Environmental & Engineering Geoscience MAY 2015
VOLUME XXI, NUMBER 2
THE JOINT PUBLICATION OF THE ASSOCIATION OF ENVIRONMENTAL AND ENGINEERING GEOLOGISTS AND THE GEOLOGICAL SOCIETY OF AMERICA SERVING PROFESSIONALS IN ENGINEERING GEOLOGY, ENVIRONMENTAL GEOLOGY, AND HYDROGEOLOGY
Environmental & Engineering Geoscience (ISSN 1078-7275) is published quarterly by the Association of Environmental and Engineering Geologists (AEG) and the Geological Society of America (GSA). Periodicals postage paid at AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7301. Phone: 844-331-7867 and additional mailing offices. EDITORIAL OFFICE: Environmental & Engineering Geoscience journal, Department of Geology, Kent State University, Kent, OH 44242, U.S.A. phone: 330-672-2968, fax: 330-672-7949, ashakoor@kent.edu. CLAIMS: Claims for damaged or not received issues will be honored for 6 months from date of publication. AEG members should contact AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7301. Phone: 844-331-7867. GSA members who are not members of AEG should contact the GSA Member Service center. All claims must be submitted in writing. POSTMASTER: Send address changes to AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7301. Phone: 844331-7867. Include both old and new addresses, with ZIP code. Canada agreement number PM40063731. Return undeliverable Canadian addresses to Station A P.O. Box 54, Windsor, ON N9A 6J5 Email: returnsil@imexpb.com. DISCLAIMER NOTICE: Authors alone are responsible for views expressed in articles. Advertisers and their agencies are solely responsible for the content of all advertisements printed and also assume responsibility for any claims arising therefrom against the publisher. AEG and Environmental & Engineering Geoscience reserve the right to reject any advertising copy. SUBSCRIPTIONS: Member subscriptions: AEG members automatically receive the journal as part of their AEG membership dues. Additional subscriptions may be ordered at $40 per year. GSA members who are not members of AEG may order for $40 per year on their annual GSA dues statement or by contacting GSA. Nonmember subscriptions are $175 and may be ordered from the subscription department of either organization. A postage differential of $10 may apply to nonmember subscribers outside the United States, Canada, and Pan America. Contact AEG at 303-757-2926; contact GSA Subscription Services, c/o American Institute of Physics, 2 Huntington Quadrangle, Suite 1N01, Melville, NY 11747-4502. Single copies are $50.00 each. Requests for single copies should be sent to AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7301. Phone: 844-331-7867. © 2015 by the Association of Environmental and Engineering Geologists All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without permission in writing from AEG. THIS PUBLICATION IS PRINTED ON ACID-FREE PAPER EDITORS ABDUL SHAKOOR Department of Geology Kent State University Kent, OH 44242 330-672-2968 ashakoor@kent.edu
BRIAN G. KATZ Florida Department of Environmental Protection 2600 Blair Stone Rd. Tallahassee, FL 32399 850-245-8233 eegeditorbkatz@gmail.com
EDITORIAL BOARD ROBERT H. SYDNOR JEROME V. DEGRAFF USDA Forest Service Consulant THOMAS J. BURBEY CHESTER F. WATTS (SKIP) Virginia Polytechnic Institute Radford University SYED E. HASAN University of Missouri, Kansas City ASSOCIATE EDITORS JOHN W. BELL PAUL M. SANTI Nevada Bureau of Mines and Colorado School of Mines Geology ROBERT L. SCHUSTER U.S. Geological Survey RICHARD E. JACKSON (Book Reviews Editor) ROY J. SHLEMON R. J. Shlemon Geofirma Engineering, Ltd. & Associates, Inc. JEFFREY R. KEATON AMEC Americas GREG M. STOCK National Park Service PAUL G. MARINOS National Technical University RESAT ULUSAY Hacettepe University, Turkey of Athens, Greece CHESTER F. “SKIP” WATTS JUNE E. MIRECKI U.S. Army Corps of Radford University Engineers TERRY R. WEST Purdue University PETER PEHME Waterloo Geophysics, Inc NICHOLAS PINTER Southern Illinois University SUBMISSION OF MANUSCRIPTS Environmental & Engineering Geoscience (E&EG), is a quarterly journal devoted to the publication of original papers that are of potential interest to hydrogeologists, environmental and engineering geologists, and geological engineers working in site selection, feasibility studies, investigations, design or construction of civil engineering projects or in waste management, groundwater, and related environmental fields. All papers are peer reviewed. The editors invite contributions concerning all aspects of environmental and engineering geology and related disciplines. Recent abstracts can be viewed under “Archive” at the web site, “http://eeg.geoscienceworld.org”. Articles that report on research, case histories and new methods, and book reviews are welcome. Discussion papers, which are critiques of printed articles and are technical in nature, may be published with replies from the original author(s). Discussion papers and replies should be concise. To submit a manuscript go to http://eeg.allentrack.net. If you have not used the system before, follow the link at the bottom of the page that says New users should register for an account. Choose your own login and password. Further instructions will be available upon logging into the system. Please carefully read the “Instructions for Authors”. Authors do not pay any charge for color figures that are essential to the manuscript. Manuscripts of fewer than 10 pages may be published as Technical Notes. For further information, you may contact Dr. Abdul Shakoor at the editorial office. Cover photo Coastal bluff landslide in Quaternary glacial drift in the Dungeness littoral cell, Strait of Juan de Fuca near Port Angeles, Washington, USA. Photo Credit: David Parks; see related article on pp. 129 to 146.
Environmental & Engineering Geoscience Volume 21, Number 2, May 2015 Table of Contents
75
Sources and Changes in Groundwater Quality with Increasing Urbanization, Northeastern Illinois Hue-Hwa Hwang, Samuel V. Panno, and Keith C. Hackley
91
Sorption-Desorption Characteristics of Tetrabromobisphenol A on Humin and Sediment of Lake Chaohu, China Suwen Yang, Shengrui Wang, Binghui Zheng, Fengchang Wu, and Qiang Fu
101
Gully Erosion Mapping Using Object-Based and Pixel-Based Image Classification Methods Ayoob Karami, AsaDollah Khoorani, Ahmad Noohegar, Seyed Rashid Fallah Shamsi, and Vahid Moosavi
111
Near-Surface Geophysical Imaging of a Talus Deposit in Yosemite Valley, California Anna G. Brody, Christopher J. Pluhar, Greg M. Stock, and W. Jason Greenwood
129
Bluff Recession in the Elwha and Dungeness Littoral Cells, Washington, USA David S. Parks
147
Mine-Water Flow between Contiguous Flooded Underground Coal Mines with Hydraulically Compromised Barriers David D. M. Light and Joseph J. Donovan
Sources and Changes in Groundwater Quality with Increasing Urbanization, Northeastern Illinois HUE-HWA HWANG1 SAMUEL V. PANNO KEITH C. HACKLEY2 Illinois State Geological Survey, Prairie Research Institute, University of Illinois, Champaign, IL 61820
Key Terms: Geochemistry, Environmental Geology, Pollution, Isotope Geochemistry, Agriculture
ABSTRACT During the last decade of the twentieth century, McHenry County had the fastest-growing population in Illinois. Just north of the Chicago metropolitan area, land use in the eastern half of the county changed from row-crop agriculture to urban sprawl. Water supplies are from shallow sand and gravel aquifers and are highly vulnerable. We evaluated the change of groundwater quality in McHenry County during most of the twentieth century and identified the degree and extent of contamination, and sources, using available historic water-quality data. To evaluate historic data, we calculated background concentrations of selected ions using cumulative probability plots to identify the presence of anthropogenic contamination. Timing of groundwater contamination coincides with that of population growth and the onset of utilization of artificial N-fertilizer and road salt. Groundwater from urban areas showed greater Na+ and Cl2 contents than rural areas, which reflect more extensive applications of road salt beginning in the early 1960s. Groundwater was collected for chemical and isotope analyses from selected shallow wells with historically elevated NO32 concentration as well as from farms with livestock. The isotope data suggest N-fertilizer and soil nitrogen are the predominant sources for NO32 in shallow groundwater. Animal waste was also a source for NO32 near farms with livestock. Spatial analysis suggested that the source of NO32 in the groundwater was from surface-borne contaminants. The permeable soils and near-surface sand and gravel aquifer found in most of McHenry County provide pathways for 1
Corresponding author email: hhhwang@illinois.edu; phone: (217) 244-9876; fax: (217) 244-0424. 2 Present address: Isotech Laboratories, Inc., 1308 Parkland Ct., Champaign, IL 61821.
surface contaminants to migrate into shallow groundwater. INTRODUCTION The Chicago metropolitan area in northeastern Illinois recently has seen the most rapid increase in population and land development in the state. Kelly (2008) found that the groundwater quality in the Chicago metropolitan area has degraded since the early 1900s, and the change appeared to have been most rapid in the outlying counties. McHenry County, located on the edge of the Chicago metropolitan area (Figure 1), has experienced the fastest growth rate of any county in Illinois between 1991 and 2000 (U.S. Census Bureau, 2000). From 2001 until 2010, McHenry County ranked seventh in growth rate of all Illinois counties (U.S. Census Bureau, 2010). Because of its rapid growth over the last 20 years or so, and because its water supplies are almost entirely from groundwater, McHenry County is an excellent region to study the anthropogenic impacts on groundwater resources. The population of McHenry County increased from 35,000 in 1930 to 183,000 in 1990 and grew 42 percent between 1990 and 2000 (Hwang et al., 2007). The county population grew another 18.7 percent after 2000 and reached 308,760 in 2010 (U.S. Census Bureau, 2010). About 75 percent of its groundwater supply comes from shallow aquifers composed of sand and gravel, which are highly permeable and rapidly recharged, and are subject to surface-borne contamination (Curry et al., 1997). Approximately 13 percent of the 280 McHenry County wells in the Illinois State Water Survey (ISWS) water-quality database contained NO32 (as N) concentrations at or exceeding the U.S. Environmental Protection Agency’s (USEPA) drinking water standard of 10 mg/L (Meyer, 1998). Water-quality records from the McHenry County Health Department between 1986 and 2002 also indicated that NO32 concentrations in more than 800 wells (about 6
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
75
Hwang, Panno, and Hackley
Figure 1. Location map of the study area, McHenry County, Illinois, showing the major towns and cities of the county.
percent of the total record) were at or exceeded 10 mg/ L (Hwang et al., 2007). On a global basis, NO32 pollution in groundwater is a common problem. The most common NO32 sources in surface water and groundwater are naturally occurring atmospheric NO32, soil organic matter, septic effluent, animal waste, and synthetic and organic fertilizers (Hallberg and Keeney, 1993). Increasing applications of fertilizer and large amounts of sewage disposal since the 1960s have contributed to the amount of N loading into surface water and shallow groundwater. High NO32 levels in drinking water are hazardous to human health and have been linked to blue-baby syndrome and stomach cancer (O’Riordan and Bentham, 1993). Thus, it is important to understand the history and extent of NO32 pollution in shallow groundwater and to identify its sources.
76
The groundwater contaminant most associated with urbanization is Cl2 (Eisen and Anderson, 1979). One of the major sources for Cl2 is road salt, which is used as a de-icer in urban areas. Other sources of Cl2 include leachate from leaking landfills, septic effluent, animal waste, and basin brine seeps (Panno et al., 2005, 2006b). Other contaminants typically found in urban areas include SO422, heavy metals, and volatile organic compounds (Kelly 2008). The objectives of this investigation were to, first, evaluate the change in groundwater quality throughout the history of urban development in McHenry County, Illinois, during most of the twentieth century and the early part of the twenty-first century based on available groundwater quality data; and, second, to identify the origin of NO32 in the shallow groundwater of selected areas in McHenry County where elevated NO32 levels were detected in well-water samples.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
Changes in Groundwater Quality, Northeastern Illinois
MATERIALS AND METHODS Study Area The geology and groundwater resources of the McHenry County area have been characterized previously by researchers from the Illinois State Geological Survey and the Illinois State Water Survey (Suter et al., 1959; Csallany and Walton, 1963; Woller and Sanderson, 1976; Curry et al., 1997; and Meyer, 1998). In general, the county is covered by glacial sediments deposited during the last 730,000 years from at least three separate glacial episodes, i.e., pre-Illinois, Illinois, and Wisconsin episodes (Curry et al., 1997). The physiography of the county is referred to as the Wheaton Morainal Country and consists of a series of glacial moraines and lowlands made up of very permeable sand, or sand and gravel layers, and much less permeable diamicton layers (Horberg, 1950; Curry et al., 1997). The glacial deposits in this county are a few tens of meters up to 150 m thick and overlie bedrock composed of dolomite, limestone, and shale of the Ordovician Galena and Maquoketa Groups (Herzog et al., 1994; Curry et al., 1997). Glacially deposited sand and gravel layers comprise relatively shallow, productive aquifers that are used extensively for water resources. As a result of relatively thin, sandy soils that provide little protection to the underlying aquifers, many of the sand and gravel aquifers of this county can easily be polluted with surface-borne contaminants. Curry et al. (1997) noted that greater than 70 percent of the private and municipal wells in the county are less than 30 m deep. Where the sand and gravel deposits intersect the surface, many of the private wells are sand point wells with depths typically less than 5 m. Somewhat more deeply buried sand and gravel aquifers, generally lying beneath a sandy diamicton unit, are somewhat more protected from contamination. Even deeper, and probably even more protected are the sand and gravel aquifers that include the Pearl Formation deposited during the Illinois episode, and the pre-Illinois episode basal drift aquifer of the Banner Formation. The underlying bedrock is dolomite, which is highly fractured, and it is used as a water resource in those areas of the northeast where glacial deposits are too thin to serve as useable aquifers (Visocky et al., 1985; Curry et al., 1997).
trations greater than 10 mg/L. The water-quality database of the ISWS is based on township and range, whereas water-quality records of the MCDH are sorted by address. Computer software, ArcGIS, was used to analyze both databases to delineate the change of groundwater quality through time and in different areas of McHenry County. Drilling records stored in the Geological Record Library at the Illinois State Geological Survey were used to provide depth and stratigraphic information of the wells of interest and to make cross sections. Population data for several townships were collected and analyzed to assess the population growth rate. The land-cover map of the county (Illinois Department of Agriculture, 2000) and aerial photos were used to delineate the types of land usage. Approximately 38,000 groundwater quality records from McHenry County were retrieved from the ISWS and the MCDH. The ISWS database contains records from 1913 to 1996. The MCDH database contains records from 1986 to 2002. Merging of the two databases was not feasible because the MCDH database is based on street addresses, and the ISWS database is based on township, range, and sections. To overcome this problem, we used ArcGIS to display and analyze records from the two databases on the same map. Initially, the databases had to be edited before they could be analyzed. Specifically, erroneous records (i.e., wells located outside of McHenry County or with wrong or incomplete addresses, without depth information, and those that were not groundwater) were removed from the database. Records that did not show a definite value, such as ‘‘,1,’’ were also removed. In the ISWS database, NO32 data were reported in three different ways, as dissolved NO32, total NO32, or NO32 + nitrite (as N). All nitrate data were converted to NO32 as N for consistency. Bias in the data used in this investigation was assumed to be small given the large number of well records considered (38,000), and the culling process used (described in Hwang et al., 2007). Because more than one record per well/ location was rare and because of the very large data set used, bias within the database from multiple samples per well/location should not be an issue. In addition, the method for reporting levels for all the chemical data considered are essentially the same. Sample Collection
Water-Quality Database We initially examined the water-quality database of the ISWS and the water analysis records of the McHenry County Department of Health (MCDH) for groundwater quality analyses with NO32 concen-
We selected wells with high historical NO32 concentrations to identify the sources of NO3-N by determining the NO32-nitrogen and NO32-oxygen isotopic ratios. We collected 30 groundwater samples from private wells in Marengo-Union, Wonder Lake,
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
77
Hwang, Panno, and Hackley Table 1. Chemical composition of surface water and groundwater samples. Parameters are reported in mg/L unless otherwise indicated. Columns continue on next page. Sample ID 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Type Urban Urban Rural Rural Livestock Rural Rural Urban Rural Rural Livestock Rural Rural Urban/Crop Urban/Crop Urban Urban Urban Urban Rural Urban Urban/Crop Urban Horse manure leachate Livestock Livestock Livestock Livestock Livestock Livestock Livestock
Depth (m)
Temp. (uC)
pH
Eh (mV)
Sp. Cond. (mS/cm)
Alkalinity (CaCO3)
Na
K
Ca
Mg
Sr
Ba
20 22.5 16.5 38 17.5 20 28 16 80 110 90 88 124 60 100 100 100 30 20 61 25 53 23 0
15.9 14 14.6 11.3 12.6 15.4 12 13.4 11.5 11.2 11.2 11.1 12.3 11.2 11.3 13.5 14.4 12.2 12.4 11.2 12.1 12.3 12.3 NA
7.07 7.16 7.18 7.28 7.02 7.19 7.19 7.125 7.12 6.96 7.18 7.14 6.91 7.19 7.2 7.01 7.2 7.21 7.02 7.41 7.48 7.16 7.2 7.34
456 156 476 92 445 473 491 215 460 507 487 499 538 453 510 413 147 290 462 512 88 477 455 344
1,484 1,356 672 751 1,232 619 614 1,170 968 1,330 839 844 1,223 1,215 717 1,297 891 629 1,700 741 586 969 861 5,230
362.6 296.1 316 287 382.8 274.8 247.3 358.2 418.4 427 342.8 350.9 360.2 353.3 308.7 421.4 324.2 284.8 403.2 351 347 370 332 1620
111 53.5 10.7 16.7 23.3 4.1 7.8 99.4 37.9 92 14.3 36.4 68 58.6 3.1 204 23.8 3.2 191 6.1 15.3 42.2 34.5 109
7 4 ,1 4 119 4 9 4 ,1 4 ,1 2 2 8 5 7 6 6 7 ,5 ,5 ,5 ,5 1020
91.5 100 72.6 78.8 81.3 69.7 81 101 108 122 106 97.9 121 121 95.5 73.4 111 91.2 128 93.1 61.3 100 90.5 87
33.9 43.5 29 35.6 38.5 33 36 41.5 58.7 59.8 53 47.4 59.5 62.5 49.1 31.7 52.4 42.2 54.2 49.8 38.2 48.5 36.6 94.7
0.149 0.115 0.051 0.13 0.115 0.06 0.086 0.184 0.095 0.127 0.083 0.085 0.186 0.102 0.077 0.069 0.131 0.057 0.14 0.08 2.1 0.124 0.151 220
0.025 0.023 0.017 0.124 0.04 0.008 0.015 0.029 0.051 0.081 0.037 0.042 0.054 0.055 0.035 0.036 0.073 0.02 0.063 0.046 0.122 0.062 0.024 60
26 22 7 2.5 14 7 18.5
13.7 15.1 14.7 18.6 13.7 15.2 13.2
7.32 7.376 7.06 6.94 7.248 6.93 7.38
500 525 568 488 549 542 519
841 824 1,440 1,406 714 1,175 734
240 251 526 634 285 444 249
3.8 15.8 26.7 20.5 12 18 5.2
5 4 6 67 5 87 4
105 96.9 192 147 91.4 125 93
49.9 44.7 99.5 71.2 39 49.4 42.9
0.071 0.091 0.2 0.142 0.08 0.151 0.101
0.065 0.043 0.103 0.161 0.027 0.048 0.024
Sp. Cond 5 specific conductance; TKN 5 total Kjeldahl nitrogen; DOC 5 dissolved organic carbon.
McHenry, and near Woodstock and one manure leachate sample between December 2002 and August 2003. Cation samples were acidified in the field with ultra-pure nitric acid to a pH of less than 2. All samples were transported in ice-filled coolers to the laboratory and kept refrigerated until analysis. A horse manure leachate sample was collected to provide chemical and isotopic data as one of the nitrate sources. Sample Analysis Thirty-one collected water samples were analyzed for dissolved cations, anions, total Kjeldahl N (TKN), ammonia, D/H and 18O/16O isotopic ratios, NO32-15N, and NO32-18O analyses (Table 1). Groundwater samples from nine selected wells were also analyzed for tritium content; sample locations were selected from the shallowest wells and on the basis of
78
geographic distribution. Water samples were analyzed in the field for temperature, pH, Eh, and specific conductance with techniques described by Wood (1981). Anions and cations in the groundwater samples were analyzed at the Illinois State Geological Survey (ISGS) using atomic absorption and ion chromatography methods. Total organic carbon contents of the high-NO32 samples were analyzed at the Illinois Waste Management and Research Center. Ammonia contents were determined at the Illinois Natural History Survey using the Berthelot reaction, which involves the formation of a blue-colored indolphenol compound in a solution of ammonia salt, sodium phenoxide, and sodium hypochlorite. Following enhancement of color using sodium nitroprusside, the color intensity is measured by a Bran & Luebbe TRAACS 2000 colorimeter at 660 nm. Total Kjeldahl N (TKN) was determined at the Illinois Natural History
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
Changes in Groundwater Quality, Northeastern Illinois Table 1. Extended.
B
SiO2
HCO3
SO4
Cl
Br
F
NO3-N
NH4-N
TKN
PO4-P
Fe
Mn
DOC
,0.01 0.03 0.04 ,0.01 0.09 ,0.01 ,0.01 0.16 ,0.01 0.14 0.05 0.1 0.06 ,0.02 ,0.02 0.07 ,0.02 ,0.02 0.07 ,0.02 0.12 0.09 0.12 4.8
12.4 10.1 12.7 10.4 12.8 12.8 12.4 14 19.4 20.9 18.3 18.8 16.1 18.6 15.5 18.7 19.6 12.6 16.6 17.3 19.8 16.8 12.6 43.7
442 361 385 350 467 335 302 437 510 521 418 428 439 431 376 514 395 347 492 428 423 451 405 1975
30 35.3 18.4 ,0.1 13.7 14.3 18.4 13.7 34.8 26.2 46 28.8 53.6 41.1 64.2 48.1 76.4 56.3 42.1 36.9 1.3 34.3 28.6 2.1
219 229 15.5 8.7 135 14.9 15.5 135 59.4 169 41.6 43.9 155 167 17 170 80.4 12.7 301 19.3 2.8 83.5 52.7 440
,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 ,0.05 0.3
,0.1 ,0.1 ,0.1 ,0.1 0.2 ,0.1 ,0.1 ,0.1 0.1 0.2 0.1 0.1 0.2 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 0.4 ,0.1 ,0.1 ,0.2
7.25 0.08 2.26 ,0.02 49.4 9.61 14.6 4.55 4.39 7.53 7.21 9.12 10 12.1 7.9 6.27 0.19 4.63 14.2 6.19 ,0.02 7.7 6.3 0.19
0.01 0.03 0.04 ND 0.04 0.02 0.04 ,0.01 ,0.01 ,0.01 ,0.01 ,0.01 ,0.01 0.01 ,0.01 ,0.01 0.01 ,0.01 ,0.01 0.02 1.98 0.07 0.13 155
0.01 0.19 0.02 NA 0.04 ,0.01 20 ,0.01 ,0.01 0.19 1.35 1.04 ,0.01 0.01 ,0.01 ,0.01 0.21 ,0.01 ,0.01 0.54 2.51 0.45 0.9 256
,0.1 ,0.1 ,0.1 ,0.1 1.8 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 ,0.1 111
,0.01 0.22 ,0.01 1.09 ,0.01 ,0.01 ,0.01 0.1 ,0.01 ,0.01 ,0.01 ,0.01 ,0.01 ,0.01 ,0.01 0.03 1.21 ,0.01 ,0.01 ,0.01 2.66 ,0.01 0.08 3.68
,0.002 0.003 ,0.002 0.053 0.414 ,0.002 0.003 0.009 0.002 ,0.001 0.002 0.001 0.004 0.001 ,0.001 ,0.001 0.056 0.033 ,0.001 ,0.001 0.022 ,0.001 0.006 480
1.8 0.83 1.5 0.85 8.1 1.8 2.8 1.3 0.58 0.65 0.87 0.53 1.4 0.5 1.1 0.9 1.6 0.7 1.3 2.1 1.7 0.8 1.2 NA
,0.01 ,0.01 ,0.01 ,0.01 ,0.01 ,0.01 ,0.01
11.1 12.9 14.5 12.3 16.8 20.1 13.5
293 306 641 773 347 541 304
62.4 28 236 68.5 15.7 37.6 15.2
33 35.4 69.2 56.5 18.8 36.6 15.4
,0.1 ,0.1 0.7 0.2 ,0.1 ,0.1 ,0.1
,0.4 ,0.4 ,0.4 ,0.4 ,0.4 ,0.4 ,0.4
22.4 25.4 5.59 0.02 13 18.7 25.2
0.04 0.03 0.05 21.8 0.01 0.02 ,0.01
10.6 3.92 13.8 30 4.58 5.11 6.3
,0.2 ,0.2 ,0.2 ,0.2 ,0.2 ,0.2 ,0.2
,0.01 ,0.01 ,0.01 8.3 ,0.01 ,0.01 ,0.01
0.15 0.02 0.61 0.41 ,0.01 0.08 ,0.01
0.82 0.97 26 38 1.1 4.8 0.9
Survey using the method of Raveh and Avnemelech (1979) (Table 1). Neutron activation analysis was conducted by the Nuclear Engineering Teaching Laboratory at the University of Texas at Austin to determine concentrations of Na+, Cl2, Br2, and iodide (I2) at very low detection limits (Strellis et al., 1996; Landsberger et al., 2003) (Table 2). Because of differences in Ion Chromatograph (IC) vs. neutron activation, and the internal consistency of those data, the Cl/Br ratios were calculated from neutron activation data (Table 3). All isotope analysis was performed at the Isotope Geochemistry Laboratory of the Illinois State Geological Survey. The d18O values were determined using a modified CO2-H2O equilibration method as described in Epstein and Mayeda (1953), with modifications described in Hackley et al. (1999). The dD was determined using the Zn-reduction method described in Coleman et al. (1982) and Vennemann and O’Neil (1993), with modifications described in Hackley et al. (1999). The d13C of the dissolved
inorganic carbon (DIC) was determined using a gasevolution technique as described in Hackley et al. (2010). Analytical reproducibility for the dD, d18O, and d13C analysis is equal to or less than 61.0 per mil, 60.1 per mil, and 60.15 per mil, respectively (Table 3). Tritium was analyzed for selected samples using electrolytic enrichment (Ostlund and Dorsey, 1977) and liquid scintillation counting as described in Hackley et al. (2007). Nitrate isotopic analyses were performed at the Isotope Geochemistry Laboratory of the ISGS using an improved ion-exchange method developed by Hwang et al. (1999), which was modified from a method later published by Silva et al. (2000). Detailed procedure was described in Hwang et al. (2007). Isotope analytical results are reported in Table 3. Background Concentrations of Selected Ions In order to evaluate the data set for the presence or absence of anthropogenic contaminants, it was
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
79
Hwang, Panno, and Hackley Table 2. Halide concentrations of groundwater samples based on instrumental neutron activation analysis. Sample ID 1 Urban 2 Urban 3 Rural 4 Rural 5 Livestock 6 Rural 7 Rural 8 Urban 9 Rural 10 Rural 11 Livestock 12 Rural 13 Rural 14 Urban/crop 15 Urban/crop 16 Urban 17 Urban 18 Urban 19 Urban 20 Rural 21 Urban 22 Urban/crop 23 Urban 24 Horse manure 25 Livestock 26 Livestock 27 Livestock 28 Livestock 29 Livestock 30 Livestock 31 Livestock Precipitation{ Pristine shallow aquifer{ Septic effluent{ Animal waste{ Road salt (solution){ Road salt affected water{
Cl (mg/L)
Br (mg/L)
I (mg/L)
219* 229* 19.7 45.7 30.9 28.7 27.8 135* 12.3 169* 47.2 49.0 152 170* 16.8 167* 66.2 12.2 301* 19.4 1.42 83.5* 56.0 440 32.2 38.2 69.2* 56.5* 19.8 36.6* 15.4* — — — — — —
0.056 0.054 0.034 0.064 0.076 0.053 0.067 0.061 0.088 0.127 0.032 0.038 0.069 0.076 0.038 0.0143 0.046 0.027 0.151 0.031 0.021 0.050 0.047 0.739 0.058 0.064 0.152 0.201 0.031 0.084 ND — — — — — —
0.0026 0.0023 0.0007 0.0028 0.0260 0.0015 0.0027 0.0033 0.0009 0.0042 0.0010 0.0013 0.0023 ,0.004 0.0008 ,0.004 0.0028 0.0009 0.0110 0.0007 0.0029 0.0017 0.0051 0.1824 0.0007 0.0008 0.0126 0.0351 0.0017 0.0219 ND — — — — — —
Cl/Br Ratio 3,910 4,241 579 714 406 542 415 2,213 140 1,331 1,475 1,289 2,202 2,237 442 11,678 1,439 452 1,993 626 67.6 1,680 1,192 595 555 597 454 279 639 440 ND 20–56 (mean 23–521 (mean 65–5,404 (mean 1,245–1,654 (mean 13,497 1,164–4,225 (mean
5 5 5 5
42.6) 156) 1,164) 1,422)
5 2,340)
ND 5 not determined. *Cl concentrations determined by IC. { Data from Panno et al. (2006b).
necessary to calculate background concentration ranges of selected ions (i.e., Na+, Cl2, K+, and NO32). Background refers to pre-settlement cation and anion concentrations in groundwater that are naturally present from rock-water interaction and input from natural flora and fauna. Specifically, pristine groundwater contains no anthropogenic contaminants. There are several means by which background concentrations of ions in groundwater may be determined; these include evaluation of historic data, data from pristine areas, comparison of ion concentrations with electrical conductance and alkalinity, and cumulative probability graphs (Panno et al., 2006a, 2006b). The latter technique (cumulative probability graphs) was chosen for this investigation, and the results are presented next. The data used in these
80
calculations were collected from the ISGS database for McHenry County. In total, 380, 790, 394, and 680 wellwater samples were used for the background calculations for Na+, Cl2, K+, and NO32, respectively. The background concentration for SO422 was estimated from previous studies by the authors. RESULTS AND DISCUSSION Historical Water-Quality Data Historical groundwater quality records from both ISWS and MCDH databases were analyzed to delineate temporal and spatial trends. Temporal analysis of the database revealed that total dissolved solids, Cl2, and NO32 concentrations in groundwater
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
Changes in Groundwater Quality, Northeastern Illinois Table 3. Isotope data (units: per mil for stable isotopes, TU for tritium). Sample ID
dD
d18O
d13C (HCO32)
d18O (NO32)
d15N (NO32)
d15N (NH4+)
Tritium
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
254.5 255.9 261.4 ND 251.4 259.3 250.1 253.2 253.9 255.9 258.1 260.4 256.9 256.9 256.5 257.6 257.1 256.2 ND 257.1 253.8 255.9 250.9 228.3 249.0 251.3 245.6 247.6 250.9 242.6 248.7
28.35 28.45 29.10 ND 27.80 28.98 27.78 28.17 28.33 28.65 28.78 29.04 28.60 28.86 28.90 28.50 28.46 28.47 28.33 28.84 28.39 28.75 28.00 21.99 28.57 28.09 27.06 27.17 27.89 26.74 26.70
212.27 212.78 211.36 ND 210.13 213.08 28.63 213.98 210.12 213.32 211.69 211.92 210.62 210.40 29.96 213.31 211.00 29.03 213.08 212.08 25.22 211.37 212.39 25.96 25.04 24.73 211.63 211.46 28.87 216.14 23.82
8.1 ND 7.7 ND 7.8 5.3 5.6 4.1 8.2 5.8 7.8 6.2 8.7 7.7 8.9 6.0 ND 13.5 6.1 5.0 ND 4.2 5.0 ND 10.1 7.5 16.7 ND 5.2 8.3 6.8
7.5 ND 6.0 ND 7.8 4.2 3.3 8.6 7.7 8.9 6.1 5.9 7.6 5.1 5.7 9.0 ND 22.9 8.0 3.8 ND 5.1 10.4 ND 11.8 6.9 40.1 ND 2.7 12.9 3.4
ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND 23.2 ND ND 12.5 ND ND ND 7.7 ND ND ND
8.29 6.66 ND ND 8.73 ND ND 7.06 ND ND ND ND ND 6.51 9.69 9.67 16.29 8.57 ND ND ND ND ND ND ND ND ND ND ND ND ND
ND 5 not determined, mostly due to concentration that was too low to be analyzed for isotopic ratio.
increased from the mid-1960s. Natural or background Cl2 concentrations in uncontaminated groundwater in northern Illinois are between 1 and 15 mg/L (Panno et al., 2006a). Before 1951, only 15 percent of groundwater records contained Cl2 greater than 15 mg/L, and none was above 100 mg/L (Figure 2). The percentage of records containing Cl2 greater than 15 mg/L increased to 20 percent for 1951 to 1965, 43 percent for 1966 to 1980, and 51 percent for 1981 to 1996. All Cl2 data were divided into four depth intervals between 0 and 61 m (Hwang et al., 2007). A higher percentage of samples with Cl2 . 15 mg/L was found in wells with depth 0 to 30 m (60 percent) than wells greater than 30 m (30 percent). Results of database analysis indicated Cl2 concentration in groundwater gradually increased from 1913 to 1996. The higher percentage of groundwater records with elevated Cl2 in shallower wells also suggests the source of chloride contaminants are surface-borne. In the study area, NO32 concentrations in groundwater increased from the mid-1960’s (Figure 3). This timing coincided with the period of rapid population growth in McHenry County, and the period when
synthetic fertilizers began to be widely used by farmers for growing crops in United States (Appelo and Postma, 1994). Other potential sources include natural fauna, and wastes from humans (septic effluent) and livestock (as discharge or fertilizer) (Panno et al., 2006a). Database analysis revealed that 33 percent of records with depth less than 15 m contained NO32 concentration greater than 10 mg/L. This percentage decreases to below 10 percent for depths 15 to 30 m, and 2 percent for depths greater than 60 m. Such an inverse correlation between NO3N concentrations and depth suggests a surface-borne contaminant (Hwang et al., 2007). The distribution of elevated NO32 concentrations on a land-cover map (Illinois Department of Agriculture, 2000) revealed a correlation between elevated NO3-N concentration ($10 mg/L) and areas of cropland (Hwang et al., 2007). This correlation suggests that nitrogen compounds applied or produced in association with agricultural activities may be the major sources of NO32 in shallow groundwater for McHenry County. In some cases, elevated NO32 was also found in proximity to lakes and rivers, which is probably due
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
81
Hwang, Panno, and Hackley
Figure 2. Historical record of Cl2 concentrations in groundwater samples from private wells of McHenry County from 1913 to 1996 (data from water-quality databases of Illinois State Water Survey and McHenry County Department of Health).
to movement of groundwater toward discharge points such as lakes and streams. Chemical Composition of Groundwater In general, groundwater samples collected for this study were calcite-saturated, Ca-Mg-HCO3–type water (Hwang et al., 2007). All samples had relatively high alkalinity values, typically between 300 and 400 mg/L as CaCO3 with a circum-neutral pH. The aquifers in the county are open, well-oxygenated systems with Eh values typically between +475 and +500 mV; an open, rapidly recharging system is also
82
supported by modern tritium concentrations in selected well-water samples (Table 3). Consequently, Fe and Mn concentrations were usually below detection limits (,0.01 mg/L). The values/concentrations of all of these parameters are what would be expected from rock-water interactions in an open system containing carbonate minerals. Potential Contaminant Sources Illinois applies on the order of 2,564 kg of NaCl/ km2/yr as road de-icer, and most of that is applied in conjunction with snow plowing and in the northern
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
Changes in Groundwater Quality, Northeastern Illinois
Figure 3. Historical record of NO3-N concentrations in groundwater samples from private wells of McHenry County from 1913 to 1996 (data from water-quality databases of Illinois State Water Survey and McHenry County Department of Health).
half of the state (Panno et al., 2005). Sodium chloride is applied to major roadways and to a grid of roadways in urban areas. Table 4 displays background concentration ranges of selected ions (i.e., Na+, Cl2, K+, NO32, and SO422) calculated using cumulative probability plots (Figure 4) compared with ionic concentration of collected water samples from different environments in McHenry County. Chloride concentrations in strictly rural and urban areas of McHenry County had ranges of 15.5 to 43.2 mg/L and 41.6 to 271 mg/L, respectively. Examination of the halide ratios (Cl/Br) of private well-water samples collected during this investigation
(Table 2), based on plots developed by Panno et al. (2006a), revealed that the source of their salinity was dominantly road salt. Panno et al. (2006a) found that the Cl/Br ratios of pristine shallow groundwater in northern and central Illinois typically ranged from 23 to 521 mg/L, with a mean around 156 mg/L. Sodium and Chloride The natural or background concentration ranges of Na+ and Cl2 in shallow groundwater of McHenry County provide a benchmark from which one may identify the presence of man-made contaminants.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
83
Hwang, Panno, and Hackley Table 4. Range of selected ionic concentration in groundwater from different environments with background range using cumulative probability plots (unit: mg/L). Component +
Na
K+
Cl2
NO32
SO422
Environment
Range
Mean
Median
Urban Rural Urban/rural* Livestock Urban Rural Urban/rural Livestock Urban Rural Urban/rural Livestock Urban Rural Urban/rural Livestock Urban Rural Urban/rural Livestock
23.8–191 3.1–53.5 3.2–99.4 3.8–26.7 2–7 4–119 4–9 4–87 41.6 to 271 15.5 to 43.2 2.8 to 164 15.4 to 69.2 0.1 to 14.2 0 to 49 0 to 12.1 0 to 25.4 26.2 to 76.4 4.8 to 57.9 1.3 to 56.3 15.2 to 236
92.6 16.8 37.3 14.8 5 27.2 6.8 25.4 152.8 24.1 72.9 37.8 6.9 13.7 5.7 15.8 42.9 28 30 66.2
80 10.7 34.5 15.8 6 4 7 5 166 21.5 59.4 35.4 7.3 9.6 5.5 18.7 42.1 24.9 34.3 37.6
Background Range 1.6–24
1.15–3.6
0.1–5.7
0.44–1.7
0.1–35{
*Near border of urban and rural areas. Estimated background range.
{
Sodium is a non-conservative ion and is derived from rainwater and snowmelt at present-day concentrations of about 0.06 mg/L (NADP, 2012). This concentration can increase with evapotranspiration (roughly 70 percent in Illinois) to about 0.2 mg/L. Ion exchange with Ca2+ in the soil zone would decrease the concentration of Na+ below what would be expected based on Cl2 concentrations. Anthropogenic sources, which includes wastes from humans (septic effluent) and livestock in rural areas, manure applied as fertilizer in rural areas, and road de-icers (NaCl) in both rural and urban areas (Panno et al., 2006a), can greatly increase the concentration of Na+ and Cl2 in groundwater. Inflection points for Na+ on the cumulative probability graph include 1.6 mg/L and 24.5 mg/L (Figure 4). The background concentration range, from ,0.1 to 1.6 mg/L, is near the lower end of the range found at Sterne’s Woods Fen located east of Crystal Lake in McHenry County by Panno et al. (1999) using the same technique (,1 to 10 mg/L). Because of the limited scale of that study, and the greater number of samples and broader range of sample locations for this investigation, we estimate pre-settlement background at between 1.6 and 24 mg/ L. We suggest that Na+ concentrations .24 (rounded to two significant figures) are an effect of urbanization and the use of road de-icers and are consistent with Na+ concentrations in well-water samples collected after 1960 identified by Hwang et al.
84
(2007). Sodium concentrations in groundwater above the inflection point of 24.5 mg/L are interpreted as an effect of sampling in both rural and urban areas. That is, groundwater in urban areas typically has a greater Na+ concentration than rural counterparts due to a greater concentration of roadways. Therefore, Na+ concentrations .24 mg/L are probably indicative of contamination by manure fertilizer, livestock, and/or road salt applied to roadways. The greatest concentrations of Na+ were found in wells sampled after 1960, when road salt was used routinely. Chloride is a conservative ion and is derived from rainwater and snowmelt at present-day concentrations of about 0.1 mg/L (NADP, 2012). This concentration can increase with evapotraspiration to about 0.33 mg/L in Illinois. Added to this is Cl2 from natural fauna and rock-water interaction in pristine areas. Anthropogenic sources, including wastes from humans (septic effluent, water softeners) and livestock in rural areas, soil amendments (KCl) and manure applied as fertilizer in rural areas, and road de-icers (NaCl) in both rural and urban areas (Panno et al., 2006a), can greatly increase the concentration of Cl2 in groundwater. Inflection points for Cl2 on the cumulative probability graph include 5.7 mg/L, 45 mg/L, and 107 mg/L (Figure 4). The lowest range of concentrations (0.1 to 5.7 mg/L) is somewhat lower than that determined by Panno et al. (2006a) (i.e., 0.1 to 15 mg/ L) using another technique. Bartow et al. (1909) found that the majority of wells screened in glacial
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
Changes in Groundwater Quality, Northeastern Illinois
Figure 4. Background concentrations of Na+ (n 5 380), K+ (n 5 262), Cl2 (n 5 755), and NO3-N (n 5 680) in the sand and gravel aquifers of McHenry County, Illinois, based on cumulative probability plots of historic and recent groundwater chemistry data.
drift in Illinois contained less than 15 mg/L of Cl2. The concentration range from .5.7 to 45 mg/L shares the same upper bound as that found in Sterne’s Woods Fen in McHenry County (Panno et al., 1999) using the cumulative probability technique. Based on the graphical results, we estimate pre-settlement background for Cl2 at between 0.1 and 5.7 mg/L. Groundwater contaminated with manure fertilizer, livestock effluent, and potash probably ranges from .5.7 to 45 mg/L. The highest range (.45 to 107 mg/L) is an effect of urbanization and the use of road de-icers and is consistent with Cl2 concentrations in well-water samples collected after 1960 (Hwang et al., 2007). Groundwater in urban areas typically has a greater Cl2 concentration than its rural counterpart due to a greater concentration of roadways. The greatest concentration of Cl2 (between 107 and 830 mg/L) was found in wells sampled between 1966 and 1996 primarily along major roadways in and around the vicinity of large towns in McHenry County.
Nitrate Nitrate is derived from rainwater and snowmelt at present-day concentrations of about 0.35 mg/L (as N) (NADP, 2012); these concentrations can increase, with evapotraspiration, to as much as 1.2 mg/L. Added to this is NO32 from natural fauna, plus anthropogenic sources, which include wastes from humans (septic effluent) and livestock, and N-based fertilizers (mostly anhydrous ammonia) and manure, applied as fertilizer, in rural areas (Panno et al., 2006b). Unlike the more conservative Cl2, NO32 is reactive and is often taken up by plants and, under reducing conditions, will undergo bacterially mediated denitrification that will convert it to nitrogen gas. Inflection points on the cumulative probability graph include 0.8 mg/L (a very large inflection point), 0.43 mg/L, 1.7 mg/L, and 22 mg/L (Figure 4). The initial range of between 0.01 and 0.08 mg/L is probably very dilute groundwater, reflecting NO32
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
85
Hwang, Panno, and Hackley
concentrations of rainwater and snowmelt. The range between .0.08 and 0.43 mg/L is probably indicative of evaporative/evapotranspiration concentration of NO32. This is consistent with work by Bartow et al (1909), who found that the majority of wells in the glacial sediment in Illinois at the end of the nineteenth century contained less than 0.4 mg/L NO32. The range between 0.44 mg/L and 1.7 mg/L reflects present-day background concentrations in areas not impacted by modern agriculture (Table 4), and the range is similar to that determined by Panno et al. (2006b) for NO32 in a southwestern Illinois sinkhole plain of 0.1 to 2.1 mg/L. Nitrate-N concentrations exceeding 1.7 mg/L probably reflect the effects of application of N-fertilizer, which became popular after 1960 (e.g., Panno et al., 2006b). This is consistent with Hwang et al. (2007), who identified a steady increase in NO32 in the rural areas of McHenry County after 1966. The greatest concentrations of NO32 (up to 49 mg/L) are found in wells less than 10 m deep from rural areas of McHenry County after 1970, which are probably associated with livestock effluent. Potassium Potassium is a naturally occurring ion in groundwater and may be derived from chemical weathering of K-rich feldspars and micas during rock-water interaction (Hem, 1985), all of which are present in the sand and gravel aquifer materials in northern Illinois (e.g., Hackley et al., 2010). Potassium in rainwater is typically very low in concentration, on the order of 0.02 mg/L in Illinois (NADP, 2012). Because K+ is efficiently sequestered by plants as a nutrient and tends to be reincorporated into clay minerals (e.g., illite), K+ concentrations are typically low in groundwater (in the low single digits). Anthropogenic sources, such as K-based fertilizers (KCl), as well as livestock and human waste, can increase the concentration of K+ to concentrations typically greater than 5 but typically less than 15 mg/ L (Panno et al., 2006a) (Table 1). Potassium concentrations in McHenry County groundwater range from 0.3 to 13 mg/L. However, two well-water samples in the ISWS database had K+ concentrations of 50 and 213 mg/L, which are a factor of 4 and 16 greater than the next highest concentration, suggesting either highly localized contamination of these wells (e.g., by potash) or a transcription error in the historic data. Neither was used in the background calculations, and their exclusion had no effect on the determination of background concentrations. Inflection points for K+ on the cumulative probability graph include 1.14 and 3.60 mg/L (Figure 4 and
86
Table 4). Such levels were observed in the row-crop– rich terrain of southwestern Illinois, where K+ concentrations ranged from ,1 to 3 mg/L in wellwater samples, and ,1 to as high as 7 mg/L in springwater samples. The greatest concentrations were found in the fall of the year (Hackley et al., 2007). Bartow et al. (1909) found that K+ concentrations exceeding 5.0 mg/L in springs and wells screened in glacial drift in Illinois were uncommon. We observed three populations of K+ concentrations: the lowest concentration range (0.1 to 1.14 mg/L) probably represents pre-settlement background concentrations; the concentration range from .1.14 to 3.6 mg/L represents present-day background concentrations; concentrations of K+ greater than 3.6 mg/L represent elevated concentrations from the application of Kbased soil amendments and discharge of livestock waste and septic effluent. The inflection point at 9.4 mg/L is an artifact of the cumulative probability plot (sparse data) and should be ignored. Sulfate Sulfate concentrations ranged from 14 to 64 mg/L but were generally between 35 and 45 mg/L; an upper background threshold for SO422 was estimated by Panno (ISGS, unpublished data) to be about 35 mg/L based on hundreds of shallow groundwater samples from throughout Illinois (Table 4). The effects of land-use changes (e.g., excavation, plowing, Nfertilizer application) can increase the SO422 concentrations as a result of the exposure and oxidation of pyrite within glacial tills, the anaerobic oxidation of pyrite in the presence of NO32 (Appelo and Postma, 1994), and the interaction of oxidation products with carbonate minerals within the aquifers and tills. Aquifer Susceptibility Because of the open nature of the sand and gravel aquifers, groundwater in McHenry County can easily be contaminated with surface-borne pollutants. An aquifer sensitivity map (Keefer, 1995) showed that the uppermost sand and gravel aquifers in many places of the McHenry County are highly susceptible to contamination by NO3-N leaching; soil leaching indices in many areas of the county were described as ‘‘very fast’’ to ‘‘fast.’’ Modern groundwater collected by this study reveals elevated concentrations of Na+, Cl2, K+, and NO32 (Table 4). Elevated Na+ and Cl2 concentrations in groundwater were encountered in both rural and urban areas, but concentrations were highest immediately adjacent to major roadways and in urban areas where there was a high density
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
Changes in Groundwater Quality, Northeastern Illinois
of roadways. Road salt was the most likely source of Na+ and Cl2 contamination based on Cl/Br ratios of groundwater with elevated Cl2 (Table 2). Sodium and Cl2 concentrations in urban areas were as high as 191 and 271 mg/L, respectively, almost 20 times that of background (Table 4). Groundwater in rural areas also had concentrations of Na+ and Cl2 well above background levels but typically at lower concentrations than in urban areas. Concentrations of K+ in McHenry County groundwater were relatively high, ranging from 0.6 to 8.2 mg/L; K+ in uncontaminated groundwater in this county ranged from ,0.6 and 3.6 mg/L, with a median concentration of 2.1 mg/L (Panno et al., 1999). The dominant source of K+ in this county is probably KCl and other K-containing soil amendments and fertilizers applied to the croplands. Nitrate concentrations from urban areas were typically elevated (0.1 to 14.2 mg/L), but concentrations were low relative to NO32 concentrations in rural areas (,0.1 to 49 mg/L; Table 4). Nitrate concentrations in the vicinity of livestock operations were as high as 25.4 mg/L. Nitrate isotope data from selected wells confirmed that the dominant source of NO32 was N-fertilizer. Elevated NO32 concentrations also correlated well with areas of greater leaching potential on an aquifer sensitivity map to nitrate leaching by Keefer (1995) (Figure 10 of Hwang et al., 2007). This correlation supports surface-borne contaminant sources for NO32. Isotopic Composition of Collected Water dD and d18O of Water Water samples collected for this study were analyzed for various isotopic ratios. The dD values of water ranged from 241.6 to 261.4 per mil, and the d18O values ranged from 26.7 to 29.1 per mil (Hwang et al., 2007). Most of the data fall on the meteoric line on a dD vs. d18O plot. The leachate sample from a horse manure pile had much higher dD and d18O values (228.3, 21.99 per mil). Since the leachate sample was collected from a small puddle next to the standing horse manure pile on the ground, higher dD and d18O values may reflect the effect of evaporation. d15N and d18O of Dissolved Nitrate 2
NO3 isotopes were examined to determine nitrate sources. The d15N values of nitrate ranged from 2.7 to 40.1 per mil, and the d18O values ranged from 4.1 to 16.7 per mil (Hwang et al., 2007). Based on the isotopic data, the predominant sources of NO32 in
Figure 5. d18O vs. d15N showing that the predominant sources of NO32 are N-fertilizer and soil organic matter, and that denitrification is actively occurring within the soil zone and/or aquifers (modified from Clark and Fritz, 1997).
the shallow groundwater samples are fertilizer and soil organic matter, despite the fact that the samples were collected from different environments, such as urban, rural, and livestock farms. Although several samples were collected near farms with livestock facilities, the only one with clear indication of manure/septic source was sample 27, which had the largest d15N (+40.1 per mil) and d18O (+16.7 per mil) values. The lack of an isotopic signature of manure for most of the livestock farm groundwater samples may be due to the widespread nature of croplands surrounding those operations, which caused the isotopic signature of manure to be diluted by that of fertilizer and soil organic nitrogen. Fertilizer application on urban lawns and parks may result in the isotopic signature of fertilizer and soil nitrogen in urban areas. A few samples showed enriched d15N and d18O values following the denitrification trajectory (Figure 5), which suggests that they have undergone various degrees of denitrification. A negative correlation was observed between d13C and d15N (Figure 6), which is consistent with the denitrification process; that is, in an anaerobic environment, micro-organisms serve as denitrifiers and reduce NO32 to oxidize organic carbon or sulfide in the following reactions (Batchelor and Lawrence, 1978; Kendall, 1998): { 4NO{ 3 z5Cz2H2 O?2N2 z4HCO3 zCO2 ð1Þ
z 14NO{ 3 z5FeS2 z4H ? 2z z2H2 O 7N2 z10SO2{ 4 z5Fe
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
ð2Þ
87
Hwang, Panno, and Hackley
and all nutrients (Table 1), and had a d15N value of +12.5 per mil (indicative of animal waste; Figure 5). Tritium Content in Water
Figure 6. d13C vs. d15N showing an inverse correlation that is consistent with microbially mediated denitrification. Two populations are visible here, one dominated by animal waste (upper trend), and a more clustered lower trend.
Denitrification reactions cause both the d15N and d18O of the residual NO32 to increase because micro-organisms preferentially consume 14N relative to 15N, and 16O relative to 18O. Denitrification through reaction 1 could also cause the d13CDIC in HCO32 to decrease because the organic carbon, which would be oxidized to form HCO32, typically has much lower d13C values. In reaction 2, while NO32 is reduced (denitrified), FeS is oxidized to form SO422, which should result in an increase in SO422 concentration. A positive correlation between SO422 concentration and d15N, as a result of denitrification process, showing N and O isotope evidence of denitrification was observed by Hwang et al. (2007). d15N of Ammonia Only three samples that contained enough ammonia were analyzed for ammonia d15N. The first sample was collected from a shallow well in which groundwater was under reducing conditions, and for which Cl2 and NH4+ (as N) concentrations were only 2.8 and 1.98 mg/L, respectively (well within background). This sample’s d15N value was 23.2 per mil and fell within the range of soil organic matter (Figure 5). The second sample was from a shallow (0.8 m deep) hand-dug well down gradient from a hog farm with elevated Cl2 and NH4+-N concentrations of 56.5 and 21.8 mg/L, respectively. The d15N value for the dug well was +7.7 per mil and within the range of animal waste. The third sample consisted of horse manure leachate and was enriched in Na+ and Cl2,
88
Tritium analyses were completed for six groundwater samples collected from depths of 6 to 30 m. The tritium measurements ranged predominantly from 6.5 to 9.7 TU, with one sample containing 16.29 TU. These tritium data imply that groundwater in the study area is relatively young, having a travel time from recharge to well depths of between less than one and 30 years. Most of the tritium values fall within or close to the range expected for recent precipitation, which is approximately 2 to 8 TU (Eberts and George, 2000; Hackley et al., 2007; and Warrier et al., 2013). The greater tritium levels measured for well site 17 may represent slightly older groundwater closer to 1960s values. Well site 17 also contained very little NO32 (0.19 mg/L), suggesting less immediate impact from surface infiltration. This well was screened in a very thin lens of sand sandwiched between a relatively thick tight till (Hwang et al., 2007), whereas all the other wells were screened in significantly thicker sand/gravel deposits, which undoubtedly have a more direct hydraulic connection to the land surface. CONCLUSIONS Temporal analysis of the groundwater quality databases revealed that Cl2 and NO32 concentrations in shallow groundwater from McHenry County have increased considerably from the mid-1960s to 2003. This time period coincides with rapid population growth in McHenry County. Database analysis also revealed higher percentage of elevated Cl2 and NO32 concentrations in wells shallower than 30 m in depth. The correlation of higher ionic concentration with shallower wells, and their relationship with calculated and estimated background concentrations of selected ions (Na+, K+, Cl2, NO32, and SO422) indicate that the sources of increased ionic concentrations were surface-borne. It is likely that Cl concentrations are the result of yearly application of road salt in urban areas. This is supported by Panno et al. (2005, 2006a), who, using Cl/Br ratios from the same samples, showed that groundwater samples collected during this present investigation were contaminated with halite. Rapid population growth in McHenry County since 1970 has resulted in expansion of urban areas and has resulted in more applications of road salt in winter seasons. Greater NO32 concentrations probably resulted from increased fertilizer use, given that extensive applications of fertilizer began in the 1960s.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
Changes in Groundwater Quality, Northeastern Illinois
Positive correlation between greater NO32 concentrations and areas of greatest leaching potential, shown on an aquifer sensitivity map, supports the hypothesis that non-point, surface-borne sources were responsible for NO32 contamination. Groundwater chemistry data from groundwater samples revealed that urban groundwater contained higher Na+ and Cl2 concentrations, and rural groundwater contained greater NO32 concentrations. Such association of specific ionic concentrations and environments illustrates the effect of land use on groundwater quality. Results of isotope analyses indicated that the predominant NO32 sources are fertilizer and soil organic nitrogen, from crop-related agricultural practices. This is to be expected for a county in which land use is dominated by croplands. Since private septic systems are common in rural areas, septic effluent may affect some of the shallow groundwater. However, from the 30 samples collected, there was no isotopic evidence of influence from septic systems. A more detailed sampling array near septic discharge systems would be needed to evaluate septic input to the shallow groundwater system. Isotope results did detect the influence of livestock manure as a source of NO32 at one location. Because there are many small-scale farms with livestock in McHenry County, its influence may be more prevalent, albeit localized, than what we observed from our data. Effects of denitrification were observed in groundwater from a few samples as indicated by the generally positive d15N and d18O trends for NO32 and the negative correlation between d13C of HCO32 and the d15N of NO32. ACKNOWLEDGMENTS This research was supported by a grant from the Illinois Groundwater Consortium under Award No. A8634 and by the Illinois State Geological Survey. The authors thank the Illinois State Water Survey and the McHenry County Department of Health for providing historical water-quality databases. Publication of this article has been authorized by the director of the Illinois State Geological Survey, Prairie Research Institute, University of Illinois. REFERENCES APPELO, C. A. AND POSTMA, D., 1994, Geochemistry, Groundwater and Pollution: A. A. Balkema Publishers, Leiden, Netherlands. BARTOW, E.; UDDEN, J. A.; PARR, S. W.; AND PALMER, G. T., 1909, The Mineral Content of Illinois Waters: Illinois State Geological Survey Bulletin 10, 1–192.
BATCHELOR, B. AND LAWRENCE, A. W., 1978, A kinetic model for autotrophic denitrification using elemental sulfur: Water Research, Vol. 12, pp. 1075–1084. CLARK, I. AND FRITZ, P., 1997, Environmental Isotopes in Hydrogeology: CRC Press, New York, 328 p. COLEMAN, M. L.; SHEPARD, T. J.; ROUSE, J. J.; AND MOORE, G. R., 1982, Reduction of water with zinc for hydrogen isotope analysis: Analytical Chemistry, Vol. 54, pp. 993–995. CSALLANY, S. AND WALTON, W. C., 1963, Yields of Shallow Dolomite Wells in Northern Illinois: Illinois State Water Survey Report of Investigation 46, 43 p. CURRY, B. B.; BERG, R. C.; AND VAIDEN, R. C., 1997, Geologic Mapping for Environmental Planning, McHenry County, Illinois: Illinois State Geological Survey Circular 559. EBERTS, S. M. AND GEORGE, L. L., 2000, Regional Ground-Water Flow and Geochemistry in the Midwestern Basins and Arches Aquifer System in Parts of Indiana, Ohio, Michigan and Illinois: U.S. Geological Survey Professional Paper 1423-C, 103 p. EISEN, C. AND ANDERSON, M., 1979, The effect of urbanization on groundwater quality—A case study: Ground Water, Vol. 17, pp. 456–462. EPSTEIN, S. AND MAYEDA, T., 1953, Variation of 18O content of waters from natural sources: Geochimica et Cosmochimica Acta, Vol. 4, pp. 213–224. HACKLEY, K. C.; LIU, C. L.; AND TRAINOR, D., 1999, Isotopic identification of the source of methane in subsurface sediments of an area surrounded by waste disposal facilities: Applied Geochemistry, Vol. 14, pp. 119–131. HACKLEY, K. C.; PANNO, S. V.; HWANG, H.-H.; AND KELLY, W. R., 2007, Groundwater Quality of Springs and Wells in the Sinkhole Plain in Southwestern Illinois: Determination of the Dominant Sources of Nitrate: Illinois State Geological Survey Circular 570, 1–39. HACKLEY, K. C.; PANNO, S. V.; AND JOHNSON, T. F., 2010, Chemical and isotopic indicators of groundwater evolution in the basal sands of a buried bedrock valley in the Midwestern United States: Implications for recharge, rock-water interactions and mixing: Geological Society of America Bulletin, Vol. 122, pp. 1047–1066. HALLBERG, G. R. AND KEENEY, D. R., 1993, Nitrate. In Alley, W. (Editor), Regional Ground-Water Quality: Van Nostrand Reinhold, New York. pp. 297–322. HEM, J. D., 1985, Study and Interpretation of the Chemical Characteristics of Natural Water: U.S. Geological Survey Water-Supply Paper 2254. HERZOG, B. L.; STIFF, B. J.; CHENOWITH, C. A.; WARNER, K. L.; SIEVERLING, J. B.; AND AVERY, C., 1994, Buried Bedrock Surface of Illinois: Illinois State Geological Survey Map 5, scale 1:500,000. HORBERG, L., 1950, Bedrock Topography in Illinois: Illinois State Geological Survey Bulletin 73, 111 p. HWANG, H.-H.; LIU, C.-L.; AND HACKLEY, K. C., 1999, Method improvement for oxygen isotope analysis in nitrates. In Geological Society of America North-Central Section Meeting, Abstracts with Programs, Vol. 31, No. 5, p. 12. HWANG, H.-H.; PANNO, S. V.; AND HACKLEY, K. C., 2007, Chemical and Isotopic Database for McHenry County Study on Groundwater Quality and Land Use: Illinois State Geological Survey Open-File Series OFS 2007-6. ILLINOIS DEPARTMENT OF AGRICULTURE, 2000, Land Cover of Illinois 1999–2000: Electronic document, available at http://www. agr.state.il.us/gis/landcover99-00.html KEEFER, D. A., 1995, Potential for Agricultural Chemical Contamination of Aquifers in Illinois: 1995 Revision: Illinois
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
89
Hwang, Panno, and Hackley State Geological Survey Environmental Geology Series 148, 28 p. KELLY, W. R., 2008, Long-term trends in chloride concentrations in shallow aquifers near Chicago: Ground Water, Vol. 46, pp. 772–781. KENDALL, C., 1998, Tracing nitrogen sources and cycling in catchment. In Kendall, C., and McDonnell, J. J. (Editors), Isotope Tracer in Catchment Hydrology: Elsevier Science B.V., Amsterdam, Netherlands, 839 p. LANDSBERGER, S.; O’KELLY, D. J.; AND PANNO, S. V., 2003, Determination of bromine, chlorine and iodine in environmental aqueous samples from epithermal neutron activation analysis and Compton suppression: Transactions of the American Nuclear Society, Vol. 89, pp. 735–736. MEYER, S. C., 1998, Ground-Water Studies for Environmental Planning, McHenry County, Illinois: Illinois State Water Survey Contract Report 630, 141 p. NADP, 2012, National Atmospheric Deposition Program: Electronic document, available at http://www.isws.illinois.edu/ hilites/nadp/ O’RIORDAN, T. AND BENTHAM, G., 1993, The politics of nitrate in the UK. In Burt, T. P.; Heathwaite, A. L.; and Trudgill, S. T. (Editors), Nitrate-Processes, Patterns and Management: John Wiley and Sons, New York. pp. 57–68. OSTLUND, H. G. AND DORSEY, H. G., 1977, Rapid electrolytic enrichment and hydrogen gas proportional counting of tritium. In Low-Radioactivity Measurements and Applications: Proceedings of the International Conference on Low-Radioactivity Measurements and Applications: October 6–10, 1975: Slvenske Pedagogike Nakladetal’stvo, Bratislava, Slovakia. pp. 95–104. PANNO, S. V.; HACKLEY, K. C.; HWANG, H.-H.; GREENBERG, S.; KRAPAC, I. G.; LANDSBERGER, S.; AND O’KELLY, D. J., 2005, Characterization and Identification of the Sources of Na-Cl in Groundwater and Surface Water, with Emphasis on the Midwestern U.S.: Illinois State Geological Survey Open-File Series 2005-1. PANNO, S. V.; HACKLEY, K. C.; HWANG, H.-H.; GREENBERG, S.; KRAPAC, I. G.; LANDSBERGER, S.; AND O’KELLY, D. J., 2006a, Characterization and identification of the sources of Na-Cl in ground water: Ground Water, Vol. 44, pp. 176–187. PANNO, S. V.; KELLY, W. R.; MARTINSEK, A.; AND HACKLEY, K. C., 2006b, Estimating background and threshold nitrate concentrations using probability graphs: Ground Water, Vol. 44, pp. 697–709.
90
PANNO, S. V.; NUZZO, V.; CARTWRIGHT, K.; HENSEL, B. R.; AND KRAPAC, I. G., 1999, Changes to the chemical composition of groundwater in a fen-wetland complex caused by urban development: Wetlands, Vol. 19, pp. 236–244. RAVEH, A. AND AVNEMELECH, Y., 1979, Total nitrogen analysis in water, soil and plant material with persulfate oxidation: Water Research, Vol. 13, pp. 911–912. SILVA, S. R.; KENDALL, C.; WILKISON, D. H.; ZEIGLER, A. C.; CHANG, C. C. Y.; AND AVANZINO, R. J., 2000, A new method for collection of nitrate from fresh water and the analysis of nitrogen and oxygen isotope ratios: Journal of Hydrology, Vol. 228, pp. 22–36. STRELLIS, D. A.; HWANG, H.-H.; ANDERSON, T. F.; AND LANDSBERGER, S., 1996, A comparative study of IC, ICP-AES and NAA measurements of chlorine, bromine and sodium in natural waters: Journal of Radioanalytical and Nuclear Chemistry, Vol. 211, pp. 473–484. SUTER, M.; BERGSTROM, R. E.; SMITH, H. F.; EMRICH, G. H.; WALTON, W. C.; AND LARSON, T. E., 1959, Preliminary Report on Ground-Water Resources of the Chicago Region, Illinois: Illinois State Water Survey Cooperative Ground-Water Report 1, 89 p. U.S. CENSUS BUREAU, 2000, Census of Population: U.S. Census Bureau, Washington, DC. U.S. CENSUS BUREAU, 2010, Census of Population: U.S. Census Bureau, Washington, DC. VENNEMANN, T. W. AND O’NEIL, J. R., 1993, A simple and inexpensive method of hydrogen isotope and water analyses of minerals and rocks based on zinc reagent: Chemical Geology, Vol. 103, pp. 227–234. VISOCKY, A. P.; SHERILL, M. G.; AND CARTWRIGHT, K., 1985, Geology, Hydrogeology, and Water Quality of the Cambrian and Ordovician Systems in Northern Illinois: Illinois State Geological Survey and Illinois State Water Survey Cooperative Groundwater Report 10. WARRIER, C. U.; BABU, M.; MANJULA, P.; AND HAMEED, A. S., 2013, Spatial and temporal variations of natural tritium in precipitation of southern India: Current Science, Vol. 105, No. 2, pp. 242–248. WOLLER, D. M. AND SANDERSON, E. W., 1976, Public Groundwater Supplies in McHenry County: Illinois State Water Survey Bulletin 60. WOOD, W. W., 1981, Guidelines for collection and field analysis of ground-water samples for selected unstable constituents. In Techniques of Water-Resources Investigations of the U.S. Geological Survey, Book 1, Chap. D2. 24 p.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 75–90
Sorption-Desorption Characteristics of Tetrabromobisphenol A on Humin and Sediment of Lake Chaohu, China SUWEN YANG1 SHENGRUI WANG BINGHUI ZHENG FENGCHANG WU State Key Laboratory of Environmental Criteria and Risk Assessment, Lake Research Center, Chinese Research Academy of Environmental Sciences, Anwai Dayangfang 8-1, Chaoyang District, Beijing 100012, China
QIANG FU Environmental Monitoring Quality Control Department, China National Environmental Monitoring Center, Anwai Dayangfang 8-2, Chaoyang District, Beijing 100012, China
Key Terms: Tetrabromobisphenol A, Sorption-Desorption, Sediment, Chaohu Lake
ABSTRACT Three components of sediments with regard to the sorption-desorption characteristics of tetrabromobisphenol A (TBBPA) in sediment water systems were investigated. Results show that the Freundlich and Langmuir model can describe the sorption behavior of TBBPA well. The calculated Cmax (maximum unit sorption quantity) values were 1.47, 2.13, and 3.65 mg/ kg for mineral group (MG), clay group (CG), and humin group (HG) sediments, respectively. HG exhibited a stronger nonlinear behavior than did CG and MG. The order of sorption capability was as follows: HG . CG . MG. Desorption capability order was the opposite. Simultaneously, it was found that precipitation was the main sorption type for TBBPA on sediment. The contribution of precipitation sorption ranged from 45 percent to 70 percent within a TBBPA concentration ranging from 0.1 to 10.0 mg/L in the supernatant. This may be attributable to anomalous changes in the compounds’ ionic activity in combination with metal cations. Sorption-desorption experiments on clay sediment were also conducted at pH levels ranging from 3 to 14 and temperatures ranging from 46C to 306C. In this regard, the sorption of TBBPA decreased as pH and temperature increased gradually. Further1
Corresponding author email: yangsw@craes.org.cn.
more, sorption and desorption reached a dynamic equilibrium at pH 11.5 and at a temperature of 306C, respectively. The release of TBBPA from sediment would be higher in summer than in the three other seasons, which may pose a potential ecological risk for aquatic life in lakes.
INTRODUCTION Tetrabromobisphenol A (TBBPA) is one of the most widely used brominated flame retardants (BFRs). Annual output of TBBPA in 2000 was 8,000 tons, which accounted for 76 percent of the total BFRs in China (Sun et al., 2008a). The demand for BFRs in China has increased by 8 percent per year recently (Shi et al., 2009). Lake Chaohu (Anhui Province) is one of the main production sites for BFRs in China (Jin et al. 2008; Xu et al. 2009). As the main BFR with respect to production and consumption, TBBPA can be released into the environment (Morris et al., 2004). Previous studies suggest that TBBPA is toxic to a variety of organisms (Darnerud, 2003), especially aquatic animals (Janer et al., 2007; Johnson-Restrepo et al., 2008). Therefore, it may pose a potential risk to the aquatic ecosystem (WHO/ICPS, 1995; Veldhoen et al., 2006; Liu and Zhou, 2008; and Nyholm et al., 2008). Scientists and governments worldwide have been committed to investigating TBBPA content in the environment as well as its movement, exposure toxicity, and metabolism in vivo in order to make appropriate regulatory recommendations (Kemmlein et al., 2003).
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 91–99
91
Yang, Wang, Zheng, Wu, and Fu Table 1. Physical and chemical properties of surface sediment in Lake Chaohu.1 Sediment Density (g/kg)
OC (%)
Sand (DW %)
Coarse Silt (DW %)
Medium Silt (DW %)
Fine Silt (DW %)
Clay (DW %)
1.1012
3.36
0–13.6
1.4–3.3
11.3–28.5
49.1–63.8
10.5–27
1
OC 5 organic carbon; DW 5 dry weight. OC was calculated by the content of the total organic carbon (TOC), which was detected by a TOC analyzer (Shimadzu TOC-5000).
TBBPA has been detected in various environmental and biota matrixes such as soil (Ravit et al. 2005; Xu et al., 2012), air (Jakobsson et al., 2002), sediment (Qu et al., 2011; Zhang et al., 2011; Feng et al., 2012), aquatic organisms (Leist et al., 2009; Yang et al., 2012; and He et al., 2013), and the human body (Cariou et al., 2008; Abdallah and Harrad, 2011; Mohamed and Abdallah, 2011; and Shi et al., 2013). The maximum TBBPA concentration in sediment from Lake Chaohu has already reached 518.3 ng/g (Yang et al., 2012), which is almost the highest value in the world. Thus, it may pose a potential danger to the aquatic ecosystem (WHO/ICPS, 1995; Nyholm et al., 2008; Debenest et al., 2010; and Yang et al., 2013). Furthermore, the amount of TBBPA sorbed in soils decreases significantly with the increase in pH from 6.0 to 9.0 (Sun et al., 2008c), and more TBBPA is then dissolved into the overlying water. This makes the study of sorption and desorption of TBBPA in sediment from Lake Chaohu very important and thus helps to identify the aquatic system risk. Sorption and desorption are important processes that control the distribution, transportation, and fate of chemicals in the aquatic environment. The extent of sorption and desorption of TBBPA on sediment directly influences its toxic effect in the aquatic ecosystem. There are two points of view around sediment organic matter (SOM) research on adsorption pollutants. One considers that SOM has been accepted as an important source of linear partition fraction (Huang et al., 1997; Xing and Pignatello, 1997; Xia and Ball, 1999; and Chiou, 2002). Another holds that the effect of SOM on sorption is limited to polar solutes as a result of their specific interactions with the limited active SOM site. Yet there is still no consistent explanation with regard to this subject (Yang et al., 2005). Significant sorption of TBBPA on three soils in the absence and presence of dissolved organic matter (DOM) has been reported from laboratory data (Sun, et al., 2008b), but an analogous study has not been conducted in lake sediment to confirm whether sediment is confirmed to be the final sink of TBBPA. Humin is the main organic mater of sediment and accounts for over 63 percent of the total organic matter in the middle and lower reaches of the Yangtze River in China (Meng et al., 2004), where Lake Chao is located.
92
In the present study humin is the main research object, acting as SOM affecting the TBBPA behavior during the adsorption and desorption process. In contrast, we observe the same process in clay without organic matter and in clay itself, and we try to clarify their respective roles, particularly the role of humin in the same pathway under different conditions. Therefore, the aim of this study was first to investigate the influence of humin, minerals, and clay from Lake Chaohu and the effects of pH and temperature on the sorption-desorption of TBBPA. Then, the rules of sorption and desorption of TBBPA in the laboratory were explored. From those results the situation related to adsorption and desorption of real sediment, which contains different percentages of organic matter components, was inferred. The conclusions of this study should be helpful to evaluate if TBBPA will be released into overlying water from sediment as well as to predict the ecological toxic risk on aquatic life under different pH and temperature conditions in natural aquatic environments. MATERIALS AND METHODS Chemicals and Materials TBBPA (4,4-isopropylidenebis (2,6-dibromophenol)) lab standard with a purity of 99.99 percent was purchased from Sigma-Aldrich, Inc. (St. Louis, MO, USA). TBBPA industrial standard was from Alfa Aesar (Beijing, China), with 97 percent purity. Acetonitrile, methanol, n-hexane, methylene dichloride, and carbon tetrachloride were all chromatographic grade from Merck Company (Shanghai, China). Humin was obtained from Perimed AB, Inc. (Stockholm, Sweden). ‘‘Surface’’ sediment samples (0–12 cm) were collected from the bottom of Lake Chaohu (31u38917.390N, 117u39934.220E) in September 2009. Gravel and plant residues were removed from the sediment samples by hand. All samples were freeze-dried and passed through a 60-mesh sieve. The traditional properties of the sediment samples are shown in Table 1. Batch Sorption and Desorption Equilibrium Experiments in Three Groups Clay, mineral, and humin were grouped, respectively, for batch sorption and desorption experiments.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 91–99
Sortion-Desorption of TBBPA
The clay group (CG) was obtained from the freezedried original sediment, passing through a 60-mesh sieve. The mineral group (MG) was made from the clay sample, with organic matter removed (Jin et al., 2008). The humin group (HG) was a mixture of 90 percent MG and 10 percent humin, which is the main organic matter of sediment in China. Generally it occupies 63– 74 percent of the total organic matter of sediment in the middle and lower reaches of the Yangtze River (Meng et al., 2004). After treatment, the organic carbon (OC) content of HG reached 6.74 percent. Sorption and desorption experiments were conducted by the batch equilibration technique in 1,000-mL beakers with Teflon film caps. 0.01 M CaCl2 and NaN3 solutions were prepared for the reactor system to maintain constant ionic strength and to inhibit microbial activity, respectively. The sediment samples of CG, MG, and HG were weighed (10 g) into the glass beaker, and 490 mL of background solution was added to each beaker. TBBPA (0.1 g) was dissolved in 200 mL of the mixture solution with water and methanol at a 10:1 (vol.:vol.) concentration to configure a 500 mg/L TBBPA stock solution. In the preparation of different concentrations of TBBPA the concentrations of methanol were controlled to lower than 0.05 percent in order to avoid the co-solvent effects. Then six levels of initial solutions ranging from 0.1 to 10 mg/L were added to the beakers. After kinetic experiments 12 hours was identified as the adsorption balance time. The beakers were shaken at 150 rpm for 12 hours at 25 6 0.5uC and centrifuged for 20 minutes at 4,000 rpm. One milliliter of the supernatant and 5 g sediment were removed into the sampling vial for pretreatment and further high-performance liquid chromatography (HPLC) analysis. The controls containing solutes without sediment were also conducted to evaluate TBBPA loss. Results showed that the loss of TBBPA is less than 1 percent, which is negligible. Desorption experiments were conducted after the completion of the sorption experiments, then the mixtures were centrifuged. The supernatant was discarded. The sorbents (three groups) were washed in deionized water three times to remove surface precipitation. After that, 500 mL of fresh background solution was added to the beakers, which were oscillated continuously for the same period. After being shaken and centrifuged, the supernatant and sorbent samples were taken for pre-treatment and analysis. Effect of pH and Temperature on Sorption and Desorption of TBBPA The sorption and desorption experiments were also conducted at six different temperature points ranging from 4uC to 30uC and at seven pH levels, ranging
from 3 to 14 on the sediment of the CG, according to the similar procedure used for the batch sorptiondesorption experiments. Temperature was controlled in a temperature-controlled shaker. pH was adjusted with 1 M HCI and 1 M NaOH. After being shaken and centrifuged, the supernatant and sorbent were determined by HPLC. Sample Pre-treatment and Analytical Technique The pre-treatment of supernatant samples was carried out by the liquid-liquid extraction method. Supernatant samples were passed through a 0.45-mm membrane filter, and 6 M HCl was used to adjust the sample to a pH of 2.0. Five hundred milliliters of this liquid was put into a 1,000-mL tap funnel, and 15, 15, and 10 mL of CH2C12 were added at different time intervals. After being mixed, the solution was allowed to stand until layers were formed. The CH2Cl2 liquid at the lower layer was taken. After being extracted, the liquids obtained through this method were combined and concentrated to about 0.5 mL with a rotary evaporator. The sample was dried with a nitrogen blower. Methanol was added to 1 mL, and the sample was stored at 4uC for further chromatographic analysis. ASE 300 (Accelerated Solvent Extractor ASE 300, DINEX, Inc., USA) was used to execute the pretreatment of sediment samples. The procedure is as follows: sediment samples of three groups were put into the extraction pool of the ASE with 34 mL of n-hexane and methylene dichloride solvent (4:1 vol./ vol.). All extracted solutions were collected, concentrated, and purified by a bonded C18 reverse-phase silica gel solid phase extraction column for HPLC analysis. TBBPA determination was performed by HPLC (Agilent 1200) using ultraviolet detection and isocratic elution (Sun et al., 2008c). TBBPA showed linearity with a correlation coefficient of 0.9993, ranging from 80 to 2,000 ng/mL. The mean relative standard deviation was less than 10.0 percent. Data Analysis TBBPA sorption thermodynamics were described in Freundlich and Langmuir isotherm formulas (Azizian et al., 2007; Mittal et al., 2007). The Freundlich isotherm formula was expressed as S~Kf Ce 1=n ,
ð1Þ
where S is the organic chemical concentration absorbed by a solid substance (mg/kg); Kf and n are the
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 91–99
93
Yang, Wang, Zheng, Wu, and Fu
The OC value of different types of sediment is obtained from Table 1 and Figure 1. RESULTS AND DISCUSSION Sorption and Desorption of TBBPA on the Three Sediment Groups
Figure 1. Sorption isotherms of TBBPA on the three groups. HG was the humin group; OC 5 6.74 percent. CG was the clay group; OC 5 3.63 percent. MG was the mineral group.
Freundlich sorption coefficients; and Ce is the equilibrium concentration of organic matter in liquid phase at the sorption equilibrium (mg/L). The Langmuir isotherm formula is expressed as ð2Þ
Qe ~KCe Cmax =(1zKCe ),
where Qe is the TBBPA concentration absorbed by sediment (mg/kg); Ce is the TBBPA equilibrium concentration in liquid phase at sorption equilibrium (mg/L); Cmax is the maximum unit sorption quantity (mg/kg); and K is the sorption coefficient (per g). The apparent sorption quantity Ca can be calculated from the equilibrium concentration Ce and the initial concentration C0 (mg/L) according to ð3Þ
Ca ~(C0 {Ce )V =m,
where V is the total volume of sorption solution (L) and m is the mass of sorbent added to the solution (g). The standardized OC partition coefficient, KOC, can be calculated from the Kf and OC content as follows: ð4Þ
KOC ~Kf =OC|100:
The Langmuir sorption isotherms of TBBPA on three groups of sediment are shown in Figure 1. The order of TBBPA equilibrium sorption quantity was HG . CG . MG. The isotherm of MG was more near to linear within the entire range of concentrations, indicating that the sorption partition of TBBPA between minerals and surface water had a fixed partition coefficient. As the mineral surface had some hydroxyl groups, linear sorption may occur between TBBPA and the polar surfaces through hydration functions (Sun et al., 2008c). The Langmuir sorption isotherms of HG and CG were L-type, and the sorption was unimolecular and nonlinear within the TBBPA concentrations ranging from 0.1 to 2.0 mg/L, but linear from 2.0 to 10.0 mg/L. It was found that CG and HG have different sorption modes at 2 mg/L. Solubility of TBBPA in water was about 2 mg/L at room temperature at pH 7.0. When the TBBPA concentration was lower than 2 mg/L, sorption by CG and HG was nonlinear, but it was linear at higher concentrations. The maximum unit sorption quantity Cmax on CG, HG, and MG was in the range of 8.53 to 10.86 mg/kg, which was a bit lower than that of 24 mg/kg on fluvo-aquic soil reported by Sun et al. (2008c), which was two- to threefold higher than that of the three groups in this research. This difference may be attributable to the physical and chemical properties of sediment or to a lack of consideration of precipitation. TBBPA adsorption behavior also can fit a Freundlich model where the value of 1/n is close to 0.5. In the HG group it was 0.467, which was lower than 0.5 (Table 2), indicating that TBBPA is easily adsorbed by the three types of sediment. The sorption capacity order based on the 1/n value was HG . CG . MG as well. This order is exactly consistent with the organic matter content of the three treatment groups in
Table 2. Fitting parameters of Freundlich and Langmuir sorption isothermal models. Freundlich
Clay Humin Mineral
94
Langmuir 2
Kf
1/n
R
Koc
Cmax (mg/kg)
K
R2
2.13 3.65 1.47
0.573 0.467 0.622
0.998 0.996 0.968
6,339 5,415 —
8.56 10.86 8.53
0.407 0.642 0.230
0.943 0.941 0.989
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 91–99
Sortion-Desorption of TBBPA
descending order, for HG as 6.74 percent, and for CG as 3.36 percent. From the adsorption results it was found that nonlinear extent and quantities of sorption were improved with increasing organic matter content, while the KOC value (Table 2) between CG and HG has only a 15 percent difference. That means that correlation of KOC between sediment/soil and humin is not significant, indicating that sorption of TBBPA is related to humic acid (HA) but also to other factors (He et al., 2005). This is consistent with previous research (Yang et al., 2005). Related research (Sun et al., 2008b) showed that the sorption isotherms of TBBPA on three soils were linear. However, nonlinear curves emerged when DOM was added to the reactor system, in which the organic matter content was estimated to be within the range of 6 percent to 10 percent of the total sorbent. These values indicated that the SOM may be a predominant cause for nonlinear sorption of TBBPA in sediment. The unit maximum sorption quantity Cmax was calculated from the Langmuir model based on the apparent sorption quantity that did not contain the amount of precipitation. TBBPA concentrations of the three groups after washing with deionized water were also detected. The results were very different from those for apparent sorption quantity. The main difference comes from the different precipitation amount that is a part of the adsorption quantity (Javert and Heath, 1991). In this study it was physical adsorption because it could be washed off by water. From the experimental results it was found that within the TBBPA solution concentration ranging from 0.1 to 10.0 mg/L; the mean percentages of TBBPA precipitation in the three groups were 69 percent for MG, 45 percent for CG, and 70 percent for HG. As shown in Figures 1 and 2, from 0.1 to 1.0 mg/L of TBBPA the apparent sorption quantities of MG and HG were close to the sorption quantities in the sediment after washing; that is to say, the precipitation was little. However, the apparent sorption quantity of CG was twofold higher than the mean sorption quantity after washing. Within this range, the order of apparent sorption quantities of the three groups was HG . CG . MG. The order of sorption quantities after washing precipitation on sediment was HG . MG . CG. Thus, TBBPA precipitation quantity in CG was higher than that in the other two groups. At 2 mg/L, TBBPA concentrations of HG and MG after washing suddenly decreased, whereas that of CG increased, while their precipitation was higher than lower concentrations. The order of TBBPA concentrations of the three groups after washing was CG . HG . MG, and they were only 70 percent, 31 percent, and 37 percent of
their apparent sorption quantities, respectively. From 2.0 to 10.0 mg/L, the sorption quantity of MG after washing increased linearly, and the mean sorption and precipitation quantities were 33 percent and 66 percent of the apparent sorption quantity, respectively. The sorption quantity of HG after washing reached a maximum at 5 mg/L, which was close to its apparent sorption quantity. At 10 mg/L, the sorption quantity on CG and HG after washing decreased to 18 percent and 33 percent of their apparent sorption concentration, while the precipitation occupied 82 percent and 67 percent, respectively. In general, as an ionic organic compound, TBBPA’s precipitation mechanism in the three groups was quite different from low concentration to a higher one. The sorption process of ionic organic compounds is closely related to their concentration of metal cations, which have a combining activity with ionic organic compounds, leading to precipitation (Chiou, 2002). When precipitation occurs, the aqueous calcium and sodium concentrations decrease. Sun et al. (2008c) also found that ionic strength had a strong effect on the sorption quantity of TBBPA. The ionic strengths in this study were 0.01 M CaCl2, which may be the predominant mechanism of TBBPA deposited on the surface of sediments. In the analysis of adsorption of non-polar solutes on sediment mineral, the competitive capability on mineral surface between water and polar solutes is considered weaker than that between water and non-polar solutes (Chiou, 1995, 2002). However, a significant effect between aqueous solution and TBBPA has occurred, and it might be attributable to the effect of cation that causes the shape of precipitation in this research. In the desorption experiments, it was significant that the desorption ability of TBBPA was much lower than its sorption ability. The mean desorption quantities for MG, HG, and CG were 2.9 percent, 0.3 percent, and 0.5 percent of the max unit sorption quantity, respectively. For each group after washing, the apparent desorption on MG was far greater than that on CG and HG (Figure 2). The results showed that the TBBPA in MG was more easily desorbed. The TBBPA desorption capacities of HG and CG were similar and were one order of magnitude less than the desorption quantity of MG. Effect of pH on Sorption and Desorption of TBBPA The CG of Lake Chaohu was selected to evaluate the sorption law of TBBPA on surface sediment in the pH range from 3 to 14 (Figure 3). As pH increased, the sorption quantities in sediment tended to decrease gradually. Particularly at pH levels of 3 and 7, there was a sharp decrease of sorption quantity in the
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 91–99
95
Yang, Wang, Zheng, Wu, and Fu
Figure 2. The percentage of precipitation, sorption quantity, and desorption quantity of the three groups of sediment before and after washing. The horizontal coordinates were original TBBPA concentration levels added to the supernatant solutions. The total precipitation and sorption percentage after washing on the three groups were 100 percent, respectively. DCSS was the desorption concentration in the supernatant. SCTW was the sorption concentration on the three groups after washing (parts of physical adsorption and parts of organic matter partition). PP was the precipitation of TBBPA on the surface of the sorbent (for the three groups).
sediment. Sorption quantities at pH 3 and 7 are 32.39 mg/kg and 8.56 mg/kg, respectively. Under alkaline conditions ranging from pH levels of 8 to 14, the sorption amount of TBBPA on sediment gradually decreased from 7.54 mg/kg to 0.45 mg/kg.
Meanwhile, the TBBPA concentrations of the supernatants increased. The sorption quantity at pH 3 was 72 times higher than that at pH 14. In the desorption experiments, it was found that TBBPA only has a little desorption. TBBPA concentration in solution increased slowly from a pH level of 3 (0.012 mg/kg) and reached its highest point at a pH level of 14 (0.97 mg/kg). It was calculated that the percentages of desorption quantity in the supernatant were 0.04 percent, 0.3 percent, and 215.7 percent of those in the sediment at pH levels of 3, 7, and 14, respectively. Results show that the sorption quantity of TBBPA in sediment was almost equal to the desorption quantity in supernatant at a pH 5 11.5. TBBPA is an ionic compound with two phenolic hydroxyl groups, each with two bromine atoms. The pK1 and pK2 are 7.5 and 8.5, respectively (WHO/ ICPS, 1995). It is slightly acidic and can be easily sullied and dissolved in water under alkaline conditions. At lower pH levels, TBBPA exists mainly in the molecular form, so physical sorption easily occurs on the sediment surfaces with large specific surface areas. The surface of the sediment in Lake Chaohu is a hydrous oxide–type ledikite/turface surface (Jin, 1995). It can attract ions from or release ions into solution. This is mainly caused by proton dissociation and association in the exposed OH2 groups. H+ dissociation and association depends on the H+ activity in solution and the concentration of the solution. Lower solution pH and an increasing number of positive charges lead to greater amounts of TBBPA absorbed through electrostatic interaction. Inversely, at higher pH, the negative charge concen-
Figure 3. The sorption-desorption of TBBPA on sediment from Lake Chaohu at different pH levels and temperatures. TBBPA concentration (vertical axes) is the concentration of TBBPA adsorption in the sediment and the concentration of TBBPA desorption in the overlying solution (apparent sorption quantity).
96
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 91–99
Sortion-Desorption of TBBPA
tration at the sediment surface is higher, and in this situation it retains ionic TBBPA and reacts with metal cation so as to form salt or precipitation, thus sharply decreasing the non-sorption quantity. Bisphenol A has analogous molecular structure and physicalchemical properties, which create similar sorption and desorption behavior under different pH conditions (Zeng et al., 2006; Li et al., 2007; and Sun et al., 2008c). Generally, the content of cations is high in natural lakes (Jin, 1995). From the results of the pHdependent sorption and desorption procedure, it is inferred that TBBPA can form precipitation combined with suspended solids or sinking to the surface of sediment when pH is in the neutral or acidic range. For the minor desorption quantity at those situations, the TBBPA in the sediment may pose little ecological risk in lakes. However, it may pose an apparent ecological risk in some alkaline lakes, where pH is always close to 10 (Jiang et al., 1988; He et al., 1996), or in a special evolution stage. For instance, the pH can get to 8 or 9 in the period of algae bloom bursting (Cui et al., 2008; Jia et al., 2011). Effect of Temperature on Sorption and Desorption of TBBPA The TBBPA sorption kinetics at various temperatures are shown in Figure 3. As temperature increased from 4uC to 30uC, the TBBPA concentration in the supernatant gradually increased and the sorption quantity of sediment decreased, both linearly. The rate of increase of TBBPA concentration in the supernatant was slightly faster than that of the decrease of sorption quantity in the sediment, which was 330 mg/kg-at 30uC, accounting for 9.2 percent of that at 4uC. As TBBPA is an ionic compound, temperature has a major influence on the sorption of polar organic compounds by the sediment, and sorption should be a process of heat release, as increased temperature both improves solubility in water and decreases sorption on the sediment (Chiou, 2002). Desorption of TBBPA on the sediment at different temperatures is shown in Figure 3. The sorption quantity on sediment rapidly decreased with increasing temperature, whereas that of the supernatant slowly increased. The desorption quantity at 4uC was 11.1 percent of that at 30uC, where it was almost equal to the sorption quantity on sediment. In general, the water solubility of organic compounds could be improved with the rising temperature, yet its sorption quantity would be decreased (Liu and Ji, 1996). Being an exothermic process, the sorption quantity keeps decreasing with decreasing temperature (Kozak et al., 1983).
CONCLUSION Sorption and desorption behaviors of TBBPA in three types of sediment were investigated in this study. The results show that the Freundlich model can describe the sorption behavior of TBBPA well. The magnitude of the sorption capability was as follows: HG . CG . MG. Desorption capability of the sort was the opposite. It was found that precipitation sorption was the main sorption type of TBBPA on sediment. Its percentage ranged from 45 percent to 70 percent of the total sorption. In sorption-desorption experiments on clay sediment it was indicated that the sorption of TBBPA decreased with increasing solution pH and temperature, ranging from 3 to 14 and 4uC to 30uC, respectively. Moreover, sorption and desorption reached a dynamic equilibrium at a pH level of 11.5 and at a temperature of 30uC, respectively. This shows that the release of TBBPA from the sediment is higher in summer than in the other three seasons in Lake Chaohu. The results provide a better understanding of the transportation and potential ecological risk of TBBPA for aquatic life in lakes. ACKNOWLEDGMENTS This study was financially supported by the State Major Water Project (SMWP, 2012zx07503-003). The authors thank the SMWP as well as the members of the project steering group. We also thank three anonymous reviewers for reviewing the manuscript and for their helpful comments. REFERENCES ABDALLAH, M. A.; AND HARRAD, S., 2011, TetrabromobisphenolA, hexabromocyclododecane and its degradation products in UK human milk: Relationship to external exposure: Environment International, Vol. 37, No. 2, pp. 443–448. AZIZIAN, S.; HAERIFAR, M.; AND BASIRI-PARSA, J., 2007, Extended geometric method: A simple approach to derive adsorption rate constants of Langmuir–Freundlich kinetics: Chemosphere, Vol. 68, No. 11, pp. 2040–2046. CARIOU, R.; ANTIGNAC, J. P.; ZALKO, D.; BERREBI, A.; CRAVEDI, J. P.; MAUME, D.; MARCHAND, P.; MONTEAU, F.; RIU, A.; ANDRE, F.; AND LE BIZEC, B., 2008, Exposure assessment of French women and their newborns to tetrabromobisphenolA: Occurrence measurements in maternal adipose tissue, serum, breast milk and cord serum: Chemosphere, Vol. 73, No. 7, pp. 1036–1041. CHIOU, C. T., 1995, Comment on: ‘Thermodynamics of organic chemical partition in soils’: Environmental Science Technology, Vol. 29, No. 5, pp. 1421–1423. CHIOU, C. T., 2002, Partition and Adsorption of Organic Contaminants in Environmental Systems: Wiley-Interscience, Hoboken, NJ. 213 p. CUI, F.; HUANG, Z.; LIU, Z.; AND FU, M., 2008, Relationship among chlorophyll a, dissolved oxygen and pH in the period of algae bloom: Waste Water Engineering, Vol. 34, pp. 177–178.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 91–99
97
Yang, Wang, Zheng, Wu, and Fu DARNERUD, P. O., 2003, Toxic effects of brominated flame retardants in man and in wildlife: Environment International, Vol. 29, No. 6, pp. 841–853. DEBENEST, T.; GAGNE´, F.; PETIT, A. N.; ANDRE´, C.; KOHLI, M.; AND BLAISE, C., 2010, Ecotoxicity of a brominated flame retardant (tetrabromobisphenol A) and its derivatives to aquatic organisms: Comparative Biochemistry Physiology Part C: Toxicology Pharmacology, Vol. 152, No. 4, pp. 407–412. FENG, A. H.; CHEN, S. J.; CHEN, M. Y.; HE, M. J.; LUO, X. J.; AND MAI, B. X., 2012, Hexabromocyclododecane (HBCD) and tetrabromobisphenol A (TBBPA) in riverine and estuarine sediments of the Pearl River Delta in southern China, with emphasis on spatial variability in diastereoisomer- and enantiomer-specific distribution of HBCD: Marine Pollution Bulletin, Vol. 64, No. 5, pp. 919–925. HE, M. C.; DENG, H. Z.; LI, X. R.; AND YING, Y., 2005, Sorption of phenanthrene by humic acids from different soils: Journal Environmental Sciences, Vol. 25, No. 7, pp. 982–988. HE, M. J.; LUO, X. J.; YU, L. H.; WU, J. P.; CHEN, S. J.; AND MAI, B. X., 2013, Diasteroisomer and enantiomer-specific profiles of hexabromocyclododecane and tetrabromobisphenol A in an aquatic environment in a highly industrialized area, South China: Vertical profile, phase partition, and bioaccumulation: Environmental Pollution, Vol. 179, pp. 105–110. HE, Z.; JIANG, H.; AND BI, F., 1996, Restudies on the hydrochemistry and hydrobiology of Dali Lake: Journal Dalian Fish College, Vol. 11, No. 2, pp. 1–13. (In Chinese.) HUANG, W.; YOUNG, T.; SCHLAUTMAN, M. A.; YU, H.; AND WEBER, W. J., Jr., 1997, A distributed reactivity model for sorption by soils and sediments. 9. General isotherm nonlinearity and applicability of the dual reactive Domin model: Environmental Science Technology, Vol. 31, No. 6, pp. 1703–1710. JAKOBSSON, K.; THURESSON, K.; RYLANDER, L.; SJO¨DIN, A.; ˚ ., 2002, Exposure to polyHAGMAR, L.; AND BERGMAN, A brominated diphenyl ethers and tetrabromobisphenol A among computer technicians: Chemosphere, Vol. 46, No. 5, pp. 709–716. JANER, G.; HAKKERT, B. C.; SLOB, W.; VERMEIRE, T.; AND PIERSMA, A. H., 2007, A retrospective analysis of the two-generation study: What is the added value of the second generation?: Reproductive Toxicology, Vol. 24, No. 1, pp. 97–102. JAVERT, C. T. AND HEATH, J. K., 1991, Sediment- and saturated-soilassociated reactions involving an anionic surfactant (dodecylsulfate). 1. Precipitation and Micelle Formation: Environmental Science Technology, Vol. 25, No. 6, pp. 1031–1038. JIA, X.; SHI, D.; SHI, M.; LI, R.; SONG, L.; FANG, H.; YU, G.; LI, X.; AND DU, G., 2011, Formation of cyanobacterial blooms in Lake Chaohu and the photosynthesis of dominant species hypothesis: Acta Ecology Sinica, Vol. 31, No. 11, pp. 2969–2977. JIANG, D.; ZHAI, P.; MA, P.; DUAN, R.; YANG, Y.; AND CHEN, Y., 1988, Changes of contains for elements in water and sediment of Chenghai, a lake of Yunnan Plateau, in dry and rainy seasons and group of actinomycetes in it: Journal Yunnan University, Vol. 10, No. 2, pp. 153–159. JIN, X., 1995, Lake in China Research of their Environment: China Ocean Press, Beijing, China, 318 p. JIN, X.; WANG, S.; ZHAO, H.; ZHOU, X.; AND WU, F., 2008, Effect of organic matter on DOM sorption on lake sediments: Environment Geology, Vol. 56, No. 2, pp. 391–398. JOHNSON-RESTREPO, B.; ADAMS, D. H.; AND KANNAN, K., 2008, Tetrabromobisphenol A (TBBPA) and hexabromocyclododecanes (HBCDs) in tissues of humans, dolphins, and sharks from the United States: Chemosphere, Vol. 70, No. 11, pp. 1935–1944.
98
KOZAK, J.; WEBER, J. B.; AND SHEETS, T. J., 1983, Adsorption of prometryn and metolachlor by selected soil organic matter fractions: Soil Sciences, Vol. 136, No. 2, pp. 93–101. LEIST, T. G.; POTTER, D.; HEBERT, C. E.; AND LETCHER, R. J., 2009, Temporal trends and spatial distribution of non-polybrominated diphenyl ether flame retardants in the eggs of colonial populations of Great Lakes herring gulls: Environmental Science Technology, Vol. 43, No. 2, pp. 312–317. LI, J.; ZHOU, B.; SHAO, J.; YANG, Q.; LIU, Y.; AND CAI, W., 2007, Influence of the presence of heavy metals and surface-active compounds on the sorption of bisphenol A to sediment: Chemosphere, Vol. 68, No. 7, pp. 1298–1303. LI, Y. N. AND ZHOU, Q. X., 2008, Metabolic transformation and ecotoxicology of tetrabromobisphenol A: A review: Journal Ecology, Vol. 27, pp. 263–268. (In Chinese.) LIU, W. P. AND JI, J., 1996, Sorption and desorption—the key factor of pesticides fate in soil-water environment: China Environmental Sciences, Vol. 16, No. 1, pp. 25–30. MENG, F. D.; JIANG, X.; AND JIN, X. C., 2004, Physical-chemical characteristics of the sediments in lakes from the middle and lower reaches of the Yangtze River: Research Environmental Sciences, Vol. 17, pp. 24–29. MITTAL, A.; KURUP, L.; AND MITTAL, J., 2007, Freundlich and Langmuir adsorption isotherms and kinetics for the removal of tartrazine from aqueous solutions using hen feathers: Journal Hazardous Material, Vol. 146, pp. 243–248. MOHAMED, A. E. AND ABDALLAH, S. H., 2011, TetrabromobisphenolA, hexabromocyclododecane and its degradation products in UK human milk: Relationship to external exposure: Environment International, Vol. 37, No. 2, pp. 443–448. MORRIS, S.; ALLCHIIN, C. R.; ZEGERS, B. N.; HAFTKA, J. J. H.; BOON, J. P.; BELPAIRE, C.; LEONARDS, P. E. G.; VAN LEEUWEN, S. P. J.; AND DE BOER, J., 2004, Distribution and fate of HBCD and TBBPA brominated flame retardants in North Sea estuaries and aquatic food webs: Environmental Science Technology, Vol. 38, No. 21, pp. 5497–5504. NYHOLM, R. J.; ANNA, N.; LEIF, N.; PETER, H.; AND PATRIK, L. A., 2008, Maternal transfer of brominated flame retardants in zebrafish (Danio rerio): Chemosphere, Vol. 73, pp. 203–208. QU, G. B.; SHI, J. B.; WANG, T.; FU, J.; LI, Z.; WANG, P.; RUAN, T.; AND JIANG, G., 2011, Identification of tetrabromobisphenol A diallyl ether as an emerging neurotoxicant in environmental samples by bioassay-directed fractionation and HPLC-APCIMS/MS: Environmental Science Technology, Vol. 45, No. 11, pp. 5009–5016. RAVIT, B.; EHRENFELD, J. G.; AND HA¨GGBLOM, M. M., 2005, Salt marsh rhizosphere affects microbial biotransformation of the widespread halogenated contaminant tetrabromobisphenol-A (TBBPA): Soil Biology Biochemistry, Vol. 37, No. 6, pp. 1049– 1057. SHI, Z. X.; JIAO, Y.; HU, Y.; SUN, Z. W.; ZHOU, X. Q.; FENG, J. F.; LI, J. G.; AND WU, Y. N., 2013, Levels of tetrabromobisphenol A, hexabromocyclododecanes and polybrominated diphenyl ethers in human milk from the general population in Beijing, China: Science Total Environment, Vol. 452–453, pp. 10–18. SHI, Z. X.; WU, Y. N.; LI, J. G.; ZHAO, Y. F.; AND FENG, J. F., 2009, Dietary exposure assessment of Chinese adults and nursing infants to tetrabromobisphenol A and hexabromocyclododecanes: Occurrence measurements in foods and human milk: Environmental Science Technology, Vol. 43, pp. 4314–4319. SUN, Z.; MAO, L.; FENG, Z.; YU, Y.; FENG, J.; AND YU, H., 2008a, Sorption and desorption behaviours of tetrabromobisphenol A on fluvo-aquic soil: Environmental Sciences, Vol. 29, No. 10, pp. 2874–2878. (In Chinese.)
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 91–99
Sortion-Desorption of TBBPA SUN, Z.; MAO, L.; FENG, Z.; YU, Y.; AND YU, H., 2008b, Effects of humic acid on sorption of tetrabromobisphenol A by soils: China Environmental Sciences, Vol. 28, No. 8, pp. 748–752. SUN, Z. H.; YU, Y. J.; MAO, L.; ZHENG, F.; AND YU, H. X., 2008c, Sorption behaviour of tetrabromobisphenol A in two soils with different characteristics: Journal Hazardous Material, Vol. 160, pp. 450–461. VELDHOEN, N.; BOGGS, A.; WALZAK, K.; AND HELBING, C. C., 2006, Exposure to tetrabromobisphenol-A alters TH-associated gene expression and tadpole metamorphosis in the Pacific tree frog Pseudacris regilla: Aquatic Toxicology, Vol. 78, pp. 292–302. WHO/ICPS, 1995, Environmental Health Criteria 172. Tetrabromobisphenol A and Derivatives: World Health Organization, Geneva, Switzerland, 12–38 p. XIA, G. AND BALL, W. P., 1999, Adsorption-partitioning uptake of nine low-polarity organic chemicals on a natural sorbent: Environmental Science Technology, Vol. 33, No. 2, pp. 262–269. XING, B. AND PIGNATELLO, J. J., 1997, Dual-mode sorption of low polarity compounds in glassy poly(vinylchloride) and soil organic matter: Environmental Science Technology, Vol. 31, No. 3, pp. 792–796. XU, J.; GAO, Z. S.; XIAN, Q. M.; YU, H. X.; AND FENG, J. F., 2009, Levels and distribution of polybrominated diphenyl ethers (PBDEs) in the freshwater environment surrounding a PBDE manufacturing plant in China: Environmental Pollution, Vol. 157, pp. 1911–1916.
XU, T.; WANG, J.; LIU, S. Z.; LU¨, C.; SHELVER, W. L.; LI, Q. X.; AND LI, J., 2012, A highly sensitive and selective immunoassay for the detection of tetrabromobisphenol A in soil and sediment: Analytica Chimica Acta, Vol. 751, pp. 119–127. YANG, K.; ZHU, L.; LOU, B.; AND CHEN, B., 2005, Correlations of nonlinear sorption of organic solutes with soil/sediment physicochemical properties: Chemosphere, Vol. 61, No. 1, pp. 116–128. YANG, S. W.; WANG, S. R.; WU, F. C.; YAN, Z. G.; AND LIU, H. L., 2012, Tetrabromobisphenol A: tissue distribution in fish, and seasonal variation in water and sediment of Lake Chaohu: Environmental Sciences Pollution Research, Vol. 19, pp. 4090–4096. YANG, S. W.; XU, F. F.; ZHENG, B. H.; AND WU, F. C., 2013, Multibiomarker responses upon exposure to tetrabromobisphenol A in the freshwater fish Carassius auratus: Aquatic Toxicology, Vol. 142–143, pp. 248–256. ZENG, G.; ZHANG, C.; HUANG, G.; YU, J.; WANG, Q.; LI, J.; XI, B.; AND LIU, H., 2006, Adsorption behavior of bisphenol A on sediment in Xiangjiang River, Central-south China: Chemosphere, Vol. 65, No. 9, pp. 1490–1499. ZHANG, P.; LI, Y.; LI, J.; AND LI, Y., 2011, Exposure and distribution of TBBPA in the water and sediments from Lake Chaohu: Inner Mongolia Science Technology Economy, Vol. 5, pp. 51–55. (In Chinese)
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 91–99
99
Gully Erosion Mapping Using Object-Based and Pixel-Based Image Classification Methods AYOOB KARAMI Faculty of Natural Resources, Hormozgan University, Minab Road, Bandar Bbbas, Hormozgan Province, P.O. Box 3995, Iran
ASADOLLAH KHOORANI1 Faculty of Natural Resources, Hormozgan University, Minab Road, Bandar Bbbas, Hormozgan Province, P.O. Box 3995, Iran
AHMAD NOOHEGAR Faculty of Natural Resources, Tehran University, Karaj, Iran
SEYED RASHID FALLAH SHAMSI College of Agriculture, Shiraz University, Shiraz Province, P.O. Box 71454, Iran
VAHID MOOSAVI Faculty of Natural Resources, Yazd University, Yzad Province, Iran
Key Terms: Gully Erosion, Object-Based Classification, Digital Mapping, IRS-P6, Iran
ABSTRACT Gully erosion mapping is a crucial step to monitor the erosion process and to study its current and future local impacts. Gully erosion mapping through fieldwork is difficult, time-consuming, and costly. This article compares various pixel-based image classification (PBC) algorithms, such as ISODATA, Maximum Likelihood Classification, and Support Vector Machine, with the object-based image analysis (OBIA) technique for gully erosion mapping on IRS-P6 images. Six models defined by classification types, classifiers, and feature spaces were built for comparison. The results show that OBIA classification performed better than PBC in terms of accuracy. We also found that the improvement of OBIA was primarily due to employing textural and shape features and optimized feature space, while the use of standard feature space did not improve OBIA. In addition, OBIA significantly reduced the salt-and-pepper effect that obscures the features on the output maps compared to the PBC maps (which had more salt-and-pepper effects). It seems that object-based techniques have yielded better results because of their focus on the shape of gully 1
Corresponding author email: khoorani@hormozgan.ac.ir.
networks rather than on their spectral heterogeneity. In order to improve the accuracy, a priority may be gained by fully exploring the use of membership function and hierarchical approach with multi-scale segmentation for gully mapping. In future studies we propose to determine how these factors can affect the performance of OBIA in terms of gully mapping. This study provides information on the location of gullies, gully dynamics over a period of time, and the degree of land degradation (gully density) for developing and implementing soil conservation measures. INTRODUCTION Gullies in the Fars Province of Iran are large and deep natural ditches or channels in a landscape formed by running water. Recent studies reveal that gully erosion is often a main source of sediment production (Valentin et al., 2005) and can vary to between 10 percent and 94 percent of total sediment yield caused by water erosion (Poesen et al., 2003). Gully erosion causes adverse environmental impacts and high economic costs by negatively affecting agricultural production, water quality, and facilities (Valentin et al., 2005; Taruvinga, 2008). Gully erosion generally is considered as an indicator of desertification and land degradation (UNEP, 1994). Hence, detailed identification and monitoring and mapping of gully development over time are essential requirements for estimating sediment production, soil
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 101–110
101
Karami, Khoorani, Noohegar, Shamsi, and Moosavi
conservation measures, and identification of vulnerable areas for gully formation, land degradation, and environmental impacts (Shruthi et al., 2011). Gully erosion mapping is a time-consuming, difficult job when fieldwork is used alone. Aerial photo visual interpretation is also time-consuming and costly and is limited by the interpreter skills (Vrieling, 2007; Taruvinga, 2008; and Shruthi et al., 2011). Nowadays an alternative for this type of visual interpretation technique is the automatic extraction of gully erosion from satellite imagery. The remote sensing approach is the only practical method for mapping gully features because of the large area and complexity of the size, shape, and occurrence of the gully features (Knight et al., 2007). Automatic soil erosion mapping using remote sensing techniques was initiated by pixel-based or per-pixel image classification (PBC), which only uses the surface reflectance values contained in pixels (Shruthi et al., 2011). PBC uses multi-spectral classification techniques to assign a pixel to a class considering only the spectral similarities with a class. Various PBC methods, such as Maximum Likelihood Classification (MLC), Mahalanobis distance classifier, and Support Vector Machine (SVM), are employed for thematic mapping and quantitative analysis of gully erosion (Valentin et al., 2005; Taruvinga, 2008). Gullies are complex features to map since their spectral heterogeneity is associated with the presence of bare soil, vegetation, or shadow- or moisturerelated brightness differences (Taruvinga, 2008; Shruthi et al., 2011). The spectral heterogeneity of gullies themselves causes their spectral similarity to other land covers, tending to produce ‘‘speckled’’ or ‘‘salt-and-pepper’’ image classification results (Tzotsos et al., 2008; Whiteside et al., 2011). In addition, previous studies have shown that PBC techniques such as MLC and SVM algorithms could not separate water erosion features at an acceptable level of accuracy as a result of the spectral similarities with other non-erosion features (Solaimani and Hadian Amri, 2008; Taruvinga, 2008; Pirie, 2009; Torkashvand and Alipour, 2009; Shruthi et al., 2011; and Mararakanye and Nethengwe, 2012). Gully erosion features have shape, length, topological entities, and textural characteristics that make it possible to treat them as spatial objects that can be characterized based not only on their geometric properties but also on their spatial relationship with surrounding features. A review of the literature indicates that the potential for gully erosion mapping using object-based image analysis (OBIA) from space-borne imagery has not been thoroughly explored. Since gullies have both measurable geometric
102
and spatial properties, object analysis, as opposed to individual pixels, may be more appropriate to address the spectral ambiguity problems and may be more suitable for knowledge-driven analysis (similar to visual image interpretation). Some investigations in gully erosion mapping by OBIA have already been performed (Shruthi et al., 2011). Knight et al. (2007) used ASTER imagery to map alluvial gullies associated with large tropical rivers in Australia, while Eustace et al. (2009) used high-resolution LiDAR data to successfully map gully extent and density using OBIA. Shruthi et al. (2011) investigated the use of OBIA to extract gully erosion features from IKONOS and GEOEYE-1 data using a combination of topographic, spectral, shape (geometric), and contextual information. Mararakanye and Nethengwe (2012) investigated and tested the OBIA technique for gully feature identification in the Limpopo Province in South Africa. Knight et al. (2007) gained approximately 50 percent accuracies for the gully class, showing that an object-based approach does not automatically lead to superior results. Shruthi et al. (2011) found that OBIA gully mapping is quicker and more objective than traditional image-digitization methods. Based on previous studies (e.g., Baatz and Scha¨pe, 2000; Platt and Rapoza, 2008), factors such as texture and shape, membership function, and hierarchical approach with multi-scale segmentation are important to include in the analysis to improve the accuracy and efficiency of OBIA. This study represents the first attempt to conduct a detailed comparison of OBIA and PBC for mapping gullies in Lamerd, Fars Province, Iran, using mediumresolution IRS-P6 data. Accuracies for the classifications are produced and compared using statistical tests. STUDY AREA AND RESEARCH DATA The study area (14 3 11 km) is located in Lamerd Township, Fars Province, in the southwestern region of Iran (Figure 1). The study area has an arid climate, with nine clear, dry seasons and an average annual rainfall of 250 mm. The rainy season is from December to March, with highest amount and intensities of rain occurring in the month of February. The terrain is flat, with slopes measuring between 0 percent and 2 percent and with elevations ranging from 389 m to 2,165 m (most ranging between 400 and 600 m). Most soils (texture) in the area consist of clay loam that are generally saline soil. The range of elevation, mean annual precipitation and temperature, geological formation, land use, landform, and
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 101–110
Gully Erosion Mapping
Figure 1. Study area situated over the southwestern region of Fars Province (Fars, Iran). (a) Location. (b) IRS-P6 data showing the gully networks.
climate type and soil order are included in Table 1. Gully erosion is the most significant erosion feature in the area, with gullies measuring as deep as 2.5 m and about 10 m wide. The gullies are U-shaped and dendritic (Figure 2). The Data A multi-spectral data set of Indian Remote Sensing satellite (IRS-P6), dated August 2008, have been used in this research, and include blue, green, red, and NIR bands of 23.5 m and a Panchromatic band of 5.8 m. The imagery was radiometrically and geometrically corrected and rectified to the world geodetic survey 1984 datum (WGS84) and the Universal Transverse
Table 1. General characteristics of the study area (after from Kompani-Zare et al., 2011). Name Mean elevation (m) Mean annual precipitationa (mm) Mean annual temperature (uC) Geology Land use Landform type Climate typec Soil orders
Lamerd 400 250 24 Quaternary Farming-poor range Flood plain BWh, BSh Aridsols, Entisols
a
Based on Lamerd Station, 1993 to 2010. Poor range: the range land with mean coverage of less than 20 percent. c BWh 5 arid-desert-hot arid; BSh 5 arid-steppe-hot arid. b
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 101–110
103
Karami, Khoorani, Noohegar, Shamsi, and Moosavi
Figure 2. Gullies in the Lamerd region (southwestern region of Iran). (a) U-shaped gully. (b, c) Presence of dry and alive vegetation in gullies. (d) Dendritic gully system on IRS-P6 data.
Mercator coordinate system. Image-to-map registration using a second-order polynomial transformation led to a 0.3-pixel root mean square error for multispectral imagery and an 0.5-pixel error for panchromatic imagery. A visual assessment confirmed that all image sources were aligned with ancillary data layers of higher spatial accuracy (e.g., road network and drainage network). Radiometric processing was applied to the satellite imagery. Absolute atmospheric correction of the imagery was not performed because of the lack of
simultaneously acquired ground-based spectral data or appropriate meteorological data available in the study area. Instead, a relative correction using the Dark-Object Subtraction method was used to reduce the atmospheric scattering effects (Chavez, 1988). Training/Test Data Set and Field Works A commonly accepted practice for assessing products derived from coarse resolution data using semi-automatic classification techniques is the use of a higher
Figure 3. Procedure of ground truth map preparation.
104
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 101–110
Gully Erosion Mapping Table 2. Classification models built for PBC and OBIA comparison.
Model
Classifier
1
PBC
ISODATA
2
PBC
MLC
3
PBC
4 5 6
PBC OBIA
SVM—RBF SVM—liner SVM—polynomial SVM—sigmoid NN NN NN
Data Set Set Set Set Set Set Set Set Set Set Set Set Set Set Set Set
1 2 3 4 1 2 3 4 4 4 4 4 4 4 4
Feature Space SFS SFS SFS SFS SFS SFS SFS SFS SFS SFS SFS SFS SFS SFS OFS
NN 5 nearest neighbor; MLC 5 maximum likelihood; SFS 5 standard feature space; OFS: 5 optimized feature space; Set 1 5 B1, B2, B3, B4; Set 2 5 B1, B2, B3, B4, BPCA1; Set 3 5 B1, B2, B3, B4, BPCA2; Set 4 5 B1, B2, B3, B4, BPCA1, BPCA2.
resolution satellite data as a reference (Taruvinga, 2008). ‘‘Ground truth’’ maps that have gully and non-gully classes were created using traditional gully mapping methods on the IRS-P6 panchromatic image subset and Google Earth images, as summarized in Figure 3. The ground truth map was produce during an intensive field validation effort to check and confirm the existing gully erosion map. In this research two training data sets were used, presenting the gullies and non-gullies, confirmed through field checks and delineated as a polygonal training area on screen. METHODS Image Classification In addition to the original IRS-P6 bands, four sets of test data were generated using ENVI software (ENVI, 2008) in order to evaluate the impact of including additional bands in ISODATA and MLC classifications. In the first set, only the original bands of green, red, NIR, and SWIR were used (Set 1 5 B1, B2, B3, B4). In the second set, the first principle component (PC1) was added to the original bands (Set 2 5 B1, B2, B3, B4, BPCA1). In the third data set, the PC2 was included in the original bands (Set 3 5 B1, B2, B3, B4, BPCA2). Finally, in the fourth data set, both PCs were included in the original data set (Set 4 5 B1, B2, B3, B4, BPCA1, BPCA2). To evaluate the performance of OBIA and PBC techniques, a set of image classification models (Table 2) was constructed. The models are constructed through a multiple comparison procedure. All four data sets (Sets 1, 2, 3, and 4) were classified using
ISODATA and MLC models, and then the data set that produced the highest accuracy was selected for SVM and OBIA. The PBC classification models (models 1, 2, and 3) were conducted using ENVI 4.3, while others were conducted using eCognition package (Definiens, 2004). PBC The PBC was performed using the ISODATA, MLC, and SVM algorithms. The ISODATA algorithm is the most frequently and widely used unsupervised classifier, and it was used in model 1 to test the traditional unsupervised PBC. The imagery was initially classified into four classes with maximum iterations of five and a convergence threshold of 0.95, after which it was coded into two classes. MLC is one of the most powerful and frequently used parametric supervised PBC methods (Huang et al., 2002; Yan et al., 2006; Qian et al., 2007; Dixon and Candade, 2008; Kavzoglu and Colkesen, 2009; Otukei and Blaschke, 2010; and Ouyang et al., 2011), and it was used in model 2. The MLC calculates a statistical (Bayesian) probability function from the inputs for classes established from training sites. Each pixel is then assigned to the class to which it most likely belongs. SVM is a group of theoretically superior machine learning algorithms and one of the latest additions to the existing catalog of image classification techniques that support gully mapping (Taruvinga, 2008). Essentially, SVM is based on fitting a separating hyper-plane that provides the best separation between two classes in a multidimensional feature space. In order to represent more complex shapes than linear hyperplanes, a variety of kernels, including the linear, polynomial, the radial basis function (RBF), and the sigmoid, can be used (Petropoulos et al., 2012). In addition, a penalty parameter can be introduced to the SVM classifier to allow form in classification during the training process. The input parameters required for running SVMs in ENVI include the gamma (c) in the kernel function, the penalty parameter, the number of pyramid levels to use, and the classification probability threshold value. Very little guidance exists in the literature concerning the criteria to be used in selecting the kernel-specific parameters (e.g., Carrao et al., 2008; Li and Liu, 2010). The c parameter was set to a value equal to the inverse of the number of the spectral bands of the imagery, whereas the penalty parameter was set to its maximum value (i.e., 100), forcing no misclassification during the training process. The pyramid
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 101–110
105
Karami, Khoorani, Noohegar, Shamsi, and Moosavi Table 3. Parameter value used in multi-resolution segmentation algorithm. Image Segmentation Parametersa Scale
Color/Shape
Smoothness/Compactness
20 30 60
0.6/0.4 0.5/0.5 0.8/0.2
0.7/0.3 0.6/0.4 0.4/0.6
a
Image layers used: B1, B2, B3, B4, Bpca1, Bpca2.
parameter was set to a value of zero, whereas a classification probability threshold of zero was used, meaning that all image pixels had to be classified into one class. OBIA In the OBIA, the basic processing units are image objects or segments, not single pixels (Dehvari and Heck, 2009). There are unique advantages to OBIA. Multi-scale approaches are one advantage, and the possibility of using geometric or contextual features in both segmentation and classification is another. The segments are regions (groups of pixels) that are generated by one or more criteria of homogeneity. The segmentation algorithm used in eCognition software is a bottom-up region merging algorithm that begins with 1-pixel objects. The procedure includes a pairwise clustering process to merge smaller objects into larger ones with uniform texture and color, as well as an optimization process defined by a set of parameters, such as scale, color, and shape. When the spectral and spatial heterogeneity of one object reaches a defined threshold, the procedure stops its growth. The size of an image object (segment) is determined by a scale parameter (a dimensionless integer). A larger scale parameter causes larger image segments (Aksoy and Ercanoglu, 2011). Two other optimization parameters, color (color 5 1-shape) and shape values, weighted from 0 to 1, are the other important parameters in segmentation. The color value refers to the spectral homogeneity and is very important for creating meaningful objects. The shape criterion defines the textural homogeneity of the resulting image objects, and it is divided into two groups, such as smoothness and compactness. The smoothness criterion is used to optimize image objects with regard to smoothness of borders, while the compactness criterion is used to optimize image objects with regard to compactness. Image segmentation is the first step in OBIA. Different groups of possible parameters were tested to
106
Table 4. Feature bands composition of the optimized space. Feature
Descriptions
Mean layer value Width Length:width ratio
The overall brightness of all bands Main width of the object The ratio of main length to main width of objects Distance to line that shows the main gully’s direction Average area represented by segments Standard deviation area represented by segments
Distance to line Average area Standard deviation
identify a good scale comparing segmented objects with uniform visual properties of the imagery. After several trial-and-error attempts to find the appropriate multi-resolution segmentation parameters, the selected values were determined to be 60, 0.8, 0.2, 0.4, and 0.6 for scale, color, shape, smoothness, and compactness, respectively (Table 3). The segmentation parameters used to provide optimal classification results are shown in Table 3. A nearest neighbor classifier (NN) was also adopted to conduct supervised classification in data set 4 (models 5 and 6). The NN classifies each image object into the class that has the sample object closest to it in a given feature space (Baatz et al., 2004). After segmenting the images the OBIA variables were selected. The object features allow for contextual relationships between image objects to be incorporated into the OBIA. Feature Spaces Different feature spaces were explored in order to test the influence of texture and shape features on classification. Typically in PBC methods, the image classification algorithm employs standard feature space (SFS) comprising blue, green, red, and nearinfrared bands. Nevertheless, classification sometimes also takes advantage of features other than standard spectral features. This is especially important for the OBIA, as objects represent more features than pixels. Features of objects include spectrum, shape, texture, and context, while features of pixels are limited to spectrum and texture. To evaluate the effect of shape and texture features on classification, optimized feature space (OFS) and SFS were compared. The optimized features were obtained from the feature space optimization procedure, which calculates a subset of feature space with the greatest separate distance at a given dimension (Baatz et al., 2004). The optimized feature space of the OBIA (model 6) is listed in Table 4.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 101–110
Gully Erosion Mapping Table 5. Kappa coefficient and overall accuracy of each model derived from error matrixes.
Model
Classifier
Data Set
1
ISODATA
2
MLC
3
SVM—RBF SVM—liner SVM—polynomial SVM—sigmoid NN NN NN
4 5 6
Set Set Set Set Set Set Set Set Set Set Set Set Set Set Set
Overall Accuracy
Kappa Coefficient
54.19 54.8 54.2 54.8 73.7 79.2 78.11 82 77.9 78.64 80 78.6 75 79 89.6
0.15 0.16 0.15 0.16 0.47 0.57 0.55 0.62 0.54 0.56 0.59 0.54 0.5 0.57 0.82
1 2 3 4 1 2 3 4 4 4 4 4 4 4 4
PBC
Accuracy Assessment In order to assess the classification accuracy of the models, the error matrices were established to make a comparison between the classification results and the truth. The class attributes of points on the ground truth map were compared to those on the classified image for each model. Thus, 15 matrices were established, and the heuristics are shown in Table 5. The McNemar’s test has been employed for pairedsample nominal scale data to assess whether statistically significant differences exist between the classification results (Foody, 2004; Whiteside et al., 2011; and Duro et al., 2012). This non-parametric test that is based on a chi-square (x2) statistics (Eq. 1) was used to assess whether there is a statistically significant difference between the PBC and the OBIA results. x2 ~
( f12 {f21 )2 f12 zf21
supervised model 2, which had an overall accuracy (82 percent) and kappa coefficient (0.62) but was significantly lower than that of model 6. In addition, Figure 4 shows that an OBIA model significantly reduced the salt-and-pepper effect when compared to the PBC models (models 2 and 3). In this case, OBIA was superior to PBC in terms of extracting ground objects, showing more accuracy in classification, and was more noticeable in shapes. These results are consistent with those of other studies concerning the comparison between OBIA and PBC (Qian et al, 2007; Dehvari and Heck, 2009; Ouyang et al., 2011; and Duro et al., 2012) (Figure 4).
ð1Þ
where f12 represents the number of the pixels, incorrectly classified by the first classifier, while correctly classified by the second classifier; and f21 represents the number of the pixels, correctly classified by the first classifier, incorrectly classified by the second classifier (Foody, 2004; Petropoulous et al., 2012). RESULTS Overall accuracy and kappa coefficients of the models ranged from 54 percent to 90 percent and from 0.15 to 0.82, respectively (Table 5). The OBIA model 6 reached an overall accuracy of 89.6 percent and a kappa coefficient of 0.82, which was the highest among all models. The most accurate PBC model was
Models 1 and 2 were compared, and the algorithm’s performances were tested. Additional bands of the original data were also evaluated. The results are given in Table 5 and show that the calculated kappa coefficients are different between the two algorithms of four different data sets (Sets 1, 2, 3 and 4). Model 2, with data set 4, has a kappa coefficient of 0.62 and an overall accuracy of 82 percent, which was the highest achieved. When the classification methods were considered, the patterns observed Figure 4 indicate that model 2 (MLC) generally provided a superior result to that of model 1 (ISODATA). When the data sets were considered, it was evident from Table 5 that substantial improvements were achieved when model 2 was developed by adding more variables instead of using only four bands. There were significant differences in the accuracy values when PCA1 and PCA2 bands were added, respectively, to model 2. However, there were minor differences in the accuracy values when PCA bands were added to model 1. The output maps that provided the highest accuracy result obtained from the data sets for models 1 and 2 were also compared using the model 3 accuracy assessment. The classification accuracy assessment results produced from models 2 and 3 for data set 4 are shown in Table 5. For model 3 (i.e., SVMs), higher accuracy was produced from the SVM with sigmoid kernel, which showed an overall accuracy of 80 percent and a kappa coefficient of 0.59. The assessment results also showed that the highest accuracy of model 2 marginally outperformed the highest accuracy of model 3. Since the overall classification results using the kappa coefficient were close, the significance of the results was compared using McNemar’s test statistic. A 2 3 2 contingency matrix has been constructed for the correctly and
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 101–110
107
Karami, Khoorani, Noohegar, Shamsi, and Moosavi
Figure 4. Comparison of PBC and OBIA classification: (a) ISODATA classification (model 1); (b) MLC classification (model 2); (c) SVM classification (model 3); and (d) OBIA classification (model 6).
incorrectly classified pixels and then evaluated using McNemar’s test. Results suggest a chi-square test statistic value of 4.67, which exceeded the chi-square critical value of 3.84 (alpha 5 0.05). Thus, the relatively higher superiority of the MLC approach over that of SVM was accepted. Therefore, the most accurate PBC model was supervised model 2 (MLC). Texture and Shape Features Before considering the effect of texture and shape features, a comparison was made between PBC models and model 5 to evaluate the OBIA model
108
(i.e., excluding OFS that may introduce shape and texture features). All models adopted the SFS. Their proximate accuracy (Table 5) suggests that use of the base OBIA model without employing texture and shape features will not result in higher accuracy than PBC. Model 6 was compared with other models to estimate the effect of shape and textural features. Model 6 uses optimized feature spaces, including textural features, while other models use spectral features alone (Table 4). Apparently texture and shape features influenced the classification accuracy, and OBIA can acquire more accurate classification.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 101–110
Gully Erosion Mapping
General Comparison between PBC and OBIA The most accurate OBIA model (model 6) was compared with the most accurate PBC model (i.e., model 2). The OBIA obtained a higher accuracy (89.6 percent) than the PBC (82 percent), and the salt-andpepper feature on the output map was removed. However, reducing the salt-and-pepper effect is not always advantageous. The OBIA reduces the saltand-pepper effect by merging pixels into objects and small objects into large objects, but it may lose some important detail information in that process. Based on comparison, McNemar’s test indicated that the observed difference between pixel-based and objectbased classifications was statistically significant (p , 0.05).
bands, while this was not the case for model 1. PCA has proven to increase classification accuracies of PBC algorithms (Taruvinga, 2008). Therefore, it is evident that the impact of input variables varies for different algorithms. The results indicate that textural and shape features and OFS are important factors in terms of improving OBIA accuracy, while the use of SFS did not achieve a more accurate result. Apparently texture and shape features influence the classification accuracy, and OBIA can generate a more accurate classification using these features (Shruthi et al., 2011). The improved classification using OBIA can be attributed primarily to its use of objects to reduce the spectral variability in heterogeneous land cover types, such as degraded land. CONCLUSION
DISCUSSION The OBIA method (model 6) provided statistically significant results with higher accuracies than did the PBC models. This is consistent with findings within the literature (Shruthi et al., 2011; Mararakanye and Nethengwe, 2012). This result suggests that OBIA has potential as an alternative method (over PBC approaches) for extracting gullies from IRS-P6 data captured over an arid environment in Fars Province, Iran. The improved classification using OBIA can be attributed primarily to its use of objects to reduce the spectral variability in land cover types that are heterogeneous. The research also shows that the PBC techniques, such as ISODATA, MLC, and SVM algorithms, could not separate gully features at an acceptable level of accuracy as a result of the spectral similarities (spectral ambiguity) with other non-erosion features (Solaimani and Hadian Amri, 2008; Taruvinga, 2008; Torkashvand and Alipour, 2009; Shruthi et al., 2011; and Mararakanye and Nethengwe, 2012). The presence of bare soil, vegetation, or shadow- or moisturerelated brightness differences in gullies causes their spectral similarity to other land covers (Taruvinga, 2008; Shruthi et al., 2011). The PBCs produced a speckled salt-and-pepper appearance that created a confusing output map, while the OBIA showed none of this speckle in the output map. This unfavorable result in the PBCs may be decreased by the addition of ancillary information (e.g., DEM, land use map to mask out spectrally similar features such as urban build-up areas) prior to the mapping of gullies. Methods to improve accuracies of PBCs include post-classification editing, such as filtering and manual removal. Improvement was achieved when model 2 was developed by adding more variables (PC bands) instead of using only four
Locating and quantifying gully areas within a catchment is a major challenge for the monitoring and reduction of sediment movement to reduce sediment and nutrient discharge into the surface runoff and water bodies. Accurate and detailed spatial information on gully location and extent at an appropriate spatial scale is an essential part of evaluating the impacts of the gullies on erosional sedimentation in the catchment. This article describes a comparison between OBIA and PBC for mapping gully erosion features from satellite imagery. The results of this study show a significant difference in the accuracy between PBC and OBIA in terms of mapping gullies. We also found that the improvement of OBIA was primarily due to employing textural and shape features and OFS, while the use of SFS did not improve OBIA. Membership function and hierarchical approach with multi-scale segmentation are also important factors for improving the accuracy and efficiency of OBIA (Baatz and Scha¨pe, 2000; Platt and Rapoza, 2008; and Ouyang et al., 2011). To improve the accuracy, a priority may be gained by fully exploring the use of membership function and hierarchical approach with multi-scale segmentation for gully erosion mapping. In future studies, we propose to determine how these factors can affect the performance of OBIA for gully erosion mapping. ACKNOWLEDGMENTS We wish to thank the Geographical Organization of Iran Army for providing the IRS-P6 imagery grant. We also thank Mr. Dehghan (Watershed Management Office of Lamerd) and his team for providing logistical support during the fieldwork. Thanks to the anonymous reviewers for their valuable comments.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 101–110
109
Karami, Khoorani, Noohegar, Shamsi, and Moosavi
REFERENCES AKSOY, B. AND ERCANOGLU, M., 2011, Landslide identification and classification by object-based image analysis and fuzzy logic: An example from the Azdavay region (Kastamonu, Turkey): Computers Geosciences, doi:10.1016/j.cageo.2011.05.010. BAATZ, M.; BENZ, U.; DEHGHANI, S.; HEYNEN, M.; HO¨LTJE, A.; HOFMANN, P.; LINGENFELDER, I.; MIMLER, M.; SOHLBACH, M.; WEBER, M.; AND WILLHAUCK, G., 2004, eCognition Professional User Guide: Definiens Imaging GmbH, Mu¨nchen, Germany. BAATZ, M. AND SCHA¨PE, A., 2000, Multiresolution segmentation— an optimization approaches for high quality multi-scale image segmentation: Wichmann-Verlag, Heidelberg, Germany, pp. 12–23. CARRAO, H.; GONCALVES, P.; AND CAETANO, M., 2008, Contribution of multispectral and multitemporal information from MODIS images to land cover classification: Remote Sensing Environment, Vol. 112, pp. 986–997. CHAVEZ, P. S., Jr., 1988, An improved dark-object subtraction technique for atmospheric scattering correction of multispectral data: Remote Sensing Environment, Vol. 24, pp. 459–479. DEHVARI, A. AND HECK, R. J., 2009, Comparison of object-based and pixel based infrared airborne image classification methods using DEM thematic layer: Journal Geography Regional Planning, Vol. 2, pp. 86–96. DIXON, B. AND CANDADE, N., 2008, Multispectral landuse classification using neural networks and support vector machines: One or the other, or both?: International Journal Remote Sensing, Vol. 29, pp. 1185–1206. DURO, D. C.; FRANKLIN, S. E.; AND DUBE´, M. G., 2012, A comparison of pixel-based and object-based image analysis with selected machine learning algorithms for the classification of agricultural landscapes using SPOT-5 HRG imagery: Remote Sensing Environment, Vol. 118, pp. 259–272. ENVI, 2008, ENVI On-Line Software User’s Manual, ITT Visual Information Solutions. Electronic document, available at http://www.exelisvis.com/ EUSTACE, A.; MATTHEW, P.; AND CHRISTIAN, W., 2009, Give me the dirt: Detection of gully extent and volume using highresolution LiDAR. In Jones, S. and Reinke, K. (Editors), Innovations in Remote Sensing and Photogrammetry: Springer-Verlag, Berlin, Germany, pp. 255–269. FOODY, G. M., 2004, Thematic map comparison: Evaluating the statistical significance of differences in classification accuracy: Photogrammetric Engineering Remote Sensing, Vol. 70, pp. 627–633. HUANG, C.; DAVIS, L. S.; AND TOWNSHEND, J. R. G., 2002, An assessment of support vector machines for land cover classification: International Journal Remote Sensing, Vol. 23, pp. 725–749. KAVZOGLU, T. AND COLKESEN, I., 2009, A kernel functions analysis for support vector machines for land cover classification: International Journal Applied Earth Observation Geoinformation, Vol. 11, pp. 352–359. KNIGHT, J.; SPENCER, J.; BROOKS, A.; AND PHINN, S., 2007, Largearea, high-resolution remote sensing based mapping of alluvial gully erosion in Australia’s tropical rivers. In Proceedings of the 5th Australian Stream Management Conference: Australian Rivers: Making a difference, New South Wales, Australia, pp. 199–204. KOMPANI-ZARE, M.; SOUFI, M.; HAMZEHZARGHANI, H.; AND DEHGHANI, M., 2011, The effect of some watershed, soil characteristics and morphometric factors on the relationship between the gully volume and length in Fars Province, Iran: Catena, Vol. 86, pp. 50–159.
110
LI, D.-C. AND LIU, C.-W., 2010, A class possibility based kernel to increase classification accuracy for small data sets using support vector machines: Expert Systems Applications, Vol. 37, pp. 3104–3110. MARARAKANYE, N. AND NETHENGWE, N. S., 2012, Gully features extraction using remote sensing techniques: South African Journal Geomatics, Vol. 1, pp. 109–118. OTUKEI, J. R. AND BLASCHKE, T., 2010, Land cover change assessment using decision trees, support vector machines and maximum likelihood classification algorithms: International Journal Applied Earth Observation Geoinformation, Vol. 12, pp. 27–31. OUYANG, Z.-T.; ZHANG, M.-Q.; XIE, X.; AND SHEN, Q., 2011, A comparison of pixel-based and object-oriented approaches to VHR imagery for mapping saltmarsh plants: Ecological Informatics, Vol. 6, pp. 136–146. PETROPOULOS, G. P.; KALAITZIDIS, C.; AND VADREVU, K. P., 2012, Support vector machines and object-based classification for obtaining land-use/cover cartography from Hyperion hyperspectral imagery: Computers Geosciences, Vol. 41, pp. 99–107. PLATT, R. V. AND RAPOZA, L., 2008, An evaluation of an objectoriented paradigm for land use/land cover classification: Professional Geographer, Vol. 60, pp. 87–100. POESEN, J.; NACHTERGAELE, J.; VERSTRAETEN, G.; AND VALENTIN, C., 2003, Gully erosion and environmental change: Importance and research needs: Catena, Vol. 50, pp. 91–133. QIAN, J.; ZHOU, Q.; AND HOU, Q., 2007, Comparison of pixel-based and object-oriented classification methods for extracting built-up areas in aridzone: ISPRS Workshop on Updating Geo-Spatial Databases with Imagery & The 5th ISPRS Workshop on DMGISs, Urumchi, Xingjizng, China, pp. 163–170. SHRUTHI, R. B. V.; KERLE, N.; AND JETTEN, V., 2011, Object-based gully feature extraction using high spatial resolution imagery: Geomorphology, Vol. 134, pp. 260–268. SOLAIMANI, K. AND HADIAN AMRI, M. A., 2008, Application of IRS-1D data in water erosion features detection (case study: Nour Roud catchment, Iran): Pakistan Journal Biological Sciences, Vol. 11, pp. 1893–1900. TARUVINGA, K., 2008, Gully Mapping using Remote Sensing (Case Study in KwaZulu-Natal, South Africa): M.S. Thesis, University of Waterloo, Ontario, Canada, 121 p. TORKASHVAND, A. M. AND ALIPOUR, H. R., 2009, Investigation of the possibility to prepare supervised classification map of gully erosion by RS and GIS: World Academy Science, Engineering Technology, Vol. 55, pp. 291–293. TZOTSOS, A.; IOSIFIDIS, C.; AND ARGIALAS, D., 2008, A Hybrid Texture-Based and Region Based Multiscale Image Segmentation a5Algorithm: Springer, New York, pp. 93–112. UNEP, 1994, United Nations Convention to Combat Desertification United Nations Environmental Programme. Geneva. VALENTIN, C.; POESEN, J.; AND LI, Y., 2005, Gully erosion: impacts, factors and control: Catena, Vol. 63, pp. 132–153. VRIELING, A., 2007, Mapping Erosion from Space: Doctoral Thesis, Wageningen University, ISBN: 978-90-8504-587-8. pp. 167. WHITESIDE, T. G.; BOGGS, G. S.; AND MAIER, S. W., 2011, Comparing object-based and pixel-based classifications for mapping savannas: International Journal of Applied Earth Observation and GeoInformation, Vol. 13, No. 6, pp. 884–893. YAN, G.; MAS, J.-F.; MAATHUIS, B. H. P.; XIANGMIN, Z.; AND VAN DIJK, P. M., 2006, Comparison of pixel-based and object-oriented image classification approaches – A case study in a coal fire area, Wuda, Inner Mongolia, China: International Journal of Remote Sensing, Vol. 27, No. 18, pp. 4039–4055.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 101–110
Near-Surface Geophysical Imaging of a Talus Deposit in Yosemite Valley, California ANNA G. BRODY CHRISTOPHER J. PLUHAR1 Department of Earth and Environmental Sciences, California State University, Fresno, 2576 East San Ramon Avenue, Mail Stop ST-24, Fresno, CA 93740
GREG M. STOCK National Park Service, Yosemite National Park, 5083 Foresta Road Box 700, El Portal, CA 95389
W. JASON GREENWOOD Advanced Geosciences, Inc., 2121 Geoscience Drive, Austin, TX 78726
Key Terms: Geophysics, Ground Penetrating Radar, Seismic Refraction, Electrical Resistivity, Mass Wasting, Rock Fall
ABSTRACT Talus at the base of cliffs in Yosemite Valley, CA, represents rock fall and debris avalanche accumulation occurring since the glacial retreat after the last glacial maximum. This ongoing mass wasting subjects humans and infrastructure to hazards and risk. In order to quantify post-glacial rock-fall rates, talus volumes are needed for the deposits of interest. We used three nearsurface geophysical methods (ground penetrating radar, electrical resistivity, and seismic refraction) to locate the basal contact of talus below Glacier Point, near Curry Village in the eastern Yosemite Valley. The coarseness of the talus deposit limited our ability to use these methods in some areas, and the geometry at the base of the cliff restricted our ability to conduct seismic refraction and electrical resistivity across the talusbedrock boundary there. Nonetheless, we were able to detect the basal boundary of talus on top of both bedrock and glacio-fluvial sediment fill. Geophysical imaging revealed an apparent onlapping relationship of talus over aggrading post-glacial sediment fill, and our data support the proposition of approximately 5 m of valley floor aggradation since deglaciation. The bedrock-talus contact is characterized by a dip of 52– 646, consistent with the dip of the cliff surface above the talus apex. Ground penetrating radar and resistivity
1
Corresponding author email: cpluhar@csufresno.edu.
were the most diagnostic methods, in addition to being the most rapid and easiest to implement on this type of deposit. INTRODUCTION Yosemite Valley, located in the central Sierra Nevada of California (Figure 1), provides an outstanding natural laboratory for studying rock fall in isolation from the complicating influences of other mass wasting processes. The 1-km-tall sheer granitic walls of Yosemite Valley, sculpted by alpine glaciers during the Pleistocene and mostly devoid of soils, have subsequently been modified almost exclusively by rock-fall processes. Rock falls present a threat to the approximately four million people that visit Yosemite National Park annually, as well as to infrastructure and facilities (Guzzetti et al., 2003; Stock et al., 2013). Between 1857 and 2011, 15 people were killed and at least 85 seriously injured by such events in Yosemite Valley (Stock et al., 2013). An inventory of historical rock falls in Yosemite (Stock et al., 2013) forms the basis for numerous studies of rock-fall–triggering mechanisms and volume-frequency relations (e.g., Wieczorek et al., 1995, 1999; Wieczorek and Ja¨ger, 1996; Dussauge-Peisser et al., 2002; Dussauge et al., 2003; and Guzzetti et al., 2003). This historical inventory is valuable but suffers from variable reporting rates through time, incomplete reporting of small events, coarse estimates of rock-fall volumes for most catalogued events, and the relatively short duration of the observation record (155 years). As a result, measures of long-term (thousands of years) rock-fall activity are needed to evaluate historical activity, to
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
111
Brody, Pluhar, Stock, and Greenwood
Figure 1. Location map of studied talus deposit in Yosemite Valley, Yosemite National Park (YNP), CA, shown in hillshade (illumination angle 5 315u, azimuth 5 45u) derived from a 1 3 1–m LiDAR-based digital elevation model. Red box indicates study area beneath Glacier Point shown in Figure 2.
examine the possibility of changes in rock-fall rate with time, and to examine geologically relevant volume-frequency distributions of rock fall. The long-term record of rock-fall activity in Yosemite Valley is preserved in the rock-fall debris, or talus, that has accumulated beneath the valley walls. These deposits consist of volumes of many millions of cubic meters and reach heights of more than 100 m above the floor of Yosemite Valley. Because the floor of Yosemite Valley is wide (,1 km) and very low gradient (,3 m/km), there is very little post-depositional modification or degradation of talus slopes. Thus, the talus deposits in Yosemite Valley offer a unique opportunity to quantify longterm rock-fall activity (e.g., Wieczorek and Ja¨ger, 1996). Critically, such quantification relies on assumptions about the state of the valley floor immediately following deglaciation. It is reasonable to presume that each major glacial advance down the valley removed accumulated talus from the previous interglacial period, such that the talus deposits record the accumulation since ice last retreated from the valley. If correct, the talus in Yosemite Valley would have accumulated for only the past 15,000– 17,000 years, the approximate age of local glacial retreat at the end of the Last Glacial Maximum (LGM) (Huber, 1987; Wieczorek and Ja¨ger, 1996; and Stock and Uhrhammer, 2010). Since deglacia-
112
tion, sparse data suggest approximately 5 m of aggradation of the valley floor with glacio-fluvial sediments (Cordes et al., 2013). In order to evaluate these presumptions, we employed near-surface geophysical imaging techniques to map the subsurface extent of a talus deposit in Yosemite Valley. Our research builds upon successful work in the European Alps using ground penetrating radar (GPR), seismic refraction (SR), and two-dimensional–resistivity (2DR) methods to define shallow subsurface (,30-m) contacts between bedrock and talus (e.g., Otto and Sass, 2006; Sass, 2006, 2007). Here we demonstrate that these methods can help constrain the subsurface extent of thick talus accumulations against both a steeply dipping bedrock contact and underlying glacio-fluvial sediments. STUDY AREA Geologic Setting of Yosemite Valley Topographic relief in Yosemite Valley derives from creation of the Sierra Nevada batholith during Mesozoic Farallon–North America subduction and arc volcanism (Bateman, 1992), erosion during the Paleogene (Wakabayashi and Sawyer, 2001), and rejuvenation of relief since the mid-Miocene (e.g., Huber, 1981; Wakabayashi and Sawyer, 2001).
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
Near-Surface Geophysical Imaging of Talus Deposit
Multiple Quaternary glaciations deepened and modified the drainage network of the Sierra Nevada (Wahrhaftig and Birman, 1965; Huber, 1987). The most recent period of glaciation, locally called the Tioga glaciation, peaked between 28 and 17 ka (Bursik and Gillespie, 1993; Phillips et al., 2009), corresponding with the global LGM. Unlike previous glaciations, which filled Yosemite Valley to the rim, the Tioga glaciation only extended part way up the valley walls (Matthes, 1930; Huber, 1987; and Wieczorek et al., 2008). Matthes (1930) mapped the extent of the Tioga glaciation in Yosemite Valley, denoting the farthest advancement by the presence of a probable terminal moraine near Bridalveil Meadow. Deglaciation of the valley occurred beginning about 19,000 years before present (BP), with most of the valley free from ice by 15 ka (Smith and Anderson, 1992; Stock and Uhrhammer, 2010). Below the Tioga trimline, the steep cliffs were scoured by glacial erosion, with some cliffs still retaining glacial polish. In contrast, areas above the Tioga trimline are less steep and have been weathered for a much longer interval, promoting rock falls from those areas (Bronson and Watters, 1987; Wieczorek et al., 2000, 2008; and Guzzetti et al., 2003). We chose the Curry Village talus cone for geological, hazard assessment, and survey feasibility reasons. The relatively simple talus accumulation here appears to consist entirely of blocks from fragmentaltype rock falls (i.e., no large rock avalanches) and is also free from other possible modes of accumulation, such as debris slides or debris flows, ensuring that we understand the processes creating the deposit. In addition, it is located adjacent to areas in which something is known of the subsurface (Cordes et al., 2013; National Park Service, 2013). The project also contributes to hazard assessment of the populated Curry Village, with its history of damaging rock falls. This project permitted the quantification of geological rock-fall rates (Brody, 2011) for comparison to historical rates. Finally, the survey location represents one of only a few areas in Yosemite Valley where geophysical equipment could be deployed effectively. Many of the active talus cones consist entirely of cobbles to boulders, with little option for inserting electrical resistivity electrodes and Betsy SeisgunTM shots or coupling to GPR antennae. Rock-fall source areas for talus deposits near Curry Village consist of the Half Dome Granodiorite and Granodiorite of Glacier Point (Peck, 2002). At Glacier Point, numerous joint sets (Wieczorek and Snyder, 1999; Weizorek et al., 2008; and Matasci et al., 2011) provide planes of weakness from which the rock falls of the study area often originate. The most prominent sets are nearly vertically oriented and
moderately east-dipping regional-scale joints. The intersection of these dominant features with other joint sets is responsible for the overall structure of the cliffs at this location (Matasci et al., 2011). The most numerous joints present at Glacier Point are sheeting (exfoliation) joints that have formed subparallel to the topographic surface (Wieczorek and Snyder, 1999; Stock et al., 2011). The studied talus deposit is located on the floor of Yosemite Valley east of Curry Village, beneath Glacier Point. In this area, the cliff below Glacier Point is a curving, glacially polished bedrock slab with a surface slope (dip) of approximately 60–65u, known locally as the Glacier Point Apron. Large deposits of talus flank the base of the Glacier Point Apron. These deposits are up to 130 m thick and extend as much as 370 m outward from the base of the cliff. Talus clast size at the study area increases with distance from the apex of the deposit as a result of ‘‘gravity sorting,’’ characteristic of talus slopes formed by fragmental-type rock falls (Evans and Hungr, 1993). Sand- to cobble-sized debris dominates the upper several meters of the slope proximal to the cliff face, with larger boulders up to tens of cubic meters in volume on the distal portion of the slope. Although talus near Curry Village has accumulated since 15–17 ka, at least 28 historical rock falls and rock slides recorded from above Curry Village have contributed to the overall talus volume there (Figure 2; Stock et al., 2013). This includes several notable and well-documented rock falls since 1998, with volumes ranging from about 213 to 5,637 m3 (Wieczorek and Snyder, 1999; Wieczorek et al., 2008; and Stock et al., 2011, 2013). Geophysical Survey Line To map the basal contact of the talus deposit, we employed geophysical techniques on a survey line positioned along the boundary between two talus cones near Curry Village (Figure 1, inset). We selected this location as a result of (1) the likelihood of imaging the basal contact of talus against crystalline bedrock and glacio-fluvial sediment fill in this relatively thinner portion of the talus deposit and (2) the ability to insert geophysical equipment into the finer-grained talus debris to the necessary depth (30– 60 cm) below grade. The survey line originates at 1,262 m in elevation at the base of the Glacier Point Apron cliff face, extends north toward the valley floor along a strike of N68uE, and ends at 1,218 m in elevation beyond the distal edge of the talus. The surface slope averages approximately 12.5u along the profile line, with a maximum of about 17u in the upper portion near the cliff face. At the southern end
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
113
Brody, Pluhar, Stock, and Greenwood
Figure 2. Talus deposits beneath Glacier Point Apron in eastern Yosemite Valley, shown in plan view (A) and oblique view (B) as viewed from the northeast. Areal extent of studied talus deposit shown in blue; geophysical survey line denoted by red line. The parking lot indicated caps a landfill site, the margin of which lies more than ten meters to the southeast of the geophysical profile line.
of the survey line, adjacent to the bedrock cliff, the deposit consists of sands, gravel, and cobbles (Figure 3A). Downslope from this, the upper section of the survey line is relatively steep and is composed primarily of gravel and cobble-sized clasts. Approximately halfway down the slope, a small seasonal stream traverses the profile line in two locations, demonstrating that the talus slope experiences minor modification by other processes (Figure 3B). The middle section of the survey line has a lower gradient and is dominated by gravel and cobbles, with numerous large boulders present (Figure 3C). The lower section of the profile exhibits a sandy texture and is crossed by the same stream present in the upper section of the survey line (Figure 3D). The survey line terminates west of a parking lot, beyond the approximate surficial contact between talus material and valley sediment fill (Figure 2). The gravel-surface parking lot caps a former landfill, the subject of a subsurface investigation (National Park Service, 2013) that allowed some verification of our geophysical interpretations. METHODS We used three near-surface geophysical methods to locate the basal contact of the talus deposit: GPR, SR, and 2DR. These techniques have been employed on talus slopes in the Swiss Alps, demonstrating the feasibility of geophysically imaging talus-bedrock contacts in some situations (Hoffmann and Schrott, 2003; Otto and Sass, 2006; and Sass, 2006, 2007). In addition to being non-invasive, under the right circumstances these techniques can offer rapid results and correlatable features between methods (Otto and Sass, 2006; Sass, 2006, 2007). We employed all three geophysical methods along the same survey line in
114
order to compare results and refine the overall interpretation. Where possible, we validated our geophysics interpretation with borehole data from the landfill investigation. As a result of the protected status of Yosemite National Park, no other invasive subsurface investigation was permitted. GPR Methods We used the common offset method of GPR reflection surveying (Neal, 2004). An important assumption in GPR data presentation is that radar reflections originate from directly beneath the survey equipment. Corrections must be made for dipping reflectors or reflections from above-ground features (e.g., large boulders or trees). In this study, correction was made manually rather than by migration techniques (e.g., Porsani et al., 2006). We conducted the GPR survey using a Sensors & Software pulseEKKO PRO unit with both 50-MHz and 100-MHz antennas in bistatic configuration. The lower frequency antenna provides deeper penetration (approximately 45–50 m) but lower (coarser) resolution, while the higher frequency antenna enhances resolution but reduces the maximum depth penetration (approximately 35–40 m) (Jol, 1995; Smith and Jol, 1995). Using both antennas allowed comparison and maximization of data quality at different depths. Transmitting and receiving antennas were set at 1 m apart (Sensors & Software, 1999a), with radar traces collected at 0.5-m intervals along the survey line. GPR data were acquired during late October, the driest part of the year, reducing the effect of near-surface attenuation by soil moisture. We processed GPR data using Sensors & Software EKKO View Deluxe 4 software and applied basic processing methods, including DEWOW filtering and constant gain (Fisher et al., 1992; Sensors & Software,
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
Near-Surface Geophysical Imaging of Talus Deposit
Figure 3. Field images of the eastern margin of the studied talus deposit near Curry Village showing the location of geophysical survey line (red line): (A) Upper portion of talus slope near cliff face; (B) Traversing the small seasonal stream bed; (C) Gradual slope near boundary of talus against glacio-fluvial sediment fill; (D) Fluvial sediment from seasonal stream at distal edge of talus slope.
1999b). Using elevations from a 1-m LiDAR-derived digital elevation model (DEM), we applied a topographic correction to the GPR data, as there is significant relief along the survey line. We converted from travel time to depth using a velocity of 0.14 m/ns, an average for the expected material in the talus deposit (Otto and Sass, 2006), in which velocity in dry soil/dry sand 5 0.15 m/ns and velocity through granite 5 0.13 m/ns (Sensors & Software, 2006). SR Methods We conducted the SR survey using two 24-channel Geometrics Inc. Geode model seismographs and 48 geophones at variable spacing along the survey line. We used 3-m spacing for the southern portion (closer to the cliff face), where the talus was thought to be relatively thinner, and 5-m spacing for the northern portion (closer to the valley floor), resulting in a total
line length of 200 m. Offset shots added another 40 m to this survey line length. Seismic energy was derived from gunpowder blasts triggered with a modified Betsy SeisgunTM. We fired shots at 21 sites along the geophone array as well as at offset locations out from the northern end of the profile line toward the center of Yosemite Valley. Shots were detonated at 0.5–1 m depth in hand-augered backfilled holes. Offset shots were not possible on the south end of the survey line because of the steep bedrock cliff south of the apex of the talus slope. Four to eight stacked shot traces at each geophone for each shot point enhanced the desired signal and reduced non-coherent noise (e.g., automobile traffic, footfalls of hikers, wind, etc.). The SR survey was completed during late spring and summer, when ground conditions were relatively dry and the water table was expected to be at a low level. We conducted a separate seismic velocity experiment on site bedrock in order to independently
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
115
Brody, Pluhar, Stock, and Greenwood
measure the P-wave velocity in this material. For this experiment we epoxied six metal disk geophone mounts directly to the bedrock face and used sledgehammer strikes on the bedrock face as the seismic energy source. We analyzed the seismic refraction data using the Geometrics Inc. SeisImager/2DTM software package, which includes the PickWinTM and PlotRefaTM modules. We hand-picked first arrivals from each raw stacked geophone trace. This information was converted into travel time curves for each shot location along the survey line and was later combined into a single data file. To calculate the depth of potential refractor(s), we applied three methods: timeterm inversion, network-raytracing, and tomography. Surface topography was incorporated using elevations from the 1-m LiDAR-derived DEM. Producing a tomographic inversion was particularly important for this study because (1) lateral variations in seismic velocities are expected within the talus deposit as a result of locally variable densities, and (2) the steeply dipping talus-bedrock contact is an expected and critical feature of interest to the study. We iterated the tomographic inversion twice, per Geometrics’ recommendation, and applied network-raytracing to assess the misfit between the final model and the original data (Geometrics, 2006). 2DR Methods Electrical resistivity values are highly affected by several variables, including lithology, the presence of water and/or ice, the amount and distribution of pore space in the material, and temperature (Reynolds, 1997). The expected resistivity for granite is approximately 300–3,000,000 V-m, while talus is expected to produce a range of values between 100 and 5,000 Vm, and valley fill is expected to range from 10 to 1,000 V-m (Loke, 2000; Sass, 2007). The presence of moisture or groundwater reduces resistivity values compared to dry values, resulting in the large ranges for any given material. Given these variations of many orders of magnitude, the contacts between talus and crystalline bedrock or glacio-fluvial valley fill can potentially be identified on the basis of large contrasts in resistivity values. We acquired 2DR data using an Advanced Geosciences, Inc., SuperStingR1TM resistivity IP/SP system. The resistivity array consisted of a 28electrode passive cable spaced at 6-m intervals and connected to stainless-steel electrode stakes. The resulting profile length of 168 m was moved in a 50 percent roll-along–type array to maximize linear coverage. To ensure proper electrical coupling with the ground, the soil around each electrode stake was
116
wetted with salt water. Current was applied to the subsurface using a dipole-dipole roll-along survey. We increased the sampling detail along the upper end of the profile by increasing the number of unique dipole-dipole pairs, as imaging the interface of talus against crystalline bedrock and glacio-fluvial valley fill was a primary objective of the study. The SuperstingR1 handled the following tasks: autoranging of current and voltage to maximize signal levels, data stacking with standard deviation, and automatic switching of all electrode geometries with the switchbox28 system. We analyzed the 2D-resistivity data using Advanced Geosciences, Inc. EarthImager 2D Resistivity and IP Inversion software. This program solves for the best-fitting smooth model solution from surficial apparent resistivity data. Topographic corrections obtained from the LiDAR-derived DEM were applied to the electrode positions to increase the accuracy of the final model. Resistivity modeling begins with an initial model based on the average raw data value, followed by iterative forward and inverse modeling. During these iterations, the software compares the resulting synthetic data from a forward model to the measured results and iteratively varies the inverse model resistivity values to decrease the misfit between the model result and the measured data. If model convergence is not achieved by a root mean square (RMS) error of less than 10% percent and L2 close to 1, then a small amount of misfit raw data is removed (,5 percent of the total) and the model is started over (Advanced Geosciences, Inc., 2013). RESULTS GPR Results Evaluation of the 50-MHz and 100-MHz GPR data revealed numerous radar reflectors beneath the talus surface (Figure 4). A pair of features at the south end of the profile (between position 0 and 45 m along the profile at times 0 to 1,100 ns) dip steeply toward the valley. This is clearly evident in the 50MHz results and less so in the 100-MHz data. We interpret the pair of features to be the bedrock-talus contact and a sheeting joint parallel to that about 5 m beneath it. However, it is critical to evaluate whether these signals could be spurious: the result of the radar wave bouncing off the cliff face either through an airwave or a direct ground surface wave. In other words, the processing software assumes that all energy returning from a radar pulse originates from reflections directly underfoot, even though returns could originate from anywhere in a shell of equal
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
Near-Surface Geophysical Imaging of Talus Deposit
Figure 4. GPR data corrected for surface elevation, DEWOW filtered, using constant gain and a radar velocity of 0.140 m/ns. (A) 50-MHz results. (B) 100-MHz results. Note that steeply dipping reflectors, such as the bedrock contact with talus, must be geometrically corrected (see Figure 5).
travel time around the GPR apparatus. In order to test these alternate interpretations, we performed a simple velocity calculation to determine whether the reflector could have resulted from an airwave or direct ground wave using the equation D 5 v 3 T/2, where D is the one-way distance to the reflector, v is velocity, and T is two-way travel time. From this calculation, it is evident that the signal is not the result of an airwave bouncing off surface bedrock at the south end of the profile, since the resulting velocity (0.08–0.1 m/ns) is much slower than that of a radar wave through air (0.3 m/ns). The same logic makes a refracted airwave unlikely. The interpretation that the signal is the result of a ground wave bouncing off surface bedrock at the south end of the profile is permissible based on the range of possible surface soil velocities, but this is unlikely for two
reasons. First, the ground wave explanation cannot easily account for the parallel reflectors. Second, the radar wave velocity needed to explain this as a ground wave is somewhat slower than recommended values (Sass and Wollny, 2001; Otto and Sass, 2006; Sass, 2006, 2007; and Sensors & Software, 2006) for dry porous material such as the talus on the profile line surface at the time of the survey. Furthermore, previous studies have succeeded in locating bedrocktalus contacts and joints within bedrock, demonstrating the feasibility of imaging such structures (e.g., Toshioka et al., 1995; Sass and Wollny, 2001; Porsani et al., 2006; and Sass, 2006). Since the steeply dipping reflectors at the south end of the GPR profile appear to be real, with the upper one corresponding to the inferred bedrock-talus interface, a geometric correction is required in order
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
117
Brody, Pluhar, Stock, and Greenwood
Figure 5. GPR dipping reflector correction. (A) All reflected energy is initially assumed to return from directly beneath the GPR, regardless of reflector dip. (B) Geometric correction for dipping reflectors stems from the fact that angle of incidence equals angle of reflection. A non-directional radar pulse will return significant reflected energy perpendicularly from the dipping structure rather than from the same dipping reflector directly beneath the instrument.
to render its true orientation. This correction is based on the fact that a radar wave is emitted as a nondirectional pulse, but the GPR records the signal as if it derived from returns perpendicular to the ground surface. Since a strong signal, such as that in the 50MHz results, would be expected if the radar wave had been reflected off a subsurface feature perpendicular to the angle of incidence (Figure 5), we calculated the true orientation of the dipping reflector along the southern end of the profile (from position 0–20 m). This correction yields a true dip of the talus-bedrock interface between 52u and 64u from horizontal, consistent with field measurements of the dip angle of the bedrock cliff adjacent to the apex of the talus slope. There are also multiple, parallel, strong reflections within the middle portion of the profile (between position 110 and 170 m along the profile at travel times of 1,100–1,200 ns and calculated subsurface elevations of 1,210–1,218 m), which appear to shallow toward the north. This component of the data could represent the onlap of talus over glacio-fluvial sediment fill (Figure 6). As previously stated, we assume that after deglaciation ca. 15 to 17 ka, the valley floor was relatively flat bottomed. As talus accumulated, the deposit should have prograded
118
northward into the valley, synchronous with aggradation of the valley floor with glacio-fluvial sediment; this would result in a contact between talus and fill that dips toward the cliff (Figure 6). The multiple, apparently bedded reflectors are either bedded sediment fill with talus deposited on top or coarse bedding within the lower portion of the talus deposit (Figure 5). The magnitude of the dip of this feature is a function of the radar wave velocity chosen but in this case is consistent with other geophysical data. These alternative interpretations and the dip of these reflectors are further developed in the Discussion section. In addition, the GPR data reveal a zone lacking strong internal reflectors between 190 and 235 m along the profile line and beneath approximately 1,100-ns travel time (Figure 4) in both the 50-MHz and 100-MHz results. Since attenuation increases with increasing water content (Neal, 2004), such a feature could originate from attenuation by soil moisture or groundwater, but could alternately result from the presence of bedrock lacking internal structure (Sass, 2007). Numerous concave-down hyperbolas are also visible throughout the profile (e.g., between positions 135 and 140 m at time 900 ns),
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
Near-Surface Geophysical Imaging of Talus Deposit
Figure 6. Schematic of post-glacial talus and glacio-fluvial deposition, showing distal edge of talus prograding out onto aggrading glacio-fluvial sediment fill. Steady deposition of both talus and glacio-fluvial sediment (shown here in three time snapshots) produces a ‘‘back dip’’ of the basal talus contact that dips toward the cliff.
which we interpret as internal reflections from large boulders within the talus. SR Results Tomographic modeling along the profile line (Figure 7B) generated 18 layers with seismic velocities ranging between 300 m/s at the surface to 5,000 m/s at depth. Examination of raypaths (Figure 7C) through
the tomographic model provides information about the regions of accuracy of the model. The dense clustering of raypaths down to 30–40-m depths and away from the ends of the survey line demonstrates that the model is likely well constrained in these regions. At greater depths and at the survey line termini, the raypaths become increasingly diffuse, suggesting that the model is less well determined in these domains. Overall, the tomographic model is considered to be a good representation of the subsurface to a depth of 30–40 m in the middle and north end of the model along the profile line. Accuracy is compromised on the south end of the profile as a result of (1) the lack of offset shots on the southern end of the SR profile, (2) the absence of geophones directly on exposed bedrock, and/or (3) the probable steep dip of the bedrock-talus contact, which makes refracted seismic energy less likely to pass into bedrock and back to the surface in measurable amounts. Thus, the SR tomographic model does not represent the southern extreme end of the profile accurately in the area of greatest interest. The SR tomographic model is consistent with published seismic velocities for different earth materials as well as with our own measurements of bedrock P-wave velocity. Seismic velocities for talus are expected to range from 100 to 4,600 m/s (Reynolds, 1997). The measured surface velocity of 387 m/s in the upper several meters of the profile line is consistent with materials such as dry unconsolidated soil and sand, as observed along the profile line surface during the survey. P-wave velocity values in the best-fit tomographic model increase with depth in most parts of the model to approximately 1,500– 2,400 m/s. This is typical of materials such as floodplain alluvium and is in good agreement with the surficial velocity values from Gutenberg et al. (1956) across the middle of Yosemite Valley. At the subsurface south end of the profile, the SR tomographic model approaches an average velocity of 5,000 m/s, which is within the acceptable range for the velocity associated with granites (4,600–6,200 m/s; e.g., West, 1995), velocities on granite measured in Yosemite Valley (5,250 m/s: Gutenberg et al., 1956; 5,900 m/s: Zimmer et al., 2012), and values derived from our independent surficial bedrock seismic velocity survey at the study site (4,840–5,971 m/s). Given published results and our observed seismic velocities, a talus-bedrock boundary would be identified in the seismic refraction data by the presence of a strong velocity gradient. This is the case because by its very nature, the tomographic model produces continuously varying velocities with no velocity discontinuities. Accordingly, we identified several
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
119
Brody, Pluhar, Stock, and Greenwood
Figure 7. (A) Compiled travel time curves for seismic refraction survey line. (B) Color tomographic seismic refraction model. (C) Monochrome tomographic model with raypaths (colored lines).
120
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
Near-Surface Geophysical Imaging of Talus Deposit
features of the tomographic model. There is a strong velocity gradient (between 1,344 and 4,217 m/s) within the southern segment of the profile near the cliff face (between 0 and 40 m along the survey profile) that is of the approximate range to be a talusbedrock boundary dipping steeply northward toward the valley. This gradient separates relatively high seismic velocities (4,478–4,999 m/s) of the magnitude of bedrock from lower velocities (1,344–2,911 m/s) of the magnitude of talus. Despite the stated accuracy problems in the tomographic model in this region, we consider it probable that this feature represents the talus-bedrock contact for two reasons. First, the surface location of the talus-bedrock contact at the top of the talus slope is known, and second, this strong velocity gradient dips 60u toward the valley, similar to the 60u–65u slope of the exposed cliff above the talus slope. There is another strong seismic velocity gradient between 822 and 1,876 m/s present within the middle portion of the profile (between 120 and 160 m horizontal position along the profile). This feature appears to decrease in depth toward the north and then flattens out into the valley, similar to the prominent GPR reflectors in this region. The position at the ground surface and velocity difference across this feature are consistent with the interface between talus and glacio-fluvial sediment fill. This feature is corroborated by both of the other geophysical methods employed. 2DR Results The RMS value indicates the amount of data misfit in the inverted resistivity section. While an RMS error value of ,5 percent is ideal for processing, RMS values of ,10 percent are deemed acceptable for these data, per the recommendation of Advanced Geosciences, Inc. (AGI). The overall RMS error for our 2DR model was 9.82 percent, while repeat measurement errors on individual data points were ,2 percent in nearly all cases. Measured voltage values were ..1 mV in nearly all cases, while injected currents were fairly low at several mA. These indicators suggest that for this lownoise location survey results are robust, despite very high contact resistances of thousands of Ohms for the survey hardware. Evaluation of the 2DR data reveals very strong variations in resistivity values along the profile (Figure 8A). At the surface of the southern end of the profile, resistivity values range in the tens of thousands to more than 100,000 V-m, while the nearsurface section of the northern end of the profile range from ,40 to a few thousand V-m. The most prominent feature in the 2DR is the nearly horizontal
boundary along the 45–165-m section of the survey line, separating resistivity values in the tens or hundreds of thousands of V-m near the ground surface (red, orange, and yellow on Figure 8A) from values in the thousands of V-m below that (greens and yellows on Figure 8A). Although mainly horizontal, this resistivity boundary shallows northward toward the ground surface at 150–180 m along the survey profile. The high resistivity values at the southern end of the survey profile (Figure 8A) are consistent with dry talus observed at the ground surface. There is no distinct talus-bedrock contact identified here in the 2DR, partly as a result of the impossibility of collecting surface data across the talus-bedrock boundary and partly because of the steep dip of the talus-bedrock contact. This boundary is simply not within the model space of the inversion. Consequently, the 2DR survey did not permit significant imaging of the talus-bedrock contact. The middle portion of the profile exhibits a wide variation in resistivity values, ranging from ,2,000 to ,6,000 V-m at depth to ,6,000 to .100,000 V-m near the surface. The lower values deeper in the profile are indicative of materials such as low-resistivity, moisture-retaining fines and clay, moist sand, and gravel up to intermediate-resistivity dry sand. The high values above this boundary are consistent with moreporous and drier higher-resistivity talus near the ground surface. The nearly horizontal boundary between medium and high resistivity values in the middle of the profile corresponds to the GPR reflectors in the same area (Figure 9A). Therefore, this strong contrast in resistivities is interpreted as the basal contact of talus against glacio-fluvial sediment fill. The low-resistivity (blue) region in the subsurface (Figures 8A and 9A) northern end of the 2DR profile could result from groundwater or moist fine-grained sediments. DISCUSSION In previous studies that imaged talus-bedrock contacts using GPR (Sass and Wollny, 2001; Otto and Sass, 2006; and Sass, 2006, 2007), two different approaches were used to locate the boundary. In some locations marked contrasts in dielectric constant between bedrock and talus produce distinct GPR reflections (e.g., Sass, 2006). In other cases, these materials have similar dielectric constants, such that the bedrock surface is noticeable as a boundary between talus showing distinct internal reflectors and bedrock that does not (e.g., Sass, 2007). Which of the cases will be displayed is dependent upon whether the bedrock is massive, without extensive jointing or
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
121
Brody, Pluhar, Stock, and Greenwood position from cliff face (meters)
Elevation above mean sea level (m)
SSW ephemeral stream crossings
NNE
surficial talus edge
ephemeral stream crossings
A. Inverted Resistivity Section Iteration = 4 RMS = 9.82% L2 = 0.71 Electrode Spacing = 6 m
predicted apparent resistivity
B. Predicted Versus Measured Apparent Resistivity Crossplot
measured apparent resistivity (ohm-m)
Figure 8. 2DR results. (A) Inverted resistivity section. (B) Cross plot of predicted versus measured apparent resistivities.
bedding, and whether there is a marked contrast in dielectric properties of the two contacting materials. Our data exhibit a bedrock surface showing a clear GPR reflection, but they equivocally display a bedrock surface marked by a lack of internal GPR reflections. The uppermost steeply north-dipping distinguishable reflector on the south end of the GPR survey (closest to the cliff face) is interpreted to be the basal contact of talus against crystalline bedrock (between positions 0 and 45 m at travel times of 0–1,100 ns). Furthermore, an additional, parallel reflector is interpreted to be a surface parallel sheeting joint, which is a common feature on the Glacier Point Apron. Fractures and joints in bedrock have been successfully imaged in multiple studies (Toshioka et al., 1995; Sass and Wollny, 2001; Porsani et al., 2006; and Sass, 2006). The dip of the corrected GPR feature (52u–64u from horizontal) is similar to the local cliff angle, which supports the idea of a bedrock reflector. This reflector is clearly not an airwave reflecting off of the cliff face, and a ground
122
wave bouncing off of the cliff face is also unlikely, since there are two parallel reflectors evident in the GPR at this location. On the other hand, the lack of GPR reflectors along profile line positions 190–235 m and .10 m in depth is consistent with Sass’s (2007) method for identifying bedrock. However, this zone could instead signify strong GPR attenuation due to groundwater, clayey lithologies, etc. Correspondence of multiple geophysical methods is necessary for accurate interpretation in this region and will be discussed below. The series of GPR reflectors evident in the middle portion of the profile (between positions 110 and 170 m) probably signifies the boundary of talus with glacio-fluvial sediment fill. However, it is uncertain whether the top or bottom of this series of reflectors represents the base of talus. Previous work (Otto and Sass, 2006; Sass, 2006, 2007) demonstrates the presence of internal reflectors in both rock-fall talus and debris avalanche deposits. Additional uncertainty stems from the apparent southward dip of the
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
Near-Surface Geophysical Imaging of Talus Deposit Ω-m
A
1X105
14132
elevation above sea level (meters)
1997
282
39.9
SSW
position from cliff face (meters)
NNE B
4999 4739 4478 4217 3956 3604 3433 3172 2911 2650 2389 2128 1867 1606 1344 1083 622 561 300 m/s
Figure 9. Geophysical data overlays. In both figures, the red dotted line represents the corrected position of the bedrock-talus contact and the glaciofluvial valley fill-talus contact, identified from GPR. See text for further explanation. (A) 2DR overlay on 50-MHz GPR. (B) Seismic refraction tomographic inversion overlay on 50-MHz GPR.
reflectors. This could be explained by the progressive onlap of rock-fall debris onto aggrading post-glacial fluvial deposits in the valley (Figure 6) or an inaccurate choice for the local radar wave velocity for this part of the profile. At locations 175–190 m along the profile, the talus deposit ends and there is indication of ,5 m of aggradation (Figure 4). This talus apron edge in the GPR is consistent with that identified in the high-resolution LiDAR-derived DEM. Elsewhere in Yosemite Valley, there is evidence of approximately 5 to 7 m of aggradation since deglaciation (Cordes et al., 2013), rendering these results mutually supportive. As previously stated, limitations of the SR profile geometry reduced the ability to accurately image the southern end of the profile. This explains the absence in the SR model of known 5,000+ m/s bedrock at the surficial extreme southern end of the profile. Despite this, the SR survey provided subsurface constraints on the basal contact of talus against crystalline bedrock. Our data suggest that the strong, steeply dipping, seismic velocity gradient in the southern portion of the
profile line closest to the cliff face, ranging from 1,344 to 4,217 m/s, likely represents this boundary. Another important feature of the SR tomographic model is the presence of low-velocity material thinning northward toward the valley. The velocities of this triangular-shaped body (in cross section) are consistent with talus. If the lower boundary of this body is taken to be 1,342–1,724 m/s then it corresponds to the strong, stratified GPR reflectors at profile line positions 130–190 m and depths around 5–10 m. Here, the feature appears to ramp up northward toward the ground surface. The edge of the talus deposit at the ground surface is known to be at profile position ,175 m. Therefore, it can be inferred that low-velocity talus material to the south indistinguishably grades directly into surficial, unconsolidated, post-glacial, fluvial sediment fill to the north along the profile line (Figure 9A). Overall, the interpretations of these prominent features in the tomographic model suggest a triangular-shaped body of the main talus deposit (Figure 7).
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
123
Brody, Pluhar, Stock, and Greenwood
The 2DR survey (Figure 8) yielded results that are broadly consistent with those of the other geophysical methods. In the Swiss Alps, Sass (2007) determined the location of bedrock based on the presence of a strong electrical contrast between the talus material and the bedrock. In addition, he suggested that it is impossible to assign a resistivity value to the bedrock interface because of the smooth contrasts and variation in resistivity of the bedrock itself. Therefore, it can be difficult to identify the bedrock-talus interface based on 2DR data alone. Within the inverted resistivity section, there are few indications of the basal contact of talus at the southern end of the section, and the bedrock-talus contact likely lies outside the inversion model space. Interpretation Based on Multiple Geophysical Data Sets The subsurface elevation of the basal contact of talus against crystalline bedrock or glacio-fluvial sediment fill was obtained from comparison of the GPR, SR, and 2DR processed data (Figure 9). Overlaying the SR or 2DR sections at 50 percent transparency on top of the 50-MHz GPR section highlights similarities in the results, leading to a high confidence in the processed geophysical data. In the combined SR-GPR image (Figure 9B) the SR velocity gradient along the southern end of the profile (closest to the cliff face) roughly corresponds to the corrected dipping reflector in the GPR section. In addition, strong correlation is also evident farther north along the survey profile line into the valley (profile position 120–185 m), where there is a southward dip of the pronounced velocity gradient in the SR model and similarly trending reflectors in the GPR section (Figure 9B). In the 2DR-GPR overlay (Figure 9A), interpretation of the basal contact of talus with crystalline bedrock is difficult, since the location of the corrected dipping GPR reflector falls outside the zone of 2DR coverage. However, striking similarities are apparent between the 2DR and GPR results further north along the profile line toward the valley (Figure 9A). Here, the orange-green boundary in the resistivity model correlates very well with the GPR reflectors, though it does not dip southward as clearly. The difference in dip of the feature could result from GPR imaging lithologic features, while resistivity revealed groundwater/moisture contrasts. Alternatively, this difference may arise from the user choice of radar wave velocity, but since the SR agreed well with the GPR, it is difficult to ascribe the difference to GPR processing choices alone. The north end of the survey line (profile positions 190–235 m) shows a strong corre-
124
spondence between low resistivity (blue color) and the zone of no internal GPR reflectors (,10 m below grade). A nearby borehole and groundwater investigation conducted in May 2012 (National Park Service, 2013) indicates that the groundwater table was located at about 1,206–1,208 m in elevation in the region around profile positions 190–240 m. This elevation is very similar to that of the featureless zone in the GPR imaged during October 2009, making radar wave attenuation a likely explanation for the GPR. Similarly, the low resistivity in the same part of the profile is readily explained by the presence of groundwater. If the GPR attenuation and low resistivity are ascribed to groundwater, this would imply relatively similar groundwater elevations at the times of data collection. This feature is absent from the SR tomography model because there were no geophones in the region of offset shots at 200–240 m along the profile line. As a result, any anomalous velocities in this region would be smeared out along raypaths further southward into the tomographic model. The former landfill lies upgradient of the lowresistivity GPR-attenuation zone. Sampling results in the vicinity indicate no unusual solutes in the groundwater (National Park Service, 2013), excluding the landfill as a possible source for observed features in this part of the geophysical profile. In short, the GPR attenuation and low-resistivity zone is readily explained by groundwater. Based on our geophysical data, the best interpretation of the overall talus geometry is that it reaches a maximum of ,40 m in thickness at about 50–80 m from the bedrock face at the south end of the survey line (Figure 9). The talus pinches out against bedrock at the south end of the survey line in the GPR data, in accordance with surface observations. The northward termination of talus at about 180 m along the survey line is evident at the ground surface and is consistent with the geophysical data sets (Figure 9). The 2DR shows this northward pinchout especially well (Figure 9A). Boreholes in the region north of the interpreted distal extent of talus encounter primarily glacio-fluvial sediment fill with only occasional large boulders, consistent with ‘‘outlier’’ boulders that are commonly observed beyond the edge of the active talus slope (Evans and Hungr, 1993). The apparent bedding in the GPR at survey locations 120–185 m likely results from either bedded glacio-fluvial sediment fill or coarse bedding in talus, but we cannot easily explain the southward dip if it is sediment fill. Alternatively, the apparent bedding in the GPR results could be restored to near horizontal with a different (slower) choice of radar velocity. However, this would also shallow the structure in general and would no longer agree with the boundaries also seen
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
Near-Surface Geophysical Imaging of Talus Deposit
within the SR and 2DR, making this alternative less appealing. Thus, we prefer the hypothesis that growth of the talus cone occurred simultaneously with aggradation of glacio-fluvial sediment fill, creating a south-dipping contact between sediment fill and talus. Optimization of Geophysical Surveys on Talus GPR GPR provided the most detailed subsurface information and was simplest to use in terms of field data acquisition and processing. Dry ground conditions at the time of the survey minimized radar attenuation, thereby optimizing our results. As was the case with other studies on talus (Sass and Wollny, 2001; Otto and Sass, 2006; and Sass, 2006, 2007), GPR provided good penetration depth and resolution of subsurface structures, allowing for a detailed interpretation. Multiple crossing GPR lines may have further improved the confidence of our interpretation, and we suggest this for future geophysical surveys on talus. One potential source of error is the requirement to choose an average radar velocity to convert from travel time to depth. Despite likely velocity variations throughout the talus deposit, an average velocity (representative of the various materials present) was applied to the processing as a result of the limitations of the processing software and available velocity structure information. SR The SR survey proved to be the most challenging and least diagnostic of the three geophysical methods. There were limited areas suitable for augering holes for Betsy SeisgunTM shot locations and significant physical restrictions in terms of the geometry of the SR survey. As a result of the location of the profile line against the bedrock cliff face, offset shots were not possible on the southern end of the SR survey. Imaging of the talus-bedrock interface without any shots or geophones on the bedrock side of this contact severely limited SR imaging of this interface. At a minimum, future work of this type should consider including geophones epoxied to the bedrock cliff face in order to constrain the cliff face boundary position and its high seismic velocity. Another shortcoming of the geometry of this survey resulted from the difficulty in achieving the angle of critical refraction with shots close to the cliff face. For seismic waves to be refracted, raypaths must approach the refracting boundary at an angle such that energy is refracted along the boundary. Thus, distant shots were required to achieve significant refracted energy, but this energy attenuates with
distance. For all of these reasons, it was technically challenging to image the bedrock-talus interface at depth at the south end of the survey line using SR. Another limitation of the SR result stemmed from processing software limitations. The SR processing software is designed for relatively simple layered geology, so that default tomographic models have their lowest velocity at all points at the ground surface and highest velocity at the maximum depth of the model. It was not possible to specify known seismic velocities as boundary conditions before tomographic inversion, such as our ,6,000-m/s bedrock, at the south end of the profile at the ground surface as well as at depth all along the southern end. 2DR The 2DR survey was the most rapid of the three geophysical methods. The electrodes were easy to position in the ground, but high contact resistances in this type of formation tended to reduce data quality. The necessity to wet each electrode location with salt water was somewhat cumbersome in this terrain. The possibility of imaging the bedrock-talus contact at depth was prevented by the steep dip of the contact and the inability to set the 2DR survey across the surface expression of the bedrock-talus boundary. On the other hand, the 2DR result provided excellent corroboration with GPR imaging of the talussediment fill contact. This suggests that combining 2DR and GPR may be ideal for geophysical surveying in deposits and geometries similar to this study area. CONCLUSIONS This study applied near-surface geophysical methods to map the extent of subsurface talus in a region of active accumulation by rock fall. These data are helpful for quantifying the total volume of talus beneath Glacier Point, to supplement the historical record of rock falls, and to put modern process rates into context (Brody, 2011), but the goal of this study was to evaluate the feasibility of using geophysics to map the subsurface in these materials. To constrain the basal contact of talus against crystalline bedrock or glacio-fluvial sediment fill, GPR, SR, and 2DR were used to define this interface. Of the three nearsurface geophysical techniques utilized, GPR provided the most detailed image of the subsurface of the talus deposit (Figure 4). 2DR produced strong contrasts between talus and valley fill but could not resolve the talus-bedrock interface, as a result of its steep dip at the edge of the survey line. The physical geometry of rock-fall–generated talus cones is not
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
125
Brody, Pluhar, Stock, and Greenwood
amenable to the best practices of SR surveys. The lack of offset shots in the region of greatest interest— near the talus-bedrock contact—strongly limits SR applicability in some regions. Nonetheless, there were strong congruencies among the SR, GPR, and 2DR data. Overlaying the SR or 2DR results over the GPR section strengthened the interpretation of a basal contact of talus against crystalline bedrock and glacio-fluvial valley fill, with a maximum deposit thickness of approximately 45 m (Figure 9). At the survey line location, the bedrock-talus contact dips at about the same angle as the Glacier Point Apron surface. The contact between talus and glacio-fluvial sediment fill seems to dip southward, suggesting an onlapping relationship with sediment fill as the valley floor aggraded about 5 m after deglaciation (Figure 6). The surface location of the edge of talus agrees with the predicted location from the geophysics. The results of this study suggest that GPR and 2DR may be sufficient to accurately image subsurface contacts in geologic materials such as these. ACKNOWLEDGMENTS We gratefully acknowledge Horacio Ferriz (California State University, Stanislaus), Greg Johnston (Sensors & Software, Inc.), Craig Lippus (Geometrics, Inc.), and Brad Carr (Advanced GeoSciences, Inc.) for technical advice and assistance. Horacio Ferriz also graciously provided the GPR and resistivity equipment. Chad Carlson, Joey Luce, Preston Ward, Wayne Nick, and Dustin Smith assisted with field work. Andy Shriver and Robin Trayler assisted with GIS and figure preparation, respectively. Jerry DeGraff and John Wakabayashi provided valuable project advisement and review of early manuscript drafts, and the final manuscript benefited from valuable input from three anonymous reviewers. This research would not have been possible without grants from the California State University, Fresno College of Science and Mathematics (FacultySponsored Student Research Grant), the Association of Environmental and Engineering Geologists (Sacramento Section student research grant and Norman R. Tilford Field Studies Scholarship), the Geological Society of America (Roy J. Shlemon Scholarship), and the Northern California Geological Society (Richard Chambers Memorial Scholarship). REFERENCES ADVANCED GEOSCIENCES, INC., 2013, Instruction Manual for EarthImager 2D Version 2.1.7 Resistivity and IP Inversion Software: Geometrics, Inc. Austin, TX.
126
BATEMAN, P. C., 1992, Plutonism in the Central Part of the Sierra Nevada Batholith, California: U.S. Geological Survey Professional Paper 1483, 186 p. BRODY, A. G., 2011, Quantifying Post-Glacial Talus Deposition in Yosemite Valley Using ARCGIS and Near-Surface Geophysics, and the Implications on Hazard Assessment: M.S. Thesis, California State University, Fresno, 169 p. BRONSON, B. R. AND WATTERS, R. J., 1987, The effects of long-term slope deformations on the stability of granitic rocks of the Sierra Nevada, Nevada, California. In Proceedings of the 23rd Engineering Geology and Soils Engineering Symposium: Utah State University, Logan, UT, pp. 203–217. BURSIK, M. I. AND GILLESPIE, A. R., 1993, Late Pleistocene glaciation of Mono Basin, California: Quaternary Research, Vol. 39, pp. 24–35. CORDES, S. E.; STOCK, G. M.; SCHWAB, B. E.; AND GLAZNER, A. F., 2013, Supporting evidence for a 9.6 6 1 ka rock fall originating from Glacier Point in Yosemite Valley, California: Environmental Engineering Geoscience, Vol. 19, pp. 345– 361. DUSSAUGE, C.; GRASSO, J.-R.; AND HELMSTETTER, A., 2003, Statistical analysis of rockfall volume distributions: Implications for rockfall dynamics: Journal Geophysical Research, Vol. 108, pp. 2–1–2-11. DUSSAUGE-PEISSER, C.; HELMSTETTER, A.; GRASSO, J.-R.; HANTZ, D.; DESVARREUX, P.; JEANNIN, M.; AND GIRAUD, A., 2002, Probabilistic approach to rock fall hazard assessment: Potential of historical data analysis: Natural Hazards Earth System Sciences, Vol. 2, pp. 15–26. EVANS, S. G. AND HUNGR, O., 1993, The assessment of rockfall hazard at the base of talus slopes: Canadian Geotechnical Journal, Vol. 30, pp. 620–636. FISHER, S. C.; STEWART, R. R.; AND JOL, H. M., 1992, Processing ground penetrating radar (GPR) data: GPR Processing CREWES Research Report, Vol. 4, pp. 11-1–11-22. GEOMETRICS, 2006. SeisImager/2D Manual Version 3.2: Pickwin v. 3.2 and Plotrefa v. 2.8: Advanced Geoseiences. Inc. San Jose, CA. GUTENBERG, B.; BUWALDA, J. P.; AND SHARP, R. P., 1956, Seismic explorations on the floor of Yosemite Valley, California: Bulletin Geological Society America, Vol. 87, pp. 1051–1078. GUZZETTI, F.; REICHENBACH, P.; AND WIECZOREK, G. F., 2003, Rockfall hazard and risk assessment in the Yosemite Valley, California, USA: Natural Hazards Earth System Sciences, Vol. 3, pp. 491–503. HOFFMANN, T. AND SCHROTT, L., 2003, Determining sediment thickness of talus slopes and valley fill deposits using seismic refraction—a comparison of 2D interpretation tools: Zeitschrift Geomorphologie N.F. Supplement, Vol. 132, pp. 71–87. HUBER, N. K., 1981, Amount and Timing of Late Cenozoic Uplift and Tilt of the Central Sierra Nevada, California—Evidence from the Upper San Joaquin River Basin: U.S. Geological Survey Professional Paper 1197, 28 p. HUBER, N. K., 1987, The Geologic Story of Yosemite National Park: U.S. Geological Survey Bulletin 1595. JOL, H. M., 1995, Ground penetrating radar antennae frequencies and transmitter powers compared for penetration depth, resolution and reflection continuity: Geophysical Prospecting, Vol. 43, pp. 693–709. LOKE, M. H., 2000, Electrical Imaging Surveys for Environmental and Engineering Studies—A Practical Guide to 2D and 3D Surveys: Unpublished Short Training Course Notes, Penang, Malaysia, Universiti Sains Malaysia, 61 p. MATASCI, B.; CARREA, D.; JABOYEDOFF, M.; PEDRAZZINI, A.; STOCK, G. M.; AND OPPIKOFER, T., 2011, Structural characterization of rockfall sources in Yosemite Valley from remote sensing
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
Near-Surface Geophysical Imaging of Talus Deposit data: Toward more accurate susceptibility assessment: Proceedings of the 14th Pan-American Conference on Soil Mechanics and Geotechnical Engineering Conference: Toronto, Canada, October 2011. MATTHES, F. E., 1930, Geologic History of the Yosemite Valley: U.S. Geological Survey Professional Paper 160, 225 p. NATIONAL PARK SERVICE, 2013, Curry Landfill Site Remedial Investigation/RCRA Facility Investigation (RI/RFI) Report: Yosemite National Park, Yosemite, Ca, October 2013, 282 p. NEAL, A., 2004, Ground-penetrating radar and its use in sedimentology: Principles, problems and progress: Earth Science Reviews, Vol. 66, pp. 261–330. OTTO, J. C. AND SASS, O., 2006, Comparing geophysical methods for talus slope investigations in the Turtman valley (Swiss Alps): Geomorphology, Vol. 76, pp. 257–272. PECK, D. L., 2002, Geologic Map of the Yosemite Quadrangle, Central Sierra Nevada, California: U.S. Geological Survey Geologic Investigations Series Map I-2751, scale 1:62,500. PHILLIPS, F. M.; ZREDA, M.; PLUMMER, M. A.; ELMORE, D.; AND CLARK, D. H., 2009, Glacial geology and chronology of Bishop Creek and vicinity, eastern Sierra Nevada, California: Geological Society America Bulletin, Vol. 121, pp. 1013–1033. PORSANI, J. L.; SAUCK, W. A.; AND JUNIOR, A. O. S., 2006, GPR for mapping fractures and as a guide for the extraction of ornamental granite from a quarry: A case study from southern Brazil: Journal Applied Geophysics, Vol. 58, pp. 177–187. REYNOLDS, J. M., 1997, An Introduction to Applied and Environmental Geophysics: Wiley, New York, 796 pp. SASS, O., 2006, Determination of the internal structure of alpine talus deposits using different geophysical methods (Lechtaler Alps, Austria): Geomorphology, Vol. 80, pp. 45–58. SASS, O., 2007, Bedrock detection and talus thickness assessment in the European Alps using geophysical methods: Journal Applied Geophysics, Vol. 62, pp. 254–269. SASS, O. AND WOLLNY, K., 2001, Investigations regarding alpine talus slopes using ground-penetrating radar (GPR) in the Bavarian Alps, Germany: Earth Surface Processes Landforms, Vol. 26, pp. 1071–1086. SENSORS & SOFTWARE, INC., 1999a, Ground Penetrating Radar Survey Design: Mississauga, ON, Canada. SENSORS & SOFTWARE, INC., 1999b, Practical Processing of GPR Data: Mississauga, ON, Canada. SENSORS & SOFTWARE, INC., 2006, pulseEKKO PRO User’s Guide: Mississauga, ON, Canada. SMITH, D. G. AND JOL, H. M., 1995, Ground penetrating radar: Antenna frequencies and maximum probably depths of penetration in Quaternary sediments: Applied Geophysics, Vol. 33, pp. 93–100. SMITH, S. J. AND ANDERSON, R. S., 1992, Late Wisconsin paleoecologic record from Swamp Lake, Yosemite National Park, California: Quaternary Research, Vol. 38, pp. 91–102. STOCK, G. M.; BAWDEN, G. W.; GREEN, J. K.; HANSON, E.; DOWNING, G.; COLLINS, B. D.; BOND, S.; AND LESLAR, M., 2011, High-resolution three-dimensional imaging and analysis
of rock falls in Yosemite Valley, California: Geosphere, Vol. 7, pp. 573–581, doi:10.1130/GES00617.1. STOCK, G. M.; COLLINS, B. D.; SANTANIELLO, D. J.; ZIMMER, V. L.; WIECZOREK, G. F.; AND SNYDER, J. B., 2013, Historical Rock Falls in Yosemite National Park, California (1857–2011): U.S. Geological Survey Data Series 746, 17 p. and data files: http://pubs.usgs.gov/ds/746. STOCK, G. M. AND UHRHAMMER, R. A., 2010, Catastrophic rock avalanche 3600 years BP from El Capitan, Yosemite Valley, California: Earth Surface Processes Landforms, Vol. 35, pp. 941– 951. TOSHIOKA, T.; TSUCHIDAA, T.; AND SASAHARA, K., 1995, Application of GPR to detecting and mapping cracks in rock slopes: Journal Applied Geophysics, Vol. 33, pp. 119–124. WAHRHAFTIG, C. AND BIRMAN, J. H., 1965, The Quaternary of the Pacific Mountain System in California. In Wright, H. E., Jr. and Frey, D. G. (Editors), The Quaternary of the United States—A Review Volume for the 7th Congress of the International Association for Quaternary Research: Princeton University Press, Princeton, NJ, pp. 299–340. WAKABAYASHI, J. AND SAWYER, T. L., 2001, Stream incision, tectonics, uplift, and evolution of topography of the Sierra Nevada, California: Journal Geology, Vol. 109, pp. 539–562. WEST, T. R., 1995, Geology Applied to Engineering: Long Grove, IL, Waveland Press, 560 p. WIECZOREK, G.; NISHENKO, S. P.; AND VARNES, D. J., 1995, Analysis of rock falls in the Yosemite Valley, California. In Daemen, J. J. and Schultz, R. A. (Editors), 35th US Symposium on Rock Mechanics: A. A. Balkema, Daemen, pp. 85–89. WIECZOREK, G. F. AND JA¨GER, S., 1996, Triggering mechanisms and depositional rates of postglacial slope-movement processes in the Yosemite Valley, California: Geomorphology, Vol. 15, pp. 17–31. WIECZOREK, G. F.; MORRISSEY, M. M.; IOVINE, G.; AND GODT, J., 1999, Rock-Fall Potential in the Yosemite Valley, California: U.S. Geological Survey Open File Report 99-578. WIECZOREK, G. F. AND SNYDER, J. B., 1999, Rock Falls from Glacier Point above Camp Curry, Yosemite National Park, California: U.S. Geological Survey Open-File Report 99-385. WIECZOREK, G. F.; SNYDER, J. B.; WAITT, R. B.; MORRISSEY, M. M.; UHRHAMMER, R. A.; HARP, E. L.; NORRIS, R. D.; BURSIK, M. I.; AND FINEWOOD, L. G., 2000, Unusual July 10, 1996 rock fall at Happy Isles, Yosemite National Park, California: Geological Society America Bulletin, Vol. 112, pp. 75–85. WIECZOREK, G. F.; STOCK, G. M.; REICHENBACH, P.; SNYDER, J. B.; BORCHERS, J. W.; AND GODT, J. W., 2008, Investigations and hazard assessment of the 2003 and 2007 Staircase Falls, rock falls, Yosemite National Park, California, USA: Natural Hazards Earth System Sciences, Vol. 8, pp. 421–432. ZIMMER, V. L.; COLLINS, B. D.; STOCK, G. M.; AND SITAR, N., 2012, Rock fall dynamics and deposition: An integrated analysis of the 2009 Ahwiyah Point rock fall, Yosemite National Park, USA: Earth Surface Processes Landforms, Vol. 37, pp. 680– 691, doi:10.1002/esp.3206.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 111–127
127
Bluff Recession in the Elwha and Dungeness Littoral Cells, Washington, USA DAVID S. PARKS1 Washington State Department of Natural Resources, 311 McCarver Road, Port Angeles, WA 98362
Key Terms: Environmental Geology, Land-Use Planning, Erosion, Landslides
ABSTRACT The spatial distribution and temporal variability of retreat rates of coastal bluffs composed of unconsolidated glacial deposits are of interest to landowners who occupy bluff-top properties as well as coastal resource managers who are responsible for protecting marine habitats such as forage fish spawning beaches that are dependent on bluff-derived sediments. Assessment of bluff retreat and associated sediment volumes contributed to the nearshore over time is the first step toward development of a coastal sediment budget for bluff-backed beaches using data sources including aerial photography (1939, 2001), GPSbased beach profile data (2010–2013), and airborne LiDAR (2001, 2012). These data are analyzed in context to determine alongshore rates of bluff retreat and associated volume change for the Elwha and Dungeness littoral cells in Clallam County, WA. Recession rates from 2001 to 2012 range from 0 to 1.88 m/yr in both drift cells, with mean values of 0.26 ± 0.23 m/yr (N = 152) in Elwha and 0.36 ± 0.24 m/yr (N = 433) in Dungeness. Armored sections show bluff recession rates reduced by 50 percent in Elwha and 80 percent in Dungeness, relative to their respective unarmored sections. Dungeness bluffs produce twice as much sediment per alongshore distance as do the Elwha bluffs (average, 7.5 m3/m/yr vs. 4.1 m3/m/ yr, respectively). Historical bluff recession rates (1939– 2001) were comparable to those from 2001–2012. Rates derived from short timescales should not be used directly for predicting decadal-scale bluff recession rates for management purposes, as they tend to represent shortterm localized events rather than long-term sustained bluff retreat. INTRODUCTION Coastal bluffs are a dominant geomorphic feature of the shorelines of the Strait of Juan de Fuca, 1
Corresponding author email: david.parks@dnr.wa.gov.
Washington State, USA, and are the primary source of sediment contributed to mixed sand and gravel beaches in the region (Schwartz et al., 1987; Shipman, 2004; Finlayson, 2006; and Johannessen and MacLennan, 2007). The spatial and temporal distribution of bluff recession from wave-, wind-, precipitation-, and groundwater-induced erosion is poorly understood and documented for the southern shore of the Strait of Juan de Fuca and has led to underestimating the potential hazards to infrastructure (e.g., roads, houses) posed by eroding bluffs over time (Figures 1 and 2). Efforts to protect infrastructure and limit the rates of bluff erosion by constructing shoreline revetments have historically ignored the physical and ecological effects of sediment starvation of beaches caused by shoreline hardening (Shipman et al., 2010). The disruption of sediment movement from bluffs to beaches has caused the loss of suitable habitats for critical marine species, including forage fish and juvenile salmonids (Rice, 2006; Shipman et al., 2010; Shaffer et al., 2012; and Parks et al., 2013). The importance of understanding the long-term littoral sediment budget has been underscored by the recent removal of two dams on the Elwha River and the subsequent introduction of approximately 6.4 3 106 m3 of sediment into the nearshore environment within the first 2 years (between September 2011 and September 2013) (East et al., 2014; Gelfenbaum et al., in review; and Warrick et al., in review). Relatively few studies of coastal bluff recession have been completed for the shoreline areas of the Strait of Juan de Fuca, and the studies that have been completed have used a variety of methods, leading to difficulty in comparing results. In the Elwha littoral cell (herein referred to as ‘‘drift cell’’), the U.S. Army Corps of Engineers (USACE) completed an evaluation of bluff recession rates and sediment volume supply to the nearshore environment as part of an environmental assessment for a shoreline armoring and beach nourishment project on Ediz Hook in Port Angeles (USACE, 1971). Using Government Land Office and National Geodetic Survey shoreline maps, the USACE estimated a gradual reduction in bluff recession rates from 1.5 m/yr (1850–1885) to 1.3 m/yr
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
129
Parks
Figure 1. (A) Homes threatened by receding bluffs, Dungeness drift cell. (B) Seawall installed at bluff toe to protect Port Angeles City Landfill from bluff retreat, Elwha drift cell.
(1885–1926), decreasing to 1.1 m/yr (1926–1948) and then to 0.2 m/yr (1948–1970). Each successive reduction in bluff recession rates since 1930 has been attributed to construction and maintenance of a
130
multitude of shoreline armoring projects at the base of the Elwha bluffs (USACE, 1971). The USACE (1971) study also shows a reduction in sediment volumes provided by the Elwha bluffs over
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
Coastal Bluff Recession
Figure 2. Map of the study area showing direction of net alongshore sediment transport within the Elwha and Dungeness drift cells in Clallam County, WA.
time. Prior to the construction of the Elwha Dam in 1911, the estimated sediment supply from the bluffs was 2.22 3 105 m3/yr. After construction of the Elwha Dam and prior to construction of shoreline armoring along the Elwha bluffs in 1929, the estimated sediment supply from the bluffs was nearly the same, measuring 2.06 3 105 m3/yr. Between 1929 and 1961, when substantial shoreline armoring along the bluffs was installed and maintained, the bluff sediment supply decreased to 0.73 3 105 m3/yr. Following the completion of a major shoreline armoring project along the bluffs in 1961, bluff sediment supply was estimated to have further declined to 0.31 3 105 m3/ yr. The reduction of bluff-supplied sediment over this entire time period, 1.91 3 105 m3/yr, represents an 85 percent reduction in the coastal sediment supply to Ediz Hook (Galster, 1989), which is essentially equivalent to the pre-dam fluvial sediment supply estimated by Randle et al. (1996). Bluff erosion rates to the east of the Dungeness drift cell along the Strait of Juan de Fuca were evaluated through land-parcel surveys by Keuler (1988). Bluff recession rates of up to 0.30 m/yr and sediment production rates of 1–5 m3/m/yr were observed in areas exposed to wave attack associated with long fetches. On the west side of Whidbey Island, at the eastern limit of the Strait of Juan de Fuca, Rogers et al. (2012) determined long-term bluff
erosion rates of 0–0.08 m/yr using cosmogenic 10Be concentrations in lag boulders to date shoreline positions over time scales of 103–104 years. In this study, estimates of short- and long-term bluff recession rates and associated sediment volumes contributed to the Elwha and Dungeness drift cells along the Central Strait of Juan de Fuca between 1939 and 2012 are derived from historical aerial photography, GPS beach profiles, and airborne LiDAR, and the relative contribution of bluff-derived sediment supply to the nearshore, in the context of a coastal sediment budget recently rejuvenated by the removal of two dams on the Elwha River, is presented. STUDY AREA The study area is located on the southern shore of the Central Strait of Juan de Fuca near the city of Port Angeles, WA (Figure 2). The study area is divided into two distinct shoreline segments that encompass separate but adjacent littoral cells with bluff-backed beaches: the Elwha bluffs extend along the central portion of the Elwha drift cell, and the Dungeness bluffs extend along the western portion of the Dungeness drift cell (Figure 3). Each drift cell contains an updrift segment of eroding coastal bluffs to the west that supply sediment via longshore littoral
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
131
Parks
Figure 3. (A) Photograph of the Dungeness bluffs looking west from Dungeness Spit. (B) Photograph of the Elwha bluffs west from Ediz Hook. Note the armoring placed mid-beach in front of the bluffs in photograph B.
transport to long spits at the down-drift end to the east. The Elwha bluff segment is 4.9 km long and supplies sediment to Ediz Hook. The Dungeness bluff segment is 13.6 km long and supplies sediment to
132
Dungeness Spit. A fundamental difference between the two drift cells is that the Elwha River discharges into the Strait of Juan de Fuca updrift of the Elwha bluffs, while the Dungeness River empties into the Strait of Juan de Fuca on the lee side of Dungeness
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
Coastal Bluff Recession
Spit (Figure 2). Therefore, the Elwha drift cell is composed of both river- and bluff-derived sediments, while the Dungeness drift cell is composed of only bluff-derived sediments. The Strait of Juan de Fuca is a wind-dominated marine system that exhibits net easterly longshore sediment transport within the intertidal zone of the study area (Galster and Schwartz, 1989; Schwartz et al., 1989; Warrick et al., 2009; and Miller et al., 2011). Winds in the Central Strait of Juan de Fuca are dominantly west and northwesterly, with a minor component of north and northeasterly winds (Miller et al., 2011). Therefore, both the Elwha and Dungeness drift cells exhibit net easterly littoral sediment transport (USACE, 1971; Galster and Schwartz, 1989; and Schwartz et al., 1989). The wave climate of the Central Strait of Juan de Fuca is similarly dominated by west to northwest wind waves and west to northwest swells from the Pacific Ocean. Maximum wave heights within the study area range up to 3 m, whereas average heights are 0.5 m (USACE, 1971; Gelfenbaum et al., 2009; Warrick et al., 2009; and Miller et al., 2011). Gelfenbaum et al. (2009) have modeled the distribution of significant wave heights within the Central Strait of Juan de Fuca, and given a 2-m swell at the entrance to the Strait of Juan de Fuca, nearshore wave heights of 1 m are shown throughout the study area, but with significant alongshore variability in wave height due to wave focusing or sheltering and in wave direction due to refraction. Tides within the Strait of Juan de Fuca are mixeddiurnal, with two high and low tides per day. Tidal elevations range between 21.0 m and +3.7 m in elevation (NAVD 88) (Zilkoski et al., 1992; NOAA, 2013). A precipitation gradient exists from west to east within the study area as the result of a rain-shadow effect of the Olympic Mountains. Average annual precipitation (1971–2000) in the Elwha drift cell is 660 mm vs. 406 mm in the Dungeness drift cell (Drost, 1986; NCDC, 2014). Maximum rainfall intensities within the Elwha drift cell are 117 mm/hr vs. 71 mm/hr in the Dungeness (Drost, 1986; NCDC, 2014). Precipitation occurs primarily as rain, with the wettest months between October and April and a seasonal dry period between May and September. Freezing temperatures occur within the study area between October and May, and snowfall intermittently occurs in the period between November and April. The surficial geology of the study area is dominantly composed of Pleistocene continental glacial deposits overlying pre-Fraser non-glacial sediments associated with an Elwha River source (Schasse et al.,
2000; Polenz et al., 2004) and Eocene marine sedimentary rocks (Schasse et al., 2000; Schasse and Polenz, 2002; Schasse, 2003; and Polenz et al., 2004). Pleistocene glacial deposits occurring within the study area include recessional outwash, glaciomarine drift, and glacial till. Groundwater recharge occurs along the Olympic Mountains and discharges into the Strait of Juan de Fuca. Local groundwater recharge occurs within lowelevation glacial landforms adjacent to the coastal bluffs and discharges at varying elevations on the bluffs controlled by local aquitards (i.e., beds of lowpermeability materials composed of dense silt, clay, and till) (Drost, 1986; Jones, 1996). The shoreline within the study area exhibits steeply sloping to vertical and overhanging coastal bluffs up to 80 m high created by changes in relative sea level from post-glacial rebound following Cordilleran glacial retreat; erosion of the shoreline in the study area began around 5,400 years before the present time (Downing, 1983; Dethier et al., 1995; Booth et al., 2003; Schasse, 2003; Mosher and Hewitt, 2004; and Polenz et al., 2004). Bluff recession within the study area is dominated by shallow landsliding in the form of topples, debris avalanches, flows, and slides (Varnes, 1978). Other types of gravitational failures are also present, including stress release fracturing (Bradley, 1963), cantilever, and Culmann-type (near-vertical planar) failures (Carson and Kirkby, 1972). These types of shallow mass wasting processes are common in sea cliffs composed of weakly lithified sediments (Hampton, 2002). Aeolian erosion during dry periods (in the form of ravel) is also observed. Aerial-, boat-, and ground-based surveys of the study area have determined the absence of deep-seated (Varnes, 1978) landslides consistent with existing geologic mapping (Schasse et al., 2000; Schasse and Polenz, 2002; Schasse, 2003; and Polenz et al., 2004). Processes driving shallow landsliding include over-steepening and subsequent failure of bluffs from wave-induced erosion at the bluff-base and the development of high pore-water pressures within hillslopes during storms. Land use above the bluffs varies throughout the study area from dense urban development in the Elwha drift cell within the City of Port Angeles to native second-growth forest within the Dungeness drift cell. Vegetation within the study area ranges from dense stands of mature second- and thirdgrowth Douglas fir forest to open grass associated with urban lawn-scapes. The sediment budget of the Elwha drift cell has substantially declined as a result of human-induced changes. The construction of coastal revetments began in the Elwha drift cell shortly after the
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
133
Parks
construction of two dams on the Elwha River in the early 20th century (Galster, 1989). In 1929, a coastal revetment was installed between Dry Creek and Ediz Hook to protect an industrial waterline that supplied water from the Elwha River to paper mills on Ediz Hook. Within 6 years of the placement of coastal defense works, Ediz Hook began to erode as a result of the reduction in sediment supply from bluffs (Galster, 1989). Galster (1989) estimated that in the Elwha drift cell, 15 percent of the sediment supplying Ediz Hook originated from the Elwha River, and 85 percent was supplied from coastal bluff erosion prior to construction of Elwha River Dams and coastal revetments. Galster (1989) estimated that coastal armoring in the Elwha drift cell resulted in an 89 percent reduction of sediment volume supplied to Ediz Hook. In 1975, the USACE and the City of Port Angeles armored the shoreline of Ediz Hook and began a program of beach nourishment that continues to the current time. In 2005, the City of Port Angeles constructed a 122 m– long concrete, steel, and rock seawall at the Port Angeles Landfill. Currently, 68 percent of the Elwha bluffs are armored with rip-rap or constructed seawalls. In contrast, less than 1 percent of the length of the Dungeness bluffs is armored. In 2012, the Elwha Dam on the Elwha River was completely removed, and, as of 2014, the Glines Canyon Dam has also been completely removed, resulting in the delivery of 6.4 3 106 m3 of predominantly fine sediment to the nearshore of the Elwha littoral cell within the first 2 years since dam removal began in September 2011 (East et al., 2014; Gelfenbaum et al., in review; and Warrick et al., in review). This sediment volume represents approximately 30 percent of the total sediment stored in both reservoirs. It is estimated that within 7–10 years following the complete removal of both Elwha River Dams, the long-term annual sediment contribution from the Elwha River to the nearshore will be approximately 2.5 3 105 m3/yr (Gilbert and Link, 1995; Bountry et al., 2010). Understanding the relative contribution of bluff erosion to the overall sediment budget of the Elwha drift cell will help with efforts to manage the longterm coastal environment once the reservoir sediments released by dam removal have been transported out of the fluvial network and into the Strait of Juan de Fuca. METHODS Bluff-Face Change Mapping Short- and long-term coastal bluff recession rates for the Elwha and Dungeness drift cells were
134
determined by analyzing data from historical aerial photographs and existing airborne LiDAR data. In order to make comparisons of the bluffs between the two data types, two-dimensional cross-shore transects were established in each drift cell at 30-m intervals, except where interrupted by coastal streams or ravines (Figure 4). Transects extend across the beach and up the bluff face, to at least the bluff crest, along which retreat distances could be calculated. Bluff retreat was measured between consecutive surveys at the bluff crest for aerial photos and at selected elevations across the bluff face for LiDAR data. Long-Term Bluff Change Bluff recession rates for 1939–2001 were determined by calculating the distance between bluff crest positions on geo-referenced historical aerial photographs. Prior to analysis, aerial photographs were scanned, geo-referenced, and imported into ArcGIS v. 10.1 (ESRI, Redlands, CA), and bluff crest positions were digitized for study segment areas unobstructed by vegetation. Distances between the 1939 and 2001 bluff crest positions were measured at each transect location. Recession rates for 2001–2012 were determined from the differences in horizontal position of selected elevations on bluff-face profiles extracted from digital elevation models (DEMs) available from recent airborne LiDAR data sets using methods outlined in Hapke (2004), Young and Ashford (2009) and Young et al. (2009, 2010, 2011). For this analysis, we used a 2001 bare earth DEM (2-m grid) from the Puget Sound LiDAR Consortium (PSLC, 2001) that covered the entire survey area, 2012 Clallam County LiDAR (1-m grid; Yotter-Brown and Faux, 2012) for the Dungeness drift cell, and 2012 LiDAR data (0.5m grid) from the U.S. Geological Survey (Woolpert, 2013) for the Elwha drift cell. DEMs were imported into ArcGIS and evaluated using the 3D Analyst toolset. At each transect location a two-dimensional topographic profile from the mid-beach to the bluff crest was extracted from each DEM. The net horizontal distance between the two profiles was measured at 6-m vertical intervals between the bottom and top of the bluff face. The difference in total cross-sectional area between the 2001 and 2012 topographic profiles was measured and multiplied by a unit width to estimate a volume of sediment lost between the two DEMs. Statistical evaluation of the data for bluff recession and sediment volume contributions from the airborne LiDAR DEMs was performed using exploratory data analysis methods (Schuenemeyer and Drew, 2011). Bluff recession distance values were tested for spatial
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
Coastal Bluff Recession
Figure 4. Map showing bluff and beach transect locations for the Elwha bluffs (A) and Dungeness bluffs (B5west, C5east).
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
135
Parks
Figure 5. Maximum observed bluff recession rates (m/yr) in the Dungeness drift cell for the time periods 1939–2001 (derived from aerial photography) and 2001–2012 (derived from airborne LiDAR).
trend and normalized using a lognormal transformation. Summary statistics were then computed using the de-trended values. Sources of error include internal error in the LiDAR data acquisition and processing technique as well as differences in grid size of the LiDAR-derived DEMs.
1.5-m data point spacing. The difference values along the entire transect were averaged to yield a single value of average elevation change per transect. The average elevation change was multiplied by the 20-m length of the profile and an alongshore unit width of 1 m to yield a volume change per alongshore meter (m3/m) for the 20 m of upland beach.
Beach Profile Change Monitoring To assess general trends in beach elevation change (m/yr) and to estimate rates of sediment flux on the beaches (m3/m/yr), two-dimensional, cross-shore topographic beach profiles at 12 locations, eight along the Dungeness bluffs and four along the Elwha bluffs, were surveyed between 2010 and 2013 with a ProMark 800 and 200 Real-Time Kinematic Global Positioning System (RTK-GPS). Elwha and Dungeness drift cell beach profiles were collected in all seasons. Profiles were oriented normal to the slope of the beach, extending from the base of coastal bluffs to the low water limit. Elevation measurements were recorded along each transect at horizontal intervals of approximately 1.5 m. RTK-GPS measurement accuracy ranged from 1 to 5 cm based on repeat measurements of fixed control points across the study area. Sediment volume changes were calculated using the upper 20 m of each profile, which was the extent of overlap between all surveys. The elevation difference between each pair of profiles was calculated every 0.5 m, with a linear interpolation between the original
136
RESULTS Bluff-Face Change Long-Term Bluff Change Observed rates of coastal bluff recession are highly variable across both drift cells (Figures 5–7). Table 1 provides data results from sections of each drift cell with unobstructed views of the bluff edge in aerial photography from 1939 and 2001 and includes identical shoreline reaches used for a comparison of rates derived from airborne LiDAR from 2001 and 2012. The data show a recent decrease in mean recession rates in the Elwha drift cell (20.22 m/yr) and a slight increase in mean recession rates in recent years in the Dungeness drift cell (+0.1 m/yr). Table 2 provides data results that extend along the full length of the bluffs in each drift cell. The maximum observed rate of recession between 2001 and 2012 in both drift cells was 1.88 m/yr, associated with housing development in the Dungeness drift cell (Figure 1A) and erosional hotspots along the Port
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
Coastal Bluff Recession
Figure 6. Maximum observed bluff recession rates (m/yr) in the Elwha drift cell for the time periods of 1939–2001 (derived from aerial photography) and 2001–2012 (derived from airborne LiDAR).
Figure 7. Box plot of recession rates (m/yr) by drift cell and shoreline type (created in ABOXPLOT; Bikfalvi, 2012). The central line within the box represents the sample median, while the circle represents the sample mean. The upper and lower limits of the box represent the 50th percentile of the population and the whiskers the 75th percentile. Dots beyond the upper and lower whiskers represent outliers of the population.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
137
Parks Table 1. Recession rates (m/yr) from aerial photography (1939–2001) and airborne LiDAR (2001–2012) for unobstructed bluff-edge reaches of each drift cell.
Drift Cell
Period
Minimum (m/yr)
Mean (m/yr)
Maximum (m/yr)
Standard Deviation (m/yr)
No. of Transects
Length (m)
Dungeness
1939–2001 2001–2012 1939–2001 2001–2012
0.0 0.1 0.2 0.0
0.40 0.50 0.42 0.20
1.00 0.90 0.60 0.55
0.20 0.17 0.10 0.10
181 181 75 75
5,639 5,639 2,469 2,469
Elwha
Angeles landfill revetment in the Elwha drift cell (Figure 1B). The mean recession rate in the Dungeness was 0.36 m/yr vs. 0.26 m/yr for the Elwha drift cell (Table 2) for the 2001–2012 period. In both drift cells, armored sections of bluffs showed significantly lower rates of recession than did unarmored sections: 80 percent less in the Dungeness drift cell and 50 percent less in the Elwha drift cell (Table 2 and Figure 7). Unarmored bluff sections demonstrated very similar mean rates of recession between drift cells: 0.37 m/yr for Dungeness and 0.40 m/yr for Elwha (Table 2 and Figure 7). Unarmored sections of bluffs directly down-drift and adjacent to armored sections experienced the highest rates of bluff recession in the Elwha drift cell (1.88 m/ yr) and higher than mean rates (1.0 m/yr) in the Dungeness drift cell (Figure 7). Sediment volumes eroded from bluffs in the Dungeness drift cell were almost double those observed in the Elwha drift cell per transect (Table 3 and Figures 8– 10). The mean sediment production rate in the Dungeness drift cell was 25.4 m3 per transect vs. 13.8 m3 per transect in the Elwha drift cell. Rates of sediment production from unarmored sections of bluffs were similar between drift cells. Mean values for sediment production from unarmored sections of bluffs in the Dungeness drift cell were 25.8 m3 per transect vs. 22.0 m3 per transect for the Elwha drift cell (Table 3). Sediment production rates for armored sections of bluffs were twice as high in the Elwha drift cell (11.9 m3 per transect) compared to the Dungeness drift cell (5.8 m3 per transect) (Table 3 and Figure 10). At the drift cell scale, the Dungeness bluffs produced approximately five times the volume of
sediment of the Elwha bluffs, on average (1.03 3 105 m3/yr vs. 2.0 3 104 m3/yr, respectively), on an annual basis over the 2001–2012 period (Table 4). When normalized for length, the Dungeness bluffs contributed approximately 55 percent more sediment than did the Elwha bluffs to the nearshore (7.5 m3/m/ yr vs. 4.1 m3/m/yr, respectively) on an annual basis for the 2001–2012 period (Table 5). Beach Sediment Volume Changes Annual beach sediment volume changes as well as the net 3-year change at the 12 transect locations (eight along the Dungeness bluffs; four along the Elwha bluffs) are shown in Figure 11 and Tables 6 and 7. With the exception of transect EB-1 (where the effects of sediment supply from the Elwha River are evident), the general trend in beach sediment volume has been one of net loss over the 3-year period occurring between 2010 and 2013. In the Elwha drift cell, annual beach transect elevation changes ranged from 20.72 (net loss) to +1.19 m/yr (net gain) (mean 5 20.13 6 0.52 m/yr). The greatest loss at all Elwha transects occurred during the 2010–2011 period. In the Dungeness drift cell, annual beach transect elevation changes ranged from 21.05 m/ yr to +0.22 m/yr (mean 5 20.19 6 0.29 m/yr). DISCUSSION Bluff Recession Rates Rates of bluff recession observed in this study in the Elwha drift cell generally agree with rates
Table 2. Recession rates (m/yr) by drift cell and shoreline type, 2001–2012.
Drift Cell
Shoreline Type
Dungeness
Unarmored Armored All Unarmored Armored All
Elwha
138
Minimum (m/yr)
Mean (m/yr)
Maximum (m/yr)
Standard Deviation (m/yr)
No. of Transects
Length (m)
0.0 0.0 0.0 0.0 0.0 0.0
0.37 0.08 0.36 0.40 0.21 0.26
1.88 0.46 1.88 1.88 0.58 1.88
0.79 0.40 0.24 1.30 0.40 0.23
423 10 433 60 92 152
13,320 305 13,625 1,829 3,048 4,877
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
Coastal Bluff Recession Table 3. Sediment volume contribution per transect (m3) by drift cell and shoreline type, 2001–2012.
Drift Cell Dungeness
Elwha
Shoreline Type Unarmored Armored All Unarmored Armored All
Minimum (m3)
Mean (m3)
Maximum (m3)
Standard Deviation (m3)
No. of Transects
Length (m)
0.0 0.0 0.0 0.0 0.0 0.0
25.8 5.8 25.4 22.0 11.9 13.8
163.3 9.6 124.8 143.6 41.2 159.9
24.3 3.8 31.7 30.1 7.9 35.9
423 10 433 60 92 152
13,320 305 13,625 1,829 3,048 4,877
measured by the USACE (USACE, 1971) in the Elwha drift cell in the decade between 1960 and1971, but they are elevated over those observed by Keuler (1988) in the Dungeness drift cell and are substantially higher than the long-term rates observed by Rogers et al. (2012) for Eastern Strait of Juan de Fuca shorelines. Rates of bluff erosion documented in this study are also consistent with rates observed along the west coast of the United States exposed to open-ocean wave climates (Collins and Sitar, 2008; Pettit et al., 2014). Rates of bluff recession observed between 2001 and 2012 may represent higher-than-average erosion rates due to high storm frequency and intensity occurring during this period: two time intervals, the winters of 2007 and 2009, represent two of the wettest and windiest periods on record for this location (NCDC, 2014). Additionally, the 2001–2011 period experienced four high-tide events that exceeded the 50-year recurrence interval for extreme high water levels in the Central Strait of Juan de Fuca (NOAA, 2013). Bluff recession rates observed in the Dungeness and Elwha drift cells in this study have immediate application to land-use planning for residential and
commercial construction activities adjacent to the coastal bluffs. Given a typical design life of a single family home of 100 years, applying the observed mean bluff recession rates (Table 1) provides a minimum setback distance between a structure and the edge of the bluff of 42 m in the Elwha drift cell and of 40 m in the Dungeness drift cell, based on mean long-term (1939–2001) recession rates. It should be noted that these rates of observed bluff recession fall closely in line with estimates of 0.47 m/yr published for the Elwha drift cell by Polenz et al. (2004) and likely represent the long-term post-glacial average bluff recession rate for glacial deposits on the south shore of the Central Strait of Juan de Fuca. Extending past observed bluff recession rates into the future is likely a simplistic and inaccurate method to determine future bluff recession (Hapke and Plant, 2010). Probabilistic methods of predicting bluff erosion (Lee et al., 2001; Walkden and Hall, 2005; and Hapke and Plant, 2010) that accommodate spatial and temporal variability could be applied to the Dungeness and Elwha drift cells and would likely be more accurate than using hindcast observations of bluff recession to predict future erosion rates.
Figure 8. Sediment volume (m3) per transect in the Dungeness drift cell (2001–2012).
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
139
Parks
Figure 9. Sediment volume (m3) per transect in the Elwha drift cell (2001–2012).
Figure 10. Box plot of sediment volume contributions (m3/transect) by drift cell and shoreline type (created in ABOXPLOT; Bikfalvi, 2012). The central line within the box represents the sample median, while the circle represents the sample mean. The upper and lower limits of the box represent the 50th percentile of the population and the whiskers the 75th percentile. Dots beyond the upper and lower whiskers represent outliers of the population.
140
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
Coastal Bluff Recession Table 4. Annual sediment volume contribution (m3/yr) by drift cell, 2001–2012.
Drift Cell
Mean Mean + 1 Standard No. of (m3/yr) Deviation (m3/yr) Transects Length (m)
Dungeness 103,000 Elwha 20,000
232,000 49,000
433 152
13,625 4,877
However, the data necessary to employ these procedures (e.g., wave and tidal height distributions along the bluffs) are not currently available. Sediment Volume Change Annual sediment volume contributions within the Elwha drift cell from this study (2.0 3 104 m3/yr; Table 4) are consistent with the flux of 3.1 3 104 m3/ yr determined by USACE (1971). The calculated length-normalized rate of 4.1 m3/m/yr for the Elwha drift cell is substantially less (255 percent) than the rate observed (7.5 m3/m/yr) for the Dungeness drift cell, which is consistent with a previous study by Keuler (1988) that measured sediment contribution rates for the exposed areas of the Strait of Juan de Fuca ranging between 6.0 and 12.0 m3/m/yr. Bluff-supplied sediment volume estimates for the Elwha drift cell from this study can help refine the coastal sediment budget post–dam removal. Since shore-protection works in the Elwha drift cell will remain after the Elwha Dams have been removed, a significant component of the Elwha drift cell sediment budget will remain impaired after the sediment supply from the Elwha River has been restored. Randle et al. (1996) estimates that the pre-dam fluvial sediment contribution to the Strait of Juan de Fuca was about 1.9 3 105 m3/yr. In the Elwha drift cell, the current upper estimate of annual sediment volume contribution to the nearshore from bluff erosion is approximately 2.0 3 104 m3/yr–4.9 3 104 m3/yr (Table 4), or about 11–26 percent of the pre-dam annual sediment contribution from the Elwha River. The current annual sediment volume contribution from bluff erosion in the Elwha drift cell represents a 90 percent reduction from the 1911 prearmoring estimate (2.2 3 105 m3/yr; USACE, 1971) but is roughly approximate to the 1960 post-armoring estimate (3.1 3 104 m3/yr; Galster, 1989).
Comparing the sediment production rates between the Dungeness and Elwha bluffs demonstrates the level of impairment within the Elwha drift cell. When normalized for drift cell length, the Elwha bluffs produce 56 percent less sediment volume than do the Dungeness bluffs on an annual basis. Comparing the measured rates of sediment production from bluffs (Table 5) versus sediment volume change in beach transects (Tables 6 and 7 and Figure 11) demonstrates the imbalance in the sediment supply relative to available sediment transport. In most years, the amount of available sediment volume contributed from bluffs to the beach is substantially less than the average rate of sediment loss, leading to beach lowering and resulting in accelerated bluff erosion. Management Implications Bluff recession rates were shown to vary depending on the time of measurement and length of time observed. It is not appropriate to extrapolate shortterm measurements into long-term rates, especially if the length of measurement is less than the time span of the rate being reported (e.g., producing an annual rate from ,1 year of observation). For instance, a measurement taken over a month when there was a large bluff failure could result in large overestimates of bluff recession on an annual basis if there was no further change for the remainder of the year. Moreover, using the maximum measured recession distance to calculate an annual recession rate will result in an even-greater overestimate and could give a false impression of how much the bluff is actually retreating. The maximum recession distance is measured for a specific point along the bluff and may not represent the trends observed over the larger area. It would be more correct to calculate a mean bluff recession distance for a given area measured over a long period of time (i.e., years to decades). The longterm rates should then be qualified with the amount of recession that may occur during a given event (e.g., the average maximum recession distance). As an example, for land-use management, it would be more appropriate to use a long-term mean recession rate over the horizon of interest to obtain a setback distance, with an added buffer based on event-scale recession.
Table 5. Annual length-normalized sediment contribution (m3/m/yr) by drift cell, 2001–2012.
Drift Cell
Mean (m3/m/yr)
Mean + 1 Standard Deviation (m3/m/yr)
Maximum (m3/m/yr)
No. of Transects
Length (m)
Dungeness Elwha
7.5 4.1
17.0 10.0
11.3 14.5
433 152
13,625 4,877
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
141
Parks
Figure 11. Length-normalized sediment volume change (m3/m) in the highest 20 m of each beach topographic profile during four winter-towinter time intervals. EB-1 through BL-1 were winter surveys; BL-2 through DB-4 were summer surveys. Note that intervals 1–3 are annual, whereas interval 4 spans 3 years.
It should be emphasized that the bluff recession distances reported in this study are derived from selected elevations across the bluff-face profile, which may not be seen by the homeowner at the bluff top. While the trends are not likely to significantly change, results will differ according to the methods used to analyze bluff-face change. Other methods of calculating bluff recession distances (e.g., contour change analysis, volume change analysis) are expected to provide different results than the profile-based methods used herein, and the potential to produce alongshore averaging of bluff recession rates over appropriate alongshore length scales may result in less spatially variable rates that
are more conducive to land-use zoning, buffers, and development setbacks. The bluff-face profile method has the potential to accentuate the localized erosion signals due to a lack of continuity along the bluff to enable alongshore averaging commensurate with the observed signals of change obtained at finer scale along the bluff face. While land-use planners and coastal managers are in need of long-term erosion rates for prudent resource management, property owners experience localized erosion and tend to be most interested in and concerned about the magnitude of bluff recession occurring along relatively small increments of space along their bluff-top property boundary.
Table 6. Beach topographic profile sediment volume changes for the Elwha drift cell. Note that the right-most column is net change between 2010 and 2013, while all others are annual intervals. 2010–2011 Profile EB-1 EB-2 EB-3 EB-4 Average
142
2011–2012
2012–2013
2010–2013
Volume Change Change Rate Volume Change Change Rate Volume Change Change Rate Volume Change Change Rate (m3/m) (m/yr) (m3/m) (m/yr) (m3/m) (m/yr) (m3/m) (m/yr) 213.54 27.77 212.33 211.88 211.38
20.69 20.40 20.66 20.72 20.62
2.42 20.17 1.87 0.41 1.13
0.12 20.01 0.09 0.02 0.06
24.89 5.82 22.06 21.52 3.87
1.19 20.28 20.11 20.08 0.18
13.77 213.77 212.66 212.98 26.41
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
0.23 20.23 20.22 20.23 20.11
Coastal Bluff Recession Table 7. Beach topographic profile sediment volume changes for the Dungeness drift cell. Note that the right-most column is net change between 2010 and 2013, whereas all others are annual intervals. 2010–2011 Profile BC-1 BC-2 BL-1 BL-2 DB-1 DB-2 DB-3 DB-4 Average
2011–2012
2012–2013
2010–2013
Volume Change Change Rate Volume Change Change Rate Volume Change Change Rate Volume Change Rate (m3/m) (m/yr) (m3/m) (m/yr) (m3/m) (m/yr) Change (m3/m) (m/yr) 23.87 2.83 210.47 24.38 28.56 1.11 212.23 219.44 26.88
20.20 0.15 20.43 20.22 20.43 0.06 20.79 21.05 20.36
25.37 27.85 28.57 25.22 22.07 24.30 0.48 21.67 24.32
20.24 20.34 20.38 20.29 20.12 20.24 0.03 20.09 20.21
Chronic sediment supply deficits in the Elwha drift cell due to shoreline armoring have resulted in significant habitat impairment on intertidal beaches for forage fish and juvenile salmonids (Shaffer et al., 2012; Parks et al., 2013). The removal of the two Elwha River Dams will restore a significant component of the Elwha littoral cell sediment supply. It is currently unknown to what degree and over what timescale Elwha River sediments will be stored on intertidal beaches within the Elwha drift cell. Local shoreline managers have an unprecedented opportunity to optimize storage of Elwha River sediments on intertidal beaches through implementation of selected shoreline armoring removal and large-woody debris placement strategies prior to the complete delivery of Elwha River reservoir sediments into the intertidal environment over the next 5–7 years. In contrast to the impaired habitat function of Elwha drift cell due to sediment starvation from shoreline armoring and Elwha River Dams, the Dungeness drift cell exhibits less than 1 percent by length armored shoreline and highly functioning forage fish spawning habitat (Shaffer et al., 2012; Parks et al., 2013). The intact littoral sediment supply processes from coastal bluff erosion within the Dungeness drift cell are maintaining suitable forage fish habitat (Parks et al., 2013) and expanding the Dungeness Spit through sediment deposition (Schwartz et al., 1987). CONCLUSIONS Rates of coastal bluff recession in the Dungeness and Elwha drift cells over the 1939–2012 period were highly variable in space and time and ranged between 0.31 m/yr and 1.88 m/yr. Differences between maximum near-term bluff erosion rates observed from 2001–2012 LiDAR and long-term (1939–2001) observations from digitized historical photography
22.63 1.02 23.67 4.73 21.22 0.76 2.02 3.10 0.51
20.16 0.06 20.23 0.20 20.05 0.03 0.08 0.14 0.01
211.86 24.01 222.72 24.88 211.84 22.42 29.73 218.01 210.68
20.20 20.07 20.36 20.08 20.20 20.04 20.17 20.31 20.18
were the result of individual medium-scale landslides. The presence of shoreline armoring is a controlling factor on the rate of bluff recession, with armored bluffs showing a reduced recession rate compared with unarmored bluffs. The volume of sediment produced by a unit length of unarmored bluff shoreline is greater than that of armored bluffs by factors of two (Elwha) and five (Dungeness), respectively. Sediment volumes contributed by bluffs in the Elwha drift cell between 2001 and 2012 represent 11– 29 percent of the estimated fluvial sediment contribution to the nearshore from the Elwha River prior to dam construction in 1911. Annual sediment volumes contributed by bluffs in the Elwha drift cell between 2001 and 2012 represent approximately 8–20 percent of the current estimate (Gilbert and Link, 1995; Bountry et al., 2010) of the long-term, post-dam removal annual fluvial sediment contribution to the nearshore from the Elwha River of about 2.5 3 105 m3/yr. This study confirms that alteration to bluffs, in this case armoring, drastically affects bluff recession rates and sediment volume contributions to the nearshore. Armored sections of bluffs showed significantly lower rates (280 percent, Dungeness; 253 percent, Elwha) of recession than did unarmored sections. Unarmored sections of bluffs directly down-drift and adjacent to armored sections experienced the highest rates of bluff recession in the Elwha drift cell (1.88 m/yr) and higher than mean rates (1.0 m/yr) in the Dungeness drift cell. It was beyond the scope of this study to determine why there was a difference in sediment production rates between the Elwha and Dungeness drift cells. Geology, groundwater effects, wave-approach angle, wave energy, and land use are all possible factors explaining the observed differences, and these should be further investigated in future studies.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
143
Parks
While wave run-up and erosion at the base of coastal bluffs is a dominant driving factor of erosion throughout both drift cells, portions of each drift cell also exhibited erosion in the upper one-third of the bluff profile driven by a combination of precipitation, local groundwater discharge, and relatively permeable glacial strata overlying impermeable glacial strata. The observed upper-bluff erosion driven by groundwater and precipitation appears to be spatially and temporally isolated from wave erosion, especially where shore protection works are in place, and this trend will continue whether the shoreline is armored or not. At present, it remains challenging to make reliable projections of bluff recession that may guide development setback distances for the future, given the coarse resolution of a multi-decadal interval (aerial photos for 1939–2001) and only one higher-resolution decadal interval (airborne LiDAR data for 2001– 2012). The combination of chronic recession rates and event-based erosion magnitudes is important for decision makers, and the most reliable rates will come from a longer-term high-resolution data set that must be developed over time. The results of this study provide estimates for minimum setback distances between structures and bluff edges based on long-term mean recession rates measured over the scale of an entire drift cell. This type of information provides the scientific basis that land-use planners and government regulators need in order to develop sound long-term management policies for bluff development. Recession distances measured for a specific point along the bluff may not represent the trends observed over the larger drift-cell area and over a longer period of time. It would be more correct to calculate a mean bluff recession distance for a given area measured over a long period of time (i.e., years to decades). The long-term rates should then be qualified with the amount of recession that may occur during a given event (e.g., the average maximum recession distance). As an example, for land-use management, it would be more appropriate to use a long-term mean recession rate over the horizon of interest to obtain a setback distance with an added buffer based on event-scale recession. Repeat surveys performed at relatively short intervals would enable a better determination of the relative importance of a variety of mechanisms contributing to bluff erosion, such as surface runoff (and associated land-clearing and development practices), wind, precipitation, groundwater discharge, soil saturation, wave height and direction, total water level, beach width and elevation, and littoral sediment supply. All of these factors play a role in bluff retreat
144
dynamics, and measurement of these parameters combined with high-resolution bluff-face topography and differences over time will enable the development of improved process-based bluff erosion models (Lee et al., 2001; Castedo et al., 2012). ACKNOWLEDGMENTS This study benefited from discussions with Anne Shaffer (Coastal Watershed Institute), Jon Warrick (USGS), and Hugh Shipman (WDOE) on coastal processes and sediment budgets along the Central Strait of Juan de Fuca. Jesse Wagner and Wade Raynes (Western Washington University) and Clinton Stipek (University of Washington) provided field and technical support. Western Washington University and Peninsula College provided field equipment and student interns. Anne Shaffer (Coastal Watershed Institute) provided vital overall support, coordination, and integration with other project components. Diana McCandless, Washington Department of Ecology Coastal Mapping Program, provided analysis of beach erosion data. Heather Baron, Matt Brunengo, Kerry Cato, Wendy Gerstel, Amanda Hacking, Michael W. Hart, George Kaminsky, and Keith Loague provided helpful reviews. We want to sincerely thank Ruth Jenkins, John Warrick, Chris Saari, Paul Opionuk, Pam Lowry, Connie and Pat Schoen, Hearst Cohen, Malcolm Dudley, Nippon Paper, and the Lower Elwha S’Klallam Tribe for access across private property. Dungeness National Wildlife Refuge personnel and volunteers provided access and transportation. Student interns were funded by the U.S. Environmental Protection Agency under grant number PC00J29801-0 awarded to the Washington Department of Fish and Wildlife (contract number 10-1744) and managed by the Coastal Watershed Institute. Funding for student interns and GPS equipment used to collect beach profiles were provided by the Clallam County Marine Resources Committee and by the Environmental Protection Agency grant listed above. Any opinions, findings and conclusions, or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the Environmental Protection Agency or the Washington Department of Natural Resources. REFERENCES BIKFALVI, 2012, ABOXPLOT, Advanced Boxplot Routine for MATLAB: Electronic document, available at http://alex. bikfalvi.com/research/advanced_matlab_boxplot/ BOOTH, D. B.; TROOST, K. G.; CLAGUE, J. J.; AND WAITT, R. B., 2003, The Cordilleran ice sheet: Development Quaternary Science, Vol. 1, pp. 17–43.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
Coastal Bluff Recession BOUNTRY, J. A.; FERRARI, R.; WILLE, K.; AND RANDLE, T. J., 2010, 2010 Survey Report and Area-Capacity Tables for Lake Mills and Lake Aldwell on the Elwha River, Washington: U.S. Department of Interior, Bureau of Reclamation, Technical Service Center, Report No. SRH-2010-23, 66 p. BRADLEY, W. C., 1963, Large-scale exfoliation in massive sandstones of the Colorado Plateau: Geological Society America Bulletin, Vol. 74, pp. 519–528. CARSON, M. A. AND KIRKBY, M. J., 1972, Hillslope Form and Processes: Cambridge University Press, Cambridge, MA, 475 p. CASTEDO, R.; MURPHY, W.; LAWRENCE, J.; AND PAREDES, C., 2012, A new process-response coastal recession model of soft rock cliffs: Geomorphology, Vol. 177–178, pp. 128–143. COLLINS, B. D. AND SITAR, N., 2008, Processes of coastal bluff erosion in weakly lithified sands, Pacifica, California, USA: Geomorphology, Vol. 97, No. 3–4, pp. 483–501. DETHIER, D. P.; PESSL, F., JR.; KEULER, R. F.; BALZARINI, M. A.; AND PEVEAR, D. R., 1995, Late Wisconsinan glaciomarine deposition and isostatic rebound, northern Puget Lowland, Washington: Geological Society America Bulletin, Vol. 107, No. 11, pp. 1288–1303. DOWNING, J., 1983, The Coast of Puget Sound: Its Processes and Development: University of Washington Press, Seattle, WA, 126 p. DROST, B. W., 1986, Water Resources of Clallam County, Washington: U.S. Geological Survey Water Resources Investigations Report 83-4227, 263 p. EAST, A. E.; PESS, G. R.; BOUNTRY, J. A.; MAGIRL, C. S.; RITCHIE, A. C.; LOGAN, J. B.; RANDLE, T. J.; MASTIN, M. C.; MINEAR, J. T.; DUDA, J. J.; LIERMANN, M. C.; MCHENRY, M. L.; BEECHIE, T. J.; AND SHAFROTH, P. B., 2014, Large-scale dam removal on the Elwha River, Washington, USA: River channel and floodplain geomorphic change: Geomorphology: In Press, doi:0.1016/j.geomorph.2014.08.028 FINLAYSON, D., 2006, The Geomorphology of Puget Sound Beaches: Puget Sound Nearshore Partnership Report No. 2006-02, Washington Sea Grant Program, University of Washington, Seattle, WA, 45 p. GALSTER, R. W., 1989, Ediz Hook—A Case History of Coastal Erosion and Mitigation. Engineering Geology in Washington, Vol. 2: Washington Division of Geology and Earth Resources Bulletin 78, Washington Department of Natural Resources, Olympia, WA, pp. 1177–1186. GALSTER, R. W. AND SCHWARTZ, M. L., 1989, Ediz Hook—A case history of coastal erosion and rehabilitation. In Schwartz, M. L. and Bird, E. C. F. (Editors), Artificial beaches: Journal Coastal Research, Special Issue 6, pp. 103–113. GELFENBAUM, G.; STEVENS, A.; ELIAS, E.; AND WARRICK, J., 2009, Modeling sediment transport and delta morphology on the dammed Elwha River, Washington State, USA: Coastal Dynamics: Paper No. 109, 15 p. GELFENBAUM, G. R.; STEVENS, A.; MILLER, I.; AND WARRICK, J. A., 2014, Large-scale dam removal on the Elwha River, Washington, USA: Coastal geomorphic change: Geomorphology: in review. GILBERT, J. AND LINK, R. A., 1995, Alluvium Distribution in Lake Mills, Glines Canyon Project and Lake Aldwell, Elwha Project, Washington: U.S. Department of Interior, Elwha Technical Series PN-95-4, 60 p. HAMPTON, M. A., 2002, Gravitational failure of sea cliffs in weakly lithified sediment: Environmental Engineering Geoscience, Vol. 8, No. 3, pp. 175–191. HAPKE, C., 2004, The measurement and interpretation of coastal cliff and bluff retreat. In Hampton, M. and Griggs, G. (Editors), Formation, Evolution, and Stability of Coastal
Cliffs—Status and Trends: U.S. Geological Survey Professional Paper 1693, pp. 39–50. HAPKE, C. J. AND PLANT, N., 2010, Predicting coastal cliff erosion using a Bayesian probabilistic model: Marine Geology, Vol. 278, pp. 140–149. JOHANNESSEN, J. AND MACLENNAN, A., 2007, Beaches and Bluffs of Puget Sound: Puget Sound Nearshore Partnership Report No. 2007-04, Seattle District, U.S. Army Corps of Engineers, Seattle, WA, 34 p. JONES, M. A., 1996, Delineation of Hydrogeologic Units in the Lower Dungeness River Basin, Clallam County, Washington: U.S. Geological Survey, Water Resources Investigation Report 95-4008, 11 p. KEULER, R. F., 1988, Map Showing Coastal Erosion, Sediment Supply, and Longshore Transport in the Port Townsend 30-by60-Minute Quadrangle, Puget Sound Region, Washington: U.S. Geologic Survey Miscellaneous Investigation Map I1198-E, scale 1:100,000. LEE, E. M.; HALL, J. W.; AND MEADOWCROFT, I. C., 2001, Coastal cliff recession: The use of probabilistic prediction methods: Geomorphology, Vol. 40, pp. 253–269. MILLER, I. M.; WARRICK, J. A.; AND MORGAN, C., 2011, Observations of coarse sediment movements on the mixed beach of the Elwha Delta, Washington: Marine Geology, Vol. 282, No. 3–4, pp. 201–214. MOSHER, D. C. AND HEWITT, A. T., 2004, Late Quaternary deglaciation and sea-level history of eastern Juan de Fuca Strait, Cascadia: Quaternary International, Vol. 121, pp. 23– 39. NATIONAL CLIMATE DATA CENTER (NCDC), 2014, Climate Summaries for Port Angeles and Sequim, Washington: Electronic document available at http://www.ncdc.noaa.gov/ NATIONAL OCEANIC and ATMOSPHERIC ADMINISTRATION (NOAA), 2013, Tidal Data for Port Angeles, Washington, Station 9444090; Electronic document available at http://tidesandcurrents.noaa. gov/stationhome.html?id59444090 PARKS, D.; SHAFFER, A.; AND BARRY, D., 2013, Nearshore drift-cell sediment processes and ecological function for forage fish: Implications for ecological restoration of impaired Pacific Northwest marine ecosystems: Journal Coastal Research, Vol. 29, No. 4, pp. 984–997. PETTIT, M. M.; THOMAS, M. A.; AND LOAGUE, K., 2014, Retreat of a coastal bluff in Pacifica, California: Environmental Engineering Geoscience, Vol. 20, No. 2, pp. 153–162. POLENZ, M.; WEGMANN, K. W.; AND SCHASSE, H. W., 2004, Geologic Map of the Elwha and Angeles Point 7.5-Minute Quadrangles, Clallam County, Washington: Washington Division of Geology and Earth Resources, Open File Report 2004-14. PUGET SOUND LIDAR CONSORTIUM (PSLC), 2001, Topographic LiDAR: PSLC, Clallam County, WA. RANDLE, T. J.; YOUNG, C. A.; MELENA, J. T.; AND OUELLETTE, E. M., 1996, Sediment Analysis and Modeling of the River Erosion Alternative: U.S. Bureau of Reclamation, Pacific Northwest Region, Elwha Technical Series PN-95-9, 138 p. RICE, C. A., 2006, Effects of shoreline modification on a northern Puget Sound beach: Microclimate and embryo mortality in surf smelt (Hypomesus pretiosus): Estuaries Coasts, Vol. 29, No. 1, pp. 63–71. ROGERS, H. E.; SWANSON, T. W.; AND STONE, J. O., 2012, Longterm shoreline retreat rates on Whidbey Island, Washington, USA: Quarternary Research: doi:10.1016/j.yqres.2012.06.001 SCHASSE, H. W., 2003, Geologic Map of the Washington Portion of the Port Angeles 1:100,000 Quadrangle: Washington Division of Geology and Earth Resources, Open File Report 2003-6,
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
145
Parks Washington Department of Natural Resources, Olympia, WA. SCHASSE, H. W. AND POLENZ, M., 2002, Geologic Map of the Morse Creek 7.5-Minute Quadrangle, Clallam County, Washington: Washington Division of Geology and Earth Resources, Open File Report 2002-8, Washington Department of Natural Resources, Olympia, WA. SCHASSE, H. W.; POLENZ, M.; AND WEGMANN, K. W., 2000, Geologic Map of the Carlsborg 7.5-Minute Quadrangle, Clallam County, Washington: Washington Division of Geology and Earth Resources, Open File Report 2000-7, Washington Department of Natural Resources, Olympia, WA. SCHUENEMEYER, J. H. AND DREW, L. J., 2011, Statistics for Earth and Environmental Scientists: John Wiley & Sons, Inc., Hoboken, NJ, 407 p. SCHWARTZ, M. L.; FABBRI, P.; AND WALLACE, R. S., 1987, Geomorphology of Dungeness Spit, Washington, USA: Journal Coastal Research, Vol. 3, No. 4, pp. 451–455. SCHWARTZ, M. L.; WALLACE, R. S.; AND JACOBSEN, E. E., 1989, Net shore-drift in Puget Sound, Engineering Geology in Washington, Vol. 2, Olympia, WA: Washington Division Geology Earth Resources Bulletin 78: Washington Department of Natural Resources, pp. 1137–1146. SHAFFER, J. A.; CRAIN, P.; KASSLER, T.; PENTTILA, D.; AND BARRY, D., 2012, Geomorphic habitat type, drift cell, forage fish and juvenile salmon; are they linked? Journal Environmental Science Engineering, Vol. 1, No. 5a, pp. 688–703. SHIPMAN, H., 2004, Coastal bluffs and sea cliffs on Puget Sound, Washington. In Hampton, M. A. and Griggs, G. B. (Editors), Formation, Evolution and Stability of Coastal Cliffs—Status and Trends: U.S. Geological Survey Professional Paper 1693, 123 p. SHIPMAN, H.; DETHIER, M. N.; GELFENBAUM, G.; FRESH, K. L.; AND DINICOLA, R. S. (Editors), 2010, Puget Sound shorelines and the impacts of armoring: Proceedings of the State of the Science Workshop, May 2009, Menlo Park, CA: U.S. Geological Survey Scientific Investigations Report 20105254, 262 p. U.S. ARMY CORPS OF ENGINEERS (USACE), 1971, Report on Survey of Ediz Hook for Beach Erosion and Related Purposes. Port Angeles, Washington: Department of the Army, Seattle District, Corps of Engineers, 96 p. VARNES, D. J., 1978, Slope movement types and processes. In Clark, M. (Editor), Landslide Analysis and Control: Special
146
Report 176, Transportation Research Board, National Academy of Sciences, National Research Council, Washington, DC, pp. 11–33. WALKDEN, M. J. A. AND HALL, J. W., 2005, A predictive Mesoscale model of the erosion and profile development of soft rock shores: Coastal Engineering, Vol. 52, pp. 535–563. WARRICK, J. A.; BOUNTRY, J. A.; EAST, A. E.; MAGIRL, C. S.; RANDLE, T. J.; GELFENBAUM, G. R.; RITCHIE, A. C.; PESS, G. R.; LEUNG, V.; AND DUDA, J. J., 2014, Large-scale dam removal on the Elwha River, Washington, USA: Source to sink sediment budget and synthesis. Geomorphology: in review. WARRICK, J. A.; GEORGE, D. A.; GELFENBAUM, G.; RUGGIERO, P.; KAMINSKY, G. M.; AND BEIRNE, M., 2009, Beach morphology and change along the mixed grain-size delta of the dammed Elwha River, Washington: Geomorphology, Vol. 111, pp. 136– 148. WOOLPERT, INC., 2013, Elwha River, WA LiDAR, United States Geological Survey (USGS). Contract No.: G10PC00057. Woolpert, Inc.)Englewood, CO, 11 p. YOTTER-BROWN, C. W. AND FAUX, R., 2012, LiDAR Remote Sensing, Jefferson and Clallam Counties, Washington: Puget Sound LiDAR Consortium, Seattle, WA, 29 p. YOUNG, A. P. AND ASHFORD, S. A., 2007, Quantifying sub-regional seacliff erosion using mobile terrestrial LiDAR: Shore Beach, Vol. 75, No. 3, pp. 38–43. YOUNG, A. P.; GUZA, R. T.; O’REILLY, W. C.; FLICK, R. E.; AND GUTIERREZ, R., 2011, Short-term retreat statistics of a slowly eroding coastal cliff: Natural Hazards Earth System Sciences, Vol. 11, pp. 205–217. YOUNG, A. P.; OLSEN, M. J.; DRISCOLL, N.; GUTIERREZ, R.; GUZA, R. T.; FLICK, R. E.; JOHNSTONE, E.; AND KUESTER, F., 2009, Comparison of airborne and terrestrial LiDAR estimates of seacliff erosion in Southern California: Photogrammetric Engineering Remote Sensing, Vol. 76, No. 4, pp. 421–426. YOUNG, A. P.; RAYMOND, J. H.; SORENSON, J.; JOHNSTONE, E. A.; DRISCOLL, N. W.; FLICK, R. E.; AND GUZA, R. T., 2010, Coarse sediment yields from seacliff erosion in the Oceanside Littoral Cell: Journal Coastal Research, Vol. 26, No. 3, pp. 580–585. ZILKOSKI, D. B.; RICHARDS, J. H.; AND YOUNG, G. M., 1992, Results of the general adjustment of the North American vertical datum of 1988, (NAVD 88): American Congress Surveying Mapping, Surveying Land Information Systems, Vol. 52, No. 3, pp. 133–149.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 129–146
Mine-Water Flow between Contiguous Flooded Underground Coal Mines with Hydraulically Compromised Barriers DAVID D. M. LIGHT1 JOSEPH J. DONOVAN1 Department of Geology & Geography, West Virginia University, 98 Beechurst Avenue, 330 Brooks Hall, P.O. Box 6300, Morgantown, WV 26506-6300
Key Terms: Mining Hydrogeology, Mine Flooding, Recharge
ABSTRACT Groundwater flow entering closed contiguous underground coal mines may be strongly influenced by leakage across inter-mine barriers. This study examines a complex of multiple closed and flooded mines that developed into a nearly steady-state groundwater flow system within 10 to 50 years after closure. Field waterlevel observations, mine geometry, barrier hydraulic conductivity, recharge rates, and late-stage storage gains were parameterized to match known pumping rates and develop a fluid mass balance. Vertical infiltration (recharge and leakage) estimates were developed using a depth-dependent model based on the assumption that most vertical infiltration is focused in areas with ,75 m of overburden. A MODFLOW simulation of the nearly steady-state flow conditions was calibrated to hydraulic heads in observation wells and to known pumping rates by varying barrier hydraulic conductivity. The calibrated model suggests significant head-driven leakage between adjacent mines, both horizontally through coal barriers and vertically through inter-burden into a shallower mine in an overlying seam. Calibrated barrier hydraulic conductivities were significantly greater than literature values for other mines at similar depths in the region. This suggests that some barriers may be hydraulically compromised by un-mapped entries, horizontal boreholes, or similar features that act as drains between mines. These model results suggest that post-mining inter-annual equilibrium conditions are amenable to quantitative description using mine maps, sparse observation-well data, accurately estimated pumping rates, and depth-dependent vertical infiltration estimates. Results are applicable to planning for post1 Phone: 304-293-5603; Fax: 304-293-6522; emails: dlight@mix.wvu. edu; jdonovan@wvu.edu.
flooding water-control schemes, although hydraulic testing may be required to verify model results. INTRODUCTION Underground mines can be classified into two groups: above drainage and below drainage. Abovedrainage mines can be further divided according to the direction of mining: up-dip or down-dip. Up-dip mines are ‘‘free-draining.’’ Infiltration that reaches these mines flows down-dip along the mine floor and discharges at portals and other connections to the surface, while infiltration that enters down-dip abovedrainage mines, and all below-drainage mines, flows to the lowest parts of the mine, resulting in mine flooding. Both groundwater inflow rates and accurate mine maps are essential for predicting the duration of flooding and subsequent mine-water discharge to the surface. Groundwater-inflow estimation for closed underground coal mines constrains recharge to areas of relatively shallow overburden and neglects leakage to deeper mined areas (Winters and Capo, 2004; McDonough et al., 2005; and McCoy et al., 2006). Published recharge rates applied to mines with relatively small areas of thin overburden cover, therefore, are generally minimum estimates of mine inflows. Mine maps and accurate groundwater-inflow rates (recharge and leakage) are essential to predict the time required for a mine to flood (Younger and Adams, 1999; Whitworth, 2002). Inflow rates and maps alone, however, often yield inaccurate estimates of flooding times for individual mines that are directly adjacent to, and therefore potentially connected to, other mines. In some cases, groundwater-elevation and mine-pool data for multiple mines show highly similar pool behavior between mines, suggesting inter-connection. As a result, an improved understanding of the hydrogeological interactions between adjacent mines that stems from the development of more realistic mine-inflow models and groundwaterflow models depicting conditions in multiple adjacent
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
147
Light and Donovan
mines will help clarify and improve predictions of the mine-flooding process. Such information will benefit post-closure operations by allowing more robust sizing, design, and location of mine-water extraction pumps and treatment plants, as well as the development of plans for mine-water control. Purpose The purposes of this research are to improve the understanding of post-flooding hydrogeological interactions between contiguous underground coal mines and to present a method for estimating mine inflow that includes vertical infiltration in areas with relatively thick (.100 m) overburden. The improved understanding stems from a steady-state groundwater-flow model that was conceptualized using mined areas, inter-mine barrier thicknesses, and a water budget that is based on known pumping volumes and estimated mine-water inflows. Inter-mine coal-barrier hydraulic conductivities were calibrated using known groundwater elevations and used to calculate horizontal flow between mines. Mine inflows were determined using a depth-dependent vertical infiltration model that is based on published recharge rates and overburden thicknesses. The depth-dependent model offers improved vertical infiltration estimation over earlier methods, especially when the depth of mining becomes relatively deep. Background Underground mining creates void space, removes support for overburden, and changes stress fields, frequently resulting in subsidence of overlying strata (Singh and Kendorski, 1981; Booth, 1986). Subsidence features have been categorized into zones that consist primarily of collapsed and rubblized roof rock, vertical fractures, bedding-plane separations, and sagging yet otherwise constrained strata (Singh and Kendorski, 1981; Kendorski, 1993). After mine closure, groundwater extraction ceases, and voids created by mining and subsidence begin to resaturate, resulting in an anthropogenic aquifer (Adams and Younger, 2001). Flooding in these coal-mine aquifers is marked by the initial development of a phreatic surface or ‘‘pool’’ in the deepest portion of the mine (Donovan and Fletcher, 1999), which, with continued flooding, migrates up-dip toward shallower mined areas. Flooding ceases when the pool level reaches the elevation of a ‘‘spill point’’ (Younger and Adams, 1999); alternately, mine inflows may be balanced by loses to barrier leakage or by groundwater-extraction pumping. Flooding progress tends to follow a decaying exponential curve
148
over time, with flooding rates decreasing as the pool level approaches the elevations of either groundwater sources or spill points (Whitworth, 2002). The duration of flooding varies and is controlled by recharge rates as well as the status of adjacent mines. Shallow mines tend to receive more recharge than deeper ones (Winters and Capo, 2004) and therefore tend to flood more rapidly. Considerable research has been conducted on the hydrogeology of closed underground coal mines, including the chemistry (Banks et al., 1997), volume (Pigati and Lopez, 1999), and seasonality (Pigati and Lopez, 1999; Light, 2001) of mine-water discharges. Others have examined mine aquifer properties such as porosity (Hawkins and Dunn, 2007), specific yield (McCoy, 2002), hydraulic conductivity (Aljoe and Hawkins, 1992), and retention time (Winters and Capo, 2004; Sahu and Lopez, 2009). Flooding histories have been utilized to develop models for prediction of mine flooding (Younger and Adams, 1999; Whitworth, 2002). Recharge-rate estimates for flooding and flooded mines vary from ‘‘the miner’srule-of-thumb’’ (Stoertz et al., 2001) to calculations that are based on discharge volumes (Winters and Capo, 2004; McDonough et al., 2005), pumping records (Hawkins and Dunn, 2007), and numeric modeling (Stoner et al., 1987; Williams et al., 1993). Recharge is commonly restricted to areas of relatively shallow overburden (,18 m, McDonough et al., 2005; ,75 m Winters and Capo, 2004), while leakage is typically not considered a significant source of groundwater for mine aquifers, although it has been shown to occur and even been quantified (McCoy et al., 2006; Leavitt, 1999). Neglecting leakage suggests that deep mines should be ‘‘dry’’ or have limited groundwater inflow, and it results in recharge rates that are significantly greater than published values. This would indicate that leakage should have been included in estimations of inflows to deeper mines. For the purposes of this investigation, recharge and leakage will be un-differentiated and referred to as vertical infiltration. Unconfined storage in coal mines occurs mainly in the area near the ‘‘beach,’’ where the phreatic surface intersects the floor of the mine (Hawkins and Dunn, 2007). Its value has been estimated for different extraction methods based on surface subsidence, coal seam thickness, and the height of roof collapse (McCoy, 2002). It has also been estimated using pumping rates and corresponding changes in hydraulic head (Hawkins and Dunn, 2007). Confined storage, similar to vertical infiltration in relatively deep mined areas, is commonly neglected, although it could represent a significant volume of water in areas of confined groundwater. Inter-mine coal barrier
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
Coal Mine Interaction
leakage rates have also been estimated (McCoy et al., 2006; Hawkins and Dunn, 2007). STUDY AREA The study area for this research includes seven Pittsburgh coal mines located within the Pittsburgh basin, Greene County, PA (Figures 1 and 2). The mines were operated for various periods, but all closed between 1964 and 2004 and as of spring 2013 were in the final stages of flooding, fully flooded, or managed by pumping to control mine-pool levels. Both the fully flooded mines (Crucible and Nemacolin) and the late-stage flooding mines (Pitt Gas and Gateway) contain pools with elevations above the surface of the adjacent Monongahela River (Figures 2 and 3). Mine water is pumped to treatment plants from two locations in the study area (Dilworth and Robena), and also from adjacent mines (Shannopin and Warwick #2), in order to manage pool levels in those mines. The study area is bordered by other Pittsburgh bed mines (Clyde, Humphrey, Shannopin, and Warwick #2) and is partially overlain by a mine in the Sewickley coal bed (Warwick #3) (Figure 2). There are no known surface discharges within the study area, although groundwater began discharging from an adjacent mine (Clyde, Figure 2) during early 2013 after temporary cessation of pumping operations in that mine. The water level in one mine (Mather) is currently unknown, but the mine is believed to be fully flooded with a pool elevation midway between those in adjacent mines (Gateway and Dilworth). Geologic and Hydrogeologic Setting The Pittsburgh coal basin, located within the Appalachian Plateau physiographic province (Fenneman, 1938), is bounded by the outcrop of the Pennsylvanian-age Pittsburgh coal bed in parts of southwestern Pennsylvania, southeastern Ohio, and northern West Virginia (Figure 1). The Pittsburgh coal is the basal unit of the Monongahela Group (Figure 4), which also contains the Uniontown Formation. The coal bed varies in thickness but averages 2.0 m, with minor variance in the study area. The Pittsburgh Formation consists of alternating layers of sandstone, limestone, dolomitic limestone, calcareous mudstones, shale, siltstone, and coal (Edmunds et al., 1999). The Sewickley coal, which lies stratigraphically above the Pittsburgh coal by approximately 30 m, is also mined in the basin (Figures 1 and 4), but it is neither as thick nor as extensive as the Pittsburgh coal (Hennen and Reger, 1913). The Dunkard Group overlies the Mononga-
Figure 1. Extent of the Pittsburgh coal seam (light shading) with areas of underground mining in the Pittsburgh (medium shading) and Sewickley (dark shading) seams, in addition to Greene County, PA (dashed line).
hela Group and varies in thickness up to 365 m (Edmunds et al., 1999). Structural dip of all these strata is typically less than five degrees (Beardsley et al., 1999). Rocks in the Appalachian Plateaus Province tend to have low primary porosity and permeability (Stoner, 1983). Groundwater flow is primarily through networks of stress-relief fractures and bedding-plane separations, which occur along valley walls and parallel to valley bottoms (Wyrick and Borchers, 1981; Kipp and Dinger, 1987). Hydraulic conductivity and storativity tend to decrease with depth (Stoner, 1983), and only a small portion of natural groundwater flow extends to depths greater than 50 m (Stoner et al., 1987). The removal of coal by underground mining and consequent subsidenceinduced re-distribution of overburden have considerable impacts on the un-disturbed groundwater flow regime (Stoner, 1983; Booth, 1986). Underground coal mining can also impact surface water by reducing runoff and increasing baseflow (Stoner, 1987). Mining-induced subsidence tends to create large voids and rubble zones with greatly increased hydraulic conductivity (Singh and Kendorski, 1981; Aljoe and Hawkins, 1992; Kendorski, 1993) compared to native coal and overburden (Hobba, 1991). Above rubblized areas, vertical hydraulic conductivity is similarly increased, but this effect decreases with increasing height above the rubble (Palchik, 2003). Post-closure flooding yields coal-mine aquifers (Younger and Adams, 1999), which tend to be locally heterogeneous
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
149
Light and Donovan
Figure 2. Underground Pittsburgh seam mines in the study area with structure contours of coal bottom (5 m interval), locations of monitoring wells, inter-mine barriers, and mine-water treatment plants (CLY 5 Clyde; CRU 5 Crucible; DIL 5 Dilworth; GAT 5 Gateway; HUM 5 Humphrey; MAT 5 Mather; NEM 5 Nemacolin; PIT 5 Pitt Gas; ROB 5 Robena; SHA 5 Shannopin; and WAR 5 Warwick #2).
with preferential flow paths (Aljoe and Hawkins, 1992), due to overburden subsidence, coal pillar geometry, and spatial distribution of highly transmissive main entries (Figure 5). Yet on a mine-wide scale, water levels in different locations within flooded unpumped mines are often fairly uniform (Aljoe and Hawkins, 1992; Figure 3). Coal-mine aquifers and overlying units can be hydrostratigraphically characterized using overburden subsidence zones (Kendorski, 1993; Figure 4). The caved zone contains jumbled overburden collapsed into the mine to heights of 6 to 10t, where t is the thickness of the coal seam, while strata in the overlying fractured zone contain vertical fractures and bedding-plane separations extending to heights of 24
150
to 30t above the mine floor. The dilated zone shows bedding-plane separations, increased porosity, and horizontal transmissivity, yet due to the absence of through-going fractures, it acts as the principal aquitard between overlying strata and the fractured and caved zones below. If overburden is sufficiently thick, a constrained zone consisting of gently sagging strata may also be present. The surface fracture zone contains extended and enlarged pre-existing fractures from the ground surface to 15 m depth. These subsidence zones were developed to describe overburden re-distribution over longwall panels, but similar re-distribution is likely to occur in areas of room-and-pillar mining, especially where pillars are fully extracted (Peng, 1986). The distribution of
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
Coal Mine Interaction
Figure 3. Groundwater elevations indicate fairly uniform pool levels in mines with multiple observation wells (CRU1 5 Crucible; DIL1 5 Dilworth; GAT1 and GAT2 5 Gateway; NEM1 5 Nemacolin; PIT1 5 Pitt Gas; and ROB1 5 Robena). The average surface elevation in the Monongahela River (Maxwell pool) and mine-water treatment plant control levels are indicated by arrows.
subsided materials and overburden thickness has implications for mine-aquifer water budgets including recharge and leakage. In relatively shallow areas (,75 m; e.g., Winters and Capo, 2004) where the surface fracture and fractured zones intersect, recharge rates will be highest , while in areas containing thicker overburden, aquifer inflow will occur as leakage through the dilated zone with relatively low rates. Groundwater movement through overburden is inferred to be predominantly downward into mine voids and collapsed overburden in the caved zone, which are much higher in hydraulic conductivity relative to un-mined coal and rocks. Inter-mine groundwater flow occurs horizontally through coal barriers separating mines (e.g., McCoy et al., 2006; Hawkins and Dunn, 2007), although vertical flow between mines may occur where over- or underlying seams have been mined (Miller, 2000). Flow between mines follows pressure gradients toward discharge locations. Hydraulic conductivity (K) within coal-mine aquifers is related to the degree of overburden alteration and subsidence and is significantly increased over K within native coal (Harlow and Lecain, 1993). Within the caved zone, K in un-collapsed rooms and mains
can be very high, while in ‘‘gob’’ (collapsed) areas, collapsed overburden may reduce K values. Shale and other thinly layered rocks tend to collapse in small pieces, resulting in poorly connected void space, while sandstone and similarly massive rocks tend to collapse in large blocks, leaving significant void space (Palchik, 2002). Strata in the fractured zone will have high vertical hydraulic conductivity (KV) relative to horizontal hydraulic conductivity (KH) (Palchik, 2002), yet the number and size of vertical fractures decrease with increasing distance above the mine void, resulting in a similar reduction in KV (Palchik, 2003). METHODOLOGY Groundwater-Head Data Groundwater elevations were calculated for six monitoring wells (Figure 2) using depth-to-water measurements and pressure transducers. Between September 2000 and November 2005, pressure measurements were made using vented transducers, while after November 2005 measurements were recorded primarily with sealed transducers and corrected using barometric-pressure data collected
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
151
Light and Donovan
Figure 5. Plan view of typical mine map showing variation in mining methods, main entries, and un-mined coal (pillars).
Figure 4. Generalized hydrostratigraphy of the Pittsburgh Formation, Upper Pennsylvanian Monongahela Group (after Edmunds et al., 1999). Scale is approximate.
within the study area. Pressures were recorded hourly and then converted to average daily water levels using a database. Three of the monitoring wells are previously existing rock-dust boreholes (GAT1, GAT2, and NEM1; Figure 2), while the other three wells (CRU1, PIT1, and MAT1; Figure 2) were all drilled for the purpose of monitoring mine-pool elevations. Historical (pre-September 2000) groundwater-elevations for the pools within Gateway and Robena (GAT1 and ROB1, respectively; Figure 2) are from unpublished file data. While limited, recent groundwater elevations for the pumped mines Dilworth and Robena (DIL1 and ROB1, respectively; Figure 2) were provided by treatment-plant operators. Geospatial Analysis Mine outlines and areas, inter-mine coal-barrier dimensions, and overburden isopachs for the study area were mapped using a geographic information system (GIS). Pittsburgh coal bed mine maps were obtained from the Pennsylvania Department of
152
Environmental Protection (PADEP), digitized, and geo-rectified using mining features depicted on the maps and located in the field using a global positioning system (GPS). Inter-mine barrier segments were measured, and their areas and lengths were used to estimate average width. All barriers in the study are assumed to be 2 m high, the approximate thickness of the Pittsburgh coal in this area (Edmunds et al., 1999). The Pittsburgh coal bed structure was developed using kriged base-of-coal elevations from mine maps to create a grid. This coalbed structure grid was subtracted from the 10 m digital elevation model (DEM) to create an overburden isopach. Because vertical infiltration rates are dependent upon overburden thickness, the latter is a factor in estimating the volume of groundwater that reaches the mine aquifer. Fluid Mass Balance A water budget or fluid mass balance (FMB) for mines i in the study area was developed to improve understanding of the flow regime. The FMB includes vertical infiltration (VI), extraction pumping (P), storage changes (DS), surface discharge (Q), and barrier leakage (LB): n X ðVIi zLBi zPi zDSi {Qi Þ~0
ð1Þ
i~0
Vertical infiltration is the primary source of groundwater, while extraction pumping, surface discharge,
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
Coal Mine Interaction Table 1. Vertical infiltration rates for regional coal mines.
Source
Year
Coal
State
Outcrops
Average Depth (m)
Hawkins and Dunn McDonough et al. Stoertz et al. Stoner et al. Williams et al. Winters and Capo Delmont Export Coal Run Irwin Guffey Marchand Banning McCoy Barrackville Clyde Jamison #9 Joanne Jordan Robena Shannopin Wyatt Overburden , 18 m* Overburden , 75 m*
2007 2005 2001 1987 1993 2004
LK; LF P MK P P
PA PA OH PA PA
Yes Yes Yes
,18 15
P P P P P P P
PA PA PA PA PA PA PA
Yes Yes Yes Yes Yes Yes Yes
P P P P P P P P P P
WV PA WV WV WV PA PA WV PA PA
No No No No Yes No Yes Yes No No
VI (mm/d)
Method
0.36 4.65 0.67 0.45 0.25
Pumping records Measured discharge Miner’s rule-of-thumb Numeric model Numeric model
31 37 37 69 85 94 96
0.72 0.59 0.46 0.43 0.76 0.30 0.32
Measured Measured Measured Measured Measured Measured Measured
149 136 207 169 130 174 139 101 166 166
0.05 0.19 0.03 0.03 0.11 0.04 0.06 0.21 725 3.3
discharge discharge discharge discharge discharge discharge discharge
2002 Fluid mass Fluid mass Fluid mass Fluid mass Fluid mass Fluid mass Fluid mass Fluid mass DDVIM DDVIM
balance balance balance balance balance balance balance balance
P 5 Pittsburgh; LK 5 Lower Kittaning; LF 5 Lower Freeport; MK 5 Middle Kittanning; DDVIM 5 depth-dependent vertical infiltration model. *VI applied only to areas with overburden less than these thicknesses.
and addition to storage are all sinks. Barrier leakage may be a source or sink depending upon whether flow is into or out of the study area. In order to estimate mine inflow within the relatively deep mines of the study area, a depth-dependent vertical infiltration model was developed using published recharge rates for mines in the Pittsburgh coal (Table 1). The model uses a constant rate equal to the miner’s rule of thumb (0.67 mm/d) of Stoertz et al. (2001) for depths from 0 to 30 m and an exponentially declining rate for depths below 30 m: VI(d) ~VI(0) ðdƒd1 Þ
ð2Þ
VI(d) ~eVI(0) e{l(d) ðdwd1 Þ
ð3Þ
where VI(d) is the recharge rate at depth d below land surface, VI(o) is the maximum vertical infiltration rate in shallow aquifers, d1 is the maximum depth at which the surface fracture and fracture zones intersect, l is a location-specific vertical infiltration decline parameter, and e is a fit parameter. VI(o) (0.67 mm/d) is similar to the vertical infiltration rate reported for unmined areas in Greene County, PA (Stoner, 1983), and roughly 40 percent of the average vertical
infiltration rate for aquifers in the Monongahela River basin of northern West Virginia (Kozar and Mathes, 2001). The depth-dependent vertical infiltration model was applied to the overburden isopach, yielding a vertical infiltration estimate for the study area. Groundwater is extracted for treatment from two mines within the study area, and there are no known surface discharges. Increases in confined storage within fully flooded areas of Gateway and Pitt Gas mines were determined using the daily average change in water-level elevation during 2012 (Figure 3) and a confined-storage coefficient estimate of 0.001. Specific yield for the small unconfined area within Pitt Gas was calculated (McCoy, 2002): Sy ~
Em bCs b
ð4Þ
where Sy is specific yield, Em is the coal extraction ratio, Cs is the volume of void space remaining after surface subsidence, b is the height of the coal bed, and b is the height of caved overburden. Barrier leakage estimates (LBi) were calculated using head differences between adjacent mines Dhj, with barrier heights b; barrier segments j; barrier segment widths wj and
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
153
Light and Donovan
lengths Xj; and barrier hydraulic conductivity KB (similar to McCoy et al., 2006): LBi ~
n X
KB bXj
j~0
Dhj wj
ð5Þ
Groundwater-Flow Model Development Groundwater-flow modeling was applied to better understand the interactions between adjacent flooding and flooded coal mines. The goal of the model is to use groundwater-elevation heads for calibration and pumping rates at treatment plants to define the water budget. The numeric model depicts nearsteady-state conditions in 2012, during which all mines in the study area except two flooding mines had attained post-flooding hydraulic equilibrium. Groundwater elevations within the two exceptions (Gateway and Pitt Gas) were within 10 m of anticipated equilibrium elevation. The pumping and water-level data available for Dilworth and Robena are from 2011, yet they are thought to be representative of average conditions in those mines, as their pool levels are maintained below control elevations and do not vary significantly from year to year, nor do the average annual pumping volumes. The flow model thus depicts average post-flooding groundwater control conditions, but it does not account for seasonal or inter-annual variability. RESULTS Groundwater Hydrographs Groundwater-elevation hydrographs for the study area are shown in Figure 3. The hydrographs for Crucible and Nemacolin indicate approximate equilibrium with intra-annual fluctuations attributed to seasonal variations in vertical infiltration, precipitation, and evapotranspiration rates (Pigati and Lopez, 1999; Light, 2001), as well as barrier leakage to adjacent mines. The fact that water levels have equilibrated without surface discharge or pumping control indicates that these mines must lose water entirely to barrier leakage. Their relative increases in groundwater elevations between 2007 and 2009 are attributed to the effects of post-closure flooding in adjacent Dilworth mine, decreasing inter-mine head differences and barrier leakage from these two mines into Dilworth. Dilworth mine-pool-level control pumping began during 2008 and resulted in stabilization of the pools in Crucible and Nemacolin. The Robena hydrograph indicates that extraction pumping for managing its pool level have made it a
154
groundwater sink for Nemacolin (Figures 2 and 3). The stable pool elevation within Mather in 2001–2002 shows the mine was only partially flooded during that period, and that any inflow from infiltration was lost by barrier leakage to adjacent Gateway and/or Dilworth (Figures 2 and 3). While water-level data are unavailable, the pool level in Mather is believed to have begun rising when the pool in Gateway reached the elevation of the barrier separating those mines. The flooding rate in Mather most likely increased following the 2004 closure of Dilworth. The pool in Dilworth is currently (2013) maintained by pumping below 225.5 m (Figure 3). A stable pool elevation prevailed in Pitt Gas prior to 2007 and was maintained by cross-barrier horizontal boreholes that were installed to drain Pitt Gas mine water into Gateway. In early 2007, the groundwater level in Gateway reached the elevation of those drains, initiating flooding within Pitt Gas. After 2007, Pitt Gas and Gateway flooded in tandem, with fluctuations in the flooding rate attributed to variation in seasonally affected vertical infiltration as well as groundwater heads in adjacent mines (Figure 3). Late in 2011, the pool elevation in Pitt Gas reached the elevation of its roof, resulting in accelerated flooding as water filled all mine voids and moved upward into low-porosity overburden fractures. In early 2013, the flooding rate continued to increase, and water levels in both mines were above the surface elevation of the Monongahela River (Figures 2 and 3). The potential for surface discharge from either Gateway or Pitt Gas at elevations above 233 m exists, as does the possibility that vertical infiltration to these mines will be entirely offset by barrier leakage to adjacent mines (similar to the case in Crucible and Nemacolin). Geospatial Analysis Mines in the study range from 2.3 to 80 km2 in area, while overburden thickness varies from less than 10 m to more than 300 m, averaging 166 m (Table 2). The mines contain relatively little area with thin overburden. Less than 0.01 percent of the total area contains overburden under 18 m thick, and overburden is less than 75 m thick in only 1.5 percent of the study area (Figure 6). Pitt Gas, the smallest and shallowest mine, accounts for roughly 1 percent of the total mined area and is the only mine with overburden less than 18 m thick. Inter-mine coal barriers vary in average width from 22 to 80 m and in length from 1600 to 7600 m (Figure 2 and Table 3). The barriers separating Gateway from Mather and Nemacolin from Robena are the longest and narrowest, while the barriers separating Mather from Dilworth and
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
Coal Mine Interaction Table 2. Mine area and overburden distribution statistics.
Table 3. Measured barrier dimensions (refer to Figure 2 for barrier locations).
Overburden Thickness (m) 6
2
Mine
Area (10 m )
Min.
Max.
Avg.
Pitt Gas Crucible Nemacolin Mather Robena Dilworth Gateway All mines
2.3 22.6 39.6 19.9 79.0 34.1 41.0 238
8.30 35.5 49.8 77.5 31.4 36.4 31.0 8.30
153 222 240 265 328 283 309 328
85 133 142 144 175 177 192 166
Nemacolin from Dilworth are relatively short and wide (Table 3). Fluid Mass Balance Initial vertical infiltration estimates were made by applying average daily extraction volumes for the pumps in Dilworth and Robena (Table 4 and Figure 2) first to mined areas with overburden less than 18 m thick (i.e., McDonough et al., 2005) and then to mined areas with overburden less than 75 m thick (i.e., Winters and Capo, 2004). Both estimates resulted in vertical infiltration rates that were considerably greater than those reported in similar studies (Table 1), indicating that vertical infiltration to deeper mined areas is a significant portion of the FMB. Published recharge rates for mines in the Pittsburgh coal (Table 1) were used to determine the form of the depth-dependent vertical infiltration model for depths below d1 (Eq. 3; Table 5 and Figure 7). Applying the depth-dependent vertical infiltration model to the overburden isopach produced a vertical
Barrier ID
Mines
Total Length (m)
Average Width (m)
C1 C2 G M N1 N2
Crucible-Dilworth Crucible-Nemacolin Gateway-Mather Mather-Dilworth Nemacolin-Dilworth Nemacolin-Robena
4,395 5,920 6,290 1,600 2,000 7,570
60 67 22 80 61 36
infiltration estimate that exceeds total pumping by approximately 60 percent (Table 4). This discrepancy suggests that barrier leakage to adjacent mines outside the study area may also occur. The pool level in Clyde mine (Figure 2) was approximately 10 m above the groundwater elevation in Gateway in fall 2012, which indicates that Clyde could only act as a source, not as a sink, of barrier leakage for Gateway. Similarly, the pool in Warwick #2 is maintained by pumping at an elevation of ,230 m, well above the mine-water control elevation in Robena mine (215 m). However, both Humphrey and Shannopin mines (Figure 2) contain pools at lower elevations (157 and 190 m, respectively) than the control elevation in Robena, yet they are also separated from Robena by relatively wide barriers of limited length, and likely neither mine receives significant leakage from Robena. Warwick #3 mine is in the Sewickley seam, about 30 m above the Pittsburgh bed, and its location straddles the barrier pillar between Robena and Shannopin mines (Figures 2 and 4). It was closed due to significant groundwater inflow through vertical fractures connecting it to the underlying Shannopin mine (Miller, 2000). The pool in Shanno-
Figure 6. Cumulative distribution of mine area versus overburden thickness. See Table 2 for overburden statistics.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
155
Light and Donovan Table 4. Vertical infiltration, pumping, storage, and leakage rates. DS
Mine
VI
Crucible C1 C2
BL
2,981 1,774 1,207
Dilworth Gateway G CL
2,468 2,284
29,240 2471
2,420 0
Mather M
2,247
Nemacolin N1 N2
4,464
4,668 416 5,242
Pitt Gas Robena R
2202
725 5,387
2673
20,556
22,700
27,943
Total
211,940
27,943
All values are m3/d; negatives offset VI.
pin has since been lowered by pumping to an elevation below 190 m (2012) to allow new mining in the Sewickley seam, making it a potential sink for leakage from Robena. It is interpreted that fractures between both Robena and Shannopin and the overlying Warwick #3 mine provide pathways for vertical leakage between Robena and Shannopin via the Sewickley seam workings. Any groundwater leaked into Shannopin is removed by pumping. Leakage from Robena to Shannopin via Warwick #3 is estimated thus:
Figure 7. Estimates of groundwater vertical infiltration to underground mines, fitted using Eq. 2 and Eq. 3: VI(d) 5 VI(0) (d # d1), VI(d) 5 VI(0) e2l(d) (d . d1) (l solid line and lmin dashed). The density function (dotted line) describes overburden thickness for mines within the study area.
These leakage estimates suggest that groundwater in Nemacolin should leak primarily to Robena, while a small portion leaks to Dilworth. Similarly, Crucible is expected to leak most of its groundwater to Dilworth, with some going to Nemacolin. Groundwater Flow Modeling
where LW is vertical leakage from Robena to Warwick #3 (Table 4). Estimated additions to confined storage amounted to 471 m3/d within Gateway and 27 m3/d in Pitt Gas, while a rate of 175 m3/d was added to storage within the approximately 50,000 m2 unconfined area of Pitt Gas (Tables 4 and 6). Daily pumping volumes for Dilworth and Robena were estimated by averaging annual total volumes (Table 7). Barrier leakage estimates were made for the two mines with multiple adjacent mines, Crucible and Nemacolin, assuming that the barriers are intact, homogeneous, and without hydraulically compromised areas (Table 8).
Data regarding coal-mine aquifers are often limited to the spatial extent of mining, sparse groundwaterhead measurements, and discharge volumes, while conditions within mines and of inter-mine barrier pillars are unknown. This lack of information requires a number of assumptions in order to conceptualize groundwater flow within and between mines that comprise coal-mine aquifers. Generally, all groundwater originates as vertical infiltration downward into the mines and flows toward groundwater extraction pumps in Dilworth and Robena. Vertical infiltration is inferred to be dependent upon overburden thickness, with the highest vertical infiltration rates occurring in Pitt Gas and Robena below stream valleys, while the lowest rates occur under hills and ridges (Figure 8). Relatively small volumes of the groundwater infiltrating Pitt Gas and Gateway are assumed to be retained as storage within these mines,
Table 5. Parameters for Equations 2 and 3.
Table 6. Parameters for Sy calculations.
LW ~
n X
ðVIi zPi zDSi Þ
ð6Þ
i~0
VI(o) (mm/d)
d1 (m)
l
e
lmin
b (m)
b (m)
Em
Cs
0.67
33
0.021
2
0.023
20
2.0
0.80
0.80
156
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
Coal Mine Interaction Table 7. Monthly (2011) extraction volumes for pumps in the study area (1,000 m3). Mine
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Annual
Robena Dilworth Total
260 250
40.9 368
0.68 303
0 344
175 300
295 344
216 407
0 231
0 0
0 323
0 249
0 251
988 3,372 4,360
while the remainder leaks through the barrier between Gateway and Mather, joining vertical infiltration received by the latter mine before leaking into Dilworth through barrier segment M (Figure 9). Vertical infiltration entering Crucible leaks to both Dilworth and Nemacolin; similarly, vertical infiltration that enters Nemacolin leaks to Dilworth and Robena (Figure 8). Based on anecdotal reports by miners (Miller, 2000) and on mass balance discrepancy, some additional flow is suspected to occur upward through vertical fractures from Robena to Warwick #3 in the Sewickley seam (Figure 9). It is assumed that the Gateway/Clyde barrier (north), the east barriers of Crucible and Nemacolin (east), and the deep mining faces of Gateway and Robena (west) all are effectively no-flow boundaries. Groundwater movement to and/or from surrounding un-mined coal is assumed to be insignificant relative to other portions of the FMB. Similarly, barrier leakage between the study area and surrounding mines (Clyde, Warwick #2, Humphrey, and Shannopin) is thought to be small and have no effect on overall flow directions. Modeling Approach The U.S. Geological Survey (USGS) Modular Finite-Difference Flow Model (MODFLOW-2000) (Harbaugh et al., 2000) was employed to create a steady-state model of post-flooding groundwater conditions under pumping control in the year 2012. Pre- and post-processing were conducted utilizing Groundwater Vistas version 6.22. At this time, all mines in the study area except Pitt Gas and Gateway are thought to have been fully flooded and at seasonally fluctuating, but inter-annual steady state. Table 8. Barrier leakage estimates for mines with multiple adjacent mines (refer to Figure 2 for barrier locations). Barrier ID
Mines
Dh (m)
LB (m3/d)
C1 C2
Crucible-Dilworth Crucible-Nemacolin Crucible total Nemacolin-Dilworth Nemacolin-Robena Nemacolin total
19 2.7
223.8 40.8 264.5 83.4 801.5 884.9
N1 N2
16.3 24.3
The goal of the model was to determine groundwaterflow paths and rates within and between mines in the study area. The model employs a 100 3 100 m three-layer grid rotated 16 degrees to align with most inter-mine barriers (Figure 10). Internal coal pillars .10,000 m2 area were also considered no-flow regions. Flow is, however, known to occur across narrow inter-mine barriers separating the mines; the magnitude and direction of this leakage were obtained by calibration using horizontal flow barrier (HFB) cells. HFB cells allow modeling of barrier thicknesses greater or less than grid spacing and variation of local barrier hydraulic conductivities KB (Figure 10). Initial KB values were 0.078 m/d, a value based on field calculations of McCoy et al. (2006). All three layers are confined (LAYCON 5 3) and represent groundwater flow within the mined area, as well as in overlying collapse and fracture zones, e.g., well beneath the shallow groundwater-flow system. The un-flooded portion of Robena up-dip of its water-table surface was not modeled. Vertical infiltration and barrier leakage occurring in this region were added to adjacent active cells in order to maintain the FMB. Boundaries and Parameterization Boundaries for the model include a recharge (vertical infiltration) boundary at the top of model layer 1, no-flow cells at the bottom of layer 3, and noflow cells representing un-mined coal at the perimeter of the model. A single constant-head cell was located in reasonable proximity to the pumps in both Dilworth and Robena, at the elevation of the average pool control elevation maintained in these mines during 2011 (Table 9), as an aid in calibration. The constant-head cells were removed once calibration was achieved. MODFLOW WEL-package (specified-flux) cells were utilized to simulate pumping from Dilworth and Robena; movement of groundwater into storage within nearly flooded Gateway and Pitt Gas; and upward leakage from Robena into the overlying Warwick #3 mine and, ultimately, into Shannopin to the south (Figure 10). Average daily pumping rates for Dilworth and Robena were estimated using
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
157
Light and Donovan
Figure 8. Vertical infiltration rates applied to the groundwater-flow model. This and later maps have been rotated from geographic north to align with the model grid.
operator-supplied values for 2011 (Table 7). The calculated daily increase in storage within Pitt Gas and Gateway was distributed across 3,951 WEL cells (Table 4 and Figure 10). Barrier leakage into Robena from Nemacolin and vertical infiltration to Robena in excess of pumping from Robena were distributed among 233 WEL cells in Robena to simulate leakage to Warwick #3 (Table 4). Layers 2 and 3 were assigned isotropic K values of 1000 m/d to simulate large conduits associated with main entries and highly conductive gob zones. In layer 1, KH was assigned a value of 1.0 m/d, while KV was assigned a value of 100 m/d, reflecting the fact that layer 1 is thought to contain significant vertical fracturing. The top of layer 1 is where groundwater enters active cells in the model by vertical infiltration. The per-cell infiltration rate was calculated at 100 3
158
100 m2 grid scale using local overburden thickness and the depth-dependent vertical infiltration relationship (Figures 7 and 8). Calibration Although groundwater elevations vary seasonally in all these mines, average annual elevations within monitoring wells during 2012 (Table 9 and Figure 3) were used for calibration. Calibration was accomplished by iteratively adjusting KB of individual intermine barriers until modeled heads were within 1.0 m of target values (Table 9). The calibration process also required a reduction in the volume of groundwater extracted by WEL cells for DS within Gateway and Pitt Gas (Table 4). A head change criterion of 1025 m and mass balance error of 0.007 percent were considered sufficient for convergence.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
Coal Mine Interaction
Figure 9. Conceptual model of groundwater flow across leaky barriers.
Model Results Calibration required increasing KB values for barrier sections by one to three orders of magnitude over initial estimates (Table 10). The calibrated potentiometric contours indicate flow within individual mines from relatively high vertical infiltration areas towards leaky barriers, pumps, and the WEL cells, which simulate leakage into the overlying Warwick #3 mine (Figure 11). These contours deflect
at leaky inter-mine barriers as a result of differences in conductivity between mines and barriers. In short, the barriers tend to maintain individual pools within each mine that may receive leakage or leak into one or more adjacent mines. The potentiometric contours may be analyzed to show the locations of flow divides that partition the study area into a number of catchments, while particle traces indicate that groundwater may move through multiple mines before discharging (Figure 11).
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
159
Light and Donovan
Figure 10. Boundary condition types and locations within the groundwater-flow model. WEL cells for storage and leakage are located in layer 1; extraction wells, constant heads, and targets are all in layer 3.
DISCUSSION Results indicate that post-mining hydrogeology within flooded and flooding underground mine complexes is amenable to numerical modeling.
Known data, including groundwater elevations, mine maps, and pumping volumes, can be combined with vertical infiltration estimates to allow calculation of barrier leakage rates and flow patterns within and between adjacent mines. Results also indicate the
Table 9. Observed and modeled groundwater-elevation heads in meters. Target
Min.
Max.
Avg.*
s
Modeled
CRU GAT1 GAT2 NEM PIT DIL ROB
236.3 228.0 227.5 233.6 229.4 213.6 209.0
237.5 233.4 233.0 234.9 235.0 219.9 213.1
237.0 230.5 230.1 234.3 231.8 217.4 211.3
0.35 1.46 1.62 0.33 1.42 1.5 1.1
237.0 230.6 230.6 234.4 230.9 217.0** 211.0**
*Year 2011 for DIL and ROB, 2012 for all others. **Values assigned to constant-head cells during initial calibration.
160
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
Coal Mine Interaction Table 10. Calibrated KB values. Inter-Mine Barrier
K (m/d)
% McCoy*
C1 C2 G M N1 N2
0.53 2.00 0.55 25.00 0.30 0.49
700 2,600 700 32,000 400 600
*Average K for intact coal barriers: 0.078 m/d (McCoy et al., 2006).
potential for compromised barriers with leakage rates significantly greater than would be observed due to homogeneous barrier leakage alone. Calibrated KB values suggest that coal barriers within the study area are more conductive than those in the Pittsburgh seam studied by McCoy et al.
(2006). It is likely that these barriers are hydraulically compromised by un-mapped entries between mines, boreholes, or subsidence. The actual KB values for intact coal barriers may well be similar to those determined by McCoy et al. (2006), but the significantly greater calibrated KB values are the result of averaging relatively low-KB barrier segments with relatively highly conductive compromised barrier sections. The distribution of barrier leakage out of Nemacolin and Crucible into adjacent mines indicates variation in barrier hydraulic properties and geometry. The calibrated KB values for barriers N1 and N2 are similar (Table 10), which suggests that the significantly greater barrier leakage from Nemacolin to Robena than from Nemacolin to Dilworth (Table 4) results from the greater length and narrower width of N2 relative to N1 (Table 3), as well as the steeper head gradient between Nemacolin and Ro-
Figure 11. Calibrated steady-state hydraulic heads for layer 3. Symbology as for Figure 9.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
161
Light and Donovan
bena (the N2 barrier) compared to Nemacolin and Dilworth (the N1 barrier; Table 9 and Figure 11). The C1 and C2 barriers are of similar width, but C2 is longer and more conductive; nevertheless, Crucible leaks more water into Dilworth than it does to Nemacolin, which suggests that the higher head gradient between Crucible and Dilworth is the primary control on barrier leakage out of Crucible (Tables 3, 4, 9, and 10). The calibrated KB for barrier M is an order of magnitude higher than all other KB values calculated in the model (Table 10), yet M is also the widest and shortest barrier (Table 3). These and other observations are interpreted as strong evidence that many barrier sections in this study area are hydraulically compromised and not exhibiting simple matrix or fracture flow. Calibrated groundwater-elevation contours indicate flow toward the pumps in Dilworth and Robena and toward WEL cells in Robena, which simulate leakage to Warwick #3, and also locate several flow divides within the study area (Figure 11). The locations of the flow divides reflect variation in barrier hydraulic characteristics and geometry and outline catchments that illustrate the partitioning of groundwater between the different sinks. The catchments show that groundwater infiltrating any individual mine may flow through multiple adjacent mines before reaching a sink (Figure 11). For example, vertical infiltration entering Pitt Gas flows though Gateway, Mather, and most of Dilworth before being extracted from Dilworth, while vertical infiltration that enters Crucible may leak directly to Dilworth, leak to Nemacolin, and then to Dilworth, or leak to Nemacolin, flow through Robena, and then pass through Warwick #3 in route to pumps in Shannopin. The calibrated groundwaterelevation contours also depict relatively low head gradients within individual mines as well as significant differences between KB and K in the collapsed zone, indicated by the deflection of contour lines near barriers. Both mimic shallow hydraulic gradients observed in underground mine pools (Aljoe and Hawkins, 1992). The model indicates that groundwater elevations in some contiguous flooded mines may achieve seasonally varying, inter-annual equilibrium when barrier leakage from these mines to adjacent mines is sufficient to offset vertical infiltration. Crucible and Nemacolin maintain relatively constant groundwater elevations by discharging to adjacent mines. The current conditions in Mather are unknown, but groundwater elevations in that mine are similarly thought to at equilibrium as inflowing water leaks to Dilworth. During this study, Pitt Gas and Gateway were still flooding yet leaking considerable volumes of water to Mather. At present, it is uncertain whether
162
these mines will achieve steady state by barrier leakage or ultimately discharge to the surface. The depth-dependent vertical infiltration model yields infiltration rates that decrease exponentially with increasing depth, whereas earlier methods tended to apply uniform recharge rates to shallow areas while assuming vertical infiltration is negligible in relatively deep (.75 m) mined areas. Applying recharge only to thin overburden areas (,75 m) resulted in rates that were orders of magnitude greater than values reported for relatively shallow mines. The vertical infiltration model therefore offers an improved method when deep mining becomes a significant portion of the total mined area. Yet, there is some uncertainty in the vertical infiltration model. Within the study area, modeled vertical infiltration exceeds extraction pumping by roughly 40 percent (Table 4). The model can be adjusted to site-specific information by changing the l value (Eq. 3). A minimum l (lmin) value was attained by setting vertical infiltration equal to pumping and additions to storage within Gateway and Pitt Gas and ignoring barrier leakage into or out of the study area, yet calibrating the groundwater-flow model to lmin requires groundwater flow from Robena to Nemacolin against the head gradient. It is likely that the actual l value is between 0.021 and 0.023 within the study area, yet further refinement of l is considered unwarranted given uncertainties in barrier leakage rates, vertical leakage from Robena to Warwick #3, and the potential for barrier leakage between the study area and surrounding mines. CONCLUSIONS
N N N
N
Post-closure mine flooding often results in complex hydrogeological conditions among groups of adjacent mines. These conditions are influenced by vertical infiltration, barrier leakage, and pumping rates. The post-mining hydrogeology of mine complexes is amenable to numerical modeling given known data, including groundwater-elevation heads, pumping rates, and the geospatial extent of mining. Current recharge estimation for underground mines assumes that recharge only occurs in areas with relatively thin (,75 m) overburden and neglects leakage to deeper mined areas. This restriction results in increasingly high recharge rates as the depth of mining increases. The depth-dependent vertical infiltration model offers an improved method for estimating recharge to underground mines, especially as the area of relatively deep mined area (.75 m) increases. The model is amenable to modification for site-specific conditions in other mine complexes.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
Coal Mine Interaction
N
N
N
Calibrated coal-barrier hydraulic conductivity values are greater than those reported by McCoy et al. (2006) by 33 to 253. The causes for these increases are unknown, but it is speculated that un-mapped entries between mines, boreholes, or other conditions have resulted in hydraulic compromise of barrier integrity. The calibrated groundwater-flow model indicates that barrier leakage is sufficient to offset vertical infiltration within individual mines, making it possible for groundwater extraction pumps in one or more mines to control pool elevations in multiple adjacent mines. The model further indicates that vertical leakage may play a role in the FMB of mines that are overlying or underlying other mined coal seams. Vertical leakage is especially likely when the inter-burden separating mined seams lies within the fractured zone. The results of this study have implications for other flooding and flooded underground mines, including the post-closure treatment of mine water. Failure to consider post-flooding hydrogeological conditions such as potential inter-mine connections among adjacent mines may result in poorly sited pumps, undersized wastewater treatment plants, and underestimation of water-treatment budgets.
ACKNOWLEDGMENTS The authors would like to thank T. D. Light, J. Hawkins, and two anonymous reviewers for comments that improved the manuscript. REFERENCES ADAMS, R. AND YOUNGER, P. L., 2001, A strategy for modeling ground water rebound in abandoned deep mine systems: Groundwater, Vol. 39, No. 2, pp. 249–261. ALJOE, W. W. AND HAWKINS, J. W., 1992, Application of aquifer testing in surface and underground mines: Groundwater Management, Vol. 13, pp. 541–555. BANKS, D.; YOUNGER, P. L.; ARNESEN, R.; IVERSEN, E. R.; AND BANKS, S. B., 1997, Mine-water chemistry: The good, the bad and the ugly: Environmental Geology, Vol. 32, No. 3, pp. 157–174. BEARDSLEY, R. W.; CAMPBELL, R. C.; AND SHAW, M. A., 1999, Chapter 20. Appalachian Plateaus. In Shultz, C. H. (Editor), The Geology of Pennsylvania: Pennsylvania Geological Survey, Harrisburg, PA, and Pittsburgh Geological Survey, Pittsburgh, PA, pp. 286–297. BOOTH, C. J., 1986, Strata-movement concepts and the hydrogeological impact of underground coal mining: Groundwater, Vol. 24, No. 4, pp. 507–515. DONOVAN, J. J. AND FLETCHER, J., 1999, Hydrogeological and Geochemical Response to Mine Flooding in the Pittsburgh Coal Basin, Southern Monongahela River Basin: Project WV-132, Final Report to the U.S. Environmental Protection Agency, Morgantown, WV, p. 47.
EDMUNDS, W. E.; SKEMA, V. W.; AND FLINT, N. K., 1999, Chapter 10. Pennsylvanian. In Shultz, C. H. (Editor), The Geology of Pennsylvania: Pennsylvania Geological Survey, Harrisburg, PA, and Pittsburgh Geological Survey, Pittsburgh, PA, pp. 148–169. FENNEMAN, N. M., 1938, Physiography of Eastern United States: McGraw-Hill, New York, 714 p. HARBAUGH, A. W.; BANTA, E. R.; HILL, M. C.; AND MCDONALD, M. G., 2000, The U.S. Geological Survey Modular GroundWater Model—User Guide to Modularization Concepts and the Ground-Water Flow Process: U.S. Geological Survey Open-File Report 00-92, p. 121. HARLOW, G. E., JR. AND LECAIN, G. D., 1993, Hydraulic Characteristics of, and Groundwater Flow in, Coal Bearing Rocks of Southwestern Virginia: U.S. Geological Survey Water-Supply Report 2388, 36 p. HAWKINS, J. W. AND DUNN, M., 2007, Hydrologic characteristics of a 35-year-old underground mine pool: Mine Water and the Environment, Vol. 26, No. 3, pp. 150–159. HENNEN, R. V. AND REGER, D. B., 1913, Marion, Monongalia, and Taylor Counties: West Virginia Geological and Economic Survey, Morgantown, WV, 844 p. HOBBA, W. A., 1991, Relation of Fracture Systems to Transmissivity of Coal and Overburden Aquifers in Preston County, West Virginia: U.S. Geological Survey Water-Resources Investigations Report 89-4137, 24 p. KENDORSKI, F. S., 1993, Effect of high-extraction coal mining on surface and ground waters. In Peng, S. S. (ed.), Proceedings, 12th International Conference on Ground Control in Mining: West Virginia University, Morgantown, WV, pp. 412–425. KIPP, J. A. AND DINGER, J. S., 1987, Stress-relief facture control of ground-water movements in the Appalachian Plateaus. In NWWA Focus Conference on Eastern Regional Ground-Water Issues, July 1987: Burlington, VT, 6 p. KOZAR, M. D. AND MATHES, M. V., 2001, Aquifer-Characteristics Data for West Virginia: U.S. Geological Survey WaterResources Investigations Report 01-4036, 74 p. LEAVITT, B. R., 1999, Mine flooding and barrier pillar hydrology in the Pittsburgh basin. In Proceedings 16th Annual International Pittsburgh Coal Conference. Pittsburgh, PA, Paper 6-4. LIGHT, D. D. M., 2001, A Hydrogeological and Hydrogeochemical Characterization of the Truetown Mine Discharge: Unpublished M.S. Thesis, Department of Geological Sciences, Ohio University, Athens, OH, 271 p. MCCOY, K. J., 2002, Estimation of Vertical Infiltration into Deep Pittsburgh Coal Mines of WV-PA: A Fluid Mass Balance Approach: Unpublished M.S. Thesis, Department of Geology and Geography, West Virginia University, Morgantown, WV, 151 p. MCCOY, K. J.; DONOVAN, J. J.; AND LEAVITT, B. R., 2006, Horizontal hydraulic conductivity estimates for intact coal barriers between closed underground mines: Environmental & Engineering Geoscience, Vol. 12, No. 3, pp. 273–282. MCDONOUGH, K. M.; LAMBERT, D. C.; MUGUNTHAN, P.; AND DZOMBAK, D. A., 2005, Hydrologic and geochemical factors governing chemical evolution of discharges from an abandoned, flooded, underground coal mine network: Journal of Environmental Engineering, Vol. 131, No. 4, pp. 643–650. MILLER, L., 2000, Warwick Coal Mines: Greene County, Pennsylvania: Rita Miller Publishing, 256 p. PALCHIK, V., 2002, Influence of physical characteristics of weak rock mass on height of caved zone over abandoned subsurface coal mines: Environmental Geology, Vol. 42, No. 1, pp. 92–101. PALCHIK, V., 2003, Formation of fractured zones in overburden due to longwall mining: Environmental Geology, Vol. 44, No. 1, pp. 28–38.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164
163
Light and Donovan PENG, S. S., 1986, Coal Mine Ground Control, 2nd Ed.: John Wiley & Sons, New York, 491 p. PIGATI, E. AND LOPEZ, D. L., 1999, Effect of subsidence on recharge at abandoned mines generating acidic drainage: The Majestic Mine, Athens County, OH: Mine Water and the Environment, Vol. 18, No. 1, pp. 45–66. SAHU, P. AND LOPEZ, D. L., 2009, Using time series analysis of coal mine hydrographs to estimate storage, retention time, and mine-pool interconnection: Mine Water and the Environment, Vol. 28, No. 3, pp. 194–205. SINGH, M. M. AND KENDORSKI, F. S., 1981, Strata disturbance prediction for mining beneath surface water and waste impoundments. In Peng, S. S. (ed.), Proceedings of the 1st Conference on Ground Control in Mining: West Virginia University, Morgantown, WV, pp. 76–89. STOERTZ, M. W.; HUGHES, M. L.; WANNER, N. S.; AND FARLEY, M. E., 2001, Long-term water quality trends at a sealed, partially flooded underground mine: Environmental & Engineering Geoscience, Vol. 12, No. 1, pp. 51–65. STONER, J. D., 1983, Probable hydrologic effects of subsurface mining: Groundwater Monitoring Review, Vol. 3, No. 1, pp. 128–137.
164
STONER, J. D.; WILLIAMS, D. R.; BUCKWALTER, T. K.; FELBINGER, J. K.; AND PATTISON, K. L., 1987, Water Resources and the Effects of Coal Mining, Greene County, Pennsylvania: Pennsylvanian Geological Survey Water Resource Report 63, 4th Series, 166 p. WHITWORTH, K. R., 2002, The monitoring and modelling of mine water recovery in UK coalfields. In Younger, P. L. and Robins, N. S. (Editors), Mine Water Hydrogeology and Geochemistry: Geological Society of London Special Publication 198, pp. 61–73. WILLIAMS, D. R.; FELBINGER, J. K.; AND SQUILLACE, P. J., 1993, Water Resources and the Hydrologic Effects of Coal Mining in Washington County, Pennsylvania: U.S. Geological Survey Open-File Report 89-620. WINTERS, R. W. AND CAPO, R. C., 2004, Ground water flow parameterization of an Appalachian coal mine complex: Groundwater, Vol. 42, No. 5, pp. 700–710. WYRICK, G. G. AND BORCHERS, J. W., 1981, Hydrologic Effects of Stress-Relief Fracturing in an Appalachian Valley: U.S. Geological Survey Water-Supply Paper 2177, 51 p. YOUNGER, P. L. AND ADAMS, R., 1999, Predicting Mine Water Rebound: Environment Agency R & D Technical Report W179, Bristol, UK, 109 p.
Environmental & Engineering Geoscience, Vol. XXI, No. 2, May 2015, pp. 147–164