Environmental & Engineering Geoscience AUGUST 2017
VOLUME XXIII, NUMBER 3
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 & Engineering Geologists (AEG) and the Geological Society of America (GSA). Periodicals postage paid at AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7303 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 43701-7303. 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 43701-7303. 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 digital access to the journal as part of their AEG membership dues. Members may order print subscriptions for $60 per year. GSA members who are not members of AEG may order for $60 per year on their annual GSA dues statement or by contacting GSA. Nonmember subscriptions are $295 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 844-331-7867; contact GSA Subscription Services, Geological Society of America, P.O. Box 9140, Boulder, CO 80301. Single copies are $75.00 each. Requests for single copies should be sent to AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7303. © 2017 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 ABDUL SHAKOOR Department of Geology Kent State University Kent, OH 44242 330-672-2968 ashakoor@kent.edu
EDITORIAL BOARD JEROME V. DEGRAFF CSU Fresno CHESTER (SKIP) F. WATTS Radford University THOMAS OOMMEN Michigan Technological Univ. SYED E. HASAN University of Missouri
THOMAS J. BURBEY Virginia Polytechnic Institute ABDUL SHAKOOR Kent State University BRIAN G. KATZ Florida Department of Environmental Protection
ASSOCIATE EDITORS JOHN W. BELL Nevada Bureau of Mines and Geology RICHARD E. JACKSON Geofirma Engineering, Ltd. JEFFREY R. KEATON AMEC Americas PAUL G. MARINOS National Technical University of Athens, Greece JUNE E. MIRECKI U.S. Army Corps of Engineers PETER PEHME Waterloo Geophysics, Inc NICHOLAS PINTER Southern Illinois University
PAUL M. SANTI Colorado School of Mines ROBERT L. SCHUSTER U.S. Geological Survey ROY J. SHLEMON R. J. Shlemon & Associates, Inc. GREG M. STOCK National Park Service RESAT ULUSAY Hacettepe University, Turkey CHESTER F. “SKIP” WATTS Radford University TERRY R. WEST Purdue 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.
EDITORS
BRIAN G. KATZ Florida Department of Environmental Protection 2600 Blair Stone Rd. Tallahassee, FL 32399 850-245-8233 eegeditorbkatz@gmail.com
Cover photo An image of a railroad corridor through a canyon in southeastern Nevada, taken by an unmanned aerial vehicle (UAV), showing an unstable slope with rockfalls that have periodically blocked the access road and railroad tracks – photo courtesy of field team from Michigan Tech Research Institute (MTRI) – see article on page 165.
Environmental & Engineering Geoscience Volume 23, Number 3, August 2017 Table of Contents
153
Sediment and Phosphorus Inputs from Perennial Streams to Lake Whatcom, Washington State Katherine R. Beeler and Robert J. Mitchell
165
Rockfall Hazard Rating System: BeneďŹ ts of Utilizing Remote Sensing El Hachemi Bouali, Thomas Oommen, Stanley Vitton, R¨udiger Escobar-Wolf, and Colin Brooks
179
A New Application of CurvaTool Semi-Automatic Approach to Qualitatively Detect Geological Lineaments Sabrina Bonnetto, Anna Facello, and Gessica Umili
191
Effects of Igneous Intrusions on Coal Port Structure, Methane Desorption and Diffusion within Coal, and Gas Occurrence Ming-Yi Chen, Yuan-Ping Cheng, Hong-Xing Zhou, Liang Wang, Fu-Chai Tian, and Kan Jin
209
Channel Geomorphic Evolution after Dam Removal: Is Scale Important? Timothy Land, Arpita Nandi, and Ingrid Luffman
221
A Proposed Risk-Based Strategy for Bridges Potentially Affected by Rock Scour William L. Niemann, Issac C. Wait, and Jeffrey R. Keaton
Sediment and Phosphorus Inputs from Perennial Streams to Lake Whatcom, Washington State KATHERINE R. BEELER Associated Earth Sciences, Inc., 29111/2 Hewitt Avenue, Suite 2, Everett, WA 98201
ROBERT J. MITCHELL1 Western Washington University, Department of Geology, 516 High Street, Bellingham, WA 98225
Key Terms: Lake, Mass Wasting, Sedimentation, Suspended Solids, Phosphorus, Discharge ABSTRACT Relationships among suspended sediment, phosphorus, and discharge vary temporally and spatially in the Lake Whatcom watershed, a 125-km2 , high-relief, moderately developed, forested basin in northwestern Washington State. The lake is subject to a Total Maximum Daily Load to limit phosphorus inputs. Phosphorus, which largely enters the lake adsorbed to suspended sediment in streams, has led to increased algae growth and depletion of dissolved oxygen. We used the results of high-resolution storm event sediment and phosphorous sampling in five streams to examine the effects of varying watershed features on loading and to develop sedimentdischarge and phosphorus-discharge models to estimate phosphorus loading to the lake during the 2013 water year. During most storm events, the sediment peak preceded the discharge peak. The magnitude of hydrograph rise was the best predictor of the maximum sediment concentration during the event. Of the five basins studied, a large, forested watershed that contains steep slopes susceptible to mass wasting yielded the most sediment per area. The highest phosphorus yield was from a smaller, lower-relief watershed containing 29 percent residential development. Our sediment and phosphorous yields were comparable to estimates from similar streams in the Puget Sound region in northwest Washington State. Total suspended solids and total phosphorus were significantly correlated to discharge in most streams in the watershed, but variability within and among storm events resulted in uncertainty when calculating fluxes based on discharge.
1 Corresponding
author email: robert.mitchell@wwu.edu
INTRODUCTION Understanding phosphorus transport in watersheds is important because elevated phosphorus is one of the most common causes of lake impairment in the United States (EPA, 2014). Algae growth resulting from elevated phosphorus inputs can cause dissolved oxygen concentrations to decrease as bacteria metabolize algal carbon. Both algae and dissolved oxygen depletion are degrading the water quality in Lake Whatcom, located in northwestern Washington State (WA; Figure 1). Quantifying phosphorous loading and designing mitigation strategies are challenging because much of the phosphorus enters the lake adsorbed to suspended sediment in perennial streams that discharge to the lake (Matthews et al., 2014). The amount of suspended sediment in streams is controlled by sediment availability and sediment transport capacity, which, in turn, depend on a wide range of hydrologic and watershed factors (Asselman, 1999; Gellis, 2012). Sediment can be eroded from the banks and bed of the channel or from hillslopes and roads in the surrounding watershed (Lu and Richards, 2008). A study of the Issaquah Creek watershed (Nelson and Booth, 2002)—which is located 30 km southeast of Seattle, WA, and is similar to the Lake Whatcom watershed in size, relief, landcover, and climate—found that its main sources of sediment were landslides, channel bank erosion, and road-surface erosion. Landslides were the dominant sediment source in forested areas and contributed the greatest mass of sediment to the creek. Sediment transport largely occurs during periods of high discharge, such that the amount of sediment moved during occasional high-flow events often exceeds the total transport during longer periods of low flow (Swanson et al., 1982). A review of longterm suspended sediment records found that among 77 Pacific Northwest catchments, an average of 52.8 percent of the annual suspended sediment load was produced during the 15 days of the year with the highest streamflow (Gonzalez-Hidalgo et al., 2010). As of 1998, Lake Whatcom has been subject to a Total Maximum Daily Load (TMDL) to limit
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
153
Beeler and Mitchell
Figure 1. Location of the Lake Whatcom watershed in northwest Washington State and the five basins used in this study.
phosphorus inputs. The Washington State Department of Ecology used a landcover-based phosphorus loading model coupled with a lake water quality and hydrodynamic model to determine the total amount of phosphorus that can be discharged to Lake Whatcom without causing dissolved oxygen concentrations to drop below acceptable levels (Pickett and Hood, 2008). The phosphorus loading model was based on weather and land use conditions from the 2003 water year (WY) and calibrated to measured streamflow and phosphorus concentrations. Loading goals were set to represent the amount of phosphorus the lake could assimilate during a year based on the precipitation, temperature, and wind conditions that oc-
154
curred during WY 2003 (Hood, 2013). The study was limited in that 2002–2003 was an unusually dry year, the calibration included a relatively small amount of phosphorus data, and sediment loading was not calculated. We aimed to determine how streamflow variability during rainstorm events and basin characteristics affect sediment and phosphorus loading to Lake Whatcom. Storm data are particularly useful because the amount of sediment and phosphorus transport is normally higher during periods of elevated streamflow. We used the results of high-resolution storm event sampling along with hydrologic and other watershed data to determine relationships among total
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
Sediment and Phosphorus Inputs to Lake Whatcom
phosphorus (TP), total suspended solids (TSS), and stream discharge within the Lake Whatcom watershed. We examined how these relationships varied among storm events and among different streams in the watershed and developed a set of linear stream discharge models to estimate fluxes of phosphorus and suspended sediment to the lake during WY 2013. Understanding sediment and nutrient inputs will ultimately assist water managers in modeling and mitigating water quality issues in Lake Whatcom. STUDY SITE Lake Whatcom is of glacial origin and has a surface area of 20.3 km2 . The southern portion of the lake (basins 3N and 3S; Figure 1) contains about 96 percent of the 0.95-km3 lake volume (Mitchell et al., 2010). The area of the Lake Whatcom watershed is about 126 km2 , not including the lake itself (Pickett and Hood, 2008). The watershed has high relief, with elevations ranging from 95 to 1,027 m above mean sea level (Mitchell et al., 2010). The underlying geology consists of two bedrock units: the Chuckanut Formation and the Darrington Phyllite, as well as unconsolidated glacial and alluvial sediments (Lapen, 2000). Soils in the watershed are mainly classified as loam (Miller and White, 1998). Most of the watershed (81 percent) is vegetated with a combination of deciduous, evergreen, and mixed forest. The watershed was extensively logged near the beginning of the 1900s and up through the 1940s (WADNR, 1997). Currently, about 80 percent of the forest cover is mature; the remaining 20 percent is immature as a result of periodic harvesting over the past 40 years (WADNR, 1997; Kennedy et al., 2010). Residential development covers an additional 7 percent of the watershed, with the remainder mostly consisting of shrubland, grassland, and wetland. Less than 1 percent of the watershed area is used for agriculture (NOAA, 2011). Developed areas are mostly concentrated at the northwest end and along the central west side of Lake Whatcom. These areas also contain the highest density of paved roads; unpaved logging roads dominate the upper regions of the watershed (Figure 1; WADNR, 2016). The Lake Whatcom watershed is divided into basins, five of which (Anderson, Austin, Brannian, Silver Beach, and Smith Creeks) are analyzed in this study (Figure 1). Water enters Lake Whatcom primarily through surface runoff, groundwater, and direct precipitation onto the lake and, to a small degree, through intermittent diversion via a pipeline from the Middle Fork of the Nooksack River. Diverted water enters a settling pond and then flows to Lake Whatcom via Anderson Creek (Figure 1; Tracy, 2001). Surface water inputs comprise
perennial and intermittent streams, surface runoff directly into the lake, and engineered drainage systems (Delahunt, 1990; Pickett and Hood, 2008). The lake level is partially controlled by a dam at the head waters of Whatcom Creek, the only natural surface outlet of the lake at the northwest end of Basin 1 (Figure 1). Outputs from Lake Whatcom include evaporation, the outlet at Whatcom Creek, and water removed for municipal and industrial use. The lake serves as the drinking water source for approximately 100,000 people in the city of Bellingham, WA, and the surrounding areas (Hood, 2013). Precipitation in the Lake Whatcom watershed is distributed across frequent, low-intensity rainfall events and occasional high-intensity storms, especially between the months of October and April, which typifies the region’s maritime climate. Average annual rainfall is higher in the southern part of the watershed than in the northern part as a result of storm patterns and an orographic effect caused by the high relief in the watershed. Between WY 2002 and WY 2013, the average recorded yearly rainfall ranged from 101 cm at the north end of the lake to 151 cm at the south end of the lake. Precipitation varies from year to year in the Lake Whatcom watershed. The Lake Whatcom TMDL is based on conditions from WY 2003, which was the driest year in the period from 2002 to 2013 (approximately 77 percent of the 30-year regional precipitation mean). In contrast, our sampling period (WY 2013) occurred during one of the wettest of recent years (approximately 130 percent of the 30-year regional precipitation mean).
Effect of Phosphorus Inputs on Lake Water Quality Phosphorus is the limiting nutrient for biological productivity in Lake Whatcom, controlling the growth of algae and other vegetation such that increased inputs of phosphorus lead to more algae growth (Matthews et al., 2002). As bacteria metabolize algal carbon, they consume large amounts of oxygen from the water (Coveney and Wetzel, 1989). Resulting problems include loss of aquatic habitat and release of other contaminants due to anoxic lake conditions. From a drinking water standpoint, algal blooms may be problematic because they can clog water intake filters, and some algae can produce unpleasant tastes and odors in drinking water (Matthews et al., 2014). In watershed settings, phosphorus mainly occurs adsorbed to sediment particles (Lee et al., 2012). In the Lake Whatcom watershed, more than half of the adsorbed phosphorus is thought to be available for biota to use and thus has the potential to contribute to algae growth (Liang, 1994; Groce, 2011). Therefore, for management
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
155
Beeler and Mitchell
purposes, it is important to characterize stream sediment discharging to the lake.
automated nutrient analyzer. Detection limits were 2 mg/L for TSS and 5 g/L for TP. Storm Parameters
MATERIALS AND METHODS Sampling and Laboratory Analysis We collected discharge data and water samples for TSS and TP analyses near the mouth of each creek in five basins (Figure 1) in accordance with U.S. Geological Survey (USGS) protocols for measuring fluvial sediment (Rantz, 1982; Edwards and Glysson, 1999). A stream gauge at each stream recorded the water level (stage) at 15-minute intervals. Discharge is measured regularly (weekly to several times per year) at each of the gauging stations. Updated stage-discharge rating curves, unique to each stream, were used to produce discharge hydrographs at 15-minute resolution. We used Teledyne Isco automated water samplers to collect a series of samples over the duration of each storm event. In most streams, coarser particles are concentrated near the streambed, whereas fine sediment tends to be uniformly distributed throughout the water column (Edwards and Glysson, 1999). The sample intake was placed approximately 20 cm above the bottom of the stream, with the goal of obtaining representative suspended sediment concentrations for the selected location along the stream. For each event, our objectives were to collect samples during both the rising leg and the falling leg of the hydrograph and to sample near the time of peak discharge. At Anderson, Austin, and Brannian Creeks, sample collection, discharge measurements, and stage monitoring all occurred at the same location. Samples were typically collected at intervals of equal flow volume, yielding 10–30 discrete samples per event. At Smith Creek, the sampling location was located about 30 m downstream from the gauging station. Here, the sampling interval was adjusted manually (1 to 4 hours) based on weather forecasts, readings from the water-level and velocity sensor at Smith Creek, and real-time stage data from Anderson and Olsen Creeks. This sampling strategy provided water quality data at a wide range of discharge values. The City of Bellingham, WA, provided precipitation data recorded at 15-minute intervals at four stations in the Lake Whatcom watershed. Stream water samples, including lab blanks and replicates, were analyzed for TSS and TP at the Institute for Watershed Studies (IWS) state-certified water quality lab at Western Washington University (WADOE, 2014; No. A543-12). The analysis for TSS involved running samples through a filter, determining the mass of the residue, and dividing by the sample volume. The analysis for TP was conducted on an OI Analytical FS3100 156
We examined hydrograph, precipitation, TSS, and TP data in more detail in the Smith Creek basin (Figure 1) because it is one of the largest undeveloped and forested basins in the watershed and thus serves as a baseline for understanding natural sedimentation processes. Storm event size was quantified by calculating peak discharge, magnitude and duration of rise, eventflow volume, and precipitation magnitude. In calculating event-flow volume, we estimated and removed baseflow by drawing a straight line from the start of the rising leg to a point on the falling leg where flow began to level off and integrated under the resulting storm hydrograph. If another event began before the falling leg reached an inflection point, the endpoint of the baseflow line was set at the start of rise of that next event. We used the maximum recorded TSS and TP to quantify the sediment and phosphorus response to each storm event. These parameters are typically underestimates because the exact moment of maximum concentration is likely to occur between automated sampling intervals (typically 1 to 2 hours). Precipitation data were collected at the North Shore weather station located northwest of the Smith Creek basin (Figure 1). Calculation of Sediment and Phosphorus Fluxes We analyzed the data and calculated the sediment and phosphorus fluxes from five streams in the Lake Whatcom watershed over WY 2013 using R, an opensource statistical analysis package (R Core Team, 2012). We plotted TSS and TP against discharge, applied a logarithmic transformation to TSS and TP to linearize the relationship, and fit a linear model to the transformed data (Helsel and Hirsch, 2002; USFS, 2007). The TSS data were uncensored and contain negative values (for samples that contain very little sediment, the mass before filtering may exceed the mass after filtering as a result of the limitations of the balance). When developing the TSS-discharge models, we added a constant (3.3) before transformation to avoid taking the logarithm of a negative number. We used the linear relationships to calculate three TSS values and three TP values for each 15-minute interval throughout the water year: one at the lower 95 percent confidence interval, one at the mean, and one at the upper 95 percent confidence interval. Duan’s smearing estimator was applied to correct for re-transformation bias when calculating TSS and TP from the log-transformed model (Duan, 1983). The bias occurs because regression predicts the mean of a normal distribution, and the transformed
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
Sediment and Phosphorus Inputs to Lake Whatcom
Figure 2. TP vs. TSS correlations for five streams in the Lake Whatcom watershed. Dashed lines represent 95 percent confidence intervals.
mean of the distribution is not equivalent to the mean of the transformed distribution (USFS, 2007). We estimated the sediment and phosphorus loading from each stream by multiplying flow volumes by TSS and ˜ TP concentrations (Glysson, 1987; Gray and Simoes, 2008). We also converted results from kilograms to tons and divided by the watershed area to determine yields in tons per square kilometer per year.
Basin Characteristics We compiled and analyzed digital watershed data for the five gauged sub-basins to determine whether sediment and phosphorus loading relate to spatial watershed features including basin area, relief, slope, drainage density, bedrock type, paved and unpaved roads, and degree of urban development. Sub-basins and spatial watershed features were delineated in ArcGIS 10.1 using a LiDAR bare earth terrain map with 2-m resolution (PSLC, 2006), landcover (NOAA, 2011), geology (WADNR, 2010), and roads (WADNR, 2016).
Correlation Analysis Correlation analysis is a method used to examine the monotonic relationship between two variables. We used Kendall’s tau ( ) rank-based correlations, calculated in R, to test for significant correlations between discharge, TSS, and TP over each stream’s full data set and within individual storm events. Kendall’s tau is resistant to the effects of outliers and well suited to data sets that exhibit a skewed distribution (Helsel and Hirsch, 2002). We compared the Smith Creek storm events to one another and tested for correlations among precipitation, discharge, and water quality parameters. The test statistic ranges from −1 to + 1; the closer to ± 1, the stronger the correlation. The p-value indicates statistical significance; significant correlations have p-values
of less than 0.05. Kendall’s tau values around 0.7 or above are considered strong correlations. RESULTS Sediment and phosphorus were correlated to one another in all of the sampled streams, but the relationship between them varied throughout the watershed (Figure 2). Sediment and phosphorus were significantly correlated to discharge in all of the sampled streams (Figure 3). The correlation between sediment and discharge tended to be stronger (higher Kendall’s tau) than the correlation between phosphorus and discharge. Although correlations between sediment, phosphorus, and discharge were statistically significant for most sites, there was a high degree of variability within each site. Ratios of phosphorus to sediment tended to be relatively high in Silver Beach Creek (Figure 2). Silver Beach Creek had higher levels of sediment and phosphorus relative to discharge when compared to Anderson, Austin, Brannian, and Smith Creeks (Figure 3). Among the five basins, mean calculated sediment fluxes for WY2013 were highest at Smith Creek, and phosphorus fluxes were highest at Austin Creek (Table 1). The Smith Creek watershed produced the most sediment per square kilometer, and the Silver Beach Creek watershed produced the most phosphorus per square kilometer (Table 2). We also calculated fluxes from the five basins by month, revealing that the highest sediment loads came from Smith Creek during large winter storm events, and most phosphorus loading from the five streams occurred between November and May, the rainier months of the year. We examined hydrograph, TSS, and TP data in more detail in the Smith Creek basin because it serves as a forested baseline basin, with the lowest percentage of development of the five basins studied (Table 3). In the Smith Creek basin, about 87 percent of the forest cover is mature and about 13 percent immature as a result
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
157
Beeler and Mitchell Table 1. Calculated suspended sediment and phosphorus fluxes from five basins of the Lake Whatcom watershed, WY 2013. Suspended Sediment (kg/yr) n Anderson Austin Brannian Silver Beach Smith
294 225 211 566 497
Lower 95% CI 96,500 273,000 94,500 157,000 1,190,000
Mean 119,000 405,000 136,000 240,000 1,940,000
Phosphorus (kg/yr)
Upper 95% CI 146,000 609,000 194,000 378,000 3,200,000
Lower 95% CI
Mean
Upper 95% CI
395 450 182 179 322
461 549 244 212 431
539 677 331 256 599
CI = confidence interval.
of logging activities over the last 40 years (WADNR, 1997; Kennedy et al., 2010). Peaks in stream discharge typically followed peaks in precipitation. During most events, sediment and phosphorus peaks were higher on the rising leg of the discharge hydrograph and lower on the falling leg, forming a clockwise hysteresis loop when plotted against discharge (Figure 4). Although the spatial characteristics are different for each of the five basins (Table 3), the TSS, TP, and discharge patterns (e.g., hysteresis) are similar. Sediment and phosphorus generally increased with discharge and were significantly correlated with each other in all of the Smith Creek events, but the TSS- and TP-discharge relationships were unique for each storm event (Fig-
ure 5). Sediment was significantly correlated with discharge in 18 of 22 Smith Creek events, and phosphorus was significantly correlated with discharge in 16 of 22 events. Sediment and phosphorus were usually correlated more strongly to one another than to discharge (Figures 2 and 3). Total storm event rainfall for the Smith Creek events ranged from 0.660 to 4.55 cm, as measured at the North Shore weather station (Table 4 and Figure 1). Large precipitation events tended to produce large hydrograph peaks (Table 4). Event rainfall was correlated significantly, but weakly, to peak discharge ( = 0.424), magnitude of hydrograph rise ( = 0.493), duration of rise ( = 0.35), event-flow volume ( = 0.483), and shorter
Figure 3. TSS vs. discharge and TP vs. discharge correlations for five streams in the Lake Whatcom watershed. Dashed lines represent 95 percent confidence intervals.
158
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
Sediment and Phosphorus Inputs to Lake Whatcom Table 2. Calculated suspended sediment and phosphorus yields from five basins of the Lake Whatcom watershed, WY 2013. Suspended Sediment (tons/km2 /yr) Anderson Austin Brannian Silver Beach Smith
n 294 225 211 566 497
Lower 95% CI 9.31 12.8 10.8 50.5 87.8
Mean 11.5 18.9 15.5 77.6 143
Phosphorus (kg/km2 /yr)
Upper 95% CI 14.1 28.5 22.2 122 236
Lower 95% CI 38.1 21.0 20.8 57.6 23.7
Mean 44.5 25.7 27.9 68.5 31.7
Upper 95% CI 52.0 31.6 37.8 82.6 44.1
CI = confidence interval.
term (3 days; = 0.436) and longer term (15 days; = 0.439) antecedent precipitation. Although rainfall and hydrograph rise were correlated, rainfall alone did not necessarily predict flow. For example, a 2.9-cm precipitation event in late September only resulted in a 0.28-m3 /s hydrograph rise due to precipitation loss to soil storage, whereas a similar-sized rain event in late December (2.9 cm of precipitation) increased discharge by 1.9 m3 /s because antecedent soil conditions were closer to field capacity (Table 4). When normalized to precipitation magnitude, the flow response was highest in the late winter, decreased and remained low through the summer, and increased again in the fall. Maximum sediment was significantly correlated to the magnitude of rise ( = 0.717), peak discharge ( = 0.657), event-flow volume ( = 0.506), and event rainfall ( = 0.5), but not to the duration of rise. Maximum phosphorus was correlated to event rainfall ( = 0.404), but not to any of the hydrograph magnitude parameters. DISCUSSION Given that the high-relief Smith Creek basin is almost entirely forested with the highest unpaved road density (Table 3), the stream’s main sources of sediment are likely to be the erosion of mass wasting deposits
(sediments deposited during major slope movements, such as landslides) and channel erosion. Historically, mass wasting has occurred in the Smith Creek watershed during rainstorms, but we found no evidence of failures during our sampled storm events. However, it is possible that mass wasting occurred between sampling periods. Event maximum sediment concentrations consistently occurred near hydrograph peaks and are strongly correlated with the magnitude of rise, with no unusual spikes that would signal a large mass wasting event. The observed sediment peaks almost always follow precipitation peaks. Among the events that we sampled at Smith Creek, discharge, rather than sediment and phosphorus concentration, was the main factor influencing loading. Of the hydrograph magnitude parameters, magnitude of rise was the best predictor of sediment and phosphorus response to a storm event. The relationship between hydrograph rise and maximum TSS was fairly consistent among storm events from different seasons and of different magnitudes. The correlation suggests that the increase in discharge, rather than peak flow, total flow volume, time of year, or antecedent rainfall, is the most important factor to consider when predicting stream sediment concentrations. The strong relationship between hydrograph rise and sediment loading makes sense in the context of Smith Creek sediment sources. Streams have a lot of energy to erode and suspend sed-
Table 3. Watershed parameters for five basins of the Lake Whatcom watershed.
Anderson Austin Brannian Silver Beach Smith
Area (km2 )
Drainage density (km−1 )
Main Bedrock Type
Relief (m)
Mean Slope (◦ )
Percent Developed
Paved Road Density (km−1 )
Unpaved Road Density (km−1 )
10.4 21.4 8.7 3.1c 13.6
3.18 2.95 3.17 4.23 2.22
DPa CFb DP CF CF
800 722 772 362 835
19.9 19.1 16.0 8.2 26.5
0.26 5.63 0.26 28.6 0.06
0.48 2.19 0.60 2.98 0.14
3.01 2.38 3.72 0.65 3.88
DP = Darrington Phyllite. CF = Chuckanut Formation. c The watershed area for Silver Beach Creek is based on the catchment area reported by the USGS (2014). This value was selected because of discrepancies in how the watershed was delineated in the LiDAR basins shapefiles and TMDL basins map. a
b
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
159
Beeler and Mitchell
Figure 4. Precipitation, discharge, TSS, and TP data collected during a representative storm event (15–17 November 2013) at Smith Creek.
iment in and around the channel when discharge is higher, and many of the storms with high increases in discharge also had high peak flows. The magnitude of rise also takes antecedent flow into account. A small storm that begins when discharge is already high might reach a high peak flow but produce relatively little sediment because the previous flow has already eroded the most readily available material. Magnitude of rise is better than event-flow volume for predicting sediment peaks because sediment peaks occur on the rising leg of the hydrograph and are thus largely indifferent to the slope of the recession curve, which greatly affects
flow volume estimates. The event-flow volume also has a level of uncertainty due to baseflow separation inaccuracies. Relief, slope, bedrock lithology, soil type, density of unpaved roads, and urban development likely explain the differences in sediment and phosphorus yields among the Lake Whatcom basins. The Smith Creek basin has high unpaved road density, high relief, and forested slopes (Table 3). Its steep channels are susceptible to mass wasting at large and small scales, which contributes sediment to the stream (Syverson, 1984; Buchanan and Savigny, 1990; and WADNR, 1997). Al-
Figure 5. Relationships between sediment and phosphorus concentrations and discharge at Smith Creek, with events separated by season. The overall linear model is a compilation of many distinct trends.
160
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
Sediment and Phosphorus Inputs to Lake Whatcom Table 4. Parameters calculated for Smith Creek storm events. Event Dates
W
W3
W15
Qpk
Qr
Tr
EFV
TSSmax
TPmax
22–24 Feb 2013 25–26 Feb 2013 28 Feb–3 Mar 2013 12–14 Mar 2013 6–7 Apr 2013 7–9 Apr 2013 10–12 Apr 2013 21–22 May 2013 20–21 Jun 2013 2–3 Aug 2013 29–30 Aug 2013 6–7 Sep 2013 16–17 Sep 2013 22–24 Sep 2013 28–29 Sep 2013 1–3 Nov 2013 6–8 Nov 2013 15–17 Nov 2013 30 Nov–2 Dec 2013 23–24 Dec 2013 2–4 Jan 2014 11–13 Jan 2014
2.0 1.9 2.5 2.4 1.4 1.7 1.4 1.8 4.6 1.0 0.9 0.7 1.8 2.9 2.5 3.1 2.0 1.8 1.3 2.9 2.3 4.5
2.34 2.82 2.49 2.67 4.12 5.72 2.18 1.65 4.37 1.02 1.52 1.42 2.62 3.05 4.37 4.14 2.06 2.54 2.03 2.92 2.44 4.70
3.28 5.03 7.52 7.30 4.12 5.87 7.06 5.59 5.56 1.02 2.36 4.57 4.90 5.79 11.2 4.45 7.47 9.68 6.22 3.56 6.78 9.17
0.73 0.76 2.2 1.2 0.85 1.1 1.1 0.21 0.76 0.033 0.028 0.033 0.54 0.32 1.6 0.57 0.68 1.2 0.46 1.4 0.79 2.4
0.51 0.40 1.8 0.91 0.49 0.48 0.57 0.11 0.72 0.022 0.018 0.022 0.52 0.28 1.4 0.47 0.48 0.93 0.26 1.2 0.52 2.0
15.0 4.5 32.5 32.5 2.5 9.8 3.8 8.8 13.3 8.5 9.8 7.5 4.0 14.0 6.3 7.0 22.8 12.3 6.3 20.0 9.0 29.8
25,000 16,000 390,000 110,000 7,000 57,000 18,000 3,400 18,000 460 610 650 8,700 11,000 27,000 20,000 48,000 69,000 3,900 110,000 41,000 270,000
32.4 13.3 77.3 29.2 44.8 32.9 24.8 11.7 35.8 8.95 6.30 4.55 35.3 20.8 188 34.9 17.4 57.3 17.6 80.4 29.6 163
34.5 17.4 43.7 31.1 43.1 32.2 28.5 38.3 74.4 58.1 25.2 32.0 79.7 41.9 173 57.7 59.5 70.2 24.7 98.0 34.7 142
W = event rainfall (cm); W3 = rainfall in the 3 days preceding the event (cm); W15 = rainfall in the 15 days preceding the event (cm); Qpk = discharge peak (m3 /s); Qr = hydrograph rise (m3 /s); Tr = duration of hydrograph rise (hours); EFV = event-flow volume (m3 ); TSSmax = maximum recorded TSS (mg/L); TPmax = maximum recorded TP (g/L).
though mass wasting does not appear to have occurred during the events that we sampled, erosion of existing mass wasting deposits is the likely source for the relatively high sediment yields from the Smith Creek watershed. The basins of Anderson, Austin, and Brannian Creeks also contain unpaved roads and steep slopes (Table 3), but they produce less sediment per area (Table 2). Differences in bedrock type partly account for the differences in yields. The shallow soil deposits on the Chuckanut Formation are more susceptible to shallow (infinite slope) failures because they tend to slip along the surface of the bedrock when saturated. Water permeates the soil and collects at the soil-bedrock interface, increasing pore pressure, decreasing shear strength, and causing shallow-soil failures. The Anderson and Brannian Creek basins are underlain by Darrington Phyllite rather than the Chuckanut Formation. Although landslides occur in watersheds underlain by Darrington Phyllite, soil slippage on hillslopes is less common than in the Chuckanut Formation because the phyllite is more permeable, allowing soils to drain more quickly. The phyllite, in general, is more susceptible to deep-seated landslides rather than shallow infinite slope failures (WADNR, 1997). In addition, Anderson Creek is unique in that it includes flows from the Middle Fork Nooksack River diversion, which could influence sediment yields. Fine sediment in the diverted water settles in a small lake
before entering Anderson Creek, decreasing suspended sediment concentrations at the sampling point downstream. Samples collected while the diversion was operating had relatively low ratios of TSS to discharge, suggesting that dilution is the dominant process by which the diverted water influences sediment concentrations in Anderson Creek. The effects of the diversion may partly account for the relatively weak correlation between TSS and discharge at Anderson Creek. The Austin Creek and Smith Creek basins have similar bedrock and landcover, but sediment yields were substantially higher at Smith Creek. The Austin Creek basin has more urban development, lower relief, and a lower mean slope angle than the Smith Creek basin (Figure 1 and Table 3). The Austin Creek TSSdischarge model was also highly influenced by its two largest events, which had unusually low ratios of TSS to discharge (Figure 3). The Silver Beach Creek basin has lower relief than the Anderson, Austin, Brannian, and Smith Creek basins, but it also has a somewhat higher drainage density and percentage of urban development (Table 3). Channel erosion may play a greater role in generating sediment in the smaller, urbanizing watersheds than in the larger, forested watersheds because of differences in relief, landcover, and bed material. Shallower slopes decrease the likelihood of mass wasting, and lowerrelief basins tend to have thicker soils. The presence
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
161
Beeler and Mitchell
of impervious surfaces in developed areas may result in greater runoff and higher erosion rates as the water level rises during storm events. Water may be delivered to the streams more rapidly, leading to higher storm flows and more erosion. In addition, high flows may have scoured out the channels of the larger creeks over time, leaving behind coarse sediment that is more difficult to suspend. The Silver Beach Creek basin has a relatively low density of unpaved roads, which may contribute to its relatively low sediment yields. Variation in the phosphorus-sediment ratio in different streams in the Lake Whatcom watershed could be the result of differences in water or soil chemistry. Higher TP-TSS ratios were associated with areas with more urban development and agricultural influences. The relatively high TP-TSS ratios observed in the Anderson Creek water samples could reflect relatively high concentrations of phosphorus in soil or organic inputs from pastureland and wetlands that reside in the lower reaches in the Anderson Creek basin. Although the Middle Fork Nooksack River carries fine glacial sediment containing little organic material and, thus, little particulate phosphorus, higher concentrations of phosphorus, relative to sediment, occurred even when the diversion was on, likely as a result of lateral erosion of stream channels in lower relief pasturelands and wetlands. Higher phosphorus yields from the Silver Beach Creek basin could be due to more developed soils, but they are more likely associated with anthropogenic sources such as fertilizers, detergents, and wastewater, given that the basin contains about 30 percent urban development and a high paved-road density (Table 3). Several factors influence the accuracy of sediment and phosphorus flux estimates. The quality of loading estimates depends on hydrograph quality. Stagedischarge rating curves are often uncertain at high flows because the maximum stream stage exceeds the maximum stage at which discharge has been measured. The sediment data do not include bed load, which typically makes up 5–20 percent of the total sediment load (Czuba et al., 2011). Although bedload is a component of the total sediment flux to the lake, we focused on measuring the suspended load because phosphorus tends to be adsorbed to fine sediment carried in suspension (Stone and Mudroch, 1989). Coarser sediment is more likely to settle out before or shortly after entering the lake, so bedload may not have much effect on lake water quality. In addition, finer sediment has a higher ratio of surface area to mass and may contain more organic matter, making it a better carrier of adsorbed phosphorus. Sample collection times also affect the quality of flux estimates. The linear models are sensitive to individual storms and data points, particularly at high flows, so results can vary depending on which storms and samples
162
are included. Prediction of sediment concentrations based on discharge is limited because there is not a oneto-one relationship between sediment and flow. Combining data over long periods of time and many storm events (22 in Smith Creek) produced significant correlations between TSS and discharge, but the sedimentdischarge relationship was unique for each storm event (Figure 5). Even within individual events, the relationship was most often circular (hysteretic) rather than linear or exponential. The linear models (Figure 3) assume equal sediment concentrations at equal discharges on the rising and falling legs of the storm hydrograph. In reality, sediment concentrations were usually higher on the rising leg and lower on the falling leg (Figure 4). The models do not account for mass wasting or other sudden deliveries of sediment to the stream, such as the release of built-up sediment when debris is dislodged. Mass wasting can occur at any level of discharge and may result in unusually high sediment-discharge ratios, affecting load estimates (Chleborad et al., 2006). The calculated sediment yields based on our measurements and models were at the low end of the range of yields estimated for streams in the Pacific Northwest. Sediment yields on the order of 10 tons/km2 /yr (Anderson, Austin, and Brannian Creeks) are common in the Puget Lowland. Yields of around 100 tons/km2 /yr (Silver Beach and Smith Creeks) are more typical of the mountainous catchments in the region (Czuba et al., 2012). Yields from the Lake Whatcom watershed were comparable to those from the Issaquah Creek watershed, a similar catchment located southeast of Seattle, WA (44 tons/km2 /yr; Nelson and Booth, 2002). The range of calculated phosphorus yields is reasonable when compared with the findings of Embrey and Inkpen (1998), who estimated yields in the range of 24.5–105 kg/km2 /yr for four streams in the northern Puget Sound, and the USGS SPARROW model, which calculated an average phosphorus yield of 54 kg/km2 /yr in the Puget Sound region (Wise and Johnson, 2011, 2013). CONCLUSIONS Relationships among sediment, phosphorus, and discharge varied temporally and spatially in the Lake Whatcom watershed. Transport was limited by sediment availability and varied among basins according to spatial characteristics such as topography, bedrock lithology, and landcover. Sediment and phosphorus concentrations were significantly correlated to discharge in most streams, but sediment-discharge and phosphorus-discharge relationships were not consistent within or among storm events, which resulted in uncertainty when calculating fluxes based solely on discharge. At Smith Creek, the magnitude of hydrograph
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
Sediment and Phosphorus Inputs to Lake Whatcom
rise was the best predictor of the maximum sediment and phosphorus concentrations resulting from a storm event. Improving water quality in Lake Whatcom is necessary by law because the lake is currently impaired and on the Washington State 303(d) list. Our study provides a better understanding of sediment and phosphorus dynamics in the Lake Whatcom watershed, including which factors influence the amount of sediment and phosphorus that streams deliver to the lake, which will improve modeling estimates for lake management purposes. It also highlights the challenges of predicting fluxes for management and modeling purposes in natural systems with a high degree of variability. ACKNOWLEDGMENTS Thank you to Robin Matthews, Scott Linneman, Joan Vandersypen, and Mike Hilles for their contributions to this project. The research was funded by the Western Washington University (WWU) Office of Research and Sponsored Programs and the WWU Geology Department. We conducted analyses at the Institute for Watershed Studies state-certified water quality lab at WWU and used data provided by IWS, the City of Bellingham, Brown and Caldwell, and the USGS. REFERENCES ASSELMAN, N., 1999, Suspended sediment dynamics in a large drainage basin: The River Rhine: Hydrological Processes, Vol. 13, pp. 1437–1450. BUCHANAN, P. AND SAVIGNY, K. W., 1990, Factors controlling debris avalanche initiation: Canadian Geotechnical Journal, Vol. 27, No. 25, pp. 659–675. CHLEBORAD, A. F.; BAUM, R. L.; AND GODT, J. W., 2006, Rainfall Thresholds for Forecasting Landslides in the Seattle, Washington Area—Exceedance and Probability: U.S. Geological Survey Open-File Report, 2006-1064, 31 p. COVENEY, M. F. AND WETZEL, R. G., 1989, Bacterial metabolism of algal extracellular carbon: Hydrobiologia, Vol. 173, pp. 141– 149. CZUBA, J. A.; MAGIRL, C. S.; CZUBA, C. R.; GROSSMAN, E. E.; CURRAN, C. A.; GENDASZEK, A. S.; AND DINICOLA, R. S., 2011. Sediment Load from Major Rivers into Puget Sound and its Adjacent Waters: U.S. Geological Survey Fact Sheet 20113083, 4 p. CZUBA, J. A.; MAGIRL, C. S.; CZUBA, C. R.; QURRAN, C. A.; JOHNSON, K. H.; OLSEN, T. D.; KIMBALL, H. K.; AND GISH, C. C., 2012, Geomorphic Analysis of the River Response to Sedimentation Downstream of Mount Rainier, Washington: U.S. Geological Survey Open-File Report 2012-1242, 134 p. DELAHUNT, R., 1990, Lake Whatcom Watershed On-Site Sewage Disposal Survey: Final Report: Whatcom County Health Department Office of Environmental Health. DUAN, N., 1983, Smearing estimate: A nonparametric retransformation method: Journal American Statistical Association, Vol. 78, No. 383, pp. 605–610. EDWARDS, T. K. AND GLYSSON, G. D., 1999, Field Methods for Measurement of Fluvial Sediment Techniques of Water-Resources
Investigations: U.S. Geological Survey, Reston, Virginia, Book 3, Chapter C2. EMBREY, S. S. AND INKPEN, E. L., 1998, Water-Quality Assessment of the Puget Sound Watershed, Washington, Nutrient Transport in Rivers, 1980–93: U.S. Geological Survey Water-Resources Investigations Report 97-4270, 30 p. GELLIS, A. C., 2012, Factors influencing storm-generated suspended-sediment concentrations and loads in four watersheds of contrasting land use, humid-tropical Puerto Rico: Catena, Vol. 104, pp. 35–57. GLYSSON, G. D., 1987, Sediment-Transport Curves: U.S. Geological Survey Open-File Report 87-218, 47 p. GONZALEZ-HIDALGO, J. C.; BATALLA, R. J.; CERDA, A.; AND DE LUIS, M., 2010, Contribution of the largest events to suspended sediment transport across the USA: Land Degradation Development, Vol. 21, No. 2, pp. 83–91. ˜ , F. J. M., 2008, Estimating Sediment DisGRAY, J. R. AND SIMOES charge: Appendix D of Garcia, Marcelo, ed., Sedimentation Engineering—Processes, Measurements, Modeling, and Practice: American Society of Civil Engineers Manuals and Reports on Engineering Practice 110, 22 p. GROCE, S., 2011, Soils as a Source of Bioavailable Phosphorus in the Lake Whatcom Watershed: Unpublished M.S. Thesis, Western Washington University, Bellingham, WA, 146 p. HELSEL, D. R. AND HIRSCH, R. M., 2002, Statistical Methods in Water Resources Techniques of Water Resources Investigations, Book 4, chapter A3. U.S. Geological Survey, Reston Virginia. 510 p.: Electronic document, available at https://pubs.usgs.gov/twri/twri4a3/ HOOD, S., 2013, Lake Whatcom Watershed Total Phosphorus and Bacteria Total Maximum Daily Loads: Volume 2, Water Quality Improvement Report and Implementation Strategy: Washington State Department of Ecology Publication 13-10-012, 237 p. KENNEDY, R. E.; YANG, Z.; AND COHEN, W. B., 2010, Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr—Temporal segmentation algorithms: Remote Sensing Environment, Vol. 114, No. 12, pp. 2897–2910. LAPEN, T. J., 2000, Geologic Map of the Bellingham 1:100,000 Quadrangle, Washington: Washington Department of Natural Resources, Division of Geology and Earth Resources, Open-File Report 2000-5. LEE, K. E.; LORENZ, D. L.; PETERSEN, J. C.; AND GREENE, J. B., 2012, Seasonal Patterns in Nutrients, Carbon, and Algal Responses in Wadeable Streams within Three Geographically Distinct Areas of the United States, 2007–08: U.S. Geological Survey Scientific Investigations Report 2012-5086, 55 pp. LIANG, C. W., 1994, Impact of Soil and Phosphorus Enrichment on Lake Whatcom Periphytic Algae: Unpublished M.S. Thesis, Western Washington University, Bellingham, WA. LU, H. AND RICHARDS, K., 2008, Sediment Delivery: New Approaches to Modeling an Old Problem: River Confluences, Tributaries, and the Fluvial Network, Chapter 16: John Wiley and Sons, Inc., New York. 29 p. MATTHEWS, R.; HILLES, M.; AND PELLETIER, G., 2002, Determining trophic state in Lake Whatcom, Washington (USA), a soft water lake exhibiting seasonal nitrogen limitation: Hydrobiologia, Vol. 468, No. 1, pp. 107–121. MATTHEWS, R. A.; HILLES, M.; VANDERSYPEN, J.; MITCHELL, R. J.; MATTHEWS, G. B.; AND BEELER, K., 2014, Lake Whatcom Monitoring Project 2012/2013 Report: Electronic document, available at http://cedar.wwu.edu/lakewhat_annualreps/3/ MILLER, D. A. AND WHITE, R. A., 1998, A Conterminous United States Multi-Layer Soil Characteristics Data Set for Regional
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
163
Beeler and Mitchell Climate and Hydrology Modeling, Earth Interactions, 2: Electronic document, available at http://EarthInteractions.org MITCHELL, R. J.; GABRISCH, G.; AND MATTHEWS, R. A., 2010, Lake Whatcom Bathymetry and Morphology: Electronic document, available at http://cedar.wwu.edu/bathymetry/1/ NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION (NOAA), 2011, Coastal Change Analysis Program (C-CAP) Regional Land Cover Data: Electronic document, available at http://www.csc.noaa.gov/digitalcoast/data/ccapregional NELSON, E. J. AND BOOTH, D. B., 2002, Sediment sources in an urbanizing, mixed land-use watershed: Journal Hydrology, Vol. 264, pp. 51–68. PICKETT, P. AND HOOD, S., 2008, Lake Whatcom Watershed Total Phosphorus and Bacteria Total Maximum Daily Loads: Volume 1. Water Quality Study Findings: Washington State Department of Ecology Publication 08-03-024, 145 p. PUGET SOUND LIDAR CONSORTIUM (PSLC), 2006, Lidar Bare Earth of Western Whatcom County: Electronic document, available at http://pugetsoundlidar.ess.washington.edu/index.htm R CORE TEAM, 2012, R: A Language and Environment for Statistical Computing: Electronic document, available at http://www.Rproject.org RANTZ, S. E., 1982, Measurement and Computation of Streamflow, Volume 1—Measurement of Stage and Discharge: U.S. Geological Survey Water-Supply Paper 2175, 284. STONE, M. AND MUDROCH, A., 1989, The effect of particle size, chemistry and mineralogy of river sediments on phosphate adsorption: Environmental Technology Letters, Vol. 10, pp. 501– 510. SWANSON, F. J.; JANDA, R. J.; DUNNE, T.; AND SWANSTON, D. N., 1982, Sediment Budgets and Routing in Forested Drainage Watersheds: USDA Forest Service General Technical Report PNW-141, 22 p. SYVERSON, T., 1984, History and Origin of Debris Torrents in the Smith Creek Drainage, Whatcom County, WASHINGTON: Unpublished M.S. Thesis, Western Washington University, Bellingham, WA, 84 p.
164
TRACY, K., 2001, Changes in Mirror Lake, Northwestern Washington, as a Result of the Diversion of Water from the Nooksack River: Unpublished M.S. Thesis, Western Washington University, Bellingham, WA, 107 p. UNITED STATES ENVIRONMENTAL PROTECTION AGENCY (EPA), 2014, Water Quality Assessment and Total Maximum Daily Loads Information—National Summary of State Information: Electronic document, available at http://www.epa.gov/ waters/ir UNITED STATES FOREST SERVICE (USFS), 2007, Estimating Sediment Concentration and Load: Electronic document, available at http://www.fs.fed.us/psw/topics/water/tts/loads/ Rprocedures.pdf UNITED STATES GEOLOGICAL SURVEY (USGS), 2014, National Water Information System: Electronic document, available at http://waterdata.usgs.gov/nwis WASHINGTON STATE DEPARTMENT OF ECOLOGY (WADOE), 2014, Environmental Assessment Program—Laboratory Accreditation: Electronic document, available at http://www.ecy. wa.gov/programs/eap/labs/lab-accreditation.html WASHINGTON STATE DEPARTMENT OF NATURAL RESOURCES (WADNR), 1997, Lake Whatcom Watershed Analysis: Electronic document, available at https://fortress.wa.gov/ dnr/protectionsa/ApprovedWatershedAnalyses WADNR, 2010, GIS Data and Databases: Electronic document, available at http://www.dnr.wa.gov/programs-and-services/ geology/publications-and-data/gis-data-and-databases WADNR, 2016, GIS Data: Electronic document, available at http://www.dnr.wa.gov/GIS WISE, D. R. AND JOHNSON, H. M., 2011, Surface-water nutrient conditions and sources in the United States Pacific Northwest: Journal American Water Resources Association, Vol. 47, No. 5, pp. 1110–1135. WISE, D. R. AND JOHNSON, H. M., 2013, Application of the SPARROW Model to Assess Surface-Water Nutrient Conditions and Sources in the United States Pacific Northwest: U.S. Geological Survey Scientific Investigations Report 2013-5103, 42 p.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 153–164
Rockfall Hazard Rating System: Benefits of Utilizing Remote Sensing EL HACHEMI BOUALI1 THOMAS OOMMEN STANLEY VITTON ¨ RUDIGER ESCOBAR-WOLF Michigan Technological University, 1400 Townsend Drive, Houghton, MI 49931
COLIN BROOKS Michigan Tech Research Institute, 3600 Green Court, Suite 100, Ann Arbor, MI 48105
Key Terms: Remote Sensing, Rockfall, Hazard, Landslide, Engineering Geology
portation agencies to ensure a more robust, efficient, and time-effective RHRS approach.
ABSTRACT
INTRODUCTION
Transportation corridor slopes have the potential to be hazardous to adjacent assets. The Rockfall Hazard Rating System (RHRS) is a stepwise process designed to identify potentially hazardous slopes by assigning a hazard rating that determines the order by which to mitigate and remediate slopes. The traditional RHRS approach is field-based: observations are made by a field crew who convert observations into slope ratings (preliminary and detailed). The purpose of this study is to examine the benefits of utilizing remote sensing techniques on 14 slopes within a 24-km railroad corridor in southeastern Nevada. Remote sensing allows for data acquisition in difficult-to-reach locations from various view angles. Images and data from three remote sensing technique-platform combinations are examined: optical imagery acquired via satellite, unmanned aerial vehicle, and LiDAR data acquired from a stationary sensor. Detailed RHRS slope ratings from both sets of optical images are compared to two types of field-based ratings: (1) initial field observations performed using the traditional RHRS approach and (2) average detailed rating scores from six participants (geologists and geotechnical engineers) given field notes of the 10 rating criteria for the 14 slopes. Terrestrial LiDAR is capable of monitoring slow slope deformation, with an accuracy of approximately 2 cm/yr, and identifying areas of rapid deformation. Remote sensing techniques should not entirely replace traditional field methods. Instead, developing an approach that combines the advantages of field- and remote sensing–based methodologies will enable trans-
The Rockfall Hazard Rating System (RHRS) is a procedure used to analyze and prioritize slopes along transportation corridors (roadways and railways) based on the potential hazard of rockfall occurrence. A train derailment in British Columbia, Canada, in the early 1970s was the impetus for the development of the RHRS, which is commonly used today (Brawner and Wyllie, 1975). Rating criteria based on geometric and geologic conditions of the railroad-slope environment were created to determine future rockfall mitigation and remediation efforts. Slopes were categorized by greatest potential hazard (A) to least hazardous (E). This novel approach was expanded (Wyllie et al., 1979; Wyllie, 1980, 1987) with the development of an exponential scoring system to better categorize slopes from A to E. The practicality of this proactive approach was adopted by many transportation agencies in the early 1990s when state agencies collaborated to develop the initial RHRS instructions (Pierson, 1991, 1992; Pierson and Van Vickle, 1993; and Brawner, 1994). Since then some transportation agencies have further refined the slope hazard rating criteria (Huang et al., 2009). The RHRS procedure described by Pierson and Van Vickle (1993) will be used in this article, since this version is incorporated into most recent RHRS versions. The full RHRS procedure requires completion of the following six steps: (1) slope inventory, (2) preliminary slope rating, (3) detailed slope rating, (4) project design and cost estimation, (5) project identification and development, and (6) yearly reviews and updates. The data collected in these six steps are then input into a geodatabase that also incorporates the locations of slopes adjacent to relevant transportation corridors (Step 1).
1 Corresponding
author email: eybouali@mtu.edu
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
165
Bouali, Oommen, Vitton, Escobar-Wolf, and Brooks
The preliminary slope rating (Step 2) classifies slopes into three hazard categories—A (high), B (moderate), and C (low)—based on past rockfall activity and the potential for future rockfall activity. The detailed slope rating (Step 3) further classifies high-risk slopes (from the preliminary slope rating) by assigning a score based on 10 rating criteria that include (1) slope height, (2) ditch effectiveness, (3) average vehicle risk, (4) decision sight distance, (5) roadway width, (6) structural condition, (7) rock friction (hard rock) or differential erosion rates (soft rock), (8) block size/volume, (9) climate and water presence, and (10) rockfall history. Scores range from 1 (lowest hazard) to 100 (highest hazard), per rating criteria. The detailed rating for each slope is the summation of all 10 rating criteria scores. Project design and cost estimation (Step 4) is considered prior to project implementation. Pierson and Van Vickle (1993) offer a variety of methods for project identification and development based on the RHRS procedure. The purpose of Step 5 is to formulate the best approach for rockfall mitigation and remediation construction, while Step 6 recommends rated slopes to be reviewed on an annual basis. Therefore, depending on the outcome of Step 6, changes to information recorded in Steps 1 through 5 may be required for relevant slopes. Implementation of the complete RHRS procedure can be time-consuming and expensive as a result of the amount of data acquisition and analysis (preliminary and detailed slope ratings) required, which increases drastically with scale (e.g., state-wide transportation networks), and the need for annual reviews and updates. Transportation agencies have investigated using different types of non-traditional data acquisition techniques to minimize these difficulties. Traditional data acquisitions are in situ measurements and/or field observations, in which the instrument and the user need to be on site. Non-traditional data acquisitions are remote sensing–based, in which active or passive sensors mounted upon moving or stationary instruments receive information in the form of electromagnetic waves (e.g., aerial photography). The use of remote sensing data acquisition for RHRS purposes—such as slope characterization, feature identification, and surficial change and displacement measurements—has evolved over the last 15 years. Data sources include state highway video logs (Maerz et al., 2005; Youssef et al., 2007; and Youssef and Maerz, 2012), optical photography and photogrammetry (Di Crescenzo and Santo, 2007; Lucieer et al., 2013), terrestrial laser scanning (Bauer et al., 2005; Abell´an et al., 2009, 2010), and light detection and ranging (LiDAR) technologies (Strouth and Eberhardt, 2007; Lato et al., 2009, 2012; and Lan et al., 2010). Terrestrial remote sensing techniques allow for detailed observations and accurate surficial
166
measurements of slopes. Platforms in motion, such as satellites, airplanes, terrestrial vehicles, and unmanned aerial vehicles (UAVs), allow for data acquisitions that cover relatively large areas (potentially multiple slopes at once) at various vantage points in a repeatable fashion. A 24-km section of railroad corridor in southeastern Nevada (Figure 1) was chosen to test the efficacy of applying two remote sensing methods—optical photogrammetry and LiDAR, with instruments located on different platforms and viewing angles (Table 1)—to develop an RHRS rating for slopes within the corridor. The railroad corridor follows the valley floor of a canyon system through volcanic rock consisting of rhyolite, tuff, and welded breccia, with approximately 33 percent of railroad tracks within 30 m of a slope with a height greater than 50 m. Approximately 2.5 percent of the 24-km track passes through five tunnels. The railroad corridor was studied by Justice (2015), who utilized a multi-level approach through the application of the RHRS procedure. The multi-level approach, which examined local- and regional-scale hazard assessments, was performed using satellite orthophotography and point-cloud images generated from a terrestrial LiDAR survey on three slopes within the railroad corridor. Bouali et al. (2016) used a satellitebased remote sensing technique, Interferometric Synthetic Aperture Radar (InSAR), to measure surficial displacement rates across the railroad corridor between 1992 and 2010; this study included all three slopes from the Justice (2015) study. One slope exhibited downslope displacement rates greater than 10 mm/yr, and five additional slopes were identified as potentially hazardous, based on slope distance from railroad track, slope height, slope angle, downslope displacement rate (velocity), and total displacement. This study will focus on 14 slopes (named Slope 1, Slope 2, . . ., Slope 14) that meet three geometric criteria: (1) the slope toe is located within 15 m of the railroad track, (2) the slope is at least 15 m in height, and (3) a dip greater than 25◦ toward the railroad tracks is measured on the slope face. The purpose of this study is to investigate the effectiveness of using remote sensing techniques to assess RHRS values compared to the traditional field-based method. Although some state transportation agencies continuously monitor and rate their highest priority slopes (e.g., through video logs), a synergistic field- and remote sensing–based approach will allow for measurements that are undetectable or not within view from transportation corridor heights, such as identification of rockfall source areas and unstable blocks, smallscale rock displacement detection (millimeter-scale), and identification of potential future hazardous slopes.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
Remote Sensing for RHRS
Figure 1. Twenty-four–kilometer segment of railroad corridor located in southeastern Nevada (inset). The 14 slopes (green circles) are identified by number (Slope 1, Slope 2, etc.). Background imagery was generated in ArcGIS software by Esri.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
167
Bouali, Oommen, Vitton, Escobar-Wolf, and Brooks Table 1. Types of remote sensing platforms, sensors, and output data used in this study. Platform Satellite UAV
Terrestrial
Sensors Landsat 7, WorldView 1 & 2 Nikon D800 (50 mm) mounted on Bergen Hexacopter RIEGL LMS-Z210ii
Sensor Type
Output Data
Optical Optical
Polar Orbit Remote Controlled by Operator
Photographic Images Photographic Images, 3-D Point Clouds
LiDAR
Stationary
3-D Point Clouds
Specifically, additional advancements in the use of remote sensing techniques for preliminary slope ratings, detailed slope ratings, and annual rating reviews and updates would supplement repetitive site visits yet complement the traditional RHRS field-based approach. A long-term monitoring approach that updates slope ratings in near real-time would be beneficial and could be accomplished by incorporating remote sensing techniques into RHRS procedures. METHODOLOGY The application of remote sensing techniques in acquiring information for Step 2 (preliminary slope rating), Step 3 (detailed slope rating), and Step 6 (yearly reviews and updates) can aid in the efficacy of the RHRS procedure discussed in Pierson and Van Vickle (1993). Preliminary slope ratings can be assigned to slopes based on relatively low-resolution, vertical photography acquired from optical sensors on satellites, airplanes, and UAVs. The process of assigning detailed slope ratings to “A”-level slopes, traditionally conducted by an on-site field crew, can be improved by data obtained from terrestrial LiDAR (at ground level) and oblique-angle, high-resolution photographs from UAVs. The yearly RHRS review and update can further use remote sensing data sets to reduce review time and cost. A description of an approach using traditional field methods and remote sensing techniques to provide a synergistic RHRS approach is discussed below. Preliminary Slope Rating (RHRS Step 2) The preliminary slope rating classifies a slope based on the “estimated potential for rockfall on roadway” and the “historical rockfall activity” using a threetiered class system: “A” for high, “B” for moderate, and “C” for low (p. 18, Pierson and Van Vickle, 1993). A total of eight factors are considered when performing the preliminary rating. Of these eight factors, four potential rockfall factors can be estimated using remote sensing. These factors are (1) size of rockfall material, (2) quantity of material, (3) amount available, and (4) ditch effectiveness. The four historical rockfall activ168
Motion
ity factors can also be estimated using remote sensing data if these data were acquired after the rockfall event and prior to remediation. This study will examine the four potential rockfall factors; a similar approach can be applied to the historical rockfall activity factors if relevant data were acquired. Optical photography, acquired by satellite, airplane, or UAV fly-overs, can directly aid in the quantification of factors that indicate the potential for future rockfall events that impact the adjacent transportation corridor. All four factors are area measurements (rock size, material quantity, material amount, and ditch coverage) and, therefore, can be estimated using optical photographs in a geographic information system (GIS) database. Two acquisition variables that dictate the effectiveness of optical photograph usage are image resolution and view angle. Coarse-resolution (meter-scale) images are widely available and are sufficient for identifying large unstable blocks, but there are likely to be sub–meter-scale rocks that can cause damage and will be undetectable at such coarse resolution. It is therefore important to obtain optical photographs at a resolution greater than (lower in magnitude) the smallest-sized rock deemed dangerous to traffic within the transportation corridor, most likely images with centimeter-scale resolution. The view angle from which the images are acquired is also important, as different vantage points give more information on the slope. For situations in which the top of the slope is not viewable at the transportation corridor (ground) level, a vertical view angle allows for better estimations of the material quantity and amount of material available, as well as for identification of potential large unstable blocks that were previously unobservable.
Detailed Slope Rating (RHRS Step 3) Slopes with a preliminary “A” rating are of highest priority, requiring that detailed slope ratings be conducted on these slopes first (Pierson and Van Vickle, 1993). The detailed slope rating assigns a numerical value between 1 (lowest hazard) and 100 (highest hazard) for the 10 rating criteria previously discussed. Five of the rating criteria are quantitative, with the function
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
Remote Sensing for RHRS
y = 3x being the basis of the scoring system, while discrete formulae for the exponent (x) are provided for slope height, average vehicle risk, sight distance, roadway width, and block size/volume (p. 29, Pierson and Van Vickle, 1993). The other five rating criteria (ditch effectiveness, structural condition, rock friction or differential erosion, climate and presence of water, and rockfall history) are more qualitative; scores are assigned based on observations rather than numerical measurements. The sum of scores for the 10 rating criteria equals the total hazard score for a slope, which ranges from 10 (no hazard) to 1,000 (immediate hazard). Since total hazard scores are subjective and relative, it is important to maintain consistency when conducting detailed slope ratings along transportation corridors. In addition, transportation agencies use the total hazard score as an input variable when assigning priority for mitigation and remediation strategies (Step 5). To assess the subjectivity of the detailed slope rating RHRS step, field notes for the 10 rating criteria for each of the 14 slopes were given to six participants (geologists and geotechnical engineering faculty and graduate students at Michigan Technological University). The participants, who were not given any additional information of the railroad corridor, were asked to convert the field notes (text and numeric data) into detailed slope ratings for all 14 slopes. The purpose of this exercise was to quantify, albeit simplistically, the variance of field-based detailed rating scores and, under the assumption that the field-based RHRS procedure was the baseline, to determine the accuracy of different remote sensing techniques. Remote sensing can only be beneficial if the acquired images capture enough detail to identify slope features and characteristics. Image resolution must be on par with what can be observed in the field. Additionally, remote sensing techniques can supplement fieldbased observations by acquiring images from various view angles in hard-to-reach locations on a slope (e.g., top of slope, tall vertical cliff faces, areas considered dangerous because of rockfall or unstable materials, etc.). High-resolution (centimeter-scale) UAV images, acquired at oblique view angles, allowed for additional observations at higher elevations along the slope. The combination of multiple remote sensing techniques, especially with data acquired from different types of sensors and locations, provides the ability to quantify rating criteria from the detailed slope rating procedure. Yearly Reviews and Updates (RHRS Step 6) Transportation agencies are required to maintain optimum safety throughout the transportation network. In terms of hazards posed by unstable slopes, one way
of doing this is to perform an annual review of every “A”-rated slope (RHRS Step 3). This provides a long-term monitoring procedure that is focused on the most hazardous slopes. Furthermore, any preliminarily rated slope (regardless of tiered rating: A, B, or C) that has undergone any changes (construction, maintenance, displacement, etc.) should be re-reviewed and the slope rating updated. A major benefit of utilizing remote sensing techniques in assessing the RHRS parameters is the capability of monitoring slope changes over long periods of time. Multiple acquisitions of LiDAR and optical photogrammetry can qualitatively monitor changes as well as quantitatively measure deformation rates on and around each slope. Annual-scale qualitative observations may include noting (1) the change of talus sizes at the bottom of the slope, (2) the weathering condition and erosion locations on the slope face, and (3) the presence of water in planes of weakness (faults, joints, bedding planes), among others. Quantitatively, LiDAR and optical photogrammetric techniques such as Structure from Motion (SfM) and three-dimensional pointcloud change detection allow for the calculation of surficial deformation between image pairs by calculating the change in the location of the slope surface, at two different acquisition dates, and in three-dimensional space. RESULTS AND DISCUSSION Examples of remote sensing data analyses applied to the RHRS procedure are organized by RHRS step. The effectiveness of optical satellite and UAV photography are examined in “RHRS Step 2: Preliminary Slope Rating.” Those two techniques are compared to two field-based approaches—an initial field examination and a participant survey—in “RHRS Step 3: Detailed Slope Rating.” Long-term remote sensing–based slope monitoring approach are discussed in “RHRS Step 6: Yearly Review and Update,” with a LiDAR change detection example given of Slope 1. Preliminary Slope Rating (RHRS Step 2) Preliminary slope ratings were assigned to all 14 slopes via both traditional field-based observations and using optical satellite images (Table 2). The field approach identified three “A”-rated slopes, nine “B”rated slopes, and two “C”-rated slopes. The satellite imagery approach identified one “A”-rated slope, seven “B”-rated slopes, and six “C”-rated slopes. An underestimation of preliminary slope ratings using optical satellite imagery is likely due to the use of coarseresolution imagery (meter-scale). Sub–meter-scale details are unobservable and blurry at coarse resolutions,
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
169
Bouali, Oommen, Vitton, Escobar-Wolf, and Brooks Table 2. Preliminary slope ratings for each slope using field-based observations and optical satellite imagery. Slope #
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Field Satellite
A A
B B
B B
C C
C C
B B
A B
B C
B B
B C
B C
A B
B B
B C
resulting in hazardous features (e.g., moderate-sized loose rocks, adverse joint orientations, and other evidence of displacement) that go undetected. This underestimation issue is addressed more fully in “RHRS Step 3: Detailed Slope Rating.” Although small-scale details may be unobservable with coarse-resolution imagery, other steps within the RHRS procedure can be made easier when this imagery is supplemented with optical satellite imagery. Coarse satellite imagery can be obtained free of charge and in near real-time, which can aid in large-scale rockfall detection and monitoring. Vertical optical imagery can greatly assist in the preliminary slope rating RHRS step. Figure 2 shows Slope 1—the only slope in which historic rockfalls have blocked and disrupted rail traffic—from three vantage points: (A) at ground level near the railroad tracks, (B) through near-vertical optical image acquired via satellite, and (C) through vertical optical image acquired via UAV. Many slope features were identified with satellite and UAV imagery that were undetected from ground
level. Features identified using satellite imagery include at least three large blocks at the top of the slope and the presence of the main scarp (Figure 2B [red dashed oval] and C), although accurate measurements as to the size of the main scarp were difficult to obtain. The aperture of the main scarp was estimated using UAV images (and confirmed by field work) to be about 8 m wide. Additional features detected with UAV imagery include the presence of a secondary scarp, located downslope from the main scarp, as well as large blocks and piles of loose rock (also downslope from the main scarp). Detailed Slope Rating (RHRS Step 3) A detailed slope rating is normally performed on slopes that received an “A” rating in the preliminary slope rating step. This is usually due to the sheer quantity of slopes that may be classified as “A” level, especially when dealing with state-wide transportation networks (Pierson and Van Vickle, 1993). However, since the areal extent of this project is relatively small, all
Figure 2. Slope 1. (A) Photo taken from ground level near the railroad tracks. (B) Near-vertical optical satellite image. Peak 1 from (A) is shown in white circle. Slope features that aid in the preliminary slope rating include at least three large blocks and the main scarp (red dashed oval). (C) Vertical UAV image that details the main scarp (up to 8 m wide) and a previously undetected secondary scarp, a large block also identified in (B), and piles of loose rock.
170
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
Remote Sensing for RHRS Table 3. Preliminary slope ratings based on field observations and satellite imagery (also shown in Table 2). Detailed slope ratings based on initial field observations. Rating criteria listed by number: 1 = slope height, 2 = ditch effectiveness, 3 = average vehicle risk, 4 = decision sight distance, 5 = roadway width, 6 = structural condition, 7 = rock friction, 8 = block size/volume, 9 = climate and presence of water, and 10 = rockfall history. Rating criteria in boldface were calculated using exponent formulae from Pierson and Van Vickle (1993). Slope No. Preliminary slope ratings (field/Satellite) Detailed slope rating: rating criteria (1–10)
#1 #2 #3 #4 #5 #6 #7 #8 #9 #10
Field-based detailed slope rating scores
1
2
3
4
5
6
7
8
9
10
11
12
13
14
A/A
B/B
B/B
C/C
C/C
B/B
A/B
B/C
B/B
B/C
B/C
A/B
B/B
B/C
100 27 1 0 100 81 81 47 7 62 506
100 3 1 1 100 72 62 2 7 9 334
100 9 1 0 100 81 3 9 7 3 332
16 81 1 0 100 58 8 9 7 3 269
10 81 1 13 100 36 3 5 7 3 250
76 81 1 24 100 69 21 9 7 5 327
100 70 1 8 100 81 22 100 7 3 467
71 70 1 8 100 81 3 0 7 3 329
15 81 1 1 100 81 27 9 7 3 325
100 70 1 1 100 63 27 47 7 3 350
62 27 1 0 100 81 27 27 7 27 359
81 27 1 27 100 81 27 16 7 66 427
100 81 1 1 100 70 21 9 7 3 328
13 81 1 0 100 63 81 2 7 14 308
slopes were given a detailed slope rating regardless of their preliminary slope rating. This allows for a direct comparison of how preliminary slope ratings translate into detailed slope rating hazard scores and how remote sensing data sets can assist in slope prioritization and future decision-making. Table 3 shows preliminary and detailed ratings, as well as values for rating criteria, for each slope. Slope 1 received a preliminary slope rating (field/satellite) of A/A and a detailed slope rating of 506, making it the most hazardous slope in the railroad corridor. Slopes that received an A/B preliminary slope rating corresponded to a 427–467 detailed slope rating, B/B corresponded to 325–334, B/C corresponded to 308–350, and C/C corresponded to 250–269. There is a clear distinction between slopes with an A/A or A/B rating (≥427), slopes with a B/B or B/C rating (308– 350), and slopes with a C/C rating (≤269). Table 4 displays information for all field-based detailed rating scores. The “initial score” is the detailed rating score obtained by the original field crew. The minimum and maximum “participant scores” illustrate the range in detailed rating scores assigned based on the same set of textual and numerical field notes. The initial score and participant scores were generally in agreement: the initial score fell into the participant
score range for nine slopes (Slopes 2, 4, 5, 7, 8, 9, 11, 12, and 14); the initial score was greater than the participant score range for Slopes 1 and 3; the initial score was less than the participant score range for Slopes 6, 10, and 13. The participant scores ranges were within 72 for every slope except Slope 14. Discrepancies in detailed rating scores were usually due to the interpretation of subjective textual rating criteria, especially when complex slope characteristics did not neatly compartmentalize into a pre-assigned rating criteria score. For example, the structural condition of a rock slope is assigned a rating criteria score of 3 if there are “discontinuous joints, favorable orientation,” a score of 9 for “discontinuous joints, random orientation,” a score of 27 for “discontinuous joints, adverse orientation,” and a score of 81 for “continuous joints, adverse orientation” (p. 26, Pierson and Van Vickle, 1993). Subjectivity occurred when slopes were described as having “continuous joints, random orientation,” which is not a predetermined category, and, therefore, the value of structural condition rating criteria is left for the participant to decide. This type of subjective score assignment was required for multiple slopes because selecting a quantitative value for these rating criteria was sometimes difficult as a result of complex geology.
Table 4. Detailed slope ratings. “Initial score” is identical to initial field-based observations found in Table 3. Statistics of “participant scores” (mean, standard deviation, minimum, and maximum) from the survey illustrate the subjectivity of the RHRS procedure. Slope No. Initial score Participant survey
Mean SD Minimum Maximum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
506 452 15 437 471
334 357 24 331 385
332 313 0 313 313
269 283 22 253 303
250 259 22 250 304
327 402 23 361 426
467 492 20 463 508
329 344 15 321 355
325 325 0 325 325
350 419 32 376 448
359 359 0 359 359
427 433 16 414 448
328 393 20 361 409
308 362 51 294 447
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
171
Bouali, Oommen, Vitton, Escobar-Wolf, and Brooks Table 5. Detailed slope ratings based on field observations (initial scores and participant scores, also shown in Table 4), optical satellite imagery, and optical UAV imagery. Slope No. Initial field-based scores Mean bield-based participant scores Optical satellite Optical UAV
1
2
3
4
5
6
7
8
9
10
11
12
13
14
506 452 487 528
334 357 331 396
332 313 304 n/a
269 283 244 n/a
250 259 208 n/a
327 402 369 n/a
467 492 352 n/a
329 344 293 n/a
325 325 350 n/a
350 419 265 n/a
359 359 294 n/a
427 433 357 n/a
328 393 318 n/a
308 362 282 n/a
n/a - not applicable; data not obtained by method for specified slopes.
Detailed slope ratings were also assigned using optical photography obtained from two platforms: satellite and UAV. The remote sensing–based RHRS detailed slope rating approach was conducted by downloading the data (in the case of satellite images) or acquiring data (UAV images) and then identifying rating criteria from the images obtained of each slope within the study site. Measurements were made after importing the images into a GIS for analysis. Optical satellite photography data were obtained from a combination of Landsat 7, WorldView 1, and WorldView 2 acquisitions; detailed rating scores are
listed in Table 5. Satellite-based detailed rating scores were generally lower than field-based detailed rating scores, as 11 of 14 slopes followed this trend. This score underestimation was likely due to the relatively coarse resolution (meter-scale) of the WorldView satellite imagery. Sub–meter-sized features such as joints, faults, bedding planes, zones of erosion and weathering, and the presence of water were unobservable with coarse-resolution images. Adding to the difficulty was the fact that satellite images were acquired at near vertical, and shadows were present across the slope face (Figure 3A), resulting in a loss of information.
Figure 3. Images of Slope 2 taken from three view angles: (A) optical satellite image, (B) vertical optical UAV image, and (C) oblique optical UAV image. A yellow triangle is placed in each image for geographic reference.
172
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
Remote Sensing for RHRS
Although the use of optical satellite imagery was relatively beneficial for the preliminary slope rating, the coarse resolution and view angle make relying on satellite imagery for the detailed slope rating difficult. Consistent underestimation of detailed rating scores, when compared to the field-based approach, may result in the miscalculation of potential hazard levels. It would be unacceptable if a catastrophic event (e.g., large rockfall) that went unpredicted and unmitigated occurred as a result of coarse resolution imagery. Therefore, the exclusive use of coarse-resolution satellite imagery for the detailed slope rating RHRS step is insufficient and dangerous, making it necessary to use higher resolution imagery. Optical UAV photographs were acquired using a Nikon D800 camera, with a 50-mm prime lens (collecting 36-megapixel imagery at 2 frames/s at a speed of approximately 2 m/s) onboard a Bergen Hexacopter for Slopes 1 and 2. Detailed rating scores are shown in Table 5. High-resolution imagery (centimeter-scale), obtained from multiple UAV fly-overs at different view angles (vertical and oblique), results in an overestimation of detailed rating scores when compared to both field-based methods and optical satellite photography usage. For example, vertical (Figure 3B) and oblique (Figure 3C) view angles of Slope 2 present a more complete picture of current slope conditions. A vertical view angle allows for the identification of source material for potential instabilities on the top of the slope, vegetation distribution (a vegetated region has not experienced a rockfall recently), and structural conditions, such as weathering distribution and strike of joints/faults. An oblique view angle reveals more of the upper slope face, which is difficult to see from ground level. An overestimation of detailed rating scores was the result of more robust data. Table 6 shows the increase in UAV-based detailed rating scores compared to satellite-based de-
Table 6. Comparison of satellite- and UAV-derived values for rating criteria and detailed slope rating score. Slope 1
Slope 2
Rating Criteria
Satellite
UAV
Satellite
UAV
#1 #2 #3 #4 #5 #6 #7 #8 #9 #10 Detailed Slope Rating Score
100 9 1 0 100 81 27 81 7 81 487
100 27 1 0 100 81 50 81 7 81 528
100 3 1 0 100 27 3 9 7 81 331
100 3 1 0 100 50 27 27 7 81 396
tailed ratings scores. Regarding Slope 2 (Figure 3), an increase in three rating criteria changed the detailed rating score from 331 (satellite) to 396 (UAV). An increase in the three rating criteria values—structural condition, rock friction, and block size/volume—was a direct result of better estimations due to higher resolution imagery. The use of sub-meter UAV imagery makes unstable rocks, ranging in size from tens of inches to 3 ft (about 1 m), visible; rocks of these sizes appear blurry in coarse-resolution imagery. Details of structural condition and rock friction are also made clearer using high-resolution UAV imagery, especially at an oblique view angle that reveals the slope face from the shadow zone sometimes observed from the vertical view angle (Figure 3). Yearly Reviews and Updates (RHRS Step 6) Repeat remote sensing acquisitions allow for the ability to calculate and monitor changes in slope geometry and displacement rates. A previous assessment of overall slope displacements across the railroad corridor was performed using Persistent Scatterer Interferometry and Distributed Scatterer Interferometry (PSI and DSI, respectively), an InSAR stacking technique using radar images acquired via satellites, developed by Bouali et al. (2016). In the case of a rural setting such as this railroad corridor, InSAR is capable of measuring displacement rates for a larger area, such as the general trend of slope movements over long periods of time (which may aid in potential landslide detection), compared to smaller areas, because output data can be spatially limited and detailed measurements may be lacking. Therefore, remote sensing techniques that provide high spatial data densities, such as LiDAR and optical photogrammetry, are preferred for monitoring complex and detailed changes within a slope. Since Slope 1 was considered the most hazardous slope based on preliminary and detailed slope ratings (Tables 2 through 6), LiDAR point-cloud data were acquired during each summer from 2011 to 2014 using a RIEGL LMS-Z210ii instrument, acquiring data at an angular resolution of 0.005◦ and an accuracy of around 3 cm. A detailed quantification of slope deformation was performed using a technique called “change detection,” wherein three-dimensional point clouds from two acquisitions are geometrically compared; a change, or difference, in point-cloud location indicates the occurrence of measurable displacement. Figure 4 shows the surficial changes, occurring between 2011 and 2014, mapped on a three-dimensional digital elevation model (3D DEM) of Slope 1. Blue regions (negative change) show material loss and red regions (positive change) show material accumulation. Six individual rock falls
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
173
Bouali, Oommen, Vitton, Escobar-Wolf, and Brooks
Figure 4. Change detection map of Slope 1 between 2011 and 2014. Negative change (blue) indicates material loss and positive change (red) indicates material accumulation. Six rockfall events, numbered by location, are detectable and occurred in the 3-year span. Slope 1 3D DEM was generated from the 2014 three-dimensional point cloud.
occurred on the Slope 1 face between 2011 and 2014 (numbered values in Figure 4), with substantial accumulation of material occurring at the central toe region. Detailed change detection measurements, along with similar techniques termed SfM, which utilize optical photogrammetry to create three-dimensional point clouds (Westoby et al., 2012), and InSAR can assist in the yearly review process. By capitalizing on advancements in remote sensing technologies—such as increased spatial and temporal resolution, wider variety of view angles, more accurate sensors, and a growing variety of clever data processing techniques— transportation agencies can monitor potentially hazardous slopes with a more robust, efficient, and timeeffective RHRS approach.
Additional Discussion The three RHRS steps discussed in detail were Preliminary Slope Rating (Step 2), Detailed Slope Rating (Step 3), and Yearly Review and Update (Step 6). Remote sensing techniques can also play a secondary role in information collection and analyses in the other three RHRS steps. Step 1, Slope Inventory, can be aided by analyzing DEMs generated from satellite/aerial optical or radar images (e.g., NASA’s 174
Shuttle Radar Topography Mission, German Aerospace Center Terra SAR-X and TanDEM-X, etc.). DEMs can assist in cataloging and digitizing the spatial extent of slopes adjacent to the transportation corridor. Step 4, Project Design and Cost Estimate, may benefit from detailed models derived from three-dimensional point clouds obtained from optical photogrammetry or LiDAR techniques. These models can help determine remediation designs and techniques that need to be constructed. Step 5, Project Identification and Development, is a management step that utilizes information gained from Steps 1 through 4 to determine necessary remediation projects for choice slopes. Pierson and Van Vickle (1993) provide four project identification methods: (1) score—priority given to slopes with highest detailed slope rating; (2) ratio—priority given to slopes with greatest score-to-cost ratio; (3) remedial—slopes with similar designs can be placed in a single project, which will alter slope prioritization; and (4) proximity—slopes closest to rockfall sites are given highest priority. Each of these project identification methods can use remote sensing data, directly or indirectly. Thus, remote sensing techniques are practical supplementary tools to the traditional field-based approach for RHRS. Using remote sensing has its advantages and limitations. Advantages include the following:
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
Remote Sensing for RHRS
1. It offers the ability to acquire data over hard-toreach or inaccessible terrain. 2. It offers the ability to obtain spatial coverage over geographic scales spanning several orders of magnitude, ranging from local scale (one geotechnical asset like an unstable slope) to regional scale (segment of a transportation network like the railroad corridor). 3. It offers the ability to obtain multiple acquisitions over the same area to allow for the measurement of surficial changes over long periods of time. 4. Field-based techniques can be supplemented or eliminated and replaced with remote sensing data. 5. It involves offering of a wide array of sensors acquiring data over different swaths of the electromagnetic spectrum at different spatial resolutions from various view angles. 6. It offers the ability to create a large library of historical, archival data sets that can be used with future data products. 7. Sensors can acquire one image that can be processed in a variety of ways. 8. It provides a non-invasive approach that does not disturb targets within a study area. The limitations of remote sensing include the following: 1. Shadows zones caused by mountainous topography, complex surficial geometry, or clouds (for optical images) result in a loss of data that cannot be recovered. 2. Remote sensing may not be cost-effective if used improperly. For example, acquiring high-resolution data (e.g., centimeter-scale LiDAR point clouds) for preliminary slope ratings over a state-wide transportation network would not be economical. 3. Data acquired from remote sensing techniques contain factors that affect the receiving sensors, such as noise introduced by interfering phenomena (e.g., water vapor affects two-way travel time of radar waves). Furthermore, non-uniqueness, the idea that an outcome can be achieved through many different combinations of input factors, is a serious issue that must be considered when modeling remote sensing data. 4. Some acquisition and data processing techniques require training and may include a steep learning curve. The traditional field-based RHRS approach is a more robust method, compared to remote sensing techniques. The field-based RHRS method allows field crews to interact and observe slopes in person; if only a few slopes are of interest, then the field-based approach is much simpler. If, however, a transportation agency
is tasked with monitoring a transportation corridor or an entire transportation network (e.g., local, regional, state-wide), then a remote sensing–based RHRS approach, using high-resolution optical imagery, may be preferable. For the remote sensing–based RHRS estimation, high-resolution imagery acquired from different view angles is critical. The spatial resolution and view angles obtained from satellite imagery are coarser, more limited, and often result in significant underestimation of preliminary and detailed slope rating scores. However, an approach integrating rapidly deployable UAV platforms with high-resolution optical sensors and high-resolution terrestrial LiDAR provides a readily available tool set for collecting imagery that can be used for RHRS interpretation. UAV- and LiDAR-based data collections provide much higher spatial resolution and can easily obtain multiple view angles compared to satellite-based data. Detailed slope rating scores from UAV data show promise in terms of their use as an alternate approach for field-based RHRS measurements when monitoring transportation corridors. In addition, imagery collected using UAV will provide a more methodical documentation of the site condition for the transportation agency, compared to field-based data collection. Coupled with other remote sensing techniques (e.g., InSAR and optical photogrammetry), transportation agencies would benefit from a supplementary and complimentary remote sensing RHRS approach. CONCLUSIONS RHRS is a procedure developed by Brawner and Wyllie (1975) and further expanded by geological engineers and transportation agencies (Wyllie et al., 1979; Wyllie, 1980, 1987; Pierson, 1991, 1992; Pierson and Van Vickle, 1993; and Brawner, 1994), to analyze slopes adjacent to transportation corridors and to prioritize those most likely to experience damaging rockfalls. The traditional RHRS approach is to use personnel to acquire field-based measurements, especially for Steps 1 through 3. Video logs are also commonly used for data analysis in Step 3 and for slope monitoring in Step 6. The purpose of this study, however, is to show the benefits of remote sensing data acquisition (optical satellite imagery, optical UAV imagery, and terrestrial LiDAR) and analyses for a more robust, efficient, and time-effective RHRS approach. Other remote sensing techniques, such as optical photogrammetry and InSAR, are also discussed and referenced. These remote sensing methods have a place as a supplemental data acquisition approach alongside the traditional fieldbased approach. Observations from remote sensing imagery were compared to field-based observations (used as the
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
175
Bouali, Oommen, Vitton, Escobar-Wolf, and Brooks
baseline) for the preliminary and detailed slope rating steps. Fourteen slopes along a railroad corridor in southeastern Nevada were studied. In general, observations using optical satellite imagery provided an underestimation of preliminary and detailed slope ratings. This is most likely due to the coarse-resolution imagery used (meter-scale). It was therefore concluded that high-resolution imagery is a requirement because consistent rating score underestimations may potentially lead to undetected future rockfall events, which is unacceptable. UAV imagery were obtained for Slopes 1 and 2 only because of field time limitations. When using higher resolution imagery acquired from the UAV (centimeter-scale), an overestimation of preliminary and detailed slope rating scores occurred (when compared to satellite- and field-based approaches). This result is likely due to the combination of high resolution (small slopes features and characteristics that can be observed) and various view angles (more information about the top of the slope is available) when compared to field-based approaches limited to ground level. Terrestrial LiDAR change detection successfully monitored rockfall events on Slope 1. Two threedimensional point clouds, which map the location of the slope face surface, were geometrically differenced to calculate the amount of surface change (slope deformation) that occurred between the summers of 2011 and 2014. Evidence of surficial displacement includes six locations of rock mass loss and an overall accumulation of material at the slope toe. Displacements of up to 2 to 3 m in both directions (toward and away from the LiDAR sensor) were measured over the 3-year span. Change detection, and other similar techniques (e.g., SfM, InSAR stacking), enable measurements of dynamic events that occur rapidly (e.g., rockfalls) or very slowly (e.g., landslide creep) over any length of time (e.g., daily, monthly, annually, etc.). Every slope within or adjacent to a transportation corridor has the potential to pose hazards that may affect the performance and quality of transportation assets and the safety of its users. The traditional RHRS procedure attempts to identify and prioritize the most hazardous slopes through a robust fieldbased rating system. The use of remote sensing techniques has proved beneficial by providing more information, expanding the observable study area, archiving historical data sets, and allowing for detailed analysis otherwise unavailable to field crews. By combining remote sensing techniques with traditional fieldbased approaches, transportation agencies can build a more robust, efficient, and time-effective RHRS procedure that can assist in the achievement of slope lifecycle performance goals along an entire transportation network.
176
ACKNOWLEDGMENTS This project was funded by the U.S. Department of Transportation (USDOT) through the Office of the Assistant Secretary for Research and Technology (Cooperative Agreement No. RITARS-14-H-MTU). The authors would like to express their utmost gratitude to the railroad company that allowed us access to the study site and to the on-site employees who kept us safe. The authors would also like to thank Michigan Tech Research Institute’s Richard Dobson, Ben Hart, and David Dean and Michigan Technological University’s Zachary Champion for their contributions to field data acquisition and the six Michigan Technological University participants for providing input with the RHRS survey. Additional thanks are extended to the organizations that provided the background images used in this research study, specifically the U.S. Geological Survey, the National Aeronautics and Space Administration, Google, and Digital Globe. DISCLAIMER The views, opinions, findings, and conclusions reflected in this article are the responsibility of the authors only and do not represent the official policy or position of the USDOT/OST-R or any state or other entity.
REFERENCES ´ , A.; CALVET, J.; VILAPLANA, J. M.; AND BLANCHARD, J., ABELLAN 2010, Detection and spatial prediction of rockfalls by means of terrestrial laser scanner monitoring: Geomorphology, Vol. 119, No. 3–4, pp. 162–171. ´ , A.; JABOYEDOFF, M.; OPPIKOFER, T.; AND VILAPLANA, J. ABELLAN M., 2009, Detection of millimetric deformation using a terrestrial laser scanner: Experiment and application to a rockfall event: Natural Hazards Earth System Sciences, Vol. 9, pp. 365– 372. ¨ , A., 2005, Mass movement BAUER, A.; PAAR, G.; AND KALTENBOCK monitoring using terrestrial laser scanner for rock fall management. In: Geo-Information for Disaster Management: Springer, Berlin, Germany, pp. 393–406. BOUALI, E. H.; OOMMEN, T.; AND ESCOBAR-WOLF, R., 2016, Interferometric stacking toward geohazard identification and geotechnical asset monitoring: Journal Infrastructure Systems, Vol. 22, No. 2, p. 05016001. BRAWNER, C. O., 1994, Rockfall Hazard Mitigation Methods— Participant’s Workbook, NHI Course No. 13219: U.S. Department of Transportation, Federal Highway Administration, Publication FHWA SA-93-085. BRAWNER, C. O. AND WYLLIE, D. C., 1975, Rock slope stability on railway projects: Proceedings of the American Railway Engineering Association Regional Meeting: Vancouver, British Columbia, Canada, 8 p. DI CRESCENZO, G. AND SANTO, A., 2007, High-resolution mapping of rock fall instability through the integration of photogrammetric, geomorphological and engineering-geological surveys: Quaternary International, Vol. 171–172, pp. 118–130. HUANG, S. L.; DARROW, M. M.; AND CALVIN, P., 2009, Unstable Slope Management Program: Background Research and Pro-
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
Remote Sensing for RHRS gram Inception—Phase I Final Report, Alaska Department of Transportation and Public Facilities, 90 p. JUSTICE, S. M., 2015, Application of a Hazard Rating System for Rock Slopes along a Transportation Corridor Using Remote Sensing: Master’s Thesis, Michigan Technological University, 86 p. LAN, H.; MARTIN, C. D.; ZHOU, C.; AND LIM, C. H., 2010, Rockfall hazard analysis using LiDAR and spatial modeling: Geomorphology, Vol. 118, No. 1–2, pp. 213–223. LATO, M. J.; DIEDERICHS, M. S.; HUTCHINSON, D. J.; AND HARRAP, R., 2012, Evaluating roadside rockmasses for rockfall hazards using LiDAR data: Optimizing data collection and processing protocols: Natural Hazards, Vol. 60, No. 3, pp. 831–864. LATO, M. J.; HUTCHINSON, D. J.; DIEDERICHS, M. S.; AND HARRAP, R., 2009, Engineering monitoring of rockfall hazards along transportation corridors: Using mobile terrestrial LiDAR: Natural Hazards Earth System Sciences, Vol. 9, No. 3, pp. 935–946. LUCIEER, A.; DE JONG, S. M.; AND TURNER, D., 2013, Mapping landslide displacements using Structure from Motion (SfM) and image correlation of multi-temporal UAV photography: Progress Physical Geology, Vol. 38, No. 1, pp. 1–20. MAERZ, N. H.; YOUSSEF, A.; AND FENNESSEY, T. W., 2005, New risk-consequence Rockfall Hazard Rating System for Missouri Highways using digital image analysis: Environmental Engineering Geoscience, Vol. 11, No. 3, pp. 229–249. PIERSON, L. A., 1991, Rockfall Hazard Rating System: Oregon State Highway Division, Federal Highway Administration, Report FHWA-OR-GT-92-05, 15 p. PIERSON, L. A., 1992, Rockfall Hazard Rating System: Rockfall Prediction and Control and Landslide Case Histories: Research Record 1343, Transportation Research Board, Washington, D.C., PP. 6–13.
PIERSON, L. A. AND VAN VICKLE, R., 1993, Rockfall Hazard Rating System—Participant’s Manual, NHI Course No. 130220: U.S. Department of Transportation, Federal Highway Administration, Publication FHWA SA-93-057, 112 p. STROUTH, A. AND EBERHARDT, E., 2007, The use of LiDAR to overcome rock slope hazard data collection challenges at Afternoon Creek, Washington. In: Tonon F. and Kottenstette J. T. (Editors), Laser and Photogrammetric Methods for Rock Face Characterization: American Rock Mechanics Association in Conjunction with GoldenRocks 2006, Colorado School of Mines, Golden, CO, pp. 109–120. WESTOBY, M. J.; BRASINGTON, J.; GLASSER, N. F.; HAMBREY, M. J.; AND REYNOLDS, J. M., 2012, ‘Structure-from-Motion’ photogrammetry: A low-cost, effective tool for geoscience applications: Geomorphology, Vol. 179, pp. 300–314. WYLLIE, D. C., 1980, Toppling rock slope failures: Examples of analysis and stabilization: Rock Mechanics, Vol. 13, pp. 89–98. WYLLIE, D. C., 1987, Rock slope inventory system. Proceedings of the Federal Highway Administration Rock Fall Mitigation Seminar: FHWA, Region 10, Portland, OR, 25 p. WYLLIE, D. C.; MCCAMMON, N. R.; AND BRUMUND, W. F., 1979, Use of Risk Analysis in Planning Slope Stabilization Programs on Transportation Routes: Research Record 749, Transportation Research Board, Washington, D.C. YOUSSEF, A.; MAERZ, N. H.; AND XIANG, Q., 2007, RockSee: Video image measurements of physical features to aid in highway rock cut characterization: Computers Geosciences, Vol. 33, pp. 437–444. YOUSSEF, A. AND MAERZ, N. H., 2012, Development, justification, and verification of a rock fall hazard rating system: Bulletin Engineering Geology Environment, Vol. 71, pp. 171–186.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 165–177
177
A New Application of CurvaTool Semi-Automatic Approach to Qualitatively Detect Geological Lineaments SABRINA BONETTO1 Department of Earth Sciences, University of Turin, Valperga Caluso 35, 10125 Turin, Italy; sabrina.bonetto@unito.it
ANNA FACELLO CNR-IRPI, Strada delle Cacce, 73 - 10135 Turin, Italy; anna.facello@irpi.cnr.it
GESSICA UMILI Department of Earth Sciences, University of Turin, Valperga Caluso 35, 10125 Turin, Italy; gessica.umili@unito.it
Key Terms: Geological Structure, Lineament Extraction, DTM, Semi-Automatic Survey, Argentera Massif ABSTRACT In the past years, lineament analysis has become an important analytical technique for delineation of major structural units in mineral prospecting, hydrogeology, and tectonic studies. The use of remote sensing, with progressive development of image enhancement techniques, provides an opportunity to produce more reliable and comprehensive lineament maps. In this paper, we propose the application of a semi-automatic approach based on digital terrain models (DTMs) for the extraction of potential lineaments and their detailed validation. We selected an area belonging to the Bagni di Vinadio municipality (Cuneo, NW Italy), which is part of the Argentera Massif (western Alps), as a test site. Data obtained from the code CurvaTool, developed by the authors, are successfully compared with literature information and with lineaments obtained from visual interpretation of remote sensing imagery. The CurvaTool code permits the extraction and classification of a greater number of linear features compared to visual interpretation techniques. The ability to detect features that are not perceptible by visual observation is a strong point of CurvaTool processing. In the test area, CurvaTool output data correlate with visually detected linear features and show a good correlation with regional tectonics and iso-kinematic maps from the literature.
1 Corresponding author: Sabrina Bonetto, email: sabrina.bonetto @unito.it
INTRODUCTION Detection and extraction of lineaments are important steps in analyses related to mineral prospecting, hydrogeology studies, and tectonic studies to delineate major structural units, analyze structural deformation patterns, and identify geological boundaries (Clark and Wilson, 1994; Davis and Reynolds, 1996; Rangzan et al., 2008; and Lee et al., 2012). Usually, lineament maps created by fieldwork cannot identify all the lineaments in an area, due to the limited point of view of the mapper with respect to geological structures. Fieldwork can be a time-consuming, expensive, and sometimes dangerous undertaking. Therefore, any technique that can make field work more efficient is beneficial. In the past few years, the use of remote sensing products, coupled with progressive development of image enhancement techniques, has provided scientists with a fast and relatively cheap way to gather information that complements classical field geology. Optical remote sensing data are an important source of geological information for regional mapping, tectonic structural interpretation of faults, and identification of large-scale fractures and fracture zones (Suzen and Toprak, 1998; Wladis, 1999; Marghany and Hashim, 2010; Van der Meer et al., 2012; and Hashim et al., 2013). Traditionally, lineament mapping is based on a visual or semi-automatic interpretation of geomorphological features, such as morpho-tectonic elements, drainage network offsets, and stream segment alignments, and/or spectral criterion, such as tonal changes, patterns, and textures. However, the accuracy of features detected from satellite images is affected by several factors, including the characteristics of the sensor, the characteristics of the landform, lighting conditions, and cloud
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
179
Bonetto, Facello, and Umili
coverage (Smith and Wiseb, 2007). Illumination conditions can be affected by topography: The proportion of light reflected toward the satellite varies with the relative positions of the sun, target, and viewer, the geometry of which varies with topography (Shepherd and Dymond, 2003; Marghany and Hashim, 2010; and Rahnama and Gloaguen, 2014). For this reason, the use of a digital elevation model (DEM) or a digital terrain model (DTM), alone (Simpson and Anders, 1992; Byrd et al., 1994; Collet et al., 2000; and Seleem, 2013) or in combination with remotely sensed images on regional scale (Florinsky, 1998; Chorowicz et al., 1999; and Jacques et al., 2012), provides a useful alternative technique for lineament extraction that avoids most of the limiting factors discussed above (Moore et al., 1991; Jordan et al., 2005; and Masoud and Koike, 2011). Faults and linear features can be detected and quantified using terrain parameters extracted from a DTM, such as elevation, slope, and convexity (curvature). Morphology characterization is based on slope profiles, curvature values, spatial distribution of homogeneous areas, and cumulative frequency analysis of terrain distribution (Evans, 1980; Jordan et al., 2003, 2005). For example, curvature maps and slope maps can be used to recognize change in slope gradient, and consequently to identify fault lineament distribution (Ganas et al., 2005; Jordan et al., 2005). The literature presents several methods based on gridded data in DTMs and DEMs for calculating terrain parameters (Moore et al., 1991; Wise, 2000; Shary et al., 2002; and Kienzle, 2004). In this paper, we propose the application of a new semi-automatic approach based on DTMs for the extraction of potential lineaments. The approach tested here (CurvaTool software) was originally developed to automatically detect discontinuity traces in rock outcrops to evaluate their degree of fracturing (Umili et al., 2013; Ferrero et al., 2014). In this paper, the method is expanded for feature extraction over a larger area. As a first test area, we selected the Monferrato domain (NW Italy), part of the Tertiary Piedmont Basin (TPB). Significant results have been obtained showing the effectiveness of the CurvaTool software for preliminary assessment of potential geological lineaments (Bonetto et al., 2015). Here, we further discuss the functionality and applicability of the CurvaTool method in another test area with different accessibility and geological conditions. We selected an area in the Bagni di Vinadio municipality (Cuneo, NW Italy), which is part of the Argentera Massif (western Alps), as a test area to apply the CurvaTool method using a DTM with 10 m ground spatial resolution (source: Piedmont Region GeoNetwork, 2008). We selected this mountainous area for several reasons, including problematic accessibility, interest in the geothermal features of the
180
Figure 1. Tectonic sketch map of the Argentera Massif and adjoining regions (source: Baietto et al., 2009). BFZ: Bersezio Fault Zone, BF: Bersezio Fault, VSZ: Valletta Shear Zone, FMS: Fremamorta Shear Zone. The box highlights the investigated area.
area, and availability of data from the literature. Data from the CurvaTool code (Umili et al., 2013) in the test area were compared with literature information and visually extracted lineaments from ortho-photos (source: Arpa Piemonte, data acquisition October 2000). THE ARGENTERA MASSIF The Argentera Massif (AM) is in the western Alps and belongs to the External Crystalline Massifs. It crops out in the footwall of the Penninic Frontal Thrust (Figure 1), and it is divided into two main complexes: the Tin´ee Complex (TC) and the Malinvern-Argentera Complex (MAC), which represent the western and eastern portions of the massif, respectively. The AM is characterized by high-grade metamorphic rocks (schist, paragneiss, amphibolites, diatexite, and anatectic granitoid), locally intruded by postmetamorphic granitic bodies (Fry, 1989; Bogdanoff et al., 2000). The crystalline rocks are unconformably overlain by Triassic to Early Cretaceous carbonates, which are mostly detached above Late Triassic evaporites, with the basement-cover contact mainly striking NW-SE (Guglielmetti et al., 2013). The AM is characterized by Alpine-stage ductile shear zones and strike-slip and reverse faults, resulting from brittle reactivations of networks of structures of pre-Alpine and early-Alpine age (Bogdanoff et al., 1991; Musumeci and Colombo, 2002; Corsini et al., 2004; and Guglielmetti et al., 2013). Many faults belong to a NW-SE
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
Application of CurvaTool Approach to Geological Lineaments
Figure 2. Geological map of Bagni di Vinadio area (source: Guglielmetti et al., 2013).
system that mainly consists of right-lateral, high-angle, strike-slip faults. A conjugate system of left-lateral, NE-SW–trending faults is also present, locally turning to the ENE-WSW (Baietto et al., 2009). Three main high-angle shear zones cross the AM: the Valletta shear zone (VSZ), the Bersezio fault zone (BFZ), and the Fremamorta shear zone (FMS) (Figure 1). The FMS cuts the center-southernmost portion of the AM, connecting north to the BFZ, which consists of a dense set of faults striking NW-SE (Guglielmetti, 2012). The VSZ and the BFZ run parallel to each other in the northern sector of the AM. They are oriented NW-SE and define a 3-km-wide continuous belt of high-angle strike-slip faults trending both NW-SE to NNW-SSE and NE-SW to ENE-WSW (Baietto et al., 2009; Guglielmetti et al., 2013). The VSZ corresponds to the contact between the TC and MAC and is represented by an up to 1-km-thick mylonitic rock layer formed during a pre-Alpine deformation stage with a dextral strike-slip trend (Musumeci and Colombo, 2002; Guglielmetti et al., 2013). Triassic sedimentary cover uncomfortably overlies the basement rocks along the Sespoul, La Blance, and Tortissa Thrusts (Bogdanoff et al., 2000). Seismic and global positioning system (GPS) data show that the area is still tectonically active, with crustal shortening of 2–4 mm/yr induced by N-S to NE-SW compression (Madeddu et al., 1996; Ribolini and Spagnolo, 2008), especially in the axial region of the massif (Perello et al., 2001). The continuing crustal mobility of the Argentera is also indicated by permanent scatterers (PS) data (Morelli at al., 2011).
The Bagni di Vinadio test area is in the northwestern part of the AM, corresponding to the transition between TC from the west and the MAC from the east (Figure 2). The BFZ and VSZ are the main structures in the area. Migmatitic gneisses, fine-grained aplitic granites, and minor slices of sedimentary rocks mainly occur in the center and southern portion of the test area, whereas the sedimentary cover and PermoTriassic crystalline basement crop out in the northeastern sector of the test area (Guglielmetti et al., 2013). METHODOLOGY We first analyzed the test area (Figure 3) by visual detection of linear features, and then with the CurvaTool code (Umili et al., 2013). Results were compared and related to data from the literature. We performed visual lineament extraction on a set of orthophotos, called “Flight Flood 2000,” commissioned by Piedmont Region government (http://webgis.arpa. piemonte.it/joomla_gpa_32/) after the flood of October 2000: Aerial images (ground spatial resolution of 2.5 m) were collected during autumn 2000 in the northern sector of the Piedmont and during spring 2001 in its southern sector. Visual identification of lineaments is mainly based on the experience of the operator. Two criteria were applied to identify lineaments: (i) geomorphological and (ii) tonality. The geomorphological criterion is based on the identification of morpho-tectonic and drainage elements, such as rectilinear or segmented patterns of
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
181
Bonetto, Facello, and Umili
Figure 3. Test area: Bagni di Vinadio, Argentera Massif. Base map: Ortho-photo from Flight Flood 2000 (source: Arpa Piemonte, 2000).
valleys, scarps, and ridges, drainage pattern offsets, and stream-segment alignments. The tonality criterion is based on the visual interpretation of differences in color tones and light contrasts. Tonality varies as a function of differences in vegetation, lithology, soil water content, permeability, and rock strength (O’Leary et al., 1976). Since the identification of lineaments can be affected by changes in the illumination azimuth and slope, the ortho-photos were enhanced by overlaying the DTM elevation contour map onto the photos to show the exact locations of valleys, ridges, and slope breaks. The CurvaTool code semi-automatic method for lineament identification is based on the assumption that a geological lineament can be geometrically identified as a convex or concave edge on a DTM, particularly where there is structural control of the geomorphological evolution of the analyzed area. A detailed description of the working principles and calculation methods of CurvaTool code can be found in Bonetto et al. (2015). A DTM (ground resolution of 1 point every 20 m), containing the same area as the one covered by the previously described ortho-photos, was used as input for the CurvaTool code (Umili et al., 2013). Since the area is mountainous, with an elevation difference of 182
2130 m, the DTM surface contains large numbers of recognizable crests and valleys, making the area suitable for semi-automatic linear feature extraction by the CurvaTool code. The code is based on an estimate of principal curvature values (maximum and minimum) associated with each DTM point, thus implementing the method proposed by Chen and Schmitt (1992) and extended by Dong and Wang (2005). As briefly illustrated by the flow chart in Figure 4, the user is asked for two thresholds: Tmax, which represents the minimum acceptable value of maximum principal curvature (kmax) with which to select DTM points potentially belonging to significant convex edges (e.g., crests), and Tmin, which represents the maximum acceptable value of minimum principal curvature (kmin) with which to select DTM points potentially belonging to significant concave edges (e.g., valleys). After a process of point linking, each resulting polyline is segmented, each segment is measured, and its angle with respect to north is calculated. The quality of the DTM is fundamental (Kraus, 1993; Kraus and Pfeifer, 1998): The smaller the mean distance between adjacent points, the better is the correspondence between the discretized surface and the actual ground surface, and the lower is the smoothing effect. The method works particularly well on models with a wide range of principal curvature values, that is, on surfaces with a high degree of non-planarity. Once linear feature extraction has been performed, post-processing operations are required in order to obtain significant results; therefore, the algorithm called “Filter” was created by the authors to perform operations on the linear features database (Figure 5). Postprocessing can follow two approaches, based on the degree of knowledge of the area and on the purposes of the study. The first approach is applicable in cases where no literature data are available for the studied area: In this case, a frequency analysis is performed on linear feature directions; by analyzing the resulting rosette of directions, the user can make observations useful for a preliminary tectonic assessment. Where the area is already geologically well-known, and literature data are available for the studied area, in terms of mean direction of lineaments sets, post-processing starts with a comparison with literature data: The user has to assign the minimum lineament length and the orientations of the expected clusters of lineaments (expressed by an angle with respect to north and its standard deviation). The Filter code deletes linear features shorter than the fixed length and classifies each remaining edge, attributing it to the corresponding input cluster. Nonclassified features are recorded as “others” (Bonetto et al., 2015). Mapping and statistical analysis of lineaments length can then be performed on the obtained database.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
Application of CurvaTool Approach to Geological Lineaments
Figure 4. Flow chart representing CurvaTool process.
Figure 5. Flow chart representing Filter process.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
183
Bonetto, Facello, and Umili
Figure 6. Maps of lineaments of Bagni di Vinadio obtained from visual extraction (A) and by CurvaTool processing (B).
RESULTS AND DISCUSSION Large numbers of linear features were identified in our test area: 1,848 from visual extraction and 8,465 by the CurvaTool software (Figure 6). The number of linear features extracted by CurvaTool is remarkable with respect to the number identified by visual extraction. The three-dimensional (3D) geometrical approach implemented in the software allows CurvaTool to identify all concave or convex edges of the ground surface, while visual extraction is subjective and depends on the experience and ability of the analyst and the observation scale. Moreover, some variations in surface edges are not visually detectable but can be geometrically described in terms of curvature and therefore analyzed by the code. However, a few remarks must be made on the possibility that some of the extracted linear features could be false lineaments, i.e., natural or artificial linear elements that do not represent geological lineaments. First of all, the DTM resolution plays an important role. In fact, every false lineament for which the dimensions are smaller or similar to the ground resolution is not, or is only partially, represented by the DTM; therefore, it is not detectable as a linear feature by CurvaTool. Con-
184
sidering that a 5-m-resolution DTM represents a very detailed model for our purposes, it is likely that canals and riverbanks would not be observable on the DTM surface. Moreover, the most common artificial linear elements, such as roads and railroads, are almost flat, and therefore, even if detectable on the DTM surface, they belong to areas characterized by non-significant curvature values. However, the possibility that a few linear features representing false lineaments could be detected exists; therefore, geologically based reasoning must be used in this case. Generally, main faults are not isolated structures: The area in which a fault is located is usually characterized by other structural lineaments, most of which have similar direction. Moreover, our purpose is not only to create a lineament map, but also to obtain information about the average direction of lineament sets. Therefore, a single false lineament cannot invalidate the result of a cluster analysis performed on all extracted linear features. Getting back to the results discussion, all the detected lineaments were statistically analyzed to compare them in terms of quantity, orientation, and geographical distribution. Filter was applied to the results of both the semiautomatic and visual methods to perform a cluster
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
Application of CurvaTool Approach to Geological Lineaments Table 1. Orientation of lineament sets used as input for Filter. Set
Azimuthal Direction (degrees)
Standard Deviation (degrees)
L1 L2 L3 L4
N125 N50 N0 N90
15 10 20 19.99
analysis and assign linear features into sets (Bonetto et al., 2015). Four set orientations were assigned (Table 1), according to geomorphological and structural literature information (Baietto et al., 2009; Guglielmetti et al., 2013). The acronym TELF indicates total extracted linear features. Using visual extraction (Table 2), NW-SE lineaments (L1) are the least frequent (9.42 percent of TELF), and they are mainly in the southern part of the area; NE-SW lineaments (L2) are the most numerous (19.32 percent of TELF), and they show a principal distribution in the center and southern areas. N-S (L3) and E-W lineaments (L4) show a similar frequency (nearly 18 percent of TELF both for L3 and L4), and they are mainly in the northern part of the area. More than a third of the linear features do not correspond to the orientations indicated in Filter: They represent 36 percent of TELF. For the results of the CurvaTool processing (Table 3), the NW-SE set (L1) shows a high frequency (20 percent of TELF) and uniform distribution in the test area; the NE-SW set (L2) is less frequent (11.79 percent of TELF) and is chiefly distributed in the center and southwestern part of the test area. The N-S lineaments (L3) are predominant (20.58 percent of TELF), and they show a homogeneous distribution throughout the whole area. The E-W set (L4) shows a slightly smaller frequency (20.20 percent of TELF), and it was mainly identified in the northern and southeastern parts of the area. The linear features found with CurvaTool that are not included in the range assigned to Filter represent 27 percent of total features. In Figure 7, the different sets have been separated to better compare linear features extracted by visual analysis and CurvaTool processing. Despite the high number of linear features detected, the percentage of Table 2. Statistics for visually extracted lineaments. Azimuthal Direction (degrees) Set
No. Lineaments
Minimum
Maximum
Mean
Standard Deviation
L1 L2 L3 L4
174 357 331 329
110.05 40.24 0.00 70.02
139.90 59.86 19.98 109.98
124.85 49.38 0.23 87.56
8.39 5.36 11.12 10.62
Table 3. Statistics for lineaments obtained from CurvaTool processing. Azimuthal Direction (degrees) Set
No. Lineaments
Minimum
Maximum
Mean
Standard Deviation
L1 L2 L3 L4
1693 998 1742 1710
110.10 40.10 −19.98 70.02
139.99 59.94 19.98 109.98
126.66 49.77 −0.89 90.43
8.36 5.65 11.76 11.59
unassigned lineaments identified by CurvaTool (27.43 percent) is lower than that obtained from the visual method (31.54 percent). Comparing the percentage of linear features assigned to each set, the main difference between CurvaTool and visual extraction is in the L1 and L2 sets (Table 4). The sets used in Filter correspond to the orientations of the main geological lineaments in the AM; in particular, L1 and L2 correspond to the two main observed conjugate systems that are associated with NWSE–striking faults. The NE-SW system (L2) is minor and discontinuous. Set L1 is dominant: It reactivates pre-existing shear zones and pre-Alpine foliations in the basement. Also, the basement-cover contact strikes NW-SE (Perello et al., 2001; Baietto et al., 2009). Comparing semi-automatic and visual processing, CurvaTool underestimates the importance of L1, whereas visual extraction overrates L2. This disparity is probably due to the drainage network in the test area, which is related to geological lineament orientation, thus conditioning the detection of linear features. In the test area, most of the main rivers and first-order stream channels are NW-SE elongated and have a flat floor. Since DTM points in open valleys correspond to very low and uniform values of curvature, it is likely that they were discarded from the analysis during the choice of curvature thresholds in CurvaTool. The CurvaTool technique is called semi-automatic because the user is asked for two thresholds: minimum and maximum principal curvature values that discriminate between significant and insignificant edges. Therefore, very flat areas are not considered significant for edge identification. However, where the orientation of low-order channels corresponds to a NW-SE strike, both CurvaTool and visual extraction identify linear features belonging to set L1. As reported in Ribolini and Spagnolo (2008), in some portions of the test area, for example, along the Stura River, several low-order channels run perpendicular to both the main-stem river and geological lineaments. This type of drainage pattern, where present, could be the cause of the high number of linear features assigned to L2, particularly by visual extraction, where subjectivity and main morphological elements
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
185
Bonetto, Facello, and Umili
Figure 7. Map of linear features of Bagni di Vinadio extracted and processed with Filter. Four sets are highlighted: (A) L1 (NW-SE), (B) L2 (NE-SW), (C) L3 (N-S), and (D) L4 (E-W).
186
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
Application of CurvaTool Approach to Geological Lineaments Table 4. Comparison between classified lineaments obtained from CurvaTool processing and visual interpretation.
Set L1 L2 L3 L4 Not Assigned
Percentage of Total Linear Features (8465) Extracted by CurvaTool
Percentage of Total Linear Features (1883) Extracted Visually
20.00% 11.79% 20.58% 20.20% 27.43%
11.78% 21.93% 17.57% 17.15% 31.54%
influence the process. Musso et al. (2009) noticed that, in the AM, geomorphic evolution is particularly controlled by a NE-SW normal fault system consisting of relatively short segments. Therefore, the influence of the morphological criterion on the visual interpretation is likely the reason for the high percentage of L2 lineaments identified by visual extraction, despite the secondary relevance of this set to L1. Most of the linear features detected by CurvaTool as belonging to L1 and L2 are short segments aligned along the NW-SE and NE-SW directions, respectively. Perello et al. (2001) described the NW-SE– and NE-SW–striking systems as discontinuous high-angle faults. L3 and L4 have similar percentage values in both analytical approaches. Geological mapping and literature data (Malaroda et al., 1970; Crema et al., 1971; Perello et al., 2001; Baietto et al., 2009; and Guglielmetti, 2012) indicate the presence of geological lineaments striking E-W (L4). They are usually short and discontinuous, and they frequently connect or displace faults belonging to the main NW-SE system. With regard to the ENE-WSW lineaments, they are associated with the conjugate system of NW-SE strike-slip faults (L2). CurvaTool and visual extraction also detected N-S–striking linear features; no important structures with this orientation are known at a regional scale in the study area, but detailed field data from Perello et al. (2001) reported the presence of faults with N-S orientation in the Bagni di Vinadio area, associated with low-angle shear zones. Guglielmetti (2012) identified N-S–striking morphological elements using photo-interpretation, particularly in the SE part of the AM (Terme di Vinadio area). We observed a non-homogeneous distribution of linear features and different lineament domains, particularly using CurvaTool processing (Figure 8). The spatial distribution and alignment of linear features detected by both CurvaTool and visual analysis indicate quite clearly the presence of two main lineaments, in the center and SW part of the test area. When compared to geological mapping (Malaroda et al., 1970; Crema et al., 1971; Baietto et al., 2009; and Guglielmetti, 2012), the two lineament domains correspond
respectively to the Bersezio and Valletta Faults. The anomalous concentration of NW-SE– and NE-SW– oriented linear features in the middle of these lineament systems is due to the presence of a shear zone called the “Bersezio Fault Zone” (area A in Figure 8A and B), a complex system of anastomosing faults made up of lens-shaped tectonic slices (Perello et al., 2001) formed by NW-SE (L1), NE-SW (L2), and E-W (L4) lineaments. A change in spatial distribution and frequency of linear features was observed NE of the Stura River. The NE sector (area C in Figure 8A and B) shows a homogeneous distribution of linear features with preferred NW-SE and E-W orientations, whereas the sector between this domain and the Stura River (area B in Figure 8A and B) is characterized by a predominance of NE-SW linear features. The boundary between the two sectors seems to correspond to the NW-SE–trending basement-cover contact. The southeastern areas of Bagni di Vinadio (area D in Figure 8A and B), bounded on the northwest by the Corborant River, show a predominance of L4 and L1 features. Part of this sector is geologically still included in the Bersezio Fault Zone. The domain boundaries, particularly as highlighted by CurvaTool, strike NW-SE and NE-SW, coherent with both main fault directions and iso-kinematic boundaries defined with the PS interferometric synthetic aperture radar (InSAR) technique by Morelli et al. (2011). Iso-kinematic boundaries are mainly aligned along both a NW-SE direction, parallel or subparallel to the NW-SE transpressive faults, and a NE-SW direction, subparallel to the main drainage network and normal fault system (Musso et al., 2009). In the visual approach, it is possible to observe the same domains, but the previously described limits are not as well defined compared to CurvaTool, probably because of the limited number of linear features detected (Figure 8B). CONCLUSIONS The CurvaTool code has been applied to DTMs over large areas to semi-automatically detect edges that represent potential geological linear features. To verify the results obtained by the software, CurvaTool outputs were compared to visually extracted linear features and also to geological literature data. This study demonstrates that CurvaTool processing permits extraction and classification of a greater number of linear features compared to visual interpretation. The ability to detect features not perceptible by visual observation is a strong point of CurvaTool processing. Visual interpretation is unable to detect short segments and less evident surface edges as well as CurvaTool; moreover, visual extraction is subjective and influenced by the experience of the analyst. The overall
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
187
Bonetto, Facello, and Umili
Figure 8. Main geological lineaments obtained by visual extraction (A) and CurvaTool (B), and then processed with Filter. Areas characterized by a homogeneous/specific linear feature distribution are indicated as domains A, B, C, and D.
positive aspects of this semi-automatic process include the rapidity of preliminary assessment, the capacity to identify the most interesting areas to be investigated, and the ability to analyze areas that are not directly accessible. DTM resolution has a direct influence on lineament definition and completeness. The number of points on the surveyed surface and, consequently, the amplitude of the triangles of the digital model influence the quality of the approximation of the real surface. In addition, a decrease in resolution results in “smoothing” and consequent deterioration of the edges of the surface. This reduces the range of principal curvatures and, depending on the triangulation, disrupts or alters the continuity of the edges. In the test area, CurvaTool data are consistent with visually detected linear features and show a good correlation with structural data (Malaroda et al., 1970; Perello et al., 2001; Corsini et al., 2004; and Baietto et al., 2009) and iso-kinematic maps (Morelli et al., 2011), demonstrating the applicability of the semiautomatic approach. The abundance of linear features identified by CurvaTool allows for better identification of homogeneous domains (in terms of frequency and distribution of linear features). CurvaTool processing
188
identifies the main faults and shear zones reported in the literature. The alignment of long, single segments and their high density in NW-SE–elongated areas corresponds to the locations of the Bersezio and Valletta Faults, their deformation zones, and the basementcover contact. The semi-automatic method has the potential to detect main geomorphic and structural features at a regional scale, particularly in areas where tectonic activity has a strong control on geomorphic evolution. With regard to the NW sector of the AM, where Bagni di Vinadio is located, the lower relief results in a generally higher sensitivity of the drainage network to faults and fracture systems, which determines preferential orientation of the lineaments (Ribolini and Spagnolo, 2008). Preliminary results of this research show that the application of the CurvaTool code to large areas can be a potential tool in preliminary geological and structural studies, particularly in areas that are not directly accessible or when scarce existing data are available. CurvaTool can give useful and rapid information about the orientation and spatial distribution of potential geological and geomorphological lineaments, and the
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
Application of CurvaTool Approach to Geological Lineaments
possible presence of domains with homogeneous features and lineament distributions. Based on this approach, further specific field studies can be planned to verify the results. In alpine environments, CurvaTool shows a high potential due to less Quaternary cover obscuring tectonic elements, and it is easier to validate results because of the abundance of outcrops and field data. Further research in the test area will consist of statistical analysis of the frequency distribution and length of the lineaments belonging to each set, and the influence of rock type on linear feature detection. Subsequent investigations and statistical processing of the data are needed to validate and improve the software. Testing new areas with different tectonic and geomorphologic environments is also a future priority. REFERENCES ARPA PIEMONTE, 2000, Ortho-photos–Fotoindice volo Alluvione2000 WMS. http://webgis.arpa.piemonte.it/ags101free/ services/topografia_dati_di_base/Riprese_aerofotogrammetri che_fotoindice_volo_Alluvione2000/MapServer/WMSServer. BAIETTO, A.; PERELLO, P.; CADOPPI, P.; AND MARTINOTTI, G., 2009, Alpine tectonic evolution and thermal water circulations of the Argentera Massif (south-western Alps): Swiss Journal Geosciences, Vol. 102, pp. 223–245. BOGDANOFF, S.; MENOT, R.; AND VIVINER, G., 1991, Les massifs cristallins externes des Alpes Occidentales francaises, un fragment de la zone interne varisque: Science Geologique Bulletin, Vol. 44, pp. 237–285. BOGDANOFF, S.; MICHARD, A.; MANSOUR, M.; AND POUPEAU, G., 2000, Apatite fission tracks analysis in the Argentera massif: Evidence of contrasting denudation rates in the External Crystalline Massifs of the western Alps: Terra Nova, Vol. 12, pp. 117–125. BONETTO, S.; FACELLO, A.; FERRERO, A. M.; AND UMILI, G., 2015, A tool for semi-automatic linear feature detection based on DTM: Computers & Geosciences, Vol. 75, pp. 1–12. BYRD, J. O. D.; SMITH, R. B.; AND GEISSMAN, J. W., 1994, The Teton Fault, Wyoming: Neotectonics, and mechanisms of deformation: Journal Geophysical Research, Vol. 99, No. B10, pp. 20095–20122. CHEN, X. AND SCHMITT, F., 1992, Intrinsic surface properties from surface triangulation. In G. Sandini (Editor), Proceedings of the Second European Conference on Computer Vision: Springer, Berlin, Heidelberg, pp. 739–743. CHOROWICZ, J.; DHONT, D.; AND GUNDOGDU, N., 1999, Neotectonics in the eastern North Anatolian Fault region (Turkey) advocates crustal extension: Mapping from SAR ERS imagery and digital elevation model: Journal Structural Geology, Vol. 21, pp. 511–532. CLARK, C. D. AND WILSON, C., 1994, Spatial analysis of lineaments: Computers & Geosciences, Vol. 20, No. 7–8, pp. 1237–1258. COLLET, B.; TAUD, H.; PARROT, J. F.; BONAVIA, F.; AND CHOROWICZ, J., 2000, A new kinematic approach for the Danakil block using a digital elevation model representation: Tectonophysics, Vol. 316, pp. 343–357. CORSINI, M.; RUFFET, G.; AND CABY, R., 2004, Alpine and lateHercynian geochronological constraints in the Argentera Massif (western Alps): Eclogae Geologicae Helvetiae, Vol. 97, pp. 3–15.
CREMA, G.; DAL PIAZ, G.; MERLO, C.; AND ZANELLA, E., 1971, Carta Geologica d’Italia alla scala 1:100.000, Fogli Argentera, Dronero, Demonte, n. 78, 79 e 90 e Note illustrative: Istituto Poligrafico e Zecca dello Stato–Archivi di Stato, Rome, Italy [in Italian]. DAVIS, G. H. AND REYNOLDS, S. J., 1996, Structural Geology of Rocks and Regions, 2nd ed.: John Wiley & Sons, New York. DONG, C. AND WANG, G., 2005, Curvatures estimation on triangular mesh: Journal Zhejiang University Science, Vol. 6A, Suppl. I, pp. 128–136. EVANS, I. S., 1980, An integrated system of terrain analysis and slope mapping: Zeitschrift Geomorphologie, Vol. 36, pp. 274–295. FERRERO, A. M.; MIGLIAZZA, M. R.; AND UMILI, G., 2014, Rock mass characterization by means of advanced survey methods. Keynote lecture. In Rock Engineering and Rock Mechanics: Structures in and on Rock Masses: Proceedings of EUROCK 2014, ISRM European Regional Symposium, CRC Press/Balkema, Leiden, The Netherlands, 27–29 May 2014, pp. 17-27. FLORINSKY, I. V., 1998, Combined analysis of digital terrain models and remotely sensed data in landscape investigations: Progress Physical Geography, Vol. 22, pp. 33–60. FRY, N., 1989, Southwestward thrusting and tectonics of the western Alps. In Coward, M. P., Dietrich, D., and Park, R. G. (Editors), Alpine Tectonics: Special Publication 45, Geological Society of London, London, U.K., pp. 83–109. GANAS, A.; PAVLIDES, S.; AND KARASTATHIS, V., 2005, DEM-based morphometry of range-front escarpments in Attica, central Greece, and its relation to fault slip rates: Geomorphology, Vol. 65, No. 3, pp. 301–319. GUGLIELMETTI, L., 2012. Multidisciplinary Approach of Geothermal Exploration in the Argentera Massif (Southwestern Alps): Ph.D. Thesis, Department of Earth Sciences, Torino University, Italy, 193 p. GUGLIELMETTI, L.; COMINA, C.; ABDELFETTAHA, Y.; SCHILL, E.; AND MANDRONE, G., 2013, Integration of 3D geological modeling and gravity surveys for geothermal prospection in an Alpine region: Tectonophysics, Vol. 608, pp. 1025–1036. HASHIM, M.; AHMAD, S.; JOHARI, M. A. M.; AND POUR, A. B., 2013, Automatic lineament extraction in a heavily vegetated region using Landsat Enhanced Thematic Mapper (ETM + ) imagery: Advances Space Research, Vol. 51, No. 5, pp. 874–890. JACQUES, P. D.; MACHADO, R.; AND NUMMER, A. R., 2012, A comparison for a multiscale study of structural lineaments in southern Brazil: LANDSAT-7 ETM + and shaded relief images from SRTM3-DEM: Anais Academia Brasileira Ciˆencias, Vol. 84, No. 4, pp. 931–942. JORDAN, G.; CSILLAG, G.; SZUCS, A.; AND QVARFORT, U., 2003, Application of digital terrain modelling and GIS methods for the morphotectonic investigation of the Kali Basin, Hungary: Zeitschrift Geomorphologie, Vol. 47, pp. 145–169. JORDAN, G.; MEIJNINGER, B. M. L.; VAN HINSBERGEN, D. J. J.; MEULENKAMP, J. E.; AND VAN DIJK, P. M., 2005, Extraction of morphotectonic features from DEMs: Development and applications for study areas in Hungary and NW Greece: International Journal Applied Earth Observation Geoinformation, Vol. 7, pp. 163–182. KIENZLE, S., 2004, The effect of DEM raster resolution on first order, second order and compound terrain derivatives: Transactions GIS, Vol. 8, No. 1, pp. 83–111. ¨ Verlag, KRAUS, K., 1993, Photogrammetry, Vol. 1, 4th ed.: Dummler Bonn, Germany. KRAUS, K. AND PFEIFER, N., 1998, Determination of terrain models in wooded areas with airborne laser scanner data: ISPRS Journal Photogrammetry & Remote Sensing, Vol. 53, No. 4, pp. 193–203.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
189
Bonetto, Facello, and Umili LEE, M.; MORRIS, W.; HARRIS, J.; AND LEBLANC, G., 2012, An automatic network-extraction algorithm applied to magnetic survey data for the identification and extraction of geologic lineaments: Leading Edge, Vol. 31, No. 1, pp. 26–31. MADEDDU, B.; BERTOUX, N.; AND STEPHAN, J. F., 1996, Champ de contrainte post-pliocene et deformation recentes dans les Alpes sud-occidentales: Bulletin Societe Geologique France, Vol. 8, pp. 797–810. MALARODA, R.; CARRARO, F.; DAL PIAZ, G. B.; FRANCESCHETTI, B.; STURANI, C.; AND ZANELLA, E., 1970, Carta Geologica del Massiccio dell’Argentera alla scala 1:50.000 e Note Illustrative: Memorie 9, Societ`a Geologica Italiana, Rome, Italy, pp. 557– 663 [in Italian]. MARGHANY, M. AND HASHIM, M. 2010, Lineament mapping using multispectral remote sensing satellite data research: Journal Applied Sciences, Vol. 5, No. 2, pp. 126–130. MASOUD, A. AND KOIKE, K., 2011, Morphotectonics inferred from the analysis of topographic lineaments autodetected from DMEs: Application and validation for the Sinai Peninsula, Egypt: Tectonophysics, Vol. 510, pp. 291–308. MOORE, I. D.; GRAYSON, R. B.; AND LADSON, A. R., 1991, Digital terrain modelling: A review of hydrological, geomorphological, and biological applications: Hydrological Processes, Vol. 5, No. 1, pp. 3–30. MORELLI, M.; PIANA, F.; MALLEN, L.; NICOLO` , G.; AND FIORASO, G., 2011, Iso-kinematic maps from statistical analysis of PSInSAR data of Piemonte, NW Italy: Comparison with geological kinematic trends: Remote Sensing Environment, Vol. 115, No. 5, pp. 1188–1201. MUSSO, A.; PIANA, F.; BERTOK, C.; D’ATRI, A.; MARTIRE, L.; PEROTTI, E.; AND VARRONE, D., 2009, Post-Eocene transpression in western Ligurian Alps: Geometry and kinematics of LimoneViozene Deformation Zone (France-Italy border): Epitome, Vol. 3, pp. 181. MUSUMECI, G. AND COLOMBO, F., 2002, Late Visean mylonitic granitoids in the Argentera Massif (western Alps, Italy): Age and kinematic constraints on the Ferriere–Mollieres Shear Zone: Comptes Rendus Geoscience, Vol. 334, pp. 213–220. O’LEARY, D. W.; FRIEDMAN, J. D.; AND POHN, H. A., 1976, Lineament, linear, lineation: Some proposed new standards for old terms: Geological Society America Bulletin, Vol. 87, No. 10, pp. 1463–1469. PERELLO, P.; MARINI, L.; MARTINOTTI, G.; AND HUNZIKER, C., 2001, The thermal circuits of the Argentera Massif (western Alps, Italy): An example of low enthalpy geothermal resources controlled by Neogene alpine tectonics: Eclogae Geologicae Helvetiae, Vol. 94, pp. 75–94. PIEDMONT REGION GEONETWORK, 2008, http://www.geoportale. piemonte.it/geocatalogorp/index.jsp.
190
RAHNAMA, M. AND GLOAGUEN, R., 2014, TecLines: A MATLABbased toolbox for tectonic lineament analysis from satellite images and DEMs, part 1: Line segment detection and extraction: Remote Sensing, Vol. 6, No. 7, pp. 5938–5958. RANGZAN, K.; CHARCHI, A.; ABSHIRINI, E.; AND DINGER, J., 2008, Remote sensing and GIS approach for water-well site selection, southwest Iran: Environmental & Engineering Geoscience, Vol. 14, pp. 315–326. RIBOLINI, A. AND SPAGNOLO, M., 2008, Drainage network geometry versus tectonics in the Argentera Massif (French–Italian Alps): Geomorphology, Vol. 93, pp. 253–266. SELEEM T. A., 2013, Analysis and tectonic implication of DEMderived structural lineaments, Sinai Peninsula, Egypt: International Journal Geosciences, Vol. 4, pp. 183–201. SHARY, P. A.; SHARAYA, L. S.; AND MITUSOV, A. V., 2002, Fundamental quantitative methods of land surface analysis: Geoderma, Vol. 107, No. 1, pp. 1–32. SHEPHERD, J. D. AND DYMOND, J. R., 2003, Correcting satellite imagery for the variance of reflectance and illumination with topography: International Journal Remote Sensing, Vol. 24, No. 17, pp. 3503−3514. SIMPSON, D. W. AND ANDERS, M. H., 1992, Tectonics and topography of the Western United States—An application of digital mapping: GSA Today, Vol. 2, No. 6, pp. 117–121. SMITH, M. J. AND WISEB, S. M., 2007, Problems of bias in mapping linear landforms from satellite imagery: International Journal Applied Earth Observation Geoinformation, Vol. 9, pp. 65–78. SUZEN, M. L. AND TOPRAK, V., 1998, Filtering of satellite images in geological lineament analyses: An application to a fault zone in Central Turkey: International Journal of Remote Sensing, Vol. 19, No. 6, pp. 1101–1114. UMILI, G.; FERRERO, A. M.; AND EINSTEIN, H. H., 2013, A new method for automatic discontinuity traces sampling on rock mass 3D model: Computers & Geosciences, Vol. 51, pp. 182– 192. VAN DER MEER, F. D.; VAN DER WERFF, H. M. A.; VAN RUITENBEEK, F. J. A.; HECKER, C. A.; BAKKER, W. H.; NOOMEN, M. F.; VAN DER MEIJDE, M.; CARRANZA, E. J. M.; BOUDEWIJN DE SMETH, J.; AND WOLDAI, T., 2012, Multi- and hyperspectral geologic remote sensing: A review: International Journal Applied Earth Observation Geoinformation, Vol. 14, No. 1, pp. 112–128. WISE, S., 2000, Assessing the quality for hydrological applications of digital elevation models derived from contours: Hydrological Processes, Vol. 14, No. 11–12, pp. 1909–1929. WLADIS, D., 1999, Automatic lineament detection using digital elevation models with second derivative filters: Photogrammetric Engineering & Remote Sensing, Vol. 65, No. 4, pp. 453–458.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 179–190
Effects of Igneous Intrusions on Coal Pore Structure, Methane Desorption and Diffusion within Coal, and Gas Occurrence MING-YI CHEN YUAN-PING CHENG1 HONG-XING ZHOU LIANG WANG FU-CHAO TIAN KAN JIN National Engineering Research Center for Coal & Gas Control, China University of Mining & Technology, Xuzhou 221116, China, and Key Laboratory of Gas and Fire Control for Coal Mines, China University of Mining & Technology, Xuzhou 221116, China; mingyichen@cumt.edu.cn
Key Terms: Mining Geology, Igneous Rock, Coal and Gas Outburst, Coal Seam Gas Occurrence ABSTRACT Igneous intrusions are distributed extensively in the Huaibei coalfield, China. In the Haizi coal mine, coal and gas outbursts have occurred 11 times under an extremely thick sill (average thickness 120 m). This paper presents the results of a study on the influences of the igneous rock on coal pore structure, methane desorption and diffusion properties, and coal seam gas occurrence. The results show that the thermal evolution effect of the igneous sill prominently increases the specific surface area and pore volume of the affected coal. Samples HZ1 and HZ2 (No. 7 and No. 9 seams, respectively) closer to the sill possess improved pore connectivity, while samples HZ3 and HZ4 (away from the sill) and sample HZ5 (without sill covering) of the No. 10 coal seam have poor pore connectivity. Moreover, the effective diffusivity and desorption indexes of the coal increase progressively closer to the sill. The thermal effect of the igneous sill promotes the development of coal pores, thus leading to better pore connectivity, more desorbed gas, and much higher gas desorption and diffusion rates. Consequently, the thermal evolution effect of the igneous sill can change the occurrence and characteristics of the entrapment effect in the underlying coal seams, thus resulting in a high probability of gas hazards or even coal and gas outbursts in the coal seam close to the igneous sill. Engineering practices show that the affected coal seams have high gas content, gas pressure, and gas emission amounts as well as a high propensity for coal and gas outburst.
1 Corresponding
author email: ypcheng@cumt.edu.cn.
INTRODUCTION Coal measure strata intruded by igneous rock can be found in many coalfields worldwide. Igneous intrusion activity changes the properties of the coal, such as their petrographic, maceral, and geochemical properties, resulting from the thermal effects associated with the intrusions (Finkelman et al., 1998; Golab and Carr, 2004; Stewart et al., 2005; Dai and Ren, 2007; Rimmer et al., 2009; Yang et al., 2012; Chen et al., 2014; and Rahman and Rimmer, 2014). The changes in the physicochemical properties of coal depend on the distance between the coal and the igneous intrusions, the duration of magmatic-derived heat, and the temperature of the intrusions, among other factors (Cooper et al., 2007; Wu et al., 2014). In general, with an increase in the metamorphic grade of coal affected by intruding magma, the porosity and adsorption capacity of the coal are enhanced, thus increasing its gas storage capacity (Saghafi et al., 2008). Meanwhile, a significant quantity of gas may be trapped in the coal seams by the magmatic rock, which can cause potential problems regarding mining safety when working with the affected coal seams. Many studies have suggested that such gas hazards are closely related to igneous intrusions (Jiang et al., 2011; Sachsenhofer et al., 2012; Wang et al., 2014b, 2014d; and Xu et al., 2014). The thermal effect of the intrusions has a significant impact on the pore structure of the coal (Mastalerz et al., 2009; Yao and Liu, 2012; and Wu et al., 2014), and it also influences the gas adsorption property and gas flow characteristics of the coal (Saghafi et al., 2008; Yao et al., 2011). Igneous intrusions have been identified as an important factor in abnormally high gas emissions (Shepherd et al., 1981; Xu et al., 2014). Previous research focused more on the changes in the pore parameters, such as pore volume and porosity, as well as in the gas adsorption capacity of the coal in the
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
191
Chen, Cheng, Zhou, Wang, Tian, and Jin
Figure 1. The location of the Huaibei coalfield (A) and structural outline map (B).
affected area. Correspondingly, there has been a lack of investigation regarding the pore structure and its influences on the gas desorption and diffusion properties of coal under the thermal effect from the igneous intrusions. The pore structure of coal changes the gas adsorption and flow capacities of coal (An et al., 2013; Cai et al., 2013), both of which also have a significant impact on coal and gas outbursts (Skoczylas et al., 2014; Jian et al., 2015). In China, several prediction indexes associated with gas desorption and diffusion properties within coal are generally used for the prevention and control of outburst hazards in coal mines (Cheng et al., 2010; Cheng et al., 2016). Therefore, it is reasonable to perform a study to understand the change in pore structure and its impact on gas desorption and diffusion properties of coal under the influence of intrusive thermal effects. Magmatic activity is frequent and widespread within the Huaibei coal field, China. The Haizi coal mine lies in the middle of the Huaibei coalfield in Anhui Province, where igneous intrusions are extensively distributed and have caused 11 coal and gas outbursts, all of which occurred under an extremely thick sill. Many studies have since been conducted on the coal seam gas occurrence, the disaster-causing mechanisms of the intrusions, and the prevention and control of coal and gas outbursts in the Haizi mine (Wang et al., 2013; Chen et al., 2014; Wang et al., 2014a, 2014b, 2014c, 2014d; and Zhang et al., 2015). Based on these studies, this paper is focused on determining the influence of the thermal effect of an igneous sill on pore structure, in-
192
cluding pore size distribution and pore shape, methane desorption and diffusion properties, and the impact of pore structure on these properties in an underlying coal seam. It is important to understand the influence of the igneous sill on coal seam gas occurrence and gas hazards in the Haizi mine. Moreover, these results can provide theoretical guidance for coal mines with similar geological conditions to prevent and control gas hazards. MAGMATIC INTRUSIONS IN THE HAIZI MINE As shown in Figure 1A, the Huaibei coalfield is located in northern Anhui Province with an area of approximately 13,000 km2 . Through multi-stage tectonic movement, it has formed complex fold and fault systems (Qu et al., 2008). It has also experienced many periods of magmatic activity. Magmatic activity was the most active during the Mesozoic Yanshan epoch and caused the greatest damage synchronous with coal metamorphism in the coalfield (Han, 1990; Yang et al., 1996). As shown in Figure 1B, Han (1990) divided Yanshanian magmatic intrusions into four phases. The first phase was composed of neutral magmatic rock, the magmatism of the second phase was intermediately acidic, the third phase contained acidic magmatic rock, and the fourth phase was composed of basic and ultrabasic magmatic rock. As shown in Figure 2A, the Haizi coal mine is bordered by the Damajia Fault and the Linhuan mine to the southeast, and the Daliujia Fault to the west in the Huaibei coalfield of Anhui Province, China.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
Thermal Effects of Igneous Rock on Coal
Figure 2. (A) Map of igneous intrusion distribution in the Haizi coal mine. (B) Histogram of coal-bearing strata of the II102 mining area in the Haizi coal mine. (C) Photos of igneous rock.
Its regional geological structure is closely related to the Subei Fault. The magmatic evolution in the Haizi mine can be summarized as follows. During the Yanshan epoch, deep crustal magma upwelled along the Subei Fault zones and subsequently flowed into the Daliujia Fault through the hanging wall of the Subei Fault and then invaded the Haizi coal mine from north to south. The intrusive magma continued to invade fault fracture zones connected with the Daliujia Fault. Finally, the magma invaded the coal-bearing strata along the coal seam roof and floor and even engulfed coal units such as the No. 5 coal seam. The igneous intrusions in the Haizi mine primarily include dikes and sills; igneous intrusions in sills are the most common. The maximum thickness of the igneous sill is approximately 170 m, the sill length along its strike is 6.5 km, and the average thickness of the sill is 120 m. This sill is the most stable in the II102 mining area. A histogram of coal-bearing strata in the II102 mining area of the Haizi mine is shown in Figure 2B, wherein the Nos. 7, 8, 9, and 10 seams are the primary mineable coal beds in the Permian strata. The igneous intrusions are dis-
tributed as a sill over the coal seams. This super-thick igneous sill occurs above and nearly parallel to the minable coal seams. It maintains great influence on the metamorphism and gas occurrence of the underlying coal seams. The igneous rock samples are shown in Figure 2C. They are characterized by a hard and complete structure with a porphyritic texture and light gray to grayishgreen appearance. The igneous rocks of the Haizi mine are neutral, being chiefly composed of diorite and diorite porphyrite (Wang et al., 2014b). The magmatic activity in the Haizi mine area is interpreted to be related to the first phase of Yanshan magmatic activity. EXPERIMENTAL WORK Sampling and Basic Physical Parameters of Samples Beneath the sill, the underlying coal seams are characterized by high gas pressure and gas content, which pose an outburst hazard. In addition, compared with the gas content of the No. 10 seam, the Nos. 7, 8, and 9
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
193
Chen, Cheng, Zhou, Wang, Tian, and Jin
Figure 3. Cross sections showing the sampling locations.
seams have greater value and also exceed the outburst prediction critical value of 8 m3 /t (Cheng et al., 2010). As shown in Figure 3, four samples were taken from the Nos. 7, 9, and 10 seams in the mining area with the sill covering (where the sill thickness is approximately 140 m), which are located in the outburst risk region. A sample was collected from a non-outburst risk region of the No. 10 seam in the mining area without the sill covering for comparison. There are significant differences in the basic physical properties among the five samples. The physical parameters of the samples from the corresponding coal seams are listed in Table 1. The proximate analysis roughly reflects the quantities of inorganic and organic matter in the samples, which was performed by using an automatic proximate analyzer following the ISO 17246:2010 standard. The vitrinite reflectance mainly reflects the degree of coal metamorphism, which was determined by following the ISO 7404-5:2009 standard. The Langmuir constant VL reflects the adsorption capacity of coal for gas, which was determined by following the GB/T 19560-2008 standard. The initial velocity of methane diffusion ( p) was determined by following the AQ1080-2009 standard. As shown in Table 1, the maximum vitrinite reflectance (Ro, max ) beneath the igneous sill varies from 1.25 percent (HZ4) to 2.74 percent (HZ1), and the ash content (Aad ) ranges from 8.17 percent to 26.46 percent, with the higher values being closer to the igneous sill. The volatile matter (Vdaf ) ranges from 8.92 percent (HZ1) to 13.5 percent (HZ4), with the higher values being farther away from the sill. Sample HZ1,
194
which is closest to the sill, has an abnormally high moisture value (Mad , 2.02 percent). This high value is likely caused by the igneous sill creating a sealing effect that traps the moisture from the thermal metamorphism of coal. Figure 4 illustrates the trends in both the maximum vitrinite reflectance (Ro, max ) and the Langmuir constant VL of coal, which tend to increase closer to the sill. The sample closest to the sill, HZ1, has the largest metamorphic grade (Ro, max = 2.74 percent) and methane adsorption capacity (VL = 45.12 cm3 /g), whereas sample HZ5, located in an area without sill covering, has the lowest values (Ro, max = 0.66 percent and VL = 17.80 cm3 /g). Therefore, the thermal evolution effect of the igneous sill significantly increases the metamorphic grade of coal close to the igneous intrusions. The methane adsorption capacities are also considerably enhanced, which are related to the influence of the thermal evolution effect on the development of coal porosity.
Experimental Methods The physical gas adsorption method (using N2 as the probe molecule) is used extensively for analyzing pore characteristics of porous substances (Nie et al., 2015; Okolo et al., 2015). N2 adsorption/desorption isotherms at 77 K were obtained using a Quantachrome Autosorb-iQ2 for particle sizes in the range of 0.2–0.25 mm. The relative pressure (P/P0 ) was
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
Thermal Effects of Igneous Rock on Coal Table 1. Summary of physical parameters of analyzed samples. Proximate Analyses (%) Sample HZ1 HZ2 HZ3 HZ4 HZ5
Langmuir Constant
Coal Seam
Distance from Sill (m)
Elevation (m)
Ro, max (%)
Mad
Aad
Vdaf
VL (cm3 /g)
PL (MPa)
p (mm Hg)
7 9 10 10 10
55 70 152 156 Without sill covering
−660 −680 −635 −618 −515
2.74 2.20 1.72 1.25 0.66
2.02 0.54 1.29 1.28 0.68
26.46 19.18 11.29 8.17 11.9
8.92 12.56 13.13 13.5 21.63
45.12 37.09 34.97 28.34 17.80
1.09 1.12 0.86 1.04 1.19
45.0 31.8 17.2 12.6 5.8
observed over the range from 0.001 to 0.995. Pore parameters were automatically calculated via computer software named ASIQwinTM . The BrunauerEmmett-Teller (BET), Barrett-Joyner-Halenda (BJH), and quenched solid state density functional theory (QSDFT) methods were used for determining the pore volume, specific surface area (SSA), and pore size distribution (PSD). A detailed discussion of these methods can be obtained from the publication of Lowell et al. (2012). Prior to the experiments, the coal samples were placed in a vacuum drying oven and then dried at 60◦ C for at least 24 hours. The bulk desorption method (Zhang, 2008) was used to obtain the methane desorption data (methane desorbed volume versus time) of coal with particle sizes of 0.5–1 mm and 1–3 mm. Each sample was weighed at 50 g and then placed in a coal sample jar. The gas tightness of all coal sample jars was verified. Next, the jars were placed in a thermostatic water bath at 333 K and then evacuated using a vacuum pump for
24 hours. Subsequently, the jars were filled with pure methane and then placed in a thermostatic water bath at 303 K for at least 48 hours. To obtain the required gas pressures of 2 MPa, excessive free gas was released during the methane adsorption equilibrium process. Once the gauge pressure of the coal sample jars remained constant, the methane desorption tests were started. A stopwatch was used to record the time to methane desorption, which occurred when the gas within each of the jars was simultaneously and instantaneously relieved. Methane desorption data were recorded for a period of 2 hours. A detailed description of the methane desorption tests also can be found in the publication of Liu et al. (2015). Desorption Equation Many gas desorption models and equations have been proposed and employed to calculate √ coal seam methane contents. The equation Qt = K1 t, which
Figure 4. Variations in multiple physical parameters of the coal samples with distance from the igneous sill.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
195
Chen, Cheng, Zhou, Wang, Tian, and Jin
is also called the Barrer-type equation, is the most common equation used to estimate the initial lost-gas volume (Diamond and Schatzel, 1998). However, the equation is not well suited for data in long gas desorption tests. The Airey-type and Winter-type equations are commonly used to study the gas desorption law of pulverized coal. By assuming gas desorption from a solid exhibiting a cracked structure, the gas desorbed volume discharged at time t can be found by Eq. 1 (Airey, 1968): n t Qt = A 1 − exp − (1) , t0 where Qt is the gas desorption volume at time t, mL/g; A is the ultimate desorption volume, mL/g; t0 is the desorption time constant; and n is a coefficient. In Winter’s theory, the desorption rate of methane from coal grains versus time can be described and fitted by Eq. 2 (Winter and Janas, 1975): −kt t Vt = Va . (2) ta After mathematical integration, the following equation of gas desorption volume and time is obtained: Qt =
V1 1−kt t , 1 − kt
(3)
where Qt is the gas desorption volume at time t, mL/g; V1 and Va are the rates of gas desorption at times t1 and ta , respectively, mL/(g·min); and kt is a constant that reflects the degree of attenuation of the desorption rate and characterizes the gas desorption law (Banerjee, 1988). Diffusion Equation According to Fick’s law and by assuming a spherical and homogeneous solid with a constant radius and smooth surface, the unipore model is derived as the following equation (Crank, 1979): ∞ 6 1 D 2 2 Mt =1− 2 exp − 2 n t , (4) M∞ n2 r n=1
where Mt is the total amount of diffusion media through the sphere within time t; M∞ is the total desorbed quantity; D is the diffusion coefficient; and r is the mean radius of the sphere particles. De can be obtained using the equation De = D/r2 (Marecka and Mianowski, 1998), where De is the effective diffusivity. 196
Equation 4 can be rewritten as the following equation: ∞ 6 1 Qt =1− 2 exp(−De n 2 2 t), Q∞ n2
(5)
n=1
where Qt is the total volume of desorbed gas diffusing through the coal particles within time t, mL/g; and Q∞ is the total volume of gas desorbed, mL/g. Q∞ can be determined from experimental methods (Chen and Cheng, 2015; Lu et al., 2015) and can also be determined by Eq. 1, being equal to the ultimate desorption volume A in Eq. 1. RESULTS Pore Analysis by the N2 Adsorption Method N2 Adsorption and Desorption Isotherms and Pore Shapes N2 adsorption and desorption isotherms for the five coal samples are illustrated in Figure 5. All of the adsorption isotherms are type II, exhibiting multi-layer adsorption. Sample HZ1, closest to the igneous sill, adsorbed the most N2 , while sample HZ5, without the sill covering, adsorbed the least N2 at the highest relative pressure, suggesting that sample HZ1 has a larger micro-porosity and adsorption capacity. Six characteristic types of hysteresis loops are provided by the International Union of Pure and Applied Chemistry (IUPAC) (Thommes et al., 2015), which are based on the original IUPAC classification of 1985. As shown in Figure 5, except for sample HZ1, the plots demonstrating the change in volume with respect to the P/P0 ratio for adsorption and desorption are essentially coincident for all the samples when the relative pressure (P/P0 ) is less than 0.45, suggesting that the smaller pores are mainly in the shapes of pores accessible via a single pore throat (Zhang et al., 2013). When the P/P0 exceeds 0.5, a distinct hysteresis loop can be observed in the isotherms of samples HZ1 and HZ2, indicating the presence of a large number of open pores and corresponding to better pore connectivity in the two samples. This conspicuous hysteresis is ascribed to the difference between condensation and evaporation processes in the coal pores (Mastalerz et al., 2012; Yang et al., 2014). According to the classification of IUPAC, the hysteresis loops of samples HZ1 and HZ2 belong to type H3 and may indicate the presence of numerous slit-shaped pores, the existence of which is thought to be a significant internal factor for causing coal and gas outburst hazards (Jiang et al., 2011). The desorption branch of sample HZ2 shows a sudden drop in the volume with a forced closure of the hysteresis loop at a relative pressure of approximately 0.45
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
Thermal Effects of Igneous Rock on Coal
Figure 5. N2 adsorption/desorption isotherms of coal samples at 77 K.
because of the so-called tensile strength effect (TSE) (Groen et al., 2003; Lowell et al., 2012). An abrupt change in volume also exists in the desorption branch of sample HZ1. However, the hysteresis loop is still observed at a relative pressure of less than 0.45 for sample HZ1, which is more likely to be the result of adsorption swelling of the coal (Kondo et al., 2005; Cai et al., 2013). The adsorption and desorption branches are essentially coincident for samples HZ3, HZ4, and HZ5, suggesting the existence of many semi-closed pores that may, for instance, include cylindrical and slit-shaped pores with one closed side in the coals, indicating poor pore connectivity. This phenomenon is in accordance with the research results in the publication of Nie et al. (2015). Pore Volume and Specific Surface Area The pore volume and SSA of the coal samples from the nitrogen adsorption tests at 77 K were calculated using the BET, BJH, and QSDFT methods, as shown in Table 2. The pore volume and SSA of sample HZ1 closest to the sill are the largest, whereas those of the
unaffected coal sample HZ5 are the smallest. On the basis of the IUPAC classification (Thommes et al., 2015), pores are divided into the following three categories: macro-pores (≥50 nm), meso-pores (2–50 nm) and micro-pores (≤2 nm). A micro-pore analysis using the CO2 adsorption method for the No. 9 and No. 10 coal seams obtained from sample locations similar to ours can be obtained from the publication of Wang et al. (2014a). Figure 6 illustrates a dramatic decrease in both the pore volume and SSA of micro-pores, mesopores, and macro-pores for the samples closer to the sill. Sample HZ5 has the lowest values of SSA and pore volume. In the study of Wang et al. (2014a), numerous thermally metamorphic pores in the coal samples closer to the extremely thick sill were depicted using a scanning electron photomicrograph; in contrast, only a few such pores were found in the samples away from the sill, and no pores were found in the samples of the unaffected coals. Research shows that thermally metamorphic pores are formed during the devolatilization of liptinite and vitrinite (Sarana and Kar, 2011). Therefore, under the thermal effect of the intrusions, both
Table 2. Pore volume and specific surface area of coal samples from nitrogen adsorption at 77 K. Sample No. HZ1 HZ2 HZ3 HZ4 HZ5 a b
BET SSA (m2 /g)
BJH SSA (m2 /g)
BJH Pore Volumea ( × 10−3 cm3 /g)
QSDFT Volumeb SSAb (m2 /g)
QSDFT Pore ( × 10−3 cm3 /g)
QSDFT Pore Diameterb (nm)
2.054 1.015 0.522 0.423 0.405
1.440 0.595 0.412 0.303 0.295
10.200 5.609 5.227 4.830 3.048
1.764 0.798 0.401 0.308 0.279
3.894 2.000 1.523 1.042 0.896
1.193 1.144 1.682 1.756 4.077
Pore diameter ranges from 3 to 300 nm. Pore diameter ranges from 0.9 to 35 nm.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
197
Chen, Cheng, Zhou, Wang, Tian, and Jin
Figure 6. Changes in specific surface area (SSA), and pore volume of coal samples with the distance from the sill. Micro-pore SSA and volume (A and B) were calculated using the Non-Local Density Functional Theory (NLDFT) model, where the pore diameter ranged from 0.35 to 1.50 nm (Wang et al., 2014a). Sample HZ2-1 is from the No. 9 coal seam, samples HZ3-1, HZ4-1, an HZ5-1 are from the No. 10 coal seam, and sample HZ5-1 is from the unaffected area. The meso-pore SSA and volume were calculated using the BJH model (C and D). The macro-pore (50–300 nm) SSA and volume were calculated using the BJH model (E and F). The gray zone represents the area with sill coverage.
the pore volume and SSA increase dramatically in the coal samples approaching the sill. The prominent enhancement of micro-pore volume and SSA leads to an increase in the gas adsorption and storage capacities, which can increase the risk of coal and gas outbursts during mining activity in the coal seams closer to the sill.
198
Pore Size Distributions Plots of the PSDs (Figure 7) show that uni-modal or multi-modal PSDs exist in the range of 1–100 nm for all coal samples. The QSDFT dV(d) plots (Figure 7A) suggest a bi-modality for sample HZ1 (peaks at ∼1 nm and ∼4 nm), whereas the BJH dV(d) plots (Figure 7B)
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
Thermal Effects of Igneous Rock on Coal
Figure 7. Pore size distributions (PSDs) using QSDFT (A) and BJH (B).
suggest uni-modality with a peak at <4 nm. Sample HZ2 appears to exhibit bi-modality, with peaks at ∼1 nm and ∼4 nm (Figure 7A) and peaks at ∼3 nm and ∼5 nm (Figure 7B). Both the QSDFT dV(d) and the BJH dV(d) plots reveal multi-modality for samples HZ3, HZ4, and HZ5. The major peaks exist in pores <5 nm
for samples HZ1 and HZ2, suggesting the enhancement of the SSA and volume of micro-pores and some mesopores. This result indicates a prominent increase in the gas adsorption and storage capacities of coals closer to the igneous sill. The pore volumes of samples HZ3 and HZ4 both exhibit relatively small increases in the
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
199
Chen, Cheng, Zhou, Wang, Tian, and Jin
Figure 8. Winter-type equation fitting curves of methane desorption data for the coal particles in the ranges of 1–3 mm (A) and 0.5–1 mm (B), and Airey-type equation fitting curves of methane desorption data for the coal particles in the ranges of 1–3 mm (C) and 0.5–1 mm (D).
range of 1–30 nm, indicating that the thermal effect has a weaker effect on pore development of coal farther away from the sill. Samples HZ5, HZ3, and HZ4 show similar PSD behaviors. Methane Desorption Index and Effective Diffusivity Fitted curves of methane desorption data for the coal particles (1–3 mm and 0.5–1 mm) using Eq. 1 and Eq. 3 are shown in Figure 8. The results are summarized in Table 3. Both the Airey-type and the Winter-type equations have high values of the correlation coefficient R2 , suggesting a high-fitting precision. For the samples with particle sizes in the range of 1–3 mm, the ultimate desorption volume A varies from 4.50 to 15.30 cm3 /g, the initial desorption rate V1 ranges from 0.088 to 1.103 cm3 /(min·g), and the initial desorption volume K1 varies from 0.194 to 3.679 cm3 /g. Sample HZ1 closest to the sill shows the largest initial desorption volume and rate (19.0 times and 12.5 times, respectively, greater than that of sample HZ5). In addition, the smaller-sized samples have larger initial values for the volume and rate of desorption, likely because of the heterogeneity of coal. The results of the effective diffusivity of the coal samples calculated using Eq. 5 are listed in Table 4. For the coal samples with particle sizes in the range of 1–3 mm, 200
the effective diffusivity De varies from 1.980 × 10−6 to 4.626 × 10−5 s−1 , and the effective diffusivity of sample HZ1 is 12.53 times greater than that of sample HZ5. Therefore, both the methane desorption amounts and effective diffusivity of the samples increase significantly closer to the sill. The enhancements in the gas desorption and diffusion properties of the coals are likely due to the significant influence of the thermal effect of the sill on the pore structure of the coal. DISCUSSION Change in Methane Effective Diffusivity On the basis of the Knudsen number, the gas diffusion within a coal pore system is primarily divided into Fick diffusion, Knudsen diffusion, and transitional diffusion (Li et al., 2012). Note that two other diffusion types, surface diffusion and crystal diffusion, are not often considered in coalbed methane studies (Nie et al., 2000). The Knudsen number (Kn ) is expressed as the ratio of the pore diameter (d) of the coal matrix to the mean molecule free path (). The mean molecule free path is represented by the following equation: KT , = √ 2d02 p
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
(6)
Thermal Effects of Igneous Rock on Coal Table 3. Fitting results for the methane desorption indexes. Airey-Type Equation Sample HZ1 HZ2 HZ3 HZ4 HZ5 a
3
Winter-Type Equation 2
Particle Size
A (cm /g)
t0 (min)
n
R
1–3 mm 0.5–1 mm 1–3 mm 0.5–1 mm 1–3 mm 0.5–1 mm 1–3 mm 0.5–1 mm 1–3 mm 0.5–1 mm
15.30 16.16 8.39 10.07 7.94 10.12 7.89 9.89 4.50 5.08
18.37 8.50 39.05 40 83 64.41 350 260.1 834 640.9
0.420 0.427 0.543 0.491 0.469 0.475 0.498 0.467 0.460 0.438
0.9991 0.9960 0.9996 1 0.9980 1 0.9993 1 0.9996 0.9998
3
V1 [cm /(min·g)]
kt
R2
K1 a (cm3 /g)
1.103 1.245 0.499 0.611 0.383 0.530 0.201 0.309 0.088 0.121
0.769 0.820 0.666 0.687 0.660 0.670 0.585 0.615 0.591 0.618
0.9774 0.9472 0.9885 0.9883 0.9938 0.9938 0.9982 0.9985 0.9992 0.9992
3.679 5.414 1.115 1.443 0.902 1.276 0.387 0.705 0.194 0.296
K1 represents the gas desorption volume in the first minute, and the K1 value is the measured result rather than a fitted value.
where K is the Boltzmann constant, 1.38 × 10−23 J/K; T is the absolute temperature, K; d0 is the effective diameter of gas molecules, nm; and p is the gas absolute pressure, MPa. When Kn is larger than 10, collisions are predominantly inter-molecular collisions between gas molecules due to the pore diameter being far greater than the mean molecule free path. In this case, Fick diffusion is the dominant diffusion mechanism. When Kn is larger than 0.1, collisions are primarily between gas molecules and coal pore walls because the pore diameter is much less than the mean molecule free path. Thus, Knudsen diffusion is the main gas diffusion mechanism. When Kn is between 0.1 and 10, collisions between gas molecules and collisions between gas molecules and pore walls are of equal importance. As such, transitional diffusion is the dominant diffusion mechanism. Considering Eq. 6, when the experiment temperature is 30◦ C (303.15 K), the mean molecule free path for methane is 86.5 nm in the case when the pressure of the coal sample jar is instantaneously relieved (p = 0.1 MPa). Based upon previous studies, it is known that the effective distance of a coal surface interacting with a methane molecule is approximately 0.55 nm, correTable 4. Results of effective diffusivity using the unipore model. Sample HZ1 HZ2 HZ3 HZ4 HZ5
Particle Size (mm) 1–3 0.5–1 1–3 0.5–1 1–3 0.5–1 1–3 0.5–1 1–3 0.5–1
De (s−1 ) 4.626 9.205 2.268 2.344 1.313 1.591 3.482 4.976 1.980 2.623
× × × × × × × × × ×
10−5 10−5 10−5 10−5 10−5 10−5 10−6 10−6 10−6 10−6
R2 0.8911 0.8901 0.9896 0.9699 0.9677 0.9689 0.9956 0.9836 0.9700 0.9753
sponding to a pore diameter of approximately 1.10 nm and a distance of the adsorption potential well of approximately 0.36 nm (Jiang et al., 2006). Additionally, the effective diameter of the methane molecule is approximately 0.33 nm. Hence, owing to the influence of the coal surface, surface diffusion occurs in the coal pores when the diameter ranges from 0.33 to 1.1 nm (Xu et al., 2015). As illustrated in Figure 9, Knudsen diffusion is the chief diffusion type in pores with a size range of 1.1– 8.65 nm, whereas transitional diffusion occurs in pores with a size range of 8.65–865 nm, and Fick diffusion occurs in pores when the pore diameter exceeds 865 nm. A pore diameter of 100 nm is generally considered to be a cutoff point for gas diffusion and seepage in coal by many scholars (Hodot, 1966). Given this cutoff point and regardless of surface diffusion, pores with a size range of 1.1–100 nm may represent the primary spaces for methane diffusion. Hence, under the experimental conditions (i.e., pressure is completely relieved, p is 0.1 MPa, and T is 303.15 K), the main diffusion modes are Knudsen diffusion and transitional diffusion. The relationship between the effective diffusivity (De ) and a volume of pores with a size range of 1.1–100 nm is shown in Figure 10. As illustrated in Figure 10A, a pore volume with a size range from 1.1 to 100 nm can be calculated by the sum of the volume of pores in the size range of 1.1–8.65 nm (QSDFT) and 8.65–100 nm (BJH). Sample HZ1, having a size range of 1.1–100 nm, has the largest pore volume and effective diffusivity (5.442 × 10−3 cm3 /g and 46.26 × 10−6 s−1 , respectively), while sample HZ5 has the smallest (1.317 × 10−3 cm3 /g and 1.98 × 10−6 s−1 ). The effective diffusivity and pore volume within the size range of 1.1–100 nm increase progressively closer to the sill. Figure 10B exhibits a positive correlation between the pore volume with a size range of 1.1–100 nm and methane effective diffusivity.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
201
Chen, Cheng, Zhou, Wang, Tian, and Jin
Figure 9. The diffusion modes in the corresponding pore diameter range under the experimental conditions (p = 0.1 MPa and T = 303.15 K).
Relationships between Methane Desorption Index and Pore Characteristics The igneous sill provides a high-temperature environment that promotes the thermo-metamorphism of coal and the formation of coal-bed gas (Wang et al., 2014c). Coupled with the entrapment effect of the extremely thick sill, the amount of gas released is reduced while large amounts of gas are stored in the affected coal seams. The presence of a larger number of micropores in coal can significantly increase the gas adsorption capacity of coal. Thus, more gas can be released at the moment of pressure relief under the same circumstances. More importantly, well-developed meso-pores and macro-pores can provide broad pore channels for gas diffusion and seepage; on the other hand, better pore connectivity favors gas migration from the coal matrix to fractures. Hence, the impact of the thermal effect of the sill on the pore structure of coal increases the risk of outburst accidents during mining of the coal seam. During the history of the Haizi mine, coal and gas outbursts have occurred five times in the No. 7 seam, once within the No. 9 seam, and once within the No. 10 seam under the extremely thick sill. Experience demonstrates that the No. 7 seam closest to the igneous sill has a greater risk of coal and gas outburst. The indexes K1 , p, and kt are effective parameters with which to predict the risk of coal and gas outbursts (Du, 1985; Cheng et al., 2010). Figure 11A shows the variations in the desorption indexes K1 , p, and kt
of the samples (1–3 mm) relative to distance from the igneous sill. The K1 index varies from 0.194 to 3.679 cm3 /g, the p index ranges from 5.8 to 45.0 mm Hg, and the kt index ranges from 0.585 to 0.769, with the higher values observed closer to the igneous sill. There is an increasing trend in the values of the K1 and p indexes as the distance to the sill decreases. Hence, due to the thermal evolution effect of the sill, larger SSA and pore volumes as well as better pore connectivity are found in coals closer to the sill, which leads to a greater initial gas desorption volume and rate (K1 and p) than in coals situated farther away from the sill or in unaffected coals. Figure 11B indicates that the kt index is positively related with macro-pore volume because kt is considered to be the desorption rate attenuation index characterizing the ratio between the gas desorption volume in the macro-pores and the micro-fractures in the first minute after the gas pressure is relieved (Guo et al., 2014). Note that sample HZ5 (No. 10 seam) from the non-outburst risk region and sample HZ4 from the outburst risk region (No. 10 seam) have similar values of kt , likely because the kt index is independent of the gas pressure in the coal seam (Du, 1985).
Effects of the Igneous Sill on Coal Seam Gas Occurrence in the Haizi Mine Figure 12A shows the distribution of igneous sills in the II101 and II102 mining areas. The thermal
Figure 10. Variations of the effective diffusivity and the total volume of the pores in the ranges of 1.1–8.65 nm (QSDFT) and 8.65–100 nm (BJH) for the coal samples (A), and the relationship between the pore volume in size range of 1.1–100 nm and the methane effective diffusivity (B).
202
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
Thermal Effects of Igneous Rock on Coal
Figure 11. Variation in the methane desorption indexes of the samples (1â&#x20AC;&#x201C;3 mm) with distance from the sill (A) and the relationship between the Kt index and the BJH macro-pore volume (B).
impact from igneous intrusions is greater for the II102 mining area, which possesses a substantially thicker sill, compared to the II101 mining area, which has few sill coverings. The measured results of gas content and gas pressure of the No. 10 coal seam in the
two mining areas are shown in Figure 12B and Figure 12C, respectively. The results show that the No. 10 coal seam in the II102 mining area has a relatively higher coal seam gas content and greater gas pressure gradient (0.0374 MPa/m). Figure 12D indicates that more gas is
Figure 12. (A) Map of igneous intrusion distribution in the Haizi coal mine. (B) Relationship between coal seam gas pressure and burial depth of the No. 10 coal seam in the II101 and II102 mining areas. (C) Relationship between coal seam gas content and burial depth of the No. 10 coal seam in the II101 and II102 mining areas. (D) Amount of absolute gas emission of different working faces in the II101 and II102 mining areas.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191â&#x20AC;&#x201C;207
203
Chen, Cheng, Zhou, Wang, Tian, and Jin
Figure 13. Sketched map of the effects of the extremely thick igneous sill on coal seam gas occurrence and gas hazards in the Haizi coal mine.
released from the working face in the II102 mining area than is released in the II101 mining area during actual mining. Hence, the comparison between these parameters for the II101 and II102 mining areas shows that the No. 10 coal seam in the II102 mining area carries a greater outburst risk. A coal and gas outburst accident occurred on the II1026 working face of the II102 mining area on April 25, 2009. The outburst volumes of coal and gas were 656 t and 13,210 m3 , respectively. No outburst accident occurred in the II101 mining area. The influence of the igneous sill on underlying coal seam gas occurrence can be summarized as follows. First, as a result of the thermal evolution of the igneous sill, the well-developed pores of coal closer to the sill favor the enhancement of adsorption and desorption capacities. The initial released gas possesses a larger expansion energy after the pressure is released, making it prone to cause coal and gas outbursts (Hu and Wen, 2013). Second, owing to the thermal effect of the igneous sill, the affected coal seams are characterized by relatively high metamorphic grade and gas production, such that their gas content and pressure are larger than those coal seams without sill covering. Third, the igneous sill can restrain coal seam gas emissions and thus exert a well trap effect on coal seam gas due to its low permeability and great thickness. Hence, the igneous sill has significant influence on the generation and storage of coal seam gas in the underlying coal seams, thus significantly increasing the risk of gas hazards in the Haizi coal mine. Additionally, the jostle effect of the intruded igneous rock on the underlying coal seams leads to the development of occasional tectonically deformed coal layers
204
(Zhang et al., 2015). Rupture of the overlying thick and hard sill can also result in pronounced mining-induced stress during coal seam extraction, which plays an important role in the occurrence of coal and rock outbursts (Wang et al., 2013). In this case, the combined effects of thermal evolution and entrapment make gas hazards more likely to occur in the underlying coal seams as a consequence of the effects of intrusive jostling and mining-induced stress (Figure 13). CONCLUSIONS 1) The metamorphic grade and methane adsorption capacity of the coal samples increase closer to the sill. Hence, the thermal effect of the igneous sill significantly promotes metamorphism of the affected coal. 2) Owing to the thermal effect of the sill, the pore volume and specific surface area of the coal samples increase progressively closer to the sill, and the coal samples present different pore shapes and pore size distributions. The hysteresis loops suggest the presence of a considerable number of slit-shaped pores in coal samples HZ1 and HZ2 (No. 7 seam and No. 9 seam), which are closer to the sill, indicating better pore connectivity, whereas many semi-closed pores exist in samples HZ3 and HZ4 (situated farther away from the sill) and in sample HZ5 (absent sill covering) of the No. 10 seam, indicating poor connectivity. In addition, the PSDs show that samples HZ1 and HZ2 exhibit a prominent enhancement of the volume of micro-pores and some meso-pores (<5 nm), suggesting significant enhancement of the gas
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
Thermal Effects of Igneous Rock on Coal
adsorption and storage capacities of the coal closer to the igneous sill. 3) Under the experimental conditions, effective diffusivity increases closer to the sill and is highly correlated with pore volumes in the size range of 1.1 to 100 nm. In addition, the values of the methane desorption indexes also tend to increase in the coals approaching the sill. The thermal effect of the sill promotes the development of pores in the affected coal, thereby leading to better pore connectivity, enhanced gas adsorption capacity, greater quantities of desorbed gas, and much higher gas desorption and diffusion rates. Consequently, the thermal influence of the igneous sill changes the characteristics of coal seam gas occurrence as a result of the entrapment effect of the extremely thick sill, making gas hazards a recurring problem in the affected coal seams.
ACKNOWLEDGMENTS This work was supported by the Fundamental Research Funds for the Central Universities (No. 2015XKMS008), the National Science Foundation of China (No. 51574229), Jiangsu Overseas Research & Training Program for University Prominent Young & Middle-Aged Teachers and Presidents, and A Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD). We appreciate the helpful comments from the editors and the reviewers.
REFERENCES AIREY, E. M., 1968, Gas emission from broken coal. An experimental and theoretical investigation: International Journal Rock Mechanics Mining Sciences & Geomechanics Abstracts, Vol. 5, pp. 475–494. AN, F. H.; CHENG, Y. P.; WU, D. M.; AND WANG, L., 2013, The effect of small micropores on methane adsorption of coals from northern China: Adsorption, Vol. 19, pp. 83–90. BANERJEE, B., 1988, Spacing of fissuring network and rate of desorption of methane from coals: Fuel, Vol. 67, pp. 1584–1586. CAI, Y. D.; LIU, D. M.; PAN, Z. J.; YAO, Y. B.; LI, J. Q.; AND QIU, Y. K., 2013, Pore structure and its impact on CH4 adsorption capacity and flow capability of bituminous and subbituminous coals from northeast China: Fuel, Vol. 103, pp. 258–268. CHEN, H. D.; JIANG, J. Y.; CHEN, X. J.; AND XU, C., 2014, Differences in coal bed methane occurrence for different regions of igneous erosion in the Haizi coal mine, Huaibei coalfield, China: Journal Natural Gas Science Engineering, Vol. 21, pp. 732–737. CHEN, J.; LIU, G. J.; LI, H.; AND WU, B., 2014, Mineralogical and geochemical responses of coal to igneous intrusion in the Pansan coal mine of the Huainan coalfield, Anhui, China: International Journal Coal Geology, Vol. 124, pp. 11–35. CHEN, X. J. AND CHENG, Y. P., 2015, Influence of the injected water
on gas outburst disasters in coal mine: Natural Hazards, Vol. 76, pp. 1093–1109. CHENG, L. B.; WANG, L.; CHENG, Y. P.; JIN, K.; ZHAO, W.; AND SUN, L. S., 2016, Gas desorption index of drill cuttings affected by magmatic sills for predicting outbursts in coal seams: Arabian Journal Geosciences, Vol. 9, pp. 1–15. CHENG, Y. P.; WANG, H. F.; WANG, L.; ZHOU, H. X.; LIU, H. B.; WU, D. M.; AND LI, W., 2010, Theories and Engineering Applications on Coal Mine Gas Control: China University of Mining and Technology Press, Xuzhou City, China, 840 p. COOPER, J. R.; CRELLING, J. C.; RIMMER, S. M.; AND WHITTINGTON, A. G., 2007, Coal metamorphism by igneous intrusion in the Raton Basin, CO and NM: Implications for generation of volatiles: International Journal Coal Geology, Vol. 71, pp. 15–27. CRANK, J., 1979, The Mathematics of Diffusion: Oxford University Press, London, U.K. 414 p. DAI, S. F. AND REN, D. Y., 2007, Effects of magmatic intrusion on mineralogy and geochemistry of coals from the FengfengHandan coalfield, Hebei, China: Energy & Fuels, Vol. 21, pp. 1663–1673. DIAMOND, W. P. AND SCHATZEL, S. J., 1998, Measuring the gas content of coal: A review: International Journal Coal Geology, Vol. 35, pp. 311–331. DU, B., 1985, The initial gas desorption rate as an index of burst of coal and gas: About the parameter of Kt (Chinese translation): Safety Coal Mines, Vol. 5, pp. 56–63. FINKELMAN, R. B.; BOSTICK, N. H.; DULONG, F. T.; SENFTLE, F. E.; AND THORPE, A. N., 1998, Influence of an igneous intrusion on the inorganic geochemistry of a bituminous coal from Pitkin County, Colorado: International Journal Coal Geology, Vol. 36, pp. 223–241. GOLAB, A. N. AND CARR, P. F., 2004, Changes in geochemistry and mineralogy of thermally altered coal, Upper Hunter Valley, Australia: International Journal Coal Geology, Vol. 57, pp. 197– 210. GROEN, J. C.; PEFFER, L. A.; AND P´EREZ-RAM´IREZ, J., 2003, Pore size determination in modified micro- and mesoporous materials. Pitfalls and limitations in gas adsorption data analysis: Microporous Mesoporous Materials, Vol. 60, pp. 1–17. GUO, J. Q.; KANG, T. H.; KANG, J. T.; ZHAO, G. F.; AND HUANG, Z. M., 2014, Effect of the lump size on methane desorption from anthracite: Journal Natural Gas Science Engineering, Vol. 20, pp. 337–346. HAN, S. F., 1990, Geological Condition and Prediction of Coal Accumulation in Huainan and Huaibei Coalfields: Geological Publishing House, Beijing, 234 p. HODOT, B., 1966, Coal and Gas Outbursts (Chinese translation): China Coal Industry Press, Beijing, 317 p. HU, Q.T. AND WEN, G.C., 2013, Mechanical Mechanism of Coal and Gas Outburst: Science Press, Beijing, 473 p. JIAN, K.; LEI, D. J.; FU, X. H.; ZHANG, Y. G.; AND LI, H. L., 2015, Effect of an electrostatic field on gas adsorption and diffusion in tectonic coal: International Journal Mining Science and Technology, Vol. 25, pp. 607–613. JIANG, J. Y.; CHENG, Y. P.; WANG, L.; LI, W.; AND WANG, L., 2011, Petrographic and geochemical effects of sill intrusions on coal and their implications for gas outbursts in the Wolonghu mine, Huaibei coalfield, China: International Journal Coal Geology, Vol. 88, pp. 55–66. JIANG, W. P.; CUI, Y. J.; ZHANG, Q.; AND LI, Y. H., 2006, The quantum chemical study on the coal surface interacting with CH4 and CO2 : Journal China Coal Society, Vol. 31, pp. 237– 240. JIANG, W. P.; SONG, X. Z.; AND ZHONG, L. W., 2011, Research on the pore properties of different coal body structure coals
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
205
Chen, Cheng, Zhou, Wang, Tian, and Jin and the effects on gas outburst based on the low-temperature nitrogen adsorption method: Journal China Coal Society, Vol. 36, pp. 609–614. KONDO, S.; ISHIKAWA, T.; AND ABE, I., 2005, Adsorption Science (Chinese translation): Chemical Industry, Beijing, 273 p. LI, X. C.; NIE, B. S.; ZHANG, R. M.; AND CHI, L. L., 2012, Experiment of gas diffusion and its diffusion mechanism in coal: International Journal Mining Science Technology, Vol. 22, pp. 885–889. LIU, H. H.; MOU, J. H.; AND CHENG, Y. P., 2015, Impact of pore structure on gas adsorption and diffusion dynamics for longflame coal: Journal Natural Gas Science Engineering, Vol. 22, pp. 203–213. LOWELL, S.; SHIELDS, J. E.; THOMAS, M. A.; AND THOMMES, M., 2012, Characterization of Porous Solids and Powders: Surface Area, Pore Size and Density: Springer Science & Business Media, New York, 339 p. LU, S. Q.; CHENG, Y. P.; QIN, L. M.; LI, W.; ZHOU, H. X.; AND GUO, H. J., 2015, Gas desorption characteristics of the high-rank intact coal and fractured coal: International Journal Mining Science Technology, Vol. 25, pp. 819–825. MARECKA, A. AND MIANOWSKI, A., 1998, Kinetics of CO2 and CH4 sorption on high rank coal at ambient temperatures: Fuel, Vol. 77, pp. 1691–1696. MASTALERZ, M.; DROBNIAK, A.; AND SCHIMMELMANN, A., 2009, Changes in optical properties, chemistry, and micropore and mesopore characteristics of bituminous coal at the contact with dikes in the Illinois Basin: International Journal Coal Geology, Vol. 77, pp. 310–319. MASTALERZ, M.; HE, L.; MELNICHENKO, Y. B.; AND RUPP, J. A., 2012, Porosity of coal and shale: Insights from gas adsorption and SANS/USANS techniques: Energy & Fuels, Vol. 26, pp. 5109–5120. NIE, B. S.; HE, X. Q.; AND WANG, E. Y., 2000, Diffusion mode of methane gas in coal pores: Mining Safety & Environment Protection, Vol. 27, pp. 14–16. NIE, B. S.; LIU, X. F.; YANG, L. L.; MENG, J. Q.; AND LI, X. C., 2015, Pore structure characterization of different rank coals using gas adsorption and scanning electron microscopy: Fuel, Vol. 158, pp. 908–917. OKOLO, G. N.; EVERSON, R. C.; NEOMAGUS, H. W.; ROBERTS, M. J.; AND SAKUROVS, R., 2015, Comparing the porosity and surface areas of coal as measured by gas adsorption, mercury intrusion and SAXS techniques: Fuel, Vol. 141, pp. 293–304. QU, Z. H.; JIANG, B.; WANG J. L.; AND LI, M., 2008, Characteristics of tectonic evolution and its controlling effects on coal and gas in Huaibei area: Coal Geology China, Vol. 20, pp. 34–37. RAHMAN, M. W. AND RIMMER, S. M., 2014, Effects of rapid thermal alteration on coal: Geochemical and petrographic signatures in the Springfield (No. 5) coal, Illinois Basin: International Journal Coal Geology, Vol. 131, pp. 214–226. RIMMER, S. M.; YOKSOULIAN, L. E.; AND HOWER, J. C., 2009, Anatomy of an intruded coal, I: Effect of contact metamorphism on whole-coal geochemistry, Springfield (No. 5) (Pennsylvanian) coal, Illinois Basin: International Journal Coal Geology, Vol. 79, pp. 74–82. SACHSENHOFER, R. F.; PRIVALOV, V. A.; AND PANOVA, E. A., 2012, Basin evolution and coal geology of the Donets Basin (Ukraine, Russia): An overview: International Journal Coal Geology, Vol. 89, pp. 26–40. SAGHAFI, A.; PINETOWN, K. L.; GROBLER, P. G.; AND VAN HEERDEN, J. H. P., 2008, CO2 storage potential of South African coals and gas entrapment enhancement due to igneous intrusions: International Journal Coal Geology, Vol. 73, pp. 74–87. SARANA, S. AND KAR, R., 2011, Effect of igneous intrusive on coal microconstituents: Study from an Indian Gondwana coalfield:
206
International Journal Coal Geology, Vol. 85, pp. 161–167. SHEPHERD, J.; RIXON, L. K.; AND GRIFFITHS, L., 1981, Outbursts and geological structures in coal mines: A review: International Journal Rock Mechanics Mining Sciences & Geomechanics Abstracts, Vol. 18, pp. 267–283. SKOCZYLAS, N.; DUTKA, B.; AND SOBCZYK, J., 2014, Mechanical and gaseous properties of coal briquettes in terms of outburst risk: Fuel, Vol. 134, pp. 45–52. STEWART, A.; MASSEY, M.; PADGETT, P.; RIMMER, S.; AND HOWER, J., 2005, Influence of a basic intrusion on the vitrinite reflectance and chemistry of the Springfield (No. 5) coal, Harrisburg, Illinois: International Journal Coal Geology, Vol. 63, pp. 58–67. THOMMES, M.; KANEKO, K.; NEIMARK, A. V.; OLIVIER, J. P.; RODRIGUEZ-REINOSO, F.; ROUQUEROL, J.; AND SING, K. S., 2015, Physisorption of gases, with special reference to the evaluation of surface area and pore size distribution (IUPAC Technical Report): Pure and Applied Chemistry, Vol. 87, pp. 1051– 1069. WANG, L.; CHENG, L. B.; CHENG, Y. P.; XU, C.; AND WANG, W., 2014a, Influence of thermal events of magma intrusion on coal seams metamorphic grade and adsorption and desorption characteristics: Journal China Coal Society, Vol. 39, pp. 1275–1282. WANG, L.; CHENG, L. B.; CHENG, Y. P.; YIN, G. Z.; CAI, C. C.; XU, C.; AND JIN, K., 2014b, Thermal effects of magmatic sills on coal seam metamorphism and gas occurrence: Bulletin Volcanology, Vol. 76, pp. 1–16. WANG, L.; CHENG, L. B.; CHENG, Y. P.; YIN, G. Z.; XU, C.; JIN, K.; AND YANG, Q. L., 2014c, Characteristics and evolutions of gas dynamic disaster under igneous intrusions and its control technologies: Journal Natural Gas Science Engineering, Vol. 18, pp. 164–174. WANG, L.; CHENG, Y. P.; AN, F. H.; ZHOU, H. X.; KONG, S. L.; AND WANG, W., 2014d, Characteristics of gas disaster in the Huaibei coalfield and its control and development technologies: Natural Hazards, Vol. 71, pp. 85–107. WANG, L.; CHENG, Y. P.; XU, C.; AN, F. H.; JIN, K.; AND ZHANG, X. L., 2013, The controlling effect of thick-hard igneous rock on pressure relief gas drainage and dynamic disasters in outburst coal seams: Natural Hazards, Vol. 66, pp. 1221–1241. WINTER, K. AND JANAS, H., 1975, Gas emission characteristics of coal and methods of determining the desorbable gas content by means of desorbometers.: In 16th International Conference on Coal Mine Safety Research, Sept. 22–26: Washington, D.C. WU, D.; LIU, G. J.; SUN, R. Y.; AND CHEN, S. C., 2014, Influences of magmatic intrusion on the macromolecular and pore structures of coal: Evidences from Raman spectroscopy and atomic force microscopy: Fuel, Vol. 119, pp. 191–201. XU, C.; CHENG, Y. P.; REN, T.; WANG, L.; KONG, S. L.; AND LU, S. Q., 2014, Gas ejection accident analysis in bed splitting under igneous sills and the associated control technologies: A case study in the Yangliu mine, Huaibei coalfield, China: Natural Hazards, Vol. 71, pp. 109–134. XU, H.; TANG, D. Z.; ZHAO, J. L.; LI, S.; AND TAO, S., 2015, A new laboratory method for accurate measurement of the methane diffusion coefficient and its influencing factors in the coal matrix: Fuel, Vol. 158, pp. 239–247. YANG, F.; NING, Z. F.; AND LIU, H. Q., 2014, Fractal characteristics of shales from a shale gas reservoir in the Sichuan Basin, China: Fuel, Vol. 115, pp. 378–384. YANG, M.; LIU, G. J.; SUN, R. Y.; CHOU, C. L.; AND ZHENG, L. G., 2012, Characterization of intrusive rocks and REE geochemistry of coals from the Zhuji coal mine, Huainan coalfield, Anhui, China: International Journal Coal Geology, Vol. 94, pp. 283–295.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
Thermal Effects of Igneous Rock on Coal YANG, Q.; WU, C. L.; AND TANG, D. Z., 1996, Coal Metamorphism in China: China Coal Industry Publishing House, Beijing, 212 p. YAO, Y. B. AND LIU, D. M., 2012, Effects of igneous intrusions on coal petrology, pore-fracture and coalbed methane characteristics in Hongyang, Handan and Huaibei coalfields, North China: International Journal Coal Geology, Vol. 96, pp. 72–81. YAO, Y. B.; LIU, D. M.; AND HUANG, W. H., 2011, Influences of igneous intrusions on coal rank, coal quality and adsorption capacity in Hongyang, Handan and Huaibei coalfields, North
China: International Journal Coal Geology, Vol. 88, pp. 135– 146. ZHANG, D. F.; GU, L. L.; LI, S. G.; LIAN, P. C.; AND TAO, J., 2013, Interactions of supercritical CO2 with coal: Energy & Fuels, Vol. 27, pp. 387–393. ZHANG, X. L.; CHENG, Y. P.; WANG, L.; AND ZHAO, W., 2015, Research on the controlling effects of a layered sill with different thicknesses on the underlying coal seam gas occurrence: Journal Natural Gas Science Engineering, Vol. 22, pp. 406–414. ZHANG, Y. X., 2008, Geochemical Kinetics: Princeton University Press, Princeton, NJ, 631 p.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 191–207
207
Channel Geomorphic Evolution after Dam Removal: Is Scale Important? TIMOTHY LAND ARPITA NANDI1 INGRID LUFFMAN Department of Geosciences, East Tennessee State University, Box 70357, 325 Treasure Lane, Johnson City, TN 37614
Key Terms: Dam Removal, Erosion, Geomorphology, Sedimentation, Field Studies, Channel Development ABSTRACT Post–dam removal geomorphologic adjustment of a stream channel has been documented in the scientific literature at watershed, hillslope, and laboratory scales. Hillslope-scale studies in channel cross sections are most common and add significant value in the dam-removal literature. This study examines geomorphic stream channel adjustment following dam removal at the hillslope scale under natural climatic conditions. A sedimentfilled silt fence dam (1 m tall, 12.65 m wide) was removed in three stages, and the width and depth of the upstream developing channel was monitored at six transects for 15 months. Headcut retreat and changes in channel sinuosity were also recorded. After the silt fence dam was removed, channel development was initiated by headcut formation, which migrated upstream at a rate of 4 cm/d for about 10 months and then gradually reached attenuation. The channel progressed through four distinct stages: Stage 1 (Initial conditions); Stage 2 (Downcutting)—wide, shallow, meandering channel incised to a maximum depth of 0.52 m, and sinuosity decreased; Stage 3 (Floodplain development)—upon reaching base level, surface runoff began to meander within the channel, widening it through bank slumps and erosion; and Stage 4 (Quasi-equilibrium)—channel development reached dynamic (quasi-) equilibrium with only minor widening at downstream transects (maximum width of the incised channel reached 0.46 m), accompanied by sediment aggradation. The stages of upstream channel development and headcut retreat pattern in this study are consistent with the findings of other studies at the laboratory and watershed scales, indicating that channel development after dam removal is scale independent.
1 Corresponding
author.
INTRODUCTION Reservoirs behind many existing dams (especially smaller, older dams) are filled with upstream sediment load. Sediment load reduces reservoir storage capacity, impairs structural integrity, and increases downstream scouring, leading to costly maintenance and risk (Wells et al., 2007). Yearly, 0.5 to 1 percent of global storage capacity is estimated to be lost as a result of sedimentation (Basson, 2009). Yearly loss of storage capacity due to sedimentation is higher than the increase in capacity by construction of new reservoirs (Schleiss et al., 2014). World-wide, the replacement cost due to sediment load deposited in reservoirs is nearly US$18 billion annually (excluding downstream impact [Basson, 2009]). Many dams have also aged beyond their design life span (Evans et al., 2000; Cantelli et al., 2004). Given the approximately 87,000 dams in existence in the United States primarily constructed between 1950 and 1980, and based on the continual aging of these structures, there will undoubtedly be many dam removals in the coming years for environmental, economic, and safety reasons. American Rivers (ASCE, 2016) has identified 1,300 dams that have been removed from U.S. rivers since 1912, of which 1,061 were removed in the last 25 years. The majority of the dams were removed because they were no longer performing the functions for which they were built or because they posed environmental or safety concerns. Moreover, negative ecological impacts are associated with the interruption of water flow, trapping of sediment, downstream bed incision, the physical barrier presented by the dam structure, and disrupted river habitat (Bednarek, 2001; Bushaw-Newton et al., 2002; and Cheng and Granata, 2007). Dam removal has gained popularity in recent decades as a way to reverse these negative ecological impacts (ASCE, 1997; Graf, 2002), and there is therefore a need to study the geomorphic effects of dam removal and restoration of natural flows on channel development and evolution (Pizzuto, 2002; Cantelli et al., 2007; and Sawaske and Freyberg, 2012). Until recently, the majority of dam removal studies have concentrated on dam failure and less so
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
209
Land, Nandi, and Luffman
on planned removal (MacDonald and LangridgeMonopolis, 1984; Singh and Scarlatos, 1988; Doyle et al., 2003; Vogel et al., 2015; Bellmore et al., 2016; Coleman et al., 2016; and Zhong et al., 2016). Recent research has focused on the effects of dam removal in instances in which pre- and post-removal conditions of controlled dam removal are documented (Evans et al., 2000; Pizzuto, 2002; Doyle et al., 2003; Wells et al., 2007; and Wang and Kuo, 2016). Post–dam removal geomorphic studies have been performed both at the large watershed scale (Pizzuto, 2002; Doyle et al., 2003; Cheng and Granata, 2007; Wells et al., 2007; Sawaske and Freyberg, 2012; East et al., 2015; and Wang and Kuo, 2016) and at the small laboratory model scale (Cantelli et al., 2004; Ferrer-Boix et al., 2010, 2014). Based on hydrologic processes, large watershed scale is greater than 5 km2 , small watershed scale is up to about 5 km2 , field scale or hillslope scale is up to about 1,000 m2 , and plot scale is under 10 m2 (McGuire et al., 2005). A comparison of the impacts of multiple dam removal projects at the watershed scale by Sawaske and Freyberg (2012) found that potential impacts of dam removal are site specific and can vary considerably with the local soil and geomorphic condition. In terms of channel development, many studies provide a qualitative assessment of geomorphic channel evolution associated with dam removal (Pizzuto, 2002; Doyle et al., 2003; Pearson et al., 2011; Greene et al., 2013; Thomas et al., 2014; Tullos et al., 2014; and Randle et al., 2015). Doyle and his team (2003) noted six stages of channel development after dam removal. These stages include the following: a) pre-removal condition with large sediment load in the reservoir; b) conditions immediately after the dam removal, generally a wide, shallow channel with low flow velocity; c) channel bed incision and formation of a narrow, deep channel with steep banks; d) channel widening and mass wasting of the banks; e) continued widening and floodplain development; and f) an equilibrium state with reduced bank height by channel bed aggradation. Relative to channel morphology, Harvey and Watson (1986) mentioned a six-stage conceptual model very similar to Doyle’s above six stages. In agreement with Doyle’s findings, Pizzuto (2002) independently studied upstream and downstream conditions following dam removal and found that upstream from the dam a channel incises through sediment fill and bank failures begin to occur in the channel. Pizzuto concluded that sediment fills are likely to erode even during low flows if the fills are composed of sand or cohesive silt and clay. More recent examples with the 210
same or similar results in upstream conditions observed by Pizzuto (2002) include the works of Wildman and MacBroom (2005), Wilcox et al. (2014), and Randle et al. (2015). At the laboratory scale, Ferrer-Boix et al. (2010) used a non-uniform mixture of sand and gravel in a flume to model post–dam removal channel bed profile evolution and the associated change in channel width. In a threestage removal process, the channel narrowed during the early stage of removal, followed by channel widening in later stages of removal. Cantelli et al. (2004) also used a laboratory flume setup to study erosion of deltaic dam reservoir sediment and channel propagation after dam removal. Results from both laboratory-scale studies to some extent parallel results from other studies at the watershed scale: channel incision and narrowing were observed in the initial stage after dam removal, followed by a period of channel widening. Thus, studies at the laboratory scale using simulated flows in a flume inlet have the potential to provide clues about the reservoir sediment response to dam removal prior to any actual dam-removal operation at a dam site (Cantelli et al., 2004; Ferrer-Boix et al., 2010). The primary limitation of laboratory-scale studies involves the lack of environmental variability because the studies are performed in controlled setting. Field measurements on hillslope channel cross sections (hillslope scale) are the most common scale of post–dam removal geomorphic assessments, and sitespecific studies have significant value in the damremoval literature (Cui et al., 2005; Sawaske and Freyberg, 2012). The reservoir load, sediment release, and channel restoration following dam removal should be addressed through site-specific analysis using conceptual models or field studies. A study by Kibler et al. (2011) on a small gravel-filled dam (>5 m) found that post–dam removal channel adjustment can vary as a result of the presence of diverse sediment size in the reservoir bed. In a cohesive sediment bank channel evolution model, Wildman and MacBroom (2005) found that incision steepens the banks to a critical threshold, which leads to mass failure, aggrading the bed with substantial sediment flux. Under a coarse-grained situation, bank failure occurred at a lower critical height, resulting in less sediment removal (Wildman and MacBroom, 2005). This indicates that for small dams of varying sizes (about 4–15 m), sediment load and post–dam removal stream channel adjustment is site specific and is very dependent on local site conditions (Tullos et al., 2014). This supports the need for a variety of field studies in diverse natural settings with differing sediment loads. Thus, to design an effective dam-removal strategy and stream-restoration effort at an existing dam site, prior experimental studies performed in a natural setting at the hillslope scale (10–1,000 m2 ), the smallest
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
Dam Removal Geomorphic Change
possible unit representing catchment scale runoff dynamics can provide valuable insights into the expected geomorphic impacts of the planned removal (Yair and Yassif, 2004; McGuire et al., 2005; and Colaiacomo and Wilcox, 2012). A favorable comparison between the results of this study and those of similar studies performed at watershed and laboratory scales would suggest that dam-removal effects are scale independent. In that case, hillslope-scale model studies prior to an actual removal may be employed as a decision tool to develop a strategy for dam removals in similar soil and geographic regions. The present study examines channel development at the hillslope scale within sediments accumulated behind a silt fence downstream of a gully system. The silty clay soils in the study area contrast with predominantly sandy soils in other studies (Rumschlag and Peck, 2007; Pearson et al., 2011; Skalak et al., 2011; and Harris and Evans, 2014). Accumulated reservoir delta sediments are the primary setting for channel development, while gullies upstream are the source of sediment and concentrated runoff. Staged removal of the silt fence mimics staged removal of small dams and can help us to understand the importance of scale in channel development. The goal of this study is to measure channel development in deltaic reservoir sediments behind a dam at the hillslope scale under natural precipitation conditions. In particular, this study examines how channel (a) development, (b) headcut retreat, and (c) sinuosity take place post–dam removal in cohesive soil in a subtropical humid climate. Additionally, the study addresses (d) scale independence in channel morphometrics. STUDY AREA AND EXPERIMENTAL SETUP The study was performed in an actively eroding hillslope with fluviokarst topography in the Valley and Ridge physiographic province in Tennessee (United States). Within the province, northeast- to southwesttrending parallel valleys and ridges consist of dolostone or limestone and shale, respectively (Moore, 1976). These rocks weather to strongly leached, less fertile, reddish-brown–colored, iron oxide–rich, fine-grained silty and clayey Ultisols (acrisols) of the CollegedaleEtowah complex (CeD3). The CeD3 soil is prone to weathering and erosion, leading to rill and gully erosion by water (Nandi and Luffman, 2012). East Tennessee has a “Humid Subtropical” climate ¨ (Cfa Koppen climate classification), characterized by warm summers, mild to cool winters, and year-round precipitation, with December, March, or April being the wettest month. Meteorological data were collected on site using a Davis Vantage Pro weather station. Average monthly precipitation during the study period
was 8.4 cm, with annual totals ranging from 90.3 cm to 116.6 cm. Average temperature during the study period ranged from 0.3◦ C in January to 22.3◦ C in July. Land cover for the study site is forest and pasture land, with surrounding tracts of land devoted to agricultural and residential uses. The research was performed in a 227-m2 area in the headwaters of Kendrick Creek in a small valley on a south-facing hillslope. Several substantial gullies have formed since the mid-1980s as a result of erosional processes triggered by pasture grazing. A silt fence dam constructed at the base of one of these gullies trapped eroded sediment, creating deltaic reservoir deposits. For the present study, channel development in these deposits, downstream of the gullies, was examined. The 1-m-tall silt fence dam was 12.65 m wide, and at capacity, it holds 10.36 m3 of sediment, measured by tallying the number of 10-L buckets filled during sediment removal in 2012. The trapped sediment had a measured composition of 6 percent sand, 44 percent silt, and 49 percent clay. According to U.S. Department of Agriculture (USDA) classification the sediment is silty clay and was classified as “CL-ML” according to the Unified Soil Classification System (USCS) classification system. The average bulk density was 1.32 gm/cc, the trapped sediment material porosity was 43 percent, and the hydraulic conductivity was 9.3 × 10−3 cm/s. The liquid limit and plasticity indices were 39 percent and 12 percent, respectively. The dominant minerals in the sediment were quartz, hematite, ferrihydrite, kaolinite, and chlorite. The trapped sediment showed negligible stratification (USDA, 2011; Nandi and Luffman, 2012). The trapped sediment and the 16-m-long stream channel behind the dam were monitored weekly from January 2014 to March 2015 (Figure 1). The dam was removed in three stages, each 30 cm (12 in.) in height, over the span of 58 days. The upper 30 cm of the dam was removed in the first stage on January 27, 2014. The next 30 cm was removed on February 10, 2014 (14 days after the Stage 1 removal), and the final cut was made on March 26, 2014 (58 days after the Stage 1 removal). Staged removal enabled a controlled release of trapped sediment (Cantelli et al., 2004), which aided in the field monitoring of channel development, headcut retreat, and change in sinuosity. Staged removal on a large scale alleviates downstream impacts from released sediment (Magirl et al., 2015) and lessens impacts to downstream ecosystems (Draut and Ritchie, 2015). Six transects across the dam and upstream channel were identified for field data collection, numbered T-0 to T-5, with T-0 closest to the dam and T-5 farthest away (Figure 1c). Change in channel morphometry was measured in three steps (summarized in Figure 2), as follows:
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
211
Land, Nandi, and Luffman
Figure 1. Study area location (a), geologic setting (b), and field experimental setup (c). Geologic units are marked on (b): Honaker dolomitic limestone (Chk), Conasauga Group consisting of Cambrian-age Maynardville limestone and Nolichucky shale (Ccu) and Knox limestone (OCk).
1) A series of time-lapse photographs were taken during field data collection for qualitative visual assessment of channel development. A post was installed 1.2 m downstream from the silt fence dam, and all photographs were taken from this vantage point to maintain accuracy.
2) Channel width and depth were measured for each transect using a 4-m stadia rod placed perpendicular to the channel along the transect. Channel width was measured directly from the stadia rod, and channel depth was measured using a meter stick placed vertically in the channel’s deepest point at each transect. 3) Headcut retreat was determined from the abrupt change in channel depth at each transect over time. 4) Sinuosity was quantified as the ratio of thalwegpath distance to down-valley axis distance (Evans et al., 2000). Additionally, the number of meanders was recorded as a measure of tortuosity. Sinuosity measurements were taken from January through September 2014, after which no further change was observed.
RESULTS Figure 2. Conceptual flow chart model of the channel morphometric assessment process.
212
Field data collected during the dam removal experiment were analyzed through qualitative visual as-
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
Dam Removal Geomorphic Change
Figure 3. Stages include pre-dam removal conditions (a), channel incision (b–e), channel widening (f–h), aggradation (i–j) and quasi-equilibrium state (k–l).
sessment of a sequence of field photos and quantitative geomorphic assessment of channel development, sinuosity, and headcut retreat upstream of the dam. Qualitative Visual Assessment of Field Photos The series of time-lapse photos shows changes to the channel as it developed (Figure 3). Figure 3a represents pre–dam removal conditions, when the silt fence dam contained the sediment load. Prior to the initial cut of the dam, the channel was initially shallow and wide. Following Stage 1 removal on Day 1, channel incision became evident (Figure 3b). Once the channel reached local base level (i.e., the height of the existing dam), the second cut was made (Stage 2 removal) on Day 14, and the headcut became prominent (Figure 3c). Thirteen days following Stage 2 removal, the
shallow and wide channel became incised and a narrow and deep channel with steep sidewalls developed (Figure 3d and e). After reaching local base level at Day 34, sidewalls began to slump as a result of erosion and mass wasting (Figure 3f). Subsequently on Day 58, the third and final cut was made (Stage 3 removal), and the channel began to widen. Channel widening continued throughout the remainder of the study (Figure 3g–l). On a few occasions the channel temporarily narrowed because of side wall (channel bank) slumping followed by widening (Figure 3h and i). Periods of aggradation were observed when sediment eroded from the upstream channel accumulated in the channel bottom and channel depth gradually decreased (Figure 3i and j). The channel remained in a quasi-equilibrium state with moderate to low activity in winter followed by increased channel activity during spring (Figure 3k and l).
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
213
Land, Nandi, and Luffman
Figure 4. Channel bed longitudinal profile development from pre– dam removal to channel-widening stage.
Quantitative Assessment of Channel Development At the initial pre-dam cut stage the bed longitudinal profile was planar to slightly convex upward in shape and with a very gentle gradient of 1 cm/m. Through the stages of dam removal and during the post–dam removal monitoring period the bed slope increased downstream near the former dam location, eventually showing a concave upward profile with a gradient of 22 cm/m (Figure 4). After each staged removal and over the course of 14 months, channel depth increased the most at locations closest to the former dam (52 cm at transects T0 and T1) (Figure 5). Upstream from the dam, channel depth increased gradually over time, showing the smallest increase at transect T5. This finding is consistent with that of Burroughs et al. (2009),
who measured greater amounts of erosion closest to the former dam, lessening with increasing upstream distance from the dam. In the first stages of data collection, during the process of dam removal, channel depth increased at a rate of 1.36 cm/d to 0.41 cm/d. Following initial channel development, the incision rate exponentially reduced from 0.35 cm/d to 0.08 cm/d (Table 1). The newly formed channel showed an overall increase in width after the initial dam-removal period, and on occasion the channel width increased rapidly as a result of local sidewall slumping. Like channel depth, the channel width at transect T0 showed the most change (77 cm). Both channel depth and channel width increased over time; however, the channel depth increased at a faster rate than did the width (Table 1). During the study period, the volume of stored sediment transported from behind the dam increased from 0.06 m3 to 2.09 m3 with channel incision and widening.
Headcut Retreat After dam removal, the first signs of channel development were noted by the formation of a headcut. However, as the study progressed, it became increasingly difficult to visually identify the headcut, preventing direct measurement in the field. By graphing the longitudinal profile (Figure 6) the headcut position became evident for each time step and was used to calculate the rate of retreat. The headcut retreated upstream in the form of a rapid near-vertical cut through transects T0, T1, T2, T3, and T5, except in transect T4, where the headcut showed a gradual drop. The average
Figure 5. Channel depth over time at all transects. Time of headcut retreat for each transect is shown with arrow. Precipitation accumulation is related to headcut migration and channel depth.
214
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
Dam Removal Geomorphic Change Table 1. Channel development stages (average of all transects).
Channel Development Stage after Doyle, 2003 Stage A/B: existing/initial condition Stage C: channel incision
Dam Removal Stage
Days
Depth (cm)
Width (cm)
Depth Change (mm/d)
Width Change (mm/d)
Sediment Volume (m3 )
1 14 58 83 121 186 228 248 276 297 324 367 389 416 447 468
12 19 24 29 27 31 31 31 33 36 36 32 31 34 40 39
0 0 2 2 11 21 24 23 24 27 37 49 49 35 35 35
0.0 13.6 4.1 3.5 2.2 1.7 1.4 1.3 1.2 1.2 1.1 0.9 0.8 0.8 0.9 0.8
0.0 0.0 0.3 0.2 0.9 1.1 1.1 0.9 0.9 0.9 1.1 1.3 1.3 0.8 0.8 0.7
0.00 0.00 0.03 0.04 0.23 0.50 0.57 0.55 0.61 0.74 1.02 1.20 1.16 0.91 1.07 1.04
1st removal 2nd removal 3rd removal
Stage D: channel widening begins
Stage E: channel widening continues
Stage F: quasi-equilibrium condition develops
rate of retreat of the cut was 4 cm/d, and the headcut retreat was not measurable after Day 276. Channel Sinuosity The downstream valley distance (15.30 m) was constant throughout the study; however, the thalweg-path distance measured between upstream and downstream points changed over time, decreasing from 16.08 m to 15.88 m during dam-removal stages and then increasing to 16.85 m during channel widening. As a result of the change in thalweg-path distance, the sinuosity index decreased from 1.05 to 1.04 immediately following
Stage 1 removal (Table 2) as the channel began to incise, but exhibited an increasing trend after the dam was completely removed (Stage 3 removal) on Day 58, as channel processes worked to widen the channel through meandering. Following the same pattern, the number of meanders decreased from four to three after Stage 1 removal and later increased to a maximum of seven meanders following Stage 3 removal.
DISCUSSION Channel Development Stages of Channel Evolution
Figure 6. Channel depth, width, their rate of change, and sediment volume behind the dam site. Over time the rate of total depth change decreased, where the rate of total width change had increased since inception during the monitoring period.
Comparison of our results to Doyle’s six stages of post–dam removal channel evolution (Doyle et al., 2003) shows that all stages were observed in our study, with similar processes observed in our study during Doyle’s Stages D and E. Doyle’s Stage A represented the initial conditions in which a large volume of sediment had accumulated behind the silt fence dam. During Stage B rainfall events, the flow of water over the sediment fill was shallow and wide with low velocity. Channel incision marked the beginning of Stage C after the dam removal, as flow became concentrated and created a deep, narrow channel. Stage D occurred when channel sidewalls began to slump, resulting in channel widening. Stage E was a continuation of the processes from the previous stage. Finally, aggradation of the channel bed and vegetation growth initiated Stage F. In contrast to Doyle’s stages, vegetation growth began during Stages C and D in summer 2014 and continued during Stage F with the arrival of spring 2015. Our results suggest that a four-stage conceptual model best
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
215
Land, Nandi, and Luffman Table 2. Change in channel meandering and sinuosity. The corresponding figure indicated how number of meanders varied with thalweg path distance.
describes channel evolution processes for this site, as follows: 1) Stage 1: Pre–dam removal conditions: a wide, shallow, meandering channel; 2) Stage 2: Dam removal and downcutting: narrow and deep channel develops with decreased sinuosity as flow incises into existing channel; 3) Stage 3: Channel widening: upon reaching base level, surface runoff begins to meander within the channel, causing channel widening; and 4) Stage 4: Quasi-equilibrium state: widening and aggradation of downstream areas (T0, T1, T2), some downcutting continues in upstream transects (T3, T4, T5) (Figure 4). Longitudinal Profile The increase in channel depth through time for transects T0 through T5 (Figure 4) captures the ongoing erosion of sediment associated with channel incision. Transects closest to the dam (T0 and T1) experienced the most dramatic depth increase, while the transect farthest from the dam (T5) showed the least change; T5 remained more shallow and showed no change during the first few weeks of the study because the headcut needed time to migrate upstream. At our study site, sediment build-up downstream due to bank failure and upstream sediment sources is apparent in the sudden decrease in channel depth at transects T0 and T1 after Day 297 (November 13, 2014) (Figure 5). Initially, erosion narrowed and deepened the channel until base level was attained, which led to widening of the channel. The channel-narrowing-then-widening process migrated upstream with the headcut such that the closer the cross section (transect) was to the dam, the faster 216
the channel narrowed, and then it began to widen. Following the stages outlined above, our findings agree with the watershed-scale channel development studied by Pizzuto (2002) and Randle et al. (2015); a channel will first incise through sediment fill, which will in turn be followed by bank failure once channel depth becomes great enough. Our findings also agree with those of Wilcox et al. (2014), in which sediment erosion occurs by knickpoint migration, channel incision, and small mass failures that propagate upstream. Comparison of our results with laboratory models by both Cantelli et al. (2004) and Ferrer-Boix et al. (2010) reflects similar processes of channel narrowing and subsequent widening after dam removal. Agreement between watershed-scale, hillslope-scale, and lab-scale models indicates that channel development is scale independent. Arguably, for erosion studies, field models are superior to lab models because natural climate processes and some physical conditions are not easily replicated in the lab. The results of the present study, therefore, suggest that field-scale models can be useful decision tools for planning purposes in larger dam-removal projects for the following reasons: 1) Channel development is scale independent, and therefore a hillslopescale model can accurately and affordably depict channel development processes at the watershed scale; and 2) Because impacts of dam removal are site specific and vary with the local soil and geomorphic condition (Sawaske and Freyberg, 2012), on-site hillslope-scale models can provide insight into channel development under very local conditions. Channel Width Versus Depth A graph of average channel width and depth for all transects shows the classic pattern of channel incision
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
Dam Removal Geomorphic Change
followed by widening (Figure 6). The rate of incision is highest immediately following Stage 1 removal and then decays exponentially. Channel width does not begin to increase until several weeks after Stage 3 removal (Day 83) and continues to increase in spurts, periodically slowed by slumping of channel walls, which fills in the channel downstream and reduces depth. The volume of accumulated sediment removed from behind the dam, estimated using transect cross-sectional area and thalweg distance, increased steadily throughout the channel incision and widening process. Sediment production continued at a steady rate toward the end of monitoring, when the channel attained the quasiequilibrium state (Figure 6, top). Headcut Retreat Several studies, including the present one, have found that after dam removal the headcut retreats upstream through the stored sediments (Doyle et al., 2003; Stewart and Grant, 2006; and Major et al., 2012). Bed incision post–dam removal, however, is not a required condition for bank erosion, as in at least one study of the Homestead Dam in Ashuelot River, southern New Hampshire, the channel eroded laterally, with a headcut appearing far upstream of the dam and impounded sediments (Gartner et al., 2015). In cohesive silt and clay sediments, such as those found at the study site, a vertical headcut (an eroding vertical face in the streambed) is likely to migrate upstream through the fill (Pizzuto, 2002). Headcut retreat was somewhat gradual during the study from one transect to the other, with an average rate of 4 cm/d. Although there was no permanent water flow in the channel, flow occurred during precipitation events at a rate sufficient to generate runoff, and the headcut retreat was detectable in relation to precipitation events. Qualitatively, this relationship is evident upon examination of Figure 5, in which a response to the two largest rainfall events is headcut migration in T5, channel incision in T4 and T2, and accretion in T3, T1, and T0. Quantitatively, channel depth is significantly correlated (using Spearman’s rho) to precipitation accumulation for T0 through T3 (r = 0.36 to 0.46 for current period, at a significance level of p < 0.05). Indicating longer-term impacts of precipitation potentially related to soil moisture conditions or the timing of field visits, precipitation accumulation in the prior measurement period is significantly correlated to channel depth for T0 through T3 and T5 (r = 0.4 to 0.61 for prior period, p < 0.05). Channel Sinuosity This study found that removal of the dam caused noticeable changes in channel meanders and sinuosity
in the upstream channel. The thalweg-path distance changed over time as channel incision was followed by channel widening. Changes in the ratio of thalwegpath distance to down-valley distance are a measure of changes in sinuosity of the channel, but to fully capture tortuosity, the number of meanders must also be measured. This is necessary because, for example, a stream with sinuosity of 1.1 has a 10 percent longer thalwegpath distance than it does down-valley distance, but it could manifest this distance through one large meander or through multiple smaller meanders. The latter case describes a stream with greater tortuosity and greater potential for meander migration. In this study the sinuosity initially decreased by 1.3 percent following Stage 1 removal and then increased by 6.4 percent as a measure of change in thalweg-path distance to down-valley distance (a 5.03 percent change from start to finish) (Table 2). Tortuosity (measured by the number of meanders) decreased from four to three meanders (a 25 percent reduction) during the incision stage, and then increased to seven meanders (a 233 percent increase) during channel widening. This indicates the presence of multiple smaller meanders, implying channel widening along the length of the channel, rather than one large meander developing in a single location. Evans et al. (2007) found similar changes in sinuosity, which can be explained by rapid downcutting that initially produces an incised and constricted channel and reduces the number of meanders and the sinuosity. The channel-widening stage that follows occurs because once the channel becomes graded, the stream begins to develop a floodplain by increased meandering, which we measured as sinuosity. Evans et al. (2007) concluded the same, correlating the decrease in sinuosity to the cutting of chute channels and erosion of bedforms after dam removal. Similar results in cohesive fine-grained sediment were found by Galay (1983), Stewart and Grant (2006), and Downs et al. (2009). This study differs from other studies of channel development following dam removal in one additional important factor: channel development in this study is dependent on precipitation because the runoff from upstream gullies is ephemeral. Interestingly, while flow is intermittent, the results from this study agree with results from studies of channel development in permanent streams (Rumschlag and Peck, 2007; Major et al., 2012). This suggests that systems with seasonal flows or flashy behavior will respond to dam removal in similar ways. These findings are consistent with those of Major et al. (2012), in which erosion occurred primarily through interactions among knickpoint migration, channel incision, and lateral erosion within 60 hours after breaching of the Marmot Dam. In the case of Rumschlag and Peck (2007), incision and vertical downcut-
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
217
Land, Nandi, and Luffman
ting preceded channel widening and slumping. These are two cases of perpetual flows that show results strikingly similar to the ephemeral setting of this study.
CONCLUSIONS This study examined the effects of dam removal at the hillslope scale by quantifying channel evolution in sediments trapped behind a silt fence dam. The silt fence dam was removed in three stages, and the following effects were observed:
• The pre–dam removal condition revealed a planar to slightly upward-convex longitudinal stream profile, with few meanders and low sinuosity. • During the initial post–dam removal stages, rapid undercutting and erosional narrowing took place, forming a deep incised channel. At this stage the channel meandering and sinuosity decreased as flow concentrated in the narrow newly formed channel. • The narrow incised channel stage was followed by channel widening, during which the sidewalls of the incised channel continued to erode. At this stage the channel longitudinal profile indicated an upwardconcave trend, which is common in graded streams. • While channel aggradation and widening continued, the undercutting decreased, a new floodplain developed, and channel sinuosity and meandering increased, indicating a quasi-equilibrium state. Dam removal causes profound changes in channel morphology. The processes observed for channel development and evolution at the hillslope scale suggest that channel adjustment after a dam removal follows a predictable sequence of fluvial geomorphic events, which mirrors the findings at both laboratory and watershed scales. Therefore, the study provides evidence that post–dam removal upstream channel adjustment is a scale-independent geomorphic process. Any likely spatial and temporal variation in channel adjustment can be attributed to different dam size, sediment load, soil type, and stream discharge, which were beyond the focus of this study. Further studies on channel evolution following dam removal should be carried out across multiple soil types, sediment loads, dam sizes, and, most importantly, in varied climate types. This research suggests that hillslope-scale physical models of channel development are appropriate and representative planning tools for large-scale dam-removal projects because of the demonstrated scale independence of channel adjustment. We hope this research serves as a reference on effects of dam removal by improving the efficiency of future dam removals. 218
REFERENCES ASCE, 1997, Guidelines for Retirement of Dams and Hydroelectric Facilities: ASCE, New York. 222 p. ASCE, 2016, American Rivers: Electronic document, available at http://www.americanrivers.org/threats-solutions/restoringdamaged-rivers/dam-removal-map/ BASSON, G. R., 2009, Management of Siltation in Existing and New Reservoirs: General Report Q. 89. Proceedings of the 23rd Congress of the International Commission on Large Dams (CIGBICOLD), Vol. 2. BEDNAREK, A. T., 2001, Undaming rivers: A review of the ecological impacts of dam removal: Environmental Management, Vol. 27, No. 6, pp. 803–814. BELLMORE, R. J.; DUDA, J. J.; CRAIG, L. S.; GREENE, S. L.; TORGERSEN, C. E.; COLLINS, M. J.; AND VITTUM, K., 2016, Status and trends of dam-removal research in the U.S.: WIRES Water, doi:10.1002/wat2.1164. BURROUGHS, B. A.; HAYES, D. B.; KLOMP, K. D.; HANSES, J. F.; AND MISTAK, J., 2009, Effects of Stronach Dam removal on fluvial geomorphology in the Pine River, Michigan, United States: Geomorphology, Vol. 110, Nos. 3–4, pp. 96–107. BUSHAW-NEWTON, K. L.; HART, D. D.; PIZZUTO, J. E.; THOMSON, J. R.; EGAN, J.; ASHLEY, J. T.; JOHNSON, T. E.; HORWITZ, R. J.; KEELEY, M.; LAWRENCE, J.; CHARLES, D.; GATENBY, C.; KREEGER, D. A.; NIGHTENGALE, T.; THOMAS, R. L.; AND VELINKSKY, D. J., 2002, An integrative approach towards understanding ecological responses to dam removal: The Manatawny Creek study: Journal American Water Resources Association. Vol. 38, No. 6, pp. 1581–1599. CANTELLI, A.; PAOLA, C.; AND PARKER, G., 2004, Experiments on upstream-migrating erosional narrowing and widening of an incisional channel caused by dam removal: Water Resources Research, Vol. 40, No. 3, W03304, doi:10.1029/ 2003WR002940. CANTELLI, A.; WONG, M.; PARKER, G.; AND PAOLA, C., 2007, Numerical model linking bed and bank evolution of incisional channel created by dam removal: Water Resources Research, Vol. 43, No. 7, W07436, doi:10.1029/2006WR005621. CHENG, F. AND GRANATA, T., 2007, Sediment transport and channel adjustments associated with dam removal: Field observations: Water Resources Research, Vol. 43, No. 3, W03444, doi:10.1029/2005WR004271. COLAIACOMO, E. J. AND WILCOX, A. C., 2012, Downstream channel response to dam removal at reach and Bedform scales, White Salmon River, Washington (Abstract 30-7): Geological Society America Abstracts Programs, Vol. 44, No. 7, p. 105. COLEMAN, N.; KAKTINS, U.; AND WOJNO, S., 2016, Dam-Breach Hydrology of the Johnstown Flood of 1889—Challenging the Findings of the 1891 Investigations Report: Heliyon, doi:10.1016/j.heliyon.2016.e00120. CUI, Y.; PARKER, G.; BRAUDRICK, C.; DIETRICH, W.; AND CLEUR, B., 2005, Dam Removal Express Assessment Models (DREAM) Part 1: Model development and validation: Journal Hydrologic Research, Vol. 44, No. 3, pp. 1–17. DOWNS, P.; CUI, Y.; WOOSTER, J.; DUSTERHOFF, S.; BOOTH, D.; DIETRICH, W.; AND SKLAR, L., 2009, Managing reservoir sediment release in dam removal projects: An approach informed by physical and numerical modelling of non-cohesive sediment: International Journal River Basin Management, Vol. 7, No. 4, pp. 433–452. DOYLE, M. W.; STANLEY E. H.; AND HARBOR, J. M., 2003, Channel adjustments following two dam removals in Wisconsin: Water Resources Research, Vol. 39, No. 1, doi:10.1029/ 2002WR001714. DRAUT, A. E. AND RITCHIE, A. C., 2015, Sedimentology of new
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
Dam Removal Geomorphic Change fluvial deposits on the Elwha River, Washington, USA, formed during large-scale dam removal: River Research Applications, Vol. 31, pp. 42–61, doi:10.1002/rra.2724. EAST, A. E.; PRESS, 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. J., 2015, Large-scale dam removal on the Elwha River, Washington, USA: River channel and floodplain geomorphic change: Geomorphology, Vol. 228, pp. 765–786. EVANS, J. E.; HUXLEY, J. M.; AND VINCENT, R. K., 2007, Upstream channel changes following dam construction and removal using a GIS/remote sensing approach: Journal American Water Resources Association, Vol. 43, No. 3, pp. 1–15. EVANS, J. E.; MACKEY, S. D.; GOTTGENS, J. F.; AND GILL, W. M., 2000, Lessons from a dam failure: Ohio Journal Science, Vol. 100, No. 5, pp. 121–131. FERRER-BOIX, C.; MARTIN-VIDE, J. P.; AND PARKER, G., 2014, Channel evolution after dam removal in a poorly sorted sediment mixture: Experiments and numerical model: Water Resources Research, Vol. 50, No. 11, pp. 8997–9019, doi:10.1002/2014WR015550. FERRER-BOIX, C.; VIPARELLI, E.; CANTELLI, A.; HAYDEL, R.; PARKER, G.; AND MARTIN-VIDE, J. P., 2010, Incision and width changes caused by dam removal. Experiments and data analysis: River Flow 2010 Dittrich, Koll, Aberle & Geisenhainer ¨ Wasserbau. (eds). Bundesanstalt fur GALAY, V. J., 1983, Causes of river bed degradation: Water Resources Research, Vol. 19, pp. 1057–1090. GARTNER, J. D.; MAGILLIGAN, F. J.; AND RENSHAW, C. E., 2015, Predicting the type, location and magnitude of geomorphic responses to dam removal: Role of hydrologic and geomorphic constraints: Geomorphology, Vol. 251, pp. 20–30. GRAF, W. L. (Editor), 2002, Dam removal research. Status and prospects. In Proceedings of the Heinz Center’s Dam Removal Research Workshop, October 23–24, 2002: The H. John Heinz III Center for Science, Economics and the Environment, Washington, DC, 164 pp. GREENE, S. L.; KRAUSE, A. J.; AND KNOX, J. C., 2013, A decade of geomorphic and hydraulic response to the LaValle Dam project, Baraboo River, Wisconsin: Journal American Water Resource Association, Vol. 49, pp. 1473–1484. HARRIS, N. AND EVANS, J. E., 2014, Channel evolution of sandy reservoir sediments following low-head dam removal, Ottawa River, northwestern Ohio, U.S.A.: Open Journal Modern Hydrology, Vol. 4, pp. 44–56. HARVEY, M. D. AND WATSON, C. C., 1986, Fluvial processes and morphological thresholds in incised channel restoration: Water Resources Bulletin, Vol. 22, No. 3, pp. 359–368. KIBLER, K.; TULLOS, D.; AND KONDOLF, M., 2011, Evolving expectations of dam removal outcomes: Downstream geomorphic effects following removal of a small, gravel-filled dam: Journal American Water Resources Association, Vol. 47, No. 22, pp. 408–423. MACDONALD, T. AND LANGRIDGE-MONOPOLIS, J., 1984, Breaching characteristics of dam failures: Journal Hydraulic Engineering, Vol. 10, No. 5, pp. 567–586, doi:10.1061/(ASCE)07339429(1984)110:5(567). MAGIRL, C. S.; HILLDALE, R. C.; CURRAN, C. A.; DUDA, J. J.; STRAUB, T. D.; DOMANSKI, M.; AND FOREMAN, J. R., 2015, Large-scale dam removal on the Elwha River, Washington, USA: Fluvial sediment load: Geomorphology, Vol. 246, pp. 669–686. MAJOR, J. J.; O’CONNOR, J. E.; PODOLAK, C. J.; KEITH, M. K.; GRANT, G. E.; SPICER, K. R.; PITTMAN, S.; BRAGG, H. M.; WALLICK, J. R.; TANNER, D. Q.; RHODE, A.; AND WILCOCK, P.
R., 2012, Geomorphic Response of the Sandy River, Oregon, to Removal of Marmot Dam: U.S. Geological Survey Professional Paper 1792, 64 p. MCGUIRE, K. J.; MCDONNELL, J. J.; WEILER, M.; KENDALL, C.; MCGLYNN, B. L.; WELKER, J. M.; AND SEIBERT, J., 2005, The role of topography on catchment-scale water residence time: Water Resources Research, Vol. 41, No. 5, W05002. MOORE, H. L., 1976, Drainage problems in carbonate terrain of East Tennessee. In Proceedings of 27th Annual Highway Geology Symposium: Orlando, FL, pp. 112–131. NANDI, A. AND LUFFMAN, I., 2012, Erosion related changes to physiochemical properties of Ultisols distributed on calcareous sedimentary rocks: Journal Sustainable Development, Vol. 5, No. 8, pp. 52–68. PEARSON, A. J.; SNYDER, N. P.; AND COLLINS, M. J., 2011, Rates and processes of channel response to dam removal with a sandfilled impoundment: Water Resources Research, Vol. 47, No. 8, W08504, doi:10.1029/2010WR009733. PIZZUTO, J. E., 2002, Effects of dam removal on river form and processes: BioScience, Vol. 52, No. 8, pp. 683–692. RANDLE, T. J.; BOUNTRY, J. A.; RITCHIE, A.; AND WILLE, K., 2015, Large-scale dam removal on the Elwha River, Washington, USA: Erosion of reservoir sediment: Geomorphology, Vol. 246, pp. 709–728, doi:10.1016/j.geomorph.2014.12.045. RUMSCHLAG, J. H. AND PECK, J. A., 2007, Short-term sediment and morphologic response of the middle Cuyahoga River to the removal of the Munroe Falls Dam, Summit County, Ohio: Journal Great Lakes Research, Vol. 33, pp. 142–153. SAWASKE, S. AND FREYBERG, D., 2012, A comparison of past small dam removals in highly sediment-impacted systems in the U.S.: Geomorphology, Vols. 151–152, pp. 50–58. SCHLEISS, A. J.; DE CESARE, G.; FRANCA, M. J.; AND PFISTER, M. (Editors), 2014, Reservoir Sedimentation. Taylor & Francis Group, London, U.K. SINGH, V. AND SCARLATOS, P., 1988, Analysis of gradual earthdam failure: Journal Hydraulic Engineering, Vol. 114, No. 1, pp. 21–42, doi:10.1061/(ASCE)0733-9429(1988)114:1(21). SKALAK, K.; PIZZUTO, J.; EGAN, J.; AND ALLMENDINGER, N., 2011, The geomorphic effects of existing dams and historic dam removals in the U.S. Mid-Atlantic region. In Papanicolaou, A. N. and Barkdoll, B. D. (Editors), Sediment Dynamics upon Dam Removal: American Society of Civil Engineers, Reston, VA USA, pp. 83–96. STEWART, G. AND GRANT, G., 2006, What can we learn from the removal of little dinky dams? In Moglen, G. E. (Editor), Managing Watersheds for Human and Natural Impacts: Proceedings of the 2005 Watershed Management Conference, July 19–22, 2005, Williamsburg, VA: American Society of Civil Engineers, Reston, VA, pp. 1–7. THOMAS, R. J.; CONSTANTINE, J. A.; GOUGH, P.; AND FUSSELL, B., 2014, Rapid channel widening following weir removal due to bed-material wave dispersion on the River Monnow, Wales: River Research Applications, Vol. 31, No. 8, pp. 1017–1027, doi:10.1002/rra.2803. TULLOS, D. D.; FINN, D. S.; AND WALTER, C., 2014, Geomorphic and ecological disturbance and recovery from two small dams and their removal: PLoS One, Vol. 9, No. 9, e108091, doi:10.1371/journal.pone.0108091. UNITED STATES DEPARTMENT OF AGRICULTURE (USDA), 2011, United States Department of Agriculture, Web Soil Survey, Washington County Soil Survey Report (2011): USDA, Washington, DC. VOGEL, A.; COURIVAUD, J.; AND JARECKA, K., 2015, Dam failures of embankment dams caused by overtopping or accidental leakage—New aspects for breach widening and deepening process: Dam Protections against Overtopping and
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
219
Land, Nandi, and Luffman Accidental Leakage, CRC Press Balkem, The Netherlands, pp. 59–71. WANG, H.-W. AND KUO, W.-C., 2016, Geomorphic responses to a large check-dam removal on a mountain river in Taiwan: River Research Applications, Vol. 32, No. 5, pp. 1094–1105. WELLS, R. R.; LANGENDOEN, E. J.; AND SIMON, A., 2007, Modeling pre-and post-dam removal sediment dynamics: The Kalamazoo River, Michigan: Journal American Water Resources Association, Vol. 43, No. 3, pp. 773–785, Doi:10.1111/j.17521688.2007.00062.x. WILCOX, A. C.; O’CONNOR, J. E.; AND MAJOR, J. J., 2014, Rapid reservoir erosion, hyperconcentrated flow, and downstream deposition triggered by breaching of 38 m tall Condit Dam, White Salmon River, Washington: Journal Geophys-
220
ical Research: Earth Surface, Vol. 119, No. 6, pp. 1376–1394, doi:10.1002/2013JF003073. WILDMAN, L. A. S. AND MACBROOM, J. G., 2005, The evolution of gravel bed channels after dam removal: Case study of the Anaconda and Union City Dam removals: Geomorphology, Vol. 71, pp. 245–262, doi:10.1016/j.geomorph. 2004.08.018. YAIR, A. AND YASSIF, N. R., 2004, Hydrological processes in a small arid catchment: Scale effects of rainfall and slope length: Geomorphology, Vol. 61, Nos. 1–2, pp. 155–169. ZHONG, Q.; WU, W.; CHEN, S.; AND WANG, M., 2016, Comparison of simplified physically based dam breach models: Natural Hazards, Vol. 84, No. 2, pp. 1385–1418, doi:10.1007/s11069016-2492-9.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 209–220
A Proposed Risk-Based Screening Strategy for Bridges Potentially Affected by Rock Scour WILLIAM L. NIEMANN1 ISAAC W. WAIT Marshall University, One John Marshall Drive, Huntington, WV 25755
JEFFREY R. KEATON Amec Foster Wheeler, 6001 Rickenbacker Road, Los Angeles, CA 90040
Key Terms: Rock Scour, Bridge, Risk-Based Screening ABSTRACT Based on an assessment of rock scour at 15 bridges across West Virginia, a risk-based screening strategy is proposed that accounts for both mode of scour and stream power. The proposed strategy is a result of research conducted to 1) characterize the hydraulic scour of rock at representative bridge sites in West Virginia with varying geological and hydrologic conditions; 2) assess the applicability of rock scour prediction techniques from National Cooperative Research Program (NCHRP) Report 717 to the types of rock and scour conditions at the bridge locations; and 3) identify strategies to better characterize scour of rock at bridge locations. Steps in the research included bridge-site selection; field inspection; determination of scour mode (i.e., quarrying/plucking or abrasion) and magnitude; rock core sampling; laboratory testing; and hydrologic and hydraulic evaluation and modeling. The abrasion-scourmode test described in NCHRP 717 for degradable rock that erodes gradually over time was applied. However, all but one bridge site in the West Virginia study included durable, fractured rock that eroded by quarrying/plucking of rock blocks. Based on the results of this study, a risk-based screening strategy is proposed for bridges potentially affected by rock scour. This screening strategy includes three tiers: Tier I shows evidence of long-term channel stability and requires no further action in assessing rock scour; Tier II requires consideration of abrasion as the only mode of scour; and Tier III requires consideration of quarrying/plucking as the primary or significant secondary mode of scour.
1 Corresponding
author email: niemann@marshall.
INTRODUCTION In 1987, the Schoharie Creek Bridge on Interstate 90 in upstate New York failed as a result of scour of the glacial till on which the bridge foundation was supported. Although not rock, the ice-compacted glacial till had been considered suitable for foundation support and had survived the largest flood on record in 1955 (2165 m3 /s) prior to failing in a smaller event (1835 m3 /s) (Keaton, 2013). In response to this bridge failure, the Federal Highway Administration (FHWA) issued a mandate for state transportation departments to inspect all bridges over water and later required examination of scour and related hydraulic conditions with the goal of improving methods and procedures for predicting scour. Many of the bridges identified following the FHWA action were founded on rock, yet available scour-prediction techniques (e.g., Arneson et al., 2012) were based on sand-bed channels, which tended to overestimate scour for rock channels. Other methods for accounting for rock scour in bridge design were also unsatisfactory. For example, the West Virginia Department of Highways (WVDOH) considered rock with a rock quality designation (RQD) of ≥50 percent to be scour resistant (Carte, 2011). Yet relying on RQD, or any other simple index parameter of rock condition, to guide the prediction of potential scour depth neglects hydraulic loading and often resulted in foundations that were constructed more deeply than professional experience suggested was actually necessary. The lack of reliable predictive methods in West Virginia and other states posed a dilemma for bridge engineers: the knowledge that rock and rock-like material could scour significantly during high-discharge events—even leading to catastrophic failure—yet without reliable methods for predicting scour in such materials. Understanding of rock scour was advanced significantly as a result of NCHRP Report 717 (i.e., Keaton et al. 2012). The report identified four modes of rock scour and developed a quantitative methodology by which to estimate the time-rate and magnitude of scour
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
221
Niemann, Wait, and Keaton
in cases in which rock is being removed gradually and progressively by one of these modes, grain-scale erosion (abrasion). The other three modes of rock scour are dissolution of soluble rocks, erosion by cavitation, and quarrying and plucking of durable rock blocks. Keaton (2013) summarizes the NCHRP findings and methodology and focused on the abrasion mode. A key principle developed in both Keaton et al. (2012) and Keaton (2103) is that rock scour is a “rock-water interaction phenomenon.” The rock-water interaction dynamic informs the methodology for evaluating abrasion: rock durability and hydraulic loading are evaluated at both the laboratory and field scales. A “geotechnical scour number” (GSN), a measure of equivalent scour depth per unit of equivalent cumulative stream power, is determined from a modified slake durability test performed on rock samples from a specific bridge site. This GSN is then compared to an “empirical scour number,” or “scour number” (SN), which is calculated as the ratio of measured scour at the bridge to the cumulative stream power over the time interval during which the measured scour occurred. Keaton et al. (2012) reported that the method offered the greatest promise for prediction of scour at bridges affected by abrasion and where repeated cross-section measurements of channel geometry are made over time. In 2011 the WVDOH specified that the NCHRP methodology described in Keaton et al. (2012) should be applied to 15 bridges representative of varying geography and geology within the state where scour was occurring. This article briefly describes the NCHRP study by Keaton et al. (2012), which formed the basis for the WVDOH study (RP-273); describes methods and results from the WVDOH study; and presents conclusions and a proposed screening strategy for bridges potentially affected by rock scour. RESEARCH APPROACH The 15 West Virginia bridge sites included in this study (Figure 1) were all founded in sedimentary rock and were selected based on their degree of agreement with an ideal set of characteristics, which included the following: 1) the presence of measurable scour; 2) availability of stream gage data, channel cross-sectional surveys over time, and bridge surface elevation data; 3) geology limited to a single rock type; 4) absence of significant stream bedload deposits concealing the bedrock surface; and 5) drill rig accessibility adjacent to abutments. Unfortunately, none of the bridge sites met all of the above criteria. In particular, historical cross sections were lacking for all 15 bridges, and stream gaging data were lacking for 11 bridges. The missing data necessitated changes to the methodology 222
Keaton et al. (2012) used to compute SN for their analysis of scour at the bridge over the Sacramento River described in NCHRP 717. The lack of historical cross sections meant that scour depth(s) could not be based on measurements from cross sections representing different times in the life of a bridge; instead, the only channel measurement available was the one performed during the time period of this project. In addition, scour could be quantified only for the time interval between initial bridge construction and the site assessment. The methods described in NCHRP 717 draw on stream gaging data reported in terms of average daily flow in order to compute cumulative stream power, reported as daily values with units of ft-lb/s/ft2 (Nm/s/m2 = J/s/m2 = W/m2 ). The lack of stream gaging data in West Virginia meant that historical stream flow over time, and thus cumulative stream power, could not be computed directly; instead it was estimated from probabilistic watershed rainfall-runoff models, as described in Appendix A. As a result of these differences in the flow data that serve as the basis for calculating cumulative stream power, cumulative stream power in this project was computed in terms of ft-lb/ft2 (J/m2 ) for “named” discharge events (i.e., the 25-year discharge) and then converted into the basis of “daily” cumulative stream power when necessary. Part of the justification for using named discharge events in the WVDOH research was the general absence of rock materials that scour gradually by abrasion. Geology of the West Virginia bridge sites is instead dominated by the presence of durable rocks that scour by quarrying at discrete thresholds of stream discharge. Subsurface samples for the West Virginia study were obtained through rock coring per ASTM D 2113 (2014) adjacent to bridge abutments (Figure 2), and surface samples were collected from rock exposed in stream channels in the immediate areas of the bridges. Modified slake durability testing (i.e., 60-minute wetting cycles, no drying between wetting cycles, and three to seven wetting cycles) was performed on all rock samples, consistent with the methodology of Keaton et al. (2012). Results showed no significant difference between the core and surface samples. Figure 3 shows example test data, while Figure 4 shows samples before and after the test. Geotechnical scour numbers (GSNs) then were determined from results of modified slake durability testing using the procedure developed by Keaton et al. (2012), illustrated in Figure 5, as well as by an alternative method (GSN) that better fit the West Virginia data (Figure 6). Table 1 summarizes the site characteristics and data sources for the 15 West Virginia bridge sites and the five sites in the NCHRP 717 study. Appendix A contains details of project methodology with respect to specific field and analytical tasks.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
Risk-Based Strategy for Scour-Affected Bridges
Figure 1. Locations of the 15 bridges used in RP-273 plotted on maps of shaded relief and general geology.
RESULTS Scour Mode and Magnitude Two of the scour modes defined by Keaton (2013) were recognized at the bridge sites evaluated for RP273: 1) abrasion affecting degradable (i.e., non-durable) rocks over time by the available cumulative stream power and 2) quarrying affecting durable rocks fractured into blocks small enough to be mobilized by tur-
bulence intensity at peak discharge and transported over time by discharges that exceed the threshold for transporting the block sizes. Field criteria consistent with those of Keaton et al. (2012) and summarized in Table 2 were used to determine the mode of scour at each bridge. In fact, the majority of West Virginia bridge sites showed some evidence for both abrasion and quarrying. Accordingly, five categories of scour were recognized for RP-273, three of them hybrid in
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
223
Niemann, Wait, and Keaton
Figure 2. Generic cross section of bridge abutment and target zone (T.Z. in figure) for rock coring. Vertical scour and depth of scour shown; horizontal scour is measured perpendicular to plane of cross section.
nature: 1) pure abrasion (PAB); 2) dominantly abrasion (DAB), 3) sub-equal abrasion and quarrying (AB/QP); 3) dominantly quarrying (DQP); and 5) pure quarrying (PQP). Only one site (Coon Creek) was judged to be affected solely by abrasion (PAB); four sites were
judged to be affected solely by quarrying (PQP). The remainder of sites fell into one of the three hybrid categories. Photographs illustrating examples of PAB, PQP, and AB/QP are shown as Figures 7, 8, and 9, respectively. Keaton et al. (2012) recognized that more than
Figure 3. Examples of late-time data from modified slake durability test define a quasi-linear trend indicating a constant rate of sample mass decay. Procedure is based on the method of Dickenson and Baillie (1999).
224
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
Risk-Based Strategy for Scour-Affected Bridges
Figure 4. Examples of samples with varying degrees of abrasion resistance when subjected to the modified slake durability test. Photos A and B show a relatively non-resistant sample before (A) and after (B) testing, while photos C and D show a relatively resistant sample before (C) and after (D) testing. Note that more resistant rocks are likely scouring by quarrying and plucking and that the abrasion being simulated in the modified slake durability test is therefore not providing useful information on scour. This emphasizes the need to identify mode of scour in the field in order to specify appropriate laboratory testing.
one scour mode may operate over the life of a bridge as scour exposes rock of different character; three of the five sites evaluated for NCHRP 717 were considered to be affected by abrasion only, whereas scour at the other two sites involved some component of quarrying (Keaton et al., 2012). Table 3 summarizes scour mode and rock type by bridge site, illustrating the range of conditions represented at the 15 bridges studied for RP-273. As would be expected, degradable rocks such as shale, claystone, and weathered sandstones scour primarily by abrasion, whereas durable rocks such as well-cemented sandstone and limestone scour predominantly by quarrying. Figure 10 illustrates the geographic distribution of scour mode on a shadedrelief and general geologic map of West Virginia. Five of the six sites in which quarrying was judged to be the exclusive (“pure”) or dominant mode of scour (PQP and DQP, respectively) are located in the northeast-
ern quarter of the state along the Allegheny Front and within the Valley and Ridge Province. In contrast, abrasion tends to be the sub-equal, dominant, or exclusive (“pure”) scour mode in the western and southwestern parts of the state in the Appalachian Plateau (AB/QP, DAB, and PAB, respectively). The authors attribute the predominance of quarrying in the northeastern quarter of the state to jointing and recrystallization associated with ancient tectonic activity. Measurements of vertical and horizontal scour and scour depth are presented in Table 4; Figure 2 illustrates each measurement type. GSN and GSN* from Modified Slake Durability Tests The GSN values calculated from the modified slake durability data are listed in Table 5. The GSN values obtained for this project range from 2.89 × 10−6 to
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
225
Niemann, Wait, and Keaton
Figure 5. Example of modified slake durability data from RP-273 expressed as equivalent scour depth and equivalent stream power. Slope of the zero-intercept line for individual samples is equal to the geotechnical scour number (GSN).
7.46 × 10−5 (6.22 × 10−8 to 1.56 × 10−6 m/W/m2 ) and a median of 1.11 × 10−5 (units of feet per ftlb/s/ft2 of daily cumulative stream power; 2.32 × 10−7 m/W/m2 ). These compare to GSN values reported by Keaton et al. (2012) ranging from 1 × 10−5 to 2.22 × 10−3 (2.09 × 10−7 to 4.64 × 10−5 m/W/m2 ), with a median of 1.7 × 10−4 (3.55 × 10−6 m/W/m2 ) expressed in the same units. RP-273 included both core and surface samples, whereas NCHRP 717 samples were collected mostly from surface exposures (Keaton, 2013). A comparison of GSN values determined from RP-273 core and surface samples indicated no significant variation between sample types for most of the bridge sites. This result supports the conclusion that differences between the NCHRP 717 and West Virginia GSN distributions cannot be attributed to the method of sample collection and that instead the significantly lower values and more limited range in GSN from RP-273 reflect the widespread presence of more durable and uniform rock across West Virginia. When mapping values of GSN and GSN* there are no apparent patterns in either their geographic or geologic distribution.
226
A median coefficient of determination (R2 ) was determined in the calculation of GSN and GSN* for each bridge site (in Table 5 the coefficient for GSN* is denoted as R2 *). Neither the R2 * nor R2 indicate a significant statistical correlation, and they are therefore not suitable for developing a predictive model for scour at the West Virginia bridges. However, the GSN*/GSN and R2 */R2 ratios (last two columns in Table 5) have average (median) values of 17.7 and 5.7, respectively, and are a clear indication that GSN* better fits the data to a much higher degree than does GSN and that it offers much greater potential for characterizing abrasion scour of durable rocks. Cumulative Stream Power Estimates of cumulative stream power, expressed in the same daily-basis units that result from experimental analysis through modified slake durability testing, are given in Table 6 for each project site. This table also includes a normalized representation of cumulative stream power at each site; cumulative daily effective stream power per year is the cumulative daily effective
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
Risk-Based Strategy for Scour-Affected Bridges
Figure 6. Representative plot of continuous slake durability test results with linear trend of late-time data determined with and without forcing a zero-intercept (GSN and GSN*, respectively).
stream power since construction divided by the number of years between bridge construction and scour depth inspection (i.e., “bridge age”). Since the rainfall estimates upon which cumulative stream power is based are probabilistically derived, deviation of actual flow conditions from those that might be expected on an average basis may be particularly pronounced for sites with only a few years since bridge construction, where averaging out over time has not yet occurred. Additionally, since eddies, vortices, and other effects of turbulent flow likely occur along abutments and around piers during the range of flows associated with unsteady conditions, a generalized assessment of average energy at the channel in the bridge opening may differ considerably from the energy applied at any single location over time.
In the absence of any scour modes other than abrasion, and if cumulative stream power could be definitively calculated rather than estimated probabilistically, the method of Keaton (2013) should show a relatively linear relationship between cumulative excess stream power and vertical scour. Instead, as shown in Figure 11, substantial variance appears to exist among the project data and fails to support the existence of such a simple model. The likely contributors to this variance include the presence of quarrying and plucking at most sites; variability in the degree of resistance to plucking related to the geology and hydraulics; and the multitude of uncertainties that accumulate along the modeling and computational path that begins with probabilistic precipitation depth-duration-frequency estimates and ends with estimation of cumulative daily stream power.
Scour Numbers
Comparison of GSN, GSN*, and SN
Scour number (SN), derived from vertical scour identified during field inspection and estimated cumulative effective stream power, is summarized in Table 7.
A comparison of SN and GSN for each of the project sites is provided in Table 8. Also included is a computed ratio of SN to GSN, which, if GSN is to be a
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
227
Niemann, Wait, and Keaton Table 1. Comparison of data sources for NCHRP 717 and WVDOH RP-273.
Project RP-273
Site
Leatherwood Road, Left Fork Leatherwood Creek Fifth Street, Fourpole Creek Little Sandy Creek Grassy Run, Prickett Creek Caldwell Run Paden Fork Audra Park, Middle Fork River Laurel Fork Roaring Creek Beverly, Tygart Valley River Coon Creek Bridge Fork Cucumber, Jacob’s Fork Clear Fork, Cedar Creek Mish Road, Mill Creek Average NCHRP717 Schoharie Creek Chipola River Mill Creek Sacramento River Montezuma Creek Average
Rock Type
Watershed Area Stream Gage Historical Cross (mi2 ) (km2 ) Data Available Sections Available
Shale/sandstone
3.34 (8.65)
No
No
Abutment
Sandstone (impure)
13.24 (34.29)
Yes
No
Abutment
Sandstone/shale Sandstone (impure)
8.40 (21.76) 2.52 (6.53)
No No
No No
Abutment Abutment
Shale/coal Sandstone (impure) Sandstone
1.46 (3.78) 1.01 (2.62) 149.31 (386.71)
No No Yes
No No No
Abutment Abutment Abutment and pier
Sandstone (impure) Sandstone Siltstone/sandstone
11.74 (30.41) 14.01 (36.29) 219.02 (567.26)
No No Yes
No No No
Abutment Abutment and pier Abutment
3.4 (8.81) 4.25 (11.01) 30.63 (79.33)
No No No
No No No
Abutment Abutment Abutment
123.42 (319.66)
Yes
No
Abutment and pier
Limestone
14.21 (36.80)
No
No
Abutment
Glacial till Limestone Siltstone Siltstone Sandstone/claystone
40 (103.6) 886 (2,295) 464 (1,202) 32.7 (97.6) 6,468 (16,752) 1,154 (2,289) 1,801 (4,665)
Yes Yes Yes Yes No
No Anecdotal Yes Yes No
Pier Pier Pier Pier Pier
Sandstone (weathered) Sandstone (impure) Sandstone (impure) Sandstone (impure)
suitable predictive surrogate for SN, should be near a value of 1.0. A value less than 1.0 would mean that the equivalent vertical scour implied by GSN for a given cumulative stream power is too large compared to the actual scour depth observed for the same cumulative stream power. A ratio value greater than 1.0 would mean GSN-derived equivalent scour depth is too small compared to the actual scour depth observed. Table 8 shows that the ratio is much greater than 1.0 for each project site (median value of over 4600). Using GSN Table 2. Field criteria used to identify mode of scour. Mode of Scour Abrasion
Quarrying and plucking
Hybrid
228
Scour Location
Field Evidence Abraded, sculpted and pitted surfaces; slaked rock material; fine-grained to granular bedload Blocky, irregular, and/or stepped channels; prominent jointing and/or bedding with spacing ∼<1 m; coarse-grained bedload composed of abundant rock fragments Combinations of the above
to predict actual scour in such cases would lead to non-conservative results (i.e., more scour would occur than predicted). The same phenomenon can be viewed graphically in Figure 12, where all of the data points except Leatherwood fall above the reference line with a slope of 1.0, indicating SN = GSN. Data reported in NCHRP 717 allowed SN/GSN ratios to be determined for three of the five sites included in that study; these ratios were calculated as 1.14, 3.52, and 183.85. The 1.14 ratio (Sacramento River, California) represents the best result in terms of successful scour prediction. The 3.52 ratio (Schoharie Creek, New York) is reasonably close to the ideal of 1.0, but the GSN was determined based on flume tests and not the modified slake durability test (Keaton et al., 2012). The SN/GSN ratio for a fourth bridge (Chipola River, Florida) was calculated as zero because no measurable scour was detected. However, Keaton et al. (2012) consider the Chipola bridge result promising as an illustration of the rock-water interaction phenomenon where available stream power was insufficient to overcome the scour resistance of the rock, as predicted by the GSN.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
Risk-Based Strategy for Scour-Affected Bridges
Ratios of SN to GSN* were also calculated for the RP-273 bridges using the GSN* values calculated by the alternative method (GSN*). Table 8 and Figure 12 show that the GSN* values agree better with the SN than do GSN values (median ratio of 189), but they are essentially uncorrelated, as indicated by an R2 value << 1.0. Using the GSN* to predict actual scour would underestimate the amount of scour at all sites except the Leatherwood Bridge, but the prediction would be better than that obtained using the GSN, and in the case of six bridges would be off by less than two orders of magnitude. Figure 7. Pure abrasion (PAB) of highly weathered sandstone at Coon Creek bridge site. Note scalloped appearance of sandstone surface. Vertical scour, here as undercutting of protruding sandstone ledge, is indicated by arrow.
Alternative Measures for SN An assessment was made of whether any relationship exists between cumulative stream power at each bridge site and alternative measures of scour, specifically depth and volume of scour and horizontal scour. None of the alternate parameters investigated appear to be more predictable with respect to cumulative stream power than does the vertical scour. DISCUSSION Relationship between GSN and Mode of Scour
Figure 8. Pure quarrying (PQP) of jointed and bedded sandstone at Roaring Creek bridge site. Vertical scour (white arrow) measured as vertical thickness of block removed at base of pier.
Figure 9. Scour by subequal abrasion and quarrying (AB/QP) at Paden Fork bridge site. Areas of abrasion (A) and quarrying (Q) are indicated. The vertical scour (arrow) is measured as the distance from base of abutment to base of scoured interval.
Table 3 serves as a conceptual model relating mode of scour and rock durability in which less durable (i.e., degradable) rocks tend to scour by abrasion and more durable rocks tend to scour by quarrying. As noted previously, the majority of bridge sites (10 of 15) evaluated in this study appeared to have been affected over time by hybrid modes of scour, with some degree of both abrasion and quarrying evident, and such sites appear to be characterized by rocks of intermediate and/or variable durability. Thus, scour mode at such sites could be either time-dependent or controlled by site-scale geologic heterogeneity. GSN and GSN* are quantitative measures of durability independent of scour mode. Thus, for RP-273, it is reasonable to consider the relationship between the mode of scour and the magnitude of GSN*. Such a correlation would be valuable in that interpretation of scour mode can be both difficult and subjective; if GSN* could be shown to correlate consistently with scour mode, it could be used to help discern the presence of particular scour mode(s). Figure 13 illustrates a plot of median GSN* for each mode of scour (the median is equal to a single, but presumably representative, GSN* value for categories in which only one site exists). A general trend exists in this figure toward higher GSN* with increasing dominance of abrasion. This result is not unexpected; after all, GSN and GSN* both measure abrasion resistance, and GSN was designed
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
229
Niemann, Wait, and Keaton Table 3. Summary of scour mode and rock type. Durable Sandstone
<<< Limestone
Rock Type Siltstone
Scour Mode PAB DAB AB/QP
Non-durable
Shale/Claystone
Sandstone— Weathered Coon Creek
Bridge Fork; Clear Fork; Laurel Fork; Paden Fork; Ritter Park; Leatherwood Cucumber; Grassy Run
DQP PQP
Sandstone—Impure
>>>
Audra Park; Roaring Creek
Mish
Caldwell Run Leatherwood
Beverly
specifically as a predictor of scour at sites at which abrasion is the mode of rock scour. Figure 13 indeed suggests that sites with higher GSN* (i.e., those that are less resistant to abrasion) are more likely to be affected by this type of scour. However, the data also reveal a notable discrepancy to this general trend: the median GSN* is higher for sites where pure quarrying (PQP) is the mode of scour than for sites where quarrying is accompanied by a comparable degree of abrasion (AB/QP) or lesser degree of abrasion (DQP). This apparent discrepancy can possibly be explained by recognition that rock scour results from rock-water interaction, meaning that both rock and hydraulic char-
acteristics must be considered (Keaton et al., 2012). For example, if fractures are closely spaced, the size of rock blocks is relatively small and the stream energy necessary for quarrying to occur is small compared to sites at which fractures are more widely spaced and blocks are larger. If rocks with a smaller, but still significant, abrasion resistance (higher relative GSN*) happen to have a high fracture density, quarrying could occur at relatively low stream velocities and not necessarily be accompanied by abrasion. If more durable rocks (lower relative GSN*) happen to have a low fracture density, the significant velocities necessary for quarrying large blocks may be attained only on an irregular basis, but
Figure 10. Geographic distribution of scour mode plotted on a simplified geologic map of West Virginia. Sites dominated by quarrying and plucking (PQP) tended to be located in Allegheny Front and Valley and Ridge Province in the eastern portion of the state, whereas some degree of abrasion was characteristic of sites in the Appalachian Plateau in the central and western portions of the state.
230
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
Risk-Based Strategy for Scour-Affected Bridges Table 4. Scour extent and mode at each bridge site. “NM” indicates locations for which measurements could not be obtained because of inaccessibility or safety concerns. “NA” indicates bridges without abutments. “?” indicates accurate measurements could not be obtained. Scour Extent
Site
Mode of Scour
Audra
PQP
Beverly
PQP
Bridge Fork Caldwell Run Clear Fork
AB/QP DAB AB/QP
Coon Creek Cucumber Grassy Run Laurel Fork Leatherwood—sandstone Leatherwood—shale Little Sandy Mish Paden Fork Ritter Park/Fifth Street
PAB DQP DQP AB/QP AB/QP AB/QP NA PQP AB/QP AB/QP
Roaring Creek
PQP
Vertical (in.) (mm)
Horizontal (ft) (m)
Location
Min
Max
Min
Max
Min
Max
Abutment Pier Abutment Pier Abutment Abutment Abutment Pier Abutment Abutment Abutment Abutment Abutment Abutment Abutment Abutment Abutment Abutment Pier Abutment Pier
30 (762) 1 (25) 4 (102) NM 2.5 (64) 12 (305) 3 (76) 4 (102) 1 (25) NM 12 (305) 2 (51) 5 (127) 7 (178) NA 2 (51) 15 (381) NA 8 (203) 6 (152) 1.5 (38)
40 (1,016) 2 (51) 6 (152) NM 4.5 (114) 20 (508) 3 (76) 4 (102) 5 (127) NM 12 (305) 4 (102) 12 (305) 14 (356) NA 4 (102) 15 (381) NA 10 (254) 7 (178) 2.5 (0.8)
40 (12.2) 2 (0.6) 10.25 (3.1) NM 15 (4.6) 12.5 (3.8) 3 (0.9) 2 (0.6) 1 (0.3) NM 12 (3.7) 24 (7.3) 18 (5.5) 5 (1.5) NA 15 (381) ? NA 10 (3) 2 (0.6) 3.5 (1.1)
40 (12.2) 2 (0.6) 10.25 (3.1) NM 15 (4.6) 24 (7.3) 3 (0.9) 2 (0.6) 5.5 (1.7) NM 12 (3.7) 24 (7.3) 18 (5.5) 5 (1.5) NA NM 6.7 (2) NA 12 (3.7) 2.5 (0.8) 12.5 (3.8)
114 (2,896) 4 (102) 4 (102) NM 1 (25) 9 (229) 2.5 (64) 1 (25) 1 (25) NM 5 (127) 2 (51) ? ? NA NM 12 (305) NA ? 3 (76) 1.5 (38)
138 (3,505) 5 (127) 12 (305) NM 6 (152) 30 (762) 2.5 (64) 8 (203) 10 (254) NM 18 (457) 10 (254) ? ? NA NM 12 (305) NA ? 10 (254) 2.5 (0.8)
stream energy would also likely be sufficient to perform abrasion. GSN* as a Potential Predictor of Scour in West Virginia Although the results obtained using the GSN* may appear more promising than those obtained using GSN in terms of better matching the SN for RP-273
Depth (in.) (mm)
sites, some important qualifications are associated with these results. First, the GSN is part of the NCHRP 717 methodology for predicting scour by PAB. As a modified version of GSN, this also should be true of GSN*. As such, it is encouraging that three of the bridges for which the ratio of Scour Number to GSN* is less than 100 include one bridge characterized by PAB (poorly consolidated sandstone at Coon Creek) and three others characterized by AB/QP (sandstone
Table 5. GSN and GSN* values calculated from modified slake durability tests. Bridge Site Paden Fork Coon Creek Roaring Creek Caldwell Run Mish Beverly Laurel Bridge Fork Cucumber Grassy Run Ritter Park Little Sandy Audra Park Leatherwood (shale) Leatherwood (sandstone) Clear Fork Median values
GSN, ft-lb/s/ft2 (W/m2 )
R2
GSN*, ft-lb/s/ft2 (W/m2 )
R2 *
GSN*/GSN
R2 */R2
4.42E-05 (6.45E-04) 8.99E-06 (1.31E-04) 1.04E-05 (1.52E-04) 1.04E-05 (1.52E-05) 2.89E-06 (4.22E-05) 1.40E-05 (2.04E-05) 1.18E-05 (1.72E-04) 5.60E-06 (8.17E-05) 1.79E-05 (2.61E-04) 6.47E-06 (9.44E05) 7.46E-05 (1.09E-03) 5.74E-05 (8.38E-04) 5.55E-06 (8.10E-05) 4.69E-05 (6.84E-04) 1.70E-05 (2.48E-04) 4.05E-06 (5.91E-05) 1.11E-05 (1.62E-04)
0.31 0.03 0.12 0.17 0.00 0.15 0.14 0.09 0.34 0.11 0.05 0.84 0.03 0.34 0.30 0.04 0.13
7.84E-05 (1.14E-03) 5.32E-04 (7.76E-03) 1.38E-04 (2.01E-03) 1.86E-04 (2.71E-03) 2.07E-04 (3.02E-03) 1.32E-04 (1.93E-03) 2.06E-04 (3.01E-03) 1.39E-04 (2.03E-03) 1.08E-04 (1.58E-03) 1.17E-04 (1.71E-03) 8.68E-05 (1.27E-03) 8.04E-05 (1.17E-03) 2.41E-04 (3.52E-03) 8.97E-04 (1.31E-02) 1.48E-04 (2.16E-03) 2.25E-04 (3.28E-03) 1.43E-04 (2.09E-03)
0.45 0.88 0.80 0.92 0.82 0.79 0.63 0.71 0.87 0.63 0.70 0.94 0.85 0.81 0.73 0.77 0.79
1.8 59.2 13.3 17.9 71.6 9.4 17.5 24.8 6.1 18.1 1.2 1.4 43.4 19.1 8.7 55.6 17.7
1.5 29.3 6.4 5.4 189.4 5.4 4.5 7.7 2.6 6.0 13.8 1.1 25.9 2.4 2.4 18.6 5.7
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
231
Niemann, Wait, and Keaton Table 6. Estimated cumulative daily effective stream power at each bridge site since it was constructed.
Site 1—Leatherwood 2—Fifth Street 2—Fifth Street (Pier) 3—Little Sandy 4—Grassy Run 5—Caldwell Run 6—Paden Fork 7—Audra Park 7—Audra Park (Pier) 8—Laurel Fork 9—Roaring Creek 9—Roaring Creek (Pier) 10—Beverly 10—Beverly (Pier) 11—Coon Creek 12—Bridge Fork 13—Cucumber 14—Clear Fork 14—Clear Fork (Pier) 15—Mish Road
Cumulative Daily Effective Stream Power Since Construction, ft-lb/s/ft2 (W/m2 )
Bridge Age (years)
Cumulative Daily Effective Stream Power Per Year, ft-lb/s/ft2 / year (W/m2 /year)
3.7 (54) 33.2 (469.9) 95.9 (1400) 3.0 (43.8) 0.5 (7.3) 11.5 (167.8) 4.6 (67.1) 159.0 (2,320) 459.4 (67.0) 0.5 (7.3) 184.6 (2,694) 533.5 (7786) 15.2 (221.8) 44.0 (642.1) 53.6 (782.2) 5.9 (86.1) 3.1 (45.2) 323.1 (4,715.3) 933.7 (13,626.3) 6.9 (100.7)
24 90 90 34 62 87 38 71 71 7 81 81 16 16 32 6 23 94 94 85
0.15 (2.16) 0.37 (5.22) 1.07 (15.55) 0.09 (1.29) 0.01 (0.12) 0.13 (1.93) 0.12 (1.77) 2.24 (32.68) 6.47 (94.4) 0.07 (0.10) 2.28 (33.26) 6.59 (96.12) 0.95 (13.86) 2.75 (40.13) 1.68 (24.4) 0.98 (14.35) 0.13 (1.97) 3.44 (50.16) 9.93 (144.96) 0.08 (1.19)
at Ritter Park/5th Street, shale at Leatherwood, and sandstone at Clear Fork). This is broadly consistent with NCHRP 717, which shows the most favorable SN/GSN ratios for less durable earth materials (i.e., higher relative GSN for marine siltstone at Sacramento and glacial till at Schoharie). These results may suggest Table 7. Scour number (vertical scour depth/cumulative daily effective stream power) at each site. Scour Number (ft of vertical Scour Number scour/daily (m of vertical ft-lb/s/ft2 ) scour/daily W/m2 )
Site 1—Leatherwood—sandstone 1—Leatherwood—shale 2—Fifth Street 2—Fifth Street (pier) 3—Little Sandy 4—Grassy Run 5—Caldwell Run 6—Paden Fork 7—Audra Park 7—Audra Park (pier) 8—Laurel Fork 9—Roaring Creek 9—Roaring Creek (pier) 10—Beverly 10—Beverly (pier) 11—Coon Creek 12—Bridge Fork 13—Cucumber 14—Clear Fork 14—Clear Fork (pier) 15—Mish Road
232
1.9E-01 2.4E-01 N/A 7.8E-03 N/A 2.0E + 00 1.2E-01 2.7E-01 1.8E-02 2.7E-04 5.5E-01 2.9E-03 3.1E-04 2.7E-02 N/A 4.7E-03 5.0E-02 N/A 7.7E-04 3.6E-04 3.6E-02
4.0E-03 5.0E-03 N/A 1.6E-04 N/A 4.2E-02 2.5E-03 5.6E-03 3.5E-04 5.6E-06 1.1E-02 6.1E-05 6.5E-06 5.6E-04 N/A 9.8E-05 1.0E-03 N/A 1.6E-05 7.5E-06 75E-04
that the mode of scour at these sites was closest to the pure abrasion for which the NCHRP 717 methodology was developed. A contradictory result is that two of the six RP-273 sites for which the SN/GSN* ratio is less than 100 are bridges for which the mode of scour was determined to be pure quarrying (Audra Park and Roaring Creek). Because no known theoretical basis is recognized for claiming that GSN* would apply to any degree of quarrying/plucking, it would be premature to imply that the current results suggest that GSN* could be useful as a reliable predictor of scour at such sites. Perhaps an apparent threshold effect exists whereby rocks possess a durable component (e.g., quartz sand particles or strong silica cementation) but still scour by abrasion because of a degradable attribute (e.g., clay or weak cementation). Such a condition would be distinct from that of more uniform and less degradable rocks that possess a true threshold value of turbulence intensity because they possess a relatively high degree of resistance to abrasion scour that tends to be most effective when applied repeatedly day after day, year after year. Scour of durable rock blocks that occurs during the peak discharge of a rare flood results in a scour depth of the dimension of one or two rock blocks; decades of gradual scour might be required to wear away that dimension of degradable rock, but the effects of the rare peak discharge on the degradable rock might be comparable to a few days or possibly a few weeks of abrasion caused by average discharge. The apparent threshold effect could help explain better regression fits to modified slake durability data using non-zero
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
Risk-Based Strategy for Scour-Affected Bridges
Figure 11. Vertical scour (abutment) versus cumulative daily excess stream power for each site. Excess stream power refers to power that discounts the contribution of the 2-year return period event. Lack of a linear relationship indicates that the abrasion model of Keaton (2015) for degradable rocks does not apply to the bridges studied for RP-273.
intercepts (i.e., GSN*). This phenomenon needs further investigation. In addition, calculation of the SN required the assumption that the time factor for calculating SN is the entire time since the bridge was constructed. It is reasonable to consider that the observed scour, particularly the quarrying mode, may have occurred over a period of a few years or even in response to the largest ďŹ&#x201A;ood of record as a single short-duration event. Such a condition would complicate the SN calculation and add uncertainty to its interpretation. The actual scour history might be dominated by a few years immediately following construction during which rock damaged or
loosened during construction is removed by quarrying. The time since removal of loosened blocks could have been a period of relative stability, in which case the calculated SN might be meaningless. More systematic observation and documentation of rock conditions at bridge sites are needed.
RECOMMENDED RISK-BASED APPROACH TO SCOUR EVALUATION The results of RP-273 indicate that the quantitative method presented by Keaton et al. (2012) for analyzing
Table 8. Values of GSN, GSN*, and ratios of these two parameters to the Scour number.
Site Audra Beverly Bridge Fork Caldwell Run Clear Fork Coon Creek Grassy Run Laurel Fork Leatherwood Leatherwood Mish Paden Fork Ritter Park/Fifth Street Roaring Creek
Scour Mode
Scour Number, ft/ daily ft-lb/s/ft2 (m/daily W/m2 )
GSN, ft/daily ft-lb/s2 /ft2 (m/daily W/m2 )
Scour Number/ GSN
GSN* (ft/daily ft-lb/s2 /ft2 ) (m/daily W/m2 )
Sandstone Siltstone Sandstone Shale Sandstone Sandstone (weathered) Sandstone Sandstone Sandstone (shaley) Shale Limestone Sandstone Sandstone
PQP PQP AB/QP DAB AB/QP PAB
1.80E-02 (3.76E-04) 2.70E-02 (5.64E-04) 5.00E-02 (1.04E-03) 1.20E-01 (2.51E-03) 7.70E-04 (1.61E-05) 4.70E-03 (9.82E-05)
5.55E-06 (1.16E-07) 1.40E-05 (2.92E-07) 5.60E-06 (1.17E-07) 5.60E-06 (1.17E-07) 4.05E-06 (8.46E-08) 8.99E-06 (1.88E-07)
3,243 1,929 8,929 21,429 190 523
2.41E-04 (5.03E-06) 1.32E-04 (2.76E-06) 1.39E-04 (2.90E-06) 1.86E-04 (3.88E-06) 2.25E-04 (4.70E-06) 5.32E-04 (1.11E-05)
DQP AB/QP AB/QP
2.00E + 00 (0.04) 5.50E-01 (0.011) 1.90E-01 (3.97E-03)
1.79E-05 (3.74E-07) 6.47E-06 (1.35E-07) 1.18E-05 (2.46E-07)
111,732 85,008 16,102
1.08E-04 (2.26E-06) 1.17E-04 (2.44E-06) 2.06E-04 (4.30E-06)
AB/QP PQP AB/QP AB/QP
3.35E-05 (7.00E-07) 3.60E-02 (7.52E-04) 2.70E-01 (5.64E-03) 7.80E-03 (1.63E-04)
4.69E-05 (9.80E-07) 2.89E-06 (6.04E-08) 4.42E-05 (9.23E-07) 7.46E-05 (1.56E-06)
0.7 12,457 6,109 105
8.97E-04 (1.87E-05) 2.07E-04 (4.32E-06) 7.84E-05 (1.64E-06) 8.68E-05 (1.81E-06)
0.04 174 3,444 90
Sandstone
PQP Median values
2.90E-03 (6.06E-05) 3.15E-02 (6.58E-04)
1.04E-05 (2.17E-07) 9.70E-06 (2.03E-07)
279 3,247
1.38E-04 (2.88E-06) 1.63E-04 (3.40E-06)
21 189
Dominant Rock Type
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221â&#x20AC;&#x201C;241
Scour Number/ GSN* 75 205 360 645 3.4 8.8 18,467 4,701 922
233
Niemann, Wait, and Keaton
Figure 12. Regressions of GSN and GSN* on scour number. Negative slopes have no physical meaning. Most data points plot above line with Slope = 1 indicating a non-conservative result, i.e., more scour observed than would be predicated.
scour of degradable rock material based on properties of the rock, as well as characteristics of the flowing water, is not directly applicable to the rock formations and drainage basins in West Virginia and likely similar areas in adjacent states. However, trends in GSN and GSN* for different modes of scour suggest that the
potential for larger amounts of scour may be identified in a way that could be helpful for departments of transportation hoping to manage potential scour at bridge sites. This scour management approach is expressed as a proposed risk-based screening strategy that considers both evidence of scour and its dominant
Figure 13. Median GSN* according to mode of scour (GSN* is in units of feet per foot-pounds per second per square feet).
234
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
Risk-Based Strategy for Scour-Affected Bridges Table 9. Scour-based categories in proposed risk-based screening of bridges. Stream Power Group Scour Tier I, Evidence of no scour II, Abrasion dominant III, Quarrying dominant
A, Low Power
B, Moderate Power
C, High Power
II A III A
No further action required in this tier II B III B
II C III C
mode as well as flow conditions expressed in terms of cumulative stream power. The risk-based screening approach results in a bridge being placed into one of three tiers: Tier I is associated with evidence of no significant scour over a period of years that is meaningfully long; thus, no further action is required at the present time; Tier II is associated with evidence of abrasion as the dominant mode of scour; and Tier III is associated with evidence of quarrying as the dominant mode of scour. For Tier II, the timerate and depth of scour can be estimated according to the procedure of Keaton et al. (2012) in NCHRP 717. For Tier III, a detailed analysis of site-specific conditions is recommended because no proven method is currently available for estimating the time-rate and depth of scour. For bridges placed into Tiers II or III, an evaluation of cumulative stream power is performed, and average annual cumulative stream power is estimated. Within Tier II or III, bridge sites can be further subdivided into Level A, B, or C, representing low, medium, and high stream power, respectively, through an evaluation of 1) a site’s average annual cumulative stream power relative to a regional distribution of such average annual cumulative stream powers and, 2) where available, a comparison of a site’s existing scour behavior to the scour behavior of other sites with similar characteristics (e.g., scour mode, block size, joint spacing, joint orientation, etc.). Over time, as an increasing number of bridge sites can be evaluated and scour behavior cataloged relative to average annual cumulative stream power, trends will likely emerge that help identify minimum cumulative stream power values below which no scour is observed. Table 9 summarizes the proposed risk-based screening. A flow diagram or decision tree developed from this project, presented in Figure 14, is intended to aid the user in placing a bridge in question into the appropriate tier. The decision tree utilizes a set of criteria based on the discussion and parameters above, as well as additional criteria from Keaton et al. (2012). Most importantly, criteria incorporated into the decision tree are designed to reflect the geotechnical and geologic conditions encountered in the field or in the laboratory for rocks present at the 15 West Virginia bridges. In order to validate the decision trees, the
criteria were applied to the all 15 West Virginia bridge sites, and scores were assigned to each site. The decision tree correctly identified the mode of scour at all 14 West Virginia bridge sites at which scour was observed (Table 10). CONCLUSIONS The primary conclusions that follow from the research results described in this article relate to the following:
• The great majority of bridge sites evaluated as part of RP-273 were located in small watersheds without stream flow data and founded on durable and jointed, blocky, sedimentary rock. These factors compromised the application of the NCHRP method, which is intended for gradual and progressive wear of degradable rock. For this reason, the results from this project do not support the generalized implementation of the NCHRP 717 approach in West Virginia. It is reasonable to assume that the same would be true in other parts of the Appalachian region in which rock types and hydraulic conditions are similar to those observed in West Virginia, in particular rock-water interactions that resulted in quarrying as a significant mode of scour. • The modified slake durability test and the parameter derived from it (GSN) comprise a measure of abrasion scour and thus do not provide useful information on the magnitude or scour rate for durable rocks that scour by quarrying. • The capacity to determine cumulative stream power is requisite to computing SN and thus to applying any methodology, including NCHRP 717, that depends on projecting future energy conditions at a site. For RP-273, in the absence of gaging or precipitation data, stream power was estimated based on watershed models, probabilistic representations of precipitation, and the computational steps required to determine average annual cumulative effective stream power. • Recognition that the NCHRP approach to predicting rock scour is inappropriate for most West Virginia bridge sites led to the design of a risk-based
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
235
Niemann, Wait, and Keaton
Figure 14. A–C) Flow diagram for assessing rock scour based on field, general geologic, and laboratory criteria. A and B are based on field criteria that are applied to any bridges site; C is based on laboratory criteria, which are applied depending on results of the field criteria.
236
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
Risk-Based Strategy for Scour-Affected Bridges
Figure 14. (Continued).
screening strategy. The proposed screening is an implementable product allowing for the grouping and prioritizing of bridge sites according to scour mode and cumulative stream power. The screening tool includes guidance and methodology for assessing mode(s) of scour at bridge sites in West Virginia and recommends action(s) to be taken based on the results of the screening. Although the approach does
not provide a direct quantitative prediction of scour like that afforded by NCHRP 717, it does allow for scour by multiple modes and recommends use of NCHRP 717 in cases in which scour is exclusively by abrasion of degradable rock. Based on the survey of other states, the proposed risk-based scheme is a more detailed, science-based approach than is being used elsewhere and could potentially serve as a model
Table 10. Assessment and distribution of average annual cumulative stream power at the 15 sites in RP-273. Average Annual Cumulative Stream Power, average annual , ft-lb/ft2 (J/m2 ) Group A, Low, average annual < 1,000 ft-lb/ft2 ( average annual < 14,594 J/m2 ) Grassy Run
686 (10,011)
Group B, Medium, 1,000 < average annual < 15,000 ft-lb/ft2 (14,594 < average annual < 218,909 J/m2 ) Laurel Fork Mish Road Little Sandy Paden Fork Caldwell Run Cucumber Leatherwood Average
5,642 (82,339) 7,015 (102,376) 7,560 (110,330) 10,436 (152,302) 11,464 (167,305) 11,669 (170,296) 13,263 (193,559) 9,578 (139,780)
Group C, High, average annual > 15,000 ft-lb/ft2 ( average annual > 218,909 J/m2 ) Fifth Street Ritter Park Bridge Fork Beverly Coon Creek Audra Park Roaring Creek Clear Fork Average
31,871 (465,122) 54,444 (794,550) 82,293 (1,200,976) 144,645 (2,110,935) 193,437 (2,829,948) 196,913 (2,873,729) 296,961 (4,333,820) 142,938 (2,086,023)
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221â&#x20AC;&#x201C;241
237
Niemann, Wait, and Keaton
for transportation agencies seeking to advance their management of rock scour. ACKNOWLEDGMENTS The authors of this report gratefully acknowledge the technical assistance, support, and insight provided by the project Technical Advisory Committee at the West Virginia Department of Transportation, including Joe Carte, P.E.; Douglas Kirk, P.E., C.F.M.; and George Chappell, M.S. The project coordination and monitoring provided by Donald Williams, P.E., and Michael Pumphrey, P.E., are also gratefully acknowledged. The authors also express appreciation for the valuable project contributions provided by Wael Zatar, Ph.D., and Kimberley Baker of the Rahall Transportation Institute, and by Nicholas Koutsunis, hydraulics consultant. Also acknowledged are the contributions from Brian Greene, P.G.; Terry Downs, P.G.; Nicole Vetter, P.G.; and Jared Govi of Gannett Fleming, Inc. Undergraduate students who provided assistance with modified slake durability testing were Andrew Harless, Michael Niemann, and Jennifer Niemann. Undergraduate students Chris Brumfield and Eli McWhorter performed surveying of stream channels and bridges in support of the hydrology and hydraulics analyses. APPENDIX A Rock Coring and Sample Collection Drilling for this project was conducted in June and July 2011. The first 10-ft (3-m) depth interval of rock below the base of each bridge abutment or pier was designated as the target zone, which in the case of horizontally bedded rocks corresponded to the same stratigraphic interval as scoured areas (Figure 2). At the Beverly Bridge, beds were steeply inclined, and this stratigraphic objective could not be met. In general, some combination of pavement, fill, native soil, and bedrock had to be drilled above the target zone. Hollow-stem augers were used to drill unconsolidated materials, followed by rock coring adjacent to each abutment using an NX wireline split-tube core barrel. The depths of individual boreholes ranged from 17 to 34 ft (5.2 to 10.4 m) below grade. A total of 694 ft (212 m) of drilling, including 30 cores representing 394 ft (120 m) of rock coring, was conducted for the project. At each bridge site one additional rock sample was collected from surface exposures in the channel at or immediately adjacent to the bridges. Samples from both of the cores plus one surface sample at each of the 15 bridges yielded a total of 45 rock samples. Rock samples for the NCHRP 717 project were collected mainly from surface exposures; core samples were also used for the Sacramento River and Mill Creek Bridges (Keaton, 2013). 238
Upon extraction, cores were immediately logged, photographed, and evaluated for RQD according to ASTM D6032 (2008b). Each 5-ft (1.5-m) length was double-wrapped, first in a 5-mil (0.127-mm) wax film (Parafilm or equal), followed by a heavy-gauge sheath of flexible plastic. The package was secured with duct tape and labeled, then placed in a protective core box for transport. These measures ensured that the fieldmoisture content of the core would be maintained until the modified slake durability testing could be performed. Modified Slake Durability Testing and Calculation of Geotechnical Scour Number (GSN) Rock core samples were subjected to testing for modified slake durability (continuous abrasion) testing in general accordance with methodology developed for NCHRP 717, which was based on a procedure developed by Dickenson and Baillie (1999). The modified test is meant to simulate abrasive action on the sample pieces similar to that which a bedrock stream channel would experience from the interaction with bedload. Sample equipment and sample preparation for the modified test are identical to those used for the standard test (ASTM D4644, 2008a). Procedural modifications include the following: 1) 60-minute submerged tumbling cycles (versus 10 minutes in the standard test); 2) elimination of oven drying between submerged tumbling cycles; 3) use of saturated-surface dry weights (ASTM D6473, 2010) rather than oven dry weights to calculate sample loss through time; and 4) measurement of sample unit weight. The elimination of drying and use of wet weights in the modified test reflects the condition that rock scour occurs below the water level of the stream and that rocks are never in a completely dry state. Furthermore, some rocks that may be susceptible to slaking in air would be destroyed by a single cycle of oven drying without providing useful information about scour by flowing water, as discovered by Dickenson and Baillie (1999) and documented by Keaton et al. (2012). The raw test data consist of wet sample weights after each submerged tumbling cycle and the cumulative times of the cycles. The test is continued for the number of 60-minute cycles required until the late-stage data define a linear trend, per Dickenson and Baillie (1999); depending on the durability of the sample. For RP-273, this required anywhere from three to seven cycles (180–480 minutes) of continuous abrasion. Figure 3 shows example raw test data. The incremental sample loss (sample mass lost to abrasion over each tumbling cycle) was then used to calculate a linear dimension by 1) determining the unit weight of the rock sample; 2) converting the loss in
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
Risk-Based Strategy for Scour-Affected Bridges
sample mass to a volume by dividing it by the unit weight; and 3) normalizing the volume by a unit area to obtain an “equivalent scour depth” (Keaton, 2013). An “equivalent cumulative stream power” was also calculated for each tumbling cycle per Keaton (2013): 1) the initial wet sample mass was normalized to 500 g so that results from different tests could be compared in terms of energy; 2) the average sample mass per cycle was determined; 3) the total equivalent distance along which the sample “traveled” for each cycle was calculated; 4) the product of the mass and distance was divided by the duration of the tumbling cycle in seconds; and 5) the result was normalized by the area of the sample drum in which the sample fragments reside during the test. The late-time data from a modified slake durability test defines a quasi-linear trend on a plot of equivalent scour depth versus equivalent cumulative stream power. When the linear portion of the plot is considered, a slope can be defined for each sample tested (Figure 5). This slope is referred to as the “GSN” (Keaton, 2013) and can be expressed as scour per unit of cumulative stream power (e.g., feet or meters of scour per unit of power). Higher GSN values (i.e., steeper slopes) reflect a greater amount of scour, or abrasion, per unit cumulative stream power and are characteristic of less abrasion-resistant samples, whereas lower values (i.e., flatter slopes) reflect a lesser amount of scour, or abrasion, for the same expenditure of cumulative stream power and are characteristic of more abrasion-resistant samples. Keaton et al. (2012) calculated the slope that passed through the origin as the GSN, whereas Keaton (2013) realized that weak degradable rocks could be differentiated from more durable rocks based on the intercept of the linear regression. A near-zero intercept was associated with degradable rocks, whereas a negative intercept was associated with durable rocks. The GSN becomes the basis by which the scour potential of different rock samples can be compared when scour is occurring gradually and progressively, as is the case with abrasion of degradable rock. Since a single or dominant mode of scour was not always evident from field inspection during RP-273, GSN was calculated for all bridge sites evaluated for this project. Regardless of scour mode, the GSN serves as a useful means of comparing the abrasion resistance of different rock types. Keaton (2013) suggested that GSN might be useful as an index parameter for weak rocks. The GSN was calculated in two ways for this project: 1) by forcing a zero-intercept for a trend line and 2) by calculating a best-fit line (i.e., not forcing a zero-intercept; referred to hereafter as GSN*). The former method follows the procedures specified by Keaton et al. (2012) and Keaton (2013), which are meant for non-durable (i.e., less abrasion-resistant)
rocks that trend toward near-zero values of equivalent steam power and equivalent scour depth. In such cases, forcing a zero-intercept both reflects the data and produces a relatively strong linear correlation (Figure 6). However, scour of rock at the majority of West Virginia bridges studied occurs by some degree of quarrying and plucking of durable, jointed rock, rather than by abrasion of degradable rock. In fact, a defining characteristic of durable rock is its resistance to abrasion. When subjected to a modified slake durability procedure, durable rocks retain significant mass, and, thus, substantial stream power is maintained throughout the duration of the test. Consequently, a best-fit line (i.e., linear regression) on the data has a negative intercept and intersects the xaxis (stream power axis) at a significant positive value (Figure 6). This x-axis intercept is known as the “implied threshold value,” and physically it may represent the minimum cumulative stream power needed to initiate abrasion scour, but the exact interpretation is unclear (Keaton et al., 2012). For the 45 rock samples tested for RP-273, the R2 was 0.13 for GSN and 0.79 for GSN*. This result is attributable to the presence of durable rock at a majority of the West Virginia bridges studied. Keaton et al. (2012) recognized the threshold effect for the limestone exposed at the Chipola River Bridge by calculating an alternative, non–zerointercept GSN equivalent to GSN*. Hydrologic and Hydraulic Analysis The underlying methods utilized to compute cumulative stream power in this research project are based on those described by Keaton et al. (2012). As illustrated in Equation 1, this approach necessitates a characterization of flow velocity, water depth, an estimate of the channel roughness, and the selection of a turbulencerelated velocity-enhancement factor that corresponds to the vorticity generated by flow around bridge piers to compute “instantaneous stream power”: nKpV 2 ␥ × (1) × V, P= 1.486 y0 1/3 where P = instantaneous stream power (ft-lb/s/ft2 ; W/m2 ); n = Manning’s roughness coefficient; Kp = turbulence-related velocity enhancement factor (1.0 for approach flow, 1.5 for flow around round-nosed piers, 1.7 for flow around square-nosed piers); V = flow velocity (ft/s; m/s); 1.486 = factor for U.S. customary units (ft1/3 /s) (1.0 factor for metric units [m1/3 /s]); ␥ = unit weight of water, = (62.4 lb/ft3 ; 9.802 kN/m3 ); and y0 = depth of approach flow (ft; m). Daily flow rates from stream gauging stations Kaeton et al. (2012) obtained average but since the sites
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
239
Niemann, Wait, and Keaton
under examination for RP-273 were generally much smaller (average watershed area = 40 mi2 [103.6 km2 ] vs. 1801 mi2 [4665 km2 ] in the NCHRP project) and ungaged, these same gage-based flow data were not available for cumulative stream power calculations in this study. Therefore, a number of different watershed modeling techniques were utilized to obtain flow rate estimates at each site, including HEC-HMS (USACE, 2010), HEC-1 (USACE, 1998), National Streamflow Statistics (USGS, 2007), and TR-55 (NRCS, 1986). After comparisons of model output at each site, HEC1 watershed models and corresponding runoff hydrographs were ultimately utilized as the basis for flow rate determination, and an HEC-RAS hydraulic model at each site was used to determine flow velocity and depth corresponding to each desired return-period flow rate. Watershed boundaries were delineated using digital elevation maps from the National Elevation Dataset (USGS, 2009) and the TOPAZ computational tool developed by Garbrecht and Martz (1999). Curve numbers were determined using the National Land Cover Database (Fry et al., 2011) data and NRCS (2013) Soil Survey Geographic soil type data, and in accordance with classifications described in NRCS (2010), Part 630 Hydrology—National Engineering Handbook. Time of concentration was determined by the Soil Conservation Service method, incorporating computations of maximum flow distance within a watershed, average watershed slope, and curve number. The National Oceanic and Atmospheric Administration (NOAA, 2013) was utilized to obtain precipitation depths for storms associated with average return periods of 2, 5, 10, 25, 50, 100, 200, and 500 years and was then input into watershed models to yield runoff hydrographs. The watersheds under study in this project have areas as small as 1 mi2 (2.59 km2 ), with average basin overland slopes of 10–98 percent. These conditions yield runoff that has been described as “flashy,” meaning that the runoff hydrograph has a steep rising limb, and flow likewise recedes quickly at the end of a storm. Where watershed modeling has to be used to estimate stream discharge, it was reasoned that many days of no flow results in problems for calculating cumulative stream power on the basis of average daily flow. At larger watersheds, such as those described in NCHRP 717, less variation might be expected between average daily flow rate reported in streamflow data and the peak flow rate during each day. The smallest of the watersheds studied by Keaton et al. (2012) was Mill Creek (33 mi2 ; 85.47 km2 ), which exhibited flashy discharge. The stream flow was modeled from gage data, but even the 500-year discharge hydrograph duration was less than 24 hours. However, the much smaller watersheds in RP-273 have additional uncertainty related to the lack of stream gage data and the possibility of spring-
240
fed base flow. Therefore, the runoff hydrographs yielded by watershed modeling in this study were not processed using the NCHRP procedures for use in cumulative stream power calculations in RP-273, and instead cumulative stream power was calculated using the same 15-minute intervals for which watershed runoff flow data (i.e., HEC-1–generated hydrographs) were available. “Cumulative stream power” (ft-lb/ft2 ; J/m2 ), , was determined for each storm event by multiplying the “instantaneous stream power” (ft-lb/s/ft2 ; W/m2 ) experienced during each interval by the duration of the runoff hydrograph interval (i.e., 900 seconds) between reported flow rates. When utilized for calculation of SN, cumulative stream power was converted into units of cumulative daily stream power over the period since the bridge was built to maintain consistency with the units utilized in GSN and GSN*. A detailed explanation of the steps used to calculate cumulative stream power is given in the RP-273 Project Report (2013). REFERENCES ARNESON, L. A.; ZEVENBERGEN, L. W.; LAGASSE, P. F.; AND CLOPPER, P. E., 2012, Hydraulic Engineering Circular No. 18: Evaluating Scour at Bridges, 5th ed.: Federal Highway Administration, FHWA-HIF-12-003-HEC-18. AMERICAN SOCIETY FOR TESTING AND MATERIALS (ASTM), ASTM D4644, 2008a, Standard Test Method for Slake Durability of Shales and Similar Weak 612 Rocks: American Society for Testing and Materials Procedure D4644-08. ASTM, ASTM D6032, 2008b, Standard Test Method for Determining Rock Quality Designation (RQD) of Rock Core: American Society for Testing and Materials Procedure D6032-08. ASTM, ASTM D6473, 2010, Standard Test Method for Specific Gravity and Absorption of Rock for Erosion Control: American Society for Testing and Materials Procedure D6473-10. ASTM, ASTM D2113, 2014, Standard Practice for Rock Core Drilling and Sampling of Rock for Site Exploration: American Society for Testing and Materials Procedure D2113-14. CARTE, J., 2011, personal communication, West Virginia Department of Highways, Geotechnical Unit, Charleston, WV. DICKENSON, S. E. and BAILLIE, M. W., 1999, Predicting Scour in Weak Rock of the Oregon Coast Range: unpublished research report, Department of Civil, Construction, and Environmental Engineering, Oregon State University, Corvallis, OR, Final Report SPR 382, Oregon Department of Transportation and Report FHWA-OR-RD-00-04. FRY, J.; XIAN, G.; JIN, S.; DEWITZ, J.; HOMER, C.; YANG, L.; BARNES, C.; HEROLD, N.; AND WICKHAM, J., 2011, Completion of the 2006 National Land Cover Database for the Conterminous United States: Photogrammetric Engineering Remote Sensing, Vol. 77, No. 9, pp. 858–864. GARBRECHT, J. and MARTZ, L. W., 1999, TOPAZ: An Automated Digital Landscape Analysis Tool for Topographic Evaluation, Drainage Identification, Watershed Segmentation, and Subcatchment Parameterization: U.S. Department of Agriculture, Agricultural Research Service. KEATON, J. R., 2013, Estimating erodible rock durability and geotechnical parameters for scour analysis: Environmental Engineering Geoscience, Vol. 19, No. 4, pp. 319–343. KEATON, J. R.; MISHRA, S. K.; and CLOPPER, P. E., 2012, Scour at Bridge Foundations on Rock: Transportation Research Board,
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
Risk-Based Strategy for Scour-Affected Bridges Washington, DC, National Cooperative Highway Research Program (NCHRP) Report 717, 174 p. NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION (NOAA), 2013, Precipitation Frequency Data Server: Electronic document, available at http://dipper.nws.noaa.gov/hdsc/pfds/ NATIONAL RESOURCES CONSERVATION SERVICE (NRCS), 1986, Technical Release 55, Urban Hydrology for Small Watersheds: U.S. Department of Agriculture, Natural Resources Conservation Service. NRCS, 2010, Part 630 Hydrology, National Engineering Handbook: U.S. Department of Agriculture, Natural Resources Conservation Service. NRCS, 2013, Natural Resources Conservation Service Soil Data Mart: Electronic document, available at USDA NRCS Soil Data Mart: http://soildatamart.nrcs.usda.gov/
Rahall Appalachian Transportation Institute, 2013, RP-273 PROJECT REPORT: Criteria for Predicting Scour of Erodible Rock in West Virginia, 2013: Electronic document, available at http://trid.trb.org/view.aspx?id=1300194 U.S. ARMY CORPS OF ENGINEERS HYDROLOGIC ENGINEERING CENTER (USACE), 1998, HEC-1 Flood Hydrograph Package User’s Manual Davis, CA. USACE, 2010, Hydrologic Modeling System HEC-HMS User’s Manual, Version 3.5. U.S. GEOLOGICAL SURVEY Office of Surface Water (USGS), 2007, The National Streamflow Statistics Program: A Computer Program for Estimating Streamflow Statistics for Ungaged Sites: Techniques and Methods: 4-A6. Rolla, MO. USGS, 2009, National Elevation Dataset—NED, 2nd ed.: USGS, Sioux Falls, SD.
Environmental & Engineering Geoscience, Vol. XXIII, No. 3, August 2017, pp. 221–241
241