Evaluating Streambank Retreat Prediction using the BANCS Model in the Valley and Ridge Physiographic Province
Rex S. Gamble
Thesis submitted to the faculty of the Virginia Polytechnic Institute and State University in partial fulfillment of the requirements for the degree of
Master of Science In Biological Systems Engineering
Theresa M. Thompson James B. Campbell William C. Hession
May 12th, 2021 Blacksburg, VA
Keywords: BANCS, NBS, Streambank Erosion
Copyright (optional – © or Creative Commons, see last page of template for information)
Evaluating Streambank Retreat Prediction using the BANCS Model in the Valley and Ridge Physiographic Province
Rex S. Gamble Academic Abstract
Excess sediment in streams is harmful to the environment, economy, and human health. Streambanks account for an estimated 7-92% of sediment and 6-93% of total- phosphorus loads to streams depending on the watershed. Stream stabilization through stream restoration has become a common practice to satisfy the 2010 Chesapeake Bay total maximum daily load (TMDL) due its value in credits received per dollar spent. Bank erosion is most commonly credited through the Bank Assessment for Non-point source
Consequences of Sediment (BANCS) framework, an empirically-derived model that predicts bankfull bank erosion rates using Bank Erodibility Hazard Index (BEHI), an indicator of bank stability, and Near-Bank Stress (NBS), an indicator of applied flow energy at bankfull discharge. This study assessed the BANCS framework in the Valley and Ridge physiographic province where it has not previously been applied. The spatial and temporal variability of erosion data was assessed to determine the impact of different erosion measurement schemes on bank erosion estimates and BANCS curves, and alternate NBS methods that capture flow energy beyond bankfull were applied. Three years of monthly erosion data on 64 streambanks were used to assess the spatial and temporal variability of erosion measurements and subsequently develop the erosion curves. Predicted erosion rates were then compared to measured erosion rates on three banks in the Valley and Ridge of
Southwest Virginia. Analysis of spatial variability suggests bank retreat measurements should be made every three channel widths to reliably quantify reach-scale load estimates.
Furthermore, a minimum monitoring period of 12 months is recommended to ensure seasonal
patterns in bank retreat are captured. These results also bring into question the efficacy of the
BANCS model as a crediting tool, as the developed statistical relationships between erosion rates, BEHI, and multiple NBS methods were not statistically significant. The limited number of significant curves had low r2 values (r2 < 0.1) indicating measures of NBS and
BEHI do not adequately explain the natural variability of bank retreat in the Valley and
Ridge of Southwest Virginia.
Evaluating Streambank Retreat Prediction using the BANCS Model in the Valley and Ridge Physiographic Province
Rex S. Gamble General Audience Abstract
While sediment naturally occurs in streams, too much sediment in these systems is harmful to the environment, economy, and human health. Streambanks contribute an estimated
7-92% of sediment pollution into streams. Stabilizing streambanks with stream restoration has become a common practice to reduce sediment for the 2010 Chesapeake Bay pollutant diet. The sediment reduction of bank stabilization is most commonly estimated with the Bank Assessment for Non-point source Consequences of Sediment (BANCS) framework, a model that predicts bank erosion rates using Bank Erodibility Hazard Index (BEHI), an indicator of bank stability, and Near-Bank Stress (NBS), an indicator of flow energy when the stream channel is full of water. This study assessed the BANCS framework in the Southwest (SW) Virginia where it has not previously been applied. In this process, the variability of the erosion data in space and time was assessed to determine the impact of different erosion measurement methodologies on bank erosion estimates and BANCS equations. Additionally, alternate NBS methods that represent flow energy below, at, and above the channel being full were tested. Three years of erosion data on 64 streambanks were used to assess the variability of erosion measurements in space and time and create new BANCS erosion equations. Predicted erosion rates using the new erosion equations were then compared to measured erosion rates on three banks in the area. Analysis of variability in space suggests bank retreat measurements should be made every three channel widths to reliably estimate erosion volume along a length of stream. Furthermore, a minimum measuring period of 12 months is recommended to ensure seasonal differences in bank retreat are captured. The results also bring into question the effectiveness of the BANCS model as a tool
to estimate sediment reduction for the Chesapeake Bay pollutant diet, as the developed equations between erosion rates, BEHI, and multiple NBS methods commonly failed to provide significant relationships. The limited number of significant curves had low r2 values (r2 < 0.1) indicating the measures of NBS and BEHI do not explain the natural variability of bank retreat in the study area.
Acknowledgements
My whole-hearted appreciation goes towards Dr. Tess Thompson, my research advisor and mentor throughout my time at Virginia Tech. Without her knowledge, guidance, and expertise, this research would not be here today. I also would like to thank my committee members Dr. Cully Hession and Dr. Jim Campbell for advising and supervising this project.
My most profound appreciate goes towards my research peers: Billy Paraszczuk,
Benjamin Smith, Daniel Smith, Coral Hendrix, and Samuel Withers. In my time here, they have all provided extraordinary support academically and socially. Without their company graduate life would not have been as fun, and without their help with field work, this project would not have been possible to complete.
I want to thank Laura Lehman who helped me ready field equipment and never failed to provide help when I asked. I also want to thank Denton Yoder for all his help with IT and CAD, as well as all the kombucha he shared. In addition, I would like to broadly thank all the faculty, students, and staff of the Biological Systems Department at Virginia Tech because my growth and well-being as a graduate student would not be possible without their knowledge, companionship, and support.
I also want to thank all the land owners who allowed me access to bank sites, in particular
Melinda Mays.
Finally, I would like to thank Josh Running and Gene Haffey of Stantec who are currently working hard to create more BANCS curves and were kind enough to show me the ropes regarding BANCS.
vi
Table of Contents
Academic Abstract ...... ii
General Audience Abstract ...... iv
Acknowledgements ...... vi
List of Figures ...... x
List of Tables ...... xvii
List of Abbreviations ...... xviii
1 Introduction ...... 1
2 Literature Review...... 6
2.1 Regulation of Sediment ...... 6
2.1.1 Chesapeake Bay Total Maximum Daily Loads ...... 6
2.1.2 Prevented Sediment Protocol...... 11
2.2 Bank Retreat ...... 15
2.2.1 Processes and Mechanisms ...... 15
2.2.2 Measuring Bank Retreat ...... 18
2.2.2.1 Methodologies & Categories ...... 18
2.2.2.2 Erosion Pin Measurements ...... 19
2.2.3 Predicting Bank Retreat ...... 22
2.2.3.1 Empirical Erosion Models ...... 22
2.2.3.1.1 Bank Assessment for Non-point source Consequences of Sediment
(BANCS)...... 23
2.2.3.2 Process-based Erosion Models ...... 35
3 Methodology ...... 37
vii
3.1 Research Sites ...... 37
3.1.1 Stroubles Creek...... 38
3.1.2 Toms Creek...... 40
3.1.3 North Fork of the Roanoke River ...... 41
3.2 Field Measurements ...... 42
3.3 Erosion Pin Measurement Variability ...... 43
3.3.1 Spatial Variability ...... 45
3.3.1.1 Bank-scale ...... 45
3.3.1.2 Reach-scale ...... 49
3.3.2 Temporal Variability ...... 50
3.3.2.2 Seasonal Variability ...... 51
3.3.2.2 Temporal Stabilization ...... 51
3.4 Assessment of the BANCS Model ...... 52
3.4.1 Input Erosion Rates ...... 53
3.4.2 NBS Measurement ...... 54
3.4.2.1 Traditional NBS ...... 54
3.4.2.2 Modified NBS ...... 55
3.4.2.2.1 Hydrograph-Based Models ...... 56
3.4.2.2.2 DuBoys Formula ...... 56
3.4.4 Comparing Erosion Curves ...... 57
4 Results & Discussion ...... 61
4.1 Erosion Pin Measurement Variability ...... 61
4.1.1 Spatial Variability ...... 61
viii
4.1.1.1 Bank-Scale ...... 61
4.1.1.2 Reach-Scale...... 66
4.1.2 Temporal Variability ...... 69
4.1.2.1 Seasonal Variability ...... 69
4.1.2.2 Temporal Stabilization ...... 73
4.1.3 Recommendations for Bank Retreat Measurement ...... 78
4.2 Assessment of the BANCS Model ...... 79
4.2.1 BEHI ...... 81
4.2.2 Traditional NBS ...... 83
4.2.3 Modified NBS...... 87
4.2.3.1 Hydrograph-Based ...... 87
4.3.3.2 Modified DuBoys...... 90
4.2.4 Assessing Erosion Curves ...... 92
4.2.4.1 Sites ...... 93
4.2.4.2 Input Erosion Rates ...... 96
4.2.4.3 NBS Methods ...... 97
4.2.4.4 Existing and Constructed Curves ...... 99
4.3 Comparison with the Universal Soil Loss Equation ...... 101
5 Conclusions ...... 105
References ...... 107
Appendices ...... 124
Appendix A: Stage and Flow Hydrographs ...... 124
Appendix B: Constructed BANCS Curves ...... 126
ix
List of Figures
Figure 2-1. Zones occupied by each stream restoration Protocol (Altland, Berg, et al., 2020). ... 10
Figure 2-2. Streambank retreat processes with subaerial processes, fluvial entrainment, and mass
wasting from left to right (Wynn, 2006)...... 16
Figure 2-3. BEHI A) scoring worksheet and B) conversion from ratios to BEHI score (Rosgen,
2006)...... 25
Figure 2-3 cont. BEHI A) scoring worksheet and B) conversion from ratios to BEHI score
(Rosgen, 2006)...... 26
Figure 2-4. NBS methods worksheet (Rosgen, 2006)...... 27
Figure 2-5. Modified method 5 (Altland, Berg, et al., 2020)...... 28
Figure 2-6. Relative erosion in relation to standardized radius of curvature (Hickin & Nanson,
1984)...... 30
Figure 3-1. Study site locations and respective watersheds in the Valley and Ridge of SW
Virginia...... 37
Figure 3-2. Images showing banks from A) SC in 2006, B) SC in 2020, C) TC, D) NFR. All
study sites have vertical banks composed of fine sediment and are highly erosive.
...... 39
Figure 3-3. Historic Stroubles Creek study reach plan view. Red crosses mark all erosion pin
columns, generally separated by 10 m. Banks W, X, Y, Z with 2-m erosion pin
spacings are circled in red...... 46
Figure 3-4. Established BANCS curves identified as A) the North Carolina curve for the
Piedmont (McQueen, 2011), B) the USFWS curve for the Coastal Plain (Rathbun,
2009)...... 60
x
Figure 4-1. Array method estimates of bank volumetric erosion rates suggesting that how one
calculates bank erosion volume significantly affects the erosion volume estimate 62
Figure 4-2. Array method results of one-way ANOVA with blocking of banks. Boxes represent
model least square mean with the 95% confidence interval indicted by the error
bars. PAEA is statistically different from BA and CA, exhibiting that the method
of calculation can significantly affect the erosion estimate...... 62
Figure 4-3. Series method bank volumetric erosion rates using A) OCCA and B) ORRA with BA
array method estimate shown as dashed line. Using one column or row of
measurement to estimate total bank erosion leads to high variability...... 65
Figure 4-4. Series method percent error one-way ANOVA with blocking of banks and mean
separation analyses results. Boxes represent model least square mean with the 95%
confidence interval indicted by the error bars. Result indicate that, while
statistically the same, using rows rather than columns to estimate bank erosion has
a lower percent error...... 66
Figure 4-5. Pin row erosion rate results from one-way ANOVA and mean separation analyses
displaying model least square mean and 95% confidence interval by height from
baseflow. Erosion rate estimates over long periods of time do not appear to be
affected by height from baseflow, indicating that measuring columns is redundant.
...... 67
Figure 4-6. Reach erosion volume estimates organized pin column spacing. Erosion volume
estimates appear to converge to a single value and decrease in variability as
spacing decreases...... 68
xi
Figure 4-7. Difference in erosion rates by season found using one-way ANOVA and mean
separation analysis for A) historic SC reach-scale dataset, B) TC bank-scale
dataset, C) modern SC bank scale dataset, and D) NFR bank-scale dataset. Boxes
represent model least square mean with the 95% confidence interval indicted by
the error bars. All sites except NFR have erosion rates affected by seasonality. ... 71
Figure 4-8. Time-average erosion rates for each measurement duration starting August 2005 for
individual A) erosion pins, B) column averages, and C) the entire reach. All spatial
scales show that the erosion rate estimate does not converge as more erosion data
is averaged...... 74
Figure 4-9. Time-averaged column erosion rates with different durations at: A) Toms Creek bank
starting January 2020, B) Stroubles Creek bank using new dataset starting January
2020, and C) North Fork of the Roanoke River starting March 2020. TC and SC
time-averaged erosion rates appear to level while NFR does not, likely due to
differences in erosion rate seasonality between the sites...... 75
Figure 4-10. Time-averaged erosion variance plotted for A) historic Stroubles Creek pins and
columns duration starting August 2005, and B) new datasets columns with TC, SC,
and NFR durations stating in January 2020, January 2020, and March 2020,
respectively. Variance decreases rapidly in the first two month and then less
rapidly the remaining months for all sites ...... 77
Figure 4-11. Bank erosion rates versus quantitative BEHI values for historic Stroubles Creek
dataset. BEHI does not appear to have a relationship with erosion rate...... 82
Figure 4-12. BANCS curves using different established NBS methods where: A) assesses all
NBS and selects the highest, B) only method 3 is used. Both regressions have low
xii
r2 values, indicating little of the variance in the erosion rate is explained by the
equation...... 85
Figure 4-13. Hydrograph-based curves using A) the number of peaks above baseflow, and B) the
percent of time above baseflow. Only the historic datapoints in red are used in the
regression. New site datapoints in grey are only plotted for comparison. Both
regressions have low r2 values, indicating little of the variance in the erosion rate is
explained by the equation...... 88
Figure 4-14. Statistically significant modified BANCS curves using A) hS per bankfull event, B)
hS and Rc per bankfull event, and C) hS and Rc for all monthly measurements. The
r2 values for all regressions were low, indicating little of the variance in the erosion
rate is explained by the equation...... 91
Figure 4-15. Constructed curve predicted erosion rate percent error for each site where the boxes
represent the mean and the errors bar showing the 95% confidence interval. The
model had the high error when predicting NFR erosion rates and the lowest error
when predicting TC erosion rates...... 95
Figure 4-16. Input erosion rates percent error one-way ANOVA with site blocking and mean
separation results where boxes represent the model least squre mean and error bars
show the 95% confidence interval. Using bankfull erosion rates as an input had a
higher erosion rate prediction error than using all monthly erosion rates as an
input...... 97
Figure 4-17. One-way ANOVA with site blocking and mean separation results where boxes
represent the model least squre mean and error bars show the 95% confidence
interval for: A) NBS categories percent error, and B) NBS methods percent error.
xiii
There is no statistical difference in erosion rate prediction error between NBS
categories or NBS methods...... 98
Figure 4-18. Existing and constructed curve predicted erosion rate percent error one-way
ANOVA with site blocking and mean separation results where boxes represent the
model least squre mean and error bars show the 95% confidence interval. Existing
BANCS curve have a much greater erosion rate percent errors than the curves
constructed in this study because they were created for use in different
physiographic provinces...... 100
Figure A-1. Historic Stroubles Creek flow hydrograph showing peaks modelled where gaps in
data occurred. Days where erosion pin measurements occurred are indicated by the
measurement event points...... 124
Figure A-2. Recent Stroubles Creek stage hydrograph. Days where erosion pin measurements
occurred are indicated by the measurement event points...... 124
Figure A-3. Toms Creek stage hydrograph. Days where erosion pin measurements occurred are
indicated by the measurement event points...... 125
Figure A-4. North Fork of the Roanoke River stage hydrograph. Days where erosion pin
measurements occurred are indicated by the measurement event points...... 125
Figure B-1. Erosion curve using bankfull erosion rates versus all NBS and selecting the highest.
...... 126
Figure B-2. Erosion curve using bankfull erosion rates versus NBS Method 2...... 126
Figure B-3. Erosion curve using bankfull erosion rates versus NBS Method 3...... 127
Figure B-4. Erosion curve using bankfull erosion rates versus NBS Method 4...... 127
Figure B-5. Erosion curve using bankfull erosion rates versus NBS Method 5...... 128
xiv
Figure B-6. Erosion curve using all monthly erosion rates versus all NBS and selecting the
highest...... 128
Figure B-7. Erosion curve using all monthly erosion rates versus NBS Method 2...... 129
Figure B-8. Erosion curve using all monthly erosion rates versus NBS Method 3...... 129
Figure B-9. Erosion curve using all monthly erosion rates versus NBS Method 4...... 130
Figure B-10. Erosion curve using all monthly erosion rates versus NBS Method 5...... 130
Figure B-11. Erosion curve using one year time-averaged erosion rates versus all NBS and
selecting the highest...... 131
Figure B-12. Erosion curve using one year time-averaged erosion rates versus NBS Method 2.
...... 131
Figure B-13. Erosion curve using one year time-averaged erosion rates versus NBS Method 3.
...... 132
Figure B-14. Erosion curve using one year time-averaged erosion rates versus NBS Method 4.
...... 132
Figure B-15. Erosion curve using one year time-averaged erosion rates versus NBS Method 5.
...... 133
Figure B-16. Erosion curve using all monthly erosion rates versus number of peaks above
baseflow. Only the historic dataset was used for regression. New site datapoints
are only plotted for comparison...... 133
Figure B-17. Erosion curve using all monthly erosion rates versus percent of time above
baseflow. Only the historic dataset was used for regression. New site datapoints
are only plotted for comparison...... 134
Figure B-18. Erosion curve using bankfull erosion rates versus Equation 8...... 134
xv
Figure B-19. Erosion curve using bankfull erosion rates versus Equation 9/10...... 135
Figure B-20. Erosion curve using all monthly erosion rates versus Equation 8...... 135
Figure B-21. Erosion curve using all monthly erosion rates versus Equation 9/10...... 136
Figure B-22. Erosion curve using one year time-averaged erosion rates versus Equation 8...... 136
Figure B-23. Erosion curve using one year time-averaged erosion rates versus Equation 9/10. 137
xvi
List of Tables
Table 2-1. Conversion from NBS method ratios to qualitative categories...... 29
Table 2-2. Existing and attempted BANCS curves, expanded from Bigham et al. (2018), where
the presence of the label ‘(BEHI)’ next to model fit indicates a regression of erosion
rate versus BEHI instead of the more traditional regression of erosion rate versus
NBS ...... 32
Table 3-1. Summary of spatial and temporal variability methodologies...... 44
Table 3-2. Methods used to calculate erosion volume. EDpin equals the erosion depth of a pin
while 퐸퐷 indicates the averaged erosion depth of a bank, column, or row. ℎ
represents the vertical distance associated with a bank, column, row, or pin with ℎ
indicating the averaged vertical distance for a bank, column, or row. 푥 indicates the
horizontal distance associated with a bank, column, row, or pin. 푉퐵 is the calculated
erosion volume of the bank...... 48
Table 4-1. Rank of array methods per bank and resulting variability for sensitivity study ...... 63
Table 4-2. Constructed curve regression statistics...... 80
Table 4-3. Measured annual erosion rates compared to predicted erosion rates...... 94
xvii
List of Abbreviations
BA Bank Average
BANCS Bank Assessment for Non-point source Consequences of Sediment
BEHI Bank Erosion Hazard Index
BMP Best Management Practice
CA Column Average
CAEA Column Average End Area
CBP Chesapeake Bay Program
CWA Clean Water Act
D.C. DOEE District of Columbia Department of Energy and Environment
DDOE District Department of Energy and the Environment
DEM Digital Elevation Model
Pennsylvania DEP Pennsylvania Department of Environmental Protection
U.S. EPA U.S. Environmental Protection Agency
GNSS Global Navigation Satellite System
LA Load Allocation
LID Low Impact Development
LiDAR Aerial or Terrestrial Scanning Laser Altimetry
MDE Maryland Department of the Environment
MS4 Municipal Separate Storm Sewer System
MUSLE Modified Universal Soil Loss Equation
N Nitrogen
NBS Near Bank Stress
xviii
NC North Carolina
NCD Natural Channel Design
NFR North Fork of the Roanoke River
NYSDEC New York State Department of Environmental Conservation
OCCA One Column Column Average
ORRA One Row Row Average
P Phosphorus
RA Row Average
PA Pin Area
PAEA Pin Average End Area
RSC Regenerative Stormwater Conveyance
RUSLE Revised Universal Soil Loss Equation
SC Stroubles Creek
StREAM Stream Research, Education, and Management
SW Southwest
TC Toms Creek
TMDL Total Maximum Daily Loads
TN Total-Nitrogen
TP Total-Phosphorus
TSS Total Suspended Solids
USFWS U.S. Fish and Wildlife Service
USLE Universal Soil Loss Equation
Virginia DEQ Virginia Department of Environmental Quality
xix
WIP Watershed Implementation Plan
WLA Wasteload Allocations
xx
1 Introduction
Sediment naturally occurs in aquatic systems but historic and current anthropogenic changes to the sediment cycle have and will cause long-lasting harm to these vital systems (Noe et al., 2020). European colonial changes to land cover and management, particularly the clearing of forests for agricultural use, caused a historic 10-fold increase in erosion which led to the mass storage of fine-sediment in stream and river valleys, referred to as legacy sediment (Kemp et al.,
2020; Noe et al., 2020). Aided and exacerbated by widespread damming during the 17th–20th centuries, these practices fundamentally changed the form of eastern North American streams from shallow, wide, anastomosing channels with wetland complexes into single-thread, meandering streams with vertical banks (Walter & Merritts, 2008). Much of the excess sediment entering streams today is from this legacy sediment still stored in valleys and floodplains, a continuation of a process started centuries ago (Meade, 1982).
While certainly an improvement on colonial times, sediment is still a problem in aquatic systems today; sediment is the third most common stream and river stressor in the United States behind phosphorus and nitrogen. It is estimated that 22% of U.S. stream and river lengths have
“poor” streambed sediments and that 16% of biologically impaired streams are impaired purely due to sediment as a stressor (U.S. EPA, 2020). Sediment affects aquatic systems as suspended sediment, through deposition, and as a vector for nutrients and contaminants. Suspended sediment has been observed to decrease light penetration and limit algal and macrophyte primary productivity (Izagirre et al., 2009; Jones et al., 2012; Yamada & Nakamura, 2002), cause damage to biota soft-tissue, alter fish behavior and predator-prey interactions (Kemp et al., 2011), and increase invertebrate drift (Culp et al., 1986). Deposition may decrease streambed particle size
(Wood & Armitage, 1997), decrease hyporheic exchange (Wharton et al., 2017), bury bed habitat
1
(Wood & Armitage, 1997), and inhibit benthic growth (Jones et al., 2012; Yamada & Nakamura,
2002). Nutrients are also transported by sediment with 73% of total-phosphorus (TP) and 18% of total-nitrogen (TN), the first and second most common stream and river stressors in the U.S., transported via sediment in the Chesapeake Bay (Noe et al., 2020). The general consensus of the scientific community is that P is usually the limiting nutrient in freshwater ecosystems while N is usually the limiting nutrient in estuarine and coastal systems (Conley et al., 2009); therefore, increased sediment loads to surface waters contribute to eutrophication and associated algal blooms. Once algae blooms can no longer sustain themselves they die and decompose, dramatically decreasing oxygen levels in the afflicted water body and causing mass aquatic die- offs (Conley et al., 2009). Some algae blooms emit toxins which kill aquatic fauna and livestock while seriously harming human health when ingested or in contact with skin (Lawton & Codd,
1991). Like nutrients, a number of contaminants can be transported by sediment such as heavy metals, PCBs, PAHs, and other organic contaminants (Noe et al., 2020).
Beyond the harm to aquatic systems, sediment pollution has water value, economic, and human health consequences. Excess suspended sediment decreases the perceived value of water for recreation (Smith et al., 1991) and increases drinking water processing costs (Dearmont et al.,
1998). Increased deposition in reservoirs and channels accelerates the decrease in reservoir capacity, shortening the life-span of dams, and increasing flooding frequency and magnitude.
Finally, related algal blooms harm fisheries, increase medical needs, and discourage tourism and recreation (Anderson, 2009).
In an effort to repair or alleviate the damage done by anthropogenic sediment movement into water bodies, U.S. regulatory agencies have adopted a system of limiting pollutant loads called total maximum daily load (TMDL), which is essentially the maximum amount of a
2
pollutant that can enter a water body with the system still meeting the water quality criteria set in the Clean Water Act (CWA). The largest of these TMDLs was implemented in 2010 for the
Chesapeake Bay and specifically targets sediment, nitrogen, and phosphorus pollution (U.S.
EPA, 2010). There are a number of best management practices (BMPs) available that receive
TMDL “credit” for pollution reduction, but stream restoration has one of the greatest credit per dollar. As such stream restoration implementation has rapidly increased in the Chesapeake Bay watershed (Schueler & Stack, 2014). While not in place when the Chesapeake Bay TMDL was originally developed, a regulatory process has now been established to estimate sediment and/or nutrient reductions from streambank erosion reduction, instream and riparian nutrient processing, and floodplain reconnection (Altland, Berg, et al., 2020; Schueler & Stack, 2014).
Of the credited benefits of stream restoration, streambank erosion accounts for the largest amount of sediment, and a sizable amount of TP, and possibly TN, loads to streams. Fox, Purvis,
& Penn, (2016) found that streambanks accounted for approximately 7-92% of sediment loads and 6-93% of TP loads in a watershed. While there is little research on the contribution of streambank retreat to nitrogen loads, it could be significant in some watersheds. For example, in the Central Claypan Region of Missouri it was estimated that streambanks contributed between
21% and 24% of TN exported from the study site (Willett et al., 2012). The Pennsylvania
Department of Environmental Protection reported that bank soil samples in the Chesapeake Bay watershed contained 400-2150 mg TN/kg soil, which could be a sizable portion of TN contributions into the Bay (Walter et al., 2007). Streambank sediments are of particular concern in the mid-Atlantic region due to the previously discussed legacy sediment which commonly enters streams through bank retreat in lower order streams (Donovan et al., 2015; Noe et al.,
2020). For example, across 40 streams in Baltimore Country, Maryland, legacy sediment
3
occupied roughly 60% of bank height and accounted for 57% (±16%) of measured erosion
(Donovan et al., 2015).
The crediting of streambank erosion through stream restoration for the Chesapeake Bay is laid out in the prevented sediment protocol, also referred to as protocol 1 (Altland, Berg, et al.,
2020). The process requires practitioners to “measure” bank erosion pre-restoration and either assume a 50% reduction post-restoration, or “measure” erosion post-restoration for at least 3 years to receive a higher percent reduction. The “measurement” method is chosen by the practitioner and must be consistent pre- and post-restoration. By far the most commonly used method is the Bank Assessment for Non-point source Consequences of Sediment (BANCS) model which is actually an erosion prediction rather than a measurement, although some stream restoration practitioners make actual measurements (Altland, Berg, et al., 2020).
The BANCS model has been largely adopted by regulatory agencies and practitioners despite criticism from academics (Castro-Bolinaga & Fox, 2018). Common issues include its focus on channel form rather than process (Castro-Bolinaga & Fox, 2018), the sensitivity of inputs (Bigham et al., 2018), an overreliance on the bankfull concept (Castro-Bolinaga & Fox,
2018), difficulty applying in areas with silty loam and fine clay soils (Sass & Keane, 2012) and with stream hydrology that is not driven by snowmelt (McMillan et al., 2017), the lack of erosion curves for many physiographic provinces, and the use of too little and short-term data to develop these erosion curves (Altland, Berg, et al., 2020). No curve has yet been created for the Valley and Ridge physiographic province and thus potential issues with BANCS have not been explored in the area. In addition to the lack of standardization in creating a new BANCS curve, there is also relatively little research into how the spatial and temporal variability of the input erosion
4
data impacts BANCS curves and if the method can be altered to rely less upon bankfull identification.
The goal of this study is to create a BANCS curve for the Valley and Ridge physiographic province and, in doing so, assess the sensitivity and error of the BANCS model predictions of bank retreat. Specific objectives of the study include: 1) assessing the spatial and temporal variability of erosion pin measurements on Stroubles Creek and the implications this variability has for sediment load predictions; 2) creating BANCS models using both tradition and proposed NBS methods; 3) evaluating the sensitivity of the BANCS framework to spatial and temporal variability of input erosion data; 4) evaluating the bank retreat predictions from the developed and existing BANCS models.
5
2 Literature Review
2.1 Regulation of Sediment
2.1.1 Chesapeake Bay Total Maximum Daily Loads
The U.S. Environmental Protection Agency (EPA) established the Chesapeake Bay
TMDL in 2010 under section 303 of the 1972 CWA. Implemented under the CWA which aims to make all waters of the United States “fishable” and “swimmable”, these TMDLs specifically aim to restore clean water to the Chesapeake Bay and its streams, creeks, and rivers. The development of the TMDL was initiated by poor water quality in the Bay. Despite 25 years of restoration work, water quality in the Bay remains poor, prompting the issuance of an executive order by President Obama on May 12, 2009, to renew federal efforts to restore and protect the
Chesapeake Bay. Delaware, Maryland, New York, Pennsylvania, Virginia, West Virginia and the District of Columbia have all been allocated pollution reduction amounts in N, P, and sediment, divided by jurisdiction and major river basin. Total watershed limits are 84.32 million kg/yr (185.9 million lb/yr) of N, 5.67 million kg/yr (12.5 million lb/yr) of P, and 2.93 billion kg/yr (6.45 billion lb/yr) of sediment, which amounts to a 25% reduction of N, 24% reduction of
P, and 20% reduction of sediment. These limits are to be met by 2025, with at least 60% completion by 2017. The six states and the District of Columbia each have submitted Watershed
Implementation Plans (WIPs) in phases to the EPA which outlines strategies to reduce the designated point and nonpoint source pollution within the state. While the states are responsible for implementing the WIPs, the EPA provides oversight to achieving the set limits (U.S. EPA,
2010).
TMDL limits are based on the maximum amount of pollutant that a water body can receive while still meeting water quality standards under the CWA so that it is not listed as
6
“impaired.” These limits were estimated in the 2009 Chesapeake Bay Watershed model, which was informed by peer-reviewed science and extensive monitoring data. The TMDL targets were calculated by summing estimated wasteload allocations (WLA) representing the point source pollution, load allocations (LA) representing the nonpoint source pollution, and a margin of safety. WLA sources include municipal wastewater, industrial discharge, combined sewer overflows, sanitary sewer overflows, permitted urban stormwater, and concentrated animal feeding operations. LA sources include agriculture, atmospheric deposition, forested lands, on- site wastewater treatment systems, nonregulated stormwater runoff, oceanic inputs, streambank and tidal shoreline erosion, tidal resuspension, and wildlife. Of all these pollutant sources, agriculture is the largest source of pollution to the Chesapeake Bay accounting for 44% of N and
P, and 65% of sediment loads (U.S. EPA, 2010).
Of the signatories to the Bay agreement, only West Virginia and the District of Columbia were successful in achieving their 2017 60% reduction goals. Maryland and Delaware have achieved their target for P and sediment, but not N. New York met their P target, but did not for
N and sediment. Both Pennsylvania and Virginia did not meet any of their 2017 reduction goals.
As of 2020, the EPA is using federal action in Delaware, Maryland, New York, and
Pennsylvania to meet TMDL goals with the most significant oversight falling on Pennsylvania
(Ritter, 2019).
Driven by the nutrient and sediment load reductions required by the Chesapeake Bay
TMDL, the practice of stream restoration has rapidly developed in several Bay states (Schueler
& Stack, 2014; Williams et al., 2017). This is exhibited when comparing the three phases of
WIPs. In Phase I only 156 km (97 mi.) of restored stream length was estimated between
Delaware, Maryland, Pennsylvania, and the District of Columbia with the other states not
7
estimating restoration lengths or not mentioning stream restoration entirely (Delaware’s
Chesapeake Interagency Workgroup, 2010; DDOE, 2010; MDE, 2010; NYSDEC, 2010;
Pennsylvania DEP, 2011; Virginia DEQ, 2010; West Virginia WIP Development Team, 2010).
By Phase II all the jurisdictions in total were planning 1054 km (655 mi.) of restoration in the
Bay by 2025 (Schueler & Stack, 2014). Finally, Phase III saw the planned length of restoration increase to 1262 km (784 mi.) by 2025 (Delaware’s Chesapeake Bay WIP Steering Committee,
2019; D.C. DOEE, 2019; MDE, 2012, MDE, 2019; NYSDEC, 2020; Pennsylvania DEP, 2019;
Virginia DEQ, 2019; West Virginia’s Chesapeake Bay Tributary Team, 2019). Regulatorily, stream restoration is categorized as a water quality BMP to be applied in both urban and non- urban areas (MDE, 2019; Michelsen, 2017; The Center for Watershed Protection, 2013). This means that stream restoration, depending on whether the project is in an urban or non-urban area, can contribute to either the urban/suburban runoff WLA or agriculture and/or natural LA. In general, urban stream restoration is more common because the Municipal Separate Storm Sewer
System (MS4) permits require urban/suburban municipalities to address stormwater discharges, as opposed to non-urban areas where addressing stormwater runoff is voluntary (MDE, 2019;
The Center for Watershed Protection, 2013; U.S. EPA, 2010). The predominance of urban stream restoration is such that the planned 2025 linear feet of urban restoration is more than double the planned non-urban restoration in the Phase II and III WIPs (Delaware’s Chesapeake
Bay WIP Steering Committee, 2019; D.C. DOEE 2019; MDE, 2012, MDE, 2019; NYSDEC,
2020; Pennsylvania DEP, 2019; Schueler & Stack, 2014; Virginia DEQ, 2019; West Virginia’s
Chesapeake Bay Tributary Team, 2019). Another reason for the rapid adoption of stream restoration is its cost-effectiveness in terms of TMDL credit per dollar spent, especially after crediting regulations for restoration were changed in 2012.
8
When the Bay TMDLs were first established in 2010, stream restoration crediting was undeveloped and uninformed; for example, all projects were assigned a single removal rate for
N, P, and sediment. This crediting method was deemed not “scientifically defensible” by an expert panel of scientists and practitioners in 2012, citing the variability associated with restoration projects. Although removal rates were updated to interim values and kept for planning purposes, the panel established a framework for stream restoration crediting that is still used in the Bay watershed as of 2021. Restoration crediting was split into four parts called
Protocols: Protocol 1, prevented sediment during stormflow, credits the annual erosion reduction that stream restoration produces; Protocol 2, instream and riparian nutrient process, credits the reduction of N through denitrification from the hyporheic exchange that stream restoration promotes; Protocol 3, floodplain reconnection, credits the nutrient and sediment reduction from reconnecting stream channels to the floodplain; and Protocol 4, dry channel regenerative stormwater conveyance (RSC) as an upland retrofit, credits the nutrient and sediment reduction associated with installing a dry RSC (Schueler & Stack, 2014). Since 2014 a Protocol 5 has been added which credits the nutrient and sediment reduction from stabilizing the channel between a stormwater outfall and the stream network (Altland, Berg, et al., 2020). The zones these
Protocols occupy are shown in Figure 2-1.
9
Figure 2-1. Zones occupied by each stream restoration Protocol (Altland, Berg, et al., 2020).
As a result of these 2012 crediting changes, stream restoration became more cost- effective. Using the removal rates before the crediting changes and only focusing on N removal,
Kenney, Wilcock, Hobbs, Flores, & Martínez (2012) concluded that water quality improvements were more costly per unit of N removed in stream restoration than alternative BMPs. Essentially, stream restoration was not an effective method for jurisdictions to achieve credit for pollutant reductions. However, studies after the crediting changes suggest the opposite. The Center for
Watershed Protection (2013) compared the cost-effectiveness of 35 BMPs in terms of cost per pound of TN, TP, and total suspended solids (TSS) removed. Included in these BMPs were both the historic and revised interim stream restoration removal rates. The study showed that of the
Chesapeake Bay Program (CBP) approved BMPs, stream restoration using the new removal rates was the most cost-effective BMP in removing P and TSS, and the third most cost-effective in removing N. Another study done by Montgomery County, Maryland in 2016 quantified the cost per unit of impervious acre reduced, a pollutant measure used by the state of Maryland, and found again that steam restoration was more cost-effective than stormwater ponds and other low impact development (LID) practices (Michelsen, 2017). It should be noted that these studies 10
were done using the recommended default removal rates of the time rather than actual measured removal rates, so these comparisons are estimates at best. Even so, it appears that from the perspective of state and local governments, stream restoration is one of the most cost-effective
BMPs for achieving the required pollutant reductions since the regulatory changes of 2012, which likely has contributed to the rapid growth of the practice in the Bay watershed.
2.1.2 Prevented Sediment Protocol
Stream restoration results in sediment reduction through the stabilization of the bed and banks of a stream (Altland, Berg, et al., 2020; Lammers & Bledsoe, 2017), which falls under the
Prevented Sediment Protocol or Protocol 1 (Altland, Berg, et al., 2020). A group of stream restoration experts from consulting firms, universities, and regulatory agencies, provided consensus recommendations for Protocol 1 covering a wide variety of topics, including definitions, qualifying conditions, how to calculate a prevented sediment credit, and erosion measurement and prediction practices. The following statements regarding Protocol 1 are from this consensus recommendation, but will vary between projects as permitting is ultimately decided by the appropriate state and federal regulatory agencies (Altland, Berg, et al., 2020).
Qualifying criteria are established which are intended to promote watershed-based approaches for stream restoration site selection. Existing qualifying criteria include; the stream reach must be greater than (30.5 m) 100 ft. in length and still be actively degrading due to previous or current upstream development, the project must use a long-term ‘comprehensive approach’ in restoration design, and special considerations for projects where floodplain reconnection for wetland creation and/or the addition of instream habitat features are possible. A
11
recent criterion is that Protocol 1 credits cannot be combined with Protocol 5 (Altland, Berg, et al., 2020).
Limits are given for the stream restoration practices of legacy sediment removal/stream valley restoration and bank armoring. A newer restoration practice is legacy sediment removal, also known as stream valley restoration, where practitioners remove the legacy sediment stored in stream valleys to attain pre-colonial stream geomorphology. This practice is credited with three options: 1) dividing the bank erosion zone into the remaining low bank sediment, where an assumed 50% sediment reduction efficiency is applied, and removed legacy sediment from the high bank, where a 100% sediment reduction is applied; 2) using a 90% sediment reduction efficiency for the entire bank erosion zone given the restoration meets specific standards; and 3) basing the credit on monitored data. It should be noted that raising the streambed of a channel to increase floodplain access is treated different regulatory; the 100% reduction in option 1 and all of option 2 are unavailable but there is a 100% sediment reduction for the portion of the bank erosion zone buried below the new channel invert (Altland, Brown, et al., 2020). Bank armoring, defined as the placement of hard structures into a stream for the purpose of limiting movement either horizontally or vertically, is characterized into three tiers that designate the amount of credit given to a project. The three tiers are as follows: 1) Non-Creditable, which are hard, permanent structures for protecting infrastructure and stabilizing banks including concrete retaining walls, sheet piling/planking, gabion walls, engineered block walls, A-jacks, and dumped rip-rap; 2) Creditable with Limits, structures of large rocks or boulders protecting limited portions of the bank or bank toe including localized stone toe protection, boulder revetments, non-biodegradable soil stabilization, and imbricated rip-rap; and, 3) Creditable, defined as structures mimicking natural streambank material that provide stream habitat,
12
function, and limited grade control such as root was revetments, live stakes, soil lifts, riffle-weir series, berm-pool cascades, J-hooks, and cross-vanes (Altland, Berg, et al., 2020). These designations are meant to encourage the use of structures and practices that resemble nature and contribute to overall ecological uplift as opposed to traditional civil engineering approaches which emphasize infrastructure protection and stability.
There are three key steps to the Prevented Sediment Protocol: 1) the estimation of pre- restoration erosion rates, converted into sediment loads, which can either be directly measured, the preferred approach, or estimated using BANCS, which is more common; 2) the conversion of erosion rates into N and P loads, calculated using a measured or default conversion factor; and,
3) the estimation of a restoration reduction efficiency, conservatively set at a 50% reduction, to calculate the credit given to a project (Altland, Berg, et al., 2020). If the project is successful, it will receive credit for preventing, not only sediment, but P and N too.
Step 1 of the Prevented Sediment Protocol is the estimation of pre-restoration erosion rates which can either be directly measured or estimated with BANCS. While the use of BANCS is more common, direct measurement is preferred by the Prevented Sediment committee. Not all banks in a reach need to be directly measured, but extrapolation from measured banks to unmeasured banks should be done carefully and measured banks should be representative of other banks in the project reach, although how to indicate a bank is ‘representative’ is not specified. While it varies, some practitioners use Rosgen’s BANCS categories to determine whether measured bank conditions match non-measured bank conditions (D. Altland, J.
Running, T. Thompson, personal communication, May 1, 2020). Once an erosion rate is found it is converted into a sediment load using Equation 1. Protocol 1 originally had a default bulk density value but, due to the variability of values between sites, it was the view of the Prevented
13
Sediment committee that taking bulk density samples should be a minimum requirement
(Altland, Berg, et al., 2020).
훴(푐퐴푅) 푆 = (1) 2,000 where S is the sediment load (ton/year) of the reach, c is the bulk density of soil (lbs/ft.3), A is the eroding bank area (ft.2), R is the bank erosion rate (ft./year), and 2,000 is a conversion from pounds to tons.
Step 2 of the Prevented Sediment Protocol is to convert streambank erosion rates into nutrient loadings, which is done by multiplying the sediment loads calculated in step 1 by the average soil TP and TN concentrations. Due to the highly variable nature of nutrient concentrations in soils, it was the view of the Prevented Sediment advisory committee that taking nutrient samples should be required and the samples should be analyzed using the Total-sorbed P
– EPA Method 3051 + 6010 (USEPA 1986) and Total N combustion testing (Bremner 1996) methods, respectively (Altland, Berg, et al., 2020).
Step 3, the estimation of stream restoration efficiency, is determined by multiplying the value from step 1 by an efficiency ratio. The default value is a conservative 50%, which can be increased if the site is monitored post-restoration for a minimum of three years after the completion of a project using the same method as the pre-restoration monitoring. The latter approach requires the project to be re-reported to replace the original report (Altland, Berg, et al.,
2020). If the monitored efficiency ratio is less than 50%, the project does not have to be rereported and credit is still allocated with the 50% efficiency. This potential for increased credit after three years is meant to incentivize project monitoring by practitioners, but the risk is that if the monitored efficiency ratio is less than 50%, then the time and resources of the practitioner will be wasted.
14
A two-stage verification inspection process is recommended for each completed project.
The first stage is a rapid visual inspection where assessors walk the project reach and look for signs of: bank erosion as shown by exposed bare earth on undercutting, a 20% average change in post-construction bank height at riffles, incising channels as indicated by loss of pools and riffle and/or an active knickpoint, flanking or scour of in-channel structures, failure/collapse of bank protection, and less than 80% cover of ground or canopy. The second stage, if failure is seen in the first stage, is to return to the site under a forensic inspection to diagnose the cause(s) of said failure(s). After the first stage inspection the project is categorized into three statuses which require different actions: functioning or showing minor compromise due to a 0-10% failing of the project reach, where no immediate action is needed and another stage one inspection will be carried out in 5 years for credit; showing major compromise due to a 20-40% failing of the project reach, where a second stage inspection is conducted, repairs/maintenance is performed, and credit may be reduced; or project failure due to a 50-100% failing of the project reach, where credit is lost and the project is abandoned.
2.2 Bank Retreat
2.2.1 Processes and Mechanisms
Bank retreat is the net recession of a streambank as a result of several combined processes. The three most commonly discussed processes are: fluvial entrainment, subaerial processes, and mass wasting. Fluvial entrainment is the detachment or entrainment of sediment particles or aggregates from the bank surface by the flow of water. Subaerial processes are the direct erosion or weakening of the bank due to weather conditions. Mass wasting is the geotechnical slope failure of a bank due to gravitational forces. Combined, these processes are
15
largely responsible for bank retreat with the relative importance of each process being spatially and temporally variable (Couper & Maddock, 2001; Lawler, 1995; Wynn et al., 2008). Figure 2-
2 visualizes how the processes may interact resulting in bank retreat; subaerial processes weaken the bank, streambank soil is eroded during a flood event, and mass wasting occurs on the upper bank due to slope instability.
Figure 2-2. Streambank retreat processes with subaerial processes, fluvial entrainment, and mass wasting from left to right (Wynn, 2006).
Fluvial entrainment, also referred to as “fluvial erosion”, commonly erodes more of the bank toe, which destabilizes the bank and leads to mass failure (Thorne, 1982). Fluvial entrainment is highly dependent on the material nature of the bank. The entrainment of non- cohesive sediment is more easily understood with erosion depending on the sediment grain size distribution, shape, and density. Non-cohesive soil erosion is more complex and less understood
(Thorne, 1982; Wynn, 2006). Important cohesive sediment factors affecting fluvial entrainment include grain size distribution, bulk density, clay type and content, organic matter content, pore water content, and chemistry. Usually, cohesive banks are more resistant to fluvial entrainment but more susceptible to subaerial processes (Wynn et al., 2006).
Although grouped as a single process there are numerous subaerial processes, all of which are controlled by climatic conditions. As a result, these processes are temporally variable 16
(Couper & Maddock, 2001; Wynn et al., 2008). While subaerial processes can cause direct erosion of the bank, the usual effect is to weaken the bank for fluvial entrainment or mass wasting, hence the alternate name of “preparatory processes” (Lawler, 1993a; Thorne, 1982).
Examples of different subaerial processes include wetting and drying, resulting in soil desiccation, and freeze-thaw cycling, the most widely discussed process (Couper & Maddock,
2001; Lawler, 1995; Wynn et al., 2008). Freezing of bank surfaces increases soil volume and also causes a migration of soil water to the bank surface, increasing soil moisture. The resulting increase in moisture content and decrease in density make the soil more susceptible to erosion
(Lawler, 1993a; Wynn, 2006). This process is usually observed in fine-grained soils (Couper &
Maddock, 2001; Wynn et al., 2008), such as legacy sediments.
Mass wasting is also referred to as “bank failure”, and occurs when the weight of the bank is greater than the shear strength of the soil. Mass wasting depends on bank geometry, soil material and stratification, and bank vegetation. While there are many geotechnical mechanisms of failure (Thorne, 1982), mass wasting usually follows an increase in bank height or bank angle due to fluvial entrainment or cracking. It is common to see mass wasting after floods due to an increase in bank moisture content and soil weight, erosion of the bank toe, and a decrease in the water confining pressure as the flood recedes (Wynn, 2006).
Apart from the three main processes, seepage and piping erosion also can be significant bank retreat processes. This is when subsurface water from the hillslope of a stream or river entrains sediment particles through soil pores or a preferential flow path, also known as a soil pipe. Piping erosion is usually the result of soil stratification with the presence of a conductive layer over a restricting layer. Research has shown that this process contributes to bank erosion and eventually gully formation. Gullies are formed when soil pipes erode to a point where
17
driving forces are greater than resisting forces in the soil, resulting in mass failure. While it is known that piping erosion contributes to bank retreat, the mechanisms and nature of the process are still being researched (Fox & Wilson, 2010).
2.2.2 Measuring Bank Retreat
2.2.2.1 Methodologies & Categories
Although bank stabilization has received a lot of interest recently there is a lack of standardized methodologies in the erosion rate measurement and prediction used to credit these projects (Bledsoe et al., 2016). This leads to inconsistencies when comparing results and higher uncertainties when allocating credit. The Prevented Sediment Protocol for the Chesapeake Bay is explicitly guilty of this because practitioners may use, but are not limited to, the following techniques to measure or predict bank retreat: digital elevation model (DEM) differencing, bank erosion pins, repeated cross-sections, bank profile measurements, BANCS assessments, and TSS monitoring (Altland, Berg, et al., 2020). Essentially, the method used to measure or predict bank retreat is dependent upon the professional judgement and/or preference of the practitioner, likely leading to higher variability in crediting.
Lawler (1993) reviewed measuring techniques for streambank erosion and identified seven common “field-intensive” methods which are classified by resolution and time scale. Long timescales are approximately 10 to 2×104 years and have low-resolutions of 1-100 years. These long timescale techniques include sedimentological evidence, botanical evidence, and serial historical maps. Intermediate timescales are 1-30 years and have a relatively low temporal resolution of 0.05-2 years. These techniques include planimetric resurvey and repeated cross- profiling. Short timescales range from a few months to 10 years and have a fine resolution of
18
days if necessary. Short timescales allow for a process-based emphasis due to the ability to relate erosion to specific hydrometeorological events and to focus on spatial variation. Methodologies include erosion pins and repeated terrestrial photogrammetric surveys. It should also be noted that since 1993 there have been advances in bank erosion measurement technologies, such as aerial or terrestrial scanning laser altimetry (LiDAR) (Nasermoaddeli & Pasche, 2008; Thoma et al., 2005), structure-from-motion (James et al., 2019), and Global Navigation Satellite System
(GNSS) surveying (Brasington et al., 2000; Schilling & Wolter, 2001). These new technologies function in the same way as some historic methodologies, namely, both create terrains at different points in time and compare changes over time to estimate the volume of sediment lost.
The three new techniques, LiDAR, GNSS surveying, and structure-from-motion, fall under a new category that the Chesapeake Bay Prevented Sediment Protocol calls “Digital Elevation
Model (DEM) Differencing” with high resolution and either an intermediate or short time scale
(Altland, Berg, et al., 2020). The proper methodology for a study is dependent on the user’s necessary timescale and resolution. For crediting purposes, the resolution should be high and the timescale should ideally be at least five years but the Chesapeake Bay recommends a minimum of 3 years (Altland, Berg, et al., 2020; Bledsoe et al., 2016). This desired timescale and resolution essentially rules out all long timescale techniques and favors short timescale methodologies.
2.2.2.2 Erosion Pin Measurements
One of the most common bank retreat measurement techniques is bank/erosion pins because it is simple, cheap, sensitive and works in almost all fluvial environments (Altland,
Berg, et al., 2020; Lawler, 1993b). Despite the commonality of the method, there is little
19
standardization, and the standard operating procedures that do exist vary widely (Altland, Berg, et al., 2020). The method involves the installation of pins perpendicularly into a bank surface and the repeated measurement of the exposed pin length to calculate erosion rates (Altland, Berg, et al., 2020; Boardman & Favis-Mortlock, 2016; Lawler, 1993b; Wynn et al., 2006). The pin itself should be made of a non-rusting and non-rotting material so there are no chemical or biological disturbances to the soil while the pin is in place (Lawler, 1993b). This means that the common practices of using rebar or wooden stakes are not acceptable; instead one should use stainless steel pins (Wynn et al., 2006). Contrary to this wisdom, the Chesapeake Bay protocol still recommends iron or steel (Altland, Berg, et al., 2020).
Site selection is an important part of an erosion pin study and is dependent on the type of study being performed (Boardman & Favis-Mortlock, 2016; Lawler, 1993b). A general recommendation is to avoid sites where pins may be tampered with by livestock or humans
(Wynn et al., 2006). The guidance related to the placement of pins in a reach vary. Thompson et al. (2006) recommends installing pins at cross-sections with a spacing of 10 m (32.8 ft.) horizontally and 0.3 m (1.0 ft.) vertically with the horizontal spacing decreasing to 2 m (6.56 ft.) on more actively eroding banks. A vertical column of pins at each cross-section starts at the baseflow water level and fits as many pins as possible on the bank using the previously mentioned spacing. On the other hand, the Chesapeake Bay protocol recommends that pins should be placed every 61-152 m (200-500 ft.) with at least 2 pins vertically placed depending on the complexity of the site. If measured pin data is to be extrapolated to non-measured banks, as is common in stream restoration, the measured pins should be representative of the unmeasured banks (Altland, Berg, et al., 2020). There is no specified manner of quantifying what banks are
‘representative’ of other banks, but practitioners commonly use the BANCS categorization
20
established by Rosgen to compare banks (D. Altland, J. Running, T. Thompson, personal communication, May 1, 2020).
Pin installation can either leave the pins flush with the bank or slightly exposed so they can easily be found when returning to the site to measure; Thompson et al., (2006) recommends exposing the pins by 4 cm. (Altland, Berg, et al., 2020). No specified time interval is given for revisiting the site as this is dependent on the scope of the investigation, but the field visits should be frequent due to vandalism or rapid erosion events and not occur during high flow conditions for safety reasons (Altland, Berg, et al., 2020; Boardman & Favis-Mortlock, 2016; Wynn et al.,
2006). The only guidance found on measurement is by Thompson et al. (2006) who recommends that, after measuring pin length with calipers, if the length exceeds 1.0 cm of the initial setting
(i.e. > 3.0 cm or < 5.0 cm), the pin should be reset to 4.0 cm with this action being recorded in notes. It is unknown how stream restoration practitioners perform measurement, but it is likely to vary between practitioners.
Any scientific measurement has error and uncertainty. Bank pin uncertainty is not at all discussed in the Chesapeake Bay protocol, and is rarely discussed in academic papers as well
(Boardman & Favis-Mortlock, 2016). When a trained individual is repeatedly measuring erosion pins, the standard deviation of the measurement is 1.5 mm (0.059 in.), but increases to 2.5 mm
(0.098 in.) when multiple individuals are performing repeated pin measurements (Boardman et al., 2015; Boardman & Favis-Mortlock, 2016). This error is compounded when erosion data is extrapolated from the pin point erosion measurements to a volume/load used for crediting. The
Chesapeake Bay does have an equation to perform this calculation, but before the equation is applied, erosion data is averaged and extrapolated from measured banks to non-measured banks
(Altland, Berg, et al., 2020). It should be noted that every method used for erosion data
21
averaging and extrapolation has associated assumptions that should be considered and stated
(Boardman & Favis-Mortlock, 2016).
2.2.3 Predicting Bank Retreat
2.2.3.1 Empirical Erosion Models
Empirical models aim to predict bank erosion through simple variables (Klavon et al.,
2017). A majority of these erosion models in literature fit the form expressed in Equation 2
(Moody et al., 2005):
퐸 = 퐾 푋 (2) where E represents erosion rate, K is a soil erosivity/resistance coefficient, and X is an indicator of flow energy (Moody et al., 2005). While the equation is usually expressed linearly due to simplicity and ease of calculation, a power function with an exponent between 1.05 and 6.8 was found to be more versatile for a wide range of X with the linear form being more acceptable for higher X ranges. Even so, there is no theoretical consensus on whether the equation should be linear or a power function (Knapen et al., 2007).
Flow energy, X, can be represented in a multitude of ways including; kinetic energy per unit area, near-bank velocity, the difference between near-bank velocity and average stream velocity, near-bank water depth, unit stream power, boundary shear stress, and the difference between applied shear stress (τa) and critical boundary shear stress at which erosion starts (τc), known as excess shear stress (Thompson, 2020). The temporal and spatial average of the critical shear stress parameter is the most common expression of flow energy with its application occurring in all forms of sediment modelling from bank erosion (Kassa, 2019; Langendoen,
22
2011; Osman & Thorne, 1988) to overland soil erosion (Knapen et al., 2007) and bedload transport (Wong & Parker, 2006).
2.2.3.1.1 Bank Assessment for Non-point source Consequences of Sediment (BANCS)
The most widely applied empirical model is BANCS, which predicts bank erosion through statistically correlating measured erosion due to bankfull floods, Bank Erodibility
Hazard Index (BEHI), and Near-Bank Stress (NBS) (Castro-Bolinaga & Fox, 2018), usually as a power function (Berg et al., 2013; McQueen, 2011; Rosgen, 2006). The BANCS model resembles the general equation for estimating bank erosion shown above with 퐾 estimated using
BEHI, a quantification of bank stability, and 푋 quantified with NBS, an approximation of flow energy. One of the largest advantages of BANCS is that both BEHI and NBS are largely quantified in the field, so predicting erosion rates requires only a single field excursion to categorize the study banks given a BANCS erosion curve exists for the area (Rosgen, 2006).
BEHI is an empirical indicator of bank stability based on the following morphological properties measured in the field: bank height to bankfull height ratio, bank angle, root depth to bank height ratio, weighted root density, surface protection, material stratification, and material
(Bigham et al., 2018; Klavon et al., 2017; McMillan et al., 2017). The scoring system used for
BEHI is shown in Figure 2-3 below. Freeze-thaw and piping erosion mechanisms are considered in this parameter through the material and stratification adjustments, respectively. In general, adjustments are difficult to assess with the bank material adjustment in particular leading to the most uncertainty (Bigham et al., 2018). It should also be noted that the adjustments in Figure 2-
3A have been changed by certain parties. For example, the optional spreadsheet provided to practitioners in the Chesapeake Bay has a clay/silt loam category that adds 5 points to account
23
for freeze-thaw, which is missing in the original sheet (Altland, Berg, et al., 2020; Rosgen,
2006). Unfortunately, since the sheet is only recommended, it can lead to inconsistencies in bank
BEHI ratings depending on which tool is used by practitioners.
NBS approximates the applied fluvial shear stress on the bank toe. The seven possible
NBS methodologies are shown in Figure 2-4 with Figure 2-5 showing the modified method 5 created by Stantec. Most NBS methods have a conversion table to convert the method ratios into their respective qualitative NBS values, shown in Table 2-1. Rosgen suggests selecting several of the best fitting methodologies and using the highest NBS of those calculated (Rosgen, 2006).
There is no discussion in the original text about which NBS method is best suited for what situation; practitioners either have to infer this information or participate in one of Rosgen’s short courses. From recent publications it is known that method 1 is most appropriate with mid- channel deposition and method 2 should only be applied when a bank is on a meander bend, but little else is specified (Rosgen et al., 2019).
How the NBS methods were conceived are unexplained but several methods have physical basis. Method 1 assumes where flows are directed to a bank, erosion will be high, hence the automatic placement into the “High”, “Very High”, or “Extreme” NBS categories. Method 2 relates to studies that show high boundary shear stresses occur along the outside of a meander bend (Perron et al., 2009). Both the original and modified method 5 are related to method 2, suggesting higher shear stresses are applied to the bank due to secondary flows when the thalweg is nearer to the bank, or when the thalweg is much lower than the bank’s toe, indicating more erosion has occurred. Method 6 uses established shear stress equations, and method 7 relates to a method of calculating boundary shear stress called law of the wall (Wilcock, 1996). Methods 3 and 4 do not have a known basis in hydraulics or fluvial geomorphology (Thompson, 2020). The
24
two most detailed and physically-based methods, method 6 and method 7, are rarely used in the field due to their relative inconvenience.
(A)
Figure 2-3. BEHI A) scoring worksheet and B) conversion from ratios to BEHI score (Rosgen, 2006). 25
(B)
Figure 2-3 cont. BEHI A) scoring worksheet and B) conversion from ratios to BEHI score (Rosgen, 2006).
26
Figure 2-4. NBS methods worksheet (Rosgen, 2006).
27
Figure 2-5. Modified method 5 (Altland, Berg, et al., 2020).
While some methods do have underpinnings in research and hydraulics, this may not be well translated to the NBS methods themselves. Method 2, for example, does not correctly reflect empirical research. Hickin & Nanson (1984) show that erosion rates peak when the radius of curvature to bankfull width is 2-3, as shown in Figure 2-6. This is not reflected in method 2’s conversion of ratios to NBS categories, where these peak ratios would be categorized between
“Very Low” or “High.” Method 7 also has inconsistencies, as the method neglects to acknowledge that the law of the wall can only be applied if the velocity profile matches the log- law equation (Rosgen, 2006; Wilcock, 1996).
28
Table 2-1. Conversion from NBS method ratios to qualitative categories. NBS Method Ratio Conversion to NBS Category 1 N/A High /Very High / Extreme Very Low: > 3.00 Low: 2.21 – 3.00 푅푎푑𝑖푢푠 표푓 퐶푢푟푣𝑖푡푢푟푒 Moderate: 2.01 – 2.20 2 퐵푎푛푘푓푢푙푙 푊𝑖푑푡ℎ High: 1.81 – 2.00 Very High: 1.50 – 1.80 Extreme: < 1.50
Very Low: < 0.20 푃표표푙 푆푙표푝푒 Low: 0.20 – 0.40
퐴푣푒푟푎푔푒 푆푙표푝푒 Moderate: 0.41 – 0.60 3 (Slopes at baseflow) High: 0.61 – 0.80 Very High: 0.81 – 1.00 Extreme: > 1.00
Very Low: < 0.40 푃표표푙 푆푙표푝푒 Low: 0.41 – 0.60
푅𝑖푓푓푙푒 푆푙표푝푒 Moderate: 0.61 – 0.80 4 (Slopes at baseflow) High: 0.81 – 1.00 Very High: 1.01 – 1.20 Extreme: > 1.20
Very Low: < 1.00 Low: 1.00 – 1.50 푀푎푥𝑖푚푢푚 푁푒푎푟 퐵푎푛푘 퐷푒푝푡ℎ Moderate: 1.41 – 1.80 5 푀푒푎푛 퐵푎푛푘푓푢푙푙 퐷푒푝푡ℎ High: 1.81 – 2.50
Very High: 2.51 – 3.00 Extreme: > 3.00
Very Low: 0.67 – 1.00 Low: 0.66 – 0.33 퐷𝑖푠푡푎푛푐푒 푡표 푇ℎ푎푙푤푒푔 푓푟표푚 푆푡푢푑푦 퐵푎푛푘 5 Moderate: 0.32 – 0.22 퐵푎푛푘푓푢푙푙 푊𝑖푑푡ℎ (Modified) High: 0.21 – 0.11
Very High: 0.10 – 0.06 Extreme: < 0.05
Very Low: < 0.80 Low: 0.80 – 1.05 푀푎푥𝑖푚푢푚 푁푒푎푟 퐵푎푛푘 푆ℎ푒푎푟 푆푡푟푒푠푠 Moderate: 1.06 – 1.14 6 퐵푎푛푘푓푢푙푙 푆ℎ푒푎푟 푆푡푟푒푠푠 High: 1.15 – 1.19
Very High: 1.20 – 1.60 Extreme: > 1.60
29
Table 2-1 cont. Conversion from NBS method ratios to qualitative categories. NBS Method Ratio Conversion to NBS Category Very Low: < 0.50 Low: 0.50 – 1.00 Moderate: 1.01 – 1.60 7 N/A High: 1.61 – 2.00 Very High: 2.01 – 2.40 Extreme: > 2.40
Figure 2-6. Relative erosion in relation to standardized radius of curvature (Hickin & Nanson, 1984).
30
A large number of BANCS curves have been developed, as shown in Table 2-2. As a note, if the r2 value in Table 2-2 is labelled as BEHI, it means that BEHI was used as the erosion predictor rather than NBS. Curves have had varying successes in the past, exhibited by the r2 values in Table 2-2 and the fact that several of the curves in Table 2-2 have negative slopes. This lack of success is likely contributed to by the absence of standardization in the creation of
BANCS curves. Erosion curves that use the BANCS framework have been created using a varying number of sites, erosion measurement techniques, years of data, and streamflows as shown in Table 2-2. There are also a number of curve where this information is unknown, and thus are not listed in Table 2-2, including the curve developed for the Chesapeake Bay watershed as of 2021. The variable success of the BANCS framework has also been attributed to a number of other factors.
The BANCS model has been heavily criticized for a number of issues including: its focus on channel form rather than process (Castro-Bolinaga & Fox, 2018); the sensitivity of BEHI and
NBS inputs (Bigham et al., 2018); the lack of correlation between NBS and erosion (Allmanová,
Vlčková, Jankovský, Jakubis, et al., 2019; Ghosh et al., 2016); an overreliance on the bankfull concept (Castro-Bolinaga & Fox, 2018); difficulty applying in areas with silty loam and fine clay soils (Sass & Keane, 2012) and with stream hydrology not driven by snowmelt (McMillan et al.,
2017); the lack of erosion curves for many physiographic provinces; and, the use of small data sets and short-term data to develop these erosion curves (Altland, Berg, et al., 2020).
31
Table 2-2. Existing and attempted BANCS curves, expanded from Bigham et al. (2018), where the presence of the label ‘(BEHI)’ next to model fit indicates a regression of erosion rate versus BEHI instead of the more traditional regression of erosion rate versus NBS Erosion Highest Area # of Years Model Fit NBS Method Measurement Observed (Author) Sites of Data Method Discharge Alabama and Florida Bankfull r2 = 0.01-0.92 74 2 2,5 Bank profiles (McMillan et al., 2017) exceeded 1.39× greater Arkansas Not Reported 24 1 Highest value Bank profiles than (Van Eps et al., 2004) bankfull 65% below California r2 = 0.37–0.77 137 1 5 Bank profiles bankfull to 1.5× (Kwan & Swanson, 2014) bankfull r2 = 0.92 Colorado 60–70% below (Corresponding BEHI 49 1 7 Bank profiles (Rosgen, 2001, 2006) bankfull equation not specified) Kansas r2 = 0.75–0.77 16 4 Highest value Bank profiles 1-2.5× bankfull (Sass & Keane, 2012) Michigan r2 = 0.37–0.67 (BEHI) 46 2 5 Bank pins Not reported (Dick et al., 2014) New York r2 = 0.23 (BEHI) 8× greater than 16 11 Highest value Bank profiles (Coryat, 2011) r2 = 0.35-0.37 bankfull New York r2 = 0.53 (BEHI) Repeated 9× greater than 9 1 5 (Markowitz & Newton, 2011) r2 = 0.20 cross-sections bankfull North Carolina r2 = 0.167 27 1 Highest value Bank profiles Not reported (Jennings & Harman, 2001) Near-bank Oklahoma 4× greater than r2 = 0.09–0.32 29 1 area/total Bank pins (Harmel et al., 1999) bankfull bankfull area Western Carpathians – Slovakia r2 = 0.72 (BEHI) 4× greater than (Allmanová, Vlčková, 18 1 5 Bank profiles r2 = 0.004-0.15 bankfull Jankovský, Allman, et al., 2019)
32
Table 2-2 cont. Existing and attempted BANCS curves expanded from Bigham et al. (2018). Erosion Highest Area # of Years Model Fit NBS Method Measurement Observed (Author) Sites of Data Method Discharge Western Carpathians – Slovakia r2 = 0.73 (BEHI) Bankfull not (Allmanová, Vlčková, 18 1 5 Bank profiles r2 = 0.51-0.66 exceeded Jankovský, Jakubis, et al., 2019) r2 = 0.84 Wyoming 60–70% below (Corresponding curve 40 1 7 Bank profiles (Rosgen, 2001, 2006) bankfull not stated) West Bengal r2 = 0.28 (BEHI) Repeated 24 5 5 Not reported (Ghosh et al., 2016) r2 = 0.28 cross-sections
33
The reliance on bankfull discharge in particular has theoretical issues. Stream restoration is usually practiced in degraded and/or unstable systems, meaning the theoretical channel- forming discharge that bankfull discharge estimates does not exist (Copeland et al., 2000). In practice this means unstable and degraded channels have less obvious bankfull indicators, making bankfull identification, and thus the BANCS methodology, difficult. A modified BEHI method not reliant on bankfull discharge has been proposed and tested in Northeastern Ohio. The procedure simply removes the study bank height to bankfull height ratio and adjusts the scoring for other measurements (Newton & Drenten, 2015).
Despite these criticisms, BANCS has been adopted by key U.S. government agencies such as the U.S. EPA, the Forest Service, and the Fish and Wildlife Service to credit bank erosion (Castro-Bolinaga & Fox, 2018; McMillan et al., 2017). In the Chesapeake Bay, BANCS can be used to predict retreat rates for TMDL crediting purposes, as opposed to directly measuring retreat, if it is the preference of the practitioner. Not all parties in the Bay are embracing BANCS; as of 2020, Protocol 1 openly encourages the direct measurement of bank erosion, and the Pennsylvania Department of Environmental Protection explicitly expressed its skepticism towards BANCS (Altland, Berg, et al., 2020).
In line with these reservations, the authors of Protocol 1 acknowledge the deficiencies in
BANCS and attempts to address them as much as possible. There is a concerted effort to make more BANCS erosion curves, particularly in the Coastal Plain and Piedmont physiographical provinces. The development of these curves is predicted to cost of $400,000 and take 2 years, if not longer, because the data used to develop BANCS curves are limited to bankfull events, meaning only one or two data points per bank can contribute to curve development a year in the
Bay region (Altland, Berg, et al., 2020; Thompson, 2020). It has been observed that BEHI is
34
sensitive to the field measurement of bank height, root depth, bank angle and bank material while the BANCS erosion equations are sensitive to the NBS method used (Bigham et al., 2018).
To combat the issue of field measurement sensitivity, Protocol 1 follows the suggestions of
Bigham et al. (2018) by recommending quality control measures such as certified, region- specific practitioner BANCS training and completion of these measurements by groups of two individuals. To address the sensitivity of NBS, there is a suggestion to research whether it would be feasible to replace NBS entirely with non-form-based, direct estimates of shear by measuring water level at the site and applying the DuBoys equation. This new method could reduce the sensitivity associated with NBS and allow the use of erosion events not associated with bankfull events. This adaptation would result in a larger range of erosion rates and accelerate the curve development process as there would be more than just one or two data points per bank per year.
2.2.3.2 Process-based Erosion Models
Among academics, there has been a push to use process-based models for streambank erosion predictions. The U.S. Department of Agriculture (USDA) developed two alternate, process-based models that quantify bank erosion with cross sections: the Bank Stability and Toe
Erosion Model (BSTEM) and the Conservational Channel Evolution and Pollution Transport
System (CONCEPTS) (McMillan et al., 2017). Both models use similar methodologies to estimate bank erosion by considering fluvial erosion, geotechnical failure, near-bank pore-water pressure, vegetative cover, channel geometry, bank material, bank material layering, and bed material. The differences between the two are that BSTEM assumes a steady and uniform flow and does not allow for erosion of the bed while CONCEPTS uses one-dimensional unsteady flow, calculates bed erosion, and has sediment transport capabilities between cross-sections.
35
However, neither model considers subaerial processes (Klavon et al., 2017). To improve
BSTEM, version 5.4 was integrated into one-dimensional HEC-RAS 5.0 by Gibson, Simon,
Langendoen, Bankhead, & Shelley (2015), which allows the simulation of bank erosion at multiple cross sections over a reach. It should be noted that both CONCEPTS and BSTEM with
HEC-RAS are designed for straight or low sinuosity channels and do not incorporate secondary flows, which is a significant limitation when potentially used in restoration design. To improve upon this limitation, the bank retreat routines in BSTEM and CONCEPTS have been incorporated into two-dimensional models which, in part, account for secondary flows, hopefully yielding more accurate applied shear stress estimates (Klavon et al., 2017). Nonetheless both
BSTEM and CONCEPTS are not commonly used in predicting stream restoration bank stability for crediting with most stakeholders preferring the simplicity of BANCS.
36
3 Methodology
3.1 Research Sites
All the sites are located in the Valley and Ridge physiographic province of SW Virginia, shown in Figure 3-1, which is characterized by northeasterly sandstone or chert ridges separated by valleys of shale or limestone. Due to sandstone and chert forming thin, acidic soils, most ridges are covered by woody vegetation, while the more fertile soils in the valleys have been cleared for agriculture or other development (Chowns, 2018). Precipitation, ranging between 914 to 1219 mm/yr (36 to 48 in./yr), is well distributed throughout the year with slightly more in the summer (Keaton et al., 2005). Streamflow is seasonally highest in spring. Streamflow in the region is closely linked with groundwater due to the karst topology with an estimated 65% to
95% of surface streamflow coming from spring discharges (Virginia DEQ, 2015).
(2) (4) (8 Kilometers)
Figure 3-1. Study site locations and respective watersheds in the Valley and Ridge of SW Virginia.
37
Like many streams in the East and Southeast, all the sites have legacy sediment stored in the floodplain and banks due to intensive land clearing during colonial times. From historic sources such as topographic maps and changes in current channel slope due to a sediment wedge near the site, it is believed that Toms Creek had a milldam directly downstream of the sites, exacerbating the build-up of legacy sediment. There is also circumstantial evidence such as soil characteristics to suggest that a mill dam was also present along Stroubles Creek. Resultantly, all the streams are characterized by incised channels with vertical banks composed of fine sediment, as seen in Figure 3-2.
3.1.1 Stroubles Creek
Stroubles Creek (SC), a tributary to the New River, drains a majority of the town of
Blacksburg and Virginia Tech campus. The recent SC erosion pin site for model validation is at coordinates 37.210891, -80.444770 is shown in Figure 3-2B and has a drainage area of 14.4 km2
(5.56 mi.2), while the older reach-scale erosion dataset, Figure 3-2A, starts approximately 360 m
(1180 ft.) downstream. According to the 2016 National Land Cover Database, the SC watershed is predominantly urban with approximately 86.3% of the drainage basin being developed. The remaining watershed is 10.6% agricultural, predominately pasture, and 2.8% forested. Upstream of the research site are two constructed ponds that is a part of the Virginia Tech campus, the
Duck Pond and Vet Med Pond, which both act as bedload sediment traps. Historically, SC was channelized and cattle had access to the stream channel until 2009, causing high rates of bank retreat in combination with continued urbanization (Wynn et al., 2010).
38
(A) (B)
(C) (D)
Figure 3-2. Images showing banks from A) SC in 2006, B) SC in 2020, C) TC, D) NFR. All study sites have vertical banks composed of fine sediment and are highly erosive.
39
The floodplain and banks are 40% McGary and 35% Purdy soils with a greater than 200 cm (80 in.) depth to a restrictive layer. McGary soil at Stroubles are silt loam and loam from 0 to
60 and 60 to 110 cm (0 to 24, and 24 to 43 in.) in depth, respectively (Wynn et al., 2008). Purdy soils horizons are loam, clay, and clay loam from 0 to 28, 28 to 86, and 86 to 201 cm (0 to 11, 11 to 34, and 34 to 79 in.) in depth, respectively. Both soils are cohesive in nature. The depth to the water table in the McGary and Purdy soils are 30 to 91 cm (12 to 36 in.) and 0 cm (0 in.), respectively. At Stroubles the water table is near the ground surface during the winter but can drop below the stream bed during the summer to become a losing stream.
Banks at this site range between 86 cm (34 in.) and 168 cm (66 in.) in height. Previous studies performed at Stroubles have noted that the banks are sensitive to subaerial processes.
Streambank erodibility was 2.9 and 2.1 times higher in winter and spring/fall than summer, respectively. Freeze-thaw cycling was identified as a significant contributing factor to bank erodibility (Wynn et al., 2008). Other winter weather conditions that contribute to higher retreat rates include increased runoff and groundwater levels due to reduced evapotranspiration.
Stroubles Creek was listed on the Virginia 303(d) list in 1998 for benthic macroinvertebrate impairment with the primary stressor being sediment from construction and bank erosion. A TMDL and implementation plan that included stream restoration were developed and constructed in 2006 and 2010, respectively (Wynn et al., 2010).
3.1.2 Toms Creek
Toms Creek (TC) is a tributary to the New River that drains a small part of the town of
Blacksburg; however, most of the watershed is rural. The monitored site shown in Figure 3-2C
(37.242178, -80.465162) has a watershed area of 32.2 km2 (12.4 mi.2); 51.7% of the watershed is
40
forested, 24.5% is developed, and 23.2% is agricultural. The percent forested area has been consistent since 1938, but approximately 24% of the watershed has changed from agricultural to developed land. Similar to Stroubles Creek, Toms Creek also has 40% McGary and 35% Purdy soils. Soils at the site are known to be silt loam between 0 and 75 cm (0 and 30 in.) in depth with
24 to 26% silt and 6.3 to 7.6% clay. The water table fluctuates seasonally in the same manner as
SC; near the ground surface during the winter and decreasing significantly during the summer.
The measured bank is higher at this site, 235 cm (93 in.). Due to proximity and similarities in soils and climate to SC, Toms Creek is also likely to be sensitive to subaerial processes, particularly freeze-thaw cycling.
3.1.3 North Fork of the Roanoke River
The North Fork of the Roanoke River (NFR) drains to the Roanoke River. The bank location in Figure 3-2D at 37.313555, -80.250970 has a 15.9 km2 (6.14 mi.2) watershed area which is currently 58.6% forested, 36.5% agricultural, 4% developed, and the rest split between barren, shrub, and herbaceous cover. The research site is open to cattle, leading to accelerated erosion at crossings; while there are no cattle crossing on the measured bank, there is a crossing approximately 21 m (68 ft.) upstream of the site. Directly adjacent to the research site is an emergent palustrine wetland believed to be predominantly groundwater fed.
The streambanks and floodplain at the NFR site are composed of Clubcalf soil, which is a cohesive soil of silt loam from 0 to 94 cm (0 to 37 in.) in depth then is a silty clay loam below.
Depth to a restrictive layer is more than 203 cm (80 in.) and the depth to the water table is 0 to
46 cm (0 to 18 in.). The existence of floodplain wetlands indicate that the site has a high water
41
table. The site has banks that are approximately 102 cm (40 in.) in height. In contrast to the SC and TC, NFR erosion rates are believed to be driven by the high water table and cattle access.
3.2 Field Measurements
Two sets of erosion pin data were used in this study: the historic SC reach dataset, and the more recent measurements along SC, the NFR, and TC. The historic Stroubles Creek data includes 249 erosion pins along a 440 m (1440 ft.) reach of Stroubles Creek, shown in Figure 3-
3. Each measured bank had a vertical column of three to five pins starting at baseflow and separated by a height of 30 cm (11.8 in.) (Wynn et al., 2006). Most pin columns were separated by 10 m (32.8 ft.) horizontally, but four severely eroding banks have erosion pins at a 2.0 m
(6.56 ft.) horizontal spacing. Pin lengths were recorded monthly when possible between August
2005 and May 2007 with a single 13 month interval measurement between May 2007 and June
2008. Overall, the dataset covers approximately three years with two years having monthly measurements and one year having one measurement over the year. The dataset is accompanied by a detailed survey of the reach pre- and post-installation, performed May 2006 and May 2007.
Surveying was done using a total station (Leica Geosystems TC407, St. Gallen, Switzerland) that collected survey points every meter down the thalweg and along the bank edges, and repeated cross-section at every pin column location.
The second dataset is from a single bank at each of the three sites: SC, NFR, and TC.
Pins at all sites were set in an array with three to four pins starting at baseflow separated by a height of 30.0 cm (11.8 in.) vertically in a column, and all columns spaced 1.00 or 2.00 m (3.28 to 6.56 ft.) horizontally. In total there were 71 pins with SC, the NFR, and TC having 21, 27, 23 pins, respectively. The channel cross section and longitudinal profile at each bank was measured
42
using a laser level (Topcon RL-H5, Livermore, CA). Stream stage was also measured at each site. The water levels were measured with HOBO readers at intervals of 10 minutes at the NFR and TC, while at SC the water levels were measured at 10 minutes intervals by the Stream
Research, Education, and Management (StREAM) Lab at Virginia Tech using a datalogger
(Campbell Scientific CR1000, Logan, UT) with a pressure transducer (Campbell Scientific
CS451, Logan, UT). Since field visits were conducted every month, if there was equipment malfunction with the HOBOs at the TC or NFR sites, missing flow was estimated using adjacent stream gages. To fill missing flow gaps in recorded SC flow data, a variable source area hydrological model was constructed using R (Schneiderman et al., 2007). The Nash–Sutcliffe model efficiency coefficient for this model was 0.536, which is considered a satisfactory model fit (Moriasi et al., 2007).
3.3 Erosion Pin Measurement Variability
Table 3-1 summarizes section 3.3 by stating the spatial and temporal questions posed, the data set(s) used, what was calculated, and how individual pins were averaged.
43
Table 3-1. Summary of spatial and temporal variability methodologies. Type of Research Question Dataset(s) Used Pin Treatment Analyses Variability Averaging pins in Historic SC bank, columns, and One-way ANOVA with Does the erosion volume calculation (2 m spaced pins) Spatial rows. blocking of banks and method affect the erosion volume (Bank-scale) Spatial interpolation mean separation estimate? New SC, TC, NFR between columns Sensitivity Analysis (All pins) and vertical pins One-way ANOVA with Historic SC blocking of banks and Can measurements of a single column (2 m spaced pins) Spatial Averaging pins in mean separation of or row of pins adequately reflect the (Bank-scale) columns and rows percent error relative to soil volume lost from an entire bank? New SC, TC, NFR bank average array (All pins) method One-way ANOVA and Spatial Does vertical pin placement affect the Historic SC No averaging, pins mean separation of rows (Reach-scale) total reach erosion rate estimate? (All pins) separated into rows (i.e. height from baseflow) How does longitudinal pin spacing and Historic SC Averaging pins in Spatial Visual observation of number along the reach affect the total (10 m spaced pins and column and then (Reach-scale) boxplot estimated reach erosion volume? four 2 m spaced pins) averaging columns Historic SC (All pins) Is there a difference in erosion rates by No averaging, pins One-way ANOVA and Temporal season? separated by season mean separation analysis New SC, TC, NFR (All pins) Time-averaging of Historic SC historic pins, (10 m spaced pins and How many months of data is needed column averages, four 2 m spaced pins) Temporal for the mean and variability of the and reach averages Visual observation
erosion rate to not vary appreciably? Time-averaging of New SC, TC, NFR new SC, TC, NFR (All pins) column averages
44
3.3.1 Spatial Variability
There were two scales examined in terms of spatial variability, the reach-scale and the bank-scale. The bank-scale study used only the banks with 2 m or less horizontal intervals from both the new and historic datasets, while the reach-scale study used the all the pins separated by
10 m horizontal intervals as well as one column of pins from each highly erosive bank with 2 m horizontal intervals. Only the historic dataset was used for the reach-scale analysis. Most of the study focused on calculating erosion volumes, but the need to compare erosion values between the new and historic dataset despite different measurement duration, especially in the bank-scale study, led to the use of a volumetric erosion rate of cubic-meters per year (m3/yr.).
3.3.1.1 Bank-scale
On the bank-scale, two questions were posed: 1) does the erosion volume calculation method affect the erosion volume estimate; and 2) can measurements of a single column or row of pins adequately reflect the soil volume lost from an entire bank. The first question was analyzed using a one-way ANOVA with blocking and separation of means. The second question was explored by estimating the total bank erosion volume per pin column and row, and comparing the results to the estimates when the complete grid of pins on a bank were used.
As stated above, only the pins with 2 m spacing or less were used for the bank-scale analysis. These pin arrays are situated along four banks in the historic dataset, referred to as banks W, X, Y, and Z, as shown in Figure 3-3. All pins in the new dataset were analyzed.
At the bank-scale, erosion volumes were calculated using eight methods. Six methods used all the pins on a bank to calculate one erosion volume, meaning each method yields one erosion volume per bank; these methods were categorized as array methods. Of the six array
45
methods, two were a form of spatial interpolation called average end area, commonly used in transportation engineering cut-and-fill calculations. The remaining methods used one column or row of pins to calculate one erosion volume from the bank, so the methods yielded one erosion volume for every column or row of pins; these methods were categorized as series methods and are more in line with how consultants perform bank erosion volume measurements. These methods are outlined in Table 3-2.
Z
W Y X N
Figure 3-3. Historic Stroubles Creek study reach plan view. Red crosses mark all erosion pin columns, generally separated by 10 m. Banks W, X, Y, Z with 2-m erosion pin spacings are circled in red.
A total of eight averaging method were used. BA, CA, RA, and PA are essentially an erosion depth multiplied by a representative area. BA is the average of all pins on the bank multiplied by the estimated bank area. CA is the average of all the pins in a column multiplied by the area said column represents, summed together for the whole bank. RA does the same as CA using rows instead of columns. PA sums each pin’s individual erosion multiplied by the area represented by the pin. CAEA spatially interpolates between averaged column erosions; rather than assuming each column represents a block of erosion, this method assumes there is a linear transition between the erosion of two adjacent columns horizontally. This method is analogous to
46
using the trapezoidal rule rather than the rectangular/midpoint rule when estimating an integral.
PAEA continues to spatially interpolate between columns while also spatially interpolating between pins in a column; rather than averaging all the pins in a column, this method assumes a linear transition between pins vertically. For both CAEA and PAEA, the boundary conditions are assumed to be zero. The equations used for all the methods are provided in Table 3-2.
The six array methods were used to calculate volumetric erosion rates from each of the seven banks (four historic, three from current study). This data was then used to address the first question above, whether the averaging method used on a bank affects the output erosion. This was tested using a one-way ANOVA with blocking per bank. A separation of means analysis was then performed to show whether any one method was statistically different from any other method. A sensitivity study was also performed on the volumetric erosion rates; the array methods were ranked one to six in ascending order on every bank and the variance of all the ranks for each method was calculated. The method with the lowest variance was viewed as the least sensitive.
The series methods were used to calculate volumetric erosion rates per pin column or row on the seven banks analyzed. Essentially these methods were using CA or RA but instead of multiplying all column or row averages by the representative area of said column or row and summing, OCCA and ORRA multiplies each column or row average by the estimated area of the bank. The output is that each individual row and column resulted in one bank volumetric erosion rate estimate. The absolute percent error was calculated for each volumetric erosion rate using the BA estimation as the known value, as shown in Equation 3. One-way ANOVA with blocking of bank and mean separation were performed on the percent errors of the two methods to find whether they were statistically different.
47
|푆푒푟푖푒푠 푀푒푡ℎ표푑 퐸푠푡푖푚푎푡푒−퐵퐴 퐸푠푡푖푚푎푡푒| 푃푒푟푐푒푛푡 퐸푟푟표푟 = × 100 (3) 퐵퐴 퐸푠푡푖푚푎푡푒
Table 3-2. Methods used to calculate erosion volume. EDpin equals the erosion depth of a pin while ̅퐸퐷̅̅̅ indicates the averaged erosion depth of a bank, column, or row. ℎ represents the vertical distance associated with a bank, column, row, or pin with ℎ̅ indicating the averaged vertical distance for a bank, column, or row. 푥 indicates the horizontal distance associated with a bank, column, row, or pin. 푉퐵 is the calculated erosion volume of the bank. Name Method Bank Erosion Volume Equation (Abbreviation) Category Bank Average Array 푉 = ̅퐸퐷̅̅̅ ℎ̅ 푥 (BA) 퐵 푏푎푛푘 푏푎푛푘 푏푎푛푘 푛 Column Average Array 푉 = ∑ ̅퐸퐷̅̅̅ ℎ 푥 (CA) 퐵 푐표푙푢푚푛,푖 푐표푙푢푚푛,푖 푐표푙푢푚푛,푖 푖=1 푛 Row Average Array 푉 = ∑ ̅퐸퐷̅̅̅ ℎ 푥 (RA) 퐵 푟표푤,푖 푟표푤,푖 푏푎푛푘 푖=1 푛 Pin Area Array 푉 = ∑ 퐸퐷 ℎ 푥 (PA) 퐵 푝푖푛,푖 푝푖푛,푖 푝푖푛,푖 푖=1 푛 (퐴푖 + 퐴푖+1) Column Average 푉 = ∑ 푥 , 퐵 2 푐표푙푢푚푛,푖 End Area Array 푖=1 (CAEA) ̅ 퐴푖 = ̅퐸퐷̅̅̅푐표푙푢푚푛,푖 ℎ푐표푙푢푚푛,푖
푛 (퐴 + 퐴 ) 푉 = ∑ 푖 푖+1 푥 , Pin Average 퐵 2 푐표푙푢푚푛,푖 End Area Array 푖=1
(PAEA) 푛 (퐸퐷푗 + 퐸퐷푗+1) 퐴 = ∑ ℎ 푖 2 푝푖푛,푗 푗=1 One Column Column Average One-Series ̅̅̅̅ (OCCA) 푉퐵 = 퐸퐷푐표푙푢푚푛ℎ푏푎푛푘푥푏푎푛푘 One Row Row Average One-Series ̅̅̅̅ (ORRA) 푉퐵 = 퐸퐷푟표푤ℎ푏푎푛푘푥푏푎푛푘
48
3.3.1.2 Reach-scale
The reach-scale spatial analysis examined whether vertical pin placement affects the total reach erosion rate estimate and how longitudinal pin spacing and number along the reach affects the total estimated reach erosion volume. As previously stated, this analysis used all the 10 m spaced pins along with one column of pins from banks W, X, Y, and Z, so that overall 46 pin columns were analyzed.
For the first question posed, pins were grouped by rows. Since the rows started at baseflow and were separated by 30 cm (11.8 in.) vertically, rows A, B, C, D and E represented a distance from baseflow of 0.0 m (0.0 ft.), 0.30 m (1.0 ft.), 0.60 m (2.0 ft.), 0.90 m (3.0 ft.), 1.2 m
(3.9 ft.) and 1.5 m (4.9 ft.), respectively. ANOVA and separation of means analyses were then performed to see whether the erosion rate estimates based on any single row was significantly different.
For the second question posed, each pin column was averaged, so that there were 46 column erosion depths throughout the 440 m (1444 ft.) reach, which represented 434.4 m2 (4676 ft.2) of eroding bank area. To simulate different pin spacings and numbers, an exercise was performed where reach erosion volumes were repeatedly calculated using, at first, one pin column to represent the whole reach, and then increasing the number of pin columns used by one to calculate reach erosion until all the pin columns were used. All column erosion values used to calculate reach erosion were averaged and then multiplied by the bank area. This is represented in Equation 4 below:
퐴 푉 = 푇퐵 ∑푛 ̅퐸퐷̅̅̅ (4) 푅 푛 푖=1 푐표푙푢푚푛,푖
49
2 2 where 푉푅 is the reach erosion volume (m ), 퐴푇퐵 is the total eroding bank area equal to 434.4 m ,
̅퐸퐷̅̅̅푐표푙푢푚푛 is the averaged erosion depth from a column of pins (m), and 푛 is the number of
columns being used to calculate erosion volume.
One pin column being used represented the largest spacing of 440 m (1440 ft.) per pin
column, while all the pin columns being used represented the smallest spacing of 9.54 m (31.1
ft.) per pin column. When using multiple pin columns to calculate reach erosion, it was assumed
that the pins were evenly spaced. Due to this constraint, the number of erosion volumes
calculated generally decreased as the number of pin columns used increased. For example, when
only one pin column was used to calculate a reach erosion, 46 reach erosion volumes were
calculated. When two evenly spaced pin columns were, it yielded 24 reach erosion volumes.
Finally, when all the pins were used to calculate a reach erosion, only one erosion volume was
yielded. After all the reach erosion volumes were calculated, values were plotted as boxplots
against number of pin columns and column spacing with the aim of quantifying at what spacing
and number of pin columns the median and variation in erosion volumes becomes consistent.
3.3.2 Temporal Variability
Two questions were posed temporally: 1) how many months of data is needed for the
mean and variability of the erosion rate to not vary appreciably, and 2) is there a difference in
erosion rates by season? The first question was explored by calculating time-averaged erosion
rates for each month cumulatively. These values were then plotted as boxplots and statistically
compared using ANOVA and mean separation. The second question was tested by separating the
data by seasons, then performing an ANOVA and separation of means analysis. All pin erosion
50
rates were calculated as the erosion depth in centimeters divided by the time since the last measurement in days. These erosion rates were then converted to centimeters per year.
3.3.2.2 Seasonal Variability
To analyze the seasonal variability, the seasons were divided by month for both historic and new datasets with Fall, Winter, Spring, and Summer being September to November,
December to February, March to May, and June to August, respectively. July 2006 in the historic dataset was removed from the data because a large flood, shown in Figure A-1, caused an outlier of high erosion rates for the summer months. All the seasonal data was plotted on a boxplot and statistically tested for significant difference with ANOVA and mean separation analyses on both the historic and new datasets.
3.3.2.2 Temporal Stabilization
To estimate the minimum number of months of erosion data needed to predict the mean, erosion rates were time-averaged, meaning that each monthly erosion rate would be calculated as the average of all previous months up to the current month. For example, the average erosion rate for the third month would be the average of the first, second, and third months. This computation was done with differing spatial scales. The historic dataset was used to perform this analysis on a reach-scale using both pin erosion rates and average column erosion rates, and the new dataset was used to examine time-averaging on the bank-scale using averaged erosion rates using only column of erosion pin measurements.
The analysis was completed using both the 2 m and 10 m pins when using pin erosion rates, but only the 10 m pins when based on column erosion rates using the historic dataset.
Time-averaged pin erosion rates for each pin per month was calculated by averaging pin erosion
51
rates for all previous months up to the current month. The time-averaged column erosion rates were then calculated by taking the column average of the time-averaged pin erosion rates in every column. The data was visualized as boxplots, and the variance of the datasets were then calculated and plotted. Another approach was to calculate the time-averaged reach erosion rate and standard deviation. This computation was done simply by taking the mean and standard deviation of all the time-averaged column erosion rates.
Statistical analysis was performed on the time-averaged pin erosion rates; analysis was not performed on the column and reach-scale data because preliminary results were fairly similar in nature between the three spatial scales and a larger dataset was preferred. An ANOVA and mean separation analysis were performed on all available months. Whole 12-month sets of data would have been preferred to represent a total year, but a large summer flood in July 2006, shown in Figure A-1, caused the average erosion and erosion variance at the 12th month to increase and there was a data gap between the 23rd and 37th month.
3.4 Assessment of the BANCS Model
An issue with the BANCS model is its use of the bankfull depth and/or discharge, both in the erosion rate measurements and the NBS methods. To address said issue, two questions are posed: 1) can the erosion rate input be altered as to not rely only on bankfull erosion events and to better reflect long-term erosion; and, 2) can NBS be calculated in a manner that does not require a bankfull discharge and can represent a broad range of flows?
To answer these questions, multiple BANCS erosion curves were constructed for the
Valley and Ridge physiographic province using the historic Stroubles Creek erosion dataset. All
52
curves were created through linear regression with the linearization of a power relationship from
Equation 5 to Equation 6:
퐸 = 푎푒푏푋 (5)
ln 퐸 = 푏푋 + ln 푎 (6) where E is the erosion rate in cm/yr, X is an indicator of applied fluvial shear, and a and b are constants. Measured erosion rates from TC, SC, and NFR were plotted in every erosion curve but these datapoints were not included in the regression analysis. It should be noted that linear regressions were also performed but no statistically significant relationships were developed.
3.4.1 Input Erosion Rates
To address the first question above, different bank retreat rates were used in the analysis, including only bank retreat measured per bankfull event, all monthly erosion rates, or bank retreat time-averaged over approximately one year. The commonly used BANCS model only plots erosion rates for each bankfull event. The erosion rates were also separated into BEHI categories, with regression only occurring within a BEHI category. Plotting erosion rates per bankfull event required the identification of bankfull stage or flow in the study reaches. This identification was done using a combination of bankfull indicators in the field and examining constructed stage-discharge relationships where available. Along with this data, flow data was needed to identify months with flood events of only bankfull flow.
The time-averaged curves aimed to use 12 month time-averaged erosion rates to capture a year’s worth of seasonal variability. While this was the ideal case, the data did not always cover
12-month periods. The erosion rates for the first year were dramatically higher in summer due to a large summer flood, so regression was attempted with and without this high erosion event
53
included in the data. Another issue was that the dataset also only went up to 23 months rather than 24 months, with the second year missing one month. Due to this, the second time-averaged year only contained 11 months rather than 12 months. Finally, a third year could be added to the time-averaged curves, but it covered 13 months. The third year was also technically not time- averaged, since monthly measurements were not made during the last year. Overall, the time- averaged regression analyses were conducted with four sets of time-averaging: three years with the first year including the summer flood; three years with the first year not including the summer flood; the first two years with the first year including the summer flood; and, the first two years with the first year not including the summer flood. Since erosion as averaged over approximately 12-months, the three year average datasets had three erosion rates per study bank and the two year average datasets had two erosion rates per study bank.
3.4.2 NBS Measurement
To address the second question above, new NBS measures were developed based on fluid mechanics and used in the regression analysis. The results were compared to the traditional NBS methods. All curves were grouped into two categories based on the fluvial shear parameter. The two categories were the traditional NBS methods, where the curves were based on already established NBS methods, or the modified BANCS model, where the existing NBS methods were substituted for another indicator of applied fluvial shear.
3.4.2.1 Traditional NBS
All the banks used were categorized into the qualitative BANCS categories for NBS and
BEHI. Because this analysis was completed with a historic dataset and the stream channel had
54
been altered during a stream restoration, the measured banks could not be placed into NBS and
BEHI categories through field observations; rather, the survey that accompanied the dataset was used to quantify these parameters. Due to the data limitation previously stated, only methods 2,
3, 4 and the modified 5 were used to calculate NBS values, which are ratio of curvature to bankfull width, ratio of pool slope to average slope, ratio of pool slope to riffle slope, and thalweg position relative to the study bank, respectively (Altland, Berg, et al., 2020; Rosgen,
2006). The methods are shown in more detail in Figure 2-4. All NBS values were kept continuous on a scale between 1 and 7 by interpolating between the calculated value and the qualitative conversion for each NBS method shown in Table 2-1. Like NBS, the BEHI for the historic SC dataset was calculated using the survey and photographs rather than in the field.
Regressions were performed for each individual NBS used, as well as by using the highest NBS found for each bank, as recommended by Rosgen (Rosgen, 2006; Rosgen et al.,
2019). In total, there were seven NBS options. With the three temporal erosion rate options, there were 15 regressions performed per BEHI category using the traditional NBS methods.
3.4.2.2 Modified NBS
Another aim of the study was to find an indicator of fluvial shear with a stronger theoretical basis than the traditional NBS methods. Four proposed alternatives were tested: the number of peaks above baseflow, the percent of time above baseflow, an estimation of shear using the DuBoys formula, and a representation of shear using DuBoys formula combined with a standardized radius of curvature. The former two methods were labelled as hydrograph-based methods while the latter two methods were labelled as modified DuBoys methods. The hydrograph-based methods were only run with all monthly erosion rates while the modified
55
DuBoys methods were run using only bankfull events, all events, and time-averaged erosion.
This resulted in eight regressions per BEHI category.
3.4.2.2.1 Hydrograph-Based Models
The two hydrograph-based curves are the number of peaks above baseflow and percent time above baseflow, which are called such because a discharge or stage hydrograph is required to calculated NBS using these methods. The hydrographs used in this study are shown in
Appendix A. For the historic SC dataset, baseflow is estimated to be 0.28 cms (1.0 cfs). A peak is defined as being at a flow series where flow decreases at least one timestep after a local maximum. The number of peaks within each measurement date was counted, which served as an estimate of applied fluvial shear. The period of time above the baseflow for any given erosion pin measurement interval was estimated and divided by the total time of the measurement interval. This fraction was then multiplied by 100 to determine the percent of time above baseflow for a given measurement interval.
3.4.2.2.2 DuBoys Formula
The most recent Chesapeake Bay recommendation suggested that NBS could be substituted with the DuBoys equation, shown in Equation 7:
휏푏 = 휌푔푅푆 (7) where 휌 is the density of water, 푔 is the acceleration due to gravity, 푅 is the hydrailuc radius calculated as cross-sectional area of flow (퐴) divided by wetted perimeter (푃), and 푆 is the stream slope. A simplification of DuBoys was used instead, which removes 휌 and 푔 since these will be constant at any site and simplified 푅 to ℎ, commonly known as the wide-channel
56
assumption, representing the hydraulic depth of the first level floodplain surface. This simplification is shown in Equation 8, where 푋 represents an indicator of fluvial shear.
푋 = ℎ푆 (8)
It has been shown that applied shear stress and bank retreat rates within a meander bend can vary depending on the radius of curvature (Hickin & Nanson, 1984; Ursic & Thorton, 2011), so the equation was modified to account for increased shear stress in meander bends by dividing by the radius of curvature (푅푐) standardized by stream width (푊), as shown in Equations 9 and
10.
ℎ푆 푋 = 푅 (9) ( 푐) 푊
ℎ푊푆 푋 = (10) 푅푐
Both of these values were calculated per bank and can be obtained with a simple survey, similar to the traditional NBS values. However, unlike the traditional NBS, these values are not fit into the seven qualitative categories.
3.4.4 Comparing Erosion Curves
Erosion curves were assessed in two ways. First, the regression equations were evaluated statistically to determine if the regression coefficients were significantly different from zero and if the equation explained the variance in the data set. Second, the predicted erosion rates from each curve were compared to the measured long-term erosion rates of the new, bank-scale erosion dataset.
When examining the regressions, curves were compared using the r2, predicted r2, and, in particular, p-value parameters. Curves that had a p-value within a 90% confidence level (p-value
< 0.10) were considered acceptable. Of those with acceptable p-values, the curve was examined
57
to determine whether the slope was positive or negative. A negative slope indicated that the erosion rate decreased with the applied fluvial shear stress, which does not conform to the expected fluvial erosion process of increasing fluvial erosion with applied shear stress.
The three banks from the new dataset, TC, SC, and NFR, were categorized by BEHI category and every used indicator of NBS so that the predicted erosion rates could be estimated from each erosion curve. For the new bank pin dataset, classifying NBS and BEHI was possible in the field, and so was performed as such after basic training provided by practitioners. The observed erosion rates for the three banks were also calculated to compare to the predicted erosion rates. This measured erosion was time-averaged so predictions could be compared to long-term erosion measurements that are needed as part of the regulatory framework. These values were plotted on the erosion curves and used to calculate a percent error for each predicted erosion rate.
Predicted erosion rates were found for several curves constructed in this study and two existing BANCS curves. Of the curves constructed in this study, predicted erosion rates were found if a significant relationship was developed (p-value < 0.10). Additionally, predicted erosion rates were estimated with the constructed erosion curve that utilized the traditional
BANCS framework; only using erosion rates from bankfull events and using all possible NBS methods and applying the more erosive method. For the hydrograph-based curves, the number of peaks or percent of time above baseflow were averaged over a year so that one prediction value could be obtained per site. The two existing curves applied, shown in Figure 3-4, are the North
Carolina (NC) curve and U.S. Fish and Wildlife Service (USFWS) curve, both of which are not included in Table 2-2 due to lack of known data. The NC curve was created for the Piedmont physiographic province and little can be found about its inception. The curve consists of 30
58
datapoints, all with unique BEHI and NBS combinations, over four BEHI categories. The
USFWS curve was created using data from Hickey Run, Watts Branch, and Oxon Run, all of which are located in the Coastal Plain physiographic province (Secrist, n.d.). There is a total of
46 datapoints over four BEHI categories with 30 unique BEHI and NBS combination. The
USFWS is the curve currently applied in the Valley and Ridge despite it being created in the
Coastal Plain physiographic province.
Once predicted erosion rates were estimated, the percent error was calculated using
Equation 11.
푃푟푒푑푖푐푡푒푑−푀푒푎푠푢푟푒푑 푃푒푟푐푒푛푡 퐸푟푟표푟 = × 100 (11) 푀푒푎푠푢푟푒푑
The percent errors were analyzed by comparing different groupings using one-way
ANOVA and mean separation analyses. Percent error groupings were between sites, input erosion rates, NBS methods, and constructed and existing curves. For all the analyses except the site grouping, site blocking was used. Additionally, only the groupings of constructed and existing curves used the existing curves; the analysis of other groupings only included curves created in this study.
59
(A)
(B)
Figure 3-4. Established BANCS curves identified as A) the North Carolina curve for the Piedmont (McQueen, 2011), B) the USFWS curve for the Coastal Plain (Rathbun, 2009).
60
4 Results & Discussion
4.1 Erosion Pin Measurement Variability
4.1.1 Spatial Variability
4.1.1.1 Bank-Scale
The results for the array methods are shown in Figure 4-1. The resultant volumetric erosion rates from the same dataset vary depending on the method used. The bank average (BA) and row average (RA) methods appear to consistently have the highest volumetric erosion rates, while the two spatial interpolation methods, the CAEA and PAEA, have the lowest volumetric erosion rates. To determine if the differences in volumetric erosion rates by method were statistically significant, ANOVA and least mean squared analyses were performed. The results in
Figure 4-2 indicates that how one calculates bank erosion volume significantly affects the erosion volume estimate. More specifically, the erosion volumes estimated using the BA and RA methods are significantly higher than the estimates using PAEA spatial interpolation methods.
The erosion estimates using the CA, PA, and CAEA methods are not statistically different from estimates using any of the other methods.
61
Figure 4-1. Array method estimates of bank volumetric erosion rates suggesting that how one calculates bank erosion volume significantly affects the erosion volume estimate
Figure 4-2. Array method results of one-way ANOVA with blocking of banks. Boxes represent model least square mean with the 95% confidence interval indicted by the error bars. PAEA is statistically different from BA and CA, exhibiting that the method of calculation can significantly affect the erosion estimate. 62
For the array methods, the two spatial interpolation methods, CAEA and PAEA, have the lowest erosion volumes, while the BA and RA methods have the highest erosion volumes. The issue with these results is that the true erosion volumes are unknown, so the accuracy of each method is also unknown. A more suitable manner to gage which array method is better for application is to find the method with the least sensitivity to a single pin, column or row on the bank. Table 4-1 show the results of the sensitivity study with the lowest variance signaling the least sensitive method, which is PAEA, followed by the BA and RA. The PAEA method being the least sensitive makes sense as the method allocates the least spatial weight to a single pin, column or row.
Table 4-1. Rank of array methods per bank and resulting variability for sensitivity study Banks BA CA RA PA CAEA PAEA W 5 3 6 4 1 2 X 6 4 5 3 2 1 Y 5 3 6 4 1 2 Z 5 3 6 4 1 2 TC 4.5 4.5 4.5 4.5 2 1 SC 4 2.5 5 2.5 6 1 NFR 3.5 5.5 3.5 5.5 2 1 Variance 0.65 1.14 0.89 0.95 3.14 0.29 (m6/yr2)
Given that an array of pins is used to measure the bank, practitioners should use the
PAEA interpolation method since it is the least sensitive to extreme values by individual pins, rows, or columns. Baring this due to difficulty, the BA and RA methods should be used since these are the least sensitive methods following PAEA. Unfortunately, most consultants do not use pin arrays to represent a bank. Many only measure bank retreat along a single column per bank, so these forms of averaging and spatial interpolation are not possible. For this reason, the series methods are discussed below.
63
The erosion volumes per pin column or row as calculated using the OCCA and ORRA methods are shown in Figure 4-3A and Figure 4-3B, respectively. The figure suggests that the estimated bank erosion volume by column or row vary, suggesting that soil loss along a single bank is not represented well by a single column or row. The mean percent error relative to the array method BA estimate for OCCA and ORRA was 32.3% and 22.9%, respectively. The
ANOVA and mean separation analyses between the percent errors of OCCA and ORRA, shown in Figure 4-4, indicate that two methods were not statistically different using a 95% confidence level. While not graphically shown, the two methods are statistically different for any confidence level less than 90%. Although not statically different, the results suggest that rows tend to vary less from the array BA estimate, indicating that soil loss from a single bank is better represented with a single row of measurements than a single column. This finding is supported by Figure 4-5 which illustrates that rate measurements based on different pin rows are not statistically different over long periods of time, supporting the idea that the vertical resolution of measures does not have to be increased.
64
(A)
(B)
Figure 4-3. Series method bank volumetric erosion rates using A) OCCA and B) ORRA with BA array method estimate shown as dashed line. Using one column or row of measurement to estimate total bank erosion leads to high variability. 65
Figure 4-4. Series method percent error one-way ANOVA with blocking of banks and mean separation analyses results. Boxes represent model least square mean with the 95% confidence interval indicted by the error bars. Result indicate that, while statistically the same, using rows rather than columns to estimate bank erosion has a lower percent error.
4.1.1.2 Reach-Scale
Figure 4-5 visualizes the results of the ANOVA and mean separation analysis of reach- scale erosion rate estimates based on a single row of erosion pins per bank. The figure suggests that different pin rows, represented by height from baseflow, result in statistically similar erosion rate estimates, using a 10% confidence level on a reach-scale and over long periods of time.
Even when using a 5% confidence level, only the erosion rate estimate from the top row, 1.2 m
(3.9 ft.) from baseflow, is significantly different than the other measurements. The top row also has notably more variability. The two differences along the top row of pins are likely due to fewer measurements occurring on the 1.2 m level; along the reach, only six of the 68 measured banks were tall enough to accommodate the 1.2 m pins. Given that bank retreat involves erosion
66
of the bank toe, followed by mass wasting of the upper bank, if one retreat process is restricted
(e.g. the bank toe is resistant to erosion), the retreat of the entire bank will be limited. Similarly, erosion of a bank toe will eventually lead the rest of the bank eroding. This cyclic retreat process is not reflected longitudinally; if one column in the bank erodes, this does not necessarily mean adjacent columns of bank will erode. This finding indicates that, if there are limited resources, practitioners should increase the number of longitudinal measurements and decrease the number of vertical measurements.
Figure 4-5. Pin row erosion rate results from one-way ANOVA and mean separation analyses displaying model least square mean and 95% confidence interval by height from baseflow. Erosion rate estimates over long periods of time do not appear to be affected by height from baseflow, indicating that measuring columns is redundant.
Figure 4-6 are boxplots of reach-scale erosion estimates by pin column spacing. The results suggest that the mean soil loss estimate stabilizes and the variability decreases at pin
67
column spacings less than or equal to 28 m (~ 92 ft.) or three channel widths. The variability decreases further at a spacing of 14 m (~ 46 ft.), which is 1.5 channel widths.
Figure 4-6. Reach erosion volume estimates organized pin column spacing. Erosion volume estimates appear to converge to a single value and decrease in variability as spacing decreases.
As indicated by the results, the mean stabilizes and variability dramatically decreases once the pin column spacing reaches 28 m (~92 ft.). At the moment, the Chesapeake Bay
Protocol 1 recommends a minimum spacing between approximately 60 m (200 ft.) and 150 m
(500 ft.) (Altland, Berg, et al., 2020), which for Stroubles Creek is a measurement spacing of 6.4 and 16 channel widths, respectively. If the recommended Bay spacings were to be used on this site, the erosion rate measured in the 440 m long reach could be represented with between three and seven pin columns with a spacing of 150 m and 63 m, respectively. Using the 150 m (500 ft.) spacing, the measured reach erosion volumes could range between 41.5 m3 (1466 ft.3) and 212.4
68
m3 (7501 ft.3), depending on the selection of measurement locations, as compared to the higher spatial resolution erosion volume of approximately 115.5 m3 (4079 ft.3) at a spacing of 28 m.
The higher measurement spacing results in a possible difference of 170.9 m3 (6035 ft.3) between erosion measurements along the same reach, with a possible underestimation of 64% or overestimation of 84% relative to the higher spatial resolution erosion volume. Using the 60 m
(200 ft.) spacing, the measured reach erosion volume could range from 85.0 m3 (3002 ft.3) to
138.0 m3 (4873 ft.3), an underestimation of 26% or overestimation of 19% relative to the higher spatial resolution erosion volume. While these results are limited to this study site, the bank stratification and riparian vegetation along this reach of Stroubles Creek was very consistent at the time these measures occurred, which implies that stream reaches with greater variability would require smaller measurement spacings to adequately quantify reach-scale bank retreat volume and rate. Therefore, it is recommended additional studies be conducted to confirm this finding at other locations and for practitioner to utilize smaller measurement spacing in the interim.
4.1.2 Temporal Variability
4.1.2.1 Seasonal Variability
Bank retreat rates by season were also calculated (Figure 4-7). The historic Stroubles dataset, as shown in Figure 4-7A, has two statistical groups showing that winter and spring erosion rates are statistically different from summer and fall erosion rates. While the results for
TC and SC from the new dataset, Figures 4-7B and 4-7C, respectively, are not consistent about spring erosion, both show winter as statistically different from summer and fall. This is in line with previous research showing higher bank erodibility during the winter months (Wynn et al.,
69
2008). The results from the NFR in Figure 4-7D do not match this trend, instead showing that there are no statistically significant differences in bank retreat rates among seasons. For SC and
TC, seasonality has a large effect on erosion rates, particularly in winter where freeze-thaw cycling and other winter condition would increase erosion in banks (Lawler, 1993a; Wynn et al.,
2008). While freeze-thaw cycling occurs in the winter at SC and TC due to the temperate climate, colder climates have banks that stay frozen in winter with freeze-thaw occurring in fall/spring. Alternatively, warmer climates have banks that do not experience freeze-thaw cycling at all because the banks never freeze.
The difference in results from NFR can be explained in a number of ways. The site climate should be quite similar to the other sites as they are all located in the Valley and Ridge physiographic province, with similar seasonal precipitation and temperatures. The NFR watershed has more agricultural and forested land than the SC and TC watersheds, with only 4% of land developed in the NFR watershed as compared to 86% and 25% in the SC and TC watersheds, respectively. While this causes the NFR to be less flashy, it does not explain the lack of seasonal variability. It is believed that the true cause relates to freeze-thaw cycling, or lack thereof. There are two possibilities, either the dataset does not have enough winter data, or freeze-thaw cycling is simply less important at NFR.
70
(A) (B)
(C) (D)
Figure 4-7. Difference in erosion rates by season found using one-way ANOVA and mean separation analysis for A) historic SC reach-scale dataset, B) TC bank-scale dataset, C) modern SC bank scale dataset, and D) NFR bank-scale dataset. Boxes represent model least square mean with the 95% confidence interval indicted by the error bars. All sites except NFR have erosion rates affected by seasonality.
71
The first possibility is that the new dataset only includes one month of winter data for the
NFR, December 2020, while three months of winter data is included in the SC and TC data,
January, February and December 2020. Since the one winter month in NFR was the beginning of winter, it is possible that the dataset missed the full effect of freeze-thaw cycling that occurs when temperatures go below freezing more often. Measured atmospheric temperature data inidcates the NFR site dropped below freezing 60 times in the measurement period while air temperatures at the SC and TC sites dropped below freezing 89 times during the measurement period. The other possibility is that freeze-thaw cycling is less of a driving force at NFR because the bank has a lower silt content than TC and SC. The pore size distribution in silty soils promotes the movement of soil water to the bank face due to capillary action, creating large ice crystals and physically disturbing the soil structure as the soil freezes overnight. It is also likely that the NFR site is driven by another mechanism of erosion entirely. The water table at NFR was high throughout the study and field observations indicated bank instability at NFR was caused by piping erosion (Fox & Wilson, 2010). At the site, the two pin rows closest to baseflow consistently only had erosion directly around the pin; it is believed that the pins themselves created preferential flowpaths for shallow groundwater and piping around the pin. If this is the case, the measurement of erosion with pins increased the erosion rate of the banks. This may explain why the erosion rates measured at NFR were consistently higher than those at TC and SC regardless of season, and despite the watershed being more rural and the study bank height being lower. Thus, it is believed that the bank retreat measured at the NFR is more influenced by continuous piping erosion, exacerbated by the erosion pins, rather than seasonal freeze-thaw like the other two sites.
72
4.1.2.2 Temporal Stabilization
Time-averaged erosion rates along Stroubles Creek at the pin-, column-, and reach-scales are shown in Figure 4-8. Each scale exhibits the same trend; as more data is added over time, the mean does not entirely stabilize, but the variance/standard deviation decreases and becomes relatively constant.
Figure 4-9 displays the time-averaged, column-averaged erosion rate results at the TC,
SC and NFR banks sites, comparable to Figure 4-8B. In Figure 4-8B, the historic SC reach, the data jumps around due to a larger flood event in July 2006, shown in Figure A-1, and seasonality when months 6, 12, 21-22 are included in the temporal averaging with an overall increase in time-averaged erosion rate over the measurement duration. Figures 4-9A and 4-9B, the TC and
SC bank sites, show a general decrease in erosion rates as more months are added to the temporal averaging; high erosion rates at the start of the measurement duration due to winter freeze-thaw cycling are balanced by more moderate erosion rates as the year progresses. On the other hand,
Figure 4-9C, the NFR bank site, has little change to erosion rates as more months are added to the temporal averaging. This discrepancy in time-averaged erosion results between the NFR and the SC and TC sites are due to the effect of seasonality on each respective bank, as discussed in section 4.1.2.1. Unlike the historic dataset, all the new bank sites in Figure 4-9 saw no large flooding in the recoding period, as shown in Figures A-2, A-3, and A-4.
73
(A) (B)
(C)
Figure 4-8. Time-average erosion rates for each measurement duration starting August 2005 for individual A) erosion pins, B) column averages, and C) the entire reach. All spatial scales show that the erosion rate estimate does not converge as more erosion data is averaged. 74
(A) (B)
(C)
Figure 4-9. Time-averaged column erosion rates with different durations at: A) Toms Creek bank starting January 2020, B) Stroubles Creek bank using new dataset starting January 2020, and C) North Fork of the Roanoke River starting March 2020. TC and SC time-averaged erosion rates appear to level while NFR does not, likely due to differences in erosion rate seasonality between the sites.
75
Figure 4-10A plots the variance of the historic SC dataset using pins and columns. Even though average erosion rates shift up and down in Figure 4-8B, the variability in Figure 10-A rapidly decrease in the first year of measurement, particularly in with first two months of measurements averaged. There is a large increase in variability in the month 12 due to the large
July 2006 flood in that recording period, shown in Figure A-1. It is believed that if this flood did not occur, the variability would have stabilized during the 12th month. Figure 4-10B affirms this trend, with the variance of TC and SC decreasing rapidly in the first two months and stabilizing by the 12th month. Although NFR does not record 12 months of data, Figure 4-10B appears to show the same trend in NFR as SC and TC up to month 10. Variability decreases when measuring an entire year of retreat because as more months of erosion are averaged into the data, the measurement moves towards the central tendency and extreme retreat rates of particular months are balanced by more moderate erosion of other months.
The results show that at least 12 months of data is necessary for the mean erosion rate to be similar, and 12 months of data is needed for erosion rate variability to be similar. These results indicate that erosion is better represented when using time-averaging rather than erosions per bankfull event as is currently the practice. This finding is supported by previous research regarding erosion pins which indicates that using pins over longer time intervals is more appropriate than over short time intervals (Boardman & Favis-Mortlock, 2016). While this research is directly discussing erosion pins, this trend is likely appropriate for any short-interval, single location erosion measurement such as repeated cross-sections and bank profiles. Using 12- months of data allows for the measurement of bank retreat to account for seasonal changes in erosion rates, which is of particular importance in regions where freeze-thaw cycling occurs.
76
(A)
(B)
Figure 4-10. Time-averaged erosion variance plotted for A) historic Stroubles Creek pins and columns duration starting August 2005, and B) new datasets columns with TC, SC, and NFR durations stating in January 2020, January 2020, and March 2020, respectively. Variance decreases rapidly in the first two month and then less rapidly the remaining months for all sites 77
In Figures 4-8, 4-9, and 4-10 above there is temporal jump in data between the 23rd and
37th month, which is the one year interval where erosion was only measured once. Figure 4-8B suggests that the erosion rate of the one year interval reading is similar to the erosion rate of a one year time-averaged reading. In terms of measurement methodology, the result suggests that instead of taking monthly measurements and averaging the values for a year, one could simply take one reading per year. Of course, if one were to do this, the erosion rates from outlier flood events could not be excluded from the analysis and pin losses would be much more likely.
4.1.3 Recommendations for Bank Retreat Measurement
The spatial portion of this study consisted of bank-scale and reach-scale analyses. On the bank-scale, if measuring uses point measurements such as erosion pins, it is recommended that an array of measurements is performed and erosion volumes are calculated by spatially interpolating both horizontally and vertically. If only one series of measurements can be performed on a bank, it is preferable to measure one row rather than one column. On the reach- scale, it is recommended to measure bank retreat every three bankfull widths.
Temporally, it is recommended retreat measurements use at least 12 months of data to capture retreat seasonality. These measurements may either be taken at shorter intervals such as months and averaged, or the interval itself may be 12 months. Be aware that while longer intervals may require less effort, if using erosion pins, the likelihood of pin losses are higher. It is also recommended to avoid the use of erosion pins if there are indications of piping erosion on the bank.
78
4.2 Assessment of the BANCS Model
BEHI results are shown and discussed in the context of this study. Erosion curves were created changing the input erosion rate and the NBS values. Three input erosion rates were used: erosion rates at bankfull discharges only, all monthly erosion rates, and one year time-averaged erosion rates. Two broad categories of NBS were used: traditional NBS, which used NBS methods 2, 3, 4 and modified 5; and modified NBS methods, which were further divided into hydrograph-based and modified DuBoys categories. The hydrograph-based NBS values were number of peaks above baseflow and percent time above baseflow. Due to the nature of the hydrograph-based curves, the only erosion rate input used was all monthly erosion rates. The modified DuBoys curves substituted a modification of the Duboys formula and the same modification divided by a standardized radius of curvature for NBS, which are Equations 8 and
9/10, respectively. Each section discusses specific curves as necessary with Table 4-2 displaying all regression statistic results and Appendix B showing all 23 constructed curves. Percent error in erosion estimates was calculated for all statistically significant curves, the traditional BANCS curve, and two existing, in-use BANCS curves. The percent errors were compared between sites, input erosion rates, NBS methods, and constructed and existing curves.
79
Table 4-2. Constructed curve regression statistics. Erosion Curve Regression Statistics Erosion Input NBS p-value r2 Predicted r2 Only Bankfull All NBS, highest 0.143 0.0237 0 Erosion selected Only Bankfull Only Method 2 0.742 0.0012 0 Erosion Only Bankfull Only Method 3 0.056 0.0400 0 Erosion Only Bankfull Only Method 4 0.583 0.0034 0 Erosion Only Bankfull Only Method 5 0.668 0.0020 0 Erosion All Monthly All NBS, highest 0.324 0.0017 0 Erosion selected All Monthly Only Method 2 0.186 0.0030 0 Erosion All Monthly Only Method 3 0.964 0 0 Erosion All Monthly Only Method 4 0.727 0.0002 0 Erosion All Monthly Only Method 5 0.853 0.0001 0 Erosion One Year Time- All NBS, highest 0.424 0.0057 0 Averaged Erosion selected One Year Time- Only Method 2 0.510 0.0039 0 Averaged Erosion One Year Time- Only Method 3 0.489 0.0043 0 Averaged Erosion One Year Time- Only Method 4 0.870 0.0002 0 Averaged Erosion One Year Time- Only Method 5 0.705 0.0013 0 Averaged Erosion All Monthly Number of Peaks 0.003 0.0165 0.010 Erosion Above Baseflow All Monthly Percent of Time Above 0.018 0.0103 0.002 Erosion Baseflow Only Bankfull Modified DuBoys 0.042 0.0450 0.008 Erosion (Equation 8) Modified DuBoys with Only Bankfull Radius of Curvature 0.061 0.0403 0 Erosion Adjustment (Equations 9-10) All Monthly Modified DuBoys 0.226 0.0025 0 Erosion (Equation 8)
80
Table 4-2 cont. Constructed curve regression statistics. Erosion Curve Regression Statistics Erosion Input NBS p-value r2 Predicted r2 Modified DuBoys with All Monthly Radius of Curvature 0.0444 0.0073 0.0009 Erosion Adjustment (Equations 9-10) One Year Time- Modified DuBoys 0.328 0.0085 0 Averaged Erosion (Equation 8) Modified DuBoys with One Year Time- Radius of Curvature 0.651 0.0019 0 Averaged Erosion Adjustment (Equations 9-10)
4.2.1 BEHI
The banks measured in this study only fell within the High and Very High BEHI categories. For all the regressions these two BEHI categories were combined into a High/Very
High BEHI category. While combining these two categories has been done in previous studies, for example the curves from Yellowstone National Park, Colorado (Rosgen, 2006), and the
USFWS curve (Rathbun, 2009), there are other justifications for this action. For one, there is no natural increase in erosion rate based on the continuous BEHI values. Figure 4-11 displays the bank-averaged erosion rates versus the continuous, quantitative BEHI values. On the continuous
BEHI scale, High BEHIs are between 30 and 39.5 and Very High BEHIs are between 40 and 45
(Rosgen, 2006). Figure 4-11 indicates no increase in erosion rate with BEHI score. In fact, the mean erosion rate for all High banks is slightly greater than the mean erosion rate for all Very
High bank.
81
Figure 4-11. Bank erosion rates versus quantitative BEHI values for historic Stroubles Creek dataset. BEHI does not appear to have a relationship with erosion rate.
In this study, whether a bank is High or Very High largely depended upon the ratio of study bank height to bankfull height. Banks where the ratio was greater than 1.6 were largely
Very High, while banks where the ratio was less than 1.6 were largely High. Although this measurement makes physical sense with higher ratios indicating the channel is more incised with limited floodplain access and thus higher boundary shear stress during floods, the High and Very
High categories do not appear to be an effective distinction for Stroubles Creek. The commonality of combining these two categories (Rathbun, 2009; Rosgen, 2006) suggests this issue is not unique to Stroubles Creek.
82
4.2.2 Traditional NBS
Since the High and Very High BEHI categories were combined, a total of 15 traditional
BANCS erosion curves were developed, shown in Figures B-1 to B-15. Each curve has a unique combination of NBS method and erosion rate temporal input, which will be discussed below. As stated before, curves were assessed initially through regression statistics, shown in Table 4-2, and whether the regression slope was positive or negative. Of all the curves which were in the
90% confidence level, none had a negative slope.
The standard BANCS curve involves measuring the erosion for each bankfull flow event and then plotting those values versus the highest NBS value for that bank. Figure 4-12A displays the erosion associated with bankfull events from the historic SC, SC, TC, and NFR sites with a regression performed only using the historic SC data. Because the Stroubles Creek watershed is highly urbanized, multiple bankfull discharges occur each year (Annable et al., 2012). The regression analysis did not yield a statistically significant relationship between NBS and measured erosion and the relationship explained only 2% of the variance in the erosion measurements (r2 = 0.0237, p = 0.143, Figure 4-12A). There is significant scatter in the data and an only weak relationship between erosion rate and NBS.
Regression analysis between bankfull erosion and each of the individual NBS methods, instead of just the highest value for each erosion event, was also explored. Of these curves, only the relationship of erosion rate versus NBS Method 3 resulted in a regression with a positive slope that was significantly different from zero, although the relationship explained only 4% of the variance in the dataset (p = 0.0558, r2 = 0.04; Figure 4-12B). Method 3 uses the ratio of pool water surface slope to average water surface slope at baseflow, which has no known association with boundary shear stress at bankfull discharge in hydraulics or fluvial geomorphology. An
83
unfortunate aspect of this NBS method is that measuring water surfaces to calculate water slopes is time-consuming and difficult relative to the other NBS methods, decreasing use amongst practitioners. Another issue is that this method relies on the assumption that the measured stream has stable riffle-pool morphology. There are many stream reaches being restored in the Valley and Ridge Physiographic Province which are either step-pool or plane bed in the Montgomery-
Buffington stream classification system, meaning these streams would have different bedforms from pool-riffle systems. Even if the stream being studied is riffle-pool, highly disturbed streams are likely to have underdeveloped riffles and pools that are hard to identify and subsequently measure.
One of the noted limitations of the BANCS method is that only bankfull events can be used to develop the relationship, limiting the amount of data available for developing the erosion rating curves to a single data point per bank every 1-2 years. Methods of incorporating erosion rates from storms other than bankfull events were explored, including all recorded monthly erosion rates regardless of flood event magnitude during recording and annual time-averaged erosion rates. Regression results showed that there were no statistically significant relationships between erosion rate from these other selected flood events and NBS. This is not surprising as the established NBS methods were specifically calibrated for bankfull flows.
84
(A)
(B)
Figure 4-12. BANCS curves using different established NBS methods where: A) assesses all NBS and selects the highest, B) only method 3 is used. Both regressions have low r2 values, indicating little of the variance in the erosion rate is explained by the equation.
85
As stated in the methodology section, four sets of time-averaged erosion data were evaluated, including: three years of erosion data with the first year including the summer flood; three years with the first year not including the summer flood; two years with the first year including the summer flood; and, two years with the first year not including the summer flood.
Erosion data using three years of time-averaged data had three erosion rates per bank, while data using two years of time-averaged data had two erosion rates per bank. None of the time-averaged erosion curves produced a significant relationship with any of the traditional NBS methods or the highest selected NBS when all methods were applied.
It should be noted that despite the curve 4-12B being statistically significant (p-value <
0.10), the equation explains little of the variance in the dataset, as indicated by the low r2 values of 0.040. This low r2 value is comparable to some of the previous BANCS curves shown in
Table 2-2, but much lower than other curves. The low r2 values for the curves in Alabama and
Florida, Oklahoma, and Western Carpathians, were attributed to the study not being in a snowmelt driven hydrology, the use of only one NBS method, the presence of cohesive banks, and large flooding (Allmanová, Vlčková, Jankovský, Allman, et al., 2019; Harmel et al., 1999;
McMillan et al., 2017). Stroubles Creek has both hydrology not driven by snowmelt and cohesive banks, but large flooding was excluded from the traditional BANCS curves. On top of this, the predicted r2 values are all much lower than the r2 values, indicating that these equations are not suitable for prediction, which is the intended application. It is suggested that current and future BANCS curves be presented with information on the significance and fit of the data by the equation, which is uncommon at this point in time.
Of all the applied NBS methods, method 3 has the strongest relationship with erosion rate at bankfull on Stroubles Creek. Due to method 3 being more inconvenient to quantify,
86
practitioners typically do not utilize it, possibly leading to the less robust BANCS curves. This possibility highlights the impractical nature of having seven NBS methods that users can pick and choose between. As such, it is recommended that users in the Valley and Ridge apply method 3 if erosion is measured per bankfull event. Despite this conclusion, the lack of a strong relationship in general indicates either that NBS does not adequately quantify fluvial shear or that bank retreat along Stroubles Creek is not primarily driven by bankfull fluvial shear.
4.2.3 Modified NBS
Modified BANCS methods are split into hydrograph-based methods and modified
DuBoys methods. Due to BEHI being categorized into a single “High/Very High” category, there were a total of eight modified BANCS regressions.
4.2.3.1 Hydrograph-Based
The hydrograph-based methods yielded stronger relationships to the observed erosion than the NBS methods, as shown in Figures 4-13A and 4-13B. Both curves are within the 95% confidence level and have non-zero predicted r2 values (Figure 4-13A, p-value = 0.0264, r2 = 0.
0165; Figure 4-13B, p-value = 0.0175, r2 = 0.0103), which is an improvement to the curves developed using established NBS methods.
87
(A)
(B)
Figure 4-13. Hydrograph-based curves using A) the number of peaks above baseflow, and B) the percent of time above baseflow. Only the historic datapoints in red are used in the regression. New site datapoints in grey are only plotted for comparison. Both regressions have low r2 values, indicating little of the variance in the erosion rate is explained by the equation.
88
Similar to the previous curves, while Figures 4-13A and 4-13B are statistically significant, the equations explain little of the variance in the dataset. Despite the p-values being lower than those of the traditional BANCS curves, the r2 is lower with the values of 0.0165 and
0.0103 as opposed to of 0.04 and 0.0437. This indicates that the equations of the hydrograph- based curves explain less of the variances than the traditional NBS methods above. The lower p- values can be attributed to the larger sample size rather than a stronger relationship with erosion rate. The predicted r2 values are non-zero for the hydrograph-based curves, but are still both much lower than the r2 values, indicating that these equations are not suitable for prediction.
However, the low r2 values are not surprising. While the two hydrograph-based methods of using number of peaks and percent of time above baseflow can approximate the applied fluvial shear during the monthly intervals, they are hardly a robust indicator of said shear. A more appropriate measurement would be to convert the hydrograph into a time-series of applied shear using DuBoys equation, and then using the shear time-series to estimate the applied fluvial shear during the time interval. Of course, this process would be more time-consuming as it would require a survey of every bank used to create the curve, further data processing of the hydrograph with the survey data to estimate the shear time-series for each bank, and computations to represent that time-series of shear for one time-interval. While it is believed this would decrease variability and increase the quality of the relationship, it will still not be perfect because bank retreat is also driven by other erosion processes such as freeze-thaw cycling, as indicated in Figures 4-7A, 4-7B, and 4-7C.
While the regression statistics for these hydrograph-based models are promising and indicate that the bank erosion predictions should be paired with flow or stage data, the curves are not practical for regulatory use. The great advantage of the traditional BANCS model is that,
89
once a curve is constructed, it only requires field work and limited surveying to estimate a bank retreat rate for a stream reach. On the other hand, the hydrograph-based curves would require the prediction be made only where stream discharge data is available, which greatly limits the application of this technique. Thus, it is not recommended that hydrograph-based methods be adopted despite the improved theoretical basis and statistical relationships.
4.3.3.2 Modified DuBoys
The modified DuBoys curves (Equations 8 and 9-10) substituted NBS use three erosion rate inputs: erosion rates only from bankfull events, all monthly erosion rates, and one year time- averaged erosion rates. Of these three erosion rate inputs, only the one year time-averaged erosion rates did not produce any statistically significant relationships. Figure 4-14 displays three of the remaining four curves that did meet the regression criteria. Using the modified DuBoys alone was only successful when performing the regression with bankfull erosion as shown in
Figure 4-14A. Of the three modified DuBoys curves, the regression statistics in Figure 4-14A are the best, with the lowest p-value and highest r2 and predicted r2. The fact that this substitute for
NBS is only viable when paired with erosion per bankfull events indicates that the method is still linked to bankfull, which this study is attempting to rely less upon. With the modified DuBoys and standardized radius of curvature regression, several curves that used bankfull events and all monthly erosion rates were statistically significant, as shown in Figures 4-14B and 4-14C, respectively. Comparing regression statistics indicates that using bankfull erosion rates are better for this NBS supplement; using all monthly erosion rates explained less than 1% of the data variance.
90
(A)
(B)
(C)
Figure 4-14. Statistically significant modified BANCS curves using A) hS per bankfull event, B) hS and Rc 2 per bankfull event, and C) hS and Rc for all monthly measurements. The r values for all regressions were low, indicating little of the variance in the erosion rate is explained by the equation.
91
As with the other regression relationships, the curves in Figure 4-14 are statistically significant but their equations explain little of the variance in the dataset. Both the per bankfull curves, Figures 4-14A and 4-14B, have r2 values on par with traditional BANCS curves, and the curve that uses all monthly erosion events, Figure 4-14C, has the lowest r2 of all the statistically significant curves so far. This indicates that while all the curves in Figure 4-14 do not explain the variance in the data well, the modified per bankfull event equations are slightly better. As is the case with the other curves presented, all the predicted r2 values are much lower than the r2 values, indicating that these equations are not suitable for prediction
The above results have two implications. Firstly, the modified DuBoys method may be more appropriate for erosion curves in the Valley and Ridge when plotting per bankfull event because modified DuBoys has better regression statistics than any NBS method. While this method maybe more time intensive due to the need to survey and post-process the data, it will eliminate the need to perform multiple NBS methods to find the highest value. Secondly, if one wishes to represent more than bankfull erosion, using all monthly erosion rates with the modified
DuBoys and standardized radius of curvature is the best method. Again, this method to replace
NBS will be more time intensive due to surveying and post-processing, but curve creators would not have to wait for a bankfull event to add datapoints to a curve. So, the curve creation process itself may accelerate.
4.2.4 Assessing Erosion Curves
Table 4-3 compares measured one year time-averaged erosion rates at the current study sites to the predicted erosion rates from several of the constructed curves and currently used
BANCS curves. This section compares results between sites, input erosion rates, NBS methods,
92
and constructed and existing curves using the absolute percent error. Overall, all the curves developed in this study did a good job in estimating future erosion, with all predicted rates either in the same or within one order of magnitude of measured values, and no single prediction having greater than 80% error from the measured erosion rate. This is not the case for the existing BANCS curves which consistently overpredicted by greater than 100%.
4.2.4.1 Sites
The difference in measured and predicted erosion rates for TC and SC were generally less than ±5 cm/yr (2 in./yr) and ±10 cm/yr (4 in./yr), respectively. Interestingly, the percent error between measured and predicted erosion rates were generally higher in SC, the stream the curve was created with, than TC. Further, the values predicted for SC were always lower than the measured erosion, while there was a mix of positive and negative values for TC. Why the constructed curves underpredicted the erosion rate of SC is less clear, especially since the curve itself was constructed using Stroubles Creek pre-restoration when retreat rates should be higher.
The percent error between measured and predicted values were highest for the NFR. Like SC, all the curves underpredicted the erosion rate. The general trends for the sites are shown in Figure 4-
15 with TC have the lowest percent error and NFR having the highest percent error.
93
Table 4-3. Measured annual erosion rates compared to predicted erosion rates. Erosion Curve TC SC NFR Erosion Erosion Rate Percent Error Erosion Rate Percent Error Erosion Rate Percent Error NBS Rates (cm/yr) (%) (cm/yr) (%) (cm/yr) (%)
Measured Time -Averaged 10.9 - 13.0 - 27.4 - Only All NBS, highest Bankfull 7.7 -29.2 6.0 -53.8 9.4 -65.5 selected Erosion Only Bankfull Only Method 3 8.7 -19.7 6.5 -49.7 9.7 -64.8 Erosion Only Bankfull Modified DuBoys 19.6 80.0 8.1 -37.6 8.1 -70.6 Erosion Modified DuBoys Only with Radius of Bankfull 11.7 7.9 5.8 -55.6 10.3 -62.4 Curvature Erosion Adjustment Modified DuBoys All Monthly with Radius of 12.7 16.5 8.4 -35.3 11.8 -57.1 Erosion Curvature Adjustment All Monthly Number of Peaks 9.0 -17.3 8.2 -37.1 17.9 -34.6 Erosion Above Baseflow All Monthly Percent of Time 8.6 -21.2 9.3 -28.1 9.2 -66.3 Erosion Above Baseflow USFWS BANCS Curve 61.0 461.0 30.5 134.5 76.2 178.2 (Berg et al., 2013) NC BANCS Curve 6.1 -43.9 30.5 134.5 289.6 957.0 (McQueen, 2011)
94
Figure 4-15. Constructed curve predicted erosion rate percent error for each site where the boxes represent the mean and the errors bar showing the 95% confidence interval. The model had the high error when predicting NFR erosion rates and the lowest error when predicting TC erosion rates.
The constructed curves underpredicting the erosion rate for NFR was expected, especially since it is hypothesized that the erosion at this site is dominated by a piping erosion worsened by the erosion pins. Piping erosion is meant to be accounted for by the stratification adjustment in
BEHI. There are several issues with this set-up. For one, stratification may not be visible, especially to those unexperienced in soils, and piping can occur even when stratification is not present, though it is less common (Fox & Wilson, 2010). Another issue is whether this stratification adjustment is sufficient. If the adjustment was noticed and accounted for in BEHI, it can increase the bank BEHI by one to two categories, which can either not increase the predicted erosion rate at all if categories are combined, or increase the predicted erosion up to one order of magnitude depending on the curve used, the NBS, and the initial BEHI category. BANCS is
95
fundamentally flawed when piping erosion is a significant process because stratification is accounted for in BEHI and the BANCS framework uses NBS, an indicator of boundary shear stress, to predict piping erosion when the driving force behind piping is site-specific groundwater dynamics (Fox & Wilson, 2010). Similarly, the BANCS framework uses a bank material adjustment in BEHI to account for freeze-thaw cycling, which means the driving force for the freeze-thaw process in BANCS is again applied boundary shear stress through NBS rather than climactic conditions, the true driver of any subaerial process (Wynn et al., 2008). The use of
BEHI adjustments to represent freeze-thaw cycling and piping erosion highlight a weakness with both the traditional and modified BANCS; the framework assumes fluvial entrainment is the main driving force of bank retreat while it is known that bank retreat is much more complex and incorporates numerous processes that are independent of stream flow.
4.2.4.2 Input Erosion Rates
To find whether the predicted erosion rates were significantly different when using different input erosion rates, one-way ANOVA with site blocking and mean separation analyses were performed on the percent error of the predicted erosion rates. Only two input erosion rates were used, erosion rate at bankfull flow and all monthly erosion rates, because no one year time- averaged erosion rate curves had a significant relationship. Additionally, this analysis only used curves created in this study, so the NC and USFWS BANCS curves were excluded. Figure 4-16 visualizes the results of the analysis with statistical pairs grouped using a 95% confidence level.
The two erosion rate input options are statistically different with the bankfull erosion rate option having a larger percent error. This result indicates that using all monthly erosion rates is a better predictor for annual erosion rate than using bankfull erosion rates on Stroubles Creek.
96
Figure 4-16. Input erosion rates percent error one-way ANOVA with site blocking and mean separation results where boxes represent the model least squre mean and error bars show the 95% confidence interval. Using bankfull erosion rates as an input had a higher erosion rate prediction error than using all monthly erosion rates as an input.
4.2.4.3 NBS Methods
To deduce whether the NBS category or NBS method used significantly affected the predicted erosion results, the percent errors of the predicted erosion rates were compared using a one-way
ANOVA with site blocking and mean separation analyses. Figure 4-17 displays the results of these analyses with grouping by NBS categories and NBS methods, Figures 4-17A and 4-17B, respectively. While the results in Figure 4-17A indicate that the modified hydrograph-based method has the lowest erosion rate prediction percent error and the modified DuBoys-based method has the highest prediction percent error, the difference is not statistically significant using a 95% confidence level. Figure 4-17B indicates the percent error of the NBS methods were similar using a 95% confidence level. Figure 4-17B should be taken with a grain of salt because
97
several of the NBS methods only have one contributing curve (i.e. only have three data points), and is not blocked by input erosion rates.
(A)
(B)
Figure 4-17. One-way ANOVA with site blocking and mean separation results where boxes represent the model least squre mean and error bars show the 95% confidence interval for: A) NBS categories percent error, and B) NBS methods percent error. There is no statistical difference in erosion rate prediction error between NBS categories or NBS methods. 98
4.2.4.4 Existing and Constructed Curves
One-way ANOVA and mean separation analyses were conducted to compare the percent error of erosion rate predictions between the existing and constructed erosion curves. Figure 4-18 is the result of these analyses. The percent error in erosion rate predictions based on existing and constructed curves are significantly different using a 95% confidence level with least square mean erosion rate percent errors of 318% and 41%, respectively. This finding indicates that erosion predictions in the Valley and Ridge which use the existing curves for crediting, overpredict erosion by around 300% for High and Very High BEHI banks, resulting in an over- allocation of sediment-reduction credit. While this overprediction could result in reduced water quality improvements, since the USFWS BANCS curve is currently being applied, the overprediction is not surprising because the established curves were created in the Piedmont and
Coastal Plain physiographic provinces. The results emphasize that erosion rate curves are region- specific, highlighting how imperative it is to create more erosion curves in different physiographic provinces.
99
Figure 4-18. Existing and constructed curve predicted erosion rate percent error one-way ANOVA with site blocking and mean separation results where boxes represent the model least squre mean and error bars show the 95% confidence interval. Existing BANCS curve have a much greater erosion rate percent errors than the curves constructed in this study because they were created for use in different physiographic provinces.
Apart from the percent error comparison, a notable difference between the constructed and existing curves is the amount of data available and used to create the models. This discrepancy is apparent when comparing Figure 3-4 to Figure 4-12. The constructed curves that only use bankfull erosion rates have 129 data points per regression equation while the established curves use between 4 and 17 data points per regression equation. The difference in number of datapoints is a result of how the datasets were collected. Most erosion datasets used to create
BANCS curves are collected for that express purpose so there are only a few banks monitored per BEHI and NBS category and erosion from only bankfull events is measured. In contrast, the historic SC dataset used in this study originally focused on highly erosive banks (High and Very
High BEHI categories) with erosion measured each month. Rather than having more variety of
100
BEHI and thus more curves with fewer datapoints, as is typical for a BANCS curve, the constructed curves have a large number of datapoints for the same BEHI and at the full range of flows. Essentially, in this study, even when only the standard per bankfull erosion rates are used,
BEHI and NBS combinations are represented by more banks than in established BANCS curves.
This larger data set reveals the greater variability in the erosion rates per BEHI category, resulting in lower r2 regression values. Subsequently this suggests that established BANCS curves have good regressions because of the limited bank selection and resulting lack of multiple banks representing a single BEHI and NBS combination.
4.3 Comparison with the Universal Soil Loss Equation
The Universal Soil Loss Equation (USLE) is an empirical model originally conceived by the U.S. Department of Agriculture (USDA) for predicting non-point source sediment erosion per unit area through terrain, climate, soil, land use, and land management variables. The equation was originally created in and for agricultural areas, but adaptations have been made since its inception to either improve upon the original or change the applicable land use. These include the Modified USLE (MUSLE) and Revised USLE (RUSLE). The basic form is shown in
Equation 12, with A being the soil loss per unit area, R being the rainfall and runoff factor, K being the soil erodibility factor, L being the slope-length factor, S being the slope-steepness factor, C being the cover and management factor, and P being the support practice factor
(Benavidez et al., 2018).
퐴 = 푅퐾퐿푆퐶푃 (12)
While the ULSE and BANCS do not predict the same type of erosion, they are similar in that they are both empirically derived erosion models. Also both models widely used for natural
101
resources planning and management, likely due to ease of use, flexibility, and regulatory need rather than accuracy or precision of predictions (Alewell et al., 2019; Castro-Bolinaga & Fox,
2018; McMillan et al., 2017). That is to say, without calibration and validation, both models can vastly overpredict or underpredict observed erosion values. For example, one study showed that the USLE model can underpredict or overpredict between 0.04 and 3 times observed values
(Benavidez et al., 2018). The two models even have the same limitations such as not accounting for processes after erosion such as transport or sedimentation, being spatially variable, neglecting certain erosion mechanisms, not accounting for seasonal variability, and the previously discussed errors in predictions (Alewell et al., 2019; Benavidez et al., 2018).
Nonetheless, the manner in which both models were developed and applied are different.
When the 1978 USLE equation was presented, its development was based on over 10,000 data points (USDA, 2016). In contrast, the original Colorado BANCS curves used between 3 and 22 datapoints per curve (Rosgen, 2006). The USLE model was created over several decades of empirical testing and improvements, which started in the 1940s and continue to the present. As time went on, the model added new variables to account for deficiencies and application in new locations; USLE-type models went from only considering vegetation, slope length, and slope in the 1940s, to the present when all the variables discussed in Equation 12 are considered (Laflen
& Flanagan, 2013). On the other hand, no published data is available regarding the development of BANCS and the model has remained relatively static in form since its debut in the 2006, regardless of criticism and suggestions (Allmanová, Vlčková, Jankovský, Allman, et al., 2019;
Bigham et al., 2018; Ghosh et al., 2016; McMillan et al., 2017; Newton & Drenten, 2015).
Despite the long-term and robust testing that went into the USLE, its limitations and applications have been approached more realistically. It should be acknowledged that USLE
102
reports do not offer regression statistics for comparison and the original USDA 1965 release of the equation only acknowledged the lack of erosion mechanisms (Wischmeier & Smith, 1965).
Nonetheless, the second publication discussed the equation’s limitations and emphasized that the
USLE is dependent on similarity to empirical testing conditions and that predictions are truly estimates rather than absolute values (Wischmeier & Smith, 1978). As such, the USLE should be treated as a planning and management tool to minimize long-term erosion using agricultural
BMPs, rather than a predictive tool unless the model is calibrated and validated (Benavidez et al.,
2018). Similar to the USLE, the original BANCS text acknowledges a critical limitation, namely that BANCS curves are region-specific, and subsequent texts explain other limitations not discussed in the original text (Rosgen, 2006; Rosgen et al., 2019). Despite these published limitations, BANCS curves are still applied to different regions for crediting as a direct predictor of bank retreat. While the USLE equation is used beyond the original assumptions and conditions, it is arguable that this misuse is justified because measuring overland erosion from large areas of land is very difficult (Benavidez et al., 2018). This justification does not exist for bank erosion because, while it is less convenient to measure bank erosion than to apply BANCS, measurement it is very achievable. Due to the apparent lack of rigorous data used to construct
BANCS, as well the plausibility of directly measuring or estimating bank retreat, it is recommended that BANCS be replaced by direct bank retreat measurements. Instead, BANCS should be treated as a management tool for early erosion estimates.
Knowledge of the USLE and the results in Figure 4-18 puts into perspective how well the constructed curves in this study predicted erosion rates despite lackluster r2, predicted r2, and site-specific differences at the NFR bank. As previously stated, all erosion predictions were the same order of magnitude or within one order of magnitude different from the measured bank
103
erosion values, and no prediction had greater than 80% error. This would be considered a success by the standards of other sediment models.
104
5 Conclusions
The goal of this study was to assess the sensitivity and error of the BANCS model predictions of bank retreat to input retreat rate spatial and temporal variability. Spatially on the bank-scale, using an array of pins with the PAEA method produces erosion estimates that are less sensitive to measurements of individual pins. The BA array method yields high predictions because the whole area of the bank is skewed to extreme pin measurements. Computing bank retreat rates using measurements from a single column or row overestimates total bank loss, as compared to other spatial averaging techniques. However, row measurements along a bank provides more consistent measurements over long periods of times than the current practice of having a single column of measurements per eroding bank face. On the reach-scale, erosion measurements should be made every three channel widths to best capture the mean reach erosion rate and to minimize variance in the estimated mean. For most second and third order streams in the mid-Atlantic U.S., this recommended spacing is less than the current the Chesapeake Bay
Protocol minimum recommendation of 60 m (200 ft.).
Measured erosion rates in the Valley and Ridge can be highly affected by seasonality.
The freeze-thaw cycling of streambanks with high silt content is likely the primary mechanism of erosion during winter, leading to higher erosion rates. A minimum 12 month sampling time is recommended to estimate bank retreat rates. The fact that bank retreat in this area is strongly influenced by freeze-thaw cycling and processes such as soil piping further highlights the need to measure long-term erosion rates and to incorporate more than fluvial erosion into future bank retreat modelling.
A majority of regressions between erosion rate and NBS were not statistically significant regardless of erosion rate input. The number of significant relationships increased slightly when
105
using the modified NBS instead of the traditional NBS, indicating the modified NBS methods may be better indicators of fluvial shear than traditional NBS methods. Retreat rate predictions for constructed curves and two currently used BANCS curves were compared as percent errors from observed erosions. Predictions from bankfull erosion curves were statistically higher than predictions using all monthly erosion rates. Predictions between constructed and existing curves were significantly different with existing curves generally overpredicting by 300%, suggesting that bank retreat credits are overallocated in the Valley and Ridge. This overprediction is unsurprising because the existing curves were created in other physiographic provinces.
While several erosion curves for the Valley and Ridge were created with predictions that were relatively accurate and significant relationships between erosion rate and NBS, this study puts into question the efficacy of BANCS, particularly as a crediting tool. All of the curves constructed in this study had low r2 values, showing that NBS does little to explain the variance in the erosion data. This is in contrast to currently used BANCS curves which usually have much higher r2 values but use much less data that does not reflect the temporal and spatial variability in bank retreat along a reach. The relative lack of data used to construct many of the applied
BANCS curves is further apparent when juxtaposed to the USLE, another empirical erosion model that was created over decades with thousands of datapoints.
As a whole, it is suggested that BANCS be used only as a planning tool for processes such as site selection in physiographic provinces where curves already exist and that sediment load reductions be assessed using pre- and post-restoration measurements of bank retreat. Future work to improve BANCS could include developing curves with more data that captures long- term, time-averaged bank erosion. Research also needs to establish an NBS that better represents the processes driving bank retreat in reaches that are not dependent on bankfull fluvial erosion.
106
References
Alewell, C., Borrelli, P., Meusburger, K., & Panagos, P. (2019). Using the USLE: Chances,
challenges and limitations of soil erosion modelling. International Soil and Water
Conservation Research, 7(3), 203–225. https://doi.org/10.1016/j.iswcr.2019.05.004
Allmanová, Z., Vlčková, M., Jankovský, M., Allman, M., & Jakubis, M. (2019). Predicting the
annual erosion rates on a small stream by the BANCS model. Soil and Water Research,
14(No. 4), 200–211. https://doi.org/10.17221/58/2018-swr
Allmanová, Z., Vlčková, M., Jankovský, M., Jakubis, M., & Allman, M. (2019). Bank erosion of
the Tŕstie stream: BANCS model predictions vs. real bank erosion. Journal of Hydrology
and Hydromechanics, 67(2), 121–128. https://doi.org/10.2478/johh-2018-0040
Altland, D., Berg, J., Brown, B., Burch, J., Cook, R., Fraley-Mcneal, L., Meyers, M., Running,
J., Starr, R., Sweeney, J., Thompson, T., White, J., & Blair, A. (2020). Consensus
Recommendations for Improving the Application of the Prevented Sediment Protocol for
Stream Restoration Projects Built for Pollutant Removal Credit. Chesapeake Stormwater
Network. https://chesapeakestormwater.net/download/9928/
Altland, D., Brown, T., Coleman, J., Ehrhart, B., Hartranft, J., Oberholzer, W., Parola, A., Stack,
B., Sweeney, J., & White, J. (2020). Consensus Recommendations for Crediting Floodplain
Restoration Projects Involving Legacy Sediments. Chesapeake Stormwater Network.
Benavidez, R., Jackson, B., Maxwell, D., & Norton, K. (2018). A review of the (Revised)
Universal Soil Loss Equation ((R)USLE): with a view to increasing its global applicability
and improving soil loss estimates. Hydrology and Earth System Science, 22(1995), 6059–
6086.
Berg, J., Burch, J., Cappuccitti, D., Filoso, S., Fraley-McNeal, L., Goerman, D., Hardman, N.,
107
Kaushal, S., Medina, D., Meyers, M., Kerr, B., Stewart, S., Sullivan, B., Walter, R. C., &
Winters, J. (2013). Appendix B Protocol 1 Supplemental Details.
https://www.chesapeakebay.net/channel_files/19138/
Bigham, K. A., Moore, T. L., Vogel, J. R., & Keane, T. D. (2018). Repeatability, Sensitivity, and
Uncertainty Analyses of the BANCS Model Developed to Predict Annual Streambank
Erosion Rates. Journal of the American Water Resources Association, 54(2), 423–439.
https://doi.org/10.1111/1752-1688.12615
Bledsoe, B., Lammers, R. W., Jones, J., Clary, J., Earles, A., Strecker, E., Struck, S., Leisenring,
M., & McGuire, A. (2016). Stream Restoration as a BMP: Crediting Guidance (WERF-
1T13). Water Environment Research Foundation.
Boardman, J., & Favis-Mortlock, D. (2016). The use of erosion pins in geomorphology. In A.
Goudie (Ed.), Geomorphological Techniques (4th ed.). British Society for Geomorphology.
https://doi.org/https://doi.org/10.4324/9780203430590
Boardman, J., Favis-Mortlock, D., & Foster, I. (2015). A 13-year Record of Erosion on Badland
Sites in the Karoo, South Africa. Earth Surface Processes and Landforms, 40(14), 1964–
1981. https://doi.org/10.1002/esp.3775
Brasington, J., Rumsby, B. T., & Mcvey, R. A. (2000). Monitoring and modelling morphological
change in a braided gravel-bed river using high resolution GPS-based survey. Earth Surface
Processes and Landforms, 25(9), 973–990. https://doi.org/10.1002/1096-
9837(200008)25:9<973::AID-ESP111>3.0.CO;2-Y
Castro-Bolinaga, C. F., & Fox, G. A. (2018). Streambank Erosion: Advances in monitoring,
modeling and management. Water, 10(10). https://doi.org/10.3390/w10101346
Chowns, T. (2018). Valley and Ridge Geologic Province. In New Georgia Encyclopedia. New
108
Georgia Encyclopedia. https://www.georgiaencyclopedia.org/articles/science-
medicine/valley-and-ridge-geologic-province
Conley, D. J., Paerl, H. W., Howarth, R. W., Boesch, D. F., Seitzinger, S. P., Havens, K. E.,
Lancelot, C., & Likens, G. E. (2009). Controlling Eutrophication: Nitrogen and Phosphorus.
Science, 323(5917), 1014–1015. https://doi.org/10.1126/science.1167755
Copeland, R. R., Biedenharn, D. S., & Fischenich, J. C. (2000). Channel-Forming Discharge
(ERDC/CHL CHETN-VIII-5). U.S. Army Corp of Engineers.
Coryat, M. (2011). Analysis of the Bank Assessment for Non-point Source Consequences of
Sediment (BANCS) Approach for the Prediction of Streambank Stability and Erosion along
Stony Clove Creek in the Catskills. In Syracuse University SURFACE. Syracuse University.
Couper, P. R., & Maddock, I. P. (2001). Subaerial river bank erosion processes and their
interaction with other bank erosion mechanisms on the River Arrow, Warwickshire, UK.
Earth Surface Processes and Landforms, 26(6), 631–646. https://doi.org/10.1002/esp.212
Culp, J. M., Wrona, F. J., & Davies, R. W. (1986). Response of stream benthos and drift to fine
sediment deposition versus transport. Canadian Journal of Zoology, 64(6), 1345–1351.
https://doi.org/10.1139/z86-200
Dearmont, D., McCarl, B. A., & Tolman, D. A. (1998). Costs of Water Treatment due to
Diminished Water Quality: A Case Study in Texas. Water Resources Research, 34(4), 849–
853.
Delaware’s Chesapeake Bay WIP Steering Committee. (2019). Delaware’s Phase III
Chesapeake Bay Watershed Implementation Plan Assembled. Delaware Natural Resources
and Environmental Control. https://doi.org/10.32964/tj18.8
Delaware’s Chesapeake Interagency Workgroup. (2010). Delaware’s Phase I Chesapeake Bay
109
Watershed Implementation Plan. Delaware Natural Resources and Environmental Control.
https://dnrec.alpha.delaware.gov/watershed-stewardship/nps/chesapeake/phase-i/
Dick, B. M., Hey, R., Peralta, P., Jewell, I., Simon, P., & Peszlen, I. (2014). Estimating Annual
Riverbank Erosion Rates – A Dendrogeomorphic Method. River Research and
Applications, 30, 845–856. https://doi.org/10.1002/rra
District of Columbia Department of Energy & Environment. (2019). District of Columbia’s
Phase III Watershed Implementation Plan for the Chesapeake Bay. District of Columbia
Department of Energy & Environment.
https://doee.dc.gov/sites/default/files/dc/sites/ddoe/service_content/attachments/FINAL DC
PHASE III WIP - HQ2 web.pdf
District of Columbia Department of the Environment. (2010). Chesapeake Bay Watershed
Implementation Plans. District of Columbia Department of the Environment.
https://doee.dc.gov/sites/default/files/dc/sites/ddoe/publication/attachments/Final_District_o
f_Coluimbia_WIP_Bay_TMDL.pdf
Donovan, M., Miller, A., Baker, M., & Gellis, A. (2015). Sediment contributions from
floodplains and legacy sediments to Piedmont streams of Baltimore County, Maryland.
Geomorphology, 235, 88–105. https://doi.org/10.1016/j.geomorph.2015.01.025
Fox, G. A., Purvis, R. A., & Penn, C. J. (2016). Streambanks: A net source of sediment and
phosphorus to streams and rivers. Journal of Environmental Management, 181, 602–614.
https://doi.org/10.1016/j.jenvman.2016.06.071
Fox, G. A., & Wilson, G. V. (2010). The Role of Subsurface Flow in Hillslope and Stream Bank
Erosion: A Review. Soil Science Society of America Journal, 74(3), 717–733.
https://doi.org/10.2136/sssaj2009.0319
110
Ghosh, K. G., Pal, S., & Mukhopadhyay, S. (2016). Validation of BANCS model for assessing
stream bank erosion hazard potential (SBEHP) in Bakreshwar River of Rarh region, Eastern
India. Modeling Earth Systems and Environment, 2(2), 1–15.
https://doi.org/10.1007/s40808-016-0172-0
Gibson, S., Simon, A., Langendoen, E., Bankhead, N., & Shelley, J. (2015). A Physically-Based
Channel-Modeling Framework Integrating HEC-RAS Sediment Transport Capabilities and
the USDA-ARS Bank-Stability and Toe-Erosion Model (BSTEM). In The Advisory
Committee on Water Information (Ed.), 5th Federal Interagency Hydrologic Modeling
Conference and the 10th Federal Interagency Sedimentation Conference. The Advisory
Committee on Water Information.
Harmel, R. D., Haan, C. T., & Dutnell, R. C. (1999). Evaluation of Rosgen’s Streambank
Erosion Potential Assessment in North- east Oklahoma. Journal Of The American Water
Resources Association, 35(1), 113–121.
Hickin, E. J., & Nanson, G. C. (1984). Lateral Migration Rates of River Bends. Journal of
Hydraulic Engineering, 110(11), 1557–1567. https://doi.org/10.1061/(asce)0733-
9429(1984)110:11(1557)
Izagirre, O., Serra, A., Guasch, H., & Elosegi, A. (2009). Effects of sediment deposition on
periphytic biomass, photosynthetic activity and algal community structure. Science of the
Total Environment, 407(21), 5694–5700. https://doi.org/10.1016/j.scitotenv.2009.06.049
James, M. R., Chandler, J. H., Eltner, A., Fraser, C., Miller, P. E., Mills, J. P., Noble, T., Robson,
S., & Lane, S. N. (2019). Guidelines on the use of structure-from-motion photogrammetry
in geomorphic research. Earth Surface Processes and Landforms, 44(10), 2081–2084.
https://doi.org/10.1002/esp.4637
111
Jennings, G. D., & Harman, W. A. (2001). Measurement and Stabilization of Streambank
Erosion in North Carolina. In J. C. Ascough & D. C. Flanagan (Eds.), Soil Ero- sion
Research for the 21st Century, Proceedings of the International Symposium, January 3-5,
Honolulu, HI (pp. 537–540). ASAE. https://doi.org/10.13031/2013.4596
Jones, J. I., Collins, A. L., Naden, P. S., & Sear, D. A. (2012). The Relationship between Fine
Sediment and Macrophytes in Rivers. River Research and Applications, 28, 1006–1018.
https://doi.org/10.1002/rra
Kassa, K. H. (2019). A Multi-Scale Modeling Approach to Examine Bank Erosion and Sediment
Transport in Stream Restoration: A Case Study in the North Carolina Piedmont Region.
North Carolina State University.
Keaton, J., Messinger, T., & Doheny, E. (2005). Development and Analysis of Regional Curves
for Streams in the Non-Urban Valley and Ridge Physiographic Province, Maryland,
Virginia, and West Virginia (Scientific Investigations Report 2005-5076). In United States
Geological Survey. United States Geological Survey.
Kemp, D., Sadler, P., & Vanacker, V. (2020). The human impact on North American erosion,
sediment transfer, and storage in a geologic context. Nature Communications, 11(1), 1–9.
https://doi.org/10.1038/s41467-020-19744-3
Kemp, P., Sear, D., Collins, A., Naden, P., & Jones, I. (2011). The Impacts of Fine Sediment on
Riverine Fish. Hydrological Processes, 25(11), 1800–1821.
https://doi.org/10.1002/hyp.7940
Kenney, M. A., Wilcock, P. R., Hobbs, B. F., Flores, N. E., & Martínez, D. C. (2012). Is Urban
Stream Restoration Worth It? Journal of the American Water Resources Association, 48(3),
603–615. https://doi.org/10.1111/j.1752-1688.2011.00635.x
112
Klavon, K., Fox, G., Guertault, L., Langendoen, E., Enlow, H., Miller, R., & Khanal, A. (2017).
Evaluating a process-based model for use in streambank stabilization: insights on the Bank
Stability and Toe Erosion Model (BSTEM). Earth Surface Processes and Landforms, 42(1),
191–213. https://doi.org/10.1002/esp.4073
Knapen, A., Poesen, J., Govers, G., Gyssels, G., & Nachtergaele, J. (2007). Resistance of soils to
concentrated flow erosion: A review. Earth-Science Reviews, 80(1–2), 75–109.
https://doi.org/10.1016/j.earscirev.2006.08.001
Kwan, H., & Swanson, S. (2014). Prediction of Annual Streambank Erosion for Sequoia
National Forest, California. Journal of the American Water Resources Association, 50(6),
1439–1447. https://doi.org/10.1111/jawr.12200
Laflen, J. M., & Flanagan, D. C. (2013). The Development of U.S. Soil Erosion Prediction and
Modeling. International Soil and Water Conservation Research, 1(2), 1–11.
https://doi.org/10.1016/S2095-6339(15)30034-4
Lammers, R. W., & Bledsoe, B. P. (2017). What Role Does Stream Restoration Play in Nutrient
Management? Critical Reviews in Environmental Science and Technology, 47(6), 335–371.
https://doi.org/10.1080/10643389.2017.1318618
Langendoen, E. J. (2011). Application of the CONCEPTS Channel Evolution Model in Stream
Restoration Strategies. In A. Simmon, S. J. Bennett, & J. M. Castro (Eds.), Stream
Restoration in Dynamic Fluvial Systems: Scientific Approaches, Analyses, and Tools (Vol.
194, pp. 487–502). American Geophysical Union. https://doi.org/10.1029/2010GM000986
Lawler, D. M. (1993a). Needle ice processes and sediment mobilization on river banks: the River
Ilston, West Glamorgan, UK. Journal of Hydrology, 150(1), 81–114.
https://doi.org/10.1016/0022-1694(93)90157-5
113
Lawler, D. M. (1993b). The measurement of river bank erosion and lateral channel change: A
review. Earth Surface Processes and Landforms, 18(9), 777–821.
https://doi.org/10.1002/esp.3290180905
Lawler, D. M. (1995). The impact of scale on the processes of channel-side sediment supply: a
conceptual model. Effects of Scale on Interpretation and Management of Sediment and
Water Quality. Proc. Symposium, Boulder, 1995, 226(226), 175–184.
Lawton, L. A., & Codd, G. A. (1991). Cyanobacterial (Blue‐Green Algal) Toxins and their
Significance in UK and European Waters. Water and Environment Journal, 5(4), 460–465.
https://doi.org/10.1111/j.1747-6593.1991.tb00643.x
Markowitz, G., & Newton, S. (2011). Project Report: Using Bank Assessment for Non-point
source Consequences of Sediment (BANCS) Model to Prioritize Potential Stream Bank
Erosion on Birch Creek, Shandaken, New York. The Ashokan Watershed Stream
Management Program. http://ashokanstreams.org/wp-content/uploads/2012/10/Birch-Creek-
BANCS-Project.pdf
Maryland Department of the Environment. (2019). Stream Restoration Crediting Clarification
for MS4 Permitting Purposes. Maryland Department of the Environment.
Maryland Department of the Environment, Maryland Department of Natural Resources,
Maryland Department of Agriculture, & Maryland Department of Planning. (2010).
Maryland’s Phase I Watershed Implementation Plan for the Chesapeake Bay Total
Maximum Daily Load. Maryland Department of the Environment.
https://mde.state.md.us/programs/Water/TMDL/Documents/www.mde.state.md.us/assets/d
ocument/MD_Phase_I_Plan_12_03_2010_Submitted_Final.pdf
Maryland Department of the Environment, Maryland Department of Natural Resources,
114
Maryland Department of Agriculture, & Maryland Department of Planning. (2012).
Maryland’s Phase II Watershed Implementation Plan for the Chesapeake Bay TMDL.
Maryland Department of the Environment.
https://mde.state.md.us/programs/Water/TMDL/TMDLImplementation/Documents/FINAL
_PhaseII_Report_Docs/Final_Documents_PhaseII/Final_Phase_II_WIP_MAIN_REPORT_
102612.pdf
Maryland Department of the Environment, Maryland Department of Natural Resources,
Maryland Department of Agriculture, & Maryland Department of Planning. (2019).
Maryland’s Phase III Watershed Implementation Plan to Restore Chesapeake Bay by 2025.
Maryland Department of the Environment.
https://mde.maryland.gov/programs/Water/TMDL/TMDLImplementation/Documents/Phas
e III WIP Report/Final Phase III WIP Package/Phase III WIP Document/Phase III WIP-
Final_Maryland_8.23.2019.pdf
McMillan, M., Liebens, J., & Metcalf, C. (2017). Evaluating the BANCS Streambank Erosion
Framework on the Northern Gulf of Mexico Coastal Plain. Journal of the American Water
Resources Association. https://doi.org/10.1111/1752-1688.12572
McQueen, A. (2011). Appendix D of Estimating Sediment Loads using the Bank Assessment of
Non-point source Consequences of Sediment (BANCS) Watershed.
https://mde.state.md.us/programs/Water/319NonPointSource/Documents/Watershed
Plans/Antietam Creek Plan/Appendix D_BEHI.pdf
Meade, R. H. (1982). Sources, Sinks, and Storage of River Sediment in the Atlantic Drainage of
the United States. The Journal of Geology, 90(3), 235–252.
http://www.boneandjoint.org.uk/highwire/filestream/13335/field_highwire_article_pdf/0/74
115
1.full-text.pdf
Michelsen, E. (2017). Stream & Wetland Restoration Through the Lens of MS4 Compliance and
the Chesapeake Bay TMDL : Can We Get There Without It? Anne Arundel County
Watershed Protection & Restoration Program.
Moody, J. A., Dungan Smith, J., & Ragan, B. W. (2005). Critical shear stress for erosion of
cohesive soils subjected to temperatures typical of wildfires. Journal of Geophysical
Research: Earth Surface, 110(1), 1–13. https://doi.org/10.1029/2004JF000141
Moriasi, D. N., Arnold, J. G., VanLiew, M. W., Bingner, R. L., Harmel, R. D., & Veith, T. L.
(2007). Model Evaluation Guidelines for Systematic Quantification of Accuracy in
Watershed Simulations. American Society of Agricultural and Biological Engineers, 50(3),
885−900. https://doi.org/doi: 10.13031/2013.23153
Nasermoaddeli, M. H., & Pasche, E. (2008). Application of terrestrial 3D laser scanner in
quantification of the riverbank erosion and deposition. River Flow, January 2008, 2407–
2416. http://www.tuhh.de/t3resources/wb/Publikationen/MA-
Veroeffentlichungen/nasermoaddelli/riverflow2008.pdf
New York State Department of Environmental Conservation. (2010). Final Phase I Nutrient and
Sediment Water Quality Improvement and Protection Plan. New York State Department of
Environmental Conservation. https://www.dec.ny.gov/docs/water_pdf/finalphaseiwip.pdf
New York State Department of Environmental Conservation. (2020). Amended Final Phase III
Watershed Implementation Plan: New York Chemung and Susquehanna River Basins. New
York State Department of Environmental Conservation.
https://www.dec.ny.gov/docs/water_pdf/amendedwipiii.pdf
Newton, S. E., & Drenten, D. M. (2015). Modifying the bank erosion hazard index (BEHI)
116
protocol for rapid assessment of streambank erosion in northeastern Ohio. Journal of
Visualized Experiments, 96. https://doi.org/10.3791/52330
Noe, G. B., Cashman, M. J., Skalak, K., Gellis, A., Hopkins, K. G., Moyer, D., Webber, J.,
Benthem, A., Maloney, K., Brakebill, J., Sekellick, A., Langland, M., Zhang, Q., Shenk, G.,
Keisman, J., & Hupp, C. (2020). Sediment dynamics and implications for management:
State of the science from long-term research in the Chesapeake Bay watershed, USA. Wiley
Interdisciplinary Reviews: Water, 7(4), 1–28. https://doi.org/10.1002/wat2.1454
Osman, A. M., & Thorne, C. R. (1988). Riverbank Stability Analysis I: Theory. Journal of
Hydraulic Engineering, 114(2), 134–150.
Pennsylvania Department of Environmental Protection. (2011). Pennsylvania Chesapeake
Watershed Implementation Plan. Pennsylvania Department of Environmental Protection.
https://www.dep.state.pa.us/river/iwo/chesbay/docs/refmaterials/PAChesapeakeWIP_1-11-
11.pdf
Pennsylvania Department of Environmental Protection. (2019). Pennsylvania Phase 3
Chesapeake Bay Watershed Implementation Plan. Pennsylvania Department of
Environmental Protection.
https://files.dep.state.pa.us/Water/ChesapeakeBayOffice/WIPIII/FinalPlan/PA_Phase_3_WI
P_Final.pdf
Perron, J. T., Kirchner, J. W., & Dietrich, W. E. (2009). Formation of evenly spaced ridges and
valleys. Nature, 460(7254), 502–505. https://doi.org/10.1038/nature08174
Rathbun, J. (2009). Appendix A: Standard Operating Procedure - Monitoring Stream Bank
Erosion with Erosion Pins. 1–5.
Ritter, W. F. (2019). Progress on the Chesapeake Bay TMDL and Challenges in Meeting the
117
2025 Pollution-Reduction Loads. World Environmental and Water Resources Congress
2019. https://doi.org/https://doi.org/10.1061/9780784482339.006
Rosgen, D. L. (2001). A Practical Method of Computing Streambank Erosion Rates.
Proceedings of the Seventh Federal Interagency Sedimentation Conference.
Rosgen, D. L. (2006). Streambank erosion (BANCS model). In D. Frantila (Ed.), Watershed
Assessment of River Stability (WARSSS) (First, pp. 5-53-5–81). Friesens.
Rosgen, D. L., Keane, T. D., & Geenen, D. (2019). Discussion: “Evaluating the BANCS
Streambank Erosion Framework on the Northern Gulf of Mexico Coastal Plain” by Mitchell
McMillan, Johan Liebens, and Chris Metcalf. Journal of the American Water Resources
Association, 55(1), 274–280. https://doi.org/10.1111/1752-1688.12699
Sass, C. K., & Keane, T. D. (2012). Application of Rosgen’s BANCS model for NE Kansas and
the development of predictive streambank erosion curves. Journal of the American Water
Resources Association, 10(4), 277–287. https://doi.org/doi:10.1111 ⁄ j.1752-
1688.2012.00644.x
Schilling, K. E., & Wolter, C. F. (2001). Application of GPS and GIS to Map Channel Features
in Walnut Creek, Iowa. Journal Of The American Water Resources Association, 36(6),
1423–1434.
Schneiderman, E. M., Steenhuis, T. S., Thongs, D. J., Easton, Z. M., Zion, M. S., Neal, A. L.,
Mendoza, G. F., & Walter, M. T. (2007). Incorporating Variable Source Area Hydrology
into a Curve-Number-Based Watershed Model. Hydrological Processes, 21, 3420–3430.
https://doi.org/10.1002/hyp.6556
Schueler, T., & Stack, B. (2014). Recommendations of the Expert Panel to Define Removal Rates
for Individual Stream Restoration Projects.
118
https://www.chesapeakebay.net/documents/Final_CBP_Approved_Stream_Restoration_Pan
el_report_LONG_with_appendices_A-G_02062014.pdf
Secrist, M. A. (n.d.). Hickey Run Bank Erosion Rate Curves – Why, How, and the Future. U.S.
Fish and Wildlife Service.
https://cdn.ymaws.com/www.vwea.org/resource/collection/C560DED3-7431-43D8-AFE3-
9008EDA766DD/6-Mark_Secrist,_US_Fish_and_Wildlife_Service_H.pdf
Smith, D. G., Cragg, A. M., & Croker, G. F. (1991). Water Clarity Criteria for Bathing Waters
Based on User Perception. Journal of Environmental Management, 33(3), 285–299.
https://doi.org/10.1016/S0301-4797(91)80030-9
The Center for Watershed Protection. (2013). Cost-Effectiveness Study of Urban Stormwater
BMPs in the James River Basin.
http://stat.chesapeakebay.net/?q=node/130&quicktabs_10=3.%0Ahttp://www.jswconline.or
g/content/65/6/413.short
Thoma, D. P., Gupta, S. C., Bauer, M. E., & Kirchoff, C. E. (2005). Airborne laser scanning for
riverbank erosion assessment. Remote Sensing of Environment, 95, 493–501.
Thompson, T. (2020). Appendix A of Consensus Recommendations for Improving the
Application of the Prevented Sediment Protocol for Stream Restoration Projects Built for
Pollutant Removal Credit.
Thorne, C. R. (1982). Process and Mechanisms of River Bank Erosion (R. D. Hey, J. C.
Bathurst, & R. Thorne (eds.)). John Wiley & Sons.
U.S. Department of Agriculture. (2016). USLE History. U.S. Department of Agriculture.
https://www.ars.usda.gov/midwest-area/west-lafayette-in/national-soil-erosion-
research/docs/usle-database/usle-history/#:~:text=The Universal Soil Loss Equation
119
(USLE) is hailed as one,conservation in the 20th century.&text=USLE as a complete
technology,1
U.S. Environmental Protection Agency. (2010). Chesapeake Bay Total Maximum Daily Load for
Nitrogen, Phosphorus and Sediment.
U.S. Environmental Protection Agency. (2020). National Rivers and Streams Assessment 2013–
2014: A Collaborative Survey (EPA 841-R-19-001).
https://www.epa.gov/sites/production/files/2020-12/documents/nrsa_2013-
14_final_report_2020-12-17.pdf
Ursic, M. E., & Thorton, C. I. (2011). Quantification of Shear Stress in a Meandering Native
Topographic Channel Using a Physical Hydraulic Model. Colorado State University.
Van Eps, M. A., Formica, S. J., T.L. Morris, J. M. B., & Cotter, A. S. (2004). Using a Bank
Erosion Hazard Index (BEHI) to Estimate Annual Sediment Loads from Streambank
Erosion in the West Fork White River Watershed. In J. L. D’Ambrosio (Ed.), Proceedings
of the 12-15 September 2004 Self-Sustaining Solutions for Streams, Wetlands, and
Watersheds Conference (pp. 125–132). ASAE.
Virginia Department of Environmental Quality. (2010). Commonwealth of Virginia Chesapeake
Bay TMDL Phase I Watershed Implementation Plan: Revision of the Chesapeake Bay
Nutrient and Sediment Reduction Tributary Strategy. Virginia Department of
Environmental Quality.
https://www.mwcog.org/file.aspx?A=ll2%2FlxOkwgBec1IRI9jKuPbBWpS4QRK5YctyMF
u%2BarQ%3D
Virginia Department of Environmental Quality. (2015). Chapter 3 Virginia’s Environmental
Resources. In State Water Resources Plan (pp. 25–40). Virginia Department of
120
Environmental Quality.
Virginia Department of Environmental Quality. (2019). Commonwealth of Virginia Chesapeake
Bay TMDL Phase III Watershed Implementation Plan. Virginia Department of
Environmental Quality.
https://www.deq.virginia.gov/home/showpublisheddocument?id=4481
Walter, R. C., & Merritts, D. J. (2008). Natural streams and the legacy of water-powered mills.
Science, 319(5861), 299–304. https://doi.org/10.1126/science.1151716
Walter, R. C., Merritts, D., & Rahnis, M. (2007). Estimating volume, nutrient content, and rates
of stream bank erosion of legacy sediment in the Piedmont and Valley and Ridge
physiographic provinces, southeastern and central PA. In Pennsylvania Department of
Environmental Protection.
https://www.dep.pa.gov/PublicParticipation/AdvisoryCommittees/WaterAdvisory/Chesapea
keBayManagementTeam/Documents/padeplegacysedimentreport2007waltermerrittsrahnisfi
nal.pdf
West Virginia’s Chesapeake Bay Tributary Team. (2019). West Virginia’s Phase 3 Watershed
Implementation Plan for the Chesapeake Bay Total Maximum Daily Load. WV Department
of Environmental Protection.
http://www.wvca.us/bay/files/bay_documents/1298_WV_WIP3_final_082319.pdf
West Virginia WIP Development Team. (2010). West Virginia’s Chesapeake Bay TMDL
Watershed Implementation Plan. WV Department of Environmental Protection.
http://www.wvca.us/bay/files/bay_documents/208_WV_Final_WIP_I_Nov_29_2010.pdf
Wharton, G., Mohajeri, S. H., & Righetti, M. (2017). The pernicious problem of streambed
colmation: a multi-disciplinary reflection on the mechanisms, causes, impacts, and
121
management challenges. Wiley Interdisciplinary Reviews: Water, 4(5), e1231.
https://doi.org/10.1002/wat2.1231
Wilcock, P. R. (1996). Estimating local bed shear stress from velocity observations. Water
Resources Research, 32(11), 3361–3366. https://doi.org/10.1029/96WR02277
Willett, C. D., Lerch, R. N., Schultz, R. C., Berges, S. A., Peacher, R. D., & Isenhart, T. M.
(2012). Streambank erosion in two watersheds of the Central Claypan Region of Missouri,
United States. Journal of Soil and Water Conservation, 67(4), 249–263.
https://doi.org/10.2489/jswc.67.4.249
Williams, M. R., Bhatt, G., Filoso, S., & Yactayo, G. (2017). Stream Restoration Performance
and Its Contribution to the Chesapeake Bay TMDL: Challenges Posed by Climate Change
in Urban Areas. Estuaries and Coasts, 40(5), 1227–1246. https://doi.org/10.1007/s12237-
017-0226-1
Wischmeier, W. H., & Smith, D. D. (1965). Predicting Rainfall-Erosion Losses From Cropland
east of the Rocky Mountains: Guide for Selection of Practices for Soil and Water
Conservation. Agriculture Handbook No. 282, 28, 47.
Wischmeier, W. H., & Smith, D. D. (1978). Predicting Rainfall Erosion Losses: A Guide to
Conservation Planning,. Agriculture Handbook No. 537.A.
Wong, M., & Parker, G. (2006). Reanalysis and Correction of Bed-Load Relation of Meyer-Peter
and Müller Using Their Own Database. Journal of Hydraulic Engineering, 132(11), 1159–
1168. https://doi.org/10.1061/(asce)0733-9429(2006)132:11(1159)
Wood, P. J., & Armitage, P. D. (1997). Biological Effects of Fine Sediment in the Lotic
Environment. Environmental Management, 21(2), 203–217.
Wynn, T. (2006). Streambank Retreat: A Primer. In Watershed Update (Vol. 4, Issue 1).
122
http://www.remarkableriparian.org/pdfs/pubs/streambank retreat.pdf
Wynn, T., Benham, B. L., Zeckoski, R. W., Yagow, E. R., Staley, N., Simspon, A., Kozarek, J.,
& Lehman, S. (2006). Incorporating Stream Erosion and Restoration into Watershed
Models.
Wynn, T., Henderson, M. B., & Vaughan, D. H. (2008). Changes in streambank erodibility and
critical shear stress due to subaerial processes along a headwater stream, southwestern
Virginia, USA. Geomorphology, 97(3–4), 260–273.
https://doi.org/10.1016/j.geomorph.2007.08.010
Wynn, T., Hession, W. C., & Yagow, G. (2010). Stroubles Creek Stream Restoration.
Yamada, H., & Nakamura, F. (2002). Effect of fine sediment deposition and channel works on
periphyton biomass in the Makomanai River, northern Japan. River Research and
Applications, 18(5), 481–493. https://doi.org/10.1002/rra.688
123
Appendices
Appendix A: Stage and Flow Hydrographs
Figure A-1. Historic Stroubles Creek flow hydrograph showing peaks modelled where gaps in data occurred. Days where erosion pin measurements occurred are indicated by the measurement event points.
Figure A-2. Recent Stroubles Creek stage hydrograph. Days where erosion pin measurements occurred are indicated by the measurement event points.
124
Figure A-3. Toms Creek stage hydrograph. Days where erosion pin measurements occurred are indicated by the measurement event points.
Figure A-4. North Fork of the Roanoke River stage hydrograph. Days where erosion pin measurements occurred are indicated by the measurement event points. 125
Appendix B: Constructed BANCS Curves
Figure B-1. Erosion curve using bankfull erosion rates versus all NBS and selecting the highest.
Figure B-2. Erosion curve using bankfull erosion rates versus NBS Method 2.
126
Figure B-3. Erosion curve using bankfull erosion rates versus NBS Method 3.
Figure B-4. Erosion curve using bankfull erosion rates versus NBS Method 4.
127
Figure B-5. Erosion curve using bankfull erosion rates versus NBS Method 5.
Figure B-6. Erosion curve using all monthly erosion rates versus all NBS and selecting the highest.
128
Figure B-7. Erosion curve using all monthly erosion rates versus NBS Method 2.
Figure B-8. Erosion curve using all monthly erosion rates versus NBS Method 3. 129
Figure B-9. Erosion curve using all monthly erosion rates versus NBS Method 4.
Figure B-10. Erosion curve using all monthly erosion rates versus NBS Method 5.
130
Figure B-11. Erosion curve using one year time-averaged erosion rates versus all NBS and selecting the highest.
Figure B-12. Erosion curve using one year time-averaged erosion rates versus NBS Method 2.
131
Figure B-13. Erosion curve using one year time-averaged erosion rates versus NBS Method 3.
Figure B-14. Erosion curve using one year time-averaged erosion rates versus NBS Method 4.
132
Figure B-15. Erosion curve using one year time-averaged erosion rates versus NBS Method 5.
Figure B-16. Erosion curve using all monthly erosion rates versus number of peaks above baseflow. Only the historic dataset was used for regression. New site datapoints are only plotted for comparison.
133
Figure B-17. Erosion curve using all monthly erosion rates versus percent of time above baseflow. Only the historic dataset was used for regression. New site datapoints are only plotted for comparison.
Figure B-18. Erosion curve using bankfull erosion rates versus Equation 8.
134
Figure B-19. Erosion curve using bankfull erosion rates versus Equation 9/10.
Figure B-20. Erosion curve using all monthly erosion rates versus Equation 8. 135
Figure B-21. Erosion curve using all monthly erosion rates versus Equation 9/10.
Figure B-22. Erosion curve using one year time-averaged erosion rates versus Equation 8. 136
Figure B-23. Erosion curve using one year time-averaged erosion rates versus Equation 9/10.
137