transfer period, because the
on the water quality changes in DPH (Figure 6), TN content
differences in COD and TP content between transferred
in the transferred water from UNSH was about 0.5 times
water and DPH, DWR and DPH were small. It is worth
that of DPH and in the inflow from DWR was almost
noting that, as the source of transferred water for DPH,
three times that of DPH. Even if the inflow from DWR
the content of each water quality indicator in UNSH was
was too small to be measured during the water transfer
less than that in DPH. However, water quality had already
period (Figure 5), it was still inevitable that a quite high
deteriorated at Denglou station. Increase in TN content
TN concentration entered the DPH through DWR inflow.
was evident in Denglou station, and the content of other
Thus, the increase in TN in DPH was mainly due to high
indicators, such as COD, NH4-N, TP, were also higher
Q. Li et al.
1086
Figure 6
|
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
|
51.5
|
2020
Variations in water quality factors in different periods. The black arrows refer to the flow direction. (a) The water quality in DPH and (b) the water quality of inflows of DPH.
than that in DPH. Liangji Canal and the Liuchang River
Changes in phytoplankton community structure and
serve as water conveyance channels, but are surrounded
replacement of dominant taxa
by residential communities and farmland. Moreover, spring is the primary season for wheat growth in northern
Changes in habitat will lead to changes in community struc-
China, so fertilization, fishing, and other agricultural activi-
ture, such as the replacement of dominant taxa. Cyanophyta
ties likely influence the TN content in DPH. Thus, human
played a dominant role in the community before and after
activities along the diversion route might have polluted the
water diversion. As the dominant phylum, Cyanophyta
transferred water, thus affecting the phytoplankton in
occupied a large percentage in the whole water transfer
DPH. Therefore, the water diversion and DWR jointly
cycle, and their distribution was closely related to NH4-N
affect the water quality in DPH. Water diversion not only
and TN (Figure 7). Cyanophyta decreased rapidly during
affected water quality, but also maintained the water level
the water transfer period, but gradually recovered after
in DPH directly. To ensure sufficient water volume in the
water diversion, which was related to the decrease of
next stage, the transferred water in DPH must be large
NH4-N during the water transfer period (Figure 6). In
enough to raise the water level (Figure 5). Agricultural activi-
addition to water quality change, temperature differences
ties in the spring also greatly increase the water demand
caused by seasonal changes also influenced the dramatic
during the water transfer period, and this reduces the
changes in Cyanophyta. Increases in temperature have pre-
amount of water in the lake. These frequent water level fluc-
viously been shown to be conducive to the dominance of
tuations promote material circulation in the lake and are
Cyanophyta (Markensten et al. ; Gray et al. ).
detrimental to the growth of phytoplankton (Kivrak );
Hence, Cyanophyta dominated the phytoplankton in the
therefore, they likely also affected the community structure
warm non-transfer periods, especially in S1, which had the
in DPH to some extent. Therefore, these water quality and
highest temperature among three samplings. The diversity
water level changes caused by water diversion and DWR
and the total density of phytoplankton also decreased in
inflow drove the phytoplankton succession in DPH.
S2. Dai et al. () also believed that the diversity of
Q. Li et al.
1087
Figure 7
|
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
|
51.5
|
2020
RDA of environmental factors and phytoplankton community during the whole water transfer cycle.
phytoplankton in the receiving water was lower during the
disappeared as the environment changed, but the taxa
water transfer period than the non-water transfer period,
with better adaptability to the environment were still at a
which was consistent with our results. The decrease in the
relatively stable level. When compared with the research
diversity and the total density might be related to low temp-
results before the SNWTP, the phytoplankton community
erature (Ho & Michalak ). Low temperature narrows
structure and Shannon–Wiener index in DPH changed
the niche breadth of genus with different niche require-
greatly during the same periods (Tian et al. ). Further
ments, leading to a decrease in the genera number in S2
analysis results of quantitative structure showed that the
and thus a decrease in the diversity index and richness
genera among three periods had a similarity coefficient of
(Yang et al. ). Even though the phytoplankton commu-
0.10, while no dominant genus was the same (Table 1).
nity structure and total density showed a recovery trend
Additionally, as shown in Figure 4 and Table 1, the three
after the water transfer period, there were still some differ-
same dominant genera appeared before and after the
ences.
seasonal
water transfer period, while only one dominant genus was
changes might be one of the main reasons for the total phy-
the same during and after the water transfer period. More-
toplankton density difference between S1 and S2. Water
over, the similarity coefficients of genera (0.33 and 0.50)
temperature is known to be closely related to the growth
were larger than that of dominant genera (0.04 and 0.14)
of phytoplankton (Markensten et al. ; Bergstrom et al.
during the same periods. Four in nine genera that recurred
), and the growth of some algae is known to have an
during these three periods belonged to Bacillariophyta,
obvious circadian rhythm (Straub et al. ). Before and
which maintained a stable percentage throughout the
after water diversion, the air temperature in DPH was simi-
whole water transfer cycle (Figure 2). Therefore, in a water
lar, while the water temperature in S1 was higher than in S3
transfer cycle, the change in dominant genus was more dras-
(Figure 5). Thus, the total phytoplankton density remained
tic than that of a whole phytoplankton community. The
at a low value after the water diversion.
sensitivity of each dominant genus to the environment was
Temperature
differences
caused
by
Compared with the results before the water transfer
further analyzed by the niche breadth model. In general,
period (S1), the phytoplankton genus number decreased
the species with narrower niche breadth are more environ-
(Table 1) and phytoplankton diversity increased at the
mentally sensitive. Results in Table 2 showed that the
genus level (Figure 3) after the water transfer period (S3).
dominant genera were sensitive to TP and the N/P ratio.
These
findings
indicated
that
some
sensitive
taxa
Most of the screened dominant genera belonged to the
1088
Q. Li et al.
|
Succession of phytoplankton under the influence of water transfer
Hydrology Research
|
51.5
|
2020
Cyanophyta and Chlorophyta, which were closely related to
with runoff in the rainy season, directly affecting the phyto-
the TN and TP. Thus, the TN and TP changes affected the
plankton in DPH. These findings confirmed previous
dominant genera of phytoplankton.
findings that hydrological connectivity increased community similarity (Yuan et al. ).
Runoff affects phytoplankton by affecting water quality during non-transfer periods Water quality in DPH gradually recovered after the fluctuation during the water transfer period (Figure 6). Similar water quality conditions led to similar phytoplankton structures before and after water diversion, as well as high genera repetition rates. The water quality in DPH during the non-water transfer periods was mainly influenced by runoff inflow. In the rainy season in northern China, DPH mainly received the runoff supply from DWR. After the water transfer period, the COD in DPH rose rapidly, which was mainly caused by the high COD in the DWR, because of the higher COD content in DWR than in DPH (Figure 6). Even though samples collected after the water transfer were not obtained in the rainy season, the rainfall that occurred carried enough COD into DPH to increase the COD content. TN gradually decreased after the water transfer period, which might also be related to the runoff from the DWR. When compared with the water transfer period, TN content in DWR after the water diversion decreased greatly, but it was still higher than that in DPH. Rainfall-runoff caused the DWR to carry TN into DPH continuously, thus affecting water quality. NH4-N changed a little after the water transfer period because of the similar content in DPH and DWR. TP concentration in DPH was already very low and could be easily affected. Thus, the slight decrease in TP might be caused by the low TP content in DWR after the water transfer period (S3). The third sampling (S3) occurred soon after the water transfer period, so the dilution effect of the rainfall on pollutants was not reflected. This explained why the water
CONCLUSIONS Implementation of the SNWTP led to alternations between water diversion inflow and runoff in DPH, but phytoplankton
succession
was
still
not
clear.
In
this
study,
phytoplankton community before, during, and after the water transfer period in a water transfer cycle were analyzed, and various effects of water transfer and runoff on phytoplankton were discussed in combination with the corresponding changes in water quality, temperature, and water level. Low temperature caused Chlorophyta to replace Cyanophyta as the main dominant phylum, and low temperature was also related to the decrease in the total density and diversity index of phytoplankton during the water transfer period. Water temperature affected the recovery of total phytoplankton density after the water transfer period. In a water transfer cycle, changes in dominant genus were more drastic than that of a whole phytoplankton community, and dominant genera were sensitive to TP and TN changes. Transferred water and DWR inflow jointly affected the water quality in DPH during the water transfer period and water quality in DPH was improved. DWR was responsible for the increase in TN content. At the same time, the water transfer process increased the risk of water pollution. Runoff from DWR affected the water quality in DPH during the non-transfer periods. In addition, phytoplankton in DWR entered DPH through runoff, which affected the phytoplankton community directly.
quality indicators did not return to the level before the water transfer period. Moreover, similarity coefficient of the phytoplankton community in DWR and DPH after the
ACKNOWLEDGEMENTS
water transfer period (S3) (0.45) was bigger than that during the water transfer period (S2) (0.39), and the total
This study was supported by the National Key R&D
number of phytoplankton genera in DPH and DWR in the
Program of China (Grant No. 2016YFC0401308) and the
non-transfer period were much higher than that in the
Chinese
water transfer period (Table 2). This indicated that DWR
Program of Water Pollution Control and Treatment (Grant
brought a large amount of phytoplankton into DPH along
No. 2017ZX07302004).
National
Special
Science
and
Technology
1089
Q. Li et al.
|
Succession of phytoplankton under the influence of water transfer
DATA AVAILABILITY STATEMENT All relevant data are included in the paper or its Supplementary Information.
REFERENCES Amano, Y., Sakai, Y., Sekiya, T., Takeya, K., Taki, K. & Machida, M. Effect of phosphorus fluctuation caused by river water dilution in eutrophic lake on competition between blue-green alga Microcystis aeruginosa and diatom Cyclotella sp. J. Environ. Sci. (China) 22, 1666–1673. Anneville, O., Chang, C. W., Dur, G., Souissi, S., Rimet, F. & Hsieh, C. h. The paradox of re-oligotrophication: the role of bottom–up versus top–down controls on the phytoplankton community. Oikos 128, 1666–1677. Bergstrom, A. K., Faithfull, C., Karlsson, D. & Karlsson, J. Nitrogen deposition and warming – effects on phytoplankton nutrient limitation in subarctic lakes. Global Change Biol. 19, 2557–2568. Borics, G., Nagy, L., Miron, S., Grigorszky, I., László-Nagy, Z., Lukács, B. A., G-Tóth, L. & Várbíró, G. Which factors affect phytoplankton biomass in shallow eutrophic lakes? Hydrobiologia 714, 93–104. Cao, J., Hou, Z., Li, Z., Chu, Z., Yang, P. & Zheng, B. Succession of phytoplankton functional groups and their driving factors in a subtropical plateau lake. Sci. Total Environ. 631–632, 1127–1137. Cobbaert, D., Wong, A. & Bayley, S. E. Precipitation-induced alternative regime switches in shallow lakes of the Boreal Plains (Alberta, Canada). Ecosystems 17, 535–549. Dai, J., Wu, S., Wu, X., Xue, W., Yang, Q., Zhu, S., Wang, F. & Chen, D. Effects of Water diversion from Yangtze River to Lake Taihu on the phytoplankton habitat of the Wangyu River channel. Water 10, 759–769. Da Silva, C. F. M., Torgan, L. C. & Schneck, F. Temperature and surface runoff affect the community of periphytic diatoms and have distinct effects on functional groups: evidence of a mesocosms experiment. Hydrobiologia 839, 37–50. Deng, J., Qin, B., Sarvala, J., Salmaso, N., Zhu, G., Ventelä, A.-M., Zhang, Y., Gao, G., Nurminen, L., Kirkkala, T., Tarvainen, M. & Vuorio, K. Phytoplankton assemblages respond differently to climate warming and eutrophication: a case study from Pyhäjärvi and Taihu. J. Great Lakes Res. 42, 386–396. Galir Balkić, A., Ternjej, I. & Špoljar, M. Hydrology driven changes in the rotifer trophic structure and implications for food web interactions. Ecohydrology 11, e1917–e1919. Gray, E., Elliott, J. A., Mackay, E. B., Folkard, A. M., Keenan, P. O. & Jones, I. D. Modelling lake cyanobacterial blooms Disentangling the climate-driven impacts of changing mixed depth and water temperature. Freshwat. Biol. 64, 2141–2155.
Hydrology Research
|
51.5
|
2020
Guo, C., Chen, Y., Liu, H., Lu, Y., Qu, X., Yuan, H., Lek, S. & Xie, S. Modelling fish communities in relation to water quality in the impounded lakes of China’s South-to-North Water Diversion Project. Ecol. Model. 397, 25–35. Han, D., Wang, G., Liu, T., Xue, B.-L., Kuczera, G. & Xu, X. Hydroclimatic response of evapotranspiration partitioning to prolonged droughts in semiarid grassland. J. Hydrol. 563, 766–777. He, S., Li, R. & Li, B. Dongping lake water regulation scheme of the East Route of South-to-North Water Transfer Project (in Chinese). Yellow River 36, 52–54. Heino, J. & Tolonen, K. T. Ecological niche features override biological traits and taxonomic relatedness as predictors of occupancy and abundance in lake littoral macroinvertebrates. Ecography 41, 2092–2103. Ho, J. C. & Michalak, A. M. Exploring temperature and precipitation impacts on harmful algal blooms across continental U.S. lakes. Limnol. Oceanogr. 9999, 1–18. Hu, Z., Kang, F., Zou, A., Yu, L., Li, Y., Tian, T. & Guiling, K. Evolution trend of the water quality in Dongping Lake after South-North Water Transfer Project in China. J. Groundw. Sci. Eng. 7, 333–339. Huang, J., Gao, J., Zhang, Y. & Xu, Y. Modeling impacts of water transfers on alleviation of phytoplankton aggregation in Lake Taihu. J. Hydroinform. 17, 149–162. Jaccard, P. Contribution au problème de l’immigration postglaciaire de la flore alpine: étude comparative de la flore alpine du massif de Wildhorn, du haut bassin du Trient et de la haute vallée de Bagnes (Contribution to post-glacial alpine flora migration: a comparative study of alpine flora in the Wilderhorn Range, the Upper Trent Basin and the Upper Barness Valley). Bull. Soc. Vaud. Sci. Nat. 36, 87–130. Kivrak, E. Seasonal and long term changes of the phytoplankton in the lake Tortum in relation to environmental factors, Erzurum, Turkey. Biologia 61, 339–345. Li, Y., Tang, C., Wang, C., Anim, D., Yu, Z. & Acharya, K. Improved Yangtze River diversions: are they helping to solve algal bloom problems in Lake Taihu, China? Ecol. Eng. 51, 104–116. Li, Q., Wang, G., Wang, H., Shrestha, S., Xue, B., Sun, W. & Yu, J. Macrozoobenthos variations in shallow connected lakes under the influence of intense hydrologic pulse changes. J. Hydrol. 584, 124755–124766. Liu, J., Rühland, K. M., Chen, J., Xu, Y., Chen, S., Chen, Q., Huang, W., Xu, Q., Chen, F. & Smol, J. P. Aerosol-weakened summer monsoons decrease lake fertilization on the Chinese Loess Plateau. Nat. Clim. Chang. 7, 190–194. Liu, J., Li, M., Wu, M., Luan, X., Wang, W. & Yu, Z. Influences of the south–to-north water diversion project and virtual water flows on regional water resources considering both water quantity and quality. J. Clean Prod. 244, 118920–118927. Markensten, H., Moore, K. & Persson, I. Simulated lake phytoplankton composition shifts toward cyanobacteria dominance in a future warmer climate. Ecol. Appl. 20, 752–767.
1090
Q. Li et al.
|
Succession of phytoplankton under the influence of water transfer
Ongun Sevindik, T., Tunca, H., Gönülol, A., Gürsoy, N., Küçükkaya, Ş . N. & Kinali, Z. Phytoplankton dynamics and structure, and ecological status estimation by the Q assemblage index: a comparative analysis in two shallow Mediterranean lakes. Turk. J. Bot. 41, 25–36. Özkan, K., Jeppesen, E., Davidson, T., Bjerring, R., Johansson, L., Søndergaard, M., Lauridsen, T. & Svenning, J.-C. Long-term trends and temporal synchrony in plankton richness, diversity and biomass driven by re-oligotrophication and climate across 17 Danish lakes. Water 8, 427–450. Qian, K., Liu, X. & Chen, Y. Effects of water level fluctuation on phytoplankton succession in Poyang Lake, China – a five year study. Ecohydrol. Hydrobiol. 16, 175–184. Qin, J., Cheng, F., Zhang, L., Schmidt, B., Liu, J. & Xie, S. Invasions of two estuarine gobiid species interactively induced from water diversion and saltwater intrusion. Manag. Biol. Invasion 10, 139–150. Rao, K., Zhang, X., Yi, X. J., Li, Z. S., Wang, P., Huang, G. W. & Guo, X. X. Interactive effects of environmental factors on phytoplankton communities and benthic nutrient interactions in a shallow lake and adjoining rivers in China. Sci. Total Environ. 619–620, 1661–1672. Rodger, A. W., Mayes, K. B. & Winemiller, K. O. Preliminary findings for a relationship between instream flow and shoal chub recruitment in the Lower Brazos River, Texas. Trans. Am. Fish. Soc. 145, 943–950. Scheffer, M. Ecology of shallow lakes. Popul. Commun. Biol. 22, 76–121. Shannon, C. E. A mathematical theory of communication. Bcll Syst. Tech. J. 27, 379–423. Sharov, A. N. Phytoplankton as an indicator in estimating long-term changes in the water quality of large lakes. Water Resour. 35, 668–673. Silva, T. F. G., Vinçon-Leite, B., Lemaire, B. J., Petrucci, G., Giani, A., Figueredo, C. C. & Nascimento, N. d. O. Impact of urban stormwater runoff on Cyanobacteria dynamics in a tropical urban lake. Water 11, 946–974. Snit’ko, L. V. & Snit’ko, V. P. Phytoplankton as an indicator in assessing long-term variations in water quality of lakes Bolshoye Miassovo and Turgoyak, the South Urals. Water Resour. 41, 210–217. Straub, C., Quillardet, P., Vergalli, J., de Marsac, N. T. & Humbert, J. F. A day in the life of microcystis aeruginosa strain PCC 7806 as revealed by a transcriptomic analysis. PLoS ONE 6, e16208–e16220. Sun, W., Wang, Y., Fu, Y. H., Xue, B., Wang, G., Yu, J., Zuo, D. & Xu, Z. Spatial heterogeneity of changes in vegetation growth and their driving forces based on satellite observations of the Yarlung Zangbo River Basin in the Tibetan Plateau. J. Hydrol. 574, 324–332. Tang, C., Yi, Y., Yang, Z., Zhou, Y., Zerizghi, T., Wang, X., Cui, X. & Duan, P. Planktonic indicators of trophic states for a
Hydrology Research
|
51.5
|
2020
shallow lake (Baiyangdian Lake, China). Limnologica 78, 125712–125723. Tian, C., Lu, X., Pei, H., Hu, W. & Xie, J. Seasonal dynamics of phytoplankton and its relationship with the environmental factors in Dongping Lake, China. Environ. Monit. Assess. 185, 2627–2645. Tian, C., Pei, H., Hu, W., Hao, D., Doblin, M. A., Ren, Y., Wei, J. & Feng, Y. Variation of phytoplankton functional groups modulated by hydraulic controls in Hongze Lake, China. Environ. Sci. Pollut. Res. 22, 18163–18175. Toporowska, M., Ferencz, B. & Dawidek, J. Impact of lakecatchment processes on phytoplankton community structure in temperate shallow lakes. Ecohydrology 11, e2017–e2019. Tuttle, C. L., Zhang, L. & Mitsch, W. J. Aquatic metabolism as an indicator of the ecological effects of hydrologic pulsing in flow-through wetlands. Ecol. Indicators 8, 795–806. Wang, G., Li, J., Sun, W., Xue, B., Yinglan, A. & Liu, T. a Nonpoint source pollution risks in a drinking water protection zone based on remote sensing data embedded within a nutrient budget model. Water Res. 157, 238–246. Wang, P., Yao, J., Wang, G., Hao, F., Shrestha, S., Xue, B. L., Xie, G. & Peng, Y. b Exploring the application of artificial intelligence technology for identification of water pollution characteristics and tracing the source of water quality pollutants. Sci. Total Environ. 693, 133440–133455. Weng, B., Yang, Y., Yan, D., Wang, J., Dong, G., Wang, K., Qin, T. & Dorjsuren, B. Shift in plankton diversity and structure: influence of runoff composition in the Nagqu River on the Qinghai-Tibet Plateau. Ecol. Indicators 109, 105818–105825. Yang, Y., Stenger Kovács, C., Padisák, J. & Pettersson, K. Effects of winter severity on spring phytoplankton development in a temperate lake (Lake Erken, Sweden). Hydrobiologia 780, 47–57. Yao, X., Zhang, L., Zhang, Y., Du, Y., Jiang, X. & Li, M. Water diversion projects negatively impact lake metabolism: a case study in Lake Dazong, China. Sci. Total Environ. 613–614, 1460–1468. Yuan, Y., Jiang, M., Liu, X., Yu, H., Otte, M. L., Ma, C. & Her, Y. G. Environmental variables influencing phytoplankton communities in hydrologically connected aquatic habitats in the Lake Xingkai basin. Ecol. Indicators 91, 1–12. Zhu, R., Wang, H., Chen, J., Shen, H. & Deng, X. Use the predictive models to explore the key factors affecting phytoplankton succession in Lake Erhai, China. Environ. Sci. Pollut. Res. 25, 1283–1293. Zhuang, W. Eco-environmental impact of inter-basin water transfer projects: a review. Environ. Sci. Pollut. Res. 23, 12867–12879. Zhuang, W., Ying, S. C., Frie, A. L., Wang, Q., Song, J., Liu, Y., Chen, Q. & Lai, X. Distribution, pollution status, and source apportionment of trace metals in lake sediments under the influence of the South-to-North Water Transfer Project, China. Sci. Total Environ. 671, 108–118.
First received 12 May 2020; accepted in revised form 3 July 2020. Available online 10 August 2020
1091
© 2020 The Authors Hydrology Research
|
51.5
|
2020
A framework for event-based flood scaling analysis by hydrological modeling in data-scarce regions Jianzhu Li, Kun Lei, Ting Zhang, Wei Zhong, Aiqing Kang, Qiushuang Ma and Ping Feng
ABSTRACT Flood scaling theory is important for flood predictions in data-scarce regions but is often applied to quantile-based floods that have no physical mechanisms. In this study, we propose a framework for flood prediction in data-scarce regions by event-based flood scaling. After analyzing the factors controlling the flood scaling, flood events are first simulated by a hydrological model with different areally averaged rainfall events and curve number (CN) values as inputs, and the peak discharge of each subcatchment is obtained. Then, the flood scaling is analyzed according to the simulated peak discharge and subcatchment area. Accordingly, the relationship curves between the scaling exponent and the two explanatory factors (rainfall intensity and CN) can be drawn. Assuming that the flood and the corresponding rainfall event have the same frequency, the scaling exponent with a specific flood frequency can be interpolated from these curves. Key words
| data-scarce region, event-based flood scaling, flood prediction, hydrological modeling
HIGHLIGHTS
• • •
Hydrological modeling was used to conduct flood scaling analysis. Rainfall intensity and curve number were identified to be the two main factors affecting flood scaling. A framework of flood prediction by event-based flood scaling in data-scarce regions was proposed.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.042
Jianzhu Li Kun Lei Ting Zhang (corresponding author) Qiushuang Ma Ping Feng State Key Laboratory of Hydraulic Engineering Simulation and Safety, Tianjin University, Tianjin, China E-mail: zhangting_hydro@tju.edu.cn Wei Zhong School of Management, Tianjin University of Technology, Tianjin, China Aiqing Kang State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, China
1092
J. Li et al.
|
A framework of event-based flood scaling analysis
Hydrology Research
|
51.5
|
2020
GRAPHICAL ABSTRACT
INTRODUCTION Flood scaling, which is extensively applied to flood predic-
has been tested in many river basins of different sizes from
tion in ungauged and poorly gauged river basins of
small-scale to mesoscale basins (Ogden & Dawdy ;
different sizes, is usually expressed by an equation that
Gupta et al. ; Yue & Gan ; Kroll et al. ;
relates peak discharge to catchment descriptors such as
Ayalew et al. ). Generally, the scaling exponent is less
the drainage area, slope of the watershed, land use, and rain-
than 1 in quantile-based flood scaling but might exceed 1
fall characteristics; among these descriptors, the drainage
in event-based flood scaling. Moreover, flood scaling ana-
area is the most widely used explanatory variable to predict
lyses are often conducted in homogeneous and nested
flood peak discharge (Jothityangkoon & Sivapalan ;
watersheds, in which the flood generation mechanisms are
Menabde & Sivapalan ; Galster et al. ; Al-Rawas
similar (Eash ; Ogden & Dawdy ; Furey & Gupta
& Valeo ; Ishak et al. ; Han et al. ; Farmer
, ). Accordingly, the function between the flood
et al. ; Furey et al. ; Lee & Huang ). The
quantile and watershed area can be established, and the
θ
power law function recommended worldwide is Q ¼ αA ,
scaling parameters can be obtained. However, the fitted
where Q is the peak discharge for a given drainage area A,
function may not have satisfactory precision; in this situ-
α is the scaling intercept, and θ is the scaling exponent
ation, other controlling factors, such as the corresponding
(Gupta et al. ; Medhi & Tripathi ). If a reference
quantile-based rainfall intensity, should be taken into
river basin with drainage area A1 has a flood peak discharge
consideration in addition to the drainage area.
Q1 and θ is determined, then the flood peak Q2 in an
Although the quantile-based scaling exponent is widely
ungauged basin with drainage area A2 can be estimated by
used for future flood prediction, the physical mechanism
the following equation: Q1/Q2 ¼ (A1/A2)θ.
cannot be interpreted because the peak discharges of
Most previous studies focused on the relation between
floods with the same return period in different catchments
the quantile-based peak discharge and drainage area. The
may not be generated by the same rainfall events
peak discharge shows simple scaling or multiscaling
(Liu et al. ). Therefore, Gupta et al. () proposed per-
for snowmelt-generated or rainfall-generated floods, which
forming flood scaling research by using event-based floods
means that the scaling parameters (especially the scaling
for statistical analysis. As such, observed floods that occur
exponent) remain invariant or vary, respectively, with the
simultaneously in different catchments can be selected,
flood return period (Gupta & Dawdy ). This hypothesis
and the spatial and temporal distributions of rainfall can
1093
J. Li et al.
|
A framework of event-based flood scaling analysis
Hydrology Research
|
51.5
|
2020
be considered. Furey & Gupta () presented a new index
Hourly observed flood data at the Zijingguan hydrologi-
to reflect the temporal variability of excess rainfall, and the
cal station were used for flood modeling. The corresponding
results showed that the scaling parameters depended mainly
hourly rainfall data at seven rainfall gauging stations located
on the duration and amount of excess rainfall.
in the Zijingguan catchment were obtained from the Hydrol-
Together with statistical analysis, hydrological modeling
ogy and Water Resource Survey Bureau of Hebei Province.
can provide intuitive results regarding how meteorological
The hourly data range from 1956 to 2018. Since no large
and land surface factors impact flood scaling parameters
floods occurred after 2000, some flood events before 2000
(Yang et al. ). For simulated flood events, other control-
were selected for simulation and flood scaling analysis.
ling factors of the scaling parameters, such as the channel
To establish a hydrological model, a digital elevation
slope, flow velocity, and rainfall distribution, can be ident-
model (DEM) with a resolution of 30 m was downloaded
ified by changing the hydrological model parameters
from the Geospatial Data Cloud (http://www.gscloud.cn/).
(Mantilla ; Venkata ). Therefore, it is better to use
Remotely sensed land use data from 1980 to 2000 were
physically based distributed hydrological models that con-
collected from the Institute of Geographic Sciences and
flood
Resources of the Chinese Academy of Sciences (http://
generation processes. For example, Ayalew et al. ()
www.resdc.cn/) and were classified into forestland, grass-
tain
all
the
important
parameters
reflecting
applied the CUENCAS rainfall–runoff model to three river
land, agricultural land, water area, urban area, and unused
basins in Iowa to investigate how the rainfall intensity, dur-
land. Soil data were downloaded from the Environmental
ation, hillslope overland flow velocity, and channel flow
and Ecological Science Data Center for West China
velocity influenced the scaling parameters; they found that
(http://westdc.westgis.ac.cn/).
the key factors affecting the scaling parameters were the rainfall duration and overland flow velocity.
Methods
The main aims of this paper are to (1) perform a simulated flood scaling analysis and evaluate the effects of the
HEC-HMS model calibration and validation
rainfall characteristics and land surface and (2) propose a framework for estimating the scaling exponent by hydrologi-
The Hydrologic Engineering Center-Hydrologic Modeling
cal modeling. Compared with previous studies, the novelty
System (HEC-HMS) model developed by the US Army
of this study is that a framework for determining the flood
Corps of Engineers is widely used for watershed hydrologi-
scaling exponent is proposed by hydrological modeling in
cal simulations. This model has been verified to be
data-scarce regions, especially for multiscaling. Hence, the
applicable in semi-arid regions throughout China (Wang &
proposed method can be used in other data-scarce regions
Sun ). In addition, the Soil Conservation Service curve
on a global scale.
number (SCS-CN) method is often used to calculate the amount of generated runoff, and the SCS unit hydrograph method is used for flow routing. Therefore, we selected
DATA AND METHODS
these two modules for the model. The CN value is determined based on the spatial distributions of the soil water
Data
content and vegetation types (A et al. ) and will vary considerably under wet and dry antecedent conditions
The flood scaling analysis was carried out in the Zijingguan
(Han et al. ).
catchment, situated in the northeastern part of China. The
A set of default parameters is established when the data
catchment has a drainage area of 1,760 km2. The long-term
are input into the model, after which the parameters are
average annual precipitation is approximately 650 mm,
adjusted artificially to ensure that the simulated and
70–80% of which is concentrated in the flood season (July,
observed flood hydrographs fit well. The relative error
August, and September). Flood events are mostly generated
(RE) and the Nash–Sutcliffe efficiency (NSE) coefficient,
by rainstorms with a high intensity and short duration.
which are expressed as Equations (1) and (2), respectively,
J. Li et al.
1094
|
A framework of event-based flood scaling analysis
Hydrology Research
|
51.5
|
2020
HEC-HMS model, which is applicable in semi-humid and
are used to evaluate the performance of the model.
semi-arid regions, is employed to simulate flood events due RE ¼
^ i qi q qi
(1)
to its data availability and structure (Alfy ). The simulated peak discharges and corresponding subcatchment areas are fitted with power functions, and then the scaling
P
^ i )2 (qi q NSE ¼ 1 P i )2 (qi q
(2)
^i are the observed and simulated discharges, where qi and q i is the mean observed discharge of the respectively, and q flood event. If the value of the NSE coefficient is in the range of 0.6–1, the hydrological model is considered to perform well in simulating the flood event.
exponent is obtained for each flood event. Framework for determining the scaling exponent based on simulated flood events Based on the simulated flood events and scaling analysis, a framework that links event-based scaling exponents to quantile-based scaling exponents can be proposed. There are two main assumptions: (1) the input rainfall at each rain gauge is
Flood scaling analysis
spatially uniform and (2) the rainfall events have the same
This study aims to analyze the flood scaling exponent in a
cedures are as follows:
data-scarce region. Therefore, the Zijingguan catchment
1. A typical rainfall event is selected from the observed data.
frequency as the generated flood events. The idea and pro-
2
with a drainage area of 1,760 km (Figure 1), which has
A typical rainfall event should have a large rainfall
only one hydrological station (Zijingguan station), is
amount and high rainfall intensity. Generally, the largest
selected. Scaling analysis is conducted by using the simu-
rainfall event is selected for its rare occurrence as the
lated flood peak discharges from 11 subcatchments. The
low-frequency design flood.
Figure 1
|
Subcatchments of the Zijingguan catchment.
J. Li et al.
1095
|
A framework of event-based flood scaling analysis
2. The typical rainfall event is amplified. The typical rainfall event is amplified by multiple constants, such as 1.5, 2, 2.5, … , to form hypothetical rainfall events with the same temporal distribution as the typical rainfall
Hydrology Research
|
51.5
|
2020
corresponding scaling exponent is calculated according to the relationship established in step 4. The framework is illustrated in Figure 2.
event. 3. The amplified rainfall events are used as inputs to drive the established HEC-HMS model, and the corresponding
RESULTS
flood events at the outlets of the subcatchments and Zijingguan catchment are output with various CN
Flood event simulations
values ranging from 10 to 100. 4. The flood scaling is analyzed for all the simulated flood
The Zijingguan catchment was divided into 11 subcatch-
events generated by the hypothetical rainfall events.
ments, and 17 outlets were selected with drainage areas
Then, using the relationship between the scaling expo-
ranging from 3 to 1,760 km2. Fourteen observed flood
nent and rainfall intensity (total amount), the CN value
events were selected for HEC-HMS calibration and vali-
is obtained, and the existence of simple scaling or multi-
dation, including large floods (>5-year return period with
scaling is concluded.
a peak discharge greater than 500 m3/s) and small floods
5. If the flood shows simple scaling, then the scaling exponent can be used in quantile-based flood scaling.
(<5-year return period with a peak discharge smaller than 500 m3/s). The model parameters were adjusted artificially
6. If the flood shows multiscaling, then the rainfall and the
to fit the observed flood hydrographs, and comparisons
corresponding flood events are assumed to have the same
between the simulated and observed flood events are listed
frequency. The rainfall with a specific frequency is
in Table 1 to estimate the model performance. Some simu-
obtained from rainfall frequency analysis, and the
lated and observed flood events are shown in Figure 3.
Figure 2
|
Event-based flood scaling exponent estimation framework.
J. Li et al.
1096
Table 1
|
|
A framework of event-based flood scaling analysis
Hydrology Research
|
51.5
|
2020
Comparison between the simulated and observed flood characteristics Flood events
Fpo (m3/s)
Fps (m3/s)
RE of Fp (%)
Fdo (mm)
Fds (mm)
RE of Fd (%)
NSE
Calibration
3 Aug 1956 7 Jul 1958 31 Jul 1959 6 Aug 1963 9 Aug 1964 14 Aug 1966 13 Aug 1973 28 Jul 1974
1490 905 648 4490 752.3 242 108 309
1537.4 922.9 697.4 3600 644 262.7 92.6 282.3
3.18 1.98 7.62 19.82 14.40 8.55 14.26 8.64
136.70 12.12 91.44 229.79 67.94 12.60 4.39 13.81
126.26 10.63 103.89 233.98 62.28 11.77 4.58 11.53
7.64 12.29 13.62 1.82 8.33 6.59 4.33 16.51
0.83 0.51 0.62 0.80 0.81 0.88 0.65 0.58
Validation
6 Aug 1975 25 Aug 1978 14 Aug 1979 30 Jul 1988 10 Aug 1988 28 Jul 1996
139 428 245 175 349.8 739.4
132.2 401.1 281.5 182 282.3 810.3
4.89 6.29 14.90 4.00 19.30 9.59
3.94 26.24 19.47 8.28 27.27 97.88
4.41 27.61 20.07 7.85 22.38 82.49
11.93 5.22 3.08 5.19 17.93 15.72
0.91 0.95 0.90 0.79 0.77 0.95
Note: Fp is the flood peak, Fpo is the observed flood peak, and Fps is the simulated flood peak. Fd is the flood depth, Fdo is the observed flood depth, and Fds is the simulated flood depth.
Figure 3
|
Some simulated and observed flood events: (a) 3 August 1956; (b) 6 August 1963; (c) 25 August 1978; and (d) 28 July 1996.
J. Li et al.
1097
|
A framework of event-based flood scaling analysis
Hydrology Research
|
51.5
|
2020
The RE values of the peak discharges for all 14 flood
which is less than 1. These results reflect multiscaling, and
events are less than 20%, ranging from 1.98 to 19.82%,
large flood events show greater scaling exponents and coeffi-
and the RE values of the flood depths range from 13.62
cients of determination (Figures 4 and 5). The values of θ for
to 17.93%. The NSE coefficients of the 14 flood events are
large floods range from 0.7621 to 0.9234 with an average of
between 0.51 and 0.95 with an average of 0.77. Therefore,
0.8634, while those of small floods range from 0.5396 to
the calibrated model is capable of capturing the peak dis-
0.9922 with an average of 0.7225. Although the values of θ
charges, and the RE and NSE coefficient have acceptable
for two small flood events (occurring on 13 August 1973
accuracies. The simulated flood hydrographs fit the obser-
(730,813) and 6 August 1975 (750,806)) are relatively high,
vations very well during both the calibration and the
the R 2 is approximately 0.66, which is lower than that of
validation periods, which implies that the performance of
large floods. These results imply that there must be other fac-
the HEC-HMS model is sufficient to simulate flood events
tors, such as rainfall and land surface properties, influencing
in the Zijingguan catchment and can be used in the follow-
the statistics (Fang et al. ).
ing sections to analyze the flood scaling in this area. Effects of rainfall and CN parameters on the scaling Event-based flood scaling
exponent
According to the simulated flood events in each subcatch-
Due to the large drainage area of the Zijingguan catchment,
ment, the scaling of the peak discharge for each flood
rainfall is not spatially uniform. Figure 6 shows the spatial
event was analyzed with the subcatchment area as the
distributions of the two rainfall events. Each rainfall event
only explanatory variable. The scaling parameters for each
was distributed nonuniformly over the whole catchment,
flood event are listed in Table 2. The power function fit
especially the small rainfall events. This may be the reason
the plots very well with acceptable coefficients of determi-
for the small values of the fitted R 2 in the above section.
nation except for the flood event that occurred on 30 July 1988, verifying that flood scaling exists in this catchment. The scaling exponent θ varies for different flood events in
Flood scaling with areally averaged rainfall as the model input
the range from 0.5396 to 0.9922 with an average of 0.7829, To research the effects of the spatial rainfall distribution on Table 2
|
the flood scaling properties, the areally averaged rainfall cal-
Scaling parameters for each flood event
Flood event
α
θ
R2
3 Aug 1956
1.9212
0.9045
0.9071
7 Jul 1958
0.9144
0.7621
0.5166
31 Jul 1959
1.1848
0.8663
0.9804
6 Aug 1963
4.0131
0.9015
0.9646
9 Aug 1964
1.4038
0.8225
0.8916
14 Aug 1966
0.7012
0.7853
0.7529
13 Aug 1973
0.1041
0.8577
0.6613
28 Jul 1974
1.1296
0.671
0.6007
6 Aug 1975
0.0363
0.9922
0.6601
25 Aug 1978
1.8798
0.646
0.6364
14 Aug 1979
1.2422
0.6801
0.8649
30 Jul 1988
0.6129
0.5396
0.3288
10 Aug 1988
1.1759
0.6078
0.5921
28 Jul 1996
0.9474
0.9234
0.9358
culated by the Thiessen polygon method was input at each of the seven rain gauges. Then, the simulated flood peak discharge at each subcatchment was subjected to flood scaling analysis, as shown in Figure 7 and Table 3. The peak discharge simulated by uniform rainfall shows multiscaling as well. Compared with the fitted power function determined by the observed rainfall, all coefficients of determination increase, demonstrating stronger flood scaling (Figure 8) with the lowest R 2 of 0.6149. The scaling exponent also varies with an average of 0.7829, which is the same as the value under the condition of nonuniform spatial rainfall. For small flood events, subcatchment W150 has a relatively higher peak discharge, as shown in Figure 7 because of its higher calibrated CN value (CN ¼ 65). W150 is located at the lower part of the catchment, and thus, the initial soil moisture content may be higher
J. Li et al.
1098
Figure 4
|
|
A framework of event-based flood scaling analysis
Hydrology Research
|
51.5
|
2020
Variations in θ and R 2 with the flood peak discharge (the dotted line is 500 m3/s, and the red line is the average of the small and large flood peaks, respectively). Please refer to the online version of this paper to see this figure in color: https://doi.org/10.2166/nh.2020.042|0|0|2020.
Figure 5
|
Scaling of the peak discharge with the catchment area fitted by a power function (the flood events of 1975 and 1979 represent small flood events, while those of 1963 and 1996 represent large flood events).
than that in the other subcatchments, leading to a higher
shows obvious scaling properties. This may be because the
runoff generation capacity and higher peak discharge. How-
subcatchments are unsaturated during small flood events
ever, for large flood events, the peak discharge of W150
but become saturated for large flood events.
J. Li et al.
1099
|
A framework of event-based flood scaling analysis
Figure 6
|
Spatial distributions of rainfall events over the Zijingguan catchment: (a) 6 August 1963 and (b) 14 August 1979.
Figure 7
|
Flood scaling for the simulated peak discharges with the areally averaged rainfall as input.
Hydrology Research
|
51.5
|
2020
J. Li et al.
1100
Table 3
|
|
A framework of event-based flood scaling analysis
Hydrology Research
|
51.5
|
2020
different temporal distributions. Therefore, the effects of
Scaling parameters with the areally averaged rainfall as input
Flood events
α
θ
R2
3 Aug 1956
13.531
0.6041
0.9345
7 Jul 1958
1.1047
0.7705
0.8777
31 Jul 1959
0.9179
0.8912
0.9898
6 Aug 1963
3.4961
0.9425
0.9986
9 Aug 1964
1.2534
0.8373
0.9675
14 Aug 1966
0.5259
0.8045
0.9340
13 Aug 1973
0.0753
0.7235
0.9264
28 Jul 1974
1.1235
0.8207
0.9713
6 Aug 1975
0.0806
0.8248
0.8246
25 Aug 1978
0.8298
0.8484
0.9719
14 Aug 1979
1.2297
0.6834
0.9139
30 Jul 1988
0.324
0.5884
0.6149
10 Aug 1988
1.0227
0.7357
0.9138
28 Jul 1996
1.2509
0.8849
0.9848
the rainfall intensity on flood scaling need to be studied with the same spatial and temporal distributions.
Effects of rainfall with the same temporal distribution and CN value on the scaling exponent The rainfall event that occurred on 6 August 1963 was selected as the typical event, based on which some hypothetical rainfall events were obtained by amplifying the typical rainfall event to drive the HEC-HMS model. The flood hydrographs were simulated under different initial CN values. Then, a scaling analysis was conducted, as shown in Figure 10. The larger the maximum 1 h rainfall amount and CN value are, the larger the scaling exponent. When the CN value is sufficiently large (CN 70), the exponent becomes constant regardless of the rainfall intensity. For CN values less than 70, the scaling exponent increases
Relationships between the rainfall characteristics and the scaling exponent
with
increasing
maximum
1h
rainfall
and
finally
approaches a constant. From this figure, if we determine the CN value and the design maximum 1 h rainfall, we
On the basis of the above 14 scaling exponents θ simulated
can obtain the scaling exponent and make flood predictions.
by the areally averaged rainfall, the relationships between the maximum 1 h rainfall (P1) and θ and between the total
Linking the framework to the quantile-based scaling
rainfall (P) and θ were analyzed (Figure 9). The scaling expo-
exponent
nent has no obvious relation with P1 or P, which is uncommon. This may be because the rainfall intensity and
Based on the above flood scaling analysis, flood predictions
total rainfall amount from different rainfall events have
can
Figure 8
| θ and R 2 values under the conditions of spatially uniform and nonuniform rainfall.
be
conducted
by
implementing
a
hydrological
J. Li et al.
1101
Figure 9
|
|
A framework of event-based flood scaling analysis
Hydrology Research
|
51.5
|
2020
Relationships (a) between the maximum 1 h rainfall (P1) and θ and (b) between the total rainfall amount (P) and θ.
DISCUSSION Flood scalings in ungauged basins should be analyzed by analyzing homogeneous regions with sufficient flood data. In watersheds possessing only one or a few hydrological stations, flood scaling analysis cannot be conducted by fitting the observed flood peak and watershed attributes. Therefore, hydrological modeling is an approach for performing scaling research by simulating the flood processes in each subwatershed. Figure 10
|
Relationship among the maximum 1 h rainfall (P1), CN, and θ.
During the flood scaling analysis, the spatial rainfall distribution was found to be a controlling factor for the scaling exponent. Spatially uniform rainfall resulted in better fitting
simulation according to the following steps. (1) A hydrologi-
of the simulated peak discharge and watershed area and
cal model is established first and then calibrated and
resulted in a different scaling exponent. Hence, the differ-
validated. (2) A typical rainfall event with a high rainfall
ence between the scaling exponent obtained from fitting
intensity and a large total amount is selected, and some
the observed rainfall and that obtained from fitting the
hypothetical rainfall events are formed by amplifying the
areally averaged rainfall cannot be neglected. In practice,
typical rainfall event to keep the temporal distribution of
we use the scaling exponent obtained from fitting the
the hypothetical rainfall events the same. (3) The hypotheti-
observed peak discharge. However, this may lead to a sig-
cal rainfall events are areally averaged and used to drive
nificant error.
the hydrological model to simulate the peak discharge for
Similarly, land surface conditions significantly influ-
each subcatchment. (4) Flood scaling analysis is conducted
enced the scaling exponent. However, for different rainfall
by using the simulated peak discharge and the area of
events, the relationship between θ and the maximum 1 h
each subcatchment, and the relations among the scaling
or total rainfall is not good. Since all 14 flood events were
exponent, rainfall and CN values are built, as shown in
simulated by using the same CN value, the temporal rainfall
Figure 10. (5) According to the assumption that the rainfall
distribution must have a significant influence on θ. Furey &
and the corresponding flood have the same frequency, from
Gupta () presented a variable to represent the effective
Figure 10, the scaling exponent of a specific quantile can be
rainfall duration and analyzed the relationship with the scal-
obtained by interpolation under any rainfall intensity and the
ing exponent. Therefore, the influence of the temporal
CN value.
rainfall distribution needs further research.
1102
J. Li et al.
|
A framework of event-based flood scaling analysis
Hydrology Research
|
51.5
|
2020
In addition, the simulated peak discharge of each sub-
and the National Natural Science Foundation of China
catchment was used for flood scaling analysis. Thus, the
(No. 51779165). We extend sincere thanks to the
result must be affected by the peak discharge simulation
Hydrology and Water Resource Survey Bureau of Hebei
error. In this catchment, the peak discharge simulation
Province for providing the research data.
errors ranged from 1.98 to 19.82%, with the largest simulation error obtained for the event that occurred on 6 August 1963. However, this flood event was selected to illus-
DATA AVAILABILITY STATEMENT
trate the multiscaling analysis framework. If the result is applied to a data-scarce region, the simulation accuracy
Data cannot be made publicly available; readers should con-
should be improved further.
tact the corresponding author for details.
Gupta & Dawdy () noted that floods caused by rainstorms exhibited multiscaling, and the results of our study further verify their statements. However, our results were
REFERENCES
obtained in semi-humid and semi-arid regions with multiscaling, whereas in humid regions, the initial CN value is large, and simple scaling may be inferred from the findings in this study.
CONCLUSIONS In this study, we proposed a method to conduct flood scaling analysis by hydrological modeling in data-scarce regions. The scaling exponents were very different between using the observed rainfall and the areally averaged rainfall as model inputs, and the coefficient of determination was larger for spatially uniform rainfall. Therefore, the spatial rainfall distribution has a significant impact on flood scaling. Based on how the spatial rainfall distribution and CN values influence the scaling exponent, a framework was proposed to estimate the scaling exponent in a data-scarce region. We recommend that the areally averaged rainfall of an event be used as the input to the hydrological model; accordingly, the correlation curves relating the scaling exponent, rainfall intensity, and CN values can be drawn. As long as the flood frequency and CN value are determined, the scaling exponent can be obtained from these correlation curves.
ACKNOWLEDGEMENTS This research is supported by the National Key Research and Development Program of China (2018YFC0407902)
A, Y., Wang, G., Liu, T., Xue, B. & Kuczera, G. Spatial variation of correlations between vertical soil water and evapotranspiration and their controlling factors in a semiarid region. Journal of Hydrology 574, 53–63. Alfy, M. E. Assessing the impact of arid area urbanization on flash floods using GIS, remote sensing, and HEC-HMS rainfall-runoff modeling. Hydrology Research 47, 1142–1160. Al-Rawas, G. A. & Valeo, C. Relationship between wadi drainage characteristics and peak-flood flows in arid northern Oman. Hydrological Sciences Journal 55 (3), 377–393. Ayalew, T. B., Krajewski, W. F., Mantilla, R. & Small, S. J. Exploring the effects of hillslope-channel link dynamics and excess rainfall properties on the scaling structure of peakdischarge. Advances in Water Resources 64, 9–20. Ayalew, T. B., Krajewski, W. F. & Mantilla, R. Analyzing the effects of excess rainfall properties on the scaling structure of peak discharges: insights from a mesoscale river basin. Water Resources Research 51 (6), 3900–3921. Eash, D. A. Techniques for estimating flood-frequency discharges for streams in Iowa. Center for Integrated Data Analytics Wisconsin Science Center. Fang, Q., Wang, G., Xue, B., Liu, T. & Kiem, A. How and to what extent does precipitation on multi-temporal scales and soil moisture at different depths determine carbon flux responses in a water-limited grassland ecosystem? Science of the Total Environment 635, 1255–1266. Farmer, W. H., Over, T. M. & Vogel, R. M. Multiple regression and inverse moments improve the characterization of the spatial scaling behavior of daily streamflows in the Southeast United States. Water Resources Research 51 (3), 1775–1796. Furey, P. R. & Gupta, V. K. Effects of excess rainfall on the temporal variability of observed peak-discharge power laws. Advances in Water Resources 28 (11), 1240–1253. Furey, P. R. & Gupta, V. K. Diagnosing peak-discharge power laws observed in rainfall–runoff events in Goodwin Creek experimental watershed. Advances in Water Resources 30 (11), 2387–2399.
1103
J. Li et al.
|
A framework of event-based flood scaling analysis
Furey, P. R., Troutman, B. M., Gupta, V. K. & Krajewski, W. F. Connecting event-based scaling of flood peaks to regional flood frequency relationships. Journal of Hydrologic Engineering 21 (10), 04016037. Galster, J. C., Pazzaglia, F. J., Hargreaves, B. R., Morris, D. P., Peters, S. C. & Weisman, R. N. Land use effects on watershed hydrology: the scaling of discharge with drainage area. Geology 34 (9), 713–716. Gupta, V. K. & Dawdy, D. R. Physical interpretations of regional variations in the scaling exponents of flood quantiles. Hydrological Processes 9 (3–4), 347–361. Gupta, V. K., Castro, S. L. & Over, T. M. On scaling exponents of spatial peak flows from rainfall and river network geometry. Journal of Hydrology 187 (1), 81–104. Gupta, V. K., Troutman, B. M. & Dawdy, D. R. Towards a nonlinear geophysical theory of floods in river networks: an overview of 20 years of progress. Nonlinear Dynamics in Geosciences 121–151. Gupta, V. K., Ayalew, T. B., Mantilla, R. & Krajewski, W. F. Classical and generalized Horton laws for peak flows in rainfall-runoff events. Chaos 25 (7), 075408. Han, S. J., Wang, S. L., Xu, D. & Zhang, Q. Scale effects of storm-runoff processes in agricultural areas in Huaibei Plain. Transactions of the Chinese Society of Agricultural Engineering 28 (8), 32–37. Han, D., Wang, G., Liu, T., Xue, B., Kuczera, G. & Xu, X. Hydroclimatic response of evapotranspiration partitioning to prolonged droughts in semiarid grassland. Journal of Hydrology 563, 766–777. Ishak, E., Haddad, K., Zaman, M. & Rahman, A. Scaling property of regional floods in New South Wales Australia. Natural Hazards 58 (3), 1155–1167. Jothityangkoon, C. & Sivapalan, M. Temporal scales of rainfall–runoff processes and spatial scaling of flood peaks: space–time connection through catchment water balance. Advances in Water Resources 24 (9–10), 1015–1036.
Hydrology Research
|
51.5
|
2020
Kroll, C. N., Rapant, D. F. & Vogel, R. M. Prediction of hydrologic statistics in nested watersheds across the United States. In: World Environmental and Water Resources Congress, ASCE, Baltimore, MD, pp. 2326–2335. Lee, K. T. & Huang, J. K. Influence of storm magnitude and watershed size on runoff nonlinearity. Journal of Earth System Science 125 (4), 777–794. Liu, S. Y., Huang, S. Z., Xie, Y. Y., Wang, H., Leng, G. Y., Huang, Q., Wei, X. T. & Wang, L. Identification of the nonstationarity of floods: changing patterns, causes, and implications. Water Resources Management 33 (3), 939–953. Mantilla, R. Physical Basis of Statistical Scaling in Peak Flows and Stream Flow Hydrographs for Topologic and Spatially Embedded Random Self-Similar Channel Networks. Dissertations and Theses – Gradworks. Medhi, H. & Tripathi, S. On identifying relationships between the flood scaling exponent and basin attributes. Chaos 25, 075405. Menabde, M. & Sivapalan, M. Linking space–time variability of river runoff and rainfall fields: a dynamic approach. Advances in Water Resources 24 (9), 1001–1014. Ogden, F. L. & Dawdy, D. R. Peak discharge scaling in small Hortonian watershed. Journal of Hydrologic Engineering 8 (2), 64–73. Venkata, P. M. Role of Rainfall Variability in the Statistical Structure of Peak Flows. Dissertations and Theses – Gradworks. Wang, L. & Sun, W. J. Research on HEC-HMS and Vflo rainfall characteristics simulation and comparative based on DEM data: a case study of Miyun District, Beijing. Acta Scientiae Circumstantiae 39, 3559–3564. Yang, W. C., Yang, H. B. & Yang, D. W. Classifying floods by quantifying driver contributions in the Eastern Monsoon Region of China. Journal of Hydrology 585, 124767. Yue, S. & Gan, T. Y. Scaling properties of Canadian flood flows. Hydrological Processes 23 (2), 245–258.
First received 27 March 2020; accepted in revised form 6 August 2020. Available online 11 September 2020
1104
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Response of redox zonation to recharge in a riverbank filtration system: a case study of the Second Songhua river, NE China Xiaosi Su, Yaoxuan Chen, Hang Lyu, Yakun Shi, Yuyu Wan and Yiwu Zhang
ABSTRACT Bank filtration induced by groundwater pumping results in redox zonation along the groundwater flow path. Besides the river water, recharge from other sources can change local redox conditions; therefore, redox zonation is likely to be complex within the riverbank filtration (RBF) system. In this study, hydrodynamics, hydrogeochemistry, and environmental stable isotopes were combined together to identify the redox conditions at an RBF site. The recharge characteristics and redox processes were revealed by monitoring the variations of water level, δ 2H and δ 18O, and redox indexes along shallow and deep flow paths. The results show that local groundwater is recharged from river, regional groundwater, and precipitation. The responses of redox zonation are sensitive to 2 different sources. In the river water recharge zone near shore, O2, NO 3 , Mn(IV), Fe(III), and SO4 are
reduced in sequence, the ranges of each reaction are wider in deep groundwater because of the high-velocity deep flow. In the precipitation vertical recharge zone, precipitation intermittently drives O2, NO 3 , and organic carbon to migrate through vadose zone, thereby decreasing the groundwater reducibility. In the regional groundwater lateral recharge zone in the depression cone, the reductive regional groundwater is continuously recharging local groundwater, leading to the cyclic reduction of Mn(IV) and Fe(III). Key words
| hydrodynamics, hydrogeochemistry, recharge characteristics, redox zonation, riverbank filtration system, stable isotope
HIGHLIGHTS
• • •
Hydrodynamics, hydrogeochemistry, and environmental stable isotopes were combined to explore the redox zonation in an RBF system. Redox zones and sequential redox processes involved were identified and partitioned. Analysis of the redox zonation response to flow regime and recharge of river, precipitation, and regional groundwater.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.108
Xiaosi Su Yaoxuan Chen Yiwu Zhang Institute of Water Resources and Environment, Jilin University, Changchun 130026, China Xiaosi Su College of Construction Engineering, Jilin University, Changchun 130021, China Yaoxuan Chen Yiwu Zhang College of New Energy and Environment, Jilin University, Changchun 130026, China Yaoxuan Chen Hang Lyu (corresponding author) Yuyu Wan Key Laboratory of Groundwater Resources and Environment (Jilin University), Ministry of Education, Changchun 130026, China E-mail: lvhangmail@163.com Hang Lyu Yuyu Wan Jilin Provincial Key Laboratory of Water Resources and Environment, Jilin University, Changchun 130026, China Yakun Shi No.1 Institute of Geo-environment Survey of Henan, Zhengzhou 450000, China
1105
X. Su et al.
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
INTRODUCTION Riverbank filtration (RBF) involves water extraction by
2þ 2þ concentrations of NHþ at 0.98– 4 , Mn , and Fe
pumping wells at riverside fields to enhance water supply
3.62 N mg/L,
by stimulating river recharge to groundwater. This can
respectively,
attenuate or degrade pollutants, such as suspended solids,
0.50 N mg/L, 0.40 mg/L, and 2.00 mg/L, respectively. (WHO
inorganic or organic substances, poisonous heavy metals,
). Therefore, an appraisal of the characteristics of the
pathogenic viruses, and bacteria; hence, it is considered an
redox environment and biogeochemical behavior of the sensi-
efficient and natural treatment technology for water quality
tive components during bank filtration under different sources
improvement (Hiscock & Grischek ; Tufenkji et al.
of recharge should be undertaken to gain a deeper understand-
; Trauth et al. ; Muz et al. ). Rivers are rich in
ing of the complex biogeochemical processes.
4.71–6.82 mg/L,
and
8.54–12.74 mg/L,
which exceed the WHO guideline limits of
O2 and organic matter; driven by Gibbs free energy, the
This study aims to: (1) carry out survey and analysis of
organic matter supplies electrons to the lowest unoccupied
the hydrodynamics of the river and groundwater around
molecular orbital locations during river filtration; therefore,
the Kaladian well field by combining δ 2H and δ 18O
2 O2, NO are reduced in 3 , Mn(IV), Fe(III), and SO4
analyses to identify the spatial distribution characteristics
sequence (Kedziorek & Bourg ; Farnsworth & Hering
of recharge; (2) survey the spatial distribution of the redox
). However, this process can also cause incomplete
environment, sensitive components in river water and
removal of organic carbon, suspended solids, and inorganic
groundwater, and the soluble N and C components in the
pollutants, and release of heavy metals, such as Fe, Mn, and
aquifer medium, as well as analyze their relationships with
As, from sediment or the aquifer medium to groundwater
hydrodynamic conditions; and (3) partition the redox zona-
(Gandy et al. ; Su et al. ).
tion along the groundwater flow path based on the spatial
Bank filtration controlled by groundwater pumping results in a relatively stable redox zonation along the
distribution of recharge and redox conditions, as well as clarify the hydrogeochemical processes within each zone.
groundwater flow path and previous research has mostly focused on the spatial distribution of the redox environment (Greskowiak et al. ; Kumar & Riyazuddin ). Many
MATERIALS AND METHODS
factors affecting the formation of redox zonation, such as fluctuations in river stage and groundwater level, intensity
Study area
of pumping, lithology and structure of aquifer medium, and water temperature, have all been considered in previous
The Kaladian riverside well field, located on the alluvial–
studies, yet the influence of recharge conditions of local
proluvial plain of the Second Songhua River in northeastern
groundwater, such as water from precipitation and regional
China, is characterized by flat topography with elevations
groundwater conditions have been neglected (Burt et al.
varying from 124 to 129 m above sea level. The unconfined
; Massmann et al. ; Kohfahl et al. ; Wang
aquifer is mainly composed of sand with the thickness of
et al. ). In addition to river water, groundwater recharge
17–20 m, and a continuous and stable clay-laminated layer
from other sources can also change the local redox con-
of ∼24 m thickness forms the impervious base of the uncon-
ditions, and hence, redox zonation is likely to be more
fined aquifer. The upper and lower parts of the unconfined
complex within the RBF system (Kedziorek & Bourg ;
aquifer consist of fine sand and medium sand with a thin
Buzek et al. ).
layer of silty-clay in the middle, respectively. The long-term
The Kaladian well field, located in northeastern China,
pumping wells (C1–C8) operating intermittently at a total
is characterized by regional groundwater rich in Fe and
pumping rate of 6,000–10,000 m3/day have formed a
Mn with Mn
2þ
2þ
and Fe
contents of over 6 and 10 mg/L,
respectively. The local groundwater in recent 3 years has
stable groundwater depression cone centered around well C5 and C6.
X. Su et al.
1106
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
The well field is located in a cold temperate zone that
NS1, NS2, NS3, NS4, NS5, and NS6 were established at
experiences a continental semi-arid monsoon climate with a
0, 2, 5, 8, 14, and 30 m from the shore, respectively. Further-
dry and windy spring, hot and rainy summer, cool autumn
more, six far-shore monitoring points FS1, FS2, FS3, FS4,
with early frost, and a cold and long winter. Monthly average
FS5, and FS6 were located at 80, 200, 420, 700, 850, and
temperatures in January and July can reach 17.5 and
1,110 m from the shore. Each monitoring point has a shal-
23.3 C, respectively, while the annual average temperature
low (7.5 m) and deep (14 m) well with a screen interval of
is 4.7 C. Annual mean precipitation is 425.7 mm, of which
6–7.5 and 12.5–14 m, respectively. Furthermore, two
more than 70% is concentrated from June to August, while
regional groundwater monitoring wells, RG1 and RG2, are
the annual mean evaporation is 1,689.8 mm, with the maxi-
located at 2,100 and 2,700 m, respectively, from the shore.
mum in May at 316.5 mm. The Second Songhua River, flowing from southeast to northwest, is controlled by climate
Sampling
and upstream water conservancy projects and has an annual average runoff, river surface width, and depth of 476.0 m3/s,
River and groundwater sampling
400–450 m, and 3–7 m, respectively. Herein, 22 long-term monitoring wells with a depth of
River water and groundwater samples were collected in Sep-
6–9 m were constructed to monitor the groundwater level
tember 2018. River water was collected 0.5 m below the
and quality since December 2015. Based on the monitoring
surface. Groundwater was pumped at a rate <5 L/min from
results, a dense monitoring section from the river to the
the wells, and the samples were collected after flushing the
depression cone center was constructed in August 2018, as
volume of the well pipe three times. Sampling bottles were
shown in Figure 1. The hydrodynamic and environmental
filled to the brim and then sealed. Samples for SO2 4 , NO3 ,
conditions are susceptible to change within 30 m of the
and NHþ 4 were stored in polyethylene bottles, while concen-
shore; therefore, six near-shore monitoring points NS1,
trated sulfuric acid was added to the samples for NHþ 4 to
Figure 1
|
Location of Kaladian riverside well field and distribution of monitoring wells.
1107
X. Su et al.
|
Response of redox zonation to recharge in a riverbank filtration system
generate a pH <2. Samples for Mn2þ were packed in brown
Hydrology Research
|
51.5
|
2020
In situ permeability test for riverbed sediment
glass bottles with diluted hydrochloric acid to generate a pH <2. Samples for DOC were stored in brown glass bottles and
The standpipe falling head test method, also known as the
concentrated sulfuric acid was added to generate a pH <2.
tube test, was used to determine the permeability of the riv-
Samples for testing stable isotopes were collected in
erbed sediments; four test points R1, R2, R3, and R4 were
500 mL polyethylene bottles and stored at 4 C.
arranged, respectively, at 5, 30, 150, and 300 m from point
A Hach HQ40d portable meter (Hach Company, USA)
NS1, as shown in Figure 1. The length of the standpipe
was used to test pH, Eh, and dissolved oxygen (DO) content
was 1.50 or 2.80 m for different river depths, with an inner
2þ
in water, while a Hach DR1900 was used to test Fe
HS
and
diameter of 0.04 m. After being inserted into the depth Lv
on site. An atomic absorption analyzer (Scientific
of the riverbed sediment, the water head change process in
iCE 3300, Thermo, Germany) was used to analyze Mn2þ
the tube was recorded after adding water from the upper
content, and an ion chromatograph (881 Compact IC pro,
end, and the calculation was carried out by Equation (1)
Metrohm, Switzerland) was used to test SO2 content at 4
(Chen ),
the Institute of Water Resources and Environment, Jilin University (Changchun, China). A continuous flow analyzer (San þ þ, Skalar, Netherlands) was used to measure NO 3
Lν h1 Kν ¼ , ln t2 t1 h2
(1)
and NHþ 4 contents, and a TOC analyzer (TOC-L CPH Shimadzu, Japan) was used to test DOC at the Northeast
where Kv is the vertical hydraulic conductivity of riverbed
Institute of Geography and Agroecology, Chinese Academy
sediments, m/day; Lv is the length of the riverbed sediments
of Sciences (Changchun, China). The δ 2H and δ 18O values
in the test tube, m; t1 and t2 are the instantaneous moments
were tested using a laser isotope meter (L2140-i, Picarro,
at the beginning and end of the test days; and h1 and h2
USA) at the Third Institute of Oceanography, Ministry of
correspond to the water level in the test tube, m.
Natural Resources (Xiamen, China), with accuracies of ±0.50 and ±0.20‰, respectively.
RESULTS Aquifer medium sampling River stage and groundwater level dynamics Sample sites of the vadose zone and aquifer medium were located in the vicinity of point NS5, FS1, and FS3 and
The river stage and groundwater level dynamics from Janu-
samples were collected using percussion drills in September
ary 2016 to October 2018 are shown in Figure 2. Annual
2018. The collection horizons were 0–0.5, 1.0–1.5, 2.0–2.5,
variations in the river stage ranged between 1.0 and 3.2 m,
and 6.5–7.5 m below the land surface. Samples were
and the river stage was high during July–September and rela-
stored in 500 mL glass bottles and were immediately trans-
tively low during October–March of the following year. The
ported to the laboratory at 20 C for testing total
surface of the river was frozen from December–March of the
nitrogen (TN), soil organic carbon (SOC), and ion exchange
following year with the frozen layer thickness being 1 m.
þ þ forms of NO 3 /NH4 (IEF-NO3 /NH4 ).
The trend of regional groundwater level changes was con-
TN content was tested at 120–124 C in an alkaline sub-
sistent with that of the river stage, and the annual change
strate using K2S2O8 to oxidize all forms of nitrogen in the
was generally less than 1.5 m. The local groundwater level
medium to NO 3 , then tested using a continuous flow analy-
of the well field was lower than the river level throughout
zer. The loss-on-ignition method was used to obtain the
the year, indicating year-round river filtration.
organic matter content, and then by using the Van Bemme-
In point NS1–NS6 near the shore, the groundwater level
len factor of 1.724, was converted to the SOC content. The
could quickly respond to the change in the river level, show-
þ IEF-NO 3 /NH4 in the 2 mol/L KCl extraction medium was
ing an annual variation of 2.8–3.0 m. In point FS1–FS4 far
tested using a continuous flow analyzer.
from shore, the annual variation in the groundwater level
X. Su et al.
1108
Figure 2
|
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
Water level dynamics of river and groundwater.
was low, at 0.9–1.8 m. During winter when the pumping
values at the RFZ are closer to those of the river end mem-
intensity was high, variations in the water levels in point
bers, along the direction of the groundwater runoff,
FS3 and FS4, located within the depression cone, were rela-
whereas the groundwater δ 2H and δ 18O values at the GRZ
tively large, indicating that the far-shore groundwater level
and DCZ increase gradually, and are close to the values of
was less affected by the fluctuations in the river level and
the regional groundwater end member. In the vertical dimen-
was instead mainly affected by pumping.
sion, the shallow and deep groundwater δ 2H and δ 18O values
According to the above-mentioned water level dynamics
show no obvious difference in the RFZ and GRZ within
of the river and groundwater, the NS1–FS4 monitoring sec-
200 m from the shore, while the shallow groundwater δ 2H
tion was divided into three hydrodynamic zones as follows:
and δ 18O values are obviously closer to those of the regional
(1) river filtration zone (RFZ) (River-NS6); (2) groundwater
groundwater end member than deep groundwater in the
runoff zone (GRZ) (NS6–FS3); and (3) depression cone
DCZ, which is 420–700 m from the shore.
zone (DCZ) (FS3–FS4). Spatial distribution characteristics of redox-sensitive 2
18
Spatial distribution of δ H and δ O in river water and
indexes
groundwater Redox indexes in river water and shallow groundwater River water and regional groundwater are two stable lateral recharge sources, and two major end members of the ground-
Eh/DO. The Eh values and DO contents of the river water
water in the well field. The Second Songhua River originates
were high at 152.25 mV and 8.205 mg/L, respectively, as
in the Changbai Mountains (elevation 2,750 m), and the
shown in Figure 4(a) and 4(b), indicating an oxidizing environ-
elevation difference with respect to the well field is more
ment. During river filtration, shallow groundwater Eh and DO
than 2,000 m, leading to lower δ 2H and δ 18O values in the
at the RFZ rapidly dropped to below 150.00 mV and
river water. The western part of Jilin, where the well field fea-
1.0 mg/L, respectively, gradually making the environment in
tures low terrain and shallow groundwater depth, is easily
this zone reducing in nature. Regional groundwater Eh and
2
18
affected by evaporation; hence, the δ H and δ O values of
DO were 151.62 mV and 0.68 mg/L, respectively, indicat-
the regional groundwater are relatively high.
ing a reducing environment. Average shallow groundwater
In the NS1–FS4 monitoring section, the distribution 2
18
Eh and DO at the GRZ and DCZ were 126.25 mV and
ranges and average of δ H and δ O values for shallow and
1.18 mg/L, respectively, which were higher than those of
deep groundwater lie between those for the river water and
the shallow groundwater at the RFZ and regional ground-
regional groundwater, as shown in Figure 3(a) and 3(b). In
water, with a tendency of increasing along the direction of
the horizontal dimension, the groundwater δ 2H and δ 18O
flow until reaching their highest values at the DCZ.
X. Su et al.
1109
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
Figure 3
|
Spatial distribution of δ 2H and δ 18O values in river water and groundwater along the groundwater flow path in the monitoring section.
Figure 4
|
Spatial distribution of environmental and hydrochemical indexes in shallow groundwater along the groundwater flow path in the monitoring section.
|
51.5
|
2020
1110
X. Su et al.
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
pH. The pH value of the river water was 7.72 as shown in
Meanwhile, Mn2þ and Fe2þ at the DCZ were closer to
Figure 4(c). During river filtration, pH continuously
that of the regional groundwater.
decreased to 7.28 in well NS1-1, then slowly increased along the direction of flow, while the average shallow
2 SO2 4 . The SO4 content in the river water was 29.72 mg/L
groundwater pH in the RFZ was 7.64. Average regional
as shown in Figure 4(f). Average shallow groundwater SO2 4
groundwater pH was 7.18. Average shallow groundwater
in well NS1-1–NS4-1 that is within 14 m from shore was
pH at both the GRZ and DCZ was 7.31, which is between
29.98 mg/L without obvious change. In well NS5-1, which
the shallow groundwater at the RFZ and regional ground-
is 30 m from shore, SO2 4 began to decrease, reaching its
water values, with a tendency of decreasing along the
lowest value of 8.97 mg/L in well FS1-1, which is 200 m
direction of flow; the lowest pH was recorded in the DCZ.
from shore at the GRZ. Average regional groundwater
þ þ NO 3 /NH4 . The contents of NO3 and NH4 in river water
2 SO2 4 was 58.36 mg/L. Average shallow groundwater SO4
at the DCZ was 62.29 mg/L, which is higher than the shallow groundwater at the RFZ, GRZ, and regional ground-
were 2.274 and 0.124 N mg/L as shown in Figure 4(d) and
water. Meanwhile, the content of HS was relatively low at
4(e), respectively. During river filtration, the shallow ground-
all monitoring wells and was maintained below the detec-
þ water NO 3 decreased rapidly, while NH4 increased rapidly,
until in well NS1-1 at the shore, where NO 3 then dropped to 0.762 N mg/L, having been consumed up to approximately NHþ 4
tion limit. DOC. The DOC content of the river water was 23.14 C mg/L
increased to 0.613 N mg/L, a 4-fold
as shown in Figure 4(i). During river filtration, at the RFZ and
increase, until in well NS4-1 that is 8 m from shore, where
GRZ within 80 m from the shore, average shallow ground-
70%. Meanwhile,
þ NO 3 dropped below the detection limit, while NH4 was
NO 3
water DOC of well NS1-1–FS1-1 gradually decreased along
was very low,
the direction of flow with a value of 9.33 C mg/L on average,
mostly below the detection limit, while average NHþ 4 was
which is approximately 60% less than that of the river water.
relatively high at 1.881 N mg/L. In the shallow groundwater
The average regional groundwater DOC was 16.38 C mg/L.
0.865 N mg/L. Regional groundwater
of the GRZ and the DCZ, average
NO 3
was 0.759 N mg/L,
At the GRZ and DCZ within 80–700 m from the shore, the
which was higher than the shallow groundwater at the RFZ
average shallow groundwater DOC of well FS1-1–FS4-1
and regional groundwater, while average NHþ 4 was 1.188 N
was 17.75 C mg/L, which is higher than that of the shallow
mg/L, which was between that of the shallow groundwater
groundwater at the RFZ and regional groundwater.
at RFZ and regional groundwater. Redox indexes in deep groundwater Mn2þ/Fe2þ. The contents of Mn2þ and Fe2þ in the river
Eh/DO. The Eh values and DO contents of deep ground-
water were extremely low as shown in Figure 4(g) and
water were relatively stable as shown in Figure 5(a) and
4(h). During river filtration, shallow groundwater Mn2þ and Fe2þ began to increase until reaching well NS1-1; with Mn2þ peaking at 1.12 mg/L in well NS3-1, which is 5 m from the shore. Fe2þ peaked at 3.42 mg/L in well NS6-1, which is 30 m from the shore. Both Mn2þ and Fe2þ began to decrease after peaking until they dropped to a minimum value in well FS1-1, which is 80 m from the shore. Average regional groundwater Mn2þ and Fe2þ were 5.98
5(b), respectively. Average deep groundwater Eh and DO in well NS1-2–NS6-2 that are within 30 m from shore at the RFZ were 125.32 mV and 0.63 mg/L respectively, indicating a reducing environment. Deep groundwater Eh and DO in well FS1-2–FS4-2 that are 80–700 m from shore at the GRZ and DCZ were 129.49 mV and 0.74 mg/L, respectively, and there was no significant change when compared with the deep groundwater at the RFZ.
and 12.21 mg/L, respectively. Average shallow groundwater Mn2þ and Fe2þ at the GRZ and DCZ were 3.79 and
pH. The pH value was relatively stable in deep groundwater
6.38 mg/L, respectively, ranging between the shallow
as shown in Figure 5(c) along the direction of flow. Average
groundwater at the RFZ and the regional groundwater.
deep groundwater pH in well NS1-2–FS2-2 within 420 m
X. Su et al.
1111
Figure 5
|
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
Spatial distribution of environment and hydrochemistry indexes for deep groundwater along the groundwater flow path in the monitoring section.
from shore at the RFZ and GRZ was 7.39. The average deep
mg/L, respectively, which are higher than those of deep
groundwater pH in well FS3-2–FS4-2 at the DCZ was 7.21,
groundwater at the RFZ and regional groundwater.
which was closer to that of the regional groundwater. Mn2þ/Fe2þ. As shown in Figure 5(g) and 5(h), the contents þ NO 3 /NH4 .
Compared with the shallow groundwater, the
of Mn2þ and Fe2þ in deep groundwater at the RFZ and GRZ
deep groundwater had lower NO 3 content and higher
began to increase in well NS1-2, with Mn2þ in well NS6-2,
NHþ 4
content as shown in Figure 5(d) and 5(e), respectively.
which is 30 m away from the shore, peaking at 1.18 mg/L,
þ Average deep groundwater NO 3 and NH4 in well NS1-2–
and average Fe2þ in well FS1-2 and FS2-2, which are 80–
NS6-2, which are within 30 m from shore at the RFZ,
200 m from shore peaking at 3.76 mg/L. After peaking,
were 0.394 and 2.081 N mg/L, respectively. Average NO 3
Mn2þ began to decrease in well FS1-2, which is 80 m from
NHþ 4
in well FS1-2–FS4-2, which are 80–700 m
shore, while Fe2þ did not show a significant decline, and
from shore at the GRZ and DCZ, were 0.452 and 2.230 N
both average Mn2þ and Fe2þ gradually increased in well
and
X. Su et al.
1112
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
FS2-2–FS3-2, which are at the junction of the GRZ and DCZ
Spatial distribution characteristics of carbon and
with values of 3.79 and 6.18 mg/L, respectively. The Mn2þ
nitrogen in aquifer
2þ
and Fe
|
2020
contents were closer to those in the regional
groundwater than to the water near the depression cone
Vertical distributions of N and C contents in the vadose zone
center.
and aquifer medium are shown in Figure 6. The points NS5, FS2, and FS4 are 14, 200, and 700 m from the shore and are
SO2 4 .
The
SO2 4
content in deep groundwater of well NS1-
located in the RFZ, GRZ, and DCZ, respectively.
2–FS1-2, which are within 80 m from shore at the RFZ and
In the vertical profile, the surface medium of the vadose
GRZ, remained relatively stable with an average value of
zone of each hole had a high content of fine particles, as the
24.53 mg/L as seen in Figure 5(f). Deep groundwater
surface is mostly covered by trees and crops; the surface
SO2 4 in well FS1-2 and FS2-2, which are 80–200 m from
medium forms an important growth and metabolism area
shore, decreased for a short interval, and then continued
for plants and microorganisms where organic life and
to increase along the direction of flow to reach 32.43 mg/
humus are abundant, as well as fertilizers. The average con-
L in well FS3-2 and FS4-2, which are 420 and 700 m from
tents of TN and SOC were 0.262 N and 6.273 C g/kg,
shore, respectively. Concentrations of HS were the same
respectively, and the average C:N value was 14.217, indicat-
as those for shallow groundwater, with relatively low
ing that nitrogen assimilation and mineralization might be
concentrations.
prevalent (Dhondt et al. ; Evans et al. ). The content of coarse particles in the lower part of the vadose zone and
DOC. The content of DOC in deep groundwater was higher
aquifer medium increased, while contents of TN and SOC
than that in the shallow groundwater as shown in Figure 5(i).
were lower with an average of 0.144 N g/kg and 1.495 C g/
In deep groundwater, average DOC in well NS1-2–NS6-2,
kg, respectively, and the average C:N ratio was 8.186.
which is within 30 m from shore at the RFZ, was 24.75 C
Although this C:N value indicates that ammonification
mg/L. Average DOC in well FS1-2–FS4-2, which are 80–
could occur, the organic nitrogen content within the substrate
700 m from shore at the GRZ and DCZ, was 25.85 C mg/L,
was limited. Meanwhile, the spatial distribution character-
which is higher than the value recorded for the deep ground-
istics of IEF-NHþ 4 and IEF-NO3 in the medium were
water at the RFZ and the regional groundwater; DOC at the
similar to those of TN and SOC, with a tendency to gradually
GRZ increased significantly.
decrease.
Figure 6
|
Vertical distribution of nitrogen and carbon in the vadose zone medium at point NS5, FS2, and FS4.
X. Su et al.
1113
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
In the horizontal direction, the lithological composition
sand with a hydraulic conductivity (K ) of 28.47 m/day,
and structure of pores in point NS5 and FS4 were similar, as
while the lower aquifer consisted of medium sand with a
well as the nitrogen and carbon indicators. In point FS2, the
K of 40.38 m/day.
content of coarse particles in the vadose zone was signifi-
The heterogeneity of the water-bearing medium affected
cantly higher than those in point NS5 and FS4, indicating
the different hydraulic exchange characteristics between the
that the permeability of the vadose zone of point FS2 is rela-
river and groundwater, showing two flow paths from the
tively stronger. Meanwhile, the contents of TN, SOC,
river to the center of the depression cone, as shown in
IEF-NHþ 4 , and IEF-NO3 in the medium of each layer in
Figure 7. In shallow riverbed with lower permeability, flow
point FS2 were also lower than those of point NS5 and FS4.
occurred at a shallow depth (9 m) with lower velocity and was lateral. In deeper riverbed with higher permeability,
DISCUSSION Hydrodynamic exchange between river and
flow occurred at a depth greater than 9 m (reaching 17– 19 m at the bottom of the aquifer), had a higher velocity, and was vertical to the lateral flow.
groundwater Recharge of local groundwater Flow path along the direction of filtration and runoff To more precisely clarify the recharge conditions along the The hydraulic gradient between the river and groundwater
direction of the two flows, the hydrogen and oxygen stable
acts as a driving force, while gravity flow induces ground-
isotopes of the river water and groundwater were calculated
water runoff from the river to the pumping wells (Su et al.
and analyzed based on mass conservation. Relative contri-
). As Figure 7 shows, the riverbed sediments near
butions of the river water and regional groundwater could
point NS1 were relatively denser and continuous, compris-
be estimated using an end member mixing model which
ing of silty-clay mixed with fine sand with a thickness of
can be determined using Equation (2),
0.05–0.3 m, while for tube test points R1 and R2 located at this district, the results of Kv varied from 4.82 to 6.39 m/ day. Riverbed sediments in the middle of the river contained
δ 1 n1 þ δ 2 n2 ¼ δ(n1 þ n2 ) n1 þ n2 ¼ 1
(2)
fine sand with low horizontal continuity and stratification, and the tube test points R3 and R4 showed that Kv varied
where δ1, δ2, and δ are the isotope values for river water,
from 46.11 to 59.37 m/day. Otherwise, the vertical differ-
regional groundwater, and local groundwater, respectively;
ence in lithology and permeability of the upper and lower
and n1 and n2 are the proportions of river and regional
aquifer was evident; the upper aquifer consisted of fine
groundwater, respectively.
Figure 7
|
Direction of flow during river filtration and groundwater runoff at the monitoring section.
X. Su et al.
1114
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
In this study, δ 18O was used for this calculation. The
of reduction processes indicates DOC as the electron
contribution rates of the river at the RFZ, GRZ, and DCZ
donor, as shown in Table 1. Generally, ΔG (W) values
were 84.37–93.56%, 72.80–74.82%, and 26.08–50.42%,
of aerobic respiration and denitrification are higher and
respectively, at the shallow depths; the contribution rates
closer, leading to in situ overlapping of their ranges
of the river at the three zones were 78.45–89.80%, 59.93–
(Korom ; Rivett et al. ). The reactions of Mn(IV)
64.82%, and 52.23–55.03%, respectively, at deeper depths.
and Fe(III) reduction with lower ΔG (W) values could
Thus, it can be concluded that the contribution rate of the
occur sequentially with a clear boundary. However, when
river to deep local groundwater is greater than that of the
the Eh values become low, SO2 4 reduction and methane fer-
shallow groundwater, indicating a closer relationship
mentation can occur simultaneously (Champ et al. ;
between the river and deep groundwater. Meanwhile,
Stumm & Morgan ). In this study, as the redox environ-
along the direction of groundwater flow, with increasing dis-
mental indexes and sensitive components along the
tance away from the river, the contribution rate of the river
direction of flow varied, O2, NO 3 , Mn(IV), Fe(III), and
to both shallow and deep groundwater continuously
SO2 were reduced sequentially, as shown in Figure 8(a) 4
decreased, especially at the DCZ, where the contribution
and 8(b), until reaching well FS1-1, which is 80 m from
of the regional groundwater was quite high.
the shore at shallow flow, and well FS2-2, which is 200 m
The amplitude of the contribution rate of the river at
from the shore at deep flow.
shallow flow was higher than that at deep flow, indicating
Based on this, regional groundwater makes the second
that the recharge condition was more complex for shallow
highest contribution toward the quantity of recharge, which
groundwater. In point FS2 at the GRZ, where the per-
is continuous and stable, and shows spatial variability. The
meability of the vadose zone is greater than that in point
variations in Mn2þ and Fe2þ contents in groundwater might
NS5 and FS4, precipitation could more easily pass through
reflect the influence of regional groundwater lateral recharge
the vadose zone to recharge the shallow groundwater. In
in cases when: (1) the regional groundwater has high concen-
18
points FS1 and FS2 at the GRZ, the δ O values of shallow
trations of Mn2þ and Fe2þ; (2) the Mn2þ and Fe2þ contents in
groundwater are lower than that of deep groundwater, as
groundwater at well FS1-1, which is 80 m away from shore,
shown in Figure 3, which implies that the shallow ground-
are lowest because of the sequential redox processes; and
water at the GRZ could be affected by precipitation.
(3) mixing of groundwater with two different reducibilities suggests a new environment wherein the concentrations of Mn2þ and Fe2þ in the far-shore groundwater are influenced
Redox zonation of RBF
by groundwater mixing and reaction. Precipitation recharge may also contribute to the quan-
Principle and identification processes
tity of groundwater. However, variations in NHþ 4 , NO3 ,
River water is the most important recharge component of
and DOC contents in groundwater might reflect the influ-
the RBF system. During river filtration, the main sequence
ence of the vertical recharge of precipitation, in cases
Table 1
|
Redox processes in a closed system (modified after Stumm & Morgan 1995; Champ et al. 1979) Equation
ΔG (W) kJ/eq
Aerobic respiration
CH2 O þ O2 ¼ CO2 þ H2 O
125.1
Denitrification
þ CH2 O þ 4=5NO 3 þ 4=5H ¼ CO2 þ 2=5N2 þ 7=5H2 O
118.8
Mn(IV) reduction
CH2 O þ 2MnO2 þ 4Hþ ¼ CO2 þ 2Mn2þ þ 3H2 O
81.8
Fe(III) reduction
CH2 O þ 4Fe(OH)3 þ 8Hþ ¼ CO2 þ 4Fe2þ þ 11H2 O
28.9
Sulfate reduction
þ CH2 O þ 1=2SO2 4 þ 1=2H ¼ CO2 þ 1=2HS þ 2H2 O
25.3
Methane fermentation
CH2 O ¼ 1=2CH4 þ 1=2H2 O
23.2
Reaction
X. Su et al.
1115
Figure 8
|
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
Variation trends of the redox environmental index and sensitive components along the directions of shallow and deep flows.
when: (1) contents of TN and SOC are quite low which inhi-
partition the redox zonation, the range of the river water
bits the release of NHþ 4 and contents of TN, NHþ , 4 NO3 ,
DOC to the groundwater; (2)
recharge zone (RWZ) was inferred as 0–80 m in shallow
and SOC of the vadose zone
flow and 0–200 m in deep flow. The range of the precipi-
medium are abundant, especially at the upper region of
tation vertical recharge zone (PVZ) was 80–420 m in
the aquifer; and (3) increasing NHþ 4 and SOC contents are
shallow flow and 200–420 m in deep flow. The range of
positively correlated with the particle contents of the
the regional groundwater lateral recharge zone (RGZ) was
vadose zone medium.
420–700 m at both the shallow and deep flows.
Therefore, along the directions of both shallow and deep flows, the increasing rates of NHþ 4 , NO3 , and DOC contents
were greater than those of the Mn2þ and Fe2þ contents
Range of sequential redox reactions in the river water recharge zone
between points FS1 and FS3, while the trend was opposite between points FS3 and FS4, as shown in Figure 8(a) and
Precise partitions of the RWZ were carried out and the
8(b). Thus, by adjusting the hydrodynamic zonation to
results are shown below. Vertical differences in the
1116
X. Su et al.
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
permeability of aquifer medium are known to lead to the
content. However, this range of this zone could be assigned
differences in flow velocity, which in turn, influence water
in well NS4-1–NS6-1, 8–30 m from the shore for shallow
retention time, nutrient flux, microbial community structure,
groundwater and well FS1-2–FS2-2, 80–200 m from the
acid–base condition, and redox condition between the river
shore for deep groundwater.
water and groundwater (Massmann et al. ; Su et al. ), leading to sequential reduction reactions that have
SO2 4 reduction zone
wider distribution ranges in deep groundwater. Meanwhile, this zone is the most important for producing NHþ 4 by
This zone is symbolized by an obvious decrease in SO2 4 con-
ammonification (Dhondt et al. ; Evans et al. ).
tents, as shown in Table 1; the ΔG (W) values of Fe(III) and SO2 reduction were relatively similar, leading to a 4
Aerobic respiration and denitrification zone
non-definite boundary. The HS produced by SO2 4 reduction would precipitate upon reaction with Fe2þ, which
Microorganism-induced aerobic respiration and denitrifica-
is one of the reasons for the decreasing levels of Fe2þ and
tion occurred preferentially and overlapped; they occurred
low concentrations of monitored HS as mentioned above.
within point NS2, less than 2 m from shore. As denitrifica-
The ranges could be assigned in well NS6-1–FS2-1, 30–
tion was catalyzed by denitrifying bacteria, the facultative
200 m from the shore for the shallow groundwater, and
anaerobic bacteria could reduce NO 3 to N2 under a
well FS2-2, ∼200 m from the shore for the deep groundwater.
relatively O2-rich environment and a DO content of up to 2–4 mg/L (Rivett et al. ). However, aerobic respiration
Redox conditions within the PVZ
occurred within ∼0.1–1.5 m beneath the riverbed surface, as previously reported (Revsbech et al. ; Kumar &
The contents of Mn2þ and Fe2þ reduced, but stable values
Riyazuddin ; Su et al. ).
were observed between well NS6-1 and FS1-1 for shallow
Mn(IV) reduction zone
of precipitation and regional groundwater as shown in
groundwater during RBF with a relatively little influence Figure 8(a); therefore, shallow groundwater in well FS1-1 When the Eh values of groundwater dropped to a desired
and the regional groundwater were determined to be the
range, manganese minerals in the riverbed sediment and
two Mn2þ and Fe2þ mixing end members for far-shore
aquifer medium were released into the groundwater by
groundwater. By using the end member mixing model as
organic complexation or reductive dissolution, as symbo-
shown in Equation (2), the mixing ratio of water using
content, which increased
δ 18O was calculated; then, the Mn2þ and Fe2þ mixed lines
slowly to reach a peak value along the flow direction and
for the shallow and deep flows were obtained. Results are
then decreased because of oxidation and precipitation
shown in Figure 9(a) and 9(b).
lized by the variation in Mn
2þ
(Greskowiak et al. ; Kedziorek & Bourg ).
The redox background at the PVZ, which is represented
Hence, the range of this zone could be assigned as well
by point FS2, was the SO2 reduction environment. The 4
NS2-1–NS4-1, 2–8 m from shore for shallow flow and well
measured Fe2þ concentration was lower than that in the
NS2-2–FS1-2, 2–80 m from shore for deep flow.
Fe2þ mixed line for both the shallow and deep flows, indicating the occurrence of reactions that consume Fe2þ. Possible
Fe(III) reduction zone
reasons for this might be (1) Fe2þ combined with HS produced by SO2 4 reduction and (2) oxidizing by O2 recharge
Similar to Mn(IV) reduction, under favorable Eh values of
occurred vertically through the vadose zone as Fe2þ was
groundwater, ferrous minerals in the riverbed sediment
readily oxidized, except for DOC and HS as shown in
and aquifer medium were released into the groundwater
Table 2. However, in the shallow groundwater, HS was con-
by organic complexation or reductive dissolution, as symbo-
sumed almost completely in well FS2-1 and DO content was
lized by the variation in Fe2þ content similar to that in Mn2þ
still relatively high as seen in Figure 4(a), thus O2 would act
X. Su et al.
1117
Figure 9
Table 2
|
|
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
The Mn2þ and Fe2þ concentrations at mixed lines of shallow and deep flows.
Redox processes in an open system (modified after Stumm & Morgan 1995; Champ et al. 1979) Equation
ΔG (W) kJ/eq
Aerobic respiration
CH2 O þ O2 ¼ CO2 þ H2 O
125.1
Sulfide oxidation
þ 1=2HS þ O2 ¼ 1=2SO2 4 þ 1=2H
Reaction
2þ
99.8 þ
þ O2 þ 10H2 O ¼ 4Fe(OH)3 þ 8H
96.2
Fe(II) oxidation
4Fe
Nitrification
þ 1=2NHþ 4 þ O2 ¼ 1=2NO3 þ H þ 1=2H2 O
43.3
Mn(II) oxidation
2Mn2þ þ O2 þ 2H2 O ¼ 2MnO2 þ 4Hþ
40.3
on the consumption. In the deep groundwater, SO2 4 began to reduce in well FS2-2, while DO content was maintained at low levels as seen in Figure 5(a); thus, it is inferred that HS and O2 contributed to the decreasing of Fe2þ. Unlike Fe2þ, the measured Mn2þ concentration was
Redox conditions within the regional groundwater lateral recharge zone At this zone, under conditions where NHþ 4 , NO3 , and DOC
along the flow direction did not increase, (NH4)2SO4 fertili-
mixed lines of both shallow
zer used at the surface caused significant accumulation of
and deep groundwater, indicating high Mn(IV) reduction.
SO2 4 in both shallow and deep groundwater as shown in
obviously higher than the Mn
2þ
The redox environment was still reductive despite an increase of Eh in comparison with the
SO2 4
reduction
Figures 4(f) and 5(f), indicating that vertical recharge is significant but still less than the lateral recharge. The Eh values
zone. Therefore, it is inferred that the actual lowest Eh
and DO contents along the flow direction at points FS3 and
values in shallow and deep groundwater should be located
FS4 increased, where they were located within the
between the controlling values of Mn(IV) and Fe(III)
depression cone, and except for direct vertical recharge,
reduction. Under this relatively weak reductive condition,
the intermittent state of pumping would promote ground-
water fluctuation leading to O2 dissolution by air trap and
nitrification would hardly occur in the presence of HS 2þ
and Fe . Therefore,
NO 3
could be recharged and accom-
pany O2 vertically, and both would be reduced before
diffusion (Massmann et al. ; Kohfahl et al. ). The points FS3 and FS4 represent the groundwater of
This shows that the vertical recharge
RGZ, where the measured Mn2þ and Fe2þ concentrations
affected the sequential redox zonation mainly because of
were obviously higher than the mixed lines at both shallow
the recharge of electron accepters with higher Gibbs free
and deep groundwater as shown in Figure 9(a) and 9(b),
energy and electron donors.
indicating that Mn(IV) and Fe(III) reduction were both
Fe(III) and
SO2 4 .
1118
X. Su et al.
|
Response of redox zonation to recharge in a riverbank filtration system
Hydrology Research
|
51.5
|
2020
active. The actual lowest Eh values in shallow and deep
the aquifer medium would consistently release more Mn2þ
groundwater should be located at the controlling values of
and Fe2þ into the groundwater.
Fe(III) reduction, which was lower than that of the PVZ, owing to continuous lateral recharge of the high Fe2þ reduc-
ACKNOWLEDGEMENTS
tive regional groundwater. Similar to the PVZ, under relatively reductive con2þ ditions, oxidation of HS , Fe2þ, NHþ would 4 , and Mn
The authors would like to thank the editors of Hydrology
barely occur, and small quantities of O2 and NO 3 recharged
Research and the reviewers for their thoughtful and
vertically would be rapidly reduced. In summary, at the
constructive
RGZ and PVZ, O2,
NO 3,
Mn(IV), and Fe(III) would be
comments,
which
helped
improve
the
manuscript.
reduced in sequence again, which is affected by vertical and lateral recharge including: (1) depression of groundwater reduction in different degrees; (2) recharge intensity
FUNDING INFORMATION
of electron donors; and (3) recharge intensity of electron accepters with higher Gibbs free energy.
This work was supported by the National Natural Science Foundation of China (Grant numbers: 41877178, 41372238).
CONCLUSIONS DATA AVAILABILITY STATEMENT With a background of high ferrous and manganese content in regional groundwater, river filtration and groundwater
All relevant data are included in the paper or its Supplemen-
runoff toward the groundwater depression cone in the
tary Information.
RBF system have two different groundwater flow paths (shallow and deep) due to the variation in permeability of riverbed sediments at different intervals of the river. Mean-
REFERENCES
while, recharge intensities of the river water, regional groundwater, and precipitation showed differences in the spatial distribution along the groundwater path from the river to the depression cone, influencing the redox conditions. At the RWZ with a high river recharge contribution, O2, 2 NO 3 , Mn(IV), Fe(III), and SO4 were reduced in sequence
meaning the DOC, as the electron donor, formed a strong reducing environment accompanied by an increase in NHþ 4 through ammonification. However, lateral recharge from reductive regional groundwater with high Fe2þ and Mn2þ concentrations and vertical recharge from oxidative precipi2 tation with high O2, NO 3 , SO4 , and DOC concentrations
change the redox conditions of the local groundwater from RBF. At the PVZ, the influence of vertical precipitation recharge was more significant than that of the lateral recharge, which led to higher Eh values in groundwater in the PVZ than that in the RGZ. These redox conditions provided an opportunity for Mn(IV) reduction at the PVZ and for both Mn(IV) and Fe(III) reduction at the RGZ; therefore,
Burt, T. P., Pinay, G., Matheson, F. E., Haycock, N. E., Buttruini, A., Clement, J. C., Danielescu, S., Dowrick, D. J., Hefting, M. M., Hillbrichtllkowaka, A. & Maitre, V. Water table fluctuations in the riparian zone: comparative results from a pan-European experiment. Journal of Hydrology 265 (1), 129–148. Buzek, F., Kadlecova, R., Jackova, I. & Lnenickova, Z. Nitrate transport in the unsaturated zone: a case study of the riverbank filtration system Karany, Czech Republic. Hydrological Processes 26 (5), 640–651. Champ, D. R., Gulens, J. & Jackson, R. E. Oxidationreduction sequences in ground water flow systems. Canadian Journal of Earth Sciences 16 (1), 12–23. Chen, X. Measurement of streambed hydraulic conductivity and its anisotropy. Environmental Geology 39 (12), 1317–1324. Dhondt, K., Boeckx, P., Cleemput, O. V. & Hofman, G. Quantifying nitrate retention processes in a riparian buffer zone using the natural abundance of 15N in NO-3? Rapid Communications in Mass Spectrometry 17 (23), 2597–2604. Evans, D. M., Schoenholtz, S. H., Wigington, P. J. & Stephen, M. G. Nitrogen mineralization in riparian soils along a river continuum within a multi-land-use basin. Soil Science Society of America Journal 75 (2), 719.
1119
X. Su et al.
|
Response of redox zonation to recharge in a riverbank filtration system
Farnsworth, C. E. & Hering, J. G. Inorganic geochemistry and redox dynamics in bank filtration settings. Environmental Science & Technology 45 (12), 5079–5087. Gandy, C. J., Smith, J. W. N. & Jarvis, A. P. Attenuation of mining-derived pollutants in the hyporheic zone: a review. Science of the Total Environment 373, 435–446. Greskowiak, J., Prommer, H., Massmann, G. & Nützmann, G. Modeling seasonal redox dynamics and the corresponding fate of the pharmaceutical residue phenazone during artificial recharge of groundwater. Environmental Science & Technology 40, 6615–6621. Hiscock, K. M. & Grischek, T. Attenuation of groundwater pollution by bank filtration. Journal of Hydrology 266 (3), 139–144. Kedziorek, M. A. M. & Bourg, A. C. M. Electron trapping capacity of dissolved oxygen and nitrate to evaluate Mn and Fe reductive dissolution in alluvial aquifers during riverbank filtration. Journal of Hydrology 365 (1–2), 74–78. Kohfahl, C., Massmann, G. & Pekdeger, A. Sources of oxygen flux in groundwater during induced bank filtration at a site in Berlin, Germany. Hydrogeology Journal 17 (3), 571–578. Korom, S. F. Natural denitrification in the saturated zone: a review. Water Resources Research 28, 1657–1668. Kumar, A. R. & Riyazuddin, P. Seasonal variation of redox species and redox potentials in shallow groundwater: a comparison of measured and calculated redox potentials. Journal of Hydrology 444 (10), 187–198. Massmann, G., Nogeitzig, A., Taute, T. & Pekdeger, A. Seasonal and spatial distribution of redox zones during lake bank filtration in Berlin, Germany. Environmental Geology. 54, 53–65. Muz, M., Oswald, S. E., Schäfferling, R. & Lensing, H. Temperature-dependent redox zonation, nitrate removal and attenuation of organic micropollutants during bank filtration. Water Research 162, 225–235.
Hydrology Research
|
51.5
|
2020
Revsbech, N. P., Jacobsen, J. P. & Nielsen, L. P. Nitrogen transformations in microenvironments of river beds and riparian zones. Ecological Engineering 24 (5), 447–455. Rivett, M. O., Buss, S. R., Morgan, P., Smith, J. W. & Bemment, C. D. Nitrate attenuation in groundwater: a review of biogeochemical controlling processes. Water Research 42 (16), 4215–4232. Stumm, W. & Morgan, J. J. Aquatic Chemistry: Chemical Equilibria and Rates in Natural Waters, 3rd edn. WileyInterscience, New York. Su, X., Lu, S., Gao, R., Su, D., Yuan, W., Dai, Z. & Papavasilopoulos, E. N. Groundwater flow path determination during riverbank filtration affected by groundwater exploitation: a case study of Liao River, Northeast China. Hydrological Sciences Journal 62 (14), 2331–2347. Su, X., Lu, S., Yuan, W., Nam, C. W., Dai, Z., Dong, W., Du, S. & Zhang, X. Y. Redox zonation for different groundwater flow paths during bank filtration: a case study at Liao River, Shenyang, northeastern China. Hydrogeology Journal 26, 1573–1589. Trauth, N., Musolff, A., Knöller, K., Kaden, U. S., Keller, T., Werban, U. & Fleckenstein, J. H. River water infiltration enhances denitrification efficiency in riparian groundwater. Water Research 130, 185–199. Tufenkji, N., Ryan, J. N. & Elimelech, M. The promise of bank filtration. Environmental Science & Technology 36 (21), 422–428. Wang, P., Yao, J., Wang, G., Hao, F., Shrestha, S., Xue, B., Xie, G. & Peng, Y. Exploring the application of artificial intelligence technology for identification of water pollution characteristics and tracing the source of water quality pollutants. Science of the Total Environment 693, 133440. WHO Guidelines for Drinking-Water Quality, 4th edn. World Health Organization, Geneva, Switzerland.
First received 30 July 2020; accepted in revised form 27 August 2020. Available online 14 September 2020
1120
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Coincidence probability of streamflow in water resources area, water receiving area and impacted area: implications for water supply risk and potential impact of water transfer Xingchen Wei, Hongbo Zhang, Vijay P. Singh, Chiheng Dang, Shuting Shao and Yanrui Wu
ABSTRACT Under changing environment, the feasibility and potential impact of an inter-basin water transfer project can be evaluated by employing the coincidence probability of runoff in water sources area (WSA), water receiving area (WRA), and the downstream impacted area (DIA). Using the Han River to Wei River Water Transfer Project (HWWTP) in China as an example, this paper computed the coincidence probability and conditional probability of runoff in WSA, WRA and DIA with the copulabased multivariate joint distribution and quantified their acceptable and unfavorable encounter probabilities for evaluating the water supply risk of the water transfer project and exploring its potential impact on DIA. Results demonstrated that the most adverse encounter probability (dry–dry– dry) was 26.09%, illustrating that this adverse situation could appear about every 4 years. The
Xingchen Wei Hongbo Zhang (corresponding author) Chiheng Dang Shuting Shao Yanrui Wu School of Water and Environment, Chang’an University, Xi’an, China E-mail: hbzhang@chd.edu.cn Hongbo Zhang Key Laboratory of Subsurface Hydrology and Ecological Effect in Arid Region, Ministry of Education, Chang’an University, Xi’an, China
acceptable and unfavorable probabilities in all encounters were 44.83 and 55.17%, respectively, that is the unfavorable situation would be dominant, implying flood and drought risk management should be paid greater attention in project operation. The conditional coincidence probability (dry WRA & dry DIA if dry WSA) was close to 70%, indicating a requirement for an emergency plan and management to deal with potential drought risk. Key words
| bivariate copula, coincidence probability, conditional probability, Han River to Wei River Water Transfer Project, trivariate copula
HIGHLIGHTS
• • • • •
Coincidence probability of annual runoff in all main impacted areas is investigated. Copula-based method to capture the encounter situations of water transfer project. Acceptable and unfavorable probabilities in all encounters imply water transfer risk. Conditional coincidence probability with dry water resources area was computed. An emergency water scheduling plan is required to deal with potential drought risk.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.106
Vijay P. Singh Department of Biological and Agricultural Engineering & Zachry Department of Civil and Environmental Engineering, Texas A&M University, College Station, TX, USA and National Water Center, UAE University, Al Ain, UAE
1121
X. Wei et al.
|
Coincidence probability of streamflow for water transfer risk, China
Hydrology Research
|
51.5
|
2020
GRAPHICAL ABSTRACT
INTRODUCTION Inter-basin water transfer (IBT) means building water trans-
project feasibility. Some multivariate approaches as main
fer projects that span two or more basins for transferring
analysis tools were applied to calculate synchronous–asyn-
water from a basin with abundant water resources to those
chronous probabilities of streamflow, precipitation and
in shortage and for redistributing water resources among
drought between water source area and WRA. For instance,
the basins to meet water demand in the water-deficient
He et al. () employed a Bayesian network model to
area (Zhuang ). IBTs project, as an important safeguard
investigate the rich–poor rainfall encounter risk between
to join different water systems, has been widely applied in
WSA and WRA in the middle route of the South-to-
many countries and regions, e.g. Australia, Canada, China,
North Water Transfer Project (SNWTP) in China and
India and the United States, with the purpose of supporting
implemented real-time scenario simulations, reflecting the
economic and societal development (Manshadi et al. ;
change in risk in the operation of water transfer projects.
Yan & Chen ; Du et al. ). However, it is being
Liu et al. () investigated the concurrent drought prob-
debated that the hydrological cycle has altered in some
ability between the water source and destination regions of
basins or regions under the combined influence of climate
the central route of the SNWTP using Clayton copula and
change and human activities (Zou et al. ), with the
general circulation models (GCMs) and found that the prob-
result that the feasibility of IBT projects planned or under
ability of concurrent drought events was highly likely to
construction may be questionable. The questions include
increase during 2020–2050, representing the WSA probably
whether there will be enough water to be diverted and the
having insufficient amounts of water for diversion while the
new potential influence on water use, ecological protection
WRA urgently needing the diversion of water in concurrent
and disaster control (e.g. drought and pollution) in
drought years. Du et al. () investigated the synchronous
the downstream areas of the water resources area under the
and asynchronous exceedance probabilities of wet versus
changing environment. Thus, the coincidence probability
dry conditions of precipitation in WSA and WRA of the
analysis of annual runoff in these areas, including synchro-
Shuhe to Futuan Water Transfer Project (SFWTP) using a
nous and asynchronous probabilities between or among the
bivariate copula joint distribution function, and results
water sources area (WSA), water receiving area (WRA) and
indicated that climate change had positive impacts on the
the downstream impacted area (DIA), is appealing for evalu-
exceedance probabilities, demonstrating that the project
ating the feasibility and potential influence of IBT projects.
risk was decreasing. In addition, the coincidence probability
In previous studies, the coincidence probability analysis
analysis of precipitation in WSA, WRA and DIA had been
of IBT projects played an important role in determining
reported by Yan & Chen (). This analysis quantified
1122
X. Wei et al.
|
Coincidence probability of streamflow for water transfer risk, China
Hydrology Research
|
51.5
|
2020
the synchrony and asynchrony of precipitation for the
to build the joint probability distribution between the annual
middle route of SNWTP and verified the effectiveness of tri-
runoff of three regions (WSA, WRA and DIA) and trivariate
variate Clayton copula in the study area, and obtained the
coincidence probability and conditional coincidence prob-
combination frequencies for the middle route of SNWTP,
ability were analyzed to obtain information for flood and
representing that the amount of transferable water was
drought risk management for the project.
generally assured, but the possibility for water transfer was very small if extreme deficit rainfall events occurred in the WRA.
STUDY AREA AND DATA
It is found from the literature that bivariate and trivariate copulas have been widely used in the coincidence
Han River to Wei River Water Transfer Project
probability analysis of hydrological variables in the relevant areas of IBT projects (Yan & Chen ; Du et al. ).
As an important project to resolve the water conflict in
Therein, bivariate copulas are generally used to investigate
northwest China due to increasing industrial and dom-
the hydrological combination frequencies between WSA
estic water consumption, the HWWTP connecting the
and WRA for determining the feasibility of a water transfer
upper Han River basin (WSA) and Wei River basin
project, and that between WSA and DIA for analyzing the
(WRA) in China (as shown in Figure 1) is designed to
potential influence of the project on the downstream areas
annually transfer an average of 1,000 million m3 of water
(Yan & Chen ). Trivariate copulas are employed to
in 2025 and 1,500 million m3 in 2030 from Huangjinxia
calculate the hydrological combination frequencies and con-
reservoir and Sanhekou reservoir in the upper Han
ditional frequencies among WSA, WRA and DIA by
River basin to large-scale industrial zones and urban
capturing the spatial dependence structure of hydrological
areas in the Wei River basin, with the main purpose of
variables influencing the coincidence probabilities of asyn-
alleviating watershed water conflict (Wu et al. ).
chrony and synchrony. Recent years have seen a growing
Because 70% of water is for industrial development, it
interest in applying copulas to hydrological frequency analy-
requires stable water transfer, implying high requirement
sis, which can be attributed to manifold advantages of
in controlling water supply risk, especially in the changing
copulas in modeling joint distributions, representing flexi-
environment.
bility in selecting arbitrary margins and dependence
The WSA and the DIA of HWWTP are located in the
structure, the ability to deal with three variables or more,
upper
and the separability in analyzing marginal and dependence
95,200 km2 and have a mainstream length with approxi-
Han
River
basin,
with
a
drainage
area
of
structure (Salvadori & De Michele ; Serinaldi et al.
mately 925 km (Figure 1). The basin is influenced by the
; Zhang & Singh ).
north subtropic monsoon, representing the annual mean
In this study, trivariate combination frequencies and
temperature ranging between 12 and 16 C (Li et al. )
conditional frequencies of annual runoff series in WSA,
and average annual precipitation varying from 700 to
WRA and DIA influenced by the Han River to Wei
1,800 mm, and 80% of annual precipitation falls in the wet
River Water Transfer Project (HWWTP) in China, were
period from May to October in which the corresponding
investigated through copula-based multivariate probability
runoff is about 4.11 × 1010 m3, accounting for 70% of the
distribution, with the purpose of estimating water supply
total annual runoff. Also, the runoff of the entire basin
risk of water transfer project and exploring its potential
demonstrates a large annual and interannual variability (Li
impact on the downstream area of the water source areas.
et al. ). Therein, the WSA mainly refers to the watershed
Different from the previous relevant studies on the water
above Huangjinxia and Sanhekou reservoirs and the DIA is
transfer project, it uses annual runoff data from three
downstream of them, primarily covering the watershed
gauges located in WSA, WRA and DIA, respectively and
below Shiquan gauge.
provides a reference for the scheduling of inter-basin transfer
The WRA of HWWTP mainly refers to Guanzhong
projects. The bivariate and trivariate copulas were employed
plain region located in the central part of Shaanxi Province
X. Wei et al.
1123
Figure 1
|
|
Coincidence probability of streamflow for water transfer risk, China
Hydrology Research
|
51.5
|
2020
Study area and locations of three gauges potentially impacted by the Han River to Wei River Water Transfer Project (HWWTP) in China.
and belongs to the Wei River basin. The Guanzhong plain 2
DSDJK gauges were selected during 1956–2000. Consider-
region covers a total area of 55,500 km , as the most
ing that measured runoff is greatly affected by human
economically developed area in Shaanxi Province, including
activities in WRA, representing nonstationary feature (Lan
five prefecture-level cities, i.e. Xi’an city, Baoji city,
et al. ), the natural runoff data, covering 45 years from
Xianyang city, Weinan city, Tongchuan city and one agricul-
1956 to 2000, were selected from the XY gauge. The
tural high-tech industries demonstration zone as Yangling
runoff data were obtained from hydrological manuals pub-
(Tian et al. ). The Guanzhong plain region has an aver-
lished by the Hydrological Bureaus of the Yellow River
age elevation of approximately 500 m and belongs to the
Conservancy Commission and Yangtze River Conservancy
continental monsoon climate, with the annual mean temp-
Commission. Statistical features of the annual data series
erature of 6–13 C and precipitation of 500–800 mm (Deng
are shown in Table 1.
et al. ). Data
Table 1
|
Statistical features of annual runoff during 1956–2000
Standard
According to the water transfer route of the HWWTP, the
deviation (108 m3)
Coefficient of variation
Coefficient of skewness
data used in this paper included annual runoff data from
Gauges
Mean (108 m3)
Shiquan (SQ) gauge on the WSA, Danjiangkou gauge (the
Shiquan (SQ)
327.5
143.6
0.4385
0.9964
watershed above SQ is not covered or the data deducted
Xianyang (XY)
49.8
20.6
0.4125
0.7392
that from at SQ, DSDJK) on the DIA, and Xianyang (XY)
Danjiangkou (DSDJK)
368.1
136.4
0.3703
1.0753
gauge on the WRA (Figure 1) (see Supplementary Material). Therein, the observed runoff data series from SQ and
Note: Runoff at Danjiangkou does not include the runoff from the area above Shiquan gauge.
X. Wei et al.
1124
|
Coincidence probability of streamflow for water transfer risk, China
METHODOLOGIES
Hydrology Research
|
51.5
|
2020
can be easily constructed and are capable of capturing a variety of dependence structures with several desirable properties, such as symmetry and associativity, they have
Copula functions
become very popular (Nelsen ; Hofert ). Sklar () introduced the copula theorem to model the
The widely used Archimedean family copulas include
stochastic nature of multi-dimensional processes by using
the Frank copula, Clayton copula, Ali–Mikhail–Haq
univariate marginal distributions to derive multivariate dis-
copula (AMH) and Gumbel–Hougaard (G–H) copula, with
tribution functions. Nowadays, copula models play an
a parameter θ. For bivariate copula function, Kendall’s cor-
active role in the hydrological field (Yan & Chen ;
relation coefficient τ has a certain relation with parameter θ,
Zhang & Singh ), such as drought assessment
as shown in Table 2. The appropriate copulas can be deter-
(Chanda et al. ), flood risk analysis (Favre et al. ;
mined by different values of Kendall’s τ from observations
Serinaldi et al. ; Chebana et al. ) and streamflow
(Hofert ; Xie & Wang ).
simulation (Chen et al. ). According to Sklar’s theorem,
Moreover, parameters of multi-dimensional copulas can
the copula function C can be used to describe a joint multi-
be estimated by the maximum-likelihood estimator (MLE)
variate probability distribution of n correlated variables (X1,
or inference of functions for margins (IFM) method (Joe
X2, X3,…, Xn). If variables (X1, X2, X3,…, Xn) have arbitrary
, ). In view of the well-known optimality properties
marginal distribution functions FX1 (x1 ), FX2 (x2 ), . . . , FXn (xn ),
of the MLE, it would be the preferred option for estimating
respectively, there will exist a copula function C, which
θ. Nevertheless, it is found from application that the more
can combine these marginal distribution functions to give
flexible IFM method is preferable to the MLE. Although
the joint distribution function, FX1 ,X2 ,...,Xn (x1 , x2 , . . . , xn ), as
the conceptual bases of the two methods are very similar, and the efficiency of these two methods is almost the same
follows:
in many cases, the IFM method is easier to calculate than FX1 ,X2 ,...Xn (x1 , x2 , . . . , xn ) ¼ Cθ [FX1 (x1 ), FX2 (x2 ), . . . , FXn (xn )] ¼ Cθ (u1 , u2 , ::, un )
u1 , u2 , ::, un ∈ [0, 1]
where
are
uniformly
the MLE method. For example, the trivariate copula parameter θ is estimated in two steps in the IFM method. In the first step,
(1)
the parameters αk (k ¼ 1, 2, 3) of each marginal distribution
distributed
are estimated separately via Xki, i ¼ 1,2,…,n, and this estima∧
as
tor is expressed as αk . The second step is that θ is estimated
u1 ¼ FX1 (x1 ), u2 ¼ FX2 (x2 ), . . . , un ¼ FXn (xn ), and the copula
by replacing αk for αk in the log-likelihood (as shown in
random
realizations
of
the
variables,
defined
parameter is θ. A unique copula function C(⋅) exists, when
∧
Equation (2)) and the IFM estimate of θ is as follows:
these marginal distributions are continuous (Sraj et al. ). There are many types of copula functions, such as Plackett copula (Plackett ), Archimedean copulas and elliptical
∧
θ ¼ arg max
n X
∧
∧
i¼1
(2)
copulas (Fang et al. ). Because the Archimedean copulas Table 2
|
∧
log [cθ {F1 (X1i , α1 ), F2 (X2i , α2 ), F3 (X3i , α3 ); θ}]
Types of Clayton, Frank and GH copula (Nelsen 2006) Cθ
θ0 range
Relation between θ and τ
Frank
1 (e θu 1)(e θv 1) ln 1 þ θ θ e 1
∞ < θ < ∞
τ ¼1þ
Clayton
(u θ þ v θ 1) 1=θ
θ>0
AMH
uv=[1 θ(1 u)(1 v)]
1 θ < 1
G–H
exp { [( ln u)θ þ ( ln v)θ ]
Archimedean copula
1=θ
}
θ 1
ð 4 1 θ t dt 1 t θ θ 0e 1
θ 2þθ 2 2 1 2 τ ¼ 1 1 ln (1 θ) 3θ 3 θ 1 τ ¼1 θ τ¼
1125
X. Wei et al.
|
Coincidence probability of streamflow for water transfer risk, China
Hydrology Research
Marginal distribution
expressed as follows:
The Pearson type III (P-III) distribution, Gumbel distri-
Dn ¼ max jPEi PTi j (i ¼ 1, 2, . . . , n)
bution and Lognormal distribution were employed in this
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u 2 N pffiffiffiffiffiffiffiffiffiffiffi u 1X RMSE ¼ MSE ¼ t (PEi PTi ) N i¼1
study for describing the marginal distribution of single annual runoff. The cumulative distribution function (CDF) of P-III distribution can be defined as
F(x) ¼ P(x > xp ) ¼
βα Γ(α)
ð∞
pffiffiffiffiffiffiffiffiffiffiffi AIC ¼ N ln ( MSE) þ 2k
|
51.5
|
2020
(6)
(7)
(8)
where N is the sample size, k is the number of parameters
(x δ)α 1 e β(x δ)
(3)
xp
of different distributions, PEi and PTi are the empirical frequency and theoretical frequency, respectively. Therein,
where α, β and δ are the shape, scale and location parameters of the P-III distribution, respectively.
the empirical joint probability always plays a significant role in the selection of a copula function, which is used as a criterion for judging and selecting the best theoretical dis-
The CDF of Gumbel distribution can be defined as
tribution from the Archimedean copula functions. Assuming the variables X and Y with the same length, the empirical
h i x k F(x) ¼ P(x > xp ) ¼ exp e γ
(4)
frequency of bivariate variables (X and Y ) was estimated by using the Gringorten formula (Gringorten ), as
where γ and k are the scale and the location parameters of the Gumbel distribution, respectively. The CDF of Lognormal distribution can be defined as " 1 F(x) ¼ P(x > xp ) ¼ pffiffiffiffiffiffi σ 2π
ð ∞ exp xp
( ln x μ)2 2σ 2 t
follows. H(x, y) ¼ P(X xi , Y yi ) ¼
# dt
(5)
NO: of(X xi , Y yi ) 0:44 N þ 0:12
(9)
where (xi , yi ) is the combination of the ith values in the X and Y series arranged in increasing order, i is 1: N, and N is the length of the series. For the trivariate copula function
where μ and σ are the mean and standard deviation formed by the natural logarithm of the variable x, respectively. The linear moment method (Hosking ) was employed to estimate the parameters of the above three distributions, and a fitting test was used to choose the suitable marginal distribution of a single variable.
also, the above equation was used. The theoretical frequencies can be obtained by the models of marginal and copula distributions. It is noted that according to RMSE and AIC criteria, the smaller AIC and RMSE are, the better the fitness of the distribution. Coincidence probability
Fitting test
The coincidence probability is usually defined as the probability of two or more events that happen at the same
In this paper, Kolmogorov–Smirnov (K–S) test (Dn, Massey
time. For the IBT project, it is considered that the runoff
), root mean square error (RMSE, Zhang & Singh ,
coincidence is generally defined as the simultaneous occur-
), Akaike’s information criterion (AIC, Zhang & Singh
rence of runoff in two or more basins, which represent the
, ) were employed to measure the goodness of fit
frequency P of wetness, dryness or normal of one basin
of the marginal and joint distributions which can be
when the other basin is in a condition of wetness, dryness
X. Wei et al.
1126
|
Coincidence probability of streamflow for water transfer risk, China
or normal. It is reported that the 37.5 and 62.5% quantiles
Hydrology Research
|
51.5
|
2020
Wet–normal–dry periods coincidence probability, Pwnd:
are usually used as thresholds to define the condition of wetness or dryness in precipitation coincidence (Liu & Zheng ; Yan & Chen ), and this paper used them due to
Pwnd ¼ Px,w ∧ Py,n ∧ Pz,d ¼ (X Xw , Yd YYw , ZZd ) ¼ Cθ (vw , ωw ) Cθ (vd , ωd ) Cθ (uw , vw , ωw ) þ Cθ (uw , vw , ωd )
high rainfall–runoff relation and simultaneous frequency.
(14)
For the annual runoff variable X, Xw and Xd were assigned the values of Pw ¼ 62.5% and Pd ¼ 37.5%, respect-
Dry–dry–dry periods coincidence probability, Pddd:
ively, as the threshold quantiles of wetness and dryness, respectively. The degree of wet–dry runoff can be described as wetness (X Xw ¼ X62.5%), dryness (X < Xd ¼ X37.5%) and normal (Xw > X Xd). In terms of bivariate copula, there were nine encounter situations (X and Y ), and the trivariate copula (X, Y and Z ) had 27 combinations. Some of all 36 combination functions were as follows.
¼ Cθ (ud , vd , ωd )
(15)
where u, v and ω are the marginal distributions of X, Y and Z, and subscripts w, n and d mean wet, normal and dry conditions. The composite letters note the coincidence
Wet–wet periods coincidence probability, Pww:
condition. For example, ddd presents the dry–dry–dry condition (period), and Pddd is the coincidence probability
Pww ¼ Px,w ∧ Py,w ¼ P(X Xw , Y Yw ) ¼ 1 uw vw þ Cθ (uw , vw )
Pddd ¼ Px,d ∧ Py,d ∧ Pz,d ¼ (X < Xd , YYd , ZZd )
(10)
of this condition. Here, the ∧ product notation is used to express the simultaneous occurrence of events. Other combination probabilities listed in Tables 3 and 4 were
Wet–dry periods coincidence probability, Pwd:
calculated in a similar way.
Pwd ¼ Px,w ∧ Py,d ¼ P(X Xw , Y < Yw ) ¼ vw Cθ (uw , vd )
(11)
RESULTS AND DISCUSSION
Normal–wet periods coincidence probability, Pnw:
Parameter estimation and fitting test of marginal distributions
Pnw ¼ Px,n ∧ Py,w ¼ P(Xd X < Xw , Y Yw ) ¼ uw vd Cθ (uw , vw ) þ Cθ (ud , vw )
(12)
Wet–wet–wet periods coincidence probability, Pwww: Pwww ¼ Px,w ∧ Py,w ∧ Pz,w ¼ (X Xw , Y Yw , Z Zw ) ¼ 1 uw vw ww þ Cθ (uw , vw ) þ Cθ (uw , ωw ) þ Cθ (vw , ωw ) Cθ (uw , vw , ωw ) (13) Table 3
|
The parameters of the P-III, Gumbel, and Lognormal distributions mentioned above were estimated using the linear moment method, and results are shown in Table 5. Also, the K–S test (Dn), RMSE and AIC were employed to test the feasibility of fitting these three distributions to runoff data, and results are shown in Figure 2. The statistical results in Figure 2 indicate that the P-III, Gumbel and Lognormal distributions all passed the K–S test
Nine coincidence combinations of bivariate copula Y
X
Wet
Normal
Dry
Wet
Pww ¼ Px,w ∧ Py,w
Pwn ¼ Px,w ∧ Py,n
Pwd ¼ Px,w ∧ Py,d
Normal
Pnw ¼ Px,n ∧ Py,w
Pnn ¼ Px,n ∧ Py,n
Pnd ¼ Px,n ∧ Py,d
Dry
Pdw ¼ Px,d ∧ Py,w
Pdn ¼ Px,d ∧ Py,n
Pdd ¼ Px,d ∧ Py,d
X. Wei et al.
1127
Table 4
|
|
Coincidence probability of streamflow for water transfer risk, China
Hydrology Research
|
51.5
|
2020
Twenty-seven coincidence combinations of trivariate copula Z
X
Y
Wet
Normal
Dry
Wet
Wet Normal Dry
Pwww ¼ Px,w ∧ Py,w ∧ Pz,w Pwnw ¼ Px,w ∧ Py,n ∧ Pz,w Pwdw ¼ Px,w ∧ Py,d ∧ Pz,w
Pwwn ¼ Px,w ∧ Py,w ∧ Pz,n Pwnw ¼ Px,w ∧ Py,n ∧ Pz,n Pwdn ¼ Px,w ∧ Py,d ∧ Pz,n
Pwwd ¼ Px,w ∧ Py,w ∧ Pz,d Pwnd ¼ Px,w ∧ Py,n ∧ Pz,d Pwdd ¼ Px,w ∧ Py,d ∧ Pz,d
Normal
Wet Normal Dry
Pnww ¼ Px,n ∧ Py,w ∧ Pz,w Pnnw ¼ Px,n ∧ Py,n ∧ Pz,w Pndw ¼ Px,n ∧ Py,d ∧ Pz,w
Pnwn ¼ Px,n ∧ Py,w ∧ Pz,n Pnnn ¼ Px,n ∧ Py,n ∧ Pz,n Pndn ¼ Px,n ∧ Py,d ∧ Pz,n
Pnwd ¼ Px,n ∧ Py,w ∧ Pz,d Pnnd ¼ Px,n ∧ Py,n ∧ Pz,d Pndd ¼ Px,n ∧ Py,d ∧ Pz,d
Dry
Wet Normal Dry
Pdww ¼ Px,d ∧ Py,w ∧ Pz,w Pdnw ¼ Px,d ∧ Py,n ∧ Pz,w Pddw ¼ Px,d ∧ Py,d ∧ Pz,w
Pdwn ¼ Px,d ∧ Py,w ∧ Pz,n Pdnn ¼ Px,d ∧ Py,n ∧ Pz,n Pddn ¼ Px,d ∧ Py,d ∧ Pz,n
Pdwd ¼ Px,d ∧ Py,w ∧ Pz,d Pdnd ¼ Px,d ∧ Py,n ∧ Pz,d Pddd ¼ Px,d ∧ Py,d ∧ Pz,d
Table 5
|
Estimated parameters of P-III, Gumbel and Lognormal distributions
P-III
Gumbel δ
γ
Lognormal
Gauges
α
β
k
μ
σ
SQ
4.0371
71.4811
38.9582
110.7327
263.6216
327.5376
142.0196
XY
8.1332
7.2092
8.7950
15.8512
40.6888
49.8382
20.3298
DSDJK
3.4689
73.1856
114.2435
105.0921
307.4579
368.1180
134.7853
1:36 with the critical value D0:05 ¼ pffiffiffiffiffiffi ¼ 0:2027. At SQ gauge, it 45 was seen that the AIC of the Gumbel distribution was the
tau (τ) are first calculated to describe the dependence (con-
smallest of the three models, while the P-III distribution
and the values of R and τ are shown in Table 6.
had the smallest RMSE. After consideration, the P-III distri-
For this, Pearson’s correlation coefficient (R) and Kendall’s cordance) of annual runoff for any two of the three gauges, In Table 6, it is seen that runoff series from any two of
bution was finally selected to fit the annual runoff data from
the three gauges had positive correlation, thus the G–H
the SQ gauge. For XY and DSDJK gauges, the Gumbel dis-
copula and Clyton cupola functions were considered suit-
tribution had both the smallest AIC and RMSE, thus the
able for the data series analysis.
Gumbel distribution was employed to fit at the two
Then, Kendall’s correlation coefficient τ was counted
gauges. The curves in Figure 2 represent the performances
between annual runoff data from any two of the three
of three models in fitting annual runoff data from the three
gauges, and the corresponding copula parameters θ can be
gauges visually.
computed in terms of the equations of the G–H and Clyton copulas listed in Table 2. The values of τ and θ are shown
Parameter estimation and fitting test of joint distribution
in Table 7. Furthermore, the bivariable joint distributions from the Clayton copula and G–H copula were constructed by
Various forms of the copula function have different require-
using the estimated parameters and the equations in
ments for the correlation of variables. For example, the
Table 2, respectively, while the parameters of trivariate Clay-
G–H copula and Clayton copula functions are suitable
ton and G–H copula functions need to be estimated through
for random variables with positive correlation, the Frank
Equation (2) for the construction of trivariate distributions,
copula function applicable to random variables with posi-
which is different from the bivariable joint distribution. So,
tive and negative correlations, and the AMH function is
the fitness results of all joint distributions, representing Dn,
often applied to random variables with weak correlations.
RMSE and AIC, are shown in Tables 8 and 9.
X. Wei et al.
1128
|
Coincidence probability of streamflow for water transfer risk, China
Hydrology Research
Table 6
|
|
51.5
|
2020
Pearson’s correlation coefficient (R) and Kendall’ tau (τ) between runoff series from three gauges
Gauges
R
τ
SQ–XY
0.7983
0.5993
SQ–DSDJK
0.8542
0.6721
XY–DSDJK
0.7355
0.4611
Table 7
|
Parameter values of bivariate copula functions
Θ Gauges
τ
G––H
Clayton
SQ–XY
0.5993
2.494
2.988
SQ–DSDJK
0.6721
3.049
4.098
XY–DSDJK
0.4611
1.855
1.711
Tables 8 and 9 show that bivariate and trivariate copulas, based on the G–H and Clayton copula, both passed the K–S test (Dn less than 0.207). The smallest values of RMSE and AIC (0.0811, 224.12) in Table 8 demonstrate that Clayton copula was the best-fitted bivariable joint distribution between runoff series from SQ and XY gauges and was also suitable for that between XY and DSDJK gauges, with RMSE and AIC of 0.0995 and 205.71. Differently, the best-fitted bivariable joint distribution between SQ and DSDJK gauges was the G–H copula, and the smallest values of RMSE and AIC were 0.0700 and 237.33. Additionally, it can be seen from Table 9 that the smallest values of RMSE and AIC are 0.1646 and 212.66, respectively, implying the G–H copula was the best suitable function among trivariate copula functions and can describe the joint distribution among three gauges. Coincidence probability of bivariate copula According to the chosen and established bivariate copula functions, the bivariate joint runoff distribution between any two gauges was obtained, and the joint coincidence probability was calculated under certain conditions. The joint cumulative probability distribution and contours at any two of the three coupled gauges are shown in Figure 3. From the contours in Figure 3, one can obtain the probabilities that annual runoff of any two gauges was less than a certain value at the same time, and that possible combiFigure 2
|
Marginal distributions fitted with empirical frequency and models.
nations of annual runoff from different gauges under a
X. Wei et al.
1129
Table 8
|
|
Coincidence probability of streamflow for water transfer risk, China
Hydrology Research
|
51.5
|
2020
K–S test (Dn), RMSE and AIC of bivariate copulas Dn
RMSE
AIC
Gauges
G–H
Clayton
G–H
Clayton
G–H
Clayton
SQ–XY
0.1600
0.1280
0.0854
0.0811
219.47
224.12
SQ–DSDJK
0.1279
0.1086
0.0700
0.0701
237.33
237.16
XY–DSDJK
0.1840
0.1472
0.1016
0.0995
203.82
205.71
Table 9
|
K–S test (Dn), RMSE and AIC of trivariate copulas
probabilities generally pointed to wet–wet periods coincidence probability, normal–normal periods coincidence
Trivariate copula functions
Θ
Dn
RMSE
AIC
G–H copula
3.49
0.0921
0.1646
212.66
Clayton copula
3.27
0.1160
0.1648
191.85
probability and dry–dry coincidence probability, while other probabilities were regarded as asynchronous coincidence probabilities. The synchronous and asynchronous coincidence probabilities at any two gauges were computed,
certain probability, such as the combination of annual runoff from the SQ gauge with a 30% probability and corresponding annual runoff from the XY gauge, or the combination of annual runoff from the XY gauge with a 30% probability and annual runoff from the SQ gauge. At this time, the joint cumulative probability in Figure 3 was able to analyze the worst situation of simultaneous dry periods occurring from any two gauges. For instance, it can be calculated that the joint probability was 10%, that the annual runoff was less than 2 × 1010 m3 at SQ gauge and less than 0.27 × 1010 m3 at XY gauge; that the joint probability was 90%, that the annual runoff was less than 0.86 × 1010 m3 at XY gauge and less than 7 × 1010 m3 at DSDJK gauge. For other examples, the possible combination of annual runoff at SQ and DSDJK gauges with a joint probability less than 50% was that the annual runoff was less 10
than 5 × 10
3
10
m at SQ gauge and less than 3.5 × 10
3
m at
DSDJK gauge, or that the annual runoff was less than 10
3.05 × 10
3
10
m at SQ gauge and less than 6 × 10
3
m at
DSDJK gauge. Through the constructed bivariate joint distributions, the nine coincidence probabilities of annual runoff data mentioned in Table 3 were computed at any two gauges. In the IBT project, the most adverse encounter situation
as shown in Table 10. Obviously, it is seen from the table that the synchronous coincidence probability was generally higher than asynchronous coincidence probability. It can be explained well due to the close geographical position among three gauges and similar climatic conditions. The larger synchronous coincidence probabilities indicated when there was enough water to be diverted in WSA, WRA or DIA was also likely to be wet, implying water demand in WRA for transferring water was small or the potential influence caused by the water transfer in DIA was low. Contrarily, when WSA was dry and cannot provide abundant water for diversion, water demand in WRA for transferring water was very high or the potential influence in DIA was greatly strong due to WRA or DIA being probably dry at that time. This was not expected in the IBT project, because it took on the project with a big risk, threatening the water security in WSA, WRA and DIA. Of course, when there was enough water to be diverted in WSA, WRA just happened to be dry and DIA was wet, and transferring water in the interbasin provided the benefit at this time and caused the low impact in the downstream river basin. Thus, it was clear that coincidence probability of bivariate copula failed to provide a valuable reference when considering WSA, WRA and DIA of the IBT project
was considered as dry–dry periods, which means that a shortage of water resources appeared both in the WSA
Coincidence probability of trivariate copula
and the WRA. The statistical results showed that Pdd was not too high, 30, 29.19 and 26.49% in SQ–XY, XY–DSDJK
The coincidence probability of trivariate copula certainly
and SQ–DSDJK, respectively. The synchronous coincidence
provided an effective tool to comprehensively analyze the
X. Wei et al.
1130
Figure 3
|
|
Coincidence probability of streamflow for water transfer risk, China
Hydrology Research
|
51.5
|
2020
Contours and joint cumulative probability distribution: (a) joint cumulative probability distribution of annual runoff from the SQ and XY gauges; (b) contours of the joint cumulative probability distribution of annual runoff from the SQ and XY gauges; (c) joint cumulative probability distribution of annual runoff from the SQ and DSDJK gauges; (d) contours of the joint cumulative probability distribution of annual runoff from the SQ and DSDJK gauges; (e) joint cumulative probability distribution of annual runoff from the XY and DSDJK gauges; (f) contours of the joint cumulative probability distribution of annual runoff from the XY and DSDJK gauges.
X. Wei et al.
1131
Table 10
|
|
Coincidence probability of streamflow for water transfer risk, China
SQ– XY
XY– DSDJK
SQ– DSDJK
Synchronous probability Wet–wet Normal–normal Dry–dry Total
26.78 10.50 30.00 67.29
30.43 12.06 29.19 71.68
23.33 8.38 26.49 58.19
Asynchronous Probability
8.86 1.86 8.86 5.64 1.86 5.64 32.72
5.85 1.22 5.85 7.09 1.22 7.09 28.32
9.89 4.28 9.89 6.73 4.28 6.73 41.8
Periods
Wet–normal Wet–dry Normal–wet Normal–dry Dry–wet Dry–normal Total
|
51.5
|
2020
safety in Guanzhong plain, China. Considering this, the
Coincidence probability of bivariate copula (%)
Probability
Hydrology Research
wet–wet–wet and dry–dry–dry encounter situations were determined as unfavorable circumstances, in which the wet–wet–wet situation implying flood risk and the dry– dry–dry situation implying drought risk in this study. Other encounter situations were also regarded as acceptable conditions. These results are shown in Table 11 where the values of synchronous coincidence probabilities Pwww, Pnnn and Pddd were 29.08, 7.99 and 26.09%, respectively. Therein, the most adverse encounter situation was in dry– dry–dry periods, with the coincidence probability (Pddd) of 26.09%. Through the probability, the recurrence interval was obtained in the worst encounter situation, and the value was 3.8 years, which means the most adverse situation
influence of the IBT project. Based on the G–H copula, this
(dry–dry–dry) could appear once every 3.8 years on average.
study constructed trivariate joint distributions, by which
Also, it can be seen from the table that unfavorable prob-
27 coincidence probabilities from the trivariate copula,
ability (Pwww þ Pddd) reached 55.17%, greater than the
corresponding to the combinations shown in Table 4, were
acceptable probability, implying the hydrological condition
computed.
Similarly,
probability
Pwww,
wet–wet–wet
coincidence
was very unfavorable to water transfer in more than half
normal–normal–normal
the
coincidence
of the whole operation period. It undoubtedly posed a big
probability Pnnn, and dry–dry–dry coincidence probability
challenge to the scheduling of inter-basin projects, threaten-
Pddd were considered as synchronous coincidence probabil-
ing the safety of water delivery.
ities of trivariate copula, while other probabilities were
Furthermore, in order to implement the optimal schedul-
asynchronous coincidence probabilities. Furthermore, syn-
ing of the transfer project, it was necessary to identify the
chronous coincidence probabilities and asynchronous
possible runoff in WRA and DIA when different runoff
coincidence probabilities at three gauges (SQ–XY–DSDJK)
rhythms happened in WSA. Fortunately, the joint distribution
can be obtained, as shown in Table 11.
model established by the copula function provided an effec-
As is known, the purpose of HWWTP was to balance
tive tool for solving this problem. So three equations were
the non-uniform temporal and spatial distributions of
constructed to calculate dry–dry coincidence probabilities
water resources in WSA and WRA and increase water
of annual runoff from the XY and DSDJK gauges with the
supply for guaranteeing industrial and domestic water
condition that SQ gauge was in wet, normal and dry periods.
Table 11
|
Coincidence probability of trivariate copula (%)
Wet (SQ)
Normal (SQ)
Dry (SQ)
Periods
Wet (XY)
Normal (XY)
Dry (XY)
Wet (XY)
Normal (XY)
Dry (XY)
Wet (XY)
Normal (XY)
Dry (XY)
Wet (DSDJK)
29.08
2.23
0.05
2.23
2.78
0.34
0.05
0.34
0.39
Normal (DSDJK)
2.23
2.78
0.34
2.78
7.99
2.40
0.34
2.40
3.75
Dry (DSDJK)
0.05
0.34
0.39
0.34
2.40
3.75
0.39
3.75
26.09
Sum of synchronous probability (Pwww þ Pnnn þ Pddd)
63.16
Sum of asynchronous probability (Pother)
36.84
Unfavorable probability (Pwww þ Pddd)
55.17
Acceptable probability (Pother)
44.83
1132
X. Wei et al.
|
Coincidence probability of streamflow for water transfer risk, China
Hydrology Research
|
51.5
|
2020
The conditional dry–dry coincidence probabilities were
the conditional coincidence probability of a certain magni-
obtained as follows, where X notes runoff from the SQ
tude was obtained directly from the contours. For
gauge, Y is that from the XY gauge and Z refers to the
example, when SQ gauge was dry, the conditional coinci-
runoff from the DSDJK gauge.
dence probability of dry XY and dry DSDJK was 20%, in which dry XY illustrated the annual runoff at XY gauge
Wet SQ, dry XY and dry DSDJK,
was less than 0.5 × 1010 m3 and dry DSDJK demonstrated that it was less than 2.2 × 1010 m3.
F(y, zjx) ¼ P(Y < Yd , Z < Zd jX Xw ) ¼
Cθ (vd , ωd ) Cθ (uw , vd , ωd ) ¼ 0:0104 1 uw
(16)
Also, the conditional coincidence probability of the most adverse encounter situation (dry A and dry B if dry C) for the IBT project was 69.57%, representing the encounter prob-
Normal SQ, dry XY and dry DSDJK,
ability of dry XY and dry DSDJK in the case that SQ was in the dry period. Compared with the coincidence probability
F(y, zjx) ¼ P(Y < Yd , Z < Zd jXd X < Xw ) ¼
Cθ (un , vd , ωd ) Cθ (ud , vd , ωd ) ¼ 0:4483 un ud
(dry SQ and dry XY and dry DSDJK) of trivariate copula in (17)
Table 11, the conditional coincidence probability (dry XY and dry DSDJK if dry SQ) better showed the importance of drought risk, indicating the dry–dry–dry coincidence prob-
Dry SQ, dry XY and dry DSDJK,
ability in the study area was close to 70% when the drought event occurred in WSA. Thus, it should be emphasized
F(y, zjx) ¼ P(Y < Yd , Z < Zd jX < Xd ) ¼
Cθ (ud , vd , ωd ) ud
¼ 0:6957
again that the implementation effect of designed water transfer could not be guaranteed to a large extent, once WSA was (18)
in a dry period during the operation of HWWTP.
where u, v and ω are the marginal distributions of X, Y and Z, and subscript w, n and d mean the wet, normal and dry
CONCLUSIONS
conditions, respectively. The above calculation showed that the coincidence
The Han River to Wei River Water Transfer Project
probability of wet SQ, dry XY and dry DSDJK was 1.04%;
(HWWTP) is a strategic project serving the water resources
that of normal SQ, dry XY and dry DSDJK was 44.83%;
allocation in northwest China, aiming to mitigate water
that of dry SQ, dry XY and dry DSDJK was 69.57%. Therein,
shortage in the Guanzhong Plain. It is important for
the probability that dry XY and dry DSDJK happened at the
HWWTP to estimate the water supply risk of the water transfer
same time reached 69.57%, in the case that dry SQ was
project and explore its potential impact on the downstream of
occurring, and that of dry XY and dry DSDJK arrived at
the water sources areas. To discuss this issue, this study ana-
44.83% even though normal runoff occurred at SQ gauge.
lyzed the coincidence probability of annual runoff series
They implied that water supply deficit in WRA due to the
between or among the water source area (WSA), WRA and
insufficient transfer of water caused by low watershed
the DIA. In this analysis, the Archimedean copula-based
yield in WSA, and water supply deficit in DIA due to low
method was used to capture the encounter situations of var-
flow from the upstream watershed caused by normal water
ious water yields in different areas by constructing bivariate
transfer, could disturb the operation of the HWWTP,
and trivariate joint distributions and calculated various combi-
when there was not abundant water yield in WSA.
nation frequencies, implying the water supply risk of the
To visualize the conditional coincidence probability among three gauges mentioned above, the distribution and
HWWTP in China. The main conclusions of the study can be summarized as follows.
contours of different magnitudes of runoff from XY and
The coincidence probability of annual runoff series from
DSDJK gauges in the case that SQ gauge was in dry
WSA (SQ gauge), WRA (XY gauge) and DIA (DSDJK gauge)
period were drawn, as shown in Figure 4. From the figure,
was obtained for different encounter situations by using
X. Wei et al.
1133
Figure 4
|
|
Coincidence probability of streamflow for water transfer risk, China
Hydrology Research
|
51.5
|
2020
Conditional coincidence probability of XY–DSDJK given the runoff of SQ gauge.
multivariate copula functions, which provided a valuable refer-
According to the coincidence probability of annual runoff
ence for water resources scheduling of the IBT project because
series from the three gauges, it was found that the acceptable
they covered all primary areas impacted by the project.
and unfavorable probabilities were 44.83 and 55.17%,
The most adverse encounter in all situations was the
respectively. Obviously, the unfavorable probability was
case that three hydrological gauges (WSA, WRA and DIA)
greater than the acceptable probability and should be paid
were in the dry period at the same time, i.e. dry–dry–dry,
greater attention to flood and drought risk management
and the coincidence probability was 26.09% obtained by
during the operation of the project.
the trivariate copula. Accordingly, the recurrence interval
Considering the most adverse encounter situation in
was calculated with a value of 3.8 years, illustrating that
WRA (XY gauge) and DIA (DSDJK gauge) under the con-
this adverse situation would appear about every 4 years.
dition that WSA (SQ gauge) was in the dry period, i.e. dry
1134
X. Wei et al.
|
Coincidence probability of streamflow for water transfer risk, China
XY and dry DSDJK if dry SQ, the conditional coincidence
Hydrology Research
|
51.5
|
2020
REFERENCES
probability was computed with a value of 69.57%. Different from the coincidence probability (dry–dry–dry) of trivariate copula, it better emphasized the risk of drought for the water transfer project, representing the probability of drought event simultaneously occurring in WRA and DIA was close to 70% when a drought event occurred in WSA. Therefore, it was inferred that the volume of designed transferable water could not be guaranteed to a large extent once WSA was in a dry period during the operation of HWWTP, which could lead to water shortages caused by drought in WRA and DIA could not be alleviated effectively and could further threaten regional water security. Thus, it is suggested that an emergency water resources scheduling plan and management programs should be drawn up, with the purpose of dealing with potential flood and drought risk of the IBT project.
ACKNOWLEDGEMENTS The authors gratefully acknowledge the financial support from the National Natural Science Foundation of China (No. 51979005 and 51809005), the Natural Science Basic Research Program of Shaanxi (No. 2020JM-250) and State Key Laboratory of Eco-hydraulics in Northwest Arid Region, Xi’an University of Technology (No. 2018KFKT-4).
AUTHOR CONTRIBUTIONS All authors contributed to this study. H.Z. provided the writing idea; X.W. carried out data analyses and wrote the first manuscript draft; C. D. and S. S. contributed to the analysis of results; Y.W. drew the figures; H.Z. and V.S. revised the manuscript. All authors read and approved the final manuscript.
DATA AVAILABILITY STATEMENT All relevant data are included in the paper or its Supplementary Information.
Chanda, K., Maity, R., Sharma, A. & Mehrotra, R. Spatiotemporal variation of long-term drought propensity through reliability-resilience-vulnerability based drought management index. Water Resources Research 50 (10), 7662–7676. doi:10.1002/ 2014WR015703. Chebana, F., Dabo-Niang, S. & Ouarda, T. B. M. J. Exploratory functional flood frequency analysis and outlier detection. Water Resources Research 48 (4), 4514. doi:10. 1029/2011WR011040. Chen, L., Singh, V. P., Guo, S., Zhou, J. & Zhang, J. Copulabased method for multisite monthly and daily streamflow simulation. Journal of Hydrology 528 (528), 369–384. doi:10. 1016/j.jhydrol.2015.05.018. Deng, W. B., Liu, W. J., Li, X. X. & Yang, Y. G. Source apportionment of and potential health risks posed by trace elements in agricultural soils: a case study of the Guanzhong Plain, Northwest China. Chemosphere. doi:10.1016/j. chemosphere.2020.127317. Du, H., Wang, Y., Liu, K. & Cheng, L. Exceedance probability of precipitation for the Shuhe to Futuan Water Transfer Project in China. Environmental Earth Sciences 78 (7), 1–12. doi:10.1007/s12665-019-8207-2. Fang, H. B., Fang, K. T. & Kotz, S. The meta-elliptical distributions with given marginals. Journal of Multivariate Analysis 82 (1), 1–16. doi:10.1006/jmva.2001.2017. Favre, A. C., Adlouni, S. E., Perreault, L., Thiemonge, N. & Bobee, B. Multivariate hydrological frequency analysis using copulas. Water Resources Research 40 (1), 1–12. doi:10.1029/ 2003WR002456. Gringorten, I. I. A plotting rule for extreme probability paper. Journal of Geophysical Research 68 (3), 813–814. doi:10. 1029/JZ068i003p00813. He, X. C., Kang, L., Cheng, X. J. & Ding, Y. Flood risk analysis in the middle route of South-to-North Water Diversion Project of China based on Bayesian network. South-to-North Water Diversion and Water Science & Technology 10 (4), 10–13. doi:10.3724/sp.j.1201.2012.04010. Hofert, M. Sampling Archimedean copulas. Computational Statistics & Data Analysis 52 (12), 5163–5174. doi:10.1016/j. csda.2008.05.019. Hofert, M. Efficiently sampling nested Archimedean copulas. Computational Statistics & Data Analysis 55 (1), 57–70. doi:10.1016/j.csda.2010.04.025. Hosking, J. R. L-moments: analysis and estimation of distributions using linear combinations of order statistics. Journal of the Royal Statistical Society Series B – Methodological 52 (1), 105–124. doi:10.1111/j.2517-6161. 1990.tb01775.x. Joe, H. Multivariate models and dependence concepts. Journal of the American Statistical Association 93 (443). doi:10.2307/2669872.
1135
X. Wei et al.
|
Coincidence probability of streamflow for water transfer risk, China
Joe, H. Asymptotic efficiency of the two-stage estimation method for copula-based models. Journal of Multivariate Analysis 94 (2), 401–419. doi:10.1016/j.jmva.2004.06.003. Lan, T., Zhang, H., Xu, C. Y., Singh, V. P. & Lin, K. Detection and attribution of abrupt shift in minor periods in humanimpacted streamflow. Journal of Hydrology 584, 124637. doi:10.1016/j.jhydrol.2020.124637. Li, S., Gu, S., Liu, W., Han, H. & Zhang, Q. Water quality in relation to the land use and land cover in the Upper Han River basin, China. Catena 75 (2), 216–222. doi:10.1016/j. catena.2008.06.005. Li, S., Liu, W., Gu, S., Cheng, X., Xu, Z. & Zhang, Q. Spatiotemporal dynamics of nutrients in the upper Han River basin, China. Journal of Hazardous Materials 162 (2), 1340–1346. doi:10.1016/j.jhazmat.2008.06.059. Liu, C. & Zheng, H. South-to-north water transfer schemes for China. International Journal of Water Resources Development 18 (3), 453–471. doi:10.1080/ 0790062022000006934. Liu, X., Luo, Y., Yang, T., Liang, K., Zhang, M. & Liu, C. Investigation of the probability of concurrent drought events between the water source and destination regions of China’s water diversion project. Geophysical Research Letters 42 (20), 8424–8431. doi:10.1002/2015GL065904. Manshadi, H. R. D., Niksokhan, M. H. & Ardestani, M. Water allocation in inter-basin water transfer with the virtual water approach. In: World Environmental & Water Resources Congress: Showcasing the future, American Society of Civil Engineers, Reston, VA, pp. 2510–2521. doi:10.1061/ 9780784412947.247. Massey, F. J. The Kolmogorov-Smirnov test for goodness of fit. Journal of the American Statistical Association 46 (253), 68–78. doi:10.2307/2280095. Nelsen, R. B. An Introduction to Copulas. Springer, New York. doi:10.2307/1271100. Plackett, R. L. A class of bivariate distributions. Journal of the American Statistical Association 60 (310), 516. doi:10.1080/ 01621459.1965.10480807. Salvadori, G. & De Michele, C. On the use of copulas in hydrology: theory and practice. Journal of Hydrologic Engineering 12 (4), 369–380. doi:10.1061/(ASCE) 10840699(2007)12:4(369). Serinaldi, F., Bonaccorso, B., Cancelliere, A. & Grimaldi, S. Probabilistic characterization of drought properties through
Hydrology Research
|
51.5
|
2020
copulas. Physics & Chemistry of the Earth Parts A/B/C 34 (10–12), 596–605. doi:10.1016/j.pce.2008.09.004. Sklar, C. A. Fonctions de répartition à n dimensions et leurs marges. Publication de l’Institut de Statistique de l’Université de Pair 8, 229–231. Sraj, M., Bezak, N. & Brilly, M. Bivariate flood frequency analysis using the copula function: a case study of the Litija gauge on the Sava River. Hydrological Process 29 (2), 225–238. doi:10.1002/hyp.10145. Tian, H., Wang, P., Tansey, K., Zhang, S. & Li, H. An IPSOBP neural network for estimating wheat yield using two remotely sensed variables in the Guanzhong Plain, PR China. Computers and Electronics in Agriculture 169, 105180. doi:10.1016/j.compag.2019.105180. Wu, L., Bai, T., Huang, Q., Wei, J. & Liu, X. Multi-objective optimal operations based on improved NSGA-II for Hanjiang to Wei River Water Diversion Project, China. Water 11 (6). doi:10.3390/w11061159. Xie, H. & Wang, K. Joint-probability methods for precipitation and flood frequencies analysis. Third International Conference on Intelligent System Design & Engineering Applications, Los Alamitos, CA. doi:10.1109/ ISDEA.2012.217. Yan, B. & Chen, L. Coincidence probability of precipitation for the middle route of South-to-North water transfer project in China. Journal of Hydrology 499, 19–26. doi: 10.1016/j. jhydrol.2013.06.040. Zhang, L. & Singh, V. P. Bivariate flood frequency analysis using the copula method. Journal of Hydrologic Engineering 11 (2), 150–164. doi:10.1061/(ASCE)1084-0699(2006)11:2(150). Zhang, L. & Singh, V. P. Gumbel-Hougaard copula for trivariate rainfall frequency analysis. Journal of Hydrology 12 (4), 409–419. doi:10.1061/(ASCE)1084-0699(2007)12:4(409). Zhang, L. & Singh, V. P. Copulas and Their Applications in Water Resources Engineering. Cambridge University Press, Cambridge, UK. Zhuang, W. Eco-environmental impact of inter-basin water transfer projects: a review. Environmental Science & Pollution Research. 23 (13), 12867–12879. doi:10.1007/ s11356-016-6854-3. Zou, L., Xia, J. & She, D. Analysis of impacts of climate change and human activities on hydrological drought: a case study in the Wei River Basin, China. Water Resources Management 32 (4), 1421–1438. doi:10.1007/s11269-017-1877-1.
First received 27 July 2020; accepted in revised form 17 August 2020. Available online 14 September 2020
1136
© 2020 The Authors Hydrology Research
|
51.5
|
2020
The exploration of a Temporal Convolutional Network combined with Encoder-Decoder framework for runoff forecasting Kangling Lin, Sheng Sheng, Yanlai Zhou, Feng Liu, Zhiyu Li, Hua Chen, Chong-Yu Xu
, Jie Chen and Shenglian Guo
ABSTRACT The Temporal Convolutional Network (TCN) and TCN combined with the Encoder-Decoder architecture (TCN-ED) are proposed to forecast runoff in this study. Both models are trained and tested using the hourly data in the Jianxi basin, China. The results indicate that the forecast horizon has a great impact on the forecast ability, and the concentration time of the basin is a critical threshold to the effective forecast horizon for both models. Both models perform poorly in the low flow and well in the medium and high flow at most forecast horizons, while it is subject to the forecast horizon in forecasting peak flow. TCN-ED has better performance than TCN in runoff forecasting, with higher accuracy, better stability, and insensitivity to fluctuations in the rainfall process. Therefore, TCN-ED is an effective deep learning solution in runoff forecasting within an appropriate forecast horizon. Key words
| artificial neural network, deep learning, Encoder-Decoder architecture, runoff forecasting, Temporal Convolutional Network
Kangling Lin Sheng Sheng Hua Chen (corresponding author) Jie Chen Shenglian Guo State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University, Wuhan 430072, China and Hubei Provincial Key Lab of Water System Science for Sponge City Construction, Wuhan University, Wuhan 430072, China E-mail: chua@whu.edu.cn Yanlai Zhou Chong-Yu Xu Department of Geosciences, University of Oslo, P O Box 1047, Blindern, N-0316, Oslo, Norway
HIGHLIGHTS
• • • • •
For the first time, TCN and TCN-ED models are proposed to forecast runoff. TCN-ED has better performance than TCN in runoff forecast in this study.
Feng Liu School of Computer Science, Wuhan University, Wuhan 430072, China
The concentration time is a critical threshold to the effective forecast horizon. Both models perform better in median and high flow than in low flow. It is subject to the forecast horizon for both models to forecast peak flow.
Zhiyu Li Hua Chen Department of Geography and Geographic Information Science, University of Illinois at Urbana-Champaign, Champaign, IL 61801, USA
INTRODUCTION Runoff forecasting is of considerable significance to water
operation, flood control, drought relief, and navigation.
resources management. Accurate runoff forecasting can
According to the extent of physical principles, models for
guide the hydraulic engineering construction, reservoir
runoff forecasting can be divided into two categories: process-driven models and data-driven models (Yuan
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying,
et al. ). Process-driven models represent a specific
adaptation and redistribution, provided the original work is properly cited
physical process employing experimental formulas before
(http://creativecommons.org/licenses/by/4.0/).
inputting data. Because of the valuable interpretability of
doi: 10.2166/nh.2020.100
1137
K. Lin et al.
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Hydrology Research
|
51.5
|
2020
process-driven models, they have been widely used by
and ant lion optimizer model in the prediction of monthly
hydrologists (Beven et al. ; Ren-Jun ; Douglas-
runoff. These studies prove that the ANNs, especially the
Mankin et al. ; Wang et al. ). However, given the
recurrent neural network (RNN), can capture the non-
uncertainty of the hydrological process and the limitations
linear and non-stationary dynamics of river flow effectively
of artificially constructed process-driven models, the par-
and stand out among multiple data-driven models.
ameters
and
simulation
process
are
challenging
to
Nevertheless, the ANN research on runoff forecasting
represent hydrological phenomena fully. Due to process-
mainly focused on the RNN (Chang et al. ), including
driven models’ physical meaning, researchers without a pro-
the LSTM (Hu et al. ) and further improvement based
fessional background cannot improve the models, resulting
on the LSTM (Yuan et al. ). The reason is conjectured
in slow model iteration and difficulty in introducing new
to be that the RNN, particularly the LSTM, has a promising
technologies. With the improvement of data availability
ability to process time-series data. However, the develop-
and quality, data-driven models can substitute or sup-
ment of ANNs is changing soon with each passing day,
plement to process-driven models for runoff forecasting
and there are oncoming ANNs worthy of applying for
(Yuan et al. ). Data-driven models focus on the optimal
hydrological modeling and runoff forecasting. The Temporal
mathematical relationships between a forecast object and a
Convolutional Network (TCN) is one of the novel ANNs
predictor, without considering the physical mechanism
designed for sequence modeling and prediction (Bai et al.
(Adamowski & Sun ). Therefore, data-driven models
). TCN is essentially a combination of the one dimen-
are highly transferable.
sion fully convolutional network (1D FCN) and causal
The artificial neural network (ANN), a popular data-
convolutions (Bai et al. ). The 1D FCN guarantees the
driven model, was applied in runoff modeling at the end
length of the input and output of TCN can be kept the
of the last century already (Hsu et al. ; Smith & Eli
same. The causal convolutions ensure that the future infor-
; Shamseldin ; Dawson & Wilby ). However,
mation will not be used during convolutions. The dilated
due to the limitation of theoretical research and computing
convolutions and residual modules, which dramatically
power at that time, the structure of ANNs was relatively
increase the receptive field of the convolutional neural net-
simple. Therefore, ANNs were tough to obtain high
work and make it easier to train, respectively, are also
accuracy results and dwarfed by another representative
used in TCN. With the empirical study demonstrated by
data-driven
(SVM).
Bai et al. (), TCN has been proved superior for sequence
Sivapragasam et al. () used the prediction technique
modeling tasks to LSTM. Recently, more and more studies
based on singular spectrum analysis coupled with SVM to
have confirmed the ability of TCN in time-series data proces-
predict the Tryggevælde catchment runoff data. Further-
sing such as the stock trend prediction, anomaly detection,
more, they used the flood data at Dhaka to demonstrate
and recognition of sepsis (Deng et al. ; He & Zhao
the forecast ability of SVM superior to that of the
; Moor et al. ). Therefore, TCN is introduced for
ANN-based model, particularly at longer lead days (Liong
the first time in hydrology by this research to provide
& Sivapragasam ).
more options for runoff forecasting based on ANNs.
model,
support
vector
machine
With the rise of deep learning, hydrologists have turned
Besides, most studies mentioned above considered all
to study the application of advanced deep ANNs in hydrolo-
hydrological information contained in input as equal factors
gical modeling and forecasting (Kisi ; Awchi &
for forecasting the runoff sequences (global relationship)
Srivastava ; Trajkovic ; Abdellatif et al. ; Baba
without further refining influence factors, therefore missing
et al. ; Shiri et al. ). For instance, Chang et al.
the interactions of the dependent variables (local relation-
() presented an ANN termed real-time recurrent learn-
ship) (Park et al. ). Recently, Kao et al. ()
ing for streamflow forecasting in the Da-Chia River. Hu
introduced Encoder-Decoder architecture in flood forecast-
et al. () applied a long short-term memory (LSTM) net-
ing by applying the LSTM based Encoder-Decoder model,
work to the Fen River basin and obtained good prediction
whose output sequence of models stepped into 1- up to
results. Yuan et al. () used the hybrid LSTM network
6-h-ahead. This is the first journal article on the application
1138
K. Lin et al.
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Hydrology Research
|
51.5
|
2020
of the Encoder-Decoder architecture in hydrology. The
As can be seen from Figure 1, there are 16 precipitation
Encoder-Decoder architecture is helpful to overcome some
stations, three evaporation stations, and seven runoff stations
obstacles in the application of ANNs on runoff forecasting,
in the study area. The hourly data of these monitoring sites
which has been successfully applied to many other similar
from 2009 to 2015 are obtained from the local Hydrology
themes such as the Shihmen Reservoir flood forecast (Kao
Bureau. The hourly runoff of the Qilijie station, the arithmetic
et al. ), Dadu River runoff forecast (Xu et al. ),
mean of the hourly precipitation, and hourly evaporation
and Yangtze River streamflow prediction (Liu et al. ).
among all corresponding stations in the basin, abbreviated
The Encoder-Decoder architecture is comprised of two
hereafter as the ‘runoff’, ‘precipitation’, and ‘evaporation’,
sub-models that work in a symbiotic manner (Loyola et al.
respectively, are used in this study. The runoff concentration
). An encoder captures the local relationship of influ-
time is about 12 h in this basin ( Jie et al. ).
ence factors by encoding the sequence of inputs. Then, a
In order to ease the negative effect of the different scales
decoder extracts the encoded vectors for the runoff forecast-
of data on models’ learning ability, the normal standardiz-
ing by jointly modeling the global and local relationships.
ation is applied to data preprocessing, which is defined as
Therefore, sequential regularities contained in a time series
follows:
can be learned in a more comprehensive, fine-grained way than using a single network (Bian et al. ). The objective of this study is to explore the ability and
X (t) ¼
X(t) X σ
(1)
stability of TCN and the integration of TCN with Encodermulti-step ahead times. To achieve this goal, the remainder
where X (t) is the normal standardization for input data in and σ are the tth time. X(t) is the input sequence, and X
of this study is organized as follows. The study area, hydro-
the mean and standard deviation of the input sequence,
logical data, and the data preprocessing method are given
respectively.
Decoder architecture (TCN-ED) for runoff forecasting with
first. TCN, the Encoder-Decoder architecture, the proposed data-driven runoff forecasting model, the comparison model, and metrics for the evaluation are then introduced.
METHODS
Next, the experimental results and thorough discussion are presented. Finally, the conclusions of this study are
Temporal Convolutional Network
summarized. The TCN, based on a 1D convolutional network, is a generic network structure for sequence modeling (Liu et al. ).
STUDY AREA AND DATA
With the empirical study demonstrated by Bai et al. (), TCN has been proved superior performance for sequence
The Jianxi River is the primary up-stream tributary of
modeling tasks to LSTM.
the Minjiang River in southeast China. The Qilijie station
In order to meet the requirements of sequence modeling
(118 180 16″E, 27 010 21″N) is located in the Jianxi River.
tasks, TCN utilizes causal convolutions. Therefore, outputs
The Qilijie station with a drainage area of 14,787 km2 is
are only influenced by present and past inputs in each
selected for this study (Jie et al. ). The primary soils in
layer (Moor et al. ). In addition, TCN uses a 1D FCN
this area are red, yellow, and paddy soils. The regional climate
structure, whose convolution layers have the same size as
is dominated by southeast Pacific Ocean and southwest
RNNs by adding zero paddings (He & Zhao ).
Indian Ocean subtropical monsoons and partly influenced
In order to increase the receptive field with less compu-
by regional landforms (Tang et al. ). The catchment is
tation cost, dilated convolutions have been introduced to
moist and rainy, with the mean annual rainfall from 1,800
TCN. The filter of a dilated convolution is applied over a
to 2,200 mm, most of which occurs from March to Septem-
region larger than its size by skipping input values with a
ber. The map of the study catchment is shown in Figure 1.
given
step.
The
dilated
factor
commonly
increases
K. Lin et al.
1139
Figure 1
|
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Hydrology Research
|
51.5
|
2020
Map of the study catchment.
exponentially with the depth of the network, which ensures
~ tþk is the target. ~ tþ1 , . . . , X X
the receptive field covering each input in the history (He & Zhao ). Moreover, TCN applies residual connections that combine the previous input and the result of the convolution with an addition to ease deterioration, which has been
~ tþk ¼ arg max ~ tþ1 , . . . , X X
^ tþk ^ tþ1 ,...,X X
^ tþ1 , . . . , X ^ tþk jXt jþ1 , Xt jþ2 , . . . , Xt ) p(X
(2)
proved very beneficial for deep networks (He et al. ). The encoder is to refine the information contained in the input and fix it into a context value. Encoder-Decoder architecture The Encoder-Decoder architecture was proposed by Cho et al. () who used RNNs both as the encoder and decoder. This architecture is comprised of two ANNs: the first one is to take the information as input and encode it into a context value. The second one is used for decoding the context value into the expected output sequence. The purpose of this architecture is to compress various
~ tþk ≈ arg max ~ tþ1 , . . . , X X
^ tþk ^ tþ1 ,...,X X
^ tþ1 , . . . , X ^ tþk jfencoder (Xt jþ1 , Xt jþ2 , . . . , Xt )) p(X
(3)
Then, the decoder decodes the context value and outputs the final prediction. ~ tþk ≈ gdecoder ( fencoder (Xt jþ1 , Xt jþ2 , . . . , Xt )) ~ tþ1 , . . . , X X (4)
information contained in the whole input sequences into a fixed-length vector to make tensor flow more stable (Xu et al. ).
TCN combined with Encoder-Decoder model
If the observation at the time step t is represented by Xt , the sequence to sequence forecasting problem is to predict ^ tþ1 , . . . , X ^ tþk whose length is k, accordthe next sequence X
TCN-ED is proposed in this study, which is compared with the TCN model to verify their ability in runoff forecasting.
ing to the previous j observation Xt jþ1 , . . . , Xt (Shi et al.
The precipitation, evaporation, and runoff for the first 48 h
). Among them, the most likely next sequence
are used as input to both models to forecast the runoff for
K. Lin et al.
1140
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Hydrology Research
|
51.5
|
2020
the next 24 h. The framework of the TCN-ED model
context value is copied 24 times, and then enters the
(Figure 2) is explained below.
second TCN (TCN decoder). Since the time step of the
As shown in Figure 2, the sample enters the temporal
second TCN’s input is half of the first TCN, the number of
block through 1 × 1 convolution. Each temporal block con-
temporal blocks is correspondingly reduced by one. In this
tains two layers of the causal convolutions. The zero
way, overfitting is eased on the premise that the receptive
paddings are added to make the output length between the
field of the TCN decoder can still cover the context value
layers the same, and the rectified linear unit is used to activate
sequence. The output of the TCN decoder at each time step
the output. The input of each temporal block is added to the
performs dimension reduction through two fully connected
output after 1 × 1 convolution with reference to Bai et al.
layers, and finally outputs the forecast runoff sequence.
(), to increase the nonlinearity of the residual link when
In this study, the output dimensions at each time step of
transferring information across layers. The dilation factor gen-
the TCN encoder and the TCN decoder are 256 and 128,
erally increases exponentially with the depth of the network.
respectively. The output dimensions at each time step of
Regarding the settings of Moor et al. (), for the first TCN
the first fully connected layer and the second fully con-
(TCN encoder), dilation factors correspond to temporal
nected layer are 64 and 1, respectively. With such a
blocks are 1, 2, 4, 8, and 16, respectively, to ensure that the
setting, the tensor flow among the network layers is hoped
receptive field can cover the input sequence. In order to
to be smoother under the premise of giving the models
meet the requirements of the output sequence length, the
enough parameter space.
Figure 2
|
The framework of the proposed TCN-ED. P, E, and Q represent the precipitation, evaporation, and runoff, respectively. f, k, and d represent the number of filters, filter size, and dilation factor, respectively.
1141
K. Lin et al.
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Flood forecasting model based on the proposed method
Hydrology Research
|
51.5
|
2020
128, respectively. The output dimensions of the first fully connected layer are 64. In order to output the runoff
To build a dataset to train and test the model, the prepro-
sequence with 24 forecast horizons, the output dimensions
cessed data from 2009 to 2015 are sorted in chronological
of the last fully connected layer are 24.
order, which are further divided into training samples with
In this study, Nash–Sutcliffe efficiency (NSE), volu-
a 5-year period and testing samples with a 2-year period.
metric efficiency (VE), and coefficient of determination
Therefore, the training and testing samples are provided
(R 2) (Dodge ) are employed as hydrological metrics.
with temporal continuity. Although the runoff concentration
NSE is a popular indicator in runoff forecasting defined by
time is about 12 h in the study basin, the forecast horizons
Nash & Sutcliffe (), but it is oversensitive to the extre-
are set to range from 1 to 24 h to detect the forecast ability
mal flood (Legates & McCabe ; Jiang et al. ). VE
of TCN and TCN-ED during different forecast horizons.
is an alternative indicator focusing on the measurement of
Each model continually trains and optimizes the performance
overall performance that addresses certain limitations of
by reducing the errors in the training set. The Adam optimizer
NSE (Criss & Winston ). The formulae of NSE and
(Kingma & Ba ) and mean square error (MSE) objective
VE are expressed as follows:
function (Allen ) are used in the training stage, and the number of epochs is set to 100. The test set is used to evaluate the performance of the models. The numerical calculations in this paper are accomplished on the supercomputing system in the Supercomputing Center of Wuhan University. Calculating the flow structure of each model is as follows: First, the data
P ^ 2 (Q Q) NSE ¼ 1 P Q)2 (Q VE ¼ 1
P ^ jQ Qj P Q
(5)
(6)
are preprocessed and divided into training data and testing tinuous 48 time step samples and 24 time step targets, input
^ is the predicted runoff, Q is the observed runoff, and where Q Q is the average of the observed runoff. Both indicators range
into the constructed TCN and TCN-ED, with MSE as the
from negative infinity to 1, where 1 represents a perfect fit
objective function, and Adam optimizer for model training
(Jiang et al. ). Four classes are chosen for each criterion:
in 100 epochs. Finally, evaluate the model at each forecast
‘very good’ when >0.66, ‘good’ when 0.33–0.66, ‘average’
horizon using the testing data.
when 0–0.33, and ‘poor’ when <0 (Ecrepont et al. ).
data. Second, the training data are further divided into con-
Since ANNs have strong randomness, all models are Evaluation of the model performance
recalculated 40 rounds (with different random initial weights) (Kao et al. ). With the repeated calculation
In order to objectively illustrate the improvement of the
results, the accuracy, stability, and robustness of each
Encoder-Decoder models compared with the ANN models
model are analyzed. The best model is used to forecast the
without the Encoder-Decoder architecture, TCN is set as
largest flood event in the testing stage, which is determined
the comparison model for TCN-ED with the same input
as the model that produces the highest NSE value averaging
and output. In order to make TCN as similar as possible
over 24 forecast horizons in the testing stages.
to TCN-ED, the comparison model stacks two TCN layers as well. The time steps of the first and second TCN layer are the same as the sample’s time steps, so the number of
RESULTS AND DISCUSSION
temporal blocks for both TCN layers is 5. The second TCN connects its output at the last time step to the two
Evaluation and comparison of models performances
fully connected layers for dimension reduction, and finally outputs the runoff sequence with forecast horizons from
In order to compare the performance of TCN and TCN-ED
t þ 1 to t þ 24. The output dimensions at each time step of
in the learning and forecasting phase, the minimum, mean,
the first TCN layer and the second TCN layer are 256 and
and maximum NSE and VE values averaging over 24
K. Lin et al.
1142
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Hydrology Research
|
51.5
|
2020
forecast horizons of each model in the training and testing
indicating that the Encoder-Decoder architecture can
stages with 40 rounds are shown in Table 1.
improve the stability with higher minimum accuracy.
As can be seen from Table 1, TCN and TCN-ED show
For evaluating the performance of each model intuitively
good performance, while TCN-ED maintains higher NSE
and holistically, Figure 3 shows Gaussian kernel density esti-
and VE values than TCN in the training and testing stages.
mation (GKDE) for all NSE and VE values of each model in
Especially for the minimum accuracy, TCN-ED has a
the training and testing stages, respectively. For the GKDE
4.48% improvement rate for NSE and a 4.65% improvement
curve, the accuracy when the density peak appears is the
rate for VE compared with TCN in the testing stages,
mode accuracy of the model, the sharpness of the curve
Table 1
|
The minimum, mean, and maximum NSE and VE values averaging over 24 forecast horizons of each model in the training and testing stages with 40 rounds
NSE (%)
VE (%)
TCN
TCN-ED
TCN
TCN-ED
Min
Mean
Max
Min
Mean
Max
Min
Mean
Max
Min
Mean
Max
Training
87.47
90.12
93.70
89.37
91.64
94.42
71.05
75.68
80.31
74.79
78.27
81.76
Testing
82.67
87.40
89.15
87.15
88.84
90.22
71.41
77.14
79.47
76.06
79.07
81.14
Figure 3
|
GKDE for NSE and VE values of TCN and TCN-ED in the (a) training and (b) testing stages.
K. Lin et al.
1143
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Hydrology Research
|
51.5
|
2020
represents the concentration of the accuracy. If the GKDE
The above phenomenon shows that TCN may overfit in the
curve is on the right of the X-axis and the shape is sharp, it
training stage, resulting in weaker transferability than TCN-
means that the model has stable and high-precision results.
ED whose GKDE curves change less between the training
As shown in Figure 3, distributions of NSE and VE
stage and the testing stage.
values of the two models are both left-skewed, i.e., the left tails of the GKDE curves are longer, the mass of distri-
Models performance with multi-step ahead times
butions are concentrated on the right, showing both models have a few relatively low values. For each subgraph,
The effective forecast horizons of ANNs have always been
the GKDE curve’s density peak of TCN is higher than that of
the bottleneck and difficulty of the artificial intelligence
TCN-ED. However, the value at the GKDE curve’s density
model research (Zhu et al. ; Cardenas-Barrera et al.
peak of TCN-ED is higher than that of TCN-ED, and the
; Claveria et al. ). In order to compare the accuracy
GKDE curve of TCN-ED has a longer tail in the high end
of TCN and TCN-ED models during different forecast hor-
than that of TCN, causing the GKDE curve of TCN-ED is
izons, boxplots for NSE and VE values in the testing stage
more rightward than that of TCN. Therefore, TCN-ED has
are shown in Figure 4.
a higher mode accuracy and more high-precision results
It can be seen from Figure 4 that both models perform
than TCN. Both models perform better under the VE indi-
very well as their NSE and VE values are very high and
cator from the training stage to the testing stage. However,
stable within t þ 12 forecast horizons, while exceeding the
the NSE GKDE curve’s density peak and the NSE value
t þ 12 forecast horizon, the forecast accuracy of both
at the density peak of TCN drop more than those of TCN-
models decreases gradually with the forecast horizon
ED from the training stage to the testing stage, respectively.
increasing. This shows that the forecast ability of the two
Figure 4
|
Boxplots for NSE and VE values of TCN and TCN-ED during forecast horizons from t þ 1 to t þ 24 in the testing stage.
K. Lin et al.
1144
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Hydrology Research
|
51.5
|
2020
models during different forecast horizons is closely related
seriously during the short forecast horizons. In comparison
to the concentration time of the basin, which needs to be
to TCN, TCN-ED’s context value which has been refined
further verified by more cases in more basins. It is evident
already ease the problem caused by redundant information
that TCN-ED’s forecast accuracy and stability are higher
and make TCN-ED more advantageous during the short fore-
than those of TCN at almost each forecast horizon, respect-
cast horizons. In addition, the context value is the output of
ively, especially for short forecast horizons up to t þ 12 and
the last time step of the encoder, so the learning memory is
long horizons close to t þ 24. In the process of the Encoder-
time-sensitive, and TCN-ED is superior obviously to TCN
Decoder, the encoder simulates the process of reading and
during long horizons close to t þ 24.
preprocessing in the brain. The context value with a specific
In flood control forecasting, more attention is paid to
length symbolizes the formed memory. The decoder rep-
the model’s ability to forecast large-scale floods. Therefore,
resents the phase when combining known memory and
the best model results of TCN and TCN-ED are used to
new information to react by the brain. Compared with
evaluate the largest flood event, whose peak is maximum
ANNs whose network layers are directly stacked, the Enco-
among 19 flood events in the testing stage at forecast hor-
der-Decoder ANNs are not only more conducive to our
izons t þ 6, t þ 12, t þ 18, and t þ 24, as shown in Figure 5.
understanding of the learning process but also make the
The largest flood event, which is considered moderately
tensor transmission between network layers more efficient
hazardous, has a maximal flow peak reaching 7,043 m3/s in
and stable. Therefore, TCN-ED has higher forecast accuracy
the study period, and the accumulated precipitation in the
and stability than TCN at almost every forecast horizon. For
basin during the flood rising period achieves 134 mm. It
TCN and TCN-ED, in order to keep the output length consist-
can be found from Figure 5 that because there are multiple
ent, the zero paddings are added, resulting in the redundant
rain peaks, the forecast of rising limb and flood peaks by
information which interferes with causal convolutions more
both models are unstable, causing the hydrograph jagged.
Figure 5
|
The observed and predicted floods with the maximum peak in the testing stage using the best model results of TCN and TCN-ED during horizons t þ 6, t þ 12, t þ 18, and t þ 24.
1145
K. Lin et al.
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Hydrology Research
|
51.5
|
2020
This situation becomes more and more evident as the
of the observations, respectively, which are drawn on the
forecast horizon increases. In terms of peak time, within
top of each subplot. The percentages of the predictions
t þ 12 forecast horizons, the forecast peak time of each
and their mean R 2 values over the 40 rounds in the five sec-
model is basically the same as the observed one, but at the
tions are shown on the right side of each subplot. R 2 values
t þ 24 forecast horizon, the models’ forecast peaks appear
of the whole observed and 40 predicted series in the testing
significantly later.
stage are calculated, and the maximum, median, and mini-
At the t þ 6 forecast horizon, the forecast runoff curves
mum R 2 values among 40 rounds are shown in each subplot.
of TCN and TCN-ED both fit the observed runoff curve
As Figure 6 shows, both TCN and TCN-ED have high R 2
well, which reflects the excellent forecast ability of TCN
values (>0.9) for the whole series at the t þ 6 and t þ 12 fore-
and TCN-ED at the short forecast horizon. However,
cast horizons in the testing stage, while R 2 values decrease a
TCN’s forecast peak flow is later than the observed peak
great deal at the t þ 18 and t þ 24 forecast horizons, especially
flow, which will put more pressure on flood control. In con-
at the t þ 24 horizon. The scatters are closer to the 45-degree
trast, the forecast results of TCN-ED are more practical. At
line at the t þ 6 and t þ 12 forecast horizons, while they are
the t þ 12 forecast horizon, it is obvious that TCN-ED has
farther away from the 45-degree line at the t þ 18 and t þ 24
higher accuracy in the peak flow forecast than TCN, as
forecast horizons. These indicate that the observation and pre-
TCN underestimates the peak flow compared to the
dication scatters have good fitness at the t þ 6 and t þ 12
observed. The forecast runoff of TCN near the peak flow
forecast horizons, and poor fitness at the t þ 18 and t þ 24
has a quick drop, which may be due to TCN being sensitive
forecast horizons, which is consistent with the results above.
to the fluctuation of precipitation. The context value used by
The percentage distribution and R 2 value of each flow
TCN-ED at each time step of decoding is uniformly the
section can further reflect the forecast ability of the
output of the last time step of the encoder, so the drastic
models for different magnitude flow. It can be seen from
changes in the rain sequence will not be immediately
each subplot in Figure 6 that the difference of percentage
reflected in the output sequence, which is more in line
between the two models in different sections is relatively
with the smooth change process of the runoff. At the fore-
small, and the percentage of TCN-ED is closer to that of
cast horizons t þ 18 and t þ 24, due to the lack of
the observation than TCN, especially in the low and peak
hydrological information, the forecast runoff hydrograph
flow sections. It can be seen from Figure 6 that R 2 values
lags far behind the observed with increasing ahead hours.
of both models in the low flow section are less than 0.3 at
This indicates that both TCN and TCN-ED models may
four forecast horizons, reflecting poor fitting between the
gradually lose its forecast ability when the forecast horizons
observation and prediction in the low flow section, while
are beyond the concentration time of the basin.
in the medium and high flow sections, R 2 values are greater than 0.5, indicating that both models have better forecast
Robustness of the models based on regression analysis
ability for median and high flow. It also can be found from Figure 6 that forecast horizons have a great influence on
In order to explore the robustness of the models, the
the prediction of peak and extreme peak flow. In Figure 6(a)
relationships between the observation and predictions of
and 6(b), R 2 values of both models in the peak flow and
the 40 rounds are drawn at the t þ 6, t þ 12, t þ 18, and
extreme peak flow sections are greater than 0.5, indicating
t þ 24 forecast horizons in the testing stage, which are dis-
that both models have good forecast ability for peak and
played as scatter plots with the KDE curves in Figure 6.
extreme peak flow at the t þ 6 and t þ 12 forecast horizons.
The runoff range is divided into the low flow section
However, R 2 values drop to about 0.3 in Figure 6(c), and
(178 m3/s), medium section (178–550 m3/s), high section
drop below 0.1 in Figure 6(d), indicating that both models
3
3
(550–1,374 m /s), peak flow section (1,374–2,885 m /s),
gradually lose the forecast ability to peak and extreme
and extreme peak flow section (>2,885 m3/s) by the 25th,
peak flow with forecast horizons increasing. In terms of
75th, 95th, and 99th percentiles of the observations ( Jiang
the R 2 value of each section in each subplot in Figure 6,
et al. ). The five sections contain 25, 50, 20, 4, and 1%
TCN-ED performs better than TCN as its higher R 2 value.
K. Lin et al.
1146
Figure 6
|
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Hydrology Research
|
51.5
|
2020
Relationships between the observed and predicted runoff in the testing stage using the 40 rounds results of TCN and TCN-ED at horizons (a) t þ 6, (b) t þ 12, (c) t þ 18, and (d) t þ 24.
1147
K. Lin et al.
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Hydrology Research
|
51.5
|
2020
According to the above analysis, within an appropriate
further verified in more basins, which will also help to
forecast horizon, both TCN and TCN-ED have good forecast
improve the forecast ability of ANNs in runoff
ability. For example, the appropriate forecast horizon in this
forecasting.
study is the concentration time (12 h) of the basin. Both
3. In general, TCN-ED has better performance than TCN in
models perform poorly in the low flow section, and good
runoff forecasting in this study. TCN-ED shows better
in the medium and high flow sections, while the forecasting
transferability from the training stage to the testing
for peak flow and extreme peak flow is greatly affected by
stage. NSE and VE values of TCN-ED for most forecast
the variation of the forecast horizons.
horizons are higher, and R 2 values of TCN-ED for most flow sections are higher. In addition, with the unique con-
CONCLUSIONS This study constructs TCN and TCN-ED and demonstrates their applicability in the Jianxi basin, China. In addition, with the comparison of TCN and TCN-ED, the ability of the Encoder-Decoder architecture is further illustrated for the runoff forecasting. First, with the repeated calculation
text value which maintains the continuity of hydrological information, TCN-ED shows better stability and is insensitive to fluctuations in the rainfall process. Liu et al. () proved that the LSTM combined with EncoderDecoder architecture has better performance in streamflow forecast than the LSTM. More follow-up studies are needed to verify the advantages of Encoder-Decoder architecture integrated with more ANN models.
results of each model in the training and testing stages, the accuracy and transferability of each model are analyzed. Second, the performance of each model during various forecast horizons in the testing stage is comprehensively
ACKNOWLEDGEMENTS
compared. Third, the best model results at the forecast horizons t þ 6, t þ 12, t þ 18, and t þ 24 are used to analyze the model’s forecast ability for the maximum floods. Finally, the robustness of the model is explored by the relationships
The study is financially supported by the National Key Research and Development Program (2018YFC0407904) and the Research Council of Norway (FRINATEK Project 274310).
between the observation and predictions of the 40 rounds at the t þ 6, t þ 12, t þ 18, and t þ 24 forecast horizons. The major findings of this study are summarized as follows. 1. The forecast horizon has a significant impact on the forecast ability of TCN and TCN-ED. In this study, NSE and VE values of both models are high and stable within t þ 12 forecast horizons. As the forecast horizon increases
DATA AVAILABILITY STATEMENT Data cannot be made publicly available; readers should contact the corresponding author for details.
after the t þ 12 forecast horizon, NSE and VE values decrease rapidly, indicating that the forecast ability of the models becomes poor. Since the concentration time
REFERENCES
of the study basin is about 12 h, it can be inferred that the concentration time is a critical threshold to the effective forecast horizon for both models, which needs to be further demonstrated in more basins. 2. Both models perform poorly in the low flow section, and good in the medium and high flow sections during most forecast horizons, while it is conditional to the forecast horizon in forecasting peak flow for both models. Whether this rule is universal or not also needs to be
Abdellatif, M., Atherton, W. & Alkhaddar, R. A hybrid generalised linear and Levenberg-Marquardt artificial neural network approach for downscaling future rainfall in North Western England. Hydrology Research 44 (6), 1084–1101. Adamowski, J. & Sun, K. R. Development of a coupled wavelet transform and neural network method for flow forecasting of non-perennial rivers in semi-arid watersheds. Journal of Hydrology 390 (1–2), 85–91. Allen, D. M. Mean square error of prediction as a criterion for selecting variables. Technometrics 13 (3), 469–475.
1148
K. Lin et al.
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Awchi, T. A. & Srivastava, D. K. Analysis of drought and storage for Mula project using ANN and stochastic generation models. Hydrology Research 40 (1), 79–91. Baba, A. P. A., Shiri, J., Kisi, O., Fard, A. F., Kim, S. & Amini, R. Estimating daily reference evapotranspiration using available and estimated climatic data by adaptive neuro-fuzzy inference system (ANFIS) and artificial neural network (ANN). Hydrology Research 44 (1), 131–146. Bai, S., Kolter, J. Z. & Koltun, V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. arXiv:1803.01271. Beven, K., Kirkby, M., Schofield, N. & Tagg, A. Testing a physically-based flood forecasting model (TOPMODEL) for three UK catchments. Journal of Hydrology 69 (1–4), 119–143. Bian, C., He, H., Yang, S. & Huang, T. State-of-charge sequence estimation of lithium-ion battery based on bidirectional long short-term memory encoder-decoder architecture. Journal of Power Sources 449, 227558. Cardenas-Barrera, J. L., Meng, J. L., Castillo-Guerra, E. & Chang, L. C. A neural network approach to multi-step-ahead, short-term wind speed forecasting. In: 12th International Conference on Machine Learning and Applications, Miami. Vol. 2, pp. 243–248. Chang, F.-J., Chang, L.-C. & Huang, H.-L. Real-time recurrent learning neural network for stream-flow forecasting. Hydrological Processes 16 (13), 2577–2588. Cho, K., Van Merrienboer, B., Gulcehre, C., Bougares, F., Schwenk, H. & Bengio, Y. Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv:1406.1078. Claveria, O., Torra, S. & Monte, E. Modelling tourism demand to Spain with machine learning techniques. The impact of forecast horizon on model selection. Revista De Economia Aplicada 24 (72), 109–132. Criss, R. E. & Winston, W. E. Do Nash values have value? Discussion and alternate proposals. Hydrological Processes 22 (14), 2723–2725. Dawson, C. W. & Wilby, R. An artificial neural network approach to rainfall-runoff modelling. Hydrological Sciences Journal 43 (1), 47–66. Deng, S., Zhang, N., Zhang, W., Chen, J., Pan, J. & Chen, H. Knowledge-driven stock trend prediction and explanation via temporal convolutional network. In: Companion Proceedings of the 2019 World Wide Web Conference, San Franscisco, CA, pp. 678–685. Dodge, Y. Coefficient of determination. Alphascript Publishing 31 (1), 63–64. Douglas-Mankin, K., Srinivasan, R. & Arnold, J. Soil and water assessment tool (SWAT) model: current developments and applications. Transactions of the ASABE 53 (5), 1423–1431. Ecrepont, S., Cudennec, C., Anctil, F. & Jaffrézic, A. PUB in Québec: a robust geomorphology-based deconvolution-
Hydrology Research
|
51.5
|
2020
reconvolution framework for the spatial transposition of hydrographs. Journal of Hydrology 570, 378–392. He, Y. & Zhao, J. Temporal convolutional networks for anomaly detection in time series. Journal of Physics: Conference Series 1213 (4), 042050. He, K., Zhang, X., Ren, S. & Jian, S. Deep residual learning for image recognition. In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, CA, pp. 770–778. Hsu, K. l., Gupta, H. V. & Sorooshian, S. Artificial neural network modeling of the rainfall-runoff process. Water Resources Research 31 (10), 2517–2530. Hu, C. H., Wu, Q., Li, H., Jian, S. Q., Li, N. & Lou, Z. Z. Deep learning with a long short-term memory networks approach for rainfall-runoff simulation. Water 10 (11), 1543. Jiang, S., Zheng, Y., Babovic, V., Tian, Y. & Han, F. A computer vision-based approach to fusing spatiotemporal data for hydrological modeling. Journal of Hydrology 567, 25–40. Jie, M.-X., Chen, H., Xu, C.-Y., Zeng, Q., Chen, J., Kim, J.-S., Guo, S.-l. & Guo, F.-Q. Transferability of conceptual hydrological models across temporal resolutions: approach and application. Water Resources Management 32 (4), 1367–1381. Kao, I. F., Zhou, Y. L., Chang, L. C. & Chang, F. J. Exploring a long short-term memory based encoder-decoder framework for multi-step-ahead flood forecasting. Journal of Hydrology 583, 12. Kingma, D. P. & Ba, J. Adam: a method for stochastic optimization. arXiv:1412.6980. Kisi, O. River flow forecasting and estimation using different artificial neural network techniques. Hydrology Research 39 (1), 27–40. Legates, D. R. & McCabe Jr, G. J. Evaluating the use of ‘goodness-of-fit’ measures in hydrologic and hydroclimatic model validation. Water Resources Research 35 (1), 233–241. Liong, S. Y. & Sivapragasam, C. Flood stage forecasting with support vector machines. Journal of the American Water Resources Association 38 (1), 173–186. Liu, J., Zhu, H., Liu, Y., Wu, H., Lan, Y. & Zhang, X. Anomaly detection for time series using temporal convolutional networks and Gaussian mixture model. Journal of Physics: Conference Series 1187, 042111. Liu, D., Jiang, W., Mu, L. & Wang, S. Streamflow prediction using deep learning neural network: case study of Yangtze River. IEEE Access 8, 90069–90086. Loyola, P., Liu, C. & Hirate, Y. Modeling user session and intent with an attention-based encoder-decoder architecture. In: Proceedings of the Eleventh ACM Conference on Recommender Systems, Como, Italy, pp. 147–151. Moor, M., Horn, M., Rieck, B., Roqueiro, D. & Borgwardt, K. Early recognition of sepsis with Gaussian process temporal convolutional networks and dynamic time warping. Machine Learning for Healthcare Ann Arbor, USA, 106, 2–26.
1149
K. Lin et al.
|
TCN combined with Encoder-Decoder framework for runoff forecasting
Nash, J. E. & Sutcliffe, J. V. River flow forecasting through conceptual models part I – a discussion of principles. Journal of Hydrology 10 (3), 0–290. Park, C., Kim, D. & Yu, H. An encoder–decoder switch network for purchase prediction. Knowledge-Based Systems 185, 104932. Ren-Jun, Z. The Xinanjiang model applied in China. Journal of Hydrology 135 (1–4), 371–381. Shamseldin, A. Y. Application of a neural network technique to rainfall-runoff modelling. Journal of Hydrology 199 (3–4), 272–294. Shi, X. J., Chen, Z. R., Wang, H., Yeung, D. Y., Wong, W. K. & Woo, W. C. Convolutional LSTM network: a machine learning approach for precipitation nowcasting. Advances in Neural Information Processing Systems 28, 802–810. Shiri, J., Marti, P., Nazemi, A. H., Sadraddini, A. A., Kisi, O., Landeras, G. & Fard, A. F. Local vs. external training of neuro-fuzzy and neural networks models for estimating reference evapotranspiration assessed through k-fold testing. Hydrology Research 46 (1), 72–88. Sivapragasam, C., Liong, S.-Y. & Pasha, M. Rainfall and runoff forecasting with SSA–SVM approach. Journal of Hydroinformatics 3 (3), 141–152. Smith, J. & Eli, R. N. Neural-network models of rainfallrunoff process. Journal of Water Resources Planning Management 121 (6), 499–508.
Hydrology Research
|
51.5
|
2020
Tang, Q., He, X., Bao, Y., Zhang, X., Guo, F., Zhu, H., Tang, Q., He, X., Bao, Y. & Zhang, X. Determining the relative contributions of climate change and multiple human activities to variations of sediment regime in the Minjiang River, China. Hydrological Processes 27 (25), 3547–3559. Trajkovic, S. Testing hourly reference evapotranspiration approaches using lysimeter measurements in a semiarid climate. Hydrology Research 41 (1), 38–49. Wang, G., Zhang, J., Jin, J., Pagano, T., Calow, R., Bao, Z., Liu, C., Liu, Y. & Yan, X. Assessing water resources in China using PRECIS projections and a VIC model. Hydrology and Earth System Sciences 16 (1), 231–240. Xu, J., Luo, W. & Huang, Y. Dadu river runoff forecasting via Seq2Seq. In: Proceedings of the 2019 International Conference on Artificial Intelligence and Computer Science, Hangzhou, China, pp. 494–498. Yuan, X. H., Chen, C., Lei, X. H., Yuan, Y. B. & Adnan, R. M. Monthly runoff forecasting based on LSTM-ALO model. Stochastic Environmental Research and Risk Assessment 32 (8), 2199–2212. Zhu, X. T., Wang, H., Xu, L. & Li, H. Z. Predicting stock index increments by neural networks: the role of trading volume under different horizons. Expert Systems with Applications 34 (4), 3043–3054.
First received 16 July 2020; accepted in revised form 26 August 2020. Available online 21 September 2020
1150
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Impact of urbanization on variability of annual and flood season precipitation in a typical city of North China Peijun Li, Depeng Zuo, Zongxue Xu, Xiaoxi Gao, Dingzhi Peng, Guangyuan Kan, Wenchao Sun, Bo Pang and Hong Yang
ABSTRACT Urbanization plays an important role in a global change, but there are few studies that combine land use with topography and precipitation. The urbanization in Jinan, a typical city of North China, between the 1980s and 2005 was analyzed by transition matrix analysis, and the topographic effects on land use changes were explored considering altitude, slope, and aspect. The temporal trends and abrupt changes of annual and flood season precipitation for the last 60 years were detected by multiple nonparametric detection methods, and the precipitation indices were adopted to characterize the frequency and intensity of precipitation. The relationship between urbanization and precipitation was finally investigated by grey correlation analysis. The results showed that land use
Peijun Li Depeng Zuo (corresponding author) Zongxue Xu Xiaoxi Gao Dingzhi Peng Wenchao Sun Bo Pang College of Water Sciences, Beijing Key Laboratory of Urban Hydrological Cycle and Sponge City Technology, Beijing Normal University, Beijing 100875, China E-mail: dpzuo@bnu.edu.cn
structure in Jinan experienced a dramatic change during recent decades due to the urbanization process, the conversion of cropland to forest, and the protection of spring in Jinan. The land use types that closely related to mankind’s activities and living behaviors were more concentrated in areas with lower altitude and slope, and detected more significant changes. The significant abrupt changes of precipitation generally occurred in 1989 and concentrated in urban areas and southern mountainous areas with high increment. The increase of flood season precipitation was more significant in the central urban region with a maximum increasing rate of 39.52%, and the increase in the number of storm and precipitation days was also obvious. The precipitation indices in the flood season were more closely related to changes in farmland and settlements affected by urbanization with a maximum correlation coefficient of 0.75. These findings illustrate the impact of urbanization on the variability of precipitation and support that the changes in the urban underlying surface have a certain effect on the surface energy balance. Key words
| abrupt change, land use change, North China, precipitation, topographic effects, urbanization
HIGHLIGHTS
• • •
A significant urbanization process has undergone in Jinan during recent decades. Flood season precipitation showed a more significant upward trend during the last 60 years. The rainstorm days, precipitation amount, and the probability of heavy rain in urban areas have significantly increased.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/ licenses/by-nc-nd/4.0/). doi: 10.2166/nh.2020.176
Guangyuan Kan Research Center on Flood & Drought Disaster Reduction of the Ministry of Water Resources, China Institute of Water Resources and Hydropower Research, Beijing 100038, China Hong Yang Eawag, Swiss Federal Institute of Aquatic Science and Technology, P.O. Box 611, 8600 Dübendorf, Switzerland
1151
•
P. Li et al.
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
The precipitation indices in the flood season were more closely related to changes in farmland and settlements affected by urbanization.
INTRODUCTION The urbanization process is generally accompanied by the
(UCM) to assess the impact of extensive urbanization on
rapid succession of underlying surface (Mahmood et al.
regional precipitation across the Beijing–Tianjin–Hebei
), resulting in changes of topography, thermal dynamic
region of China, which showed that extensive urbanization
conductivity, and hydraulic permeability (Oke ), which
considerably decreased precipitation over and downwind
in turn have an impact on the urban hydrological cycle,
of Beijing city.
such as an increase in regional precipitation and local
Other studies focused on the separation of climate
extreme precipitation, particularly in downwind suburban
change and urbanization on precipitation at different spatial
areas (Pathirana et al. ; Yu et al. ). Urbanization
scales. Gu et al. () separated potential contributions of
not only exacerbates the flood response, but also increases
urbanization and climate change to precipitation trends in
the frequency of heavy rains (Zhang et al. ). The
China at national, regional, and local scales using a ‘trajec-
Metropolitan Meteorological Experiment (METROMEX)
tory’-based method, which concluded that climate change
confirmed that the urban heat island circulation can trigger
was the principle factor for variations of precipitation,
and enhance convective weather, such as thunderstorms,
while urbanization generally has a greater effect on total pre-
heavy precipitation, and strong storms, which especially
cipitation than precipitation extremes. Whether the effect of
has a significant impact on urban downwind areas
urbanization on precipitation is increasing or decreasing is
(Changnon ). Furthermore, studies in Netherland
still controversial, and some studies have even reached the
showed that extreme precipitation in urban areas is more
opposite conclusion. The precipitation in the urban down-
intense compared with rural areas, which have a vital connec-
wind direction was suppressed in the study of major urban
tion with the urbanization process (Golroudbary et al. ).
areas and industrial facilities areas, because the addition of
There are kinds of methods for investigating the impact
ice nuclei and cloud condensation nuclei reduced the con-
of urbanization on regional precipitation. For statistical
version efficiency of cloud water to rainwater (Rosenfeld
methods, Lu et al. () applied four nonstationary general-
; Givati & Rosenfeld ). Moreover, a decrease in sur-
ized extreme value models to evaluate the impact of
face water storage and in water vapor supply to the upper
urbanization on extreme precipitation in the Yangtze River
atmosphere may also lead to a decrease in precipitation in
Delta metropolitan region, which found that urban
urban areas. The impacts of heat island effect and aerosol
expansion could increase the magnitudes of extreme precipi-
emissions on precipitation may be insufficient, compared
tation and its recurrence levels under different return
with the impact of changes in the hydrological character-
periods. Zhu et al. () analyzed hourly precipitation
istics of the underlying surface (Kaufmann et al. ;
data in Beijing Municipality during the period of 2011–
Wang et al. ).
2015 using the circular statistical analysis and the grange
However, most of the research focused on the five pre-
causality test technique, which indicated that impacts of
cipitation gauging stations located in the Xiaoqing River
urbanization on precipitation varied with different types of
basin, occupying only a quarter of the total area of the
urbanization. In addition, the climate model and the land
study area, which is not enough to comprehensively reflect
surface model are the alternative methods for exploring
the spatial heterogeneity of precipitation in the whole of
the impact of urbanization on regional precipitation. Wang
Jinan. Therefore, more detailed information on the relation-
et al. () coupled the Weather Research and Forecasting
ship between land use change and precipitation in the study
(WRF) model with a single-layer urban canopy model
area should be further investigated. In this study, a variety of
P. Li et al.
1152
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
methods were used to analyze the spatial pattern and tem-
North China Plain (116 110 –117 440 E, 36 010 –37 320 N)
poral trend of precipitation and the changes of land use
(Figure 1). The area under the jurisdiction of the Jinan
types in Jinan during the last 60 years, and the relationship
municipal administrative area is 8,151 km2, including
between land use change and precipitation variation in the
seven districts and two counties. The terrain is high in the
study area was explored. This study tried to reveal the evol-
south and low in the north. The terrain is complex and
ution law between the two environmental variables and
can be divided into three zones: the north is near the
point out the issues that need to be faced in the development,
Yellow River area, the middle is the piedmont plain area,
utilization, and protection of water and soil resources, which
and the south is the hilly mountain area. There are many
is of great significance for further urban development and the
rivers in Jinan, mainly the Yellow River and the Xiaoqing
coordination of the water and environment.
River. Jinan belongs to the sub-humid continental monsoon climate and has obvious monsoon characteristics, with an annual average temperature of 14.3
STUDY AREA AND DATA DESCRIPTION
C and an annual
average precipitation of 648.0 mm (Zuo et al. ). The precipitation is extremely uneven, and the inter- and intra-
Study area
annual variability is strong (Chang et al. ). In summer, the average precipitation in various regions is more than
Jinan, the capital of Shandong Province, is located in the
400 mm, which concentrates 60% of the annual precipi-
Midwest of Shandong Province, the downstream region of
tation. The average number of precipitation days in July is
the Yellow River Basin and on the eastern edge of the
about 15 days, and the number of rainstorm days with
Figure 1
|
Digital elevation model (DEM) and locations of precipitation gauging stations in Jinan.
1153
P. Li et al.
|
Impact of urbanization on variability of precipitation in a city of North China
daily precipitation above 50 mm is concentrated in July and
Hydrology Research
|
51.5
|
2020
METHODOLOGY DESCRIPTION
August, accounting for 70% of the number of rainstorm days throughout the year.
To analyze the relationship between urbanization and
According to statistics, the urban built-up area of Jinan
precipitation in Jinan, urban areas (including municipal
is expanded by 57.94 km2 during the period of 1986–2000
districts of Licheng, Shizhong, Tianqiao, Lixia, and
(Jiang et al. ). With the rapid economic and social
Huaiyin), suburban areas (Zhangqiu and Changqing dis-
development, human activities in Jinan have significantly
tricts), and rural areas (including Jiyang, Shanghe, and
intensified. According to the results of the sixth national
Pingyin) were divided based on GDP levels and distance
census in 2010 (http://www.stats.gov.cn), the permanent
from the urban core. In exploring the topographic effects
resident population of Jinan is 6.184 million, which is an
on the characteristics of spatial distribution of land use
increase of 892,300 compared with 10 years ago, with a
types, a sampling point was set at each 0.01 km2 in the
growth rate of 15.07%. With the processes of urbanization,
study area, and a scatter plot was drawn based on the attri-
population growth and economic development, land use
butes of sampling points. The areas of each land use type
patterns have also undergone tremendous change. Under
were calculated at 50 m increments of elevation and 1
the influence of the drastic changes in underlying surface
increments of slope.
conditions during the recent decades, large- and mediumsized cities in China have suffered from rainstorm flood and waterlogging, in which Jinan is a typical case (Hammond et al. ; Cheng et al. ).
Land use transition matrix Land use transition matrix is derived from the quantitative analysis of system state and state transition in system
Data description The land use type data of the study area in 1980s and 2005 were provided by the National Science and Technology Foundation Platform Project ‘Data-Sharing Network of Earth System Science’ (www.geodatda.cn). The daily precipitation data at 47 gauging stations during the period of 1950–2016 were collected for further analysis, which was provided by the Jinan Hydrology Bureau. The locations of the precipitation gauging stations are shown in Figure 1, in which 23 flood season gauging stations were mainly used to study the spatiotemporal
analysis. The matrix reflects the land use dynamic process information of the mutual transformation between the beginning and end of a certain period at a certain area. It can be used to characterize the structural characteristics of land use change and show the trend and direction of various land-based transformations in a concrete and comprehensive way (Lu et al. ). The general formula is shown in Table 1, in which P11–Pnn is the transfer area between different land use types A1–An during the period from T1 to T2. Sn1 and Sn2 indicate the n-type area in the land use period T1 and the period T2, respectively.
variability of precipitation in the flood season; the remainder of the gauging stations for annual precipitation monitoring was adopted for abrupt change detection and
Table 1
trend analysis of precipitation. The missing data of the precipitation series were filled by neighboring stations through the linear regressive method, in which all the correlation coefficients R 2 > 0.8, which indicates that the filled series can satisfy the quality requirements for this study. Furthermore, the economic and social statisticsrelated administrative regions in the Statistical Yearbook were also used in this study.
|
The description of land use transition matrix T2
T1
A1
A2
…
An
Total
A1
P11
P12
…
P1n
S11
A2
P21
P22
…
P2n
S21
…
…
…
…
…
…
An
Pn1
Pn2
…
Pnn
Sn1
Total
S12
S22
…
Sn2
Total area
P. Li et al.
1154
|
Impact of urbanization on variability of precipitation in a city of North China
Temporal trend and abrupt change detection methods
Hydrology Research
|
51.5
|
2020
decreasing trend. When the value of UFk or UBk exceeds a critical straight line, it indicates that the upward or downward trend is significant. If two curves appear at the
Nonparametric Mann–Kendall test method
intersection point, and the point appears between the critiThe Mann–Kendall method is a nonparametric statistical
cal lines, then the moment corresponding to the point is
test method used to calculate the changing characteristics
the start time of the abrupt changes (Zuo et al. ).
of hydro-meteorological factors (Mann ; Kendall ). Its advantage is that the sample does not need to follow a certain regular distribution and will not be disturbed by a small number of abnormal values. It is simple to calculate and suitable for type variables and order variables. For a time series x with n sample sizes, construct an ordered sequence:
sk ¼
k X
ri
k ¼ 2, 3, , n
(1)
Nonparametric Pettitt test method The Pettitt method is also a kind of nonparametric test method, which directly uses the order column to detect mutation points (Petitt ). For a time series x with n sample sizes, construct an ordered sequence: 8 < þ1 xi > xj ri ¼ 0 xi ¼ xj : 1 xi < xj
(j ¼ 1, 2, , i)
(5)
i¼1
If in which ri ¼
þ1 xi >xj 0 xi xj
kt0 ¼ Maxjsk j (j ¼ 1, 2, , i)
(2)
Assuming the time series are random and independent, the defined statistics are as follows: [sk E(sk )] ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi UFk ¼ p Var(sk )
(k ¼ 2, 3, , n)
(6)
The time corresponding to t0 is the start time of the abrupt changes. 6k2t0 P ¼ 2 exp 3 n þ n2
! (7)
If P 0.5, the detected abrupt changes are considered to (k ¼ 1, 2, , n)
(3)
Moving T-test method
where UF1 ¼ 0, E(sk) is the mean of the cumulative number sk, and Var(sk) is the variance of the cumulative number sk. n(n 1) E(sk ) ¼ 4 n(n 1)(2n þ 5) Var(sk ) ¼ 72
be statistically significant.
The basic test idea of the moving T-test is to deal with the problem of whether there is a significant difference between the mean values of the two subsequences in the climate series. When the difference between the mean values of
(4)
Given the significance level α is 0.05, check the normal distribution table to determine the critical value u0.05 ¼ ± 1.96. Repeat the above process in reverse order of time series, while making UBk ¼ UFk, k ¼ n, n 1, …, 1, UB1 ¼ 0, and draw a sequence curve diagram of UFk and UBk. If the value of UFk or UBk is greater than 0, it indicates that the sequence is increasing, and conversely, it indicates a
two subsequences exceeds a certain level of significance, it is considered that the mean value has changed qualitatively at this moment. For a time series x with n samples, a moment is selected as the reference point, and the samples of x1 and x2 in the two subsequences before and after the reference point define the statistics t for n1 and n2, respectively: t¼
x1 x2 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 1 s þ n1 n2
(8)
P. Li et al.
1155
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
in which
which shows that the development and change of the two
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi n1 s21 þ n2 s22 s¼ n1 n2 2
factors are closer. (9)
where s21 and s22 are the variances of the two subsequences. To improve the reliability of the calculation results and avoid the shift of abrupt changes caused by the length of the subsequence, the experimental comparison was performed by changing the length of the subsequence for several times. Finally, the value of n1 and n2 was selected as 7 and the confidence level α was set as 0.05. Therefore, the critical value u0.05 ¼ ±2.18 is determined.
RESULTS AND ANALYSIS Changes in land use types from 1980s to 2005 Spatiotemporal variation of land use types From the point of view of spatial distribution, the maps of land use type in Jinan for the periods of the 1980s and 2005 are shown in Figure 2. Farmland dominates the land use types in Jinan, reaching about 65% of the total
Grey relational analysis
area, mainly distributed in the northern and central plains affected by the terrain. The southern mountainous
Grey relational analysis (GRA) is a dynamic correlation
area is characterized as higher altitudes, the main land
analysis method that quantitatively describes the strength,
use types of which are forests and grasslands. The main
size, and order of the relationship between factors (Tan &
stream of the Yellow River runs through Jinan flowing
Deng ). During the development of the system, if the
from the central region to the northeast. The main
trends of the two factors are consistent, the correlation
urban area of Jinan City is located in the middle. The
between the two factors is higher; otherwise, it is lower.
evolution of construction land was characterized as an
When performing data analysis, x0 is set to be the reference
increase of urban area from east to west and an expansion
sequence and xi is set to be the comparison sequence.
of suburban area in the northwest and the southern
Dimensionless is calculated by means of the mean
mountainous areas.
method, which is transformed into a pure number sequence x00
and
xi0 .
The changes in various land use types between the two
The absolute value of the corresponding point
periods are calculated in Figure 3. The land use structure of
difference constitutes the difference sequence, and the maxi-
Jinan experienced a great change from 1980s to 2005, in
mum difference and the minimum difference between the
which settlements and farmland saw the most obviously
two sequences are identified. The grey correlation coeffi-
change. The area of settlements increased 173.26 km2 with
cient formula is as follows:
a growth rate of 18.5%, while the area of farmland decreased
0
ξ0i(k) ¼
0
0
190 km2. The period from the 1980s to 2005 was the period
0
min jx0(k) xi(k) j þ ξ max jx0(k) xi(k) j i k i k 0 0 Δ0i(k) þ ξ max jx0(k) xi(k) j i k
(10)
values according to different background requirements.
1 ξ N k¼1 0i(k)
urbanization is advancing rapidly, which reflects the impact of urbanization and population growth on land resources. It can be seen from the Annual Statistical Year-
A correlation degree is calculated as follows: γ 0i ¼
of Jinan City’ were promulgated and implemented. The sharp increase in the settlement area indicates that Jinan’s
in which, ξ is the resolution coefficient, 1 > ξ > 0. ξ takes
N X
when the ‘Administrative Measures for the Urban Planning
book of Jinan in 2005 that Jinan’s per capita GDP was 1,263 yuan in 1985, and dramatically reached to 31,606 (11)
yuan in 2005, with an increase of 25 times. The economy of Jinan rapidly developed during this period, and the level
in which γ0i is the correlation degree. The greater the grey
of urbanization gradually increased, which is in line with
correlation degree, the closer the geometric curve shape is,
the trend of land use change.
P. Li et al.
1156
|
Impact of urbanization on variability of precipitation in a city of North China
Figure 2
|
The spatial distributions of land use types in Jinan during the periods 1980s and 2005.
Figure 3
|
Proportions of land use types in Jinan during the periods 1980s and 2005.
Forests are the second most dominant type of land use following farmland, which showed an overall growth
Hydrology Research
|
51.5
|
2020
hydrology, but also social development and human activities.
trend during the study period, from 952.56 km2 in the 1980s to 954.12 km2 in 2005, increasing 0.16%. The area
Transition matrix of land use types
of wetlands and water showed an upward trend during the study period, and the growth rate was second only
The land use transition matrix of Jinan from 1980s to 2005
to settlements, reaching 10.12% with a total increase of
in Table 2 explains how the land use types changed in
2
21.92 km . Grassland and desert areas showed a slight
detail. The transferred area of settlement accounted for
decline with little change overall. The changes in land
17.36% of the total transferred area of all land types in the
use types during the study period were not only affected
study area; the increased area accounted for 32.83% of the
by natural factors such as topography, landforms, and
total increased area of all land use types. The increased
P. Li et al.
1157
Table 2
|
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
Transition matrix of land use changes in Jinan from the 1980s to 2005 (km2)
2005 1980s
Forest
Grassland
Farmland
Settlements
Wetland water
Desert
Total
Forest
806.96
58.22
79.22
9.62
2.04
0.29
956.34
Grassland
71.81
441.41
74.10
8.67
1.57
0.14
597.70
Farmland
75.22
87.82
4,887.42
344.98
57.47
3.74
5,456.65
Settlements
2.75
3.36
181.63
741.70
6.40
0.46
936.30
Wetland water
0.84
1.11
39.29
3.89
168.11
0.05
213.30
Desert
0.38
0.26
4.51
0.55
0.11
11.11
16.92
Total
957.95
592.19
5,266.15
1,109.41
235.72
15.79
8,177.21
area of settlement was 1.89 times the transferred area, and
Topographic effects on land use changes
the growth rate of which was significantly faster than other land use types. More than half of the increased settle-
The scatter plot was used to explore the topographic effects
ment came from the conversion of farmland. The settlement
on the characteristics of spatial distribution of land use types
area converted from farmland reached 345.04 km2, account-
(Figure 4). At the altitude of 400–600 m, the distribution of
ing for 60.62% of the total transferred area of farmland,
forests on each aspect and slope was relatively uniform,
which reflects the rapid development of urbanization in
while that was concentrated in the areas with the west
Jinan and the impact of population growth on land
aspect at the altitude of 150–250 m. The grassland was
resources. The transferred area of farmland was 1.5 times
generally distributed around 200 m, which had a good
the increased area of farmland.
adaptability to slope and aspect. The distribution of
The area of wetland and water showed an upward trend
cultivated land characterized as that where most was
during this period, and the increasing rate of 10% ranked
located in the areas with a slope of 0–4 and an altitude of
only second to settlements. During the research period, Jinan
about 0 m. Water bodies were mostly distributed in low-
invested billions of dollars in the treatment and restoration
lying areas with a slope of 0–5 and an altitude of about
of rivers and lakes under the jurisdiction of the municipal gov-
50 m, and most of the settlements were located in the
ernment, which not only improved the water quality, but also
areas of 0–10 and 50–100 m. The distribution of desert
expanded the water area, reflecting that the ‘spring city’ has
was relatively loose, which was less affected by topographic
put considerable effort into ecological protection and restor-
factors.
ation. With the issuance and promotion of the ‘Jinan City
The area statistics of land use types according to
Ecological Restoration and Urban Remediation Pilot’, the
aspects are shown in Figure 5. The distributions of farm-
government implemented the ecological restoration and pro-
land and water body were less affected by the aspect,
tection of water and paid more attention to the surrounding
showing a strong adaptability of aspect. Compared with
ecological greening of river banks and reservoir areas. It is
1980s, the area of farmland with easterly aspect had
expected that the wetland and water area will maintain
decreased significantly, while the area of water body had
steady growth in the future. The transferred area of forest is
an overall increase, especially for SW aspect. In the
not much different from the increased area, with an overall
1980s, settlements were mainly oriented toward westerly
increase less than 0.1%. Among the newly added areas, the
aspect, while that tended to expand to easterly aspect in
land use types were shifted mostly from farmland, accounting
2005. Most of the forest land was located on W and NW
for 50% of the total new increased area of forest, reflecting the
aspect, which had decreased from the 1980s to 2005. The
impact of environmental protection policies such as returning
desert on westerly and southerly aspects showed a signifi-
farmland to forests on land use structure in Jinan City.
cant decrease trend, which experienced an obvious shift
P. Li et al.
1158
Figure 4
|
|
Impact of urbanization on variability of precipitation in a city of North China
The distributions of land use types on various aspects and elevations in Jinan during the periods 1980s and 2005.
Hydrology Research
|
51.5
|
2020
P. Li et al.
1159
Figure 5
|
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
Land use changes on different aspects in Jinan during the periods 1980s and 2005.
from westerly aspect to easterly aspect. The area of grass-
implemented the Regulations of Forest Resources Protec-
land in westerly aspect significantly decreased during the
tion and Management since 2000, which severely
study period, while that in easterly aspect significantly
cracked down on deforestation and encouraged the con-
increased.
version of cropland to forest. Compared with 2000, the
With the rise of elevation, the area of each land
forest coverage increased from 19.7 to 24.1% in 2005,
use type increased first and then decreased (Figure 6).
which is consistent with the results reflected in the
Most areas of farmland, settlements, wetland and
figure. The area of settlements significantly increased
waters, and deserts were concentrated at 0 m, which
in the elevation belt of 0–150 m, which was also
are the land use types closely related to human activities.
transformed from farmland, indicating an obvious urban-
Forests were mostly distributed at the elevation of 250–
ization process. Wetland and waters and deserts were
350 m, while the peak value of the grasslands area
generally distributed in low-lying areas, which was
appeared at the elevation of about 200 m. The two
strongly constrained by elevation. Compared with the
types of land use were mostly distributed in mountainous
1980s, the area of wetland and waters greatly increased
and hilly areas, and their topographic effect was less
at the elevation of 0 m, while that of deserts significantly
than that on the other four types. Compared with the
decreased at the same elevation, indicating a consider-
1980s, the area of forest increased by nearly 30% at
able improvement of the eco-environment in the study
the elevation of 0 m, while the area of farmland
area during recent decades.
2
decreased by 108.17 km , showing the effect of the
As for the slope effect on land use changes, the dis-
Grain for Green Project in the study area. Jinan has
tributions of forest and grassland were still different
P. Li et al.
1160
Figure 6
|
|
Impact of urbanization on variability of precipitation in a city of North China
Land use changes at different elevations and slopes in Jinan during the periods 1980s and 2005.
Hydrology Research
|
51.5
|
2020
P. Li et al.
1161
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
from the other four land use types, similar to the
Spatiotemporal variability of flood season and nonflood
elevation effect on land use changes. As the slope
season precipitation
increased, the areas of forest and grassland first increased and then decreased, and the peak values of
area appeared in the regions with slope of 10 and 5 ,
Spatial patterns of flood season and nonflood season precipitation
respectively, which showed a strong suitability of the two land use types in the regions with relatively large
Based on the daily precipitation data at 24 gauging stations
slopes. Compared with the 1980s, the areas of forest
in Jinan during the period of 1966–2016, the spatial distri-
and grassland below 10 significantly decreased in 2005.
butions of average annual and flood season precipitation
Farmland, settlements, and wetland and waters were
are shown in Figure 7. Both spatial patterns are character-
mainly concentrated in the flat areas with slopes of 1–2 ,
ized as decreasing from southeast to northwest and
due to the fact that areas with slopes exceeding 5 are unsui-
southwest. The maximum average annual precipitation is
table for construction and farming activities. The area of
759.54 mm at the Zaolin Station in the southern mountai-
desert was small and scattered with stochastic distribution.
nous area, while the minimum value is 557.39 mm at the
Generally speaking, as the slope increases, the areas of var-
Baiyunhu Station in the east. The maximum value of the
ious land use types that are closely related to mankind’s
average precipitation during the flood season appeared at
producing activities and living behaviors will gradually
the Wupu Station, also located in the southern mountainous
decrease, while the distribution of land use types that are
area, reaching 579.58 mm, while the minimum value
less disturbed by humans gradually has the advantage,
appeared at the Baiyunhu Station with 408.48 mm.
showing a transition process from artificial ecosystems to natural ecosystems.
Figure 7
|
The spatial distributions of average AP and flood season precipitation in Jinan.
The annual and flood season precipitation in mountainous areas is generally higher than that in the plain areas.
P. Li et al.
1162
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
The settlements in concentrated areas are mostly in areas
Changes in precipitation between pre-change and post-
with relatively less precipitation. Compared with the
change point periods
characteristics of the terrain in the study area, the spatial distribution of average annual and flood season precipitation
According to the results of the abrupt change detection of
in Jinan has a significant relationship with the topography.
precipitation, the study period was divided into two subperiods: Period I: 1966–1989 (before the change point)
Abrupt changes of the flood season and nonflood season precipitation
and Period II: 1990–2016 (after the change point). The spatial variabilities in the differences between the prechange and post-change points for annual precipitation in
To comprehensively analyze and validate the abrupt
the study area are calculated and shown in Figure 9(a).
changes of flood season and nonflood season precipitation
For the spatial variability of flood season precipitation, the
in Jinan during the period of 1966–2016, the widely used
recorded data from the 23 flood season stations were incor-
nonparametric Mann–Kendall test method, the Pettitt test
porated to improve the spatial heterogeneity, which is
method, and the moving T-test method were adopted to
shown in Figure 9(b). Since the rainfall stations in the cen-
detect the abrupt changes of flood season and nonflood
tral city are mostly flood season stations, the flood season
season precipitation at 47 gauging stations in the study
data can well reflect the actual precipitation distribution.
area. The comparison between the abrupt change detection
The annual and flood season precipitation in the study
results obtained by the three different methods showed that
area have significantly increased during the post-change
the annual precipitation at all stations exhibited an upward
point period with different extents. The increase of annual
trend. The abrupt change points generally occurred in the
precipitation and flood season precipitation were similar
1960s, 1989, and 2002, among which most abrupt change
in spatial distributions. The increasing rates of the annual
points in 1989 were detected at the significance level 0.05
and flood season precipitation in the eastern and southwes-
(Figure 8). Most sites of the detected mutations are concen-
tern plain regions were relatively low with less than 10%,
trated in the urbanized areas of central Jinan and the
while the increasing rate of that in the central urban
southern mountainous areas, corresponding to the strong
region
change of land use structure during the rapid development
Especially for the Liujiazhuang Station close to the central
of Jinan.
city, the average flood season precipitation increased from
Figure 8
|
The abrupt change points of annual and flood season precipitation in Jinan.
and southern
mountain
was
relatively high.
P. Li et al.
1163
Figure 9
|
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
The amplitudes of annual (a) and flood season (b) precipitation during the pre-change and post-change periods in Jinan.
370.4 to 516.8 mm, which indicates that Jinan faced con-
were concentrated in the urban areas of central Jinan. The
siderable risk from flood disasters during the flood season
Xiaolipu Station, located near the central urban area,
in the central urban region over recent years.
increased from 22 to 33.5 days. The increase in the
To comprehensively study the characteristics of the
number of storm days throughout the year is obvious, and
variability of precipitation, two more indices, number of
the annual precipitation generally increased, which is
precipitation days and precipitation intensity, were adopted
shown in Figure 11. The proportion of rainstorm days at
for further analysis. The daily rainfall above 50 mm is
most stations has increased at different rates during the
counted as an indicator of storm intensity. To avoid the
flood season, and the stations with larger increases are con-
impact of the initial year difference, the precipitation data
centrated in central Jinan. The largest increase of rainstorm
of each site from 1980 to 2016 were selected for further
days was at the Guanying Station in the southern mountai-
analysis.
nous region. The proportion of heavy rain in the flood
As can be seen from Figure 10, the number of precipi-
season increased from 3.1 to 7%, and the proportion of
tation days throughout the year and flood season in each
heavy rain in the Xinglong Station, located in the central
rainfall station generally showed an increasing trend. The
urban area of Jinan, also increased by nearly 4%. The
spatial distribution of average annual precipitation days is
increase in the number of precipitation days at stations
similar to that of precipitation days in the flood season.
close to the urban area was significantly higher than the
The stations with a larger increase of annual precipitation
flood season, and the increase in precipitation indices
days were mostly concentrated in the southern mountainous
during the flood season was also significantly higher than
areas, and the Nanergao Station increased from 55.8 to 77.4
that of the whole year. It can be inferred that the threat of
days. In the flood season, the stations with a larger increase
flood disasters to the urban area of Jinan continues to
1164
Figure 10
P. Li et al.
|
|
Impact of urbanization on variability of precipitation in a city of North China
The changes of annual and flood season PDs during the pre-change and post-change periods in Jinan.
Hydrology Research
|
51.5
|
2020
1165
Figure 11
P. Li et al.
|
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
The amplitudes of rainstorm proportion during the pre-change and post-change periods in Jinan.
increase after the abrupt change year. The analysis of the precipitation days and precipitation intensity showed a significant increasing trend for annual and flood season scales, which indicates that beside the total amount, the frequency and intensity of precipitation in the study area also increased to some extent. Impact of urbanization on precipitation The arithmetic average of each index in the region is calculated to obtain the variabilities of land use types and precipitation indices between the pre-change point and post-change point periods (Figure 12). Six precipitation indices, including annual precipitation (AP), precipitation days (PDs), rainstorm days (RDs), AP in the flood season (AP_F), PDs in the flood season (PD_F), and RDs in the
Figure 12
|
The relationship between precipitation indices and land use changes in different regions of Jinan.
flood season (RD_F), were adopted in this study to describe the characteristics of precipitation. AP describes the
variabilities from frequency, and RD describes the variabil-
variabilities from rainfall amount, PD describes the
ities from intensity.
1166
P. Li et al.
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
As can be seen from Figure 12, the increase of settlement in the urban area is the largest with an increasing rate of 33.01%, followed by wetland water area with an increase of 25.94%, while the farmland in the urban area showed a significant decrease of 9.13%. The increase of urbanization is highest in the urban area, followed by suburban areas. According to the statistics of the Statistical Yearbook of Jinan City in 2005, the GDP per capita in the urban area is 44,805.6 yuan, Zhangqiu district is 22,483.6 yuan, Changqing district is 22,117.6 yuan, and Pingyin, Jiyang, and Shanghe counties are 20,992.6, 15,991.7, and 12,483.3 yuan, respectively. The level of per capita GDP in each region is consistent with the increase in the land use
Figure 13
|
Grey correlation analysis of precipitation indices and land use changes in Jinan.
type of settlements, indicating that the socioeconomic development is generally corresponds to the land use change.
and desert had the least impact on AP_F, with a correlation
Precipitation indicators show that the growth rates of
coefficient of 0.5. The fluctuation of RD_F was directly related
various precipitation indices in the urban area were generally
to the changes of cultivated land and settlements, with corre-
the most significant. Among them, the number of rainy days
lation coefficients of 0.73 and 0.7, respectively, while the
in the flood season increased by 58.95%, and the amount of
lowest correlation occurred in the grassland area. Wetland
precipitation in the flood season increased by 18.96%. Com-
and water, and settlements had a greater impact on PD than
pared with the average value of other areas in Jinan, the
other land use types, and the changes in desert were less
number of rainy days in the central district increased by
synchronized with that in PD. Generally speaking, the precipi-
15.38%, and the probability of heavy rains increased by
tation indices associated with the flood season in the study
4.73%, showing typical characteristics of the ‘rain island
area were more closely related to land use changes. Changes
effect’. The number of rainy days increased by 14.48% in
in farmland and settlements affected by urbanization were
Changqing District, while the number of heavy rains
more closely related to precipitation indices.
increased by 57.61% in Zhangqiu District. Although the
The above analysis shows that in the context of further
settlement area of Pingyin has increased significantly, the
development of the urban economy and society and drastic
actual urbanized area is relatively small, resulting in little
changes in land use types, the precipitation in urban areas
impact on precipitation. As for some stations in the south
during the flood season is more significantly affected com-
of Jiyang, precipitation indicators are generally affected
pared with other regions. This phenomenon may be the
because they are closer to the central urban area.
result of a combination of the heat island effect, the blocking
In order to further explore the relationship between land
effect, and the condensation effect (Mohammad et al. ).
use types and precipitation factors, GRA was used to calculate the correlation coefficient in each area. The indicators were firstly listed in order of the economic level of each district
DISCUSSION
from the largest to the smallest, and dimensionless was then carried out, taking the indicator of urban area as the initial
Different statistical test methods have their own advantages
value. The calculation results after dimensionless processing
and disadvantages, and the cross-verification of the results
are shown in Figure 13. The changes in land use and precipi-
obtained by the three abrupt change test methods in this
tation indices in different administrative areas were well
study could minimize the instability and maximize the
synchronized, with a range of correlation coefficients between
rationality of the determination of the abrupt change
0.5 and 0.75. Changes in the settlement had the greatest impact
point. In previous studies, it was found that the change in
on AP_F, with a correlation coefficient of 0.75, while grassland
reflection or absorption of solar radiation on the urban
1167
P. Li et al.
|
Impact of urbanization on variability of precipitation in a city of North China
Hydrology Research
|
51.5
|
2020
underlying surface has a certain effect on the surface energy
the spatial pattern and temporal trends of precipitation
balance, which is one of the physical mechanisms by which
and the changes of land use types in Jinan during the last
land use changes affect regional precipitation (Oke ;
60 years. Conclusions can be summarized as follows:
Arnfield ). Moreover, studies in Beijing and Tokyo showed that the increase of surface impermeable area
1. The land use types in Jinan experienced great changes
ratio makes the local evaporation decrease, the sensible
during the period 1980–2005, in which settlements and
heat flux increases, and the water vapor mixing in the
farmland were the most obvious changing types. The
boundary layer becomes more uniform, which affects the
area of settlements converted from farmland accounted
development of the local weather system (Zhang et al.
for 60.6% of the total transfer area of farmland, reflecting
; Souma et al. ). There are some previous studies
the rapid development of urbanization in Jinan and the
on simulating the impact of land use change on rainfall.
impact of growing population on land resources. It is
For example, the WRF model and UCM are recently used
worth mentioning that the growth rate of wetland water
in the field; however, it requires large amounts of computing
in Jinan ranked only second to the settlements, which
resources. More sophisticated models with long-term simu-
reflects the great efforts for ecological protection and
lations are required in the future to simulate the specific
restoration.
spatial correlation between land use change and precipitation (Miao et al. ; Wang et al. ).
2. The distribution of land use was characterized as a certain terrain gradient under the influence of topographic
Through the research, it could be found that the distri-
factors. In lower altitude and slope areas, settlements,
bution of cultivated land had a significant relation with the
farmland, and wetland and water were more concen-
requirements of crop growth and the convenience of cultiva-
trated,
tion, which was obviously controlled by the topographic
producing activities and living behaviors. The forest and
factors. In addition to urbanization, the transferred area of
grassland show a strong adaptability of higher altitudes
farmland is also affected by environmental protection pol-
and steeper slopes, accounting for a larger proportion
icies such as returning farmland to forest and returning
of land use types in the areas. The influences of aspect
farmland to lake. It is important to note that our study focused
on the distribution of various land use types were not
on the local recycled precipitation instead of large-scale
obvious.
which
are
closely
related
to
mankind’s
advected precipitation which is beyond the scope of this
3. The spatial pattern of precipitation in Jinan is generally
study but deserves further research. For example, Daniels
characterized as decreasing from southeast to northwest,
et al. () indicated that the impact of land use change on
which is mainly affected by monsoon and topography.
summer precipitation in the Netherlands was smaller than
The temporal variability of precipitation is obvious, and
that of climate change, but it cannot be ignored. Similar find-
the intra-annual distribution is uneven. The average AP
ings were found in the Beijing–Tianjin–Hebei region, about
and flood season precipitation in Jinan showed an overall
11% of the reduction of the total precipitation was caused
increasing trend. The significant abrupt change point of
by the decrease of local recycled precipitation induced by
precipitation generally occurred in 1989 detected by
urbanization, while the remainder 89% was caused by the
three methods. The growth rate of precipitation in eastern
reduction of large-scale advected precipitation (Li et al.
and western regions was relatively low, while that in the
). It remains a great challenge to clearly elucidate the
central urban region was relatively high. The frequency
independent effects of land use change on climate.
and intensity of precipitation in Jinan also increased significantly. 4. Compared with other regions, the urban region with the
CONCLUSIONS
largest area of settlements produced the largest increase of precipitation especially for the post-change period,
In this study, the impact of urbanization on annual and
and the water and wetland area have also increased sig-
flood season precipitation was qualitatively analyzed from
nificantly. The increase of the number of RDs and the
1168
P. Li et al.
|
Impact of urbanization on variability of precipitation in a city of North China
amount of precipitation during the flood season in the urban area were much higher than that in other areas. The changes in land use and precipitation indices in different administrative areas were well synchronized. The precipitation indices in the flood season were more closely related to changes in farmland and settlements affected by urbanization with a maximum correlation coefficient of 0.75. The results obtained in this study are of great significance for the further urban development and the coordination of the water and environment.
ACKNOWLEDGEMENTS This study is jointly supported by the National Key Research and
Development
Program
of
China
(Grant
No.
2017YFC1502703), the Beijing Natural Science Foundation (Grant No. 8202030), the National
Natural Science
Foundation of China (Grant No. 91647202), Major Science and Technology Program for Water Pollution Control and Treatment (Grant No. 2017ZX07302-04), and 111 project (B18006).
CONFLICT OF INTEREST The authors declare that they have no conflict of interest.
DATA AVAILABILITY STATEMENT All relevant data are available from an online repository or repositories (National Science and Technology Foundation Platform Project ‘Data-Sharing Network of Earth System Science’ (www.geodatda.cn)).
REFERENCES Arnfield, A. J. Two decades of urban climate research: a review of turbulence, exchanges of energy and water, and the urban heat island. International Journal of Climatology 23 (1), 1–26. Chang, X., Xu, Z., Zhao, G., Cheng, T. & Song, S. Spatial and temporal variations of precipitation during 1979–2015 in
Hydrology Research
|
51.5
|
2020
Jinan City, China. Journal of Water and Climate Change 9 (3), 540–554. Changnon, S. A. Rainfall changes in summer caused by St. Louis. Science 205 (4404), 402–404. Cheng, T., Xu, Z., Hong, S. & Song, S. Flood risk zoning by using 2D hydrodynamic modeling: a case study in Jinan city. Mathematical Problems in Engineering 5659197. Daniels, E., Lenderink, G., Hutjes, R. & Holtslag, A. Relative impacts of land use and climate change on summer precipitation in the Netherlands. Hydrology and Earth System Sciences 20 (10), 4129–4142. Givati, A. & Rosenfeld, D. Quantifying precipitation suppression due to air pollution. Journal of Applied Meteorology 43 (7), 1038–1056. Golroudbary, V. R., Zeng, Y., Mannaerts, C. M. & Su, Z. B. Detecting the effect of urban land use on extreme precipitation in the Netherlands. Weather and Climate Extremes 17, 36–46. Gu, X., Zhang, Q., Singh, V. P., Song, C., Sun, P. & Li, J. Potential contributions of climate change and urbanization to precipitation trends across China at national, regional and local scales. International Journal of Climatology 39 (6), 2998–3012. Hammond, M. J., Chen, A. S., Djordjevic, S., Butler, D. & Mark, O. Urban flood impact assessment: a state-of-the-art review. Urban Water Journal 12 (1 SI), 14–29. Jiang, L. G., Nie, X. H. & Liu, E. F. Analysis on the spatial structure of urban land use in Jinan city. Economic Geography 23 (1), 70–73 (in Chinese). Kaufmann, R. K., Seto, K. C., Schneider, A., Liu, Z., Zhou, L. & Wang, W. Climate response to rapid urban growth: evidence of a human-induced precipitation deficit. Journal of Climate 20 (10), 2299–2306. Kendall, M. G. Rank Correlation Methods, 4th edn. Charles Griffin, London, UK. Li, X., Zhang, X., Zhang, L. & Zheng, J. The varied impacts of land use/cover change on summer precipitation over eastern China. Geographical Research 36, 1233–1244. Lu, M., Xu, Y., Shan, N., Wang, Q., Yuan, J. & Wang, J. Effect of urbanization on extreme precipitation based on nonstationary models in the Yangtze River Delta metropolitan region. Science of the Total Environment 673, 64–73. Lu, D., Gao, G., Lu, Y., Xiao, F. & Fu, B. Detailed land use transition quantification matters for smart land management in drylands: an in-depth analysis in northwest China. Land Use Policy 90, 104356. Mahmood, R., Pielke, R. A. S., Hubbard, K. G., Niyogi, D., Dirmeyer, P. A., Mcalpine, C., Carleton, A. M., Hale, R., Gameda, S., Beltran-Przekurat, A., Baker, B., Mcnider, R., Legates, D. R., Shepherd, M., Du, J., Blanken, P. D., Frauenfeld, O. W., Nair, U. S. & Fall, S. Land cover changes and their biogeophysical effects on climate. International Journal of Climatology 34 (4), 929–953. Mann, H. B. Non-parametric test against trend. Econometrical 13, 245–259.
1169
P. Li et al.
|
Impact of urbanization on variability of precipitation in a city of North China
Miao, S., Chen, F., Li, Q. & Fan, S. Impacts of urban processes and urbanization on summer precipitation: a case study of heavy rainfall in Beijing on 1 August 2006. Journal of Applied Meteorology and Climatology 50 (4), 806–825. Mohammad, P., Goswami, A. & Bonafoni, S. The impact of the land cover dynamics on surface urban heat island variations in semi-arid cities: a case study in Ahmedabad city, India, using multi-sensor/source data. Sensors 19 (17), 3701. Oke, T. R. The energetic basis of the urban heat-island. Quarterly Journal of the Royal Meteorological Society 108 (455), 1–24. Pathirana, A., Denekew, H. B., Veerbeek, W., Zevenbergen, C. & Banda, A. T. Impact of urban growth-driven land use change on microclimate and extreme precipitation – a sensitivity study. Atmospheric Research 138, 59–72. Petitt, A. N. A non-parametric approach to the change point problem. Applied Statistics 28, 125–135. Rosenfeld, D. Suppression of rain and snow by urban and industrial air pollution. Science 287 (5459), 1793–1796. Souma, K., Tanaka, K., Suetsugi, T., Sunada, K., Tsuboki, K., Shinoda, T., Wang, Y., Sakakibara, A., Hasegawa, K., Moteki, Q. & Nakakita, E. A comparison between the effects of artificial land cover and anthropogenic heat on a localized heavy rain event in 2008 in Zoshigaya, Tokyo, Japan. Journal of Geophysical Research-Atmospheres 118 (20), 11600–11610. Tan, X. & Deng, J. Grey relational analysis: a new method for multivariate statistical analysis. Statistical Research 11 (3), 46–48 (in Chinese).
Hydrology Research
|
51.5
|
2020
Wang, J., Feng, J. & Yan, Z. Impact of extensive urbanization on summertime rainfall in the Beijing region and the role of local precipitation recycling. Journal of Geophysical Research-Atmospheres 123 (7), 3323–3340. Yu, M., Miao, S. & Zhang, H. Uncertainties in the impact of urbanization on heavy rainfall: case study of a rainfall event in Beijing on 7 August 2015. Journal of Geophysical Research-Atmospheres 123 (11), 6005–6021. Zhang, C. L., Chen, F., Miao, S. G., Li, Q. C., Xia, X. A. & Xuan, C. Y. Impacts of urban expansion and future green planting on summer precipitation in the Beijing metropolitan area. Journal of Geophysical Research-Atmospheres 114, D02116. Zhang, W., Villarini, G., Vecchi, G. A. & Smith, J. A. Urbanization exacerbated the rainfall and flooding caused by hurricane Harvey in Houston. Nature 563 (7731), 384. Zhu, X., Zhang, Q., Sun, P., Singh, V. P., Shi, P. & Song, C. Impact of urbanization on hourly precipitation in Beijing, China: spatiotemporal patterns and causes. Global and Planetary Change 172, 307–324. Zuo, D., Xu, Z., Yao, W., Jin, S., Xiao, P. & Ran, D. Assessing the effects of changes in land use and climate on runoff and sediment yields from a watershed in the Loess Plateau of China. Science of the Total Environment 544, 238–250. Zuo, D., Cai, S., Xu, Z., Li, F., Sun, W., Yang, X., Kan, G. & Liu, P. Spatiotemporal patterns of drought at various time scales in Shandong Province of Eastern China. Theoretical Applied Climatology 131 (1–2), 271–284.
First received 11 June 2020; accepted in revised form 28 August 2020. Available online 1 October 2020
1170
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Effect of initial plant density on modeling accuracy of the revised sparse Gash model: a case study of Pinus tabuliformis plantations in northern China Yiran Li, Xiaohua Liu, Chuanjie Zhang, Zedong Li, Ye Zhao and Yong Niu
ABSTRACT An accurate quantitative description of interception is necessary to understand regional water circulation. The revised sparse Gash model (RSGM) is currently used to estimate interception loss. Previous studies have proven that changes in initial plant density, which are caused by thinning, affect the accuracy of RSGM; however, the direct effect of initial density on modeling accuracy remains poorly understood because few studies have collected field data of the same species with various initial densities under similar site conditions. Therefore, six Pinus tabuliformis Carr. plantations with various initial densities were assessed from May to October 2016 in northern China. In summary, RSGM performs better with higher initial densities, and it cannot be suitably applied for plantations with lower initial densities, with the relative error ranging from 18.38 to 53.03%. Sensitivity analysis indicated that the predicted interception is highest sensitive to canopy structure, irrespective of initial density. The influence of climate parameters on simulated results decreased, as initial density increased. These support the notion that amending the representation of the canopy structure in the model and improving the estimation methods for determining the evaporation rate in open canopies can improve accuracy, and that the use of RSGM must first involve the consideration of initial density. Key words
| initial plant density, interception loss, revised sparse Gash model, sensitivity analysis, simulation accuracy
Yiran Li† Zedong Li Yong Niu (corresponding author) College of Forestry, Shandong Agricultural University, Shandong 271018, China E-mail: niuyong1988@126.com Yiran Li† Xiaohua Liu† School of Soil and Water Conservation, Beijing Forestry University, Beijing 100083, China Ye Zhao College of Biological Sciences and Technology, National Engineering Laboratory for Tree Breeding, Beijing Forestry University, Beijing 100083, China Chuanjie Zhang† College of Water Conservancy and Civil Engineering, Shandong Agricultural University, Shandong 271018, China †
HIGHLIGHTS
• • • • •
Application of the revised sparse Gash model (RSGM) in six Chinese pine plantations of different initial densities. RSGM still does not perform well at the sparse initial densities. Changes in initial plant density can affect the simulation accuracy of RSGM. Discuss the possibility of perfecting RSGM to improve its accuracy. Explore the changes in sensitivity of the parameters in RSGM under different initial densities.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/ licenses/by-nc-nd/4.0/). doi: 10.2166/nh.2020.007
Y. R. Li, X. H. Liu and C. J. Zhang should be considered co-first authors
1171
Y. Li et al.
|
Initial plant density effects modeling precision
Hydrology Research
|
51.5
|
2020
GRAPHICAL ABSTRACT
INTRODUCTION The development of plantations has continued to increase in
(RSGM) (Valente et al. ; Limousin et al. ), and the
recent years in China (Gao et al. ; Ma et al. ), with
so-called RS-Gash model (Cui & Jia ). Prior studies
the area under plantation increasing to 693,338 km2
have mainly focused on the applicability of either the
(National Forestry and Grassland Administration ).
original Gash model or its revised versions in plantations,
Nevertheless, several underlying issues, such as poor
natural forests, or shrub forests (Sadeghi et al. ;
quality of plantations, have emerged because of the rapid
Fernandes et al. ; Ghimire et al. ; Fathizadeh et al.
increase in green areas. With abundant forests in northern
). Similarly, some previous studies have evaluated the
China, water resources are considered a crucial restrictive
effect of rainfall characteristics or model parameters on
factor for plantation growth. In these areas with limited
the prediction accuracy of the model using only a single-
water resources, a good understanding of ecohydrological
tree species in a particular area (Sadeghi et al. ; Su
processes plays a critical role in the essential components
et al. ). In addition, despite some research on the
of water resources management and stand structure
response of model accuracy to tree density, available
optimization.
information is limited.
A substantial amount of precipitation can be intercepted
Tree densities, tree species, and canopy spatial arrange-
by and can subsequently evaporate from forest canopies
ments are significant in improving the prediction of
(van Dijk & Bruijnzeel a; Hassan et al. ). This
interception loss (Owens et al. ; Bui et al. ; Chen
intercepted precipitation plays a crucial role in both regional
et al. ). Managing initial density could be more effective
water balance in forested areas and spatial rainfall distri-
than altering species composition in decreasing the impacts
bution (Zeng et al. ; Llorens & Domingo ); in
on water resources (Licata et al. ). Furthermore, in com-
addition, it contributes to the growth of woods. Models of
parison with changing forest structure variables, it could be
interception loss (I ), such as the original Gash model
simpler and cruder in influencing the moisture cycle of
(Gash ), are important tools for researching ecohydrolo-
forest watersheds. Therefore, studies on hydrological model-
gical characteristics; these models have garnered substantial
ing that have focused on initial density are considered more
attention from scholars, and several modified or improved
significant, particularly for poor-quality plantations, because
models have now been developed, such as the sparse Gash
of unsuitable initial density or water resources acting as
model (Gash et al. ), the revised sparse Gash model
restraining
factors.
Additionally,
some
studies
have
1172
Y. Li et al.
|
Initial plant density effects modeling precision
Hydrology Research
|
51.5
|
2020
investigated the relationship between initial density changes
study were to explore the effect of initial plant densities on
caused by thinning and the accuracy of canopy interception
the predicted accuracy of RSGM and to evaluate the
models. Limousin et al. () reported that RSGM provides
impact of different initial densities on model parameters as
good estimates for the canopy interception of stands with
well as examine the possibility of perfecting the model to
different initial densities caused by thinning. Shinohara
improve its accuracy.
et al. () compared the performance of three canopy interception models (Mulder, sparse Gash, and WiMo) in forests both pre- and post-thinning and concluded that the
MATERIALS AND METHODS
canopy interception model might be unsuitable for Japanese cedar (Cryptomeria japonica) forests after intensive thin-
Site description and experimental design
ning. Ma et al. () explored the differences between the application of the sparse Gash model and the WiMo
The research site (Figure 1) is located 8 km northeast
model in pre- and post-thinning plantations, respectively,
of Mount Tai (Northern China) in the Yaoxiang National
and demonstrated that both models had better performance
Forest
in the un-thinned stands than in the post-thinned stands.
36 20 30″). The research site elevation ranges from 400 to
However, these observations only focused on the effects of
956 m (average: 710 m). This region is located in the warm
thinning-related initial density changes on canopy intercep-
temperate zone and has a semi-humid and monsoon cli-
tion prediction, and studies regarding the direct impact of
mate, with a mean annual temperature of 18.5 C and a
initial plant density variations on the modeling accuracy of
mean annual frost-free period of 198 days. The multi-year
RSGM have rarely been reported. We expected a novel find-
average precipitation is 727.9 mm, with 75% of the annual
ing when using RSGM in plantations with different initial
precipitation being concentrated from June to September.
plant densities.
The soil type is brown soil. The underlying rock is ancient
Park
(E11 050 39″–117 090 26″,
N36 170 58″–
0
Initial density is the most common index of forest man-
gneiss of a high metamorphic grade, with a thickness
agement in China. Since the beginning of the last century,
ranging from 10 to 90 cm; in most areas, thickness ranges
China has afforested large swathes of land to compensate
from 30 to 50 cm. Vegetation types belong to coniferous
for the insufficient green area. Although the area under
and deciduous broad-leaved forests in the warm temperate
afforestation is rapidly increasing, the initial plant density
zone, and the main tree species include Pinus tabuliformis,
of the same tree species in the same area is different. More-
Quercus acutissima Carr., Pinus densiflora Sieb., Robinia
over, as several familiar forest management measures to
pseudoacacia, and Castanea mollissima (Sun et al. ).
improve forest quality, such as thinning and replanting, are
In the study area, six plots (each with a projection area
infrequently used in most regions of China, these together
of 100 m2, length of 20 m, and width of 5 m) were set up
result in complications such as poor plantation quality.
according to the initial densities of P. tabuliformis planta-
Therefore, numerous studies focusing on ecohydrological
tion as follows: (1) 720 trees ha 1; (2) 1,010 trees ha 1; (3)
and biogeochemical fluxes as well as on either growth
1,334 trees ha 1; (4) 1,695 trees ha 1; (5) 1,720 trees ha 1;
characteristics or economic benefits of stands with different
and (6) 2,746 trees ha 1. We conducted a forestland survey
initial densities have provided a scientific basis for further
in March 2016 (Table 1) and installed the HOBO automatic
forest management in China. Furthermore, the accuracy
weather station (Onset, USA) surrounding the open area of
with which the model of interception loss estimates
the sample plots to collect necessary meteorological data
canopy interception is considered a significant part of
such as rainfall, wind speed, and radiation. The weather
such forest management, particularly for the response of
station collected data at 15-min intervals, and we down-
model simulation efficiency to initial plant density, which
loaded the data once a month. Because the six plots are
is a gap that cannot be ignored for the development of
relatively close in terms of distance and there is little
interception models as well as for the fine refinement man-
change in meteorological factors, only one meteorological
agement of forests. Therefore, the objectives of the present
station was installed.
Y. Li et al.
1173
Figure 1
Table 1
|
|
|
Initial plant density effects modeling precision
Hydrology Research
|
51.5
|
2020
Study plot location.
Investigation results of stand plots 1
Plot
Sp.
Sl. ( )
Id (trees ha
1
P. tabuliformis
22
2 3
TDP (trees)
Ah (m)
ADBH (cm)
c
760
8
10.12 ± 0.41
17.87 ± 0.79
0.74
23
1,010
12
8.98 ± 0.48
16.88 ± 1.03
0.75
21
1,334
14
7.44 ± 0.46
15.04 ± 0.72
0.78
4
23
1,695
16
9.04 ± 0.52
13.98 ± 0.81
0.80
5
25
1,720
17
7.01 ± 0.34
12.25 ± 0.75
0.80
6
22
2,746
28
12.18 ± 0.63
15.19 ± 0.58
0.70
)
1
ADBH, average diameter at breast height (mm); Sp., species; Sl., slope ( ); Id, initial density (trees ha ); TDP, tree density per plot (trees); Ah, average height (m); c, canopy cover.
Field measurements
collection, the collectors were arranged parallel to the slope along with the sample sites and the height from the
Total precipitation (TP) was determined according to the
collector to the ground was 50 cm, connecting the collect-
data from the automatic weather station. Due to the proxi-
ing barrel at the bottom of the pipe (Sheng et al. ; Ji &
mity of the measured sample plots to each other, only one
Cai ). We measured the amount of water in the collect-
station was placed in the central position relative to each
ing barrel after each rainfall event. TF depth (mm) was
plot.
calculated using the following equation:
Three PVC tubes (DN90) with a length of 8 m and an inner diameter of 9 cm were cut open to collect throughfall (TF) in each plot. To avoid the influence of herbs on TF
TF ¼
3 × V × 1000 3 × L × W × cos α
(1)
1174
Y. Li et al.
|
Initial plant density effects modeling precision
Hydrology Research
|
51.5
|
2020
where V is the amount of water collected in the collecting
by Valente et al. () and simplified by Limousin et al.
barrel after each rainfall event (L), L is the length of the
():
PVC tube (m), W is the inner diameter of the PVC tube (m), α is the slope of the sample plots ( ), and L × W × cos α
nX þm
2
is the cross-sectional area of the PVC tube (m ).
Ij ¼ c
j¼1
breast height, and three trees from each class were selected
PG,j þ
j¼1
Based on the survey results of the sample plots, the trees were categorized into three classes based on diameter at
m X
þ cptc
n q X j¼1
n n X X cE 0 P ) þ c P0 G þ qcStc (P G,j G R j¼1
j¼1
E 1 (PG,j P0 G ) R (3)
as samples. To measure stemflow (SF), a rubber pipe cut in the middle was stapled onto the trunks in a spiral shape. The gap between the rubber pipe and trunk was sealed
where n is the number of rainfall events during which the
with silicone, the bottom of the rubber pipe was connected
canopy reached saturation, m is the number of rainfall
to a collecting bucket, and the amount of water collected in
events that were insufficiently large to saturate the canopy,
this bucket was measured after each rainfall event. SF depth
c is the fraction of canopy cover, PG is the total rainfall is the wet canopy evaporation rate per amount (mm), E
(mm) was calculated using the following equation:
unit area of canopy cover (mm; calculated by the 1 Xn Gn SFj ¼ Mn i¼1 K M n
(2)
Penman–Monteith equation) (Gash et al. ; Valente is the mean rainfall intenet al. ; Limousin et al. ), R sity (mm/h), P0G is the minimum amount of precipitation
where M is the mean number of trunk plants in the pro-
required to saturate the canopy (mm), ptc ¼ pt =c is the pro-
jected area of the plot, n is the number of the diameter
portion of rain diverted into SF per unit area of cover,
class, Gn is the amount of SF in each diameter class (mm),
Stc ¼ St =c is the stem storage capacity per unit cover area,
Kn is the projected canopy area in each diameter class
and q is the number of saturated trunks.
(cm2), and Mn is the number of trees in each diameter class. Derivation of the parameters used in the model
REVISED SPARSE GASH MODEL It was evident that the canopy storage capacity (S) is an Formula
important parameter in the model; in the present study, S was estimated using the original regression method based
The original Gash model was proposed by Gash (). For
on the relationship between TP and TF. This method was
simplicity, some assumptions have been made: (1) a series of
proposed by Leyton et al. (); owing to its effectiveness
discrete storms are used to represent the real rainfall pat-
because of its simplicity (Liu ), this method has been
tern, and there is a sufficiently long interval between these
used more often than the method described by Klaassen
storms to allow the canopy and trunks to dry; (2) the con-
et al. () as well as has been used by Gash & Morton
ditions of rainfall and evaporation remain sufficiently
() and Gash (). André et al. () reported that
similar throughout all storms during this period; and (3)
the estimated performance of Leyton’s approach is better
there is no water drip from the canopy during wetting-up,
than a spraying laboratory experiment and mechanistic
and large amount of water on the canopy rapidly evaporates
modeling exercise.
at the end of a storm. The original Gash model was also
The minimum amount of water required to saturate the
revised based on these assumptions (Gash et al. ;
canopy was determined using the equation provided by
Valente et al. ; Limousin et al. ). In the present study, the following equation of RSGM was used, which is used for calculating canopy interception that was proposed
Limousin et al. (): E)S c ln (1 (E= R)) P0G ¼ (R=
(4)
1175
Y. Li et al.
|
Initial plant density effects modeling precision
Hydrology Research
where Sc ¼ S=c is the canopy storage capacity per unit area
|
51.5
|
2020
The rainfall distribution components of P. tabuliformis plantation of different densities varied during the study
of cover (in mm). The minimum amount of water that can fill the trunk
period (Table 2). With the same total rainfall, TF decreased
storage capacity is calculated using the following equation
and I increased with increasing initial density, but they
and is denoted as
P00G
(Valente et al. ; Limousin et al.
increased and decreased when the initial density increased from 1,720 to 2,746 trees ha 1, respectively. Typically, TF
):
and I tended to increase with increasing initial density P00G
0 R E))(S ¼ (R=( tc = ptc ) þ PG
(5)
during the study period. By contrast, SF initially increased with increasing initial density, following which it decreased
In addition, TF and SF can be simulated using the fol-
and increased further when initial density increased from
lowing equations provided by the model: q X
SFj ¼ cptc
j¼1 nþm X
q X
0 R))(P (1 (E= G,j P G,j ) qcStc
when initial density increased from 1,334 to 1,720 trees ha 1 1,720 to 2,746 trees ha 1. In general, SF decreased with increasing density.
(6)
j¼1
TFj ¼
j¼1
nþm X
PG,j
j¼1
nþm X j¼1
Ij
q X
Model parameters SFj
(7)
j¼1
We calculated the parameters of RSGM using the previously described methods (Table 3). S, estimated from the regression between TF and TP (Figure 2), increased with
Sensitivity analysis In the present study, we referred to separate sensitivity analyses in previous studies for each density (cf. Valente et al. ; Fan et al. ; Fathizadeh et al. ; Liu et al. ), R, St, and pt were tested in the range of an and S, c, E, increase and decrease by up to 50% of their original values and in terms of the results of models compared with field values.
increasing initial density. The value of S for 2,746 trees ha 1 (3.973 mm) was approximately four times greater than that for 720 trees ha 1 (0.894 mm). Values generated for Sc and P0G
ranged
from
1.207 mm
(for
720 trees ha 1)
to
1
5.706 mm (for 2,746 trees ha ) and from 1.229 mm (for 720 trees ha 1) to 5.807 mm (for 2,746 trees ha 1), respectively. Furthermore, the variation trends of Sc and P0G were the same as that of S, and the first two parameters for 2,746 trees ha 1 were approximately five times greater than those for 720 trees ha 1. Similar findings were obtained for St, pt, and P00G (St and pt were estimated from the regression
RESULTS Rainfall partitioning
between SF and TP, as shown in Figure 3); they typically increased with increasing initial density, and their values
Table 2
|
Rainfall redistribution composition of P. tabuliformis forests 1
In the present study, 8 h was considered the standard for
Plot
Sd. (trees ha
dividing rainfall events. Based on rainfall data from May
1
720
2016 to October 2016, a total of 28 rainfall events with a
2
1,010
total amount of 430.3 mm and a mean of 15.4 mm were
3
recorded in the study area. Based on individual rainfall
4
events, the maximum amount, duration, and intensity of rainfall in the plots were 73.7 mm, 37 h, and 26.32 mm/h, respectively, and the corresponding minimum values were 1.1 mm, 0.75 h, and 0.08 mm/h, respectively.
TP (mm)
TF (mm)
SF (mm)
I (mm)
430.3
357.9
1.6
70.8
316.5
20.3
93.6
1,334
296.8
23.6
109.9
1,695
286.2
19.7
124.4
5
1,720
277.2
22.2
131.0
6
2,746
280.1
25.3
124.9
)
TP, total precipitation (mm); TF, throughfall (mm); SF, stemflow (mm); I, interception loss (mm).
Y. Li et al.
1176
Table 3
|
|
Initial plant density effects modeling precision
Hydrology Research
|
51.5
|
2020
Parameters of RSGM for P. tabuliformis plantation 1
S (mm)
Sc (mm)
St (mm)
pt
c
P 0G (mm)
P 00G (mm)
720
0.894
1.207
0.035
0.006
0.74
1.229
7.244
2
1,010
1.281
1.708
0.394
0.073
0.75
1.738
7.394
3
1,334
2.361
3.027
0.404
0.081
0.78
3.027
8.234
4
1,695
2.774
3.467
0.414
0.073
0.80
3.529
9.422
5
1,720
3.076
3.797
0.363
0.075
0.81
3.865
8.865
6
2,746
3.937
5.706
0.545
0.094
0.69
5.807
11.795
Plot
Id (trees ha
1
)
1
Id, initial density (trees ha ).
Figure 2
|
Relationship between TF and TP for rainfall events. In the upper left of each figure, lowercase letters represent the sample plots with different initial densities (a ¼ 720
trees ha 1, b ¼ 1,010 trees ha rainstorm event.
1
, c ¼ 1,334 trees ha
1
, d ¼ 1,695 trees ha
1
1
, e ¼ 1,720 trees ha
1
, and f ¼ 2,746 trees ha
); the same is true in Figure 5. Each circle denotes a
ranged from 0.035 mm (for 720 trees ha 1) to 0.545 mm (for
calculations ranging from 36.0 mm (for 720 trees ha 1) to
2,746 trees ha 1) for St, from 0.006 (for 720 trees ha 1) to
104.4 mm (for 2,746 trees ha 1) and from 1.6 mm (for 720
1
0.096 (for 2,746 trees ha ) for pt, and from 7.244 mm (for
trees ha 1) to 19.6 mm (for 2,746 trees ha 1), respectively.
720 trees ha 1) to 11.795 mm (for 2,746 trees ha 1) for P00G .
Moreover, the simulated TF value decreased with an increasing initial density as a whole, differing from the trend of
Estimated results
measured values, and ranged from 306.3 mm (for 2,746 trees ha 1) to 392.7 mm (for 720 trees ha 1). In addition,
The simulated TF value was overestimated, whereas the
the simulation results indicated that irrespective of the
simulated I and SF values were less than the measured
initial density, there was substantially high I resulting from
values (Table 4). The simulated I and SF values showed a
evaporation from saturation until rainfall stopped (6.75–
uniform variation trend during the study period, with
28.33%) and after rainfall stopped (62.00–67.85%), and the
Y. Li et al.
1177
|
Figure 3
|
Initial plant density effects modeling precision
Hydrology Research
|
51.5
|
2020
Relationship between SF and TP for rainfall events. Each circle denotes a rainstorm event. In the upper left corner of each figure, lowercase letters represent sample plots with different initial densities (specific densities are the same as those in Figure 2).
Table 4
|
Simulation results of RSGM
trunks, the other model components increased with increasing initial density.
1
Id (trees ha
)
We used the mean of the absolute error and relative error
Components of the model
720
1,010
1,334
1,695
1,720
2,746
to describe the simulation accuracy of RSGM for I, TF, and
For m small storms, insufficient to saturate the canopy (mm)
2.34
3.55
5.17
5.31
5.37
18.09
SF (Figure 4). Regarding SF, an increase in initial density decreased the absolute error; in general, there was a tendency
Wetting-up the canopy, for 0.40 n storms > P0G that saturate the canopy (mm)
0.55
0.88
1.41
1.26
1.26
toward the lower absolute error of predicted TF and SF with increasing initial density, whereas the opposite was observed for I. Although there was a corresponding slight upward and
Evaporation from saturation until rainfall ceases (mm)
10.21 10.02 9.58
9.53
9.44
7.05
Evaporation after rainfall stops (mm)
22.34 30.74 49.58 63.79 70.74 70.86
Evaporation from trunks (mm)
0.75
downward trend in the absolute errors of TF and I when the initial density varied from 720 to 1,750 trees ha 1, the above-
8.22
7.77
7.31
6.49
7.18
mentioned conclusion was obtained irrespective of this trend. These relationships are presented in Figure 4(a). Similarly, as shown in Figure 4(b), for I, there was a clear negative correlation between initial density and relative error,
Simulated interception loss 36.0 (mm)
53.1
73.0
87.1
93.3
104.4
which ranged from 18.38% (for 2,746 trees ha 1) to 53.03% (for 720 trees ha 1). However, for SF, there was a general
Simulated TF (mm)
392.7 358.6 337.5 326.5 319.2 306.3
positive correlation between initial density and relative
Simulated SF (mm)
1.6
error, which ranged from 2.01% (in 720 trees ha 1) to
18.7
19.9
16.8
17.8
19.6
21.55% (in 2,746 trees ha 1). Regarding TF, although the lowest I occurred under saturated canopy conditions (1.10–
trend of relative error with initial density was as variable as
1.35%). With the exception of the amount of evaporation
the trend of absolute error, the difference was that the vari-
from canopy saturation until rainfall stopped and from
ation of the former was more similar to the convex type.
Y. Li et al.
1178
Figure 4
|
|
Initial plant density effects modeling precision
Hydrology Research
|
51.5
|
2020
The error between the simulated and measured values of interception varied with the initial density, including (a) absolute error and (b) relative error.
The relative error of TF in high-density stands was only
St, and pt – showed a positive correlation with the prec, E,
slightly smaller than that in low-density stands (10.14 and
dicted I. Moreover, with a 50% increase or decrease in all
10.50%, respectively), and the highest relative error was
the parameters tested, the representation indicates that S
noted in medium-density stands (15.85% in 1,720 trees ha 1).
was the chief parameter for the change in predictions regardless of the initial density, which can result in a maxi-
Sensitivity analysis Figure 5 illustrates the sensitivity analysis results for the vari all remaining parameters – S, ation of predicted I. Besides R,
Figure 5
|
mum change of ±30% in calculation, and the minimum influence on the prediction of interception was provided on the prediction results by pt. Subsequently, the effect of E when the parameters tended to be higher than that of R
, mean rainfall intensity; c, canopy cover; E, wet canopy evaporation rate per unit area of Sensitivity analysis of RSGM for change in predicted I. S, canopy storage capacity; R canopy cover; St, stem storage capacity; and pt, drainage partitioning coefficient. The lowercase letters on the right side of the longitudinal axis represent the samples of different initial densities in each figure, and the exact meaning is the same as in Figure 2.
1179
Y. Li et al.
|
Initial plant density effects modeling precision
changed in the range of 0 to þ50%. However, when the par ameters tested changed from 50 to 0%, the influence of R
Hydrology Research
|
51.5
|
2020
forest, and mixed White Oak forest). Therefore, we compared our estimated S with the variation range (0.1–3.0 mm) of S in
for all plots. Furthermore, higher was higher than that of E and E. initial density corresponded to lower influence of R
coniferous forests under different initial densities, as summar-
and c, the influence of R on the With a 50% decrease in R
P. tabuliformis forests remained higher than that in coniferous
prediction results was higher than that of c, and there was
forests of similar density such as Pinus sylvestris, Picea sitch-
a change in the corresponding average simulated cumulative
ensis, and Pinus elliottii. This may be due to the distinctive
I of 29.38 and 20.12% within 760 trees ha 1, and 19.69 and
broad-ovate crown profile of P. tabuliformis. The developed
1
ized by Llorens & Gallart (). We found that the S value in
17.01% within 1,010 trees ha . Further, the effect of c on when the initial density simulation was higher than that of R
sclerenchyma on the epidermis of needles and sunken pores
exceeded 1,010 trees ha 1.
carry more water than the coniferous forests evaluated in
can provide appropriate conditions for P. tabuliformis to other studies (China Agricultural Encyclopedia Forestry Volume Editorial Committee ). Remarkably, the variation
DISCUSSION
trend of the S value of these species exhibited an approximate wavy line with an increasing initial density, rather than follow-
During the study period, the average I measured in
ing a linear trend as observed in the present study. We propose
P. tabuliformis plantation plots ranged from 16.46 to
that the S value of the same coniferous species exhibits certain
30.44% of the total rainfall. Typically, I in temperate conifer-
threshold values with increasing initial density and that the S
ous forests ranges from 10 to 60% (Teklehaimanot et al. ;
value will continue to increase and decrease. Differences in
Loustau et al. ; Licata et al. ), with I measured in our
results between this and other studies may be due to the fact
study falling in the upper half of this range. The I range in our
that the maximum initial density of the P. tabuliformis planta-
study is consistent with those for the main plantation forest in
tion in this study (2,746 trees ha 1) was lower than the initial
China (14.7–31.8%) and coniferous forests (21.0–48.0%; Car-
density corresponding to the theoretical first threshold of S
lyle-Moses ; Wei et al. ), and it is within the range of
of P. tabuliformis artificial forests. Similarly, we hypothesize
the interception of P. tabuliformis in various study areas in
that if the initial density continues to increase, the S value of
China (15.7–36.9%; Fang et al. ). Valente et al. ()
P. tabuliformis artificial forests will show the same trend
obtained similar results there was a lower interception in
with increasing initial density as previously reported.
less dense forests. Results of the measurement of rainfall par-
In our study, fixed growing season parameters were used
titioning at different initial densities indicated that more
in RSGM to predict total interception, which was the same
rainfall penetrated the canopy and entered the next layer,
as those reported by Fan et al. () and Junqueira Junior
thereby affecting the hydrology of the forest ecosystem and
et al. (). Indeed, the rigor of the derivation process
contributing to plant growth, and appropriate thinning may
may be inadequate; however, it is possible to use fixed par-
improve the rainwater use efficiency of forests.
ameters to obtain satisfactory estimates of growing season
In the present study, the estimated S ranged from 0.894 to
interception (Wallace & McJannet ; Ghimire et al.
3.937 mm and was lower in low-density stands than in high-
; Fan et al. ). Because the study was conducted
density stands. According to Deguchi et al. (), the esti-
during the growing season, the canopy and climatic par-
mated S in different regions worldwide ranges from 0.25 to
ameters changed slightly. Moreover, the variation of the
1.69 mm, and this range is lower in medium-density (1,334,
canopy and climatic parameters will compensate for each
1
1,695, and 1,720 trees ha ) and high-density stands (2,746
other; therefore, the error of the final simulation of intercep-
trees ha 1). The reason for this might be that P. tabuliformis
tion may be considered minimal, as discussed by Wallace &
plantation in this study was an artificial coniferous pure
McJannet () and Fan et al. (). The type of variations
forest with a vegetation type that is different from that in the
for the final simulation results produced within a certain
above-mentioned literature (such as secondary broad-leaved
period when fixed parameters are used at different time
deciduous forest, mature mixed deciduous forest, hardwood
scales in RSGM to predict annual interception remains
1180
Y. Li et al.
|
Initial plant density effects modeling precision
Hydrology Research
|
51.5
|
2020
unclear and requires further investigation. Recently, a study
analytical Gash model to improve model accuracy. Ginebra-
used models with fixed parameters for the dormant and
Solanellas et al. () suggested that raindrop characteristics
growing seasons (Ma et al. ), respectively. The model
and leaf biomechanical properties may influence the modeling
showed good performance for both periods; however,
of the dynamic processes of forest canopy interception. Fur-
some parameters seemed to considerably vary between the
thermore, reconstructing the method of estimating S may be
two periods, implying that it is still questionable whether
the correct direction to reduce model simulation errors.
fixed canopy structures and climatic parameters can be used
Valente et al. () proposed a new procedure to estimate S
for simulating annual canopy interception or for a period span-
by inverse extrapolation. This method has been validated at
ning the dormant and growing seasons. Moreover, Zhang et al.
the single-tree scale but has not been widely used yet because
() used variable parameters to run RSGM to estimate and E repcanopy interception in a deciduous shrub; i.e., R
it may not be effective for regions without historical monitoring information.
resent the average values during all hours in each rainfall
Differences in initial plant density may affect forest
event. It remains to be demonstrated whether this approach
hydrological processes and the nutrient cycle, subsequently
can be used to obtain a good result in forests.
influencing vegetation growth. This will be accompanied
In the present study, although we used RSGM, it demon-
by differences in forest structure variables. Therefore, we
strated better performance in P. tabuliformis forests with
also considered the implication of the effects of changes in
higher-density and could not be applied well in lower-
forest structure variables on the modeling accuracy of the
density forests; moreover, the relative error decreased with
revised Gash model, although it is not detailed enough.
increasing initial density and the revised model tended to
Moreover, Fathizadeh et al. () showed that S and I
overestimate TF and underestimate SF and I, which is simi-
have strong relationships with wood area index, except
lar to the findings obtained by Limousin et al. () and Ma
LAI and c, and supported the idea that the studies for
et al. (). Although c was added to the revised version of
canopy structure variables can considerably enhance the
the model to further describe the characteristics of canopy
performance of the hydrological models. Indeed, we plan
structure and vegetation density along with S (Gash et al.
to continue the work of Fathizadeh et al. () to conduct
), it was not adequate. The indirect and complete esti-
future detailed studies on this subject; however, for now,
mation of the complex canopy structure is a significant
this study only discussed the influence of initial plant density
channel to improve the simulation accuracy of the model.
on the modeling accuracy of the revised Gash model owing
However, c can only interpret the horizontal structure of the
to the limitation of observation equipment and measures.
canopy and cannot adequately express canopy thickness,
Sensitivity analysis showed that S demonstrated absol-
vegetation density, and other structural and physiological par-
ute dominance in medium- and higher-density forests,
ameters, which may affect simulation accuracy, particularly in
which was similar to the results reported by Dykes (),
forests with lower density. van Dijk & Bruijnzeel (a) pro-
Liang (), and Su et al. (); however, S, the chief par-
posed correcting the revised model by supplementing the leaf area index (LAI) and expounding its relationship with S
ameter for simulated results, is not dependent on the and E gradually decreased as density. The influence of R
and c; they considered that LAI was linearly related to
the initial density increases. This may be due to the particu-
canopy capacity rather than to canopy cover in ‘constant phy-
lar characteristics of precipitation during the study period
siognomy and configuration’ and accordingly presented an
(relatively long rainfall duration, intermittent rainfall pat-
application of this version (van Dijk & Bruijnzeel b).
terns, and a relatively large number of moderate and
Nevertheless, this version of the model is based on the hypoth-
extreme rainfall events). Most rainfall occurred at night
esis that canopy capacity is not affected by initial density, and
during the study period, which may be the reason for the
because this hypothesis was evidently contrary to the results of
lesser contribution of the evaporation rate in the present
the present study, it may not be a highly effective solution to
study compared with that in other studies (Loustau et al.
improve the accuracy of the model from the root. The
; Deguchi et al. ; Fathizadeh et al. ). Linhoss
canopy structure must be effectively integrated into the
& Siegert () reported that storm duration, canopy
1181
Y. Li et al.
|
Initial plant density effects modeling precision
Hydrology Research
|
51.5
|
2020
storage, and solar radiation were extremely important for
appropriate to describe the evaporation rate for the open
canopy interception model sensitivity. Lower-density forests
canopy is an effective approach to improve the operational
exhibit larger forest gaps, larger crown width, and thinner
performance of RSGM in sparse forests. Moreover, it is
crown thickness than higher-density forests, which may
imperative to further demonstrate the influence of estimated
explain the reason for the climatic parameters being more
mean values of climatic parameters at different time scales
sensitive at lower densities. In addition, Linhoss & Siegert
on simulation accuracy. Most importantly, the impact of initial
() have reported that the estimation of the evaporation
planting density on the simulation results cannot be ignored
rate using the Penman–Monteith approach appears preferable
when using RSGM to quantify canopy interception. The appli-
for canopies that are not completely ventilated rather than for
cability of RSGM in sparse forest remains to be investigated
sparse forests and the methods for estimating the evaporation
until the model can be further improved or validated in the
rate in sparse forests should be investigated. Therefore, there
application area.
have been studies that have attempted to use the wet-bulb method instead of the Penman–Monteith approach to derive the evaporation rate at the single-tree or sparse forest level
ACKNOWLEDGEMENTS
when using the Gash-type models (e.g., Ma et al. ; Valente et al. ), and model performance using the former is
This work was funded by the Natural Science Foundation of
superior to the latter. This implies that improving the methods
Shandong Province of China project (No. ZR2016DB12)
used for estimating meteorological parameters in models could
and a project of the Key Laboratory of Water Saving
be a valid direction for sparse forests.
Irrigation Project of Ministry of Agriculture (Farmland Irrigation Research Institute of Chinese Academy of Agricultural Sciences, No. FIRI2018-04).
CONCLUSIONS The average cumulative measured I ranged from 16.46 to
DATA AVAILABILITY STATEMENT
30.44% of gross rainfall in P. tabuliformis plantations with different initial densities, and the interception was positively associated with initial plant density. The results obtained using the revised Gash model suggested that a relatively
Data cannot be made publicly available; readers should contact the corresponding author for details.
good consistency exists between the modeled and estimated interceptions in higher-density forests and that simulation
REFERENCES
accuracy enhances with an increase in the density. This may be attributable to the revised model being inadequate with respect to certain biological characteristics of P. tabuliformis plantations and the expression of the canopy structure of the sparse forest, and these weaknesses are magnified during the simulation of P. tabuliformis plantation. The predicted interception was higher sensitive to the canopy structure, irrespective of the initial density, and the close stands were less susceptible to the effects of the climate parameters on accuracy than the sparse stands. Therefore, it is crucial to effectively optimize and modify parameters describing the canopy structure in RSGM for improving simulative accuracy while ensuring the simplicity of the model and identifying and developing direct observation or estimation methods that are more
André, F., Jonard, M. & Ponette, Q. Precipitation water storage capacity in a temperate mixed oak-beech canopy. Hydrological Processes 22 (20), 4130–4141. doi:10.1002/hyp.7013. Bui, X. D., Shusuke, M. & Takashi, G. Effect of forest thinning on overland flow generation on hillslopes covered by Japanese cypress. Ecohydrology 4 (3), 367–378. doi:10.1002/ eco.135. Carlyle-Moses, D. E. Throughfall, stemflow, and canopy interception loss fluxes in a semi-arid Sierra Madre Oriental matorral community. Journal of Arid Environments 58 (2), 181–202. doi:10.1016/S0140-1963(03)00125-3. Chen, S., Chen, C., Zou, C. B., Stebler, E., Zhang, S., Hou, L. & Wang, D. Application of Gash analytical model and parameterized Fan model to estimate canopy interception of a Chinese red pine forest. Journal of Forest Research 18 (4), 335–344. doi:10.1007/s10310-012-0364-z.
1182
Y. Li et al.
|
Initial plant density effects modeling precision
China Agricultural Encyclopedia Forestry Volume Editorial Committee China Agricultural Encyclopedia Forestry, Volume (II). Agricultural Press, Beijing. Cui, Y. & Jia, L. A modified Gash model for estimating rainfall interception loss of forest using remote sensing observations at regional scale. Water 6 (4), 993–1012. doi:10. 3390/w6040993. Deguchi, A., Hattori, S. & Park, H. The influence of seasonal changes in canopy structure on interception loss: application of the revised Gash model. Journal of Hydrology 318 (1–4), 80–102. doi:10.1016/j.jhydrol.2005.06.005. Dykes, A. P. Rainfall interception from a lowland tropical rainforest in Brunei. Journal of Hydrology 200 (1–4), 260–279. doi:10.1016/S0022-1694(97)00023-1. Fan, J., Oestergaard, K. T., Guyot, A. & Lockington, D. A. Measuring and modeling rainfall interception losses by a native Banksia woodland and an exotic pine plantation in subtropical coastal Australia. Journal of Hydrology 515, 156–165. doi:10.1016/j.jhydrol.2014.04.066. Fang, S., Zhao, C., Jian, S. & Yu, K. Canopy interception of Pinus tabulaeformis plantation on Longzhong Loess Plateau, North-west China: characteristics and simulation. Chinese Journal of Applied Ecology 24 (06), 1509–1516. doi:10. 13287/j.1001-9332.2013.0326. Fathizadeh, O., Hosseini, S. M., Zimmermann, A., Keim, R. F. & Darvishi Boloorani, A. Estimating linkages between forest structural variables and rainfall interception parameters in semi-arid deciduous oak forest stands. Science of the Total Environment 601–602, 1824–1837. doi:10.1016/ j.scitotenv.2017.05.233. Fathizadeh, O., Hosseini, S. M., Keim, R. F. & Boloorani, A. D. A seasonal evaluation of the reformulated Gash interception model for semi-arid deciduous oak forest stands. Forest Ecology and Management 409, 601–613. doi:10.1016/ j.foreco.2017.11.058. Fernandes, R. P., Silva, R. W. D. C., Salemi, L. F., Andrade, T. M. B. D., Moraes, J. M. D., Dijk, A. I. J. M. & Martinelli, L. A. The influence of sugarcane crop development on rainfall interception losses. Journal of Hydrology 551, 532–539. doi:10.1016/j.jhydrol.2017.06.027. Gao, Y., Zhu, X., Yu, G., He, N., Wang, Q. & Tian, J. Water use efficiency threshold for terrestrial ecosystem carbon sequestration in China under afforestation. Agricultural and Forest Meteorology 195–196, 32–37. doi:10.1016/j.agrformet. 2014.04.010. Gash, J. H. C. An analytical model of rainfall interception by forests. Quarterly Journal of the Royal Meteorological Society 105, 43–55. doi:10.1002/qj.49710544304. Gash, J. H. C. & Morton, A. J. An application of the Rutter model to the estimation of the interception loss from Thetford Forest. Journal of Hydrology 38, 49–58. doi:10. 1016/0022-1694(78)90131-2. Gash, J. H. C., Lloyd, C. R. & Lachaud, G. Estimating sparse forest rainfall interception with an analytical model. Journal of Hydrology 170 (1), 79–86. doi:10.1016/0022-1694(95)02697-N.
Hydrology Research
|
51.5
|
2020
Ghimire, C. P., Bruijnzeel, L. A., Lubczynski, M. W. & Bonell, M. Rainfall interception by natural and planted forests in the Middle Mountains of Central Nepal. Journal of Hydrology 475, 270–280. doi:10.1016/j.jhydrol.2012.09.051. Ghimire, C. P., Bruijnzeel, L. A., Lubczynski, M. W., Ravelona, M., Zwartendijk, B. W. & van Meerveld, H. J. I. Measurement and modeling of rainfall interception by two differently aged secondary forests in upland eastern Madagascar. Journal of Hydrology 545, 212–225. doi:10. 1016/j.jhydrol.2016.10.032. Ginebra-Solanellas, R. M., Holder, C. D., Lauderbaugh, L. K. & Webb, R. The influence of changes in leaf inclination angle and leaf traits during the rainfall interception process. Agricultural and Forest Meteorology 285–286, 107924. doi:10.1016/j.agrformet.2020.107924. Hassan, S. M. T., Ghimire, C. P. & Lubczynski, M. W. Remote sensing upscaling of interception loss from isolated oaks: Sardon catchment case study, Spain. Journal of Hydrology 555, 489–505. doi:10.1016/j.jhydrol.2017.08.016. Ji, Y. & Cai, T. J. Canopy interception in original Korean pine forest: measurement and dividual simulation in Xiaoxing’an Mountains, northeastern China. Journal of Beijing Forestry University 37 (10), 41–49. doi:10.13332/j.1000-1522.20150084. Junqueira Junior, J. A., de Mello, C. R., de Mello, J. M., Scolforo, H. F., Beskow, S. & McCarter, J. Rainfall partitioning measurement and rainfall interception modelling in a tropical semi-deciduous Atlantic forest remnant. Agricultural and Forest Meteorology 275, 170–183. doi:10.1016/j. agrformet.2019.05.016. Klaassen, W., Bosveld, F. & Water, E. D. Water storage and evaporation as constituents of rainfall interception. Journal of Hydrology 212 (1–4), 36–50. doi:10.1016/S0022-1694(98) 00200-5. Leyton, L., Reynolds, E. R. C. & Thompson, F. B. Rainfall interception in forest and moorland. In: International Symposium on Forest Hydrology (W. E. Sopper & H. W. Lull, eds.). Pergamon, Oxford, pp. 163–178. Liang, W. Simulation of Gash model to rainfall interception of Pinus tabulaeformis. Forest Systems 23 (2), 300–303. doi:10.5424/fs/2014232-03410. Licata, J. A., Pypker, T. G., Weigandt, M., Unsworth, M. H., Gyenge, J. E., Fernández, M. E., Schlichter, T. M. & Bond, B. J. Decreased rainfall interception balances increased transpiration in exotic ponderosa pine plantations compared with native cypress stands in Patagonia, Argentina. Ecohydrology 4 (1), 83–93. doi:10.1002/eco.125. Limousin, J., Rambal, S., Ourcival, J. & Joffre, R. Modelling rainfall interception in a mediterranean Quercus ilex ecosystem: lesson from a throughfall exclusion experiment. Journal of Hydrology 357 (1–2), 57–66. doi:10.1016/j.jhydrol.2008. 05.001. Linhoss, A. C. & Siegert, C. M. A comparison of five forest interception models using global sensitivity and uncertainty analysis. Journal of Hydrology 538, 109–116. doi:10.1016/j. jhydrol.2016.04.011.
1183
Y. Li et al.
|
Initial plant density effects modeling precision
Liu, S. Estimation of rainfall storage capacity in the canopies of cypress wetlands and slash pine uplands in North-Central Florida. Journal of Hydrology 207 (1), 32–41. doi:10.1016/ S0022-1694(98)00115-2. Liu, Z., Wang, Y., Tian, A., Liu, Y., Webb, A. A., Wang, Y., Zuo, H., Yu, P., Xiong, W. & Xu, L. Characteristics of canopy interception and its simulation with a revised Gash model for a larch plantation in the Liupan Mountains, China. Journal of Forestry Research 29 (1), 187–198. doi:10.1007/s11676-017-0407-6. Llorens, P. & Domingo, F. Rainfall partitioning by vegetation under Mediterranean conditions. A review of studies in Europe. Journal of Hydrology 335 (1–2), 37–54. doi:10.1016/ j.jhydrol.2006.10.032. Llorens, P. & Gallart, F. A simplified method for forest water storage capacity measurement. Journal of Hydrology 240 (1–2), 131–144. doi:10.1016/s0022-1694(00)00339-5. Loustau, D., Berbigier, P., Granier, A. & Moussa, F. E. H. Interception loss, throughfall and stemflow in a maritime pine stand. I. Variability of throughfall and stemflow beneath the pine canopy. Journal of Hydrology 138 (3–4), 449–467. doi:10.1016/0022-1694(92)90130-n. Ma, J., Zha, T., Jia, X., Tian, Y., Bourque, C. P. A., Liu, P., Bai, Y., Wu, Y., Ren, C., Yu, H., Zhang, F., Zhou, C. & Chen, W. Energy and water vapor exchange over a young plantation in northern China. Agricultural and Forest Meteorology 263, 334–345. doi:10.1016/j.agrformet.2018.09.004. Ma, C., Li, X., Luo, Y., Shao, M. & Jia, X. The modelling of rainfall interception in growing and dormant seasons for a pine plantation and a black locust plantation in semi-arid Northwest China. Journal of Hydrology 577, 123849. doi:10. 1016/j.jhydrol.2019.06.021. Ma, C., Luo, Y. & Shao, M. Comparative modeling of the effect of thinning on canopy interception loss in a semiarid black locust (Robinia pseudoacacia) plantation in Northwest China. Journal of Hydrology 590, 125234. doi:10.1016/j. jhydrol.2020.125234. National Forestry and Grassland Administration China Forestry Statistical Yearbook. Available from: http://www. forestry.gov.cn/data.html/ (accessed 7 April 2019). Owens, M. K., Lyons, R. K. & Alejandro, C. L. Rainfall partitioning within semiarid juniper communities: effects of event size and canopy cover. Hydrological Processes 20 (15), 3179–3189. doi:10.1002/hyp.6326. Sadeghi, S. M. M., Attarod, P., Van Stan, J. T., Pypker, T. G. & Dunkerley, D. Efficiency of the reformulated Gash’s interception model in semiarid afforestations. Agricultural and Forest Meteorology 201, 76–85. doi:10.1016/j.agrformet. 2014.10.006. Sheng, H. C., Cai, T. J., Li, Y. & Liu, Y. J. Rainfall redistribution in Larix gmelinii forest on northern of Daxing’an mountains, north-east of China. Journal of Soil and Water Conservation 28 (06), 101–105. doi:10.13870/j. cnki.stbcxb.2014.06.019.
Hydrology Research
|
51.5
|
2020
Shinohara, Y., Levia, D. F., Komatsu, H., Nogata, M. & Otsuki, K. Comparative modeling of the effects of intensive thinning on canopy interception loss in a Japanese cedar (Cryptomeria japonica D. Don) forest of western Japan. Agricultural and Forest Meteorology 214–215, 148–156. doi:10.1016/j.agrformet.2015.08.257. Su, L., Zhao, C., Xu, W. & Xie, Z. Modelling interception loss using the revised Gash model: a case study in a mixed evergreen and deciduous broadleaved forest in China. Ecohydrology 9 (8), 1580–1589. doi:10.1002/eco.1749. Sun, J., Gao, P., Li, C., Wang, R., Niu, X. & Wang, B. Ecological stoichiometry characteristics of the leaf–litter–soil continuum of Quercus acutissima Carr. and Pinus densiflora Sieb. in Northern China. Environmental Earth Sciences 78 (1). doi:10.1007/s12665-018-8012-3. Teklehaimanot, Z., Jarvis, P. G. & Ledger, D. C. Rainfall interception and boundary layer conductance in relation to tree spacing. Journal of Hydrology 123 (3–4), 261–278. doi:10. 1016/0022-1694(91)90094-x. Valente, F., David, J. S. & Gash, J. H. C. Modelling interception loss for two sparse eucalypt and pine forests in central Portugal using reformulated Rutter and Gash analytical models. Journal of Hydrology 190 (1), 141–162. doi:10.1016/S0022-1694(96)03066-1. Valente, F., Gash, J. H., Nóbrega, C., David, J. S. & Pereira, F. L. Modelling rainfall interception by an olive-grove/ pasture system with a sparse tree canopy. Journal of Hydrology 581, 124417. doi:10.1016/j.jhydrol.2019.124417. van Dijk, A. I. J. M. & Bruijnzeel, L. A. a Modelling rainfall interception by vegetation of variable density using an adapted analytical model. Part 1. Model description. Journal of Hydrology 247 (3), 230–238. doi:10.1016/S0022-1694(01) 00392-4. van Dijk, A. I. J. M. & Bruijnzeel, L. A. b Modelling rainfall interception by vegetation of variable density using an adapted analytical model. Part 2. Model validation for a tropical upland mixed cropping system. Journal of Hydrology 247 (3), 239–262. doi:10.1016/S0022-1694(01)00393-6. Wallace, J. & McJannet, D. Modelling interception in coastal and montane rainforests in northern Queensland, Australia. Journal of Hydrology 348 (3–4), 480–495. doi:10.1016/j. jhydrol.2007.10.019. Wei, X., Liu, S., Zhou, G. & Wang, C. Hydrological processes in major types of Chinese forest. Hydrological Processes 19 (1), 63–75. doi:10.1002/hyp.5777. Zeng, N., Shuttleworth, J. W. & Gash, J. H. C. Influence of temporal variability of rainfall on interception loss. Part I. Point analysis. Journal of Hydrology 228 (3), 228–241. doi:10. 1016/S0022-1694(00)00140-2. Zhang, S., Li, X., Jiang, Z., Li, D. & Lin, H. Modelling of rainfall partitioning by a deciduous shrub using a variable parameters Gash model. Ecohydrology 11 (7), e2011. doi:10. 1002/eco.2011.
First received 15 December 2019; accepted in revised form 18 August 2020. Available online 14 September 2020
1184
© 2020 The Authors Hydrology Research
|
51.5
|
2020
Entropy weight method coupled with an improved DRASTIC model to evaluate the special vulnerability of groundwater in Songnen Plain, Northeastern China Bin Wang, Yanguo Teng, Huiqun Wang, Rui Zuo, Yuanzheng Zhai, Weifeng Yue and Jie Yang
ABSTRACT The Songnen Plain in Northeast China is the only remaining black soil agricultural area in the world and is an important food base for China. The groundwater resources in this area are abundant, but human activities have caused them polluted. This paper established a groundwater vulnerability assessment to characterize the influence of human activities which used an entropy weight method. The index was tested using the nitrate pollution distribution in the groundwater to verify the effectiveness of this method. The results showed that areas with high specific vulnerability were distributed in the northern and eastern parts of the Songnen Plain and were consistent with areas that showed serious nitrate pollution of the groundwater. The correlation coefficient between these areas was 0.2536, which greatly improved the vulnerability assessment without superimposing human activities in the model. The results clearly showed that human activities increased groundwater vulnerability on the Songnen Plain. The evaluation method provided a reference for similar evaluations and a basis for the protection and management of groundwater resources in this region. Key words
| DRASTIC model, entropy weight method, groundwater resource management, groundwater vulnerability, nitrate pollution, Songnen Plain
HIGHLIGHTS
• •
Establishing a groundwater vulnerability assessment method to characterize the influence of human activities using an entropy weight method. Entropy weight method coupled with an improved DRASTIC model to evaluate the special vulnerability of groundwater for water management.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.056
Bin Wang Yanguo Teng (corresponding author) Huiqun Wang Engineering Research Center of Groundwater Pollution Control and Remediation of Ministry of Education; College of Water Sciences, Beijing Normal University, Beijing 100875, China E-mail: teng1974@163.com Bin Wang Chinese Academy for Environmental Planning, Beijing 100012, China Rui Zuo Yuanzheng Zhai Weifeng Yue Jie Yang College of Water Sciences, Beijing Normal University, Beijing 100875, China
1185
B. Wang et al.
|
Evaluation method of the special vulnerability of groundwater
Hydrology Research
|
51.5
|
2020
GRAPHICAL ABSTRACT
INTRODUCTION Groundwater is an important source of water for human
et al. ; Jia et al. ; Hao et al. ). Because of its sub-
beings (Cassardo & Jones ; Salman et al. ). Histori-
terranean nature, groundwater pollution is typically a long-
cal experience shows that water shortages do not only
term and difficult-to-control issue (Ahmad & Al-Ghouti
seriously affect normal life but can also lead to disease out-
). Therefore, rational planning and utilization of ground-
breaks, wars and other unexpected hazards (Falkenmark
water resources to avoid groundwater pollution are far more
; Eliasson ; Mancosu et al. ; Egbueri ).
important than post-contamination treatment (Clemens
Given the unsustainable population and economic growth,
et al. ; Erostate et al. ; Thomann et al. ).
coupled with the failure to prevent groundwater pollution
Groundwater management encompasses a broad range
from industrial wastewater discharges and from heavy use
of activities including prevention of groundwater contami-
of pesticides and fertilizers (Bosch et al. ; Khosravi
nation. Vulnerability and pollution risk assessments to
et al. ; Li et al. ), more groundwater gets seriously
identify risk zones are the very first important steps to gen-
polluted, especially shallow groundwater resources (Gejl
erate useful information for devising strategies aimed at
et al. ; He et al. ). In many parts of China, shallow
groundwater protection to contamination. Delineating vul-
groundwater is an important source of drinking water (Lü
nerable zones helps water resource managers to divert
1186
B. Wang et al.
|
Evaluation method of the special vulnerability of groundwater
Hydrology Research
|
51.5
|
2020
groundwater development activities to other safer areas and,
urgent need to carry out an evaluation of the groundwater
hence, can minimize the cost of water treatment (Shrestha
vulnerability to nitrate on the Songnen Plain, in order to
et al. ). This article carried on the groundwater vulner-
guide protection and management of the regional ground-
ability research, which is of great importance to the
water resources. In the past, vulnerability assessments
allocation and protection of groundwater resources (Foster
were mostly focused on the water-resource scale or the
et al. ; Machiwal et al. ); it is largely defined by
basin scale. To evaluate a large-scale area, such as the Song-
the propensity and possibility of pollutants reaching a cer-
nen Plain, it requires not only the correct evaluation
tain location above the uppermost boundary of an aquifer.
parameters but also an improved model, in which the
Research on Chinese groundwater vulnerability began
weight of each evaluation index is determined for this
in the mid-1990s (Margat ; Gogu & Dassargues ).
area. In order to evaluate the vulnerability of groundwater
It has progressed rapidly since then (Ibe et al. ; Gogu
scientifically and reasonably, improve the reliability of the
et al. ; Polemio et al. ; Shirazi et al. ; Kumar
evaluation results, and provide guidance for the manage-
et al. ; Wachniew et al. ; Iván & Mádl-Szőnyi
ment of groundwater resources in the study area, using the
). The most common and classic evaluation model
hydrogeological characteristics of the Songnen Plain, this
used to evaluate groundwater vulnerability is the DRASTIC
article reconstructed evaluation model indicators and used
method, proposed by the US Environmental Protection
the entropy weight method and the cusp catastrophe
Agency in 1987 (Rahman ; Saidi et al. ; Barzegar
model to determine the weights of each index. The results
et al. ). The evaluation index of the DRASTIC model
of the modified DRASTIC model evaluation were compared
is based on the variables selected by US hydrogeological
with the nitrate distribution generated by human activities in
experts (Shirazi et al. ). Many efforts have been taken
the study area to verify the accuracy of the model.
to improve the DRASTIC model to make it applicable to different hydrogeological conditions (Nobre et al. ; Saidi et al. ; Wu et al. ; Zhang et al. ), although
STUDY AREA AND DATA
these improvements are often limited to adjusting the parameters of the model. In practice, it is dubious to extend
The Songnen Plain is one of three major plains in Northeast
the weightings of this method to other regions because of
China. It is located in the central part of the Songliao Basin,
the variable geographic setting of each evaluation area
bound by the Xing’an Mountains and Changbai Mountains
(Mendoza & Barmen ). Clearly, the geological and
and Songliao Watershed. It is traversed by the Songhua
hydrogeological conditions also vary a great deal (Denny
and the Nenjiang Rivers. The geographical coordinates of
et al. ). So, the evaluation indicators should be selected
the study area are 43 360 –49 260 N and 121 210 –128 180 E,
according to the specific conditions of the evaluation area,
defining a total area of 180,500 km2. Its location is shown
considering the natural attributes of the evaluation area,
in Figure 1.
and assigning a relative importance to each index. Thus,
The Songnen Plain is a semi-enclosed, asymmetrical
the correct weighting of variables directly determines the
basin, with a height difference of 400 m and a low gently
accuracy of the evaluation results (Singh et al. ).
inclined central region. Influenced by the difference in tec-
The Songnen Plain hosts both large-scale commercial
tonic activity among different regions, the geomorphology
grain and oil production. The region of the Second Songhua
of the region can be divided into three types: erosional land-
River has 210 million hectares of arable land, with an
forms, denuded landforms, and stacked landforms.
annual grain output of 16.7 billion kg, accounting for 61%
The Songnen Plain has a typical East Asian continental
of Jilin Province. Groundwater is an important source of
monsoon climate. It is cold and dry in winter, hot and rainy
water for crop growth in this region. However, recent
in summer. With an average annual temperature of 3.8 C,
survey data showed that the shallow groundwater of the
average annual precipitation of 484.57 mm, and an average
Songnen Plain had been variably contaminated by nitrate
annual evaporation of 1,498.1 mm. The main surface water
(Zhu et al. ; Bian et al. ). Therefore, there is an
systems are the Songhua River, the Nenjiang River, the
B. Wang et al.
1187
Figure 1
|
|
Evaluation method of the special vulnerability of groundwater
Hydrology Research
|
51.5
|
2020
Location of the study area showing the main rivers and cities on the Songnen Plain.
Second Songhua River, and their tributaries. According to the
wells typically have no anti-seepage measures, such as hard-
latest ‘Songliao Basin Water Resources Bulletin’, the volume
ening of the ground around the well. Typically, the wells are
3
of the water resources of the Songnen Plain is 9.6 billion m .
simple structures, with no buried gravel, and depths of less
A total of 1,409 groundwater sampling sites were inves-
than 20 m. The ground surface comprises mostly clay and
tigated on the Songnen Plain as part of this study. The
silt, while the aquifer is mainly fine sand; the groundwater
sampling was conducted from 2013 to 2015. The sampling
depth ranges from 1 to 10 m below the surface.
points are evenly distributed within the study area, as
Samples were collected in strict accordance with the
shown in Figure 2. Most of the sampling points were situated
technical specifications for water sample collection. The
in residential areas like villages. This is because most of the
site location was fixed by a GPS three-parameter calibration,
sampling sites were water wells used by villagers. These
water level measurement error was less than 1 cm, and the
B. Wang et al.
1188
Figure 2
|
|
Evaluation method of the special vulnerability of groundwater
Hydrology Research
|
51.5
|
2020
Distribution of the 1,409 groundwater sampling points within the study area.
well water was sampled using pre-treated sample bottle.
NO2 , As, Hg, Se, Cl , SO4 2 , F , NO3 , I , HCO3 ,
Reagents, such as volatile organics, were neutralized with
CO3 2 , CODCr, total hardness, TDS, Cd, Cu, Pb, Zn) of
two drops of concentrated hydrochloric acid. The collected
the groundwater was completed by the Experimental Test-
samples were stored at 4 C and transferred to laboratory
ing Center of the Shenyang Geological Survey. The
storage within 5 days.
detection and analysis methods were based on full-spectrum
Two sets of parallel samples were collected for quality
direct reading plasma spectroscopy (ICP-OES; Optima
control. In order to identify the groundwater quality and
5300DV Spectrometer, USA), ultraviolet spectrophotometry
to analyze the hydrochemical characteristics of the study
(COL; PerkinElmer Lambda 950, USA), and ion chromato-
þ
area, analytical testing of 31 inorganic components (Na ,
graphy (IC; Thermo Fisher ICS2000, USA), supplemented
Kþ, Ca2þ, TFe, Mn, Al, NH4 þ , PO4 3 , H2SiO3, Cr6þ, Br,
by
atomic
fluorescence
spectrometry
(AFS;
Titan
1189
B. Wang et al.
|
Evaluation method of the special vulnerability of groundwater
Hydrology Research
|
51.5
|
2020
AFS-9120, China), plasma mass spectrometry (ICP-MS;
weights solely according to the preference or judgments of
Thermo Fisher X Series, USA), volumetric (VOL), and gravi-
decision-makers, and the potential uncertainty in this
metric methods (GR). To ensure the accuracy and reliability
method is its main disadvantage. The objective weighting
of all analytical methods, each sample was analyzed 12
method calculates weights based on actual observations,
times, yielding a relative standard deviation (RSD%) of
without any consideration of the decision-makers’ prefer-
15%. The self-test rate and mutual inspection rate were
ences, and does not benefit from the expert knowledge or
both 100%, giving a total of 1,409 sets of valid data for
experience of the decision-makers. Thus, this paper uses
modeling.
weights determined using the entropy weight method. The entropy method is a measure of the uncertainty formulated in terms of probability theory (Clausius ). It is used
METHODS
to
describe
irreversible
phenomenon
involving
motion or a process and was introduced into information The main evaluation methods used in this paper were a modified DRASTIC model and an entropy weight method. The DRASTIC model is widely used to evaluate groundwater vulnerability to a wide range of contaminants. The acronym DRASTIC stands for the seven parameters used
theory by Shannon (). Nowadays, this method has been widely applied to various research fields to determine factor weights (Brunsell et al. ; Singh ; Ruddell et al. ; Islamoglu et al. ; Xu et al. ; Işık & Adalı ). Information entropy is the measurement of the
in the model, namely depth of the water table (D), net
disorder degree of a system (Harmancioglu ). It
recharge (R), aquifer media (A), soil media (S), topography
measures the amount of useful information based on the
(T), impact of the vadose zone media (I), and conductivity
data provided (Gao et al. ). When the difference in
of the aquifer hydraulic (C). These seven parameters are
value among evaluated objects of a given indicator is big,
further sub-divided into ranges or zones, representing var-
and the entropy is small, this illustrates that this indicator
ious hydrological settings and are assigned different ratings
provides useful information, and its weighting should be
from 1 (minimum impact) to 10 (maximum impact) based
high. In contrast, if the difference is small and the entropy
on a rating chart. In this method, a linear addition of vari-
is big, the relative weight will be low. Hence, entropy
ables is usually combined with the rating and weighting to
theory is an objective way of determining weights (Zou
determine the DRASTIC index value. The greater the rating and weight given to a parameter, the greater its relative importance in contributing to the aquifer vulnerability.
et al. ). According to Zou et al., the evaluation process is generally divided into three steps (Qiu ; Zou et al. ; Wu et al. ).
The DRASTIC formula is formulated as follows: Step1: normalization of the original matrix Ds ¼
7 X
(Wi × Ri ),
(1)
i¼1
Suppose m evaluation indicators and n evaluation objects form the original matrix of indicators, X ¼ (xij)m×n, as
where Ds represents the score of the DRASTIC index; Wi
follows:
represents the weight of factor i; and Ri represents the
2
score of factor i. Single-parameter sensitivity analyses explore the contribution of individual variables and use the same weights as the DRASTIC method. Therefore, it is particularly important to use an appropriate method to calculate the weight
X ¼ [xij ]m×n
x11 6 x21 6 ¼6 . 4 .. xm1
x12 x22 .. . xm2
... ... .. . ...
3 x1n x2n 7 7 .. 7 . 5 xmn
(2)
(i ¼ 1, 2 . . . , m; j ¼ 1, 2, . . . , n),
of each parameter. At present, the methods for weight calculation are divided into two types: subjective and objective
where xij presents the performance value of the ith alterna-
methods. The subjective weighting method determines
tive on the jth criterion.
B. Wang et al.
1190
|
Evaluation method of the special vulnerability of groundwater
|
51.5
|
2020
RESULTS AND DISCUSSION
Normalization of this matrix gives Equation (3): R ¼ (rij )m×n ,
Hydrology Research
(3)
Based on the survey results and the hydrogeological charac-
where rij is the data for the jth evaluation object under the
teristics of the study area, this paper obtained six
ith indicator and rij ∈ [0,1].
representative indicators of the DRASTIC model. The data
Among these indicators, when bigger is better, this
level at the sampling points. R was derived from meteorolo-
yields rij ¼
for D were from actual measurements of the groundwater gical data released by the government department. S was
xij min (xij ) max (xij ) min (xij )
(4)
(i ¼ 1, 2 . . . , m and j ¼ 1, 2, . . . , n),
derived from actual survey data. T was determined from the official digital elevation model, having 30 m × 30 m accuracy. I and C were derived from geological survey results. This article did not use the index A because attributes of A
but when smaller is better, this yields
were closely related to I and C; furthermore, if the index A
max (xij ) xij rij ¼ max (xij ) min (xij )
(5)
(i ¼ 1, 2 . . . , m and j ¼ 1, 2, . . . , n):
is used for evaluation, it will reduce the impact of the other indicators, making it difficult to accurately assess the vulnerability of the regional groundwater. According to the characteristics of each index, each attribute of each index has a score. Generally, the attribute of the index is
Step 2: calculation of entropy
more likely to cause groundwater pollution, and it will get For m indicators and n evaluation objects in the evaluation
a higher score. Limited by the availability of survey data,
problem, the entropy of the ith indicator is defined as
some indicators in this assessment could only use interp-
follows:
olation method with available data to obtain the situation
m P
without survey data area, which might bring certain uncerfij ln fij
ei ¼ i¼1
tainty to the assessment results. The spatial distribution of (i ¼ 1, 2 . . . , m and j ¼ 1, 2, . . . , n),
ln m rij and 0 < ej < 1: where fij ¼ m P rij
each indicator was plotted using the spatial analysis module in Arc GIS 10.3 software (Esri, Redlands, CA, (6)
USA). These results are shown in Figure 3. According to the characteristics of each index and the
i¼1
information it contained, the weights of each index were calIf fij is all the same value, then the entropy value of each
culated using the entropy weight method. The weight
criterion is the maximum value (ej ¼ 1). If fij is all 0, then fij
calculation results are shown in Table 1. According to the
ln fij is also 0 in value.
scores and weights of each indicator, this new ‘DRSTIC’ model was used to calculate the inherent vulnerability of the regional groundwater (Figure 4).
Step 3: calculation of the weight of entropy
To assess whether human activities have an impact on The weight of the entropy of the ith indicator can be formu-
the regional groundwater quality, this paper conducted a
lated as follows:
hydrogeological analysis of the shallow groundwater aquifer, the distributions of characteristic pollutants related to
wi ¼
1 ei m P n ei
where
n X
wi ¼ 1,
j¼1
i¼1
human activities, and hydrochemical characteristics. (7)
The Songnen Plain evolved from Mesozoic and Cenozoic subsidence basins, in which continental clastic sediments were deposited to a thickness of more than 8,000 m. Thus,
where the weight has a value 0 wi l.
the Songnen Plain is a large-scale groundwater aquifer
B. Wang et al.
1191
Figure 3
|
|
Evaluation method of the special vulnerability of groundwater
Hydrology Research
|
51.5
|
2020
Indicator scores of the DRSTIC model. (a)–(f) Results for depth of the water table (D), net recharge (R), soil media (S), topography (T), impact of the vadose zone media (I), and conductivity of the aquifer hydraulic (C), respectively.
B. Wang et al.
1192
Table 1
|
|
Evaluation method of the special vulnerability of groundwater
Weights of the DRSTIC model indicators (D, depth of the water table; R, net recharge; S, soil media; T, topography; I, impact of the vadose zone media; and C, conductivity of the aquifer hydraulic)
Hydrology Research
|
51.5
|
2020
mountainous areas to the west, north, and east. It flows through the western piedmont plain, northern high plain, and the eastern undulating high plain, collecting into three
Indicators
D
R
S
T
I
C
streams flowing to the south. The shallow groundwater
Weight
5.0
2.3
4.9
2.4
1.0
0.3
system interacts with the surface water system, comprising the Nenjiang, the Second Songhua River, and the Songhua
system comprising Quaternary pore water, Neogene fissure
Rivers. Rain moves downwards by means of subsurface
water, Paleogene fissure water, and Cretaceous pore and frac-
flow and drains to the rivers. During runoff, evaporation
ture water. The regional shallow groundwater resource
and human usage consume a large amount of water, and
receives input from precipitation and recharge from adjacent
only a small amount flows out of the study area.
Figure 4
|
Groundwater vulnerability within the study area based on the DRSTIC model.
B. Wang et al.
1193
|
Evaluation method of the special vulnerability of groundwater
Hydrology Research
|
51.5
|
2020
Runoff, circulation, and the regional aquifer character-
shallow groundwater of the study area is shown in Figure 5.
istics are all conducive to self-cleaning of the groundwater
Clearly, high nitrate concentration areas are mainly in the
system, maintaining good water quality under natural con-
northeast and southeast of the plain. Among them, values
ditions. However, according to our test results, the shallow
in the northeast are highest, reflecting pollution from agri-
groundwater in the study area is locally contaminated by
cultural areas and some industrial agglomeration areas;
nitrate. The proportion of groundwater samples with
high values in the southeast reflect industrial and densely
nitrate concentration more than 0.5 mg/L (China ground-
populated urban areas.
water quality standard class III) was 77.4%. The average
Based on sample test results, the groundwater samples
concentration of nitrate was 67.2 mg/L, and the highest con-
were classified using the Shukalev classification, yielding a
centration is 1,000.0 mg/L. The distribution of nitrate in the
map reflecting groundwater chemistry. The Shukalev
Figure 5
|
Distribution of nitrate within the shallow groundwater of the study area.
B. Wang et al.
1194
|
Evaluation method of the special vulnerability of groundwater
Hydrology Research
|
51.5
|
2020
classification is based on the concentrations of six major
in unconsolidated Quaternary deposits, the chemical
ions (Naþ, Ca2þ, Mg2þ, HCO3 , SO4 2 , Cl , Kþ incorpor-
composition
þ
ated into Na ) and groundwater salinity.
of
the
groundwater
is
complex
and
varied, suggesting that it is greatly affected by environ-
Using groundwater index values, a Piper three-line
mental conditions and human activities. From the
diagram for shallow and deep groundwater hydrochemi-
piedmont to the plains, especially in populated areas,
cal types in the evaluation area was constructed from
the chemical signature of the groundwater varies.
the groundwater survey data (Figure 6). The main water
This reflects flow velocities of downstream runoff,
chemistry type of the shallow groundwater in the
residence times of water–rock interactions, and inputs
study area was the HCO3–Ca type, accounting for 24.83% of
from
human
activities,
especially
artificial
mixing
all samples. This was followed by HCO3–Ca·Mg, HCO3–
such as groundwater exploitation, etc. (Zhang et al.
Na·Ca, HCO3–Na·Mg·Ca, HCO3·Cl–Ca, and HCO3·Cl–Na·Ca
; Li et al. , ). Human activities have led to
types. Although the shallow groundwater mainly resides
changes in groundwater chemistry in this region, and
Figure 6
|
Piper three-line diagram of shallow water chemical types within the study area (n ¼ 1,409).
B. Wang et al.
1195
|
Evaluation method of the special vulnerability of groundwater
clearly do have an impact on the regional groundwater
Hydrology Research
|
51.5
|
2020
Typically, areas with high special vulnerability of groundwater are mainly distributed in the north and east
quality. Because the regional groundwater has been affected by
of the Songnen Plain. These high vulnerability areas are con-
human activities, this paper selected additional indicators
sistent with those having serious nitrate pollution (Figure 5).
that reflect groundwater quality in the region, based on the
Human activities can cause the increase of nitrate concen-
distribution of characteristic pollutants and the driving
tration in groundwater (Jakobczyk-Karpierz et al. ;
forces of water pollution, mainly groundwater extraction
Teng et al. ). Under the same conditions, the regional
(E) and land-use (L). Water extraction data are available
nitrate concentration with high vulnerability of groundwater
from the Water Resources Bulletin of each city, while
should be higher.
land-use information comes from China’s 2015 satellite
To verify the accuracy of the evaluation results, they
remote sensing image data. In order to couple the above
were fitted with the distribution of nitrate in the shallow
indexes into the model and draw a quantitative evaluation
groundwater of the study area. A correlation analysis
conclusion, the E and l indexes were assigned by the same
showed that the correlation coefficient between these
principle as other indexes in the DRSTIC model, and the
two distributions was 0.2536. Comparing the vulner-
principle is that the attribute of the index is more likely to
ability of groundwater and the distribution of nitrate in
cause groundwater pollution, then it will get a higher
the study area, the correlation coefficient increased
score. The specific scores of E and I in this paper came
from 0.0752 to 0.2536. This shows that human activities
from the Delphi method. The spatial distribution of each
do affect the groundwater in the study area, making it
indicator was plotted using Arc GIS 10.3 (Figure 7).
more vulnerable to pollution. Comparing Figures 4 and
The entropy weight method was used to calculate the
5, it is clear that the regions with serious nitrate pollution
weights of variables E and L. The weight of E was 3.3,
and those with high intrinsic groundwater vulnerability
while the weight of L was 5.0. This gave rise to a new
do not coincide, and that the correlation coefficient
model, DRSTIC-EL. This article superimposed each indi-
between them also does not indicate an obvious relation-
cator to obtain groundwater vulnerability, as shown in
ship. This indicates that the shallow groundwater of the
Figure 8.
Songnen Plain has good self-cleaning potential. Although
Figure 7
|
Indicator scores for (a) groundwater exploitation (E) and (b) land-use (L) within the study area.
B. Wang et al.
1196
Figure 8
|
|
Evaluation method of the special vulnerability of groundwater
Hydrology Research
|
51.5
|
2020
Groundwater vulnerability within the study area based on the DRSTIC-EL model.
the correlation coefficient is not statistically significant,
conclusions of this study (Ahmed ; Albuquerque
given the large evaluation area and dataset, correlation
et al. ; Douglas et al. ; Modibo ). Despite
analysis at each grid point is only based on the software
all this, the quantitative impact of the above activities
platform, and the relationship between the vulnerability
on groundwater vulnerability is still far from clear, and
of groundwater and nitrate pollution cannot be compared
whether there are other influencing factors is not clear.
reliably at the regional scale. However, many studies
The research results of this paper are more inclined to
have shown that human activities, such as groundwater
determine
extraction, agricultural, and mining activities, lead to
from a macro perspective to provide guidance for water
increases in groundwater vulnerability, supporting the
resource management.
groundwater
vulnerability
characteristics
1197
B. Wang et al.
|
Evaluation method of the special vulnerability of groundwater
Hydrology Research
|
51.5
|
2020
CONCLUSIONS
DATA AVAILABILITY STATEMENT
A vulnerability assessment of the shallow groundwater
All relevant data are included in the paper or its Supplemen-
aquifer of the Songnen Plain was conducted using an
tary Information.
improved DRSTIC-LE model coupled with an entropy weight method to calculate the weighting of its indices. The evaluation results showed that the areas with high specific vulnerability of groundwater on the Songnen Plain are distributed to the north and east, consistent with areas affected by serious nitrate pollution of the groundwater. The correlation coefficient between them was 0.2536, an improvement over the result obtained from the direct calculation of the correlation coefficient between vulnerability and nitrate pollution. Chemical analysis of groundwater types also suggested that human activities have increased groundwater vulnerability on the Songnen Plain and made groundwater more vulnerable to pollution. This paper outlines a new way to evaluate the vulnerability of groundwater and validates this method using a case study. Our validation shows the applicability of the evaluation method to groundwater protection and management. Although the refined groundwater vulnerability assessment of each zone is difficult to achieve, in the area of regional groundwater resource management,
groundwater
vulnerability
assessment
undoubtedly provides a reliable reference in the macro aspect, which can relatively accurately measure the degree of regional groundwater vulnerability to pollution. Our assessment of groundwater vulnerability at such a large regional scale was limited by data availability. In the future or similar studies, more data should be obtained to improve the accuracy and reliability of the assessment.
ACKNOWLEDGEMENTS This work was supported by the National Natural Science Foundation of China (Nos. 41877355 and U19A20107) and Beijing Advanced Innovation Program for Land Surface Science. We thank Dr Trudi Semeniuk from Liwen Bianji, Edanz Editing China (www.liwenbianji.cn/ac), for editing the English text of a draft of this manuscript.
REFERENCES Ahmed, A. A. Using generic and pesticide DRASTIC GISbased models for vulnerability assessment of the Quaternary aquifer at Sohag, Egypt. Hydrogeology Journal 17 (5), 1203–1217. https://doi.org/10.1007/s10040-009-0433-3. Ahmad, A. Y. & Al-Ghouti, M. A. Approaches to achieve sustainable use and management of groundwater resources in Qatar: a review. Groundwater for Sustainable Development 11, 100367. https://doi.org/10.1016/j.gsd.2020. 100367. Albuquerque, M. T. D., Sanz, G., Oliveira, S. F., Martinez-Alegria, R. & Antunes, I. M. H. R. Spatio-temporal groundwater vulnerability assessment – a coupled remote sensing and GIS approach for historical land cover reconstruction. Water Resources Management 27 (13), 4509–4526. https://doi.org/ 10.1007/s11269-013-0422-0. Barzegar, R., Moghaddam, A. A., Norallahi, S., Inam, A., Adamowski, J., Alizadeh, M. R. & Nassar, J. B. Modification of the DRASTIC framework for mapping groundwater vulnerability zones. Groundwater 58 (3), 441–452. https://doi.org/10.1111/gwat.12919. Bian, J., Zhang, Z. & Han, Y. Spatial variability and health risk assessment of nitrogen pollution in groundwater in Songnen Plain. Journal of Chongqing University 38 (4), 104–111. https://doi.org/10.11835/j.issn.1000-582X.2015.04. 015. Bosch, A. P., Navarrete, F., Molina, L. & Martinezvidal, J. L. Quantity and quality of groundwater in the Campo de Dalias (Almeria, SE Spain). Water Science and Technology 24 (11), 87–96. Brunsell, N. A., Schymanski, S. J. & Kleidon, A. Quantifying the thermodynamic entropy budget of the land surface: is this useful? Earth System Dynamics 2 (1), 87–103. https://doi. org/10.5194/esd-2-87-2011. Cassardo, C. & Jones, J. A. A. Managing water in a changing world. Water 3 (2), 618–628. https://doi.org/10.3390/ w3020618. Clausius, R. The Mechanical Theory of Heat: With Its Applications to the Steam-Engine and to the Physical Properties of Bodies. John van Voorst, London. Clemens, M., Khurelbaatar, G., Merz, R., Siebert, C., van Afferden, M. & Roediger, T. Groundwater protection under water scarcity; from regional risk assessment to local wastewater treatment solutions in Jordan. Science of the Total Environment 706, 136066. https://doi.org/10.1016/ j.scitotenv.2019.136066.
1198
B. Wang et al.
|
Evaluation method of the special vulnerability of groundwater
Denny, S. C., Allen, D. M. & Journeay, J. M. DRASTIC-Fm: a modified vulnerability mapping method for structurally controlled aquifers in the southern Gulf Islands, British Columbia, Canada. Hydrogeology Journal 15 (3), 483–493. https://doi.org/10.1007/s10040-006-0102-8. Douglas, S. H., Dixon, B. & Griffin, D. Assessing the abilities of intrinsic and specific vulnerability models to indicate groundwater vulnerability to groups of similar pesticides: a comparative study. Physical Geography 39 (6), 487–505. https://doi.org/10.1080/02723646.2017.1406300. Egbueri, J. C. Groundwater quality assessment using pollution index of groundwater (PIG), ecological risk index (ERI) and hierarchical cluster analysis (HCA): a case study. Groundwater for Sustainable Development 10, 100292. https://doi.org/10.1016/j.gsd.2019.100292. Eliasson, J. The rising pressure of global water shortages. Nature 517 (7532), 6. https://doi.org/10.1038/517006a. Erostate, M., Huneau, F., Garel, E., Ghiotti, S., Vystavna, Y., Garrido, M. & Pasqualini, V. Groundwater dependent ecosystems in coastal Mediterranean regions: characterization, challenges and management for their protection. Water Research 172, 115461. https://doi.org/10. 1016/j.watres.2019.115461. Falkenmark, M. Growing water scarcity in agriculture: future challenge to global water security. Philosophical Transactions of the Royal Society A: Mathematical Physical and Engineering Sciences 371 (2002). https://doi.org/10. 1098/rsta.2012.0410. Foster, S., Hirata, R. & Andreo, B. The aquifer pollution vulnerability concept: aid or impediment in promoting groundwater protection? Hydrogeology Journal 21 (7), 1389–1392. https://doi.org/10.1007/s10040-013-1019-7. Gao, Y., Qian, H., Ren, W., Wang, H., Liu, F. & Yang, F. Hydrogeochemical characterization and quality assessment of groundwater based on integrated-weight water quality index in a concentrated urban area. Journal of Cleaner Production 260, 121006. https://doi.org/10.1016/j.jclepro. 2020.121006. Gejl, R. N., Rygaard, M., Henriksen, H. J., Rasmussen, J. & Bjerg, P. L. Understanding the impacts of groundwater abstraction through long-term trends in water quality. Water Research 156, 241–251. https://doi.org/10.1016/j.watres. 2019.02.026. Gogu, R. C. & Dassargues, A. Current trends and future challenges in groundwater vulnerability assessment using overlay and index methods. Environmental Geology 39 (6), 549–559. https://doi.org/10.1007/s002540050466. Gogu, R. C., Hallet, V. & Dassargues, A. Comparison of aquifer vulnerability assessment techniques. Application to the Neblon river basin (Belgium). Environmental Geology 44 (8), 881–892. https://doi.org/10.1007/s00254-003-0842-x. Hao, C., Zhang, W. & Gui, H. Hydrogeochemistry characteristic contrasts between low- and high-antimony in shallow drinkable groundwater at the largest antimony mine
Hydrology Research
|
51.5
|
2020
in Hunan province, China. Applied Geochemistry 117, 104584. https://doi.org/10.1016/j.apgeochem.2020.104584. Harmancioglu, N. Measuring the Information Content of Hydrological Processes by the Entropy Concept (in Turkish). PhD Thesis, Ege University, Izmir, Turkey. He, B., He, J., Wang, L., Zhang, X. & Bi, E. Effect of hydrogeological conditions and surface loads on shallow groundwater nitrate pollution in the Shaying River Basin: based on least squares surface fitting model. Water Research 163, 114880. https://doi.org/10.1016/j.watres.2019.114880. Ibe, K. M., Nwankwor, G. I. & Onyekuru, S. O. Assessment of ground water vulnerability and its application to the development of protection strategy for the water supply aquifer in Owerri, Southeastern Nigeria. Environmental Monitoring and Assessment 67 (3), 323–360. https://doi.org/ 10.1023/a:1006358030562. Işık, A. T. & Adalı, E. A. The decision-making approach based on the combination of entropy and ROV methods for the apple selection problem. European Journal of Interdisciplinary Studies 3 (3). http://doi.org/10.26417/ejis.v3i3.p81-6. Islamoglu, M., Apan, M. & Oztel, A. An evaluation of the financial performance of REITs in Borsa Istanbul: a case study using the entropy-based TOPSIS method. International Journal of Financial Research 6 (2), 124–138. http://doi.org/ 10.5430/ijfr.v6n2p124. Iván, V. & Mádl-Szőnyi, J. State of the art of karst vulnerability assessment: overview, evaluation and outlook. Environmental Earth Sciences 76 (3), 1–25. https://doi.org/ 10.1007/s12665-017-6422-2. Jakobczyk-Karpierz, S., Sitek, S., Jakobsen, R. & Kowalczyk, A. Geochemical and isotopic study to determine sources and processes affecting nitrate and sulphate in groundwater influenced by intensive human activity – carbonate aquifer Gliwice (southern Poland). Applied Geochemistry 76, 168–181. https://doi.org/10.1016/j.apgeochem.2016.12.005. Jia, Z., Bian, J. & Wang, Y. Impacts of urban land use on the spatial distribution of groundwater pollution, Harbin City, Northeast China. Journal of Contaminant Hydrology 215, 29–38. https://doi.org/10.1016/j.jconhyd.2018.06.005. Khosravi, K., Sartaj, M., Tsai, F. T. C., Singh, V. P., Kazakis, N., Melesse, A. M., Prakash, I., Bui, D. T. & Binh Thai, P. A comparison study of DRASTIC methods with various objective methods for groundwater vulnerability assessment. Science of the Total Environment 642, 1032–1049. https:// doi.org/10.1016/j.scitotenv.2018.06.130. Kumar, P., Bansod, B. K. S., Debnath, S. K., Thakur, P. K. & Ghanshyam, C. Index-based groundwater vulnerability mapping models using hydrogeological settings: a critical evaluation. Environmental Impact Assessment Review 51, 38–49. https://doi.org/10.1016/j.eiar.2015.02.001. Li, X., Lin, X., Du, J. & Cui, J. Analysis of hydrochemical evolution of phreatic water in Qiqihar City. Journal of Hydraulic Engineering 45 (7), 815–827. https://doi.org/10. 13243/j.cnki.slxb.2014.07.008.
1199
B. Wang et al.
|
Evaluation method of the special vulnerability of groundwater
Li, X., Wang, R. & Li, J. Study on hydrochemical characteristics and formation mechanism of shallow groundwater in eastern Songnen Plain. Journal of Groundwater Science and Engineering 6 (3), 161–170. https://doi.org/10.19637/j.cnki.2305-7068.2018.03.001. Li, X., Tang, C., Cao, Y. & Li, D. A multiple isotope (H, O, N, C and S) approach to elucidate the hydrochemical evolution of shallow groundwater in a rapidly urbanized area of the Pearl River Delta, China. Science of the Total Environment 724, 137930. https://doi.org/10.1016/j.scitotenv.2020. 137930. Lü, J., Qiu, H., Lin, H., Yuan, Y., Chen, Z. & Zhao, R. Source apportionment of fluorine pollution in regional shallow groundwater at You’xi County southeast China. Chemosphere 158, 50–55. https://doi.org/10.1016/j.chemosphere.2016.05. 057. Machiwal, D., Jha, M. K., Singh, V. P. & Mohan, C. Assessment and mapping of groundwater vulnerability to pollution: current status and challenges. Earth-Science Reviews 185, 901–927. https://doi.org/10.1016/j.earscirev. 2018.08.009. Mancosu, N., Snyder, R. L., Kyriakakis, G. & Spano, D. Water scarcity and future challenges for food production. Water 7 (3), 975–992. https://doi.org/10.3390/w7030975. Margat, J. Groundwater Vulnerability to Contamination. Bureau de Recherches Géologiques et Minières (BRGM). Mendoza, J. A. & Barmen, G. Assessment of groundwater vulnerability in the Rio Artiguas basin, Nicaragua. Environmental Geology 50 (4), 569–580. https://doi.org/10. 1007/s00254-006-0233-1. Modibo, S. A. Groundwater Quality Origin Mechanism, and Exploitation in the Niger Basin Underclimate Change and Pollution Impact. PhD Thesis, Jilin University, China. Nobre, R. C. M., Rotunno Filho, O. C., Mansur, W. J., Nobre, M. M. M. & Cosenza, C. A. N. Groundwater vulnerability and risk mapping using GIS, modeling and a fuzzy logic tool. Journal of Contaminant Hydrology 94 (3–4), 277–292. https://doi.org/10.1016/j.jconhyd.2007.07.008. Polemio, M., Casarano, D. & Limoni, P. P. Karstic aquifer vulnerability assessment methods and results at a test site (Apulia, southern Italy). Natural Hazards and Earth System Sciences 9 (4), 1461–1470. https://doi.org/10.5194/nhess-91461-2009. Qiu, W. Management Decision and Applied Entropy. China Machine Press, Beijing, China. Rahman, A. A GIS based DRASTIC model for assessing groundwater vulnerability in shallow aquifer in Aligarh, India. Applied Geography 28 (1), 32–53. https://doi.org/10. 1016/j.apgeog.2007.07.008. Ruddell, B. L., Brunsell, N. A. & Stoy, P. Applying information theory in the geosciences to quantify process uncertainty, feedback, scale. Eos, Transactions American Geophysical Union 94 (5), 56. http://doi.org/10.1002/ 2013EO050007.
Hydrology Research
|
51.5
|
2020
Saidi, S., Bouri, S. & Ben Dhia, H. Groundwater vulnerability and risk mapping of the Hajeb-jelma aquifer (Central Tunisia) using a GIS-based DRASTIC model. Environmental Earth Sciences 59 (7), 1579–1588. https://doi.org/10.1007/ s12665-009-0143-0. Saidi, S., Bouri, S., Ben Dhia, H. & Anselme, B. Assessment of groundwater risk using intrinsic vulnerability and hazard mapping: application to Souassi aquifer, Tunisian Sahel. Agricultural Water Management 98 (10), 1671–1682. https:// doi.org/10.1016/j.agwat.2011.06.005. Salman, S. A., Arauzo, M. & Elnazer, A. A. Groundwater quality and vulnerability assessment in west Luxor Governorate, Egypt. Groundwater for Sustainable Development 8, 271–280. https://doi.org/10.1016/j.gsd.2018. 11.009. Shannon, C. E. A mathematical theory of communications. Bell System Technical Journal 27 (3), 379–423. https://doi. org/10.1002/j.1538-7305.1948.tb01338.x. Shirazi, S. M., Imran, H. M. & Akib, S. GIS-based DRASTIC method for groundwater vulnerability assessment: a review. Journal of Risk Research 15 (8), 991–1011. https://doi.org/ 10.1080/13669877.2012.686053. Shrestha, S., Kafle, R. & Pandey, V. P. Evaluation of indexoverlay methods for groundwater vulnerability and risk assessment in Kathmandu Valley, Nepal. Science of the Total Environment 575, 779–790. https://doi.org/10.1016/j. scitotenv.2016.09.141. Singh, V. P. Hydrologic synthesis using entropy theory: review. Journal of Hydrologic Engineering 16 (5), 421–433. https://doi.org/10.1061/(asce)he.1943-5584.0000332. Singh, A., Srivastav, S. K., Kumar, S. & Chakrapani, G. J. A modified-DRASTIC model (DRASTICA) for assessment of groundwater vulnerability to pollution in an urbanized environment in Lucknow, India. Environmental Earth Sciences 74 (7), 5475–5490. https://doi.org/10.1007/s12665015-4558-5. Teng, Y., Zuo, R., Xiong, Y., Wu, J., Zhai, Y. & Su, J. Risk assessment framework for nitrate contamination in groundwater for regional management. Science of the Total Environment 697, 134102. https://doi.org/10.1016/j. scitotenv.2019.134102. Thomann, J. A., Werner, A. D., Irvine, D. J. & Currell, M. J. Adaptive management in groundwater planning and development: a review of theory and applications. Journal of Hydrology 586, 124871. https://doi.org/10.1016/j.jhydrol. 2020.124871. Wachniew, P., Zurek, A. J., Stumpp, C., Gemitzi, A., Gargini, A., Filippini, M., Rozanski, K., Meeks, J., Kvaerner, J. & Witczak, S. Toward operational methods for the assessment of intrinsic groundwater vulnerability: a review. Critical Reviews in Environmental Science and Technology 46 (9), 827–884. https://doi.org/10.1080/10643389.2016.1160816. Wu, J., Sun, J., Liang, L. & Zha, Y. Determination of weights for ultimate cross efficiency using Shannon entropy. Expert
1200
B. Wang et al.
|
Evaluation method of the special vulnerability of groundwater
Systems with Applications 38 (5), 5162–5165. https://doi. org/10.1016/j.eswa.2010.10.046. Wu, W., Yin, S., Liu, H. & Chen, H. Groundwater vulnerability assessment and feasibility mapping under reclaimed water irrigation by a modified DRASTIC model. Water Resources Management 28 (5), 1219–1234. https://doi. org/10.1007/s11269-014-0536-z. Xu, J., Feng, P. & Yang, P. Research of development strategy on China’s rural drinking water supply based on SWOT-TOPSIS method combined with AHPEntropy: a case in Hebei Province. Environmental Earth Sciences 75 (1), 58. https://doi.org/10.1007/s12665-0154885-6. Zhang, G., Deng, W., He, Y. & Ramsis, S. Hydrochemical characteristics and evolution laws of groundwater in Songnen
Hydrology Research
|
51.5
|
2020
Plain, Northeast China. Advances in Water Science 17 (1), 20–28. https://doi.org/10.14042/j.cnki.32.1309.2006.01.004. Zhang, B., Li, G., Cheng, P., Yeh, T.-C. J. & Hong, M. Landfill risk assessment on groundwater based on vulnerability and pollution index. Water Resources Management 30 (4), 1465–1480. https://doi.org/10.1007/s11269-016-1233-x. Zhu, W., Zhao, Y., Tang, W. & Du, J. Study on shallow groundwater water quality status and development trends in Songnen Plain. Journal of Anhui Agriculture Science 41 (4), 1664–1669 þ 1744. https://doi.org/10.13989/j.cnki.05176611.2013.04.024. Zou, Z., Yun, Y. & Sun, J. Entropy method for determination of weight of evaluating indicators in fuzzy synthetic evaluation for water quality assessment. Journal of Environmental Sciences 18 (5), 1020–1023. https://doi.org/10.1016/s1001-0742(06)60032-6.
First received 29 April 2020; accepted in revised form 15 July 2020. Available online 1 October 2020
1201
© 2020 The Authors Hydrology Research
|
51.5
|
2020
A field investigation on rill development and flow hydrodynamics under different upslope inflow and slope gradient conditions Pei Tian, Chengzhong Pan, Xinyi Xu, Tieniu Wu, Tiantian Yang and Lujun Zhang
ABSTRACT Few studies focus on the quantitative impact of upslope inflow rate and slope gradient on rill development and erosion processes. Field plot experiments under varying inflow rates (6–36 L min 1 m 1) and slope gradients (26, 42 and 57%) were conducted to address this issue. The results showed soil loss rates significantly demonstrated temporal variability in relevance to the
Pei Tian Tieniu Wu Key Laboratory for Geographical Process Analysis & Simulation of Hubei Province, Central China Normal University, Wuhan 430079, China
rill developing process. Rill erosion and its contribution to soil loss increased with increasing inflow rates and slope gradients by power functions. There was a threshold inflow discharge (12–24 L min 1 m 1), under which, rill erosion became the dominant erosion pattern. At the initial stage, downcutting of rill bottom and headward erosion were obvious, whereas rill broadening was significant at the actively rill developing period. Rill density increased with slope gradient increasing from 26% to 42%, and then decreased. For the 57% slope under high inflow rates (24– 1
36 L min
1
m ), gravity caused an increase in the collapse of rills. Mean rill width increased with
increasing inflow rates but decreased as slope gradients increased, while mean rill depth increased with increasing inflow rates and slope gradients. Stream power and rill flow velocity were the best hydrodynamic parameter to simulate rill erosion and rill morphology, respectively. Key words
| flow hydrodynamics, rill development, rill morphology, slope gradient, upslope inflow rate
HIGHLIGHTS
• • • •
Rill erosion and its contribution to soil loss increased with inflow rate and slope gradient by a power function. The threshold inflow rate that made rill erosion dominated decreased with increasing slope gradient. There was a threshold slope gradient between 42.3% and 57.4% where the rill network development began to weaken. Stream power was the best parameter simulating rill erosion with a good linear relationship.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.168
Chengzhong Pan (corresponding author) Xinyi Xu College of Water Sciences, Beijing Normal University, Beijing 100875, China E-mail: pancz@bnu.edu.cn Tiantian Yang Lujun Zhang School of Civil Engineering and Environmental Science, The University of Oklahoma, Oklahoma 73019, USA
1202
P. Tian et al.
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
SYMBOLS AND ABBREVIATIONS S
Slope gradient (sine of the slope angle), %
erosivity of flowing water exceeds a certain threshold of
q
Inflow discharge per unit width of the plot,
soil resistance; as a result, the micro-terrain of hillslope sur-
1
L min
m
1
face changes and gradually forms a rill (Govers et al. ;
hr
Rill flow depth, m
Knapen et al. ; Wang et al. ). The main factors
vs
Surface rill flow velocity, m s 1
affecting soil detachment and sediment transport on hill-
vr
Mean rill flow velocity, m s 1
slopes include slope gradients (Koulouri & Giourga ;
α
Correction factor in determining mean rill flow
Fu et al. ), overland flow velocity, and flow depth
velocity
(Gimenez & Govers ). During rill erosion processes,
g
Acceleration due to gravity, m s 2
the evolution of rill networks greatly affects the confluence
Re
Reynolds number
path of runoff, soil loss, and the micromorphology of the
Fr
Froude number
slope surface (Consuelo et al. ; Auerswald et al. ;
f
Darcy–Weisbach resistance coefficient 2
Fang et al. ; Tian et al. ). Especially on steep loess
τ
Shear stress, N m
W
Stream power, N m 1 s 1
(Han et al. ; Yinglan et al. ), rills develop rapidly,
1
causing high soil loss (Lei et al. ; Zhang et al. ),
hillslopes with sparse vegetation cover in northern China
φ
Unit stream power, m s
SLR
Soil loss rate, g m 2 min 1 2
RER
Rill erosion rate, g m
TRL
Total rill length, m Rill density, m m
RD
and there is a lack of effective prevention methods to control 1
min
2
rill erosion on bare steep slopes in this region (Kou et al. ). Therefore, a further investigation of rill erosion processes on steep hillslopes is urgently needed.
MRW Mean rill width, cm
Rill erosion may be affected by hydrological parameters,
MRD Mean rill depth, cm
such as rill flow stream power (Mahmoodabadi et al. a),
RWD
The rill width–depth ratio
runoff rate in rills (Wirtz et al. ), hydraulic slope, and rainfall intensity (Shen et al. ), as well as topographic factors (Kinnell ; Mahmoodabadi et al. b). The rill
INTRODUCTION
flow has a significant effect on the transportation of eroded soils, which will accordingly affect the soil erosion
Soil erosion is a critical concern for the sustainable develop-
rate (Di Stefano et al. ). For rills formed at the hillslope,
ment of agricultural regions worldwide (Bennett et al. ),
they quickly adapted their topographic characteristics in
and soil erosion could cause non-point source pollution
response to the changes of upslope inflow discharge and
(Wang et al. a, b), soil organic carbon loss (Fang
slope gradient (Nearing ; Nord & Esteves ; Stefano
et al. ), and biodiversity reduction (Li et al. ).
et al. ). Moreover, the erosivity of flowing water is clo-
Thus, soil erosion is a serious threat to the environment
sely related to upslope inflow discharge and slope
and agricultural security (Jiang et al. ). Rill erosion is
gradient, thus, upslope inflow discharge and slope gradient
one of the main patterns of soil erosion by water on sloping
are two important factors for rill erosion (Berger et al.
croplands and rangelands (Bewket & Sterk ; Kimaro
; Sun et al. ; Shen et al. ).
et al. ; Wu et al. ).
The slope gradient is an important topographic factor
Rill erosion primarily results from concentrated surface
which has direct effects on flow erosivity and infiltration,
flow (Kimaro et al. ; He et al. ), which is the main
thus indirectly influences rill erosion processes (Shen
source of sediment yield in hillslope erosion processes
et al. ; Zhang et al. ). Giménez et al. () indi-
(Cerdan et al. ; Wagenbrenner et al. ; Mirzaee &
cated that slope roughness is enhanced with the slope
Ghorbani ). Generally, rill erosion occurs when the
gradient
becoming
steeper,
whereas
flow
velocity
1203
P. Tian et al.
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
decreases until a critical hydraulic condition (Froude
formation and development process of rill morphology
number is 1) of the flow is reached. It has also been recog-
timely responds to the rill erosion process. After the emer-
nized that the critical flow hydraulic condition for rill
gence of rills, with the continuous flowing water, rills
initiation is closely correlated to the slope gradient
develop and show the phenomenon of bifurcation, merging
(Ziadat & Taimeh ; Ran et al. ). In addition,
and connecting on the slope, and finally evolve into rill net-
many researchers found an increase in the slope gradient
works (He et al. ; Shen et al. ). The indicators
can lead to an increase in rill erosion amount (Berger
describing rill morphology are usually rill length, rill
et al. ; Fang et al. ). Ziadat & Taimeh ()
width, rill depth, and rill networks (Bewket & Sterk ;
reported that soil erosion in uncultivated land was
Raff et al. ). After the formation of rill networks, over-
mostly influenced by slope gradient; thus, the determi-
land flows concentrated into rills can promote the
nation of slope gradient factor was an essential part of
development of rills, broadening the rill width and deepen-
rill erosion prediction (Sun et al. ). However, Liu
ing the rill length (Tian et al. ). Furthermore, the
et al. () observed that the relationship between slope
formation and development of rills contribute to the runoff
gradient and soil erosion varied when slope gradient
connectivity and concentration along the rill networks
reached certain thresholds. For instance, Yair & Klein
(Romero et al. ; Heras et al. ). However, the influ-
() found the amount of soil erosion decreased with
ence mechanisms of inflow discharge and slope gradient
the increase of slope gradients. Koulouri & Giourga
on rill morphological characteristics and rill network for-
() stated that when the slope reached 40%, the domi-
mation need to be further investigated.
nant factor for soil loss was no longer land use, but slope
It is important to clarify the hydrodynamic mechanism
gradient. In addition, the study in a vineyard area by Pijl
of rill erosion since sediment detachment and transport by
et al. () indicated that rittochino (vertical cultivation)
overland
with a slope of 39.5% had the most severe soil loss.
initiation is determined by the established threshold flow
flow
are
energy-consuming
processes.
Rill
Typically, power functions are commonly used to fit the
hydraulic conditions, and further rill development is con-
correlation between the slope gradient and the amount of
trolled by different thresholds (Wirtz et al. ). Rill
soil erosion (Di Stefano et al. ). In some experiments
erosion usually increases with increasing upslope inflow dis-
conducted on gentle slopes, significant relationships were
charge due to the increase of flow hydrodynamics (Di
found between rill erosion and slope gradient (Liu et al.
Stefano et al. ), such as flow pattern (Foster et al.
; Sirjani & Mahmoodabadi ), whereas in some
), flow velocity (Govers et al. ), depth (An et al.
other studies, no significant relationship was reported (e.g.,
), shear stress (Nearing et al. ), and stream power
Chaplot & Le Bissonnais ). Zhang et al. () found
(Berger et al. ). Hillslopes’ soil loss significantly
the soil detachment rate could be well predicted by a
increases after rill formation due to increases in rill flow
power function of upslope inflow discharge and slope gradi-
depth, rill flow velocity, and shear stress compared to
ent, and detachment rates were more sensitive to upslope
sheet flow (Liu et al. ). Numerous studies have verified
inflow discharge than to slope gradients. Therefore, there
that the rill flow hydraulic parameters are related to upslope
is an interaction effect between slope gradient and upslope
inflow discharge and slope gradient, whereas the relation-
inflow discharges on the amount of erosion and the critical
ships are different in the studies (Mirzaee & Ghorbani-
hydraulic conditions of rill initiation. Thus, it is necessary
Dashtaki ). For instance, Giménez & Govers () indi-
that more attention should be focused on the effects of
cated that rill flow velocity is independent of the slope
upslope inflow discharges and slope gradients on rill flow
gradient on the mobile bed, while it becomes slope-depen-
hydraulics and its dynamic mechanism at steep hillslopes.
dent on the fixed bed. The flow in rills has higher
Rill network and its morphological characteristics play
velocities, depth and sediment transport capacity than the
an important role in determining the flow characteristics
flow in the inter-rill areas (Gatto ; Mirzaee & Ghor-
of surface runoff and sediment delivery in hillslope erosion
bani-Dashtaki ). Furthermore, the rill bed form varies
processes (Zhang et al. ; Wu et al. ). In fact, the
with erosion processes, which in turn, changes the
1204
P. Tian et al.
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
hydraulics and erosion capacity of rill flow (Zhang et al.
Hydrology Research
|
51.5
|
2020
Experimental plots
; Jiang et al. ). In summary, it is still not thoroughly understood what
A fallow loess hillslope (35 m length and 25 m width) was
mechanisms are dominating the inflow discharge and
selected as the experimental field (Figure 1). The particle
slope gradient on the rill erosion process and its hydrodyn-
size distribution and basic physical and chemical properties
amic mechanisms, and the rill development and its
of the soils in the 0–20 cm surface layer are presented in
morphological characteristics, especially for steep hillslopes.
Tables 1 and 2, respectively.
Therefore, in this study, we performed field plot experiments
The runoff plot experiments were conducted on the
under combinations of different upslope inflow discharges
selected hillslope. Lei et al. () found sediment transport
and slope gradients at steep loess hillslopes. The objectives
capacity would not increase beyond certain values of rill
of this study are to: (1) quantify the effects of slope gradient
length and slope. Furthermore, considering the character-
and inflow discharge on rill erosion processes and rill mor-
istics of small erosion plots, we designed the runoff plot
phological characteristics; (2) evaluate rill flow hydraulic
of 5 m in length and 2 m in width, which is bounded by
characteristics and the hydrodynamic mechanism of rill
concrete walls or aluminum plastic plates. A specifically
erosion.
designed PVC pipe with a row of equally spaced (1.0 cm) grooves (0.5 cm width, 1.5 cm arc length) was fixed at the top of the runoff plot and set along the longitudinal direction
MATERIALS AND METHODS
of the PVC pipe to make sure the upslope inflow water flowed uniformly over the whole runoff plot. Meanwhile, a
Study site
PVC pipe was set at the bottom of the runoff plot to collect runoff sediment samples during the experiment. The sche-
The study site is located in the hilly loess area of northern
matic diagram of the runoff plot and its affiliated facilities
China, where the Guanting Reservoir is located (Figure 1).
is shown in Figure 2.
This region belongs to a transition zone between the northwest plateau of Beijing and the North China Plain,
Experimental design
which is one of the key areas implementing soil and water conservation in China. The Guanting Reservoir was built
In this study, we monitor the comparison results by setting the
in the early 1950s and long served as an important drinking
slope gradient at three levels, and inflow discharge at six levels.
water source for Beijing City. The hilly areas account for
For each slope gradient, two runoff plots with the same size
approximately 37% of the total area of Guanting Reservoir
and treatment were constructed at the selected hillslope. In
watershed (38 150 –41 14.20 N, 112 8.30 –116 20.60 E), and
total, six runoff plots with the same size of three different
most of the hilly areas are highly erodible loess, covering a
slope gradients were used in our experiments. All the exper-
total area of 500 km2 (Hu & Wang ). The main causes
iments were repeated twice to ensure the reproducibility of
leading to severe soil erosion here include erodible loessial
results. Given the fact that this experiment was performed in
soils, relatively sparse vegetation cover, and short-duration
the same area as that by Tian et al. (), six values of the
rainstorms (Tian et al. ). The climate is semi-arid conti-
inflow discharge per unit width (1 m) of the plot ranging
nental, with a mean annual temperature of 9 C (Li & Pan
from 6 to 36 L min 1 m 1 were used. Each inflow discharge
). Mean annual precipitation, most of which occurs in
was introduced at the head of the plot. Considering the
the summer months as short-duration and high-intensity
characteristics of hillslope erosion driven by short-duration
rainstorms, varies from 370 to 480 mm. Field runoff plot
rainstorms in the local area, the simulated scouring test lasts
experiments were performed at the Huailai Field Exper-
60 minutes for each experimental treatment. Rill erosion was
imental Station (40 150 32″ N, 115 370 01″ E), Beijing
the most significant on bare loess slopes with slope gradients
Normal University, which is located along the east side of
of 17.6–57.7% (Shen et al. ). The rill is prone to intersect
the Guanting Reservoir (Figure 1).
on loess hillslopes with a slope gradient over 15 (26%),
P. Tian et al.
1205
Figure 1
|
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
(a) The study area and (b) the location of the experimental hillslopes with field plots.
which is generally considered as the lower limit of steep slopes
driven by upslope inflow on a wide range of steep hillslopes,
in the hilly loess area of China. Although 25 (42%) is the upper
slope gradient (sine of the slope angle) of 26, 42, and 57%
limit of the slope gradient for cultivation in China set by the
were selected. According to the characteristics of short-term
government, the phenomenon of cultivation in steeper slopes
rainfall–runoff erosion in this area, the duration of each trial
is still common in the hilly loess area (Li & Pan ). There-
lasted 60 minutes after the initiation of surface runoff at the
fore, to investigate rill development and erosion processes
plot outlet.
P. Tian et al.
1206
Table 1
|
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
Particle size distribution for the topsoil soil (0–20 cm) of the sloping field
Particle size distribution (%) 2–1 mm
1–0.5 mm
0.5–0.25 mm
0.25–0.1 mm
0.1–0.05 mm
0.05–0.02 mm
0.02–0.002 mm
< 0.002 mm
1.29 ± 0.07
1.21 ± 0.19
3.53 ± 0.66
7.42 ± 0.15
25.34 ± 0.35
30.40 ± 0.62
11.52 ± 0.24
19.50 ± 0.59
As an example, 1.29 ± 0.07 refers to the mean with standard deviations, the same in other cells.
Table 2
|
Basic physical and chemical properties for the topsoil (0–20 cm) of the sloping field
the usage of backfill for the surface soil treatment, experiments with inflow discharges ranging from low to high were conducted.
Soil bulk
Soil water
Soil
Surface treatment
density (g cm 3)
content (%)
porosity (%)
Total N (mg g 1)
Organic C (mg g 1)
Bare slope
1.35 ± 0.50
13.0 ± 2.0
44.57 ± 2.32
0.56 ± 0.18
422 ± 5.49
As an example, 1.35 ± 0.50 refers to the mean with standard deviations, the same in other cells.
Runoff sediment samples were collected from the outlet of the plot at intervals of 4 minutes, and the duration of collecting each sample was recorded by a stopwatch. These samples were used to calculate soil loss rates and surface runoff rates during the experimental process. Surface flow velocities in rills (vs) were measured by the dye (KMnO4) tracing method with the slope segment of 0.5–1.5 m, 1.5–2.5 m, 2.5–3.5 m, and 3.5–4.5 m, respectively. The water depth in rills (hr) was measured perpendicularly to the rill bed using a thin steel ruler with 0.1 mm precision and the measurement sections were the same as the velocity measurements. The measurements of rill morphology (length, width, depth) were performed along each rill at a 5 cm interval to compute the rills’ volume. Then, the rill erosion amount can be obtained by multiplying the rills’ volume with soil bulk density. In addition, photographs were taken during the experimental process. The maps of rill networks at the end of the experiment were plotted, which included three steps. First, as soon as the scouring test was over, we used a highdefinition digital camera to take photos of the entire plot. Second, according to the distribution of rills formed on the
Figure 2
|
Field runoff plot system at the study site.
plot, we drew a sketch of the rill network on white paper according to a certain scale. Third, for each rill, we measured
Experimental procedures and measurements
its length and the width every 3–5 cm along the length of the rill. Last, the maps of rill networks were made by Adobe
Before the experiment, upslope inflow discharge was cali-
Photoshop, combining the photos from the above three steps.
brated to the designed values manually. In addition, the topsoils (0–20 cm) were sampled by cutting rings on the
Data calculation and analysis
top, middle, and bottom of the slope surface to measure the soil water content and dry bulk density. For each treatment, the antecedent soil water content and soil bulk density were controlled as 13 ± 2% and 1.3 ± 0.1 g cm 3, respectively, before the start of the experiment. To reduce
The slope gradient (S, %) in this study is the sine of the slope angle (θ, ) for field plots: S ¼ sin θ × 100%
(1)
1207
P. Tian et al.
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
The surface runoff rate at the plot outlet (qout) is calcu-
|
51.5
|
2020
Shear stress (Foster et al. ), stream power (Bagnold ), and unit stream power (Yang ) are used to reflect
lated as follows: qout ¼ (M m)=(1000ρ bt)
Hydrology Research
(2)
where qout is surface runoff rate at the plot outlet, L min 1 m 1; M is the total mass of a runoff sediment sample, g; m is the dry weight of sediment in a runoff sediment sample, g; ρ is the density of water used in the experiment, g L 1; t is the duration of collecting a runoff sediment sample, min; b is the width of the plot, m; 1,000 is the dimensional adjustment coefficient.
the hydrodynamic parameters of rill flows, which are calculated as: 8 < τ ¼ ρghr J W ¼ τvr : φ ¼ vr J
(7)
where τ is flow shear stress (N m 2); W is flow stream power (N m 1 s 1); φ is unit stream power (m s 1); ρ is the density of water (kg m 3).
The runoff coefficient (RC) corresponding to each runoff sample is obtained by the following equation (Tian et al. ):
Rill density (RD) is the total rill lengths (TRL) divided by the whole hillslope area (m m 2) (Bewket & Sterk ). The RD value is suitable for roughly describing the morphological characteristics of rills developed at hillslopes. The
RC ¼ qout =qin
(3)
where qin is upslope inflow discharge (L min 1 m 1).
formulas of TRL and RD are: RD ¼ TRL=A0
Mean rill flow velocity (vr) is estimated by the measured surface rill flow velocity (vs) multiplied by a correlation factor α as the equation: vr ¼ αvs
(4)
and 0.8 for turbulent flow (Horton et al. ; Li et al. ). Reynolds number (Re) and Froude number (Fr) can be used to reflect the rill flow regime, which are calculated as: Re ¼ vr hp r =v m ffiffiffiffiffiffiffi Fr ¼ vr = ghr
n X
TRLi
(9)
i¼1
where a is 0.67 for laminar flow, 0.7 for transitional flow,
TRL ¼
(8)
where TRL is the total lengths of all rills on the studied hillslope, m; RD is the rill density, m m 2; TRLi is the total length of a rill and its bifurcations (m); and i ¼ 1, …, n represents the number of the rills; A0 is the plane area of the studied hillslope (m2). In addition, for a certain cross section of a rill, the rill width–depth ratio (RWD) can reflect the shape of the rill cross section. For a certain cross section of a rill, RWD is
(5)
where hr is the mean rill flow depth (m); νm is kinematical viscosity (m2 s 1); and g is acceleration of gravity (m s 2). The resistance to rill flow is estimated by Darcy–Weisbach friction coefficient ( f), which is calculated as:
calculated as follows: RWD ¼ MRW=MRD
(10)
where MRW is the mean rill width, cm; MRD is the mean rill depth, cm; RWD is the rill width–depth ratio. The amount of rill erosion was estimated by volumetric measurements of rills on each soil erosion plot. In our exper-
f ¼ 8 g hr J =v2r
(6)
iments, hillslope soil loss is the sum of the amount of sheet erosion and rill erosion. Thus, sheet erosion was obtained by
where g is gravitational acceleration (m s 2); J is surface 1
subtracting the rill erosion from the hillslope soil loss. The
slope (m m ), which is equal to S in value; hr is the depth
statistical tests include variance analysis, correlation analy-
of rill flow (mm).
sis, equation simulation, and were conducted using the
P. Tian et al.
1208
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
IBM SPSS Statistics for Windows Version 19.0. The data
conditions. This indicated the temporal variation increased
graphs were drawn using ORIGIN 9.0.
with the increase of inflow discharges. With the same inflow discharge, the surface runoff rates were higher under higher slope gradient conditions. This indicated
RESULTS AND DISCUSSION
that, to a certain extent, soil infiltration rates decreased with increasing slope gradients. Overall, under the same
Temporal changes in surface runoff and soil loss
inflow discharge condition, there were no significant differences in surface runoff processes among different
Figure 3 shows the temporal variations of surface runoff rates at the runoff plot outlet during the experimental process under different slope gradients and inflow discharges. According to Figure 3, there were no significant differences in the trend of surface runoff rate over time under
treatments. To quantify the overall effect of slope gradients and inflow discharges on surface runoff production, the mean runoff coefficients during the whole experimental process for different treatments are presented in Figure 4.
different treatments. For each treatment, surface runoff
Under the same slope gradient condition, the mean
rates increased with time in the initial stage of the exper-
runoff coefficients slightly increased with the increasing
iment, and then gradually became steady. With the same
inflow discharges (Figure 4), indicating soil infiltration
slope gradient, the increasing speed of surface runoff rate
rates decreased as inflow discharges increased. Under the
was
same inflow discharge condition, the mean runoff
higher
Figure 3
|
under
higher
upslope
inflow
discharge
Temporal changes in surface runoff rates under three slope gradients and six inflow discharges. S stands for the slope gradient. (a) S ¼ 26%, (b) S ¼ 42% (c) S ¼ 57%.
P. Tian et al.
1209
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
as inflow discharges increased. This phenomenon implied that the decreasing effect of slope gradients on soil infiltration rates became weakened with the increases of inflow discharges. To explore the temporal variations of soil loss during the experiment, the changes in soil loss rates with duration for different inflow discharges and slope gradients are illustrated in Figure 5. According to Figure 5, the soil loss rates between different treatments were of obvious differences, especially for different inflow discharges. Under different conditions, the starting time for rill initiation varied. Specifically, the starting time of rill formation became earlier with the Figure 4
|
Mean runoff coefficients for different slope gradients and inflow discharges.
increasing slope gradient and inflow discharge, which indicated that rills formed more easily under steeper slopes
coefficients increased with the increase of slope gradients,
and high inflow discharges. In addition, the duration of rill
suggesting the soil infiltration rates decreased as slope gra-
initiation reduced as the slope gradient increased. For
dients increased. However, the rate of increment in mean
most treatments, the soil loss rate curves were high in the
runoff coefficients with slope gradients gradually decreased
early minutes during the experimental process, which was
Figure 5
|
Temporal changes in soil loss rates under different slope gradients and inflow discharges. (a) Slope gradient ¼ 26%, (b) slope gradient ¼ 42%, (c) slope gradient ¼ 57%.
P. Tian et al.
1210
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
mainly because the loose soil particles were relatively
Under the lowest inflow discharge (6 L min 1 m 1), the
abundant on the hillslope surface at the beginning of the
curves of soil loss rate generally declined with slight fluctu-
experiment, leading to numerous sediments being detached
ation during the entire experiment. For relatively high inflow
and transported by surface sheet flow, especially under high
discharges, the curves of soil loss rate sharply decreased to
inflow discharges. This phenomenon corresponds to the
the valley at approximately 10 to 20 minutes, which was
experimental findings of Parsons & Stone ().
mainly due to decreases in the detachment rate of soil par-
Overall, for most experimental treatments, two peak
ticles and the number of loose particles on the slope
values of soil loss rates existed during the whole exper-
surface. Then, the soil loss rate curves reached another
imental process. At the initial stage of the experiment,
crest at approximately 25 to 45 minutes after the run,
there was much infiltration on the 57% slope and the
which was mainly due to the active development of rills,
erosion capacity of the surface runoff is limited. In
contributing to a sharp increment in the soil loss rate. There-
addition, the soil loss rate was relatively low at the begin-
after, the soil loss rate curve descended and tended to be an
ning of the experiment, as shown in Figure 5(c). After 8
equilibrium since the development of rills gradually entered
minutes, there appeared a first peak value of soil loss
into a relatively steady stage.
rate, which was because the runoff increased and there were abundant erodible soil particles on the hillslope sur-
Contributions of slope gradient and inflow discharge to
face. Thus, the soil loss rate also increased rapidly
rill erosion
afterwards. As the experiment continued, the soil loss rate decreased with the reduction of erodible soil par-
For each slope gradient treatment, rill erosion amount, hill-
ticles. During 20 to 35 minutes, with the active
slope soil loss, and the contribution of rill erosion to
evolution of rill erosion, the soil loss rate increased to
hillslope soil loss gradually increased with the increase of
the second peak value. After 35 minutes, the rill develop-
inflow discharges (Table 3). This phenomenon indicated
ment entered a relatively stable stage, and the soil loss
that the erosion capacity of flowing water greatly increased
rate slowly decreased.
as inflow discharges increased. For the lowest slope gradient
Table 3
|
Rill erosion, sheet erosion, hillslope soil loss, and the contribution of rill erosion to hillslope soil loss for different slope gradients and inflow discharges
Rill erosion (kg)
Sheet erosion (kg)
Hillslope soil loss (kg)
Contribution of rill erosion to hillslope soil loss (%)
6 12 18 24 30 36
1.08 3.16 9.58 18.36 33.97 43.34
3.83 6.42 13.78 13.85 18.29 19.47
4.92 9.58 23.36 32.21 52.27 62.81
22.0 33.0 41.0 57.0 65.0 69.0
42
6 12 18 24 30 36
3.28 12.40 27.73 39.86 64.89 81.47
11.76 19.39 17.07 18.30 21.35 17.87
15.04 31.78 44.80 58.16 86.24 99.33
21.8 39.0 61.9 68.5 75.2 82.0
57
6 12 18 24 30 36
5.19 14.61 37.56 56.75 73.10 90.46
8.83 12.96 16.87 18.92 19.43 15.96
14.01 27.57 54.43 75.66 92.53 106.43
37.0 53.0 69.0 75.0 79.0 85.0
1
Slope gradient (%)
Inflow discharge (L min
26
m
1
)
1211
P. Tian et al.
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
condition (26%), sheet erosion increased with increasing
changed very slightly (15.04–99.33 kg and 14.01–106.43 kg
inflow discharges. However, with the slope becoming stee-
at slope gradients of 42 and 57%, respectively).
per, sheet erosion increased as inflow discharges increased
Soil loss rate (SLR) and rill erosion rate (RER) can be
up to 30 L min 1 m 1, and then decreased as inflow dis-
expressed by inflow discharges (q) and slope gradients (S)
charges continued to increase.
as dual power functions, respectively, as follows:
For the lowest slope gradient condition, the contribution of rill erosion to hillslope soil loss was 41.0% and 57.0%,
SLR ¼ 0:0199q1:239 S1:075
(R2 ¼ 0:954, p < 0:01, n ¼ 54)
respectively, under inflow discharges of 18 and 24
(11)
L min 1 m 1. This result showed that rill erosion became the dominant soil erosion pattern instead of sheet erosion
RER ¼ 0:0004q1:868 S1:524 (R2 ¼ 0:977, p < 0:01, n ¼ 54)
when inflow discharge increased to 24 L min 1 m 1. Therefore, we can conclude that there was a threshold of inflow discharge ranging from 18 and 24 L min 1 m 1, under which, the dominant soil erosion pattern started to change from sheet erosion to rill erosion. With the slope gradient increasing to 42%, rill erosion became the dominant erosion 1
1
(12) where SLR is soil loss rate, g m 2 min 1; RER is rill erosion, g m 2 min 1; q is inflow discharge per unit width, L min 1 m 1; and S is slope gradient, %. According to Equations (11) and (12), both soil loss and
m , which
rill erosion rates increased by a power function with an
indicated that the critical inflow discharge controlling the
increase in either inflow discharge or slope gradient. Com-
main soil erosion pattern was between 12 and 18
pared with the exponents of inflow discharge, the slope
pattern when inflow discharge was 18 L min
L min 1 m 1. However, with slope gradient increasing to
gradients were lower, which indicated that the inflow dis-
57%, rill erosion became the dominant soil erosion pattern
charge played a more important role in both soil loss and
when inflow discharge was 12 L min 1 m 1, which demon-
rill erosion rates than the slope gradient. In addition, both
strated that the critical inflow discharge controlling the
the exponents of inflow discharge and slope gradient in
main soil erosion pattern was between 6 and 12
Equation (11) were lower than those in Equation (12).
L min
1
1
m . In summary, the threshold inflow discharge
that controlled the dominant soil erosion pattern was related
Rill networks and morphology
to slope gradients. Overall, the critical inflow discharge that determined the dominant soil erosion pattern decreased as
Rill networks
slope gradients increased. For the same inflow discharge, with increasing slope
To explore the characteristics of rill networks developed at
gradients, both rill erosion and the contribution of rill ero-
different slopes, the rill networks at the end of the exper-
sion to hillslope soil loss generally increased, whereas the
iment for different inflow discharges and slope gradients
variations of sheet erosion showed no definite trend with
are illustrated in Figure 6
slope gradients. However, hillslope soil loss significantly
Based on the rills generated on the slope surface at the
increased with slope gradients varying from 26% to 42%,
end of the experiment (Figure 6), for each slope gradient
and then slightly varied as slope gradients ranged from
treatment, both the uniformity and density of rill network
42% to 57%. The variations in the contribution of rill ero-
distribution were enhanced with the increases of inflow
sion to hillslope soil loss and hillslope soil loss with slope
discharges.. In addition, the cross-merging phenomenon
gradients showed that the increase of slope gradients
between the rills became more obvious under higher
enhanced rill erosion, although it implied that there was
inflow discharges, which indicated that the rill network
probably a threshold of slope gradient where soil erosion
became more complex. Under the same inflow discharge
began to weaken. For example, rill erosion increased from
condition, the density of rill network increased with slope
3.28–81.47 kg to 5.19–90.46 kg with the increase of slope
gradient increasing from 26% to 42%, and then decreased
gradient from 42% to 57%, whereas hillslope soil loss
with the slope becoming steeper. This phenomenon was
P. Tian et al.
1212
Figure 6
|
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
Rill networks at the end of the experiment for different slope gradients and inflow discharges. (a) Slope gradient ¼ 26%, (b) slope gradient ¼ 42%, (c) slope gradient ¼ 57%.
more significant under higher inflow discharges. As the
depth of rills increased. This phenomenon was probably
slope gradient increased from 42% to 57%, the total length
different from previous studies, and was mainly related to
of rills decreased by 8.6–25.1% under inflow rates of 12–36
our experimental condition. In this study, rills were
1
L min
1
m , while the average rill depth increased by
formed only by the upslope inflow without raindrop
17.5–35.7%. Although the rill density decreased, the average
impacts. Under this circumstance, when the slope gradient
1213
P. Tian et al.
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
increased to a certain level (57%), the undercutting erosivity
the rills. In contrast, under other slope gradients and
of runoff increased significantly with the influence of grav-
inflow rate conditions, the main driving force of soil erosion
ity, making the rills continue to deepen. As a result, more
and rill development is the upslope inflow. For these treat-
runoff flowed into the deep rills and the number of rills
ments, the role of gravity on erosion and rill evolution was
formed was reduced compared to the 42% slope condition.
not obvious.
Therefore, the total length of rills formed under the 57% slope condition was less than that under the 42% slope,
Rill morphological characteristics
which was reflected in the decrease of rill density. This may also imply that there was a critical slope gradient
Different statistical indicators of rill morphology for differ-
between 42 and 57%, under which, the development of rill
ent slope gradients and inflow discharges are presented in
networks began to weaken. For the largest slope (57%) con-
Table 4.
dition, rills on the upper slope were significantly denser than
Based on the dynamic changes of total rill length, mean
those on the lower part of the slope, especially under
rill width, and the rill width–depth ratio with increasing
relatively low inflow discharges. This phenomenon was
slope gradients (Table 4), the vertical erosion in rills was
probably due to the strong erosion capacity of flowing
exaggerated with the slope becoming steeper since the grav-
water with low sediment concentration in the upper part
itational erosion occurred. This result is similar to He et al.
of the slope on the steep hillslopes.
(), in which the authors investigated the dynamic
As reported in a previous study (Zhang et al. ), the
changes of total rill length under different simulated rainfall
formation and evolution of drop pits on the slope surface
conditions. For each slope gradient treatment, total rill
was an essential procedure during rill development pro-
length, rill density, and mean rill depth gradually increased
cesses. Their finding has agreed with our experiments, in
as inflow discharges increased, while the increment rate
which the drop pit phenomenon was clearly observed at
declined with the increases in inflow discharges. However,
the initial stage of rill formation during the experimental process for all treatments. For the 26 and 42% slope, all
Table 4
rills were generally continuous at the end of the experiment. However, on the steeper slope (57%), there were some drop
|
Rill morphological characteristics for different slope gradients and inflow discharges
Slope
pits and discontinuous rills at the lower-middle part of the
gradient
Inflow discharge
runoff plot under relatively low inflow discharges at the
(S, %)
(q, L min
end of the experiment. A possible explanation for this
26
TRL
RD (m
MRW
MRD
(m)
m
(cm)
(cm)
RWD
6 12 18 24 30 36
12.24 17.85 22.52 27.74 32.68 36.36
1.22 1.79 2.25 2.77 3.27 3.64
1.8 2.9 4.7 5.5 6.2 6.5
0.8 1.1 1.7 2.4 3.1 3.4
2.31 2.59 2.86 2.27 2.01 1.91
42
6 12 18 24 30 36
15.30 22.04 26.64 33.72 38.46 44.28
1.53 2.20 2.66 3.37 3.85 4.43
1.6 2.5 3.9 4.7 5.6 6.0
1.0 1.4 2.2 2.9 3.7 4.0
1.58 1.75 1.78 1.63 1.50 1.50
57
6 12 18 24 30 36
17.24 20.10 24.64 28.36 32.22 33.15
1.72 2.01 2.46 2.84 3.22 3.32
1.4 2.0 3.6 4.5 5.3 5.8
1.4 1.9 2.6 3.3 4.4 4.7
1.05 1.08 1.37 1.38 1.20 1.23
phenomenon was related to the deep rills developed at the upper part of the slope. In this case, most of the surface runoff converged into the deep rills, and intensively scoured the deep soil layer. During this process, some of the surface runoff might turn into the interflow, which decreased the surface upslope inflow discharge over the lower part of the runoff plot. Thus, the drop pits and discontinuous rills at the lower part of the slope were unable to develop into continuous rills due to the weakened energy of flowing water. Furthermore, based on our observation, the gravitational erosion obviously occurred within the plot of 57% slope under upslope inflow rates of 24, 30, and 36 L min 1 m 1. Under these circumstances, we can conclude that gravity plays an important role in slope erosion and rill evolution, which causes an increase in the collapse of
1
m
1
)
2
)
TRL: total rill lengths of the whole plot, m; RD: rill density, m m 2; MRW: mean rill width, cm; MRD: mean rill depth, cm; RWD: the rill width-depth ratio.
1214
P. Tian et al.
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
under the same inflow discharge condition, mean rill width
important sediment preparing processes frequently occurred
generally showed a decreasing trend with the increases of
at steep hillslopes, such as headcut retreat, step-pool effects,
slope gradients. Furthermore, the rill width–depth ratio
and bank collapse, which were not adequately considered
varied slightly with the increase of inflow discharges, demon-
(Dong et al. ; Zegeye et al. ). Through the obser-
strating no obvious variation trend. This result is similar to
vations during the experiment, sediment yield at the runoff
the laboratory rill experiments conducted by Shen et al.
plot outlet primarily resulted from knickpoints, chutes,
() and He et al. (). According to a least significant
meanders, and sidewall sloughing. This was consistent
difference (LSD) test, mean rill depth under the lowest
with the findings by Stefanovic & Bryan (), who inves-
slope gradient was significantly different from that under
tigated rill development processes on loamy sand and
the highest slope gradient condition, whereas mean rill
sandy loam under laboratory conditions.
width and total rill length among three different slope gradient treatments were not significantly different.
According to the analysis above, the rill morphology was influenced by the interaction effects of inflow dis-
Under the same inflow discharge condition, rill density
charges and slope gradients. At the same time, the
first increased with the increase of slope gradients up to
regression analysis also indicated that rill morphological
42% or 57%, and then decreased as the slope gradient con-
indicators could not be well expressed by only inflow dis-
tinued to increase. For each inflow discharge treatment,
charges or slope gradients. The best fitting functions for
although both mean rill width and rill depth increased
rill morphological indicators, inflow discharges (q), and
with increasing slope gradients, the rill width–depth ratio
slope gradients (S) are presented in Table 5.
showed a decreasing trend as slope gradients increased.
Rill density, mean rill depth, and the rill width–depth
This phenomenon indicated that with the increase of slope
ratio can be described by the power function of inflow dis-
gradients, deepening rills were more significant than the
charges and slope gradients, whereas mean rill width was
widening rills because of the higher shear stress of concen-
well expressed by the linear function relationship of inflow
trated rill flow. The findings were consistent with the
discharges and slope gradients (Table 5). All the fitted
observations in rill experiments under field conditions by
equations were highly significant (p < 0.01). In addition,
Mirzaee & Ghorbani-Dashtaki (). For the same inflow
the coefficient of the inflow discharge was greater than
discharge condition, mean rill width and rill depth under
that of the slope gradient for all equations, which indicated
different slope gradients were not significantly different
that the inflow discharge factor played a more important
according to an LSD test. However, the rill density was sig-
role in the rill morphological indicator than the slope gradi-
nificantly different among six different inflow discharges
ent factor. The function form of mean rill width, inflow
according to an LSD test. For the highest slope gradient,
discharge, and slope gradient was consistent with the
the side wall collapse of rills was more significant than
result by Lei & Nearing (), who conducted flume exper-
that of lower slope gradients, mainly due to the obvious
iments under upslope inflow conditions.
phenomenon of gravity erosion under steep slopes. During the experimental process, one of the main sediment delivery processes was observed to be headwall retreat erosion at small knickpoints or steps, which contributed to the formation of continuous rills. Furthermore, for
Table 5
|
The fitted equation of rill morphological indicators, inflow discharges and slope gradients
steeper slopes (57%), bank collapse processes of rills due
Regression equation
R2
Sig.(p)
n
to the interactions of flowing water and gravity were another
RD ¼ 0.346S 0.125q 0.533
0.92
<0.01
18
significant erosion pattern resulting in soil loss. These
MRW ¼ 0.027S þ 0.157q þ 1.947
0.96
<0.01
18
0.96
<0.01
18
0.85
<0.01
18
phenomena were consistent with the observations for field
MRD ¼ 0.033S
rill experiments under inflow conditions by Wirtz et al.
RWD ¼ 32.734 S 0.798q 0.016
() and Tian et al. (). Therefore, the limitation of soil detachment or sediment transport equations was that
0.519 0.798
q
RD: rill density, (m m 2); MRW: mean rill width (cm); MRD: mean rill depth (cm); RWD: the rill width-depth ratio; S: slope gradient (%); q: inflow discharge (L min 1 m 1).
P. Tian et al.
1215
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
Hydrodynamic mechanisms of rill erosion and
to 57%, respectively. For all slope gradients, the Re value
morphology
began to exceed 2,000 when the inflow discharge increased to 24 L min 1 m 1, which implied that rill flows became tur-
Rill flow hydraulics
bulent flow according to the criterion of open channel flow (Tian et al. ). Fr was greater than 1.0 for all treatments,
Rill flow hydraulic parameters over the whole course of the
indicating that rill flows belonged to supercritical flow con-
experiment for different inflow discharges and slope gradi-
ditions. This observation is similar to the experimental
ents are presented in Table 6.
results for steep calcareous soils (Mirzaee & Ghorbani
Overall, mean rill flow velocity (vr) increased with
). In addition, Darcy–Weisbach resistance coefficient
increases in either slope gradient or inflow discharge
( f ) decreased with the increases of inflow discharges but
(Table 6); however, there were significant differences in vr
increased as the slope became steeper.
among six different inflow discharges, but slight differences among three different slope gradients. This result indicated
Correlations of rill erosion and rill flow hydrodynamics
that flow velocities in rills were mainly dependent on the magnitude of inflow discharges, which is similar to the
The variations of rill erosion rates with three hydrodynamic
observations by Nearing (), and Li et al. (), who
parameters of the rill flow, respectively, under different
found that flow velocities were largely independent of
slope gradients are illustrated in Figure 7. Average rill flow shear stress (τ), stream power (W ), and
slope gradients. Mean rill flow depth (hr) increased with increasing
unit stream power (φ) varied from 0.645 to 4.345 N m 2,
gradients
0.183 to 3.252 N m 1 s 1, and 0.061 to 0.419 m s 1, respect-
increased. Reynolds number (Re) increased as inflow dis-
ively, and the values were increasing with the increases of
charges
either upslope inflow discharge or slope gradient (Figure 7).
inflow discharges increased,
but
decreased
whereas
Re
as
slope
increased
and
then
decreased with increases of slope gradients, in which the
Rill erosion rates significantly increased with an increase of
slope gradients increased from 26% to 42% and from 42%
τ, W, and φ. The relationships between rill erosion rates and
Table 6
|
Mean rill flow hydraulic parameters for different slope gradients and inflow discharges
Slope gradient (%)
Inflow discharge (L min 1 m 1)
Rill flow velocity (m s 1)
Rill flow depth (×10 3 m)
Reynolds number
Froude number
Friction coefficient
26
6 12 18 24 30 36
0.284 0.368 0.455 0.527 0.607 0.659
2.41 3.50 4.57 5.97 7.21 8.86
544 1,021 1,653 2,498 3,477 4,640
1.9 2.0 2.2 2.2 2.3 2.2
1.28 1.11 0.94 0.92 0.84 0.87
42
6 12 18 24 30 36
0.329 0.448 0.564 0.640 0.699 0.733
2.13 3.28 4.30 5.51 6.64 8.12
556 1,166 1,927 2,798 3,690 4,728
2.3 2.5 2.7 2.8 2.7 2.6
1.46 1.22 1.01 1.00 1.01 1.13
57
6 12 18 24 30 36
0.362 0.464 0.574 0.629 0.711 0.748
1.99 3.13 3.88 4.63 5.40 6.21
571 1,152 1,768 2,311 3,050 3,689
2.6 2.6 2.9 3.0 3.1 3.0
1.70 1.63 1.32 1.31 1.20 1.24
P. Tian et al.
1216
Figure 7
|
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
Rill erosion rates versus rill flow shear stress, stream power, and unit stream power, respectively, for all treatments.
the above three hydrodynamic parameters were described
The critical hydrodynamic parameter could be obtained
well by the linear functions, which are listed as follows:
by the above functions. When no rill erosion occurred, i.e.,
8 < RER ¼ 48:57(τ 0:756) RER ¼ 53:98(τ 1:016) : RER ¼ 56:73(τ 1:433)
RER ¼ 0, then the critical value was determined. All the
(S ¼ 25:9%, R2 ¼ 0:96) (S ¼ 42:3%, R2 ¼ 0:98) (S ¼ 57:4%, R2 ¼ 0:97)
where RER is rill erosion rate (g m (N m 2), 8 < RER ¼ 60:33(W 0:218) RER ¼ 60:52(W 0:283) : RER ¼ 59:67(W 0:406)
2
hydrodynamic parameters, and further enhance the detach-
min ), τ is shear stress
). As a result, rill erosion rates increased with the
1
ment power and transport capacity of rill flows (Wang et al. increases of slope gradients. Moreover, as the slope becomes
(S ¼ 25:9%, R2 ¼ 0:97) (S ¼ 42:3%, R2 ¼ 0:98) (S ¼ 57:4%, R2 ¼ 0:98)
steeper, the gravity influenced the flow geometry in rills and increased the rill bed roughness, which led to an increase in the critical shear stress and the critical stream power for rill erosion (Nearing ). At the same time, for all treatments, the functional relations between rill erosion rates and those
W is stream power (N m 1 s 1),
three hydrodynamic parameters are analyzed, which are (S ¼ 25:9%, R2 ¼ 0:92) (S ¼ 42:3%, R2 ¼ 0:89) (S ¼ 57:4%, R2 ¼ 0:96) (15)
φ is unit stream power (m s 1).
increase in slope gradient will result in higher values of
(13)
(14)
8 < RER ¼ 997:04(φ 0:070) RER ¼ 912:31(φ 0:135) : RER ¼ 732:87(φ 0:209)
critical values increased as slope gradient increased. The
expressed as: 8 < RER ¼ 47:887(τ 0:916) RER ¼ 57:507(W 0:255) : RER ¼ 39:800(φ 0:051)
(R2 ¼ 0:89, p < 0:01, n ¼ 54) (R2 ¼ 0:64, p < 0:01, n ¼ 54) (R2 ¼ 0:92, p < 0:01, n ¼ 54) (16)
P. Tian et al.
1217
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Hydrology Research
|
51.5
|
2020
where RER is rill erosion rate (g m 2 min 1), τ is rill flow
According to Equation (17), when no rills formed, i.e.,
shear stress (N m 2), W is stream power (N m 1 s 1), φ is
MRD ¼ 0, then the critical rill flow velocity was determined.
unit stream power (m s 1). The critical shear stress, the criti-
Thus, in our experiments, the critical flow velocity for rill
cal stream power, and the critical unit stream power
occurrence could be estimated as 0.224 m s 1. Furthermore,
2
computed by Equation (16) were 0.916 N m , 0.255
this critical value was close to the measured flow velocity at
N m 1 s 1, and 0.051 m s 1, respectively. In contrast, the
the initial stage of rill development during the experiment
best hydrodynamic parameter to fit the rill erosion rate
(He et al. ).
was unit stream power with a good linear relationship.
CONCLUSIONS
Correlations of rill morphology and rill flow hydrodynamics
Plot experiments under six upslope inflow discharges (6–36 The Pearson correlation coefficients between rill morpho-
L min 1 m 1) and three slope gradients (26–57%) were con-
logical indicators and rill flow hydrodynamic parameters
ducted to explore the quantitative effects of inflow discharge
for all treatments are shown in Table 7.
and slope gradient on rill erosion and rill morphology, as
There was a negative correlation between the flow resist-
well as the hydrodynamic mechanism of rill erosion. The
ance coefficient and all the rill morphological indicators
results showed soil loss rates significantly related to the fluc-
(Table 7). However, RD, MRW, and MRD were positively
tuations during the process of rill development. Both rill
correlated with the other hydrodynamic parameters. Specifi-
erosion and its contribution to hillslope soil loss increased
cally, RWD was positively related to mean rill flow depth and
with the increases in inflow discharge and slope gradient,
Reynolds number but negatively correlated with other
which can be described by a power function. The inflow dis-
parameters. The most sensitive rill flow hydrodynamic par-
charge played a more important role than the slope gradient.
ameter to RD, MRW, MRD, and RWD was Reynolds
There was a threshold of inflow discharge (12–24
number, mean rill flow depth, mean rill flow velocity, and
L min 1 m 1), under which, rill erosion instead of sheet ero-
Froude number, respectively. Overall, the common sensitive
sion became the dominant erosion pattern, and the critical
parameter to various rill morphological indicators was shear
inflow discharge reduced as slope gradient increased. The
stress and stream power.
uniformity of rill network distribution enhanced with
In addition, there was a good linear function relation-
increasing inflow discharges. Rill density increased (17.7–
ship between mean rill depth (MRD) and rill flow velocity
25.4%) with slope gradient increasing from 26% to 42%,
(vr), which was expressed as:
and then decreased (7.5–25.1%) with the slope becoming steeper. Mean rill width increased with the increases of inflow discharge but decreased as the slope gradient (R2 ¼ 0:88, p < 0:01, n ¼ 54)
MRD ¼ 7:972(vr 0:224)
(17)
Table 7
|
increased. However, mean rill depth increased with increases in inflow discharge and slope gradient. The rill
Correlation coefficients between rill morphological indicators and rill flow hydrodynamic parameters for all treatments vr
hr
Re
Fr
f
τ
W
φ
RD
0.939**
0.919**
0.965**
0.438
0.548*
0.785**
0.818**
0.478*
MRW
0.859**
0.954**
0.934**
0.233
0.764**
0.633*
0.689*
0.298
MRD
0.971**
0.782**
0.880**
0.661*
0.326
0.944**
0.963**
0.753**
RWD
-0.376
0.871
0.579
0.785**
0.611*
0.609*
0.533*
0.780**
*: significant at p ¼ 0.05; **: significant at p ¼ 0.01; vr: rill flow velocity; hr: rill flow depth; Re: Reynolds number; Fr: Froude number; f : Darcy–Weisbach friction coefficient; τ: shear stress; W: stream power; φ: unit stream power; MRD: mean rill depth; MRW: mean rill width; RD: rill density; RWD: the rill width-depth ratio.
P. Tian et al.
1218
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
width–depth ratio decreased as slope gradients increased,
Hydrology Research
|
51.5
|
2020
REFERENCES
which indicated that deepening rills were more significant than the widening rills with the slope becoming steeper. In addition, stream power was the best hydrodynamic parameter correlated with rill erosion with a good linear relationship, and the critical stream power of rill erosion was 0.255 N m 1 s 1. The most sensitive hydrodynamic parameter to rill density, mean rill width, rill depth, and the rill width–depth ratio was Reynolds number, rill flow depth, rill flow velocity, and Froude number, respectively. Although the most sensitive hydrodynamic parameters to various rill morphological indicators were different, the common sensitive parameters were rill flow stream power and shear stress. The findings will help to understand the impacts of upslope inflow rate on morphological variations of rills, rill network development, and hydrodynamic mechanism of rill erosion on steep hillslopes. Weakening the erosion power of upslope runoff, such as taking measures to cover the soil surface, is useful for preventing rill formation and development, as well as the downslope soil loss.
ACKNOWLEDGEMENTS The authors wish to thank all of the technicians involved in field and laboratory work. This research is jointly funded by the
Natural
Science
Foundation
of
China
Project
(41907061), the Research Center on Mountain Torrent & Geologic Disaster Prevention of the Ministry of Water Resources, Changjiang River Scientific Research Institute (CKWV2019761/KY), and the Research Program from the State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau (A314021402-2005). We declare
that
we
have
no
financial
and
personal
relationships with other people or organizations that can inappropriately influence our work. The authors declare no conflicts of interest.
DATA AVAILABILITY STATEMENT All relevant data are included in the paper or its Supplementary Information.
An, J., Zheng, F., Lu, J. & Li, G. Investigating the role of raindrop impact on hydrodynamic mechanism of soil erosion under simulated rainfall conditions. Soil Science 177 (8), 517–526. Auerswald, K., Fiener, P. & Dikau, R. Rates of sheet and rill erosion in Germany – a meta-analysis. Geomorphology 111 (3), 182–193. Bagnold, R. A. An Approach to the Sediment Transport Problem From General Physics. US Geological Survey Professional Paper 442-I. USGS, Reston, VA, USA. Bennett, S. J., Gordon, L. M., Neroni, V. & Wells, R. R. Emergence, persistence, and organization of rill network. Natural Hazards 79, S7–S24. doi: 10.1007/s11069-0151599-8. Berger, C., Schulze, M., Rieke-Zapp, D. & Schlunegger, F. Rill development and soil erosion: a laboratory study of slope and rainfall intensity. Earth Surface Processes and Landforms 35 (12), 1456–1467. Bewket, W. & Sterk, G. Assessment of soil erosion in cultivated fields using a survey methodology for rills in the Chemoga watershed, Ethiopia. Agriculture, Ecosystems and Environment 97, 81–93. Cerdan, O., Le Bissonnais, Y., Couturier, A., Bourennane, H. & Souchère, V. Rill erosion on cultivated hillslopes during two extreme rainfall events in Normandy, France. Soil and Tillage Research 67 (1), 99–108. Chaplot, V. A. M. & Le Bissonnais, Y. Runoff features for sheet erosion at different rainfall intensities, slope lengths, and gradients in an agricultural loessial hillslope. Soil Science Society of America Journal 67, 844–851. Consuelo, C. R., Stroosnijder, L. & Guillermo, A. Sheet and rill erodibility in the northern Andean Highlands. Catena 70 (2), 105–113. Di Stefano, C., Ferro, V., Palmeri, V. & Pampalone, V. Flow resistance equation for rills. Hydrological Processes 31, 2793–2801. Di Stefano, C., Ferro, V., Palmeri, V. & Pampalone, V. Testing slope effect on flow resistance equation for mobile bed rills. Hydrological Processes 32 (5), 664–671. Dong, Y., Xiong, D., Su, Z., Duan, X., Lu, X., Zhang, S. & Yuan, Y. The influences of mass failure on the erosion and hydraulic processes of gully headcuts based on an in situ scouring experiment in dry-hot valley of China. Catena 176, 14–25. Fang, H., Sun, L. & Tang, Z. Effects of rainfall and slope on runoff, soil erosion and rill development: an experimental study using two loess soils. Hydrological Processes 29 (11), 2649–2658. Fang, Q., Wang, G., Liu, T., Xue, B. & Y, A. Controls of carbon flux in a semi-arid grassland ecosystem experiencing wetland loss: vegetation patterns and environmental variables. Agricultural and Forest Meteorology 259, 196–210.
1219
P. Tian et al.
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
Foster, G. R., Huggins, L. F. & Meyer, L. D. A laboratory study of rill hydraulics: II: Shear stress relationships. Transactions of the ASAE 27, 797–804. Fu, S., Liu, B., Liu, H. & Xu, L. The effect of slope on sheet erosion at short slopes. Catena 84, 29–34. Gatto, L. W. Soil freeze-thaw-induced changes to a simulated rill: potential impacts on soil erosion. Geomorphology 32, 147–160. Giménez, R. & Govers, G. Interaction between bed roughness and flow hydraulics in eroding rills. Water Resources Research 37 (3), 791–799. Gimenez, R. & Govers, G. Flow detachment by concentrated flow on smooth and irregular beds. Soil Sci Soc Am J 66 (5), 1475–1483. Giménez, R., Planchon, O., Silvera, N. & Govers, G. Longitudinal velocity patterns and bed morphology interaction in a rill. Earth Surface Processes and Landforms 29 (1), 105–114. Govers, G., Giménez, R. & Van, O. K. Rill erosion: exploring the relationship between experiments, modelling and field observations. Earth-Science Reviews 84 (3–4), 87–102. Han, D., Wang, G., Liu, T., Xue, B., Kuczera, G. & Xu, X. Hydroclimatic response of evapotranspiration partitioning to prolonged droughts in semiarid grassland. Journal of Hydrology 563, 766–777. He, J., Li, X., Jia, L., Gong, H. & Cai, Q. Experimental study of rill evolution processes and relationships between runoff and erosion on clay loam and loess. Soil Science Society America Journal 78, 1716–1725. He, J., Sun, L., Gong, H., Cai, Q. & Jia, L. The characteristics of rill development and their effects on runoff and sediment yield under different slope gradients. Journal of Mountain Science 13 (3), 397–404. He, J., Sun, L., Gong, H. & Cai, Q. Laboratory studies on the influence of rainfall pattern on rill erosion and its runoff and sediment characteristics. Land Degradation & Development 28 (5), 1615–1625. He, J., Sun, L., Gong, H. & Cai, Q. Comparison of rill flow velocity regimes between developing and stationary rills. Catena 167, 13–17. Heras, M. M.-D., Espigares, T., Merino-Martín, L. & Nicolau, J. M. Water-related ecological impacts of rill erosion processes in Mediterranean-dry reclaimed slopes. Catena 84, 114–124. Horton, R. E., Leach, H. R. & Vliet, R. V. Laminar sheet-flow. Transactions American Geophysical Union 15 (2), 393–404. Hu, C. & Wang, Y. Study on water-sediment optimum allocation in upstream basin and comprehensive measures of sediment control in Guanting reservoir. Reservoir sedimentation and general plan of water and sediment control. Journal of Sediment Research 2, 19–26 (in Chinese with English abstract). Jiang, F., Zhan, Z., Chen, J., Lin, J., Wang, M., Ge, H. & Huang, Y. Rill erosion processes on a steep colluvial deposit slope under heavy rainfall in flume experiments with artificial rain. Catena 169, 46–58.
Hydrology Research
|
51.5
|
2020
Jiang, Y., Shi, H., Wen, Z., Guo, M., Zhao, J., Cao, X., Fan, Y. & Zheng, C. The dynamic process of slope rill erosion analyzed with a digital close range photogrammetry observation system under laboratory conditions. Geomorphology 350, 106893. Kimaro, D. N., Poesen, J., Msanya, B. M. & Deckers, J. A. Magnitude of soil erosion on the northern slope of the Uluguru Mountains, Tanzania: sheet and rill erosion. Catena 75, 38–44. Kinnell, P. I. A. The impact of slope length on the discharge of sediment by rain impact induced saltation and suspension. Earth Surface Process and Landforms 34, 1393–1407. Knapen, A., Poesen, J., Govers, G., Gyssels, G. & Nachtergaele, J. Resistance of soils to concentrated flow erosion: a review. Earth Science Reviews 80 (1–2), 75–109. Kou, P., Xu, Q., Yunus, A., Dong, X., Pu, C., Zhang, X. & Jin, Z. Micro-topographic assessment of rill morphology highlights the shortcomings of current protective measures in loess landscapes. Science of the Total Environment 737, 139721. Koulouri, M. & Giourga, C. Land abandonment and slope gradient as key factors of soil erosion in Mediterranean terraced lands. Catena 69 (3), 274–281. Lei, T. & Nearing, M. A. Flume experiments for determining rill hydraulic characteristic erosion and rill patterns. Journal of Hydraulic Engineering l31 (11), 49–54 (in Chinese with English abstract). Lei, T., Zhang, Q., Zhao, J. & Tang, Z. A laboratory study of sediment transport capacity in the dynamic process of rill erosion. Transaction of American Society and Agriculture Engineers 44, 1537–1542 (in Chinese with English abstract). Li, C. & Pan, C. The relative importance of different grass components in controlling runoff and erosion on a hillslope under simulated rainfall. Journal of Hydrology 558, 90–103. Li, G., Abrahams, A. D. & Atkinson, J. F. Correction factors in the determination of mean velocity of overland flow. Earth Surface Processes and Landforms 21 (6), 509–515. Li, G., Zheng, F., Lu, J., Xu, X., Hu, W. & Han, Y. Inflow discharge impact on hillslope erosion processes and flow hydrodynamics. Soil Science Society of America Journal 80 (3), 711–719. Li, Q., Wang, G., Wang, H., Shrestha, S., Xue, B., Sun, W. & Yu, J. Macrozoobenthos variations in shallow connected lakes under the influence of intense hydrologic pulse changes. Journal of Hydrology 584, 124755. Liu, B., Nearing, M. & Risse, L. Slope gradient effects on soil loss for steep slopes. Transactions of the ASAE 37 (6), 1835–1840. Liu, G., Yang, M., Warrington, D., Liu, P. & Tian, J. Using beryllium-7 to monitor the relative proportions of sheet and rill erosion from loessal soil slopes in a single rainfall event. Earth Surface Processes and Landforms 36 (4), 439–448. Mahmoodabadi, M., Ghadiri, H., Yu, B. & Rose, C. a Morphodynamic quantification of flow-driven rill erosion parameters based on physical principles. Journal of Hydrology 514, 328–336. Mahmoodabadi, M., Ghadiri, H., Rose, C., Yu, B., Rafahi, H. & Rouhipour, H. b Evaluation of GUEST and WEPP with a
1220
P. Tian et al.
|
Upslope inflow and slope gradient on rill developing and flow hydrodynamics
new approach for the determination of sediment transport capacity. Journal of Hydrology 513, 413–421. Mirzaee, S. & Ghorbani-Dashtaki, S. Deriving and evaluating hydraulics and detachment models of rill erosion for some calcareous soils. Catena 164, 107–115. Nearing, M. Hydraulics and erosion in eroding rills. Water Resources Research 33 (4), 865–876. Nearing, M., Bradford, M. & Parker, C. Soil detachment by shallow flow at low slope. Soil Science Society of America Journal 55 (2), 339–344. Nord, G. & Esteves, M. The effect of soil type, meteorological forcing and slope gradient on the simulation of internal erosion processes at the local scale. Hydrological Processes 24 (13), 1766–1780. Parsons, A. J. & Stone, P. M. Effects of intra-storm variations in rainfall intensity on sheet runoff and erosion. Catena 67 (1), 68–78. Pijl, A., Reuter, L. E. H., Quarella, E., Vogel, T. A. & Tarolli, P. GIS-based soil erosion modelling under various steepslope vineyard practices. Catena 193, 104604. Raff, D. A., Ramírez, J. A. & Smith, J. L. Hillslope drainage development with time: a physical experiment. Geomorphology 62, 169–180. Ran, H., Deng, Q., Zhang, B., Liu, H., Wang, L., Luo, M. & Qin, F. Morphology and influencing factors of rills in the steep slope in Yuanmou Dry-Hot Valley (SW China). Catena 165, 54–62. Romero, C. C., Stroosnijder, L. & Baigorria, G. A. Sheet and rill erodibility in the northern Andean Highlands. Catena 70, 105–113. Shen, H. O., Zheng, F. L., Wen, L. L., Lu, J. & Jiang, Y. L. An experimental study of rill erosion and morphology. Geomorphology 231, 193–201. Shen, H. O., Zheng, F. L., Wen, L. L., Han, Y. & Hu, W. Impacts of rainfall intensity and slope gradient on rill erosion processes at loessial hillslope. Soil & Tillage Research 155, 429–436. Sirjani, E. & Mahmoodabadi, M. Effects of sheet flow rate and slope gradient on sediment load. Arabian Journal of Geosciences 7, 203–210. Stefano, C. D., Ferro, V., Pampalone, V. & Sanzone, F. Field investigation of rill and ephemeral gully erosion in the Sparacia experimental area, South Italy. Catena 101, 226–234. Stefanovic, J. R. & Bryan, R. B. Flow energy and channel adjustments in rills developed in loamy and sandy loam soils. Earth Surface Processes and Landforms 34, 133–144. Sun, L., Fang, H., Qi, D., Li, J. & Cai, Q. A review on rill erosion process and its influencing factors. Chinese Geographical Science 23 (4), 389–402. Tian, P., Xu, X., Pan, C., Hsu, K. & Yang, T. Impacts of rainfall and inflow on rill formation and erosion processes on steep hillslopes. Journal of Hydrology 548 (5), 24–39. Wagenbrenner, J. W., Robichaud, P. R. & Elliot, W. J. Rill erosion in natural and disturbed forests: 2. Modeling approaches. Water Resources Research 46, 1–12.
Hydrology Research
|
51.5
|
2020
Wang, Z., Yang, X., Liu, J. & Yuan, Y. Sediment transport capacity and its response to hydraulic parameters in experimental rill flow on steep slope. Journal of Soil and Water Conservation 70 (1), 36–44. Wang, D., Wang, Z., Shen, N. & Chen, H. Modeling soil detachment capacity by rill flow using hydraulic parameters. Journal of Hydrology 535, 473–479. Wang, G., Li, J., Sun, W., Xue, B., Y, A. & Liu, T. a Non-point source pollution risks in a drinking water protection zone based on remote sensing data embedded within a nutrient budget model. Water Research 157, 238–246. Wang, P., Yao, J., Wang, G., Hao, F., Shrestha, S., Xue, B., Xie, G. & Peng, Y. b Exploring the application of artificial intelligence technology for identification of water pollution characteristics and tracing the source of water quality pollutants. Science of the Total Environment 693, 133440. Wirtz, S., Seeger, M. & Ries, J. B. Field experiments for understanding and quantification of rill erosion processes. Catena 91, 21–34. Wu, S., Chen, L., Wang, N., Yu, M. & Assouline, S. Modeling rainfall-runoff and soil erosion processes on hillslopes with complex rill network planform. Water Resources Research 54 (12), 10117–10133. Wu, S., Chen, L., Wang, N., Li, J. & Li, J. Two-dimensional rainfall-runoff and soil erosion model on an irregularly rilled hillslope. Journal of Hydrology 580, 124346. Yair, A. & Klein, M. The influence of surface properties on flow and erosion processes on debris covered slopes in an arid area. Catena 1 (73), 1–18. Yang, C. T. Sediment Transport: Theory and Practice. McGraw Hill, New York, USA. Yinglan, A., Wang, G., Liu, T., Xue, B. & Kuczera, G. Spatial variation of correlations between vertical soil water and evapotranspiration and their controlling factors in a semi-arid region. Journal of Hydrology 574, 53–63. Zegeye, A. D., Langendoen, E. J., Steenhuis, T. S., Wolde, M. & Tilahun Seifu, A. Bank stability and toe erosion model as a decision tool for gully bank stabilization in sub humid Ethiopian highlands. Ecohydrology and Hydrobiology 20, 301–311. Zhang, G. H., Liu, B. Y., Nearing, M. A., Huang, C. H. & Zhang, K. L. Soil detachment by shallow flow. Transactions of the ASAE 45, 351–357. Zhang, P., Tang, H., Yao, W., Zhang, N. B. & Xi, Z. Experimental investigation of morphological characteristics of rill evolution on loess slope. Catena 137, 536–544. Zhang, P., Yao, W., Tang, H., Wei, G. & Wang, L. Laboratory investigations of rill dynamics on soils of the loess plateau of China. Geomorphology 293, 201–210. Ziadat, F. M. & Taimeh, A. Y. Effect of rainfall intensity, slope and land use and antecedent soil moisture on soil erosion in an arid environment. Land Degradation Development 24 (6), 582–590.
First received 24 May 2020; accepted in revised form 10 August 2020. Available online 24 September 2020