ASSESSING BEST MANAGEMENT PRACTICES AND IMPLEMENTATION
STRATEGIES TO IMPROVE WATER QUALITY
By
Subhasis Giri

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Biosystems Engineering-Doctor of Philosophy
2013

ABSTRACT
ASSESSING BEST MANAGEMENT PRACTICES AND IMPLEMENTATION
STRATEGIES TO IMPROVE WATER QUALITY

By
Subhasis Giri

Best management practices (BMPs) are widely accepted to control both point and
nonpoint source pollution. However, the placement and selection of BMPs plays a vital role in
pollution reduction. Therefore, the objectives of this study were (1) to identify pollution
generating areas priority areas (high, medium, and low) using four different targeting methods:
concentration impact index (CII), load impact index (LII), load per subbasin area index (LPSAI),
and load per unit area index (LPUAI), (2) to apply BMPs in the identified priority area to
evaluate effectiveness, (3) to assess the spatiotemporal variability of critical source areas (CSAs),
and (4) to identify the best BMP and implementation site considering social, economic, and
environmental issues using different spatial targeting methods. Ten BMPs were implemented in
the identified priority areas in the Saginaw River Watershed by the four targeting methods using
the Soil and Water Assessment Tool (SWAT). Analytical hierarchy process (AHP) was used to
compare influential criteria with different weights during the BMP selection process. Results
suggest that the LPSAI is the best method for sediment targeting whereas the CII is the best
method for total nitrogen (TN) and total phosphorus (TP) targeting. Terraces and native grass
were the most effective BMPs whereas conservation tillage and no-till were the least effective
BMPs both at subbasin and watershed level analysis. In regard to the spatiotemporal variability
in the CSAs, a distinct change in high priority areas ranking was observed due to native grass

implementation by the end of second year whereas a minimal change in high priority areas was
found in case of contour farming due to the greater pollution reduction capacity of native grass
compared to contour farming. Based on environmental, economic, and social issues, strip
cropping was preferred in all CSAs based on the subbasin level analysis while strip cropping and
residue management were preferred in the CSAs for the watershed outlet analysis.

Copyright by
SUBHASIS GIRI
2013

ACKNOWLEDGEMENTS

First of all, I would like to thank my major professor Dr. Amir Pouyan Nejadhashemi for
giving me an opportunity and all his guidance, support, and advice throughout the journey. I
would also like to thank my committee members, Dr. Timothy Harrigan, Dr. Steven Safferman,
and Dr. Alexandra Sasha Kravchenko for their advice and support in completing this degree.
I would also like to thank my parents for believing in me and their support throughout my
life, and my brother, Dipu, for their love and support. To all my friends, who have made my life
easier while thousands of mile away from home.
Finally, I would like to thank my lab-mates Sean, Andy, Yaseen, Edwin, Matt, Georgina,
Brad, Melissa, Fariborz, and Umesh for their love and support which is an unforgettable
experience in my life.

v

TABLE OF CONTENTS
LIST OF TABLES ............................................................................................................. ix
LIST OF FIGURES .......................................................................................................... xii
LIST OF ABBREVIATIONS ........................................................................................ xviii
1

INTRODUCTION ........................................................................................................ 1

2

LITERATURE REVIEW ............................................................................................. 5
2.1 WATER ..................................................................................................................... 5
2.2 FRESH WATER.......................................................................................................... 5
2.3 FRESH WATER ECOSYSTEMS .................................................................................... 6
2.4 WATER QUALITY ..................................................................................................... 8
2.5 IMPAIRED RIVERS AND STREAMS, LAKES, PONDS, AND RESERVOIRS..................... 10
2.6 POLLUTANT DEGRADE WATER QUALITY ............................................................... 10
2.6.1 Sediments .................................................................................................. 10
2.6.2 Nutrients.................................................................................................... 12
2.6.3 Suspended Solids ....................................................................................... 14
2.6.4 Pathogens.................................................................................................. 15
2.6.5 Organic Material ...................................................................................... 17
2.6.6 Metals and Toxic Organic Chemicals....................................................... 18
2.6.7 Pesticides .................................................................................................. 18
2.6.8 Animal Waste ............................................................................................ 21
2.7 SOURCES OF WATER POLLUTION ........................................................................... 22
2.7.1 Point Source .............................................................................................. 22
2.7.2 Nonpoint Source (NPS) ............................................................................. 23
2.8 AGRICULTURAL POLLUTANTS AND THEIR ECONOMIC AND ENVIRONMENTAL RISKS24
2.8.1 Sediment Damage ..................................................................................... 24
2.8.2 Nutrient Damage....................................................................................... 25
2.8.3 Pesticide Damage ..................................................................................... 25
2.8.4 Animal Waste Damage .............................................................................. 26
2.10 MECHANISM TO CONTROL AGRICULTURAL NON-POINT SOURCE POLLUTION ...... 26
2.11 BEST MANAGEMENT PRACTICES .......................................................................... 27
2.12 COMPUTER MODELS ............................................................................................ 36
2.12.1 Soil and Water Assessment Tool ............................................................. 36
2.12.2 Hydrologic Simulation Program –FORTRAN ........................................ 37
2.12.3 Annualized Agriculture Non-Point Source Model .................................. 37
2.12.4 Agriculture Policy/Environmental eXtender........................................... 38
2.12.5 Spreadsheet Tool for Estimating Pollutant Load.................................... 38
2.12.6 GIS Pollutant Load Application.............................................................. 39
2.12.7 Long Term Hydrologic Impact Analysis ................................................. 39
2.12.8 High Impact Targeting ............................................................................ 40
vi

2.13 TARGETING APPROACH ....................................................................................... 40
3. INTRODUCTION TO METHODOLOGY AND RESULTS .................................... 41
4. EVALUATION OF TARGETING METHODS FOR IMPLEMENTATION OF BEST
MANAGEMENT PRACTICES IN THE SAGINAW RIVER WATERSHED 44
4.1ABSTRACT ........................................................................................................... 44
4.2 INTRODUCTION................................................................................................. 45
4.3 MATERIALS AND METHODS .......................................................................... 50
4.3.1 Study Area ................................................................................................. 50
4.3.2 Model Description .................................................................................... 52
4.3.3 Data Sources ............................................................................................. 54
4.3.4 Sensitivity Analysis and Calibration Process ........................................... 58
4.3.5 Best Management Practices in SWAT....................................................... 60
4.3.6 Spatial Targeting Methods ........................................................................ 65
4.3.7 BMP Relative Sensitivity Index ................................................................. 66
4.4 RESULTS AND DISCUSSIONS ......................................................................... 67
4.4.1 Sensitivity Analysis.................................................................................... 67
4.4.2 Model Calibration..................................................................................... 68
4.4.3 Spatial Targeting Methods ........................................................................ 70
4.4.4 Comparison of Agricultural Lands in Sub-targeting Scenarios for
Sediment, TN, and TP ............................................................................ 86
4.4.5 BMP Pollutant Reduction ......................................................................... 90
4.4.6 Evaluation of Relative Sensitivity Index among Targeting Methods ...... 131
4.5 CONCLUSION ................................................................................................... 137
5. ANALYSIS OF BEST MANAGEMENT PRACTICE EFFECTIVENESS AND
SPATIOTEMPORAL VARIABILITY BASED ON DIFFERENT
TARGETING STRATEGIES.......................................................................... 142
5.1 ABSTARCT ........................................................................................................ 142
5.2 INTRODUCTION............................................................................................... 143
5.2.1 CSA Identification ................................................................................... 145
5.3 MATERIALS AND METHODS ........................................................................ 147
5.3.1 Study Area ............................................................................................... 147
5.3.2 Model Description .................................................................................. 148
5.3.3 Data Sources ........................................................................................... 149
5.3.4 Sensitivity Analysis and Calibration ....................................................... 150
5.3.5 Best Management Practices in SWAT..................................................... 150
5.3.6 Spatial targeting Methods ....................................................................... 151
5.3.7 Statistical Analysis .................................................................................. 153
5.3.8 Spatial Correlation ................................................................................. 157
5.3.9 Spatiotemporal Variability of Priority Area ........................................... 160
5.4 RESULTS AND DISCUSSIONS ....................................................................... 161
5.4.1 Determining the Most Effective BMPs for Targeting and Non-targeting
Pollutants ............................................................................................. 161
5.4.2 Spatial Correlation among the Priority Methods ................................... 173
5.4.3 Spatiotemporal Variability of Priority Areas ......................................... 177
vii

5.5 CONCLUSION ................................................................................................... 181
6. APPLICATION OF ANALYTICAL HIERARCHY PROCESS FOR EFFECTIVE
SELECTION OF AGRICULTURAL BEST MANAGEMENT PRACTICES. .. 184
6.1 ABSTARACT ..................................................................................................... 184
6.2 INTRODUCTION............................................................................................... 185
6.3 MATERIALS AND METHODS ........................................................................ 188
6.3.1 Study Area ............................................................................................... 188
6.3.2 Model Description .................................................................................. 190
6.3.3 Environmental, Economic, and Social Aspects of BMP Implementation
Plan ...................................................................................................... 193
6.3.4 Identify the best BMP type and implementation site using Analytic
Hierarchy Process ............................................................................... 195
6.4 RESULTS AND DISCUSSION.......................................................................... 205
6.4.1 Determining the Cost of Pollution Reduction Associated with BMP
Installation Both at Subbasin Level and the Watershed Outlet ........... 205
6.4.2 Identify the Best BMP Type and Implementation Site Using Analytic
Hierarchy Process ............................................................................... 213
6.5 CONCLUSION ................................................................................................... 237
7.CONCLUSIONS.......................................................................................................... 240
8. RECOMMENDATION FOR FUTURE RESEARCH ............................................... 248
APPENDICES ................................................................................................................ 250
APPENDIX A ADDITIONALMATERIALTO SECTION 5 TITTLED
“ANALYSIS OF BEST MANAGEMENT PRACTICE EFFECTIVENESS AND
SPATIOTEMPORAL VARIABILITY BASED ON DIFFERENT TARGETING
STRATEGIES” .................................................................................................... 252
APPENDIX B
ADDITIONALMATERIALTO SECTION 6 TITTLED
“APPLICATION OF ANALYTICAL HIERARCHY PROCESS FOR
EFFECTIVE SELECTION OF AGRICULTURAL BEST MANAGEMENT
PRACTICES.” ...................................................................................................... 275
REFERENCES ............................................................................................................... 284

viii

LIST OF TABLES
Table 2-1. Common BMPs in the United States described in the NRCS technical guide.28
Table 4-1. Continuous corn conventional tillage management operations. ...................... 56
Table 4-2. Continuous soybean conventional tillage management operations. ................ 57
Table 4-3. Continuous soybean conservation tillage operations. ..................................... 62
Table 4-4. Continuous soybean no tillage operations. ...................................................... 63
Table 4-5. SWAT Inputs for Residue Management. ........................................................ 64
Table 4-6. Sensitivity analysis results for flow, sediment, TN, and TP. .......................... 68
Table 4-7. SRW calibration and validation results. .......................................................... 69
Table 5-1. Contingency table between CII and LPSAI targeting methods based on the
TN targeting scenario. The number in the parenthesis represents the ratio of
counts in that category (e.g. number of subbasins identified as high priority)
to the total number of counts (total number of subbasins)............................ 157
Table 5-2. P-values for BMPs based on the AR (1) model for different targeting
methods based on sediment targeting (sediment). ........................................ 163
Table 5-3. BMP implementation area for sediment targeting scenario. ......................... 164
Table 5-4. P-values for BMPs based on the AR (1) model for different targeting
methods based on sediment targeting for non-targeted TN. ......................... 166
Table 5-5. P-values for BMPs based on the AR (1) model for different targeting
methods based on sediment targeting for non-targeted TP. .......................... 168
Table 5-6. Summary of targeting component (TN) and non-targeting components
(sediment and TP). ........................................................................................ 170
Table 5-7. BMP implementation area for TN targeting scenario. .................................. 170
Table 5-8. Summary of targeting component (TP) and non-targeting components
(sediment and TN). ....................................................................................... 172
Table 5-9. Comparison of actual vs. modeled counts based on the agreement plus
linear-by-linear association model: CII and LPSAI targeting methods
(sediment). .................................................................................................... 174
ix

Table 5-10. Comparison of actual versus modeled counts based on agreement plus
linear-by-linear association model: CII and LPSAI targeting methods (TN).176
Table 5-11. Comparison of actual versus modeled counts based on agreement plus
linear-by-linear association model for CII and LII targeting methods (TP). 177
Table 6-1. Five-year itemized cost for different BMPs used in this study (NRCS, 2011).194
Table 6-2. The BMP allocation area and associated social preferences of different BMPs.195
Table 6-3. Pairwise comparison matrix developed for subbasin 43 based on sediment
reduction at the subbasin level. ..................................................................... 200
Table 6-4. Pairwise comparison matrix developed for subbasin 43 based on TN
reduction at the subbasin level. ..................................................................... 200
Table 6-5. Pairwise comparison matrix developed for subbasin 43 based on TP
reduction at the subbasin level. ..................................................................... 200
Table 6-6. Weight vector calculation of BMPs for sediment, TN, and TP reduction for
subbasin 43 at subbasin level. ....................................................................... 201
Table 6-7. Weight vector calculation of BMPs for total BMP cost. ............................... 202
Table 6-8. Weight vector calculation of BMP application area...................................... 202
Table 6-9. Weight vector of criteria used in this study. .................................................. 203
Table 6-10. Decision matrix of BMPs for all criteria developed for subbasin level
analysis (subbasin 43). .................................................................................. 203
Table 6-11. Final weight vector of individual BMPs for subbasin 43. .......................... 203
Table 6-12. BMP implementation area for different priority area based on sediment
targeting scenarios. ....................................................................................... 209
Table 6-13. Summary of AHP identified rank one BMP effectiveness in all aspects
(social, environmental, and economic) both subbasin level and watershed outlet.
....................................................................................................................... 235
Table A-1. P-values for BMPs based on AR (1) model for different targeting methods
based on TN targeting(Sediment). ................................................................ 252
Table A-2. P-values for BMPs based on AR (1) model for different targeting methods
based on TN targeting (TN). ......................................................................... 254

x

Table A-3. P-values for BMPs based on AR (1) model for different targeting methods
based on TN targeting (TP)........................................................................... 256
Table A-4. P-values for BMPs based on AR (1) model for different targeting methods
based on TP targeting (sediment). ................................................................ 258
Table A-5. P-values for BMPs based on AR (1) model for different targeting methods
based on TP targeting (TN)........................................................................... 260
Table A-6. P-values for BMPs based on AR (1) model for different targeting methods
based on TP targeting (TP). .......................................................................... 262
Table A-7. Spatial correlation among the targeting methods for different targeting
scenarios. ....................................................................................................... 264
Table B-2. Pairwise comparison matrix developed for subbasin 43 based on sediment
reduction at the watershed outlet. ................................................................. 276
Table B-3. Pairwise comparison matrix developed for subbasin 43 based on TN
reduction at the watershed outlet. ................................................................. 276
Table B-4. Pairwise comparison matrix developed for subbasin 43 based on TP
reduction at the watershed outlet. ................................................................. 276
Table B-5. Weight vector calculation of BMPs for sediment, TN, and TP reduction
for subbasin 43 at the watershed level. ......................................................... 276
Table B-6. Decision matrix of BMPs for all criteria developed for watershed level
analysis. ......................................................................................................... 277
Table B-7. Final weight vector of individual BMPs for subbasin 43. ........................... 277
Table B-8. BMP ranking at subbasin level based on environmental, economic, and
social factors. ................................................................................................ 278
Table B-9. BMP ranking at watershed level based on environmental, economic, and
social factors. ................................................................................................ 279
Table B-10. BMP ranking at subbasin level based on environmental and economic
factors. ........................................................................................................... 280
Table B-11. BMP ranking at watershed level based on environmental and economic
factors. ........................................................................................................... 281
Table B-12. BMP ranking at subbasin level based on environmental factors. ............... 282
Table B-13. BMP ranking at watershed level based on environmental factors. ............. 283
xi

LIST OF FIGURES
Figure 4-1. Saginaw River Watershed. For interpretation of the references to color in this
and allother figures, the reader is referred to the electronic version of this thesis. ... 51
Figure 4-2. Flow calibration and validation hydrograph. ............................................................. 69
Figure 4-3. (a) CII targeting method priority areas based on sediment concentration. ................ 71
Figure 4-3. (b) CII targeting method priority areas based on TN concentration. ......................... 72
Figure 4-3. (c) CII targeting method priority areas based on TP concentration. .......................... 73
Figure 4-4. (a) LII targeting method priority areas for sediment. ................................................ 75
Figure 4-4. (b) LII targeting method priority areas for TN.......................................................... 76
Figure 4-4. (c) LII targeting method priority areas for TP. ......................................................... 77
Figure 4-5. (a) LPSAI targeting method priority areas for sediment. ........................................... 79
Figure 4-5. (b) LPSAI targeting method priority areas for TN..................................................... 80
Figure 4-5. (c) LPSAI targeting method priority areas for TP. .................................................... 81
Figure 4-6. (a) LPUAI targeting method priority areas for sediment. .......................................... 83
Figure 4-6. (b) LPUAI targeting method priority areas for TN. ................................................... 84
Figure 4-6. (c) LPUAI targeting method priority areas for TP. .................................................... 85
Figure 4-7. Distribution of high, medium, and low priority areas for sediment targeting
scenario. ..................................................................................................................... 86
Figure 4-8. Distribution of high, medium, and low priority areas for TN targeting scenario. ..... 88
Figure 4-9. Distribution of high, medium, and low priority areas for TP targeting
scenario. ..................................................................................................................... 89
Figure 4-10. (a) Sediment reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area. ................................................................... 92
Figure 4-10. (b) Sediment reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area. ................................................................... 93

xii

Figure 4-10. (c) Sediment reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area. .................................................................. 94
Figure 4-10. (d) Sediment reduction by BMPs in high priority area in subbasin for different
targeting methods without normalizing the BMPs application area. ....................... 95
Figure 4-10. (e) Sediment reduction by BMPs in medium priority area in subbasin for different
targeting methods without normalizing the BMPs application area. ....................... 96
Figure 4-10. (f) Sediment reduction by BMPs in low priority area in subbasin for different
targeting methods without normalizing the BMPs application area. ....................... 97
Figure 4-11. (a) Sediment reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area. .................................................................. 98
Figure 4-11. (b) Sediment reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area. .................................................................. 99
Figure 4-11. (c) Sediment reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area. ................................................................ 100
Figure 4-11. (d) Sediment reduction by BMPs in high priority area in subbasin for different
targeting methods after normalizing the BMPs application area. .......................... 101
Figure 4-11. (e) Sediment reduction by BMPs in medium priority area in subbasin for different
targeting methods after normalizing the BMPs application area. .......................... 102
Figure 4-11. (f) Sediment reduction by BMPs in low priority area in subbasin for different
targeting methods after normalizing the BMPs application area. .......................... 103
Figure 4-12. (a) TN reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area. ................................................................ 105
Figure 4-12. (b) TN reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area. ................................................................ 106
Figure 4-12. (c) TN reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area. ................................................................ 107
Figure 4-12. (d) TN reduction by BMPs in high priority area in subbasin for different targeting
methods without normalizing the BMPs application area. .................................... 108
Figure 4-12. (e) TN reduction by BMPs in medium priority area in subbasin for different
targeting methods without normalizing the BMPs application area. ..................... 109
Figure 4-12. (f) TN reduction by BMPs in low priority area in subbasin for different targeting
methods without normalizing the BMPs application area. .................................... 110
xiii

Figure 4-13. (a) TN reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area. ................................................................ 112
Figure 4-13. (b) TN reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area. ................................................................ 113
Figure 4-13. (c) TN reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area. ................................................................ 114
Figure 4-13. (d) TN reduction by BMPs in high priority area in subbasin for different targeting
methods after normalizing the BMPs application area. ......................................... 115
Figure 4-13. (e) TN reduction by BMPs in medium priority area in subbasin for different
targeting methods after normalizing the BMPs application area. .......................... 116
Figure 4-13. (f) TN reduction by BMPs in low priority area in subbasin for different targeting
methods after normalizing the BMPs application area. ......................................... 117
Figure 4-14. (a) TP reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area. .......................................................... 119
Figure 4-14. (b) TP reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area. ................................................................ 120
Figure 4-14. (c) TP reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area. ................................................................ 121
Figure 4-14. (d) TP reduction by BMPs in high priority area in subbasin for different targeting
methods without normalizing the BMPs application area. .................................... 122
Figure 4-14. (e) TP reduction by BMPs in medium priority area in subbasin for different
targeting methods without normalizing the BMPs application area. ..................... 123
Figure 4-14. (f) TP reduction by BMPs in low priority area in subbasin for different targeting
methods without normalizing the BMPs application area. .................................... 124
Figure 4-15. (a) TP reduction by BMPs in reach for different targeting methods after normalizing
the BMPs application area. .................................................................................... 126
Figure 4-15. (b) TP reduction by BMPs in reach for different targeting methods after normalizing
the BMPs application area. .................................................................................... 127
Figure 4-15. (c) TP reduction by BMPs in reach for different targeting methods after normalizing
the BMPs application area. .................................................................................... 128

xiv

Figure 4-15. (d) TP reduction by BMPs in high priority area in subbasin for different targeting
methods after normalizing the BMPs application area. ......................................... 129
Figure 4-15. (e) TP reduction by BMPs in medium priority area in subbasin for different
targeting methods after normalizing the BMPs application area. .................... 130
Figure 4-15. (f) TP reduction by BMPs in low priority area in subbasin for different targeting
methods after normalizing the BMPs application area. ......................................... 131
Figure 4-16. (a) BMP relative sensitivity index sediment. ......................................................... 132
Figure 4-16. (b) BMP relative sensitivity index TN. .................................................................. 134
Figure 4-16. (c) BMP relative sensitivity index TP. ................................................................... 136
Figure 5-1. Location of Saginaw River Watershed. ................................................................... 147
Figure 5-2. (a) ACF-sediment base scenario and (b) ACF-high priority area . Bar beyond the
confidence band (dashed horizontal line) show significance at that time lag. ........ 154
Figure 5-2. (c) ACF-pooled series of sediment base and high priority area and (d) Partial ACFpooled series. Bar beyond the confidence band (dashed horizontal line) show
significance at that time lag. .................................................................................... 155
Figure 6-2. AHP flowchart to determine the rank of competing alternatives. ............................ 198
Figure 6-4. (a, b, and c) BMPs sediment reduction at the watershed outlet by different targeting
methods.................................................................................................................... 208
Figure 6-4. (d, e, and f) BMPs sediment reduction at the subbasin by different targeting
methods……………………………………………………………………………209
Figure 6-5. BMPs TN reduction at outlet (a, b, and c) and subbasin (d, e, and f) by different
targeting methods. ................................................................................................... 211
Figure 6-6. BMPs TP reduction at outlet (a, b, and c) and subbasin (d, e, and f) by different
targeting methods. ................................................................................................... 212
Figure 6-7. (a) Placement of BMP rank one in subbasin considering only environmental factor
based on CII targeting methods. .............................................................................. 217
Figure 6-7. (b) Placement of BMP rank one in subbasin considering only environmental factor
based on LPSAI targeting methods. ........................................................................ 218
Figure 6-7. (c) Placement of BMP rank one in subbasin considering only environmental factor
based on LPUAI targeting methods. ....................................................................... 219
xv

Figure 6-7. (d) Placement of BMP rank one in subbasin considering environmental-economic
factors based on CII targeting methods. .................................................................. 220
Figure 6-7. (e) Placement of BMP rank one in subbasin considering environmental-economic
factors based on LPSAI targeting methods. ............................................................ 221
Figure 6-7. (f) Placement of BMP rank one in subbasin considering environmental-economic
factors based on LPUAI targeting methods. ........................................................... 222
Figure 6-7. (g) Placement of BMP rank one in subbasin considering environmental-economicsocial factors based on CII targeting methods. ........................................................ 223
Figure 6-7. (h) Placement of BMP rank one in subbasin considering environmental-economicsocial factors based on LPSAI targeting methods. .................................................. 224
Figure 6-7. (i) Placement of BMP rank one in subbasin considering environmental-economicsocial factors based on LPUAI targeting methods. ................................................. 225
Figure 6-8. (a) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering only environmental factor by CII targeting method. 226
Figure 6-8. (b) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering only environmental factor by LPSAI targeting method.
................................................................................................................................. 227
Figure 6-8. (c) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering only environmental factor by LPUAI targeting method.
................................................................................................................................. 228
Figure 6-8. (d) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering environmental- economic factors by CII targeting
method. .................................................................................................................... 229
Figure 6-8. (e) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering environmental- economic factors by LPSAI targeting
method. .................................................................................................................... 230
Figure 6-8. (f) Placement of BMP rank one in subbasin based on pollution reduction at watershed
outlet considering environmental- economic factors by LPUAI targeting method. 231
Figure 6-8. (g) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering environmental-economic-social factors by CII targeting
method. .................................................................................................................... 232
Figure 6-8. (h) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering environmental-economic-social factors by LPSAI
targeting method. ..................................................................................................... 233
xvi

Figure 6-8. (i) Placement of BMP rank one in subbasin based on pollution reduction at watershed
outlet considering environmental-economic-social factors by LPUAI targeting
method. .................................................................................................................... 234
Figure 7-1. Targeting methods recommendation for different pollutants................................... 244
Figure 7-2. Spatial correlation between targeting methods in identifying priority areas. .......... 245
Figure 7-3. Effectiveness of different BMPs in this study.......................................................... 246
Figure 7-4. Most effective BMPs in CSAs both for subbasin and outlet for all scenarios. ........ 247
Figure A-8. CII priority areas – sediment targeting scenario. (a) base scenario, (b) year one
contour farming, (c) year two contour farming, (d) base scenario, (e) year one native
grass, (f) year two native grass. ............................................................................... 265
Figure A-9. LPSAI priority areas – sediment targeting scenario. (a) base scenario, (b) year one
contour farming, (c) year two contour farming, (d) base scenario, (e) year one native
grass, (f) year two native grass. ............................................................................... 266
Figure A-10. LPUAI priority areas – sediment targeting scenario. (a) base scenario, (b) year one
contour farming, (c) year two contour farming, (d) base scenario, (e) year one native
grass, (f) year two native grass. ............................................................................... 267
Figure A-11. CII priority areas – TN targeting scenario. (a) base scenario, (b) year one contour
farming, (c) year two contour farming, (d) base scenario, (e) year one native grass,
(f) year two native grass. ......................................................................................... 268
Figure A-12. LPSAI priority areas – TN targeting scenario. (a) base scenario, (b) year one
contour farming, (c) year two contour farming, (d) base scenario, (e) year one native
grass, (f) year two native grass. ............................................................................... 269
Figure A-13. LPUAI priority areas – TN targeting scenario. (a) base scenario, (b) year one
contour farming, (c) year two contour farming, (d) base scenario, (e) year one native
grass, (f) year two native grass. ............................................................................... 270
Figure A-14. CII priority areas – TP targeting scenario. (a) base scenario, (b) year one contour
farming, (c) year two contour farming, (d) base scenario, (e) year one native grass,
(f) year two native grass. ........................................................................................ 271

xvii

LIST OF ABBREVIATIONS
ACF

Autocorrelation function

AHP

Analytical Hierarchy Process

Alpha_BF

Baseflow Recession Constant

AnnAGNPS

Annualized Agriculture Non-Point Source

APEX

Agriculture Policy/Environmental eXtender

AR (1)

Autoregressive model of order one

BASINS

Better Assessment Science Integrating Point and Nonpoint Sources

BMP

Best management practice

CDL

Cropland Data Layer

CF

Contour farming

CH_K1

Effective hydraulic conductivity

CI

Consistency Ratio

CII

Concentration Impact Index

CN

CN2

CR

Consistency Ratio

CSAs

Critical source areas

CT

Conservation tillage

DEM

Digital elevation model

Esco

Soil Evaporation Compensation Factor

2

G

The likelihood ratio statistic

GIS

Geographic Information Systems

xviii

H

High

H+M

High and medium

H+M+L

High, medium, and low

Ha

Hectare

HIT

High Impact Targeting

HRU

Hydrologic response unit

HSPF

Hydrologic Simulation Program FORTRAN

HUC

Hydrologic unit code

Kg

Kilogram

Km

2

Square Kilometer

LH-OAT

Latin Hypercube One Factor-At-a-Time

LID

Low Impact Development

LII

Load Impact Index

LOADEST

Load Estimator

LPSAI

Load per Subbasin Area Index

LPUAI

Load per Unit Area Index

L-THIA

Long Term Hydrologic Impact Analysis

MAUT

Multi-attribute Utility Theory

MAVT

Multi-attribute Value Theory

MCDA

Multi-criteria Decision Analysis

MDEQ

Michigan Department of Environmental Quality

MUSLE

Modified Universal Soil Loss Equation

NASS

National Agricultural Statistics Service
xix

NCDC

National Climatic Data Center

NG

Native grass

NHD

National Hydrography Dataset

NPS

Nonpoint source

NRCS

Natural Resources Conservation Service

NSE

Nash-Sutcliffe efficiency

NT

No tillage

OV_N

Manning’s n value for overland flow

PACF

Partial autocorrelation function

PBIAS

Percent bias

R

2

Coefficient of determination

Rchrg_DP

Fraction of Percolation from Root Zone that Recharges Deep Aquifer

RM 0

Residue management (0 kg/ha)

RM 1000

Residue management (1000 kg/ha)

RM 2000

Residue management (2000 kg/ha)

RS

Recharge structure

RSI

Relative Sensitivity Index

RUSLE

Revised Universal Soil Loss

SC

Strip cropping

SCS

Soil Conservation Service

SEDMOD

Spatially Explicit Delivery Model

SMDR

Soil Moisture Distribution and Routing

Spcon

Linear Re-entrainment Parameter for Channel Sediment Routing
xx

SRW

Saginaw River Watershed

STATSGO

State Soil Geographic Database

SWAT

Soil and Water Assessment Tool

T

Terraces

TN

Total nitrogen

TP

Total Phosphorous

USGS

United States Geological Survey

USLE_C

Universal Soil Loss Equation Cover factor

USLE_P

Universal Soil Loss Equation Practice factor

xxi

1

INTRODUCTION
st

Managing global water resources is one of the greatest challenges in the 21 century (Staddon,
2010; Garrido and Dinar, 2008). Increased global populations and extreme events caused by
climate change are creating a growing pressure on water resources (Trenberth et al., 2003;
Raisanen et al., 2004; Giorgi et al., 2004). Based on United Nations Environment Programme
(UNEP), approximately 40% of world population could live in water scarce regions by the year
2025 (UNEP, 2005) which prompts maintaining the sustainability of water resources to avoid
adverse conditions in the future. Therefore, managing sustainability of water resources in both
quantity and quality, while maintaining a substantial improvement in global food and energy
security is a challenging task. (Jordan et al., 2012). In particular, maintaining proper water
quality is needed to protect human, animal, and plant health (Giri et al. 2012).

Degradation of water quality occurs due to both point and nonpoint sources (NPS). In
general,controlling NPS pollution is more difficult due to the lack of a single identifiable source.
Instead more diffuse and complex process create a wide range of complex uncertainty associated
with the simulation process (Maciej, 2000; Ouyang et al., 2009; Ding et al., 2010; Laia et al.,
2011). In United States, NPS pollution from agriculture is the primary contributor to water
quality impairment in rivers and lakes (USEPA, 2005). The primary pollutants from agriculture
that degrade water quality are sediment, nutrients, animal wastes, and pesticides. Excessive
sediment load degrades wildlife habitat, elevates dredging cost, and reduces storage capacity
(Ritter and Shirmohammadi, 2001). Increased nutrients (nitrogen and phosphorus) accelerate
eutrophication resulting in the killing of fish and clogging of pipelines (USEPA, 1998;

1

Carpenter, 2008). Also, eutrophication creates taste and odor problems, which elevate drinking
water treatment costs (Dodds et al., 2009).

Improving water quality from NPS pollution requires a strategy, which can prevent entry of
pollutants into waterbodies. Best management practices (BMPs) are well-known methods that
minimize the pollutant loads in the runoff generated from agricultural activities (Arabi et al.,
2007). However, pollutant reduction efficiency of BMPs varies from site to site (Giri et al.,
2012), which is due to changes in topography, soil characteristics, geological formations,
climate, crops, and cultural practices (Cunningham et al., 2003). It is impractical, expensive, and
time consuming to implement BMPs on all fields in a watershed, so the placement of BMPs in
Critical Source Areas (CSAs), regions that contribute the most to water quality impairment
(White et al., 2009; Giri et al., 2012). Implementing BMPs in CSAs helps to maximize pollution
reductions either at the edge of a field or the watershed outlet.

In this study, three knowledge gaps are addressed. These help policy makers and stakeholders to
develop more socially acceptable BMP implementation plan based on different scenarios in order
to maximize environmental benefit with minimal cost.

1) Identification of CSAs using multiple targeting techniques and pollutants: A single
method to prioritize the placement of BMPs in pollution generating area (targeting
method) is not efficient at identifying CSAs for all pollutants. For example, placement of
BMPs, based on the pollutant concentration in the river (targeting method) is more
appropriate for designing a stream restoration plan than the placement of BMPs based on

2

load from the agricultural fields. Similarly, in case of pollutants, one targeting method
may or may not be suitable for identifying CSAs for all pollutants (sediment, TN, and
TP) or even just one pollutant.
2) Spatiotemporal variation of CSAs: CSAs vary with respect to time after BMP
implementation. Therefore, knowledge of CSAs variation in space and time is required
during the BMP implementation plan in order to achieve maximum environmental
benefit throughout the lifespan of the BMP.
3) Selection of BMPs: The BMP selection process is generally based on pollutant removal
efficiency as a standalone criterion, which does not reflect the practical condition or the
maximum environmental benefit. Because the stakeholder input is not considered during
the BMP selection process, therefore, the selected BMP may or may not be installed by
the stakeholder. Hence, the BMP selection process should include different factors such
as environmental, economic, and social concerns to make the selected BMPs more
effective and practical.

Based on the above discussion, water quality improvement is greatly depended on producers’
acceptance level to new practices, targeting CSAs, and identification of BMPs in order to
maximize investment benefits of an implementation plan. The main objectives of this study are
to compare the effectiveness of current targeting methods to identify CSAs, design BMP
implementation strategies based on the spatial and temporal variation of CSAs, and identify the
best BMP type and implementation site while considering social, economic, and environmental.

3

The specific objectives for this study are:
1) Identify CSAs using multiple targeting techniques and pollutants
2) Assess the sensitivity of BMPs to reduce pollutants both in subbasin level and watershed
outlet based on different targeting methods
3) Evaluate the impact of BMP implementation in CSAs at both the field and watershed
level
4) Determine the most effective BMPs both for targeted pollutants and non-targeted
pollutants
5) Assess the spatial and temporal variability of CSAs
6) Evaluate spatial correlation among targeting methods in categorization of priority areas (
high, medium, and low)
7) Determine the most effective BMPs at reducing pollutants by considering both pollutant
reduction and BMP cost
8) Identify the best BMPs and implementation sites considering social, economic, and
environmental issues

4

2

LITERATURE REVIEW

2.1 WATER
Water is the basic ingredient of life and it is essential to all living organisms. It is one of the
primary solvents and mediums of reactions. Life without water is impossible. It starts with water
and relies on water for physical processes such as survival, growth, and development (Chang,
2006). Therefore, water is considered a precious resource. While water is a renewable source; its
availability varies according to location (Pimentel et al., 1997). Protection of water leads to
protection of life. Managing global water resources is one of the greatest challenges of the 21

st

century (Staddon, 2010; Garrido and Dinar, 2008). Overall, increased global populations and
extreme weather events caused by climatic change is creating a growing pressure on water
resources (Trenberth et al., 2003; Raisanen et al., 2004; Giorgi et al., 2004). According to United
Nations, up to 40% of the world’s population could live in water scarce regions by the year 2025
(UNEP, 2005). Therefore, sustainability of water resources is essential both in quality and
quantity.
2.2 FRESH WATER
Fresh water regulates different activities of human beings such as drinking, washing, cooking,
growing food, and personal hygiene, and supports all types of aquatic and terrestrial ecosystems
(Millennium Assessment, 2005). Approximately 96.5% of all water on the earth is saline ocean
water and of the remaining water, 2.5% is fresh water, 0.07% is in saline lakes, and 0.93% is in
saline ground water (USGS, 2012). Meanwhile, the quantity of available fresh water is calculated
to be 43,750 cubic kilometers per year, which is more than the combined requirement of

5

agriculture, industry, and households (Srebotnjak et al., 2012). However, fresh water is unevenly
distributed around the world (UNEP, 2005) and the demand for fresh water is increasing in many
parts of the world due to rapid growth of populations and a significant increase in water use for
different purpose such as agriculture and industry (Morrison and Gleick, 2004). Fresh water
resources are fixed, while the fresh water demand is increasing. For example, additional 64
billion cubic meters of fresh water is required to fulfill the need of increase in nearly 80 million
world population in each year (UN-Water, 2013). Therefore, fulfilling the fresh water needs for
growing populations and concurrently maintaining fresh water for aquatic life becomes one of
the most challenging issues for scientists, technologists, policy makers, and politicians during
st

this 21 century (Postel, 2000).

2.3 FRESH WATER ECOSYSTEMS
An ecosystem is the combination of both living organisms and the nonliving environment
interacting with each other as a functional unit (Hassan et al., 2005). Living organisms consist of
plants, animals, and microorganisms, whereas the nonliving environment includes water, air, and
soil. Ecosystems provide both direct and indirect services such as food, water, fuel, flood
regulation, and soil formation (Millennium Assessment, 2005). The fresh water ecosystem is one
of the most endangered ecosystems in the world. The decrease in biodiversity in fresh water is
greater than seen in terrestrial ecosystems (Sala et al., 2000). Fresh water has been over used; as
a result greater than 20% of the fresh water fish populations have become endangered (Dudgeon
et al., 2006). The availability of fresh water to meet the demand of the growing human
population and maintain the fresh water ecosystem integrity is a growing concern (Alcamo et al.,
2008). Overall, the primary factors that affect global fresh water biodiversity are divided into

6

five categories; overexploitation, water pollution, flow modification, destruction or degradation
of habitat, and invasion by exotic species (Postel and Richter, 2003; Revenga et al., 2005). Each
is explained in more detail below.


Overexploitation is the excessive removal of water from the ecosystems resulting in
decreased fresh water biodiversity (Dudgeon et al., 2006).



Excessive nutrients, especially nitrogen and phosphorus, promote algal blooms resulting
in a decrease in dissolved oxygen content. This reduces fish and other aquatic organism
populations (Das and Gazi, 2010). Excessive nitrate in drinking water is harmful to
infants as well as livestock (USGS, 2012).



Flow modification such as dams construction affect streams and rivers flow regimes and
subsequently river biota (Dudgeon et al., 2006).



Habitat degradation is the combination of direct effects such as river excavation or
indirect effects such as changes within drainage basins (Dudgeon et al., 2006). An
example of habitat degradation is the clogging of river bottom due to excessive river
sediment loading as a consequence of forest clearance.



Introduction of exotic species also have a negative impact on fresh water. For an
example, indirect impact of exotic terrestrial plant Tamaricaceae affects stream flow in
Australia and North America (Dudgeon et al., 2006).

A mutual compatibility between humans and natural ecosystems helps in maintaining a
sustainable ecosystem, which is challenging. Part of sustainable ecosystems is acknowledging
the societal benefits obtained from a functional ecosystem (Everard and Moggridge, 2012).
Human activities are linked to different ecological processes through the act of supplying fresh

7

water (Falkenmark and Folke, 2003). For example, dams and water diversion structures provide
fresh water for domestic, commercial, industrial, and agricultural uses (Meybeck, 2004), which
changes the hydrology and direction of flow affecting the aquatic organisms and disrupting other
ecosystem services that can be obtained from the river like pollutant reduction. Therefore,
society is continuously losing ecosystem services provided by the fresh water ecosystem
throughout the world (Fitzhugh and Richter, 2004). Identifying ecosystem services such as
pollutant reduction may help directly or indirectly with social acceptance of sustaining fresh
water ecosystems (Palmer et al., 2005).
2.4 WATER QUALITY
Managing sustainability of water resources while maintaining profitable agriculture and
enhancing the environment is a challenging task. (Jordan et al., 2012). A waterbody is declared
impaired when it cannot be used for the designated purpose such as drinking, navigation,
recreation, fishing, and wildlife (USEPA, 2009). A reference water quality standard is developed
to control the water quality of the degraded waterbodies which varies from one waterbody to
another depending upon the usage.

Water quality can be improved at two levels: 1) ultimate goal and 2) periodic goal (Enderlein et
al., 2012). The ultimate goal allows the water to be used by society without any adverse effect,
which would also support the aquatic ecosystem. The periodic goal is designed to reach the
ultimate goal.

In order to implement strategies to improve water quality, identification of the cause of water
quality degradation and the location is the priliminary step (Srebotnjak et al., 2012). Many
8

physical, biological, and chemical parameters can be used as water quality criteria (UNEP GEM
Water, 2006). Therefore, there is no single water quality parameter that defines a healthy
freshwater source (Srebotnjak et al., 2012). However, a numeric water quality criterion describes
the minimum number of physical, chemical, and biological parameters required to support the
water quality of a waterbody (USDA, 2012). Physical and chemical parameters may be used to
describe the minimum concentration of a required parameters or maximum concentration of
pollutants whereas biological parameters can be used to describe possible community attributes
(USDA, 2012).

Water quality standards consists of three components: 1) designated usage, 2) criteria or
threshold, and 3) anti-degradation policy (USEPA, 2009). After setting the water quality
standard, monitoring of water is required to check the degree in which the water quality is being
maintained. Monitoring of water quality data is categorized into six types; physical, chemical,
biological integrity, microbial, habitat, and toxicity (USEPA, 2009).

In order to protect water quality in the United State, the total maximum daily load (TMDL)
program was introduced under the Clean Water Act section 303 (Riebschleager, 2008). A TMDL
is the total amount of pollutant a waterbody can receive while maintaining its water quality
standard (USEPA, 2007). The TMDL is the sum of point source load plus nonpoint source load
plus a margin of safety.

Extensive studies have been done on water quality, specifically how to maintain water quality by
either developing a water quality index, finding the source of pollution, or implementing best

9

management practices (BMPs) (Espejo et al., 2012; Zhoua et al., 2012; Kovacsa et al., 2012; Lee
et al., 2012; Santhi et al., 2006). Zhoua et al. (2012) performed a study to assess the effect of
landscape pattern on water quality of a river in China. They found that changes in the landscape
due to human activities had a major impact on flow and water quality. Espejo et al. (2012) found
that small areas in their watershed were contributing higher amounts of phosphorus to the river,
which can be minimized by introducing BMPs in these areas.

2.5 IMPAIRED RIVERS AND STREAMS, LAKES, PONDS, AND RESERVOIRS
According to the United States Environmental Protection Agency (USEPA), out of 16% of
nation’s 3.5 million miles of rivers and streams, 44% were declared as impaired and did not meet
the water quality standard (USEPA, 2009). Similarly, out of 39% of the nation’s 41.7 million
acres of lakes, ponds, and reservoirs; 64% did not support the designated usage of water, such as
swimming, boating, drinking purposes, and other recreational activities (USEPA, 2009).
2.6 POLLUTANT DEGRADE WATER QUALITY
Pollutants, which degrade the water quality, have a significant environmental impact. Therefore,
identification of these pollutants is required in order to protect water quality. These pollutants
include sediments, nutrients, silt, suspended particles, pathogens, organic materials, metals, toxic
organic chemicals, pesticides, and herbicides (USEPA, 1992).
2.6.1 Sediments
Sediment is one of the primary pollutants in degrading water quality (Ondrusek et al., 2012) and
is the major component of pollution in the United States (Hangsleben and Suh, 2006). It is
10

generally found in rivers, streams, lakes, and other waterbodies (Malone, 2009). Approximately,
US $16 billion can be attributed to sediment damage each year in United States (MARC, 2012).
Sediment consists of individual primary particles, aggregates, organic materials, and other
chemicals (Haan et al., 1994). It can be generated by either soil erosion or decomposition of
plant and animal material (MARC, 2012); however, the primary contributor is erosion (Persyn et
al., 2005), in fact 99% of all total suspended solids in waterbodies is from soil erosion (Ritter and
Shirmohammadi, 2001). The primary carriers of eroded soil are rainfall, runoff, and wind. The
major factors influencing sediment pollution are rainfall, wind, soil erodibility, slope, and crop
factor. Out of all these factors, rainfall is most difficult to control.

Different natural factors (such as extreme weather events) and human activities foster formation
as well as transformation of soil loss (Wu et al., 2012). Specially, improper agricultural
management practices, such as excessive disturbance of soil through plowing and farming in
steep areas enhance soil erosion. There are three types of soil erosion: sheet erosion, rill erosion,
and gully erosion. Sheet erosion is the detachment of thin layers of soil by raindrop impact and
shallow surface flow (NSW-DPI, 2012). This type of erosion is found commonly in the fields
without vegetation and have been plowed.Rill erosion is developed due to concentrated flow,
where the raindrop impact is insignificant (Govers et al., 2007). This type of erosion is
commonly found in loose structured soil and bare or overgrazed soil. Gully erosion is developed
by a concentrated flow when the velocity is strong enough to detach large quantities of soil
particles.

11

In a river system, excessive fine sediment influences the physical processes and ecological
functioning of riverbeds (Haynes et al., 2009). For example, clogging of the interstitial spaces in
gravel alluvial river beds by fine sands delays the hyporheic exchange of water and dissolved
constituents, which ultimately affects the oxygenation of fish eggs, nutrient cycling, and
pollutant retention (Gartner et al., 2012).
2.6.2 Nutrients
Nutrients are chemical composites, which mainly consist of nitrogen or phosphorus and serve as
an essential element for plant and animal nutrition (USGS, 2006), as well as stream biotic
activity (Mulholland and Webster, 2010). However, excess amount of nutrients cause detrimental
effects to living organisms. Excessive nutrients from NPS source are the primary cause of
eutrophication in lakes, streams, and other waterbodies in the United States (Carpenter et al.,
1998). NPS pollution are four times greater than the nutrient loading from point sources
(Carpenter et al., 1998). The average annual nutrient concentration of streams originating from
agricultural watersheds is nine times greater than forested watersheds and four times greater than
rangeland watersheds (Brown and Froemke, 2012). The primary sources of NPS nutrients in
streams are agricultural fertilizer, livestock manure, and atmospheric deposition (Carpenter et al.,
1998; Mallin, 2000; Jones et al., 2001; Driscoll et al. 2003; Dubrovsky et al., 2010). Nitrogen
and phosphorus are the two primary nutrients originating from agricultural lands, and they have
the greatest effect on water quality (USEPA, 2003). However, these two nutrients are essential
for high crop yields; yet the losses of these nutrients downstream can create negative
environmental impacts to aquatic life and human beings (Vitousek et al., 2009). According to the
United States Environmental Protection Agency (USEPA), streams that exceed NO3-N

12

concentrations greater than 10 mg/l cannot be used as public water supplies (Jha et al., 2010;
Schilling and Wolter, 2009). When phosphorus level in the aquatic environment increased 0.01
mg/l, the productivity of the aquatic plant creates odor and taste problem in water (USEPA,
2003). These excessive nutrients create a negative impact on the local ecosystems. In addition,
these nutrients can be transported downstream, which ultimately lead to the degradation of
downstream ecosystems communities (Rabalais et al., 2010). Therefore, elevated nutrient
load/concentration creates a negative impact on both local and downstream ecosystem due to
fluvial linkage of ecosystems (Covino et al., 2012).

Different chemical and biological processes alter the form of nutrients resulting in a nutrient
cycle between soil, water, atmosphere, and biological organism. The primary nutrients essential
for plant growth are nitrogen, potassium, and phosphorus. A major driving force behind the
increasing nitrogen and phosphorus input was fertilizer of industrially fixed N and mined
phosphate rocks. Approximately, 11 million tons of nitrogen, 5 million tons of potash, and 4
million tons of phosphate fertilizers are applied to cropland in the United States each year
(USDA-ARS, 1997). Apart from fertilizer application, livestock wastes contain high amount of
nutrients, hormones, and steroids (Johnson et al., 2006). These enter waterbodies when grazing
animals deposit their waste adjacent or directly into streams (Kolodziej and Sedlak, 2007). This
increased nutrient application rate, animal waste, and intensive cropping pattern may be the
cause of the increased nutrient concentration in the waterbodies (Verma et al., 2012).

Nutrients enter into water resources primarily by three different methods: runoff, runin, and
leaching (USDA, 2012). Runoff is the flow of excess water from rainwater and/or melting snow

13

after complete saturation of soil, which ultimately carries the pollutants along the soil surface.
Runin transfers nutrients directly into ground water through porous medium in soil or any rock
fracture, while leaching carryies nutrients in soil through percolation. The delivery of nutrients
downstream is a function of landscape and instream processes. The mobility of land-based
nutrients depend on several parameters such as the amount, timing, and composition of fertilizer
application, the amount, intensity, and location of precipitation, the location and extent of land
disturbances, and stormwater control efficiency (Lebo et al., 2012). Therefore, determining the
overall contributions of nutrients to waterbodies relies on the understanding of environmental
factors such as hydrology, climate variability, soil, terrain, and land use (Zhu et al., 2012).
2.6.3 Suspended Solids
Solid materials travel with most natural water in a suspension state (Alabaster, 1972). When the
water carrying the suspended particles is slow enough, then the suspended particles gradually
settle down to the bottom of the waterbodies (RAMP, 1996). A suspended solid is the material
that can be trapped by a filter. It consists of different material such as silt, decaying plant and
animal matter, industrial wastes, and sewage (LRRB, 2010). The different potential sources of
suspended solids are agriculture, dredging, flooding, forest fires, logging activities, mining,
recreational boating and navigation, roads, and urban development. Suspended solids are
composed of both organic and inorganic fractions. The organic fraction contains algae,
zooplankton, bacteria, and detritus, whereas the inorganic fraction consists of silts and clays.
Suspended solids have a negative impact on water treatment cost, aesthetic value, fishery
resources, and ecology (Bilotta and Brazier, 2008; Bilotta et al., 2012). For example, increase in
water turbidity due to suspended solids reduces light penetration into waterbodies, which
ultimately decrease photosynthesis creating an adverse effect on fish and other aquatic species
14

(Devlin and McVay, 2001). In addition, deposition of suspended solids prevents the exchange of
dissolved oxygen and carbon dioxide between flowing water and respiring eggs leading to
reduce survival and development of salmon eggs (Bilotta and Brazier, 2008).
2.6.4 Pathogens
Impairment of surface water by a pathogen places the waterbody into 303(d) list of USEPA
(Hathaway and Hunt, 2012). The elevated pathogen concentration is the primary reason for a
surface water impairment.

Pathogen concentrations are generally estimated using indicator

organisms. In agricultural watersheds, the environmental benchmark for water quality analysis is
waterborne pathogen standard (Edge et al., 2012). Bacterial pathogen primarily, Escherichia coli
(E. coli) and fecal coliform are the main cause of waterbodies impairment in the United States
(USEPA, 2008). When a stream segment exceeds the indicator organism, E. coli 394 cfu/100mL
or the geometric sample mean 126 cfu/100mL in 25% of the total samples, the stream segment is
considered as impaired (Riebschleager, 2008). The E. coli (O157:H7) can cause hemorrhagic
colitis (HC) and hemolytic uremic syndrome (HUS) in humans. Other enteropathogenic E. coli
causes diarrhea in children (Ibekwe et al., 2011). Apart from E. coli, Zoonotic waterborne
protozoa, specifically Cryptosporidium and Giardia spp can cause diarrhea (Carey et al., 2004;
Fayer, 2004). Also, Cryptosporidium oocysts and Giardia cysts can persist in adverse conditions
and contribute to health risks (Hogan et al., 2012). Therefore, pathogen contaminated
waterbodies is a vital public health concern in the multiuse ecosystems.

The potential sources of pathogens are from either point sources or nonpoint sources such as:
confined animal feeding operations, waste water treatment plants, combined sewer overflows,
slaughterhouses, meat processing facilities, tanning, textile, pulp and paper factories, and fish
15

and shellfish processing facilities (USEPA, 2001c). With a growing numbers of livestock,
increasingly intensive cattle, hog, and poultry farming operations, a variety of pathogens can be
observed from livestock fecal wastes (Olson et al., 1997; Hutchison et al., 2004; Lee and Newell,
2006; Opporto et al., 2007). The nonpoint pathogen sources consist of urban litter, excrement
from barnyards, pastures, contaminated refuse, domestic pet and wildlife excrement, feedlots and
uncontrolled manure storage areas in rural or agricultural areas, failing sewer lines in urban and
suburban areas, small confined animal operations, onsite wastewater systems, and rangelands
(Paul, 2003). Human population, housing density, and land development have strong positive
correlation with bacterial densities (Young and Thackston, 1999). Additionally, specific human
enteric diseases are linked to high cattle density found from epidemiological studies (Michel et
al., 1999; Valcour et al., 2002).

The fate and transport of pathogens depends on the different factors such as temperature and
rainfall intensity (Jokinen et al., 2012). For example, precipitation can carry pathogens from land
to waterbodies or mobilize pathogens within the flow zone by increasing the water flow to the
waterbodies (Wilkes et al., 2011). Therefore, in order to understand the fate and transport of
pathogens, it is essential to understand the interaction of pathogens with land use, hydrology, and
temperature (Auld et al., 2004; Thomas et al., 2006; Wilkes et al., 2009). This includes the
occurrence and densities of pathogens in waterbodies and the ability to specify pathogens
associated point and non-point sources (Ice, 2004; USEPA, 2005; Benham et al., 2006; Rao
et al., 2009).

16

2.6.5 Organic Material
Organic material is composed of both living and non-living carbon based materials (ASWP,
2012) and can occur naturally or may be from municipal and industrial effluent runoff (Malcolm,
1985). Natural organic material consists of complex organic materials, which exists in the
environment (Teixeiraa and Nunesb, 2011). The quality and concentration of organic matter
depends on several factors such as climate, location, and environmental (Kotti et al., 2005).
Eatherall et al. (2000) observed that the higher concentrations of dissolved organic matter in the
watershed is due to the sewage point source and the diffuse nonpoint sources during low and
high flows, respectively.

The water quality, food webs, and structural complexity of forested headwater streams are
influenced by coarse particulate organic matter and associated elements (Benstead et al. 2009).
The amount of coarse particulate organic matter that is transported from the headwater of a
stream primarily depends on rainfall, plant phenological patterns, stream flow, channel
geomorphology, rates of litter decomposition and biotic processing (Scalley et al., 2012). Like
particulate organic matter, dissolved organic matter also plays a significant role in the
biogeochemical cycles of carbon and nitrogen (White et al., 2010). Dissolved organic matter
helps in carrying different elements from one place to another. For example, mercury binds with
organic matter facilitating its transport from forested ecosystems to water ecosystems (Akerblom
et al., 2008; Brigham et al., 2009).

17

2.6.6 Metals and Toxic Organic Chemicals
Elevated concentrations of metals such cadmium, lead, cobalt, copper, mercury, chromium,
nickel, selenium, zinc, and iron in stream sediment and water is a serious and widespread
environmental problem. These metals originated from various sources such as mining, industrial
development, and increase in urban area

(Buyuksonmez et al., 2012). Several researchers

(Milner and Kochian, 2008; Das et al., 2011; Leung et al., 2008; Johri et al., 2010) have looked
at the health risks caused by metals. Metal contamination in the aquatic environment becomes a
serious concern due to their toxicity levels. For example, metals can deposit on microorganisms,
aquatic flora and fauna, and pass through the food chain to the human body causing health
hazards (Varol and Sen, 2012; Sin et al., 2001).

In general, metals are heavier than other particles and insoluble in water, they deposit on the
bottom of rivers with sediment (Tsakovski et al., 2012). Therefore, sediments play a major role
in the storage and transportation of metals in waterbodies (Lee et al., 2012). Apart from the
above factors, the fate and transport of heavy metals rely on several processes related to
redistribution, mobility, and transformation.

2.6.7 Pesticides
The term pesticide is defined as mixture of substances, which are made for preventing, killing,
and mitigating any types of pest (USEPA, 1993). Pesticides can be subdivided to herbicides,
insecticides, fungicides, miticides, and nematicides (USEPA, 2003). Pesticide from agricultural
fields is a common nonpoint source of contamination in waterbodies and is regarded as one of
the greatest stressors to stream ecosystems (Rasmussena et al., 2011). The use of pesticides to

18

protect crops from insects, fungi, and weeds is a general agricultural practice. However,
pesticides can contaminate both surface and ground water (Juniori and Silva, 2011). Agricultural
pesticides are primarily applied during growing period, early spring through fall. Apart from
agricultural pesticides, pesticides are heavily used in urban areas, especially in residential lawns,
gardens, parks, and golf courses (Glozier et al., 2012). The sources of pesticides are liquid and
solid wastes from pesticide manufacturing, residual post-applied plant pesticides, washing
equipment used with pesticides, and pesticides packing and distribution (Grynkiewicz et al.,
2003) whereas the other sources of pesticides consist of agricultural and horticultural activities,
landfills, destruction of unwanted vegetation, and forest protection activities (Babu et al., 2011).

Pesticides are a concern to both the public and water quality managers due to their negative
impacts on both aquatic life and human health (Kreuger, 1996). Pesticides enter waterbodies
through diffuse pathways such as runoff, aerial drift, leaching, volatilization, and food chain
movement (USEPA, 2003). Both pesticides in solution and pesticides attached to soil particle
can be transported during runoff. Organic carbon and clay content of soil helps in partitioning
pesticide between solution and solid phase (Katrijn et al., 2007).

Some pesticides do not degrade quickly and may persist and accumulate in waterbodies.
Pesticide contamination occurs due to different factors such as improper application, erosion,
cropping systems, inadequate equipment maintenance, inappropriate selection, leaching, artificial
drainage, and volatilization (UNEP, 1998). The threat pesticides present to water quality depends
on both the application method and location (USEPA, 2003). Peak concentration and duration of
exposure are two factors, which determine the degree of the negative impact the pesticides will

19

have on flora and fauna in surface water (Kreuger, 1998). Up to five percent of applied pesticides
can be lost through leaching, but generally, it is less than one percent (Carter, 2000). Surface
water is more susceptible to pesticide contamination than ground water due to easy entry of
pesticides into surface water . According to Gilliom et al. (2007) pesticides were found more
than 90% of the time in streams in developed watersheds (dominated by agriculture, urban and
mix land use) but only slightly greater than 50% of the time in shallow ground water. Pesticides
are characterized as highly soluble and have a low adsorption capacity to soil, which enables
them to reach waterbodies easily (Glozier et al., 2012). In addition, conversion of natural areas
into urban areas increases the impervious surface that reduces infiltration and helps in easy
transportation of pesticides.

Studies on herbicides showed that less than two percentage of the total herbicide applied ended
up in the nearby waterbodies shortly after application (Capel et al., 2001). The loss of herbicides
into a watershed depends on weather conditions, soil type, land use, intrinsic properties of the
compound (Kreuger, 1998; Capel et al., 2001; Katrijn et al., 2007).

The possible effects of pesticides on the environment and human health raised public concern,
which ultimately resulted into laws regulating pesticide concentration in different parts of the
world (Glozier et al., 2012). For example in Canada, two types of environmental thresholds are
suggested, (1) ecological threshold and (2) achievable threshold (Environment Canada, 2010).

In the United States, both Federal and State agencies are actively working to reduce pesticide
transport into waterbodies in order to prevent the negative impacts of pesticides to stream

20

ecology and human health (Lerch et al., 2011). Many studies have been conducted on pesticides
in order to understand their fate and transport (Hoffman et al. 2000; Kolpin et al. 2006; Weston
et al. 2009). A detailed understanding of watershed- scale pesticides mobility is possible through
understanding the complex biogeochemistry and interactions pesticides have with hydrologic
processes and land use (Zanardo et al., 2012). Lerch et al. (2011) stated that the (1) chemistry of
the contaminant, (2) hydrology and soils of the watershed, (3) land (i.e. herbicide use and crop
management), and (4) climate (particularly precipitation) are the primary factors controlling
watershed susceptibility to pesticide transport.

2.6.8 Animal Waste
The size, production, and amount of livestock manure has increased with the increasing number
of animal feeding operations (Jenkins et al., 2009). In fact, 32% of all land in the United States is
used as rangeland where cattle are free to access the waterbodies within it (USDA, 2005), and as
the animals graze, they deposit feces on the land, which then enter the waterbodies through
runoff (Kolodziej and Sedlak, 2007). The chance of spreading waterborne pathogens and
diseases through waterbodies has potentially increased. Animal waste consists of fecal and
urinary waste both from livestock and poultry, which includes processed water from milking
parlor, feed, bedding, and litter (USEPA, 2003). These wastes contains high concentration of
pollutants such as nutrients, steroid estrogens metals, salts, organic solids, bacteria, viruses, and
other microorganisms (Hanselman et al., 2003; Johnson et al., 2006; Mallin and Cahoon, 2003;
USDA, 1992). The presence of so many contaminants in the animal waste can create adverse
effects in both surface and ground water (USDA, 1992). These nutrients, steroids, metals, and

21

microorganisms can enter into the waterbodies through runoff or leachate from agricultural lands
(Kjaer et al., 2007; Matthiessen et al., 2006).
Poultry litter consists of a mixture of feces, bedding material, and feathers. These poultry litters
are applied to agricultural lands to provide nutrients such as nitrogen, phosphorus, and potassium
(Moore et al., 1995). Additionally, these poultry litters contain pathogenic bacteria, such as
Salmonella, Campylobacter, and fecal coliform, and appreciable concentration of the sex
hormones (Jeffrey et al., 1998; Jenkins et al., 2006). Application of poultry litter increases the
potential risk of contaminating surface and ground water with these pollutants.

2.7 SOURCES OF WATER POLLUTION
Water pollution is of great concern in both developed and developing countries. Water pollution
control is required to improve water quality. In order to do that, we need to identify the different
sources of water pollution. Water pollution occurs due to both point and nonpoint sources. Point
sources are easy to control as they come from a single identifiable source, whereas nonpoint
sources are difficult to control due to the associated complex diffuse processes (Chiwa et al.,
2012; Giri et al., 2010; Carpenter et al., 1998).

2.7.1 Point Source
The point source is the confined, single, exact point where pollution originates and discharges
into waterbodies (USEPA, 2012). A point source could be a pipe, channel, conduit, rolling stock,
concentrated animal feeding operation, municipal sewage treatment plants, or industrial facility.
Point source pollution generation is not controlled by natural factors; it is controlled by
anthropogenic processes such as industrial activities. Therefore, point source pollution can be
22

measured and controlled periodically at a single place (Carpenter et al., 1998). In order to control
the point sources in the United States, the Environmental Protection agency (EPA) introduced
the National Pollutant Discharge System (NPDES), which is under Clear Water Act of 1972
(USEPA, 2012). It is a permit-based system that regulates the discharge of point source
pollution.

2.7.2 Nonpoint Source (NPS)
Being the largest threat to water resources, NPS pollution has drawn attention from both
scientific communities as well as government agencies (Wanga et al., 2012). NPS pollution is a
spatially diverse load carried by surface and subsurface runoff to receiving waterbodies (Laia et
al., 2011). NPS pollution results from agricultural activities, precipitation, atmospheric
deposition, street runoff, infiltration, and drainage (Shi et al., 2012). NPS pollution is more
complex and harder to identify and control, due to complex uncertainties associated with the
simulation process (Ding et al., 2010; Laia et al., 2011; Maciej, 2000; Ouyang et al., 2009).
Factors affecting NPS pollution are soil, topography, climate, hydrology, and land use types (Ou
and Wang, 2008). Runoff is the primary mechanism that transports pollutants from one place to
another. For example, when runoff or snowmelt washes over the surface it carries pollutants such
as sediments, nutrients, organic matter, bacteria, oil, and metals to the receiving waterbodies.

In the United States, NPS pollution from agricultural activities is recognized as a primary source
of water pollution (USEPA, 2003). The increasing use of chemicals in agriculture to meet the
food demand of a constantly growing population degrades water quality. In order to maintain the
balance between crop productivity and the negative impact of NPS pollutant on water quality, we

23

need to understand the fate of NPS pollutants (Wali et al., 2011). In the United States, in order
to control the NPS pollution, the agricultural NPS policies, the Coastal Nonpoint Pollution
Control Program, and National Estuary Program under section 320 of the Clean Water Act were
developed (USEPA, 2012; Shortle et al., 2012).

2.8 AGRICULTURAL POLLUTANTS AND THEIR ECONOMIC AND ENVIRONMENTAL RISKS
The primary pollutants originating from agriculture are sediment, nutrients, animal wastes, and
pesticides. Increased amounts of these pollutants degrade water quality, which ultimately
increases the cost of water use. The following are the detail description of each pollutant and its
economic impact on water quality.

2.8.1 Sediment Damage
Increase in sediment content affects waterbodies in several ways. Excessive sediment decreases
water storage capabilities, degrade wildlife habitat, and increases dredging cost (Ritter and
Shirmohammadi, 2001). In the United States (US), the excessive sedimentation causes
approximately, $13.4 billion per year as an external cost (Tegtmeier and Duffy, 2004). Sediment
loads result in temperature changes and deplete oxygen levels, which is harmful to aquatic
organisms (Malone, 2009). Increases in sediment decreases the aesthetic value of water,
increases clarity problems, and hampers recreational activities (USDA, 1997). In addition, higher
sediment load reduces fish growth rates, reduce availability of food for fish (which changes
migration pattern), and also can be toxic at higher concentration (Newcombe and MacDonald,
1991). Excessive sediment raises the streambed, which ultimately increases the probability and
severity of flooding. Suspended sediment increases the purification cost of drinking water and

24

increases the water treatment cost of municipal and industrial use (USDA, 2012). In addition,
suspended sediment decreases the amount of sunlight available to the aquatic plants and clogs
the gills of fish (USEPA, 2003).

2.8.2 Nutrient Damage
Increase in nitrogen and phosphorus in the fresh water and coastal ecosystems accelerates
eutrophication, which results in fish death and clogged pipelines (Carpenter, 2008; USEPA,
1998). Additionally, recreational activities such as swimming, fishing, and boating are affected
due to excessive plant growth and odor and taste problem of the water (USEPA, 1999).
Lowering of oxygen level creates hypoxic zones, which do not support life. Also, studies have
shown that movement of nitrogen from land into surface water creates serious environmental
hazard such as death and abnormalities in amphibians (Rouse et al., 1999) and dead zones like in
the Gulf of Mexico (Ebionews, 2010). Nitrate is a potential threat to human health especially to
infants, which can prevent supply of sufficient oxygen to the blood stream called blue baby
syndrome (USDA, 2012). Increased phosphorus content contributes to odor and taste problems
and also increases the purification cost of drinking water (Malone, 2009). According to USEPA,
a total of $200 million is required to maintain the federal nitrate standard in drinking water due
to nitrate contamination of drinking water resources (USDA, 2012).

2.8.3 Pesticide Damage
All pesticides have harmful effect on humans, animals, and aquatic life (Lorenz, 2009).
Increased pesticide content kills fish, frogs, turtles, and other wildlife (Helfrich et al., 2009;
Gormley et al. 2005). In addition, increased pesticide amounts in waterbodies promote the

25

growth of algae, which results in interference of swimming, fishing, and boating (Helfrich et al.,
2009). Pesticides can cause health hazards for humans by consumption of contaminated fish
(UNEP, 1998; USDA, 2012).

2.8.4 Animal Waste Damage
Several contaminates present in animal waste can have negative effects both to surface and
ground water. Additionally, specific elements of animal waste can adversely affect terrestrial
plants, grazing animals, and air quality (USDA, 1992). Endocrine glands of aquatic organism can
be interrupted by estrogen from animal wastes at concentrations higher than 10 mg/L (Young et
al., 2004). Increases in animal waste application on agricultural lands increases the probability of
eutrophication in surface water as well as disruption of the endocrine gland in aquatic organisms
(Yonkos et al., 2010; Mallin and Cahoon , 2003; Kellog et al., 2000). Presence of estradiol in
poultry litter has potential to affect ecological as well as public health (Jenkins et al., 2009).
Additionally, estrogen in poultry litter can have a negative effect (premature udder growth and
lactation) in heifer (Shore et al., 1995). Testosterone present in animal waste (Lintelmann et al.,
2003) can create negative impacts on aquatic ecology (Jenkins et al., 2009). Phosphorus present
in animal manure can enter waterbodies and causes eutrophication (Soupir et al., 2006; Sharpley
et al., 1993), which ultimately leads to low oxygen levels, reduced aquatic species diversity,
turbidity, and poor taste and odor in waterbodies (Hansen et al., 2002).

2.10 MECHANISM TO CONTROL AGRICULTURAL NON-POINT SOURCE POLLUTION
The non-point source pollution from agricultural activities can be controlled by different
management activities such as BMPs systems, accepted agricultural practices, management

26

measures, management practice systems, resources management systems, and total management
systems (USEPA, 2012). This study discusses how BMPs can be effectively used to control
agricultural NPS.

2.11 BEST MANAGEMENT PRACTICES
BMPs are the practices and procedures that minimize the amount of pollutants in runoff from
agricultural activities while providing a viable economic option to the farmers (UNEP, 1998).
BMPs reduce NPS pollution by three mechanisms: 1) reducing pollution mass through erosion
control practices and concentration through nutrient and pesticide managements, 2) reducing
pollutant delivery to waterbodies through filter strip or different types of vegetative barrier, and
3) remediation through chemical and biochemical processes (Cunningham et al., 2003). In
general, BMPs are categorized primarily into structural and non-structural (Kaplowitz and Lupi,
2009). A structural BMP requires a construction or more permanent land use change to capture
runoff (e.g., filter strip or artificial wetland) (Sommerlot et al., 2013). A non-structural BMP
does not require construction but modifications of agricultural practices (UNEP, 1998).
Meanwhile, some researchers have classified BMPs into three categories: structural (such as
manure storage facilities, stream fencing and stabilization, and alternative watering systems),
vegetative (such as cover crops, filter strips, riparian buffers, and reforestation), and management
(such as loafing lot systems, and rotational grazing) (Cunningham et al., 2003).

Some of the common BMPs described in USDA NRCS conservation practice technical
documents are provided below. Table 2-1 describes the BMP type, definition and specification,
and purpose and condition of BMPs application.

27

Table 2-1. Common BMPs in the United States described in the NRCS technical guide.
BMP
Definition
Purpose
Condition of
Specification
application
Conservation

Crops grown

-Reduce sheet

-Applied in all

-Fallow land should not

Crop Rotation

in a planned

and rill and

cropland where

cover more than 25% of

(NRCS Code

sequence on

wind erosion

one third of the

the planned crop

328)

the same

-Improve soil

crop sequence

sequence during

field (USDA-

quality

is produced by

uncropped period

NRCS, 2010)

- Conserve

annual crops

- A planned two crop

water

sequence should contain

-Reduce energy

a warm and cool season

use by

crop

supplying
nitrogen through
nitrogen fixation
Terraces

Earthen

-Reduce

-Area having

-Capacity to control 10-

(NRCS Code

embankment,

erosion,

excessive

year 24-hour storm

600)

which

conserve

slope, excess

-The ridge should have

consists of

moisture by

runoff

a minimum width of 3

ridges and

retaining runoff

- Soil and

ft.

channel

topography are

-Maximum allowable

across the

suitable for

slope should be 2

field slope

terrace

horizontal to 1 vertical

(USDA-

construction

-Length of the terrace

28

Table 2-1 (cont’d)
NRCS, 2010)

should not exceed 3500
ft. in order to avoid the
potential failure risk

Strip

Systematic

-Reduce soil

-Applied in all

-Strips of the crops

Cropping

arrangement

erosion by water croplands and

should be placed at an

(NRCS Code

of equal

and wind

other suitable

angle perpendicular to

585)

width of row

-Protect

areas for

the water and wind

crops,

growing crops

growing crops

erosion forces

forages, and

from wind-

-At least 50% of the

small grains

borne soil

cover consists of

(USDA-

particle

erosion resistant crops

NRCS, 2010)

-Reduce amount

or sediment trapping

of sediments

cover

and water-borne

-Strip boundary should

contaminants in

be parallel to each other

the runoff

and as close as possible
to the contour

Contour

Consists of

-Reduce sheet

-Applicable on

-Minimum ridge height

Farming

ridges and

and rill erosion

sloping land

of two inches should be

29

Table 2-1 (cont’d)
(NRCS Code

furrows

-Increase

suitable for

maintained during the

330)

formed by

infiltration

annual crops

rotation period for row

different

-Reduce

spacing greater than 10

farming

sediment,

inches

operations

suspended solid,

-Minimum ridge height

such as

and

of one inch should be

tillage and

contaminants in

maintained for close

planting. It

the runoff

grown crops having row

helps in

spacing 10 inches or

changing the

less

direction of

-Row grade should not

runoff from

exceed 0.2%

down slope
to around the
hill (USDANRCS, 2010)
Conservation

Establishing

-Reduce

-Applicable on

-Perennial crop

Cover

and

sedimentation

lands require a

vegetation should

(NRCS Code

maintaining a

and soil erosion

permanent

provide full ground

327)

permanent

-Improve soil,

vegetative

coverage in the pathway

vegetative

water , and air

cover

during mowing and

cover

quality

harvesting

(USDA-

-Helps in

-Combination of

30

Table 2-1 (cont’d)
NRCS, 2010)

wildlife habitat

grasses, forbs, legumes,

and pollinator

and shrubs should be

habitat

planted in order to
promote biodiversity

Constructed

Artificial

-Improves storm

-For

-Construction of an

Wetland

ecosystem

water quality

agricultural

auxiliary spillway or

(NRCS Code

consisting of

-Improves water

waste water

inlet bypass to control

656)

hydrophytic

from waste

management

peak flow of 25-year,

vegetation

water treatment

system where

24-hour storm

(USDA-

plant,

constructed

-Provide an suitable

NRCS, 2010)

agricultural

wetland is a

inlet control structures

processing,

component

to avoid entering debris

livestock, and

into wetland

aquaculture

-Minimum of two rows

facilities

of functionally parallel
cells should be
considered in the
wetland system design

Filter Strip

Growing

-Reduce

-Applicable to

-Minimum width of the

(NRCS Code

herbaceous

sediment,

environmental

filter strip should be 20

393)

vegetation in

suspended

sensitive areas

feet

a strip to

solids, and other

require

- Slope of one percent

remove

contaminants

reduction of

or greater is preferable

31

Table 2-1 (cont’d)
contaminants

-Decrease

sediment,

for the area up-stream

from

dissolved

suspended

of filter strip

overland flow contaminants in

solids, and

-Maximum of four

(USDA-

dissolved

inches plant spacing is

contaminants

preferable in the filter

irrigation water

NRCS, 2010)

strip
Residue and

Distribution

-Reduce soil

- Applicable to

-Soil tillage intensity

Tillage

of amount

erosion,

all cropland

rating value should be

Management

and

especially rill

within 20

(NRCS Code

orientation of

and sheet

-Crops having row

329)

crop and

erosion

spacing less than 15

other plant

- Increase the

inches should have a

residue on

organic matter

minimum 10 inches

soil surface

of the soil

crop stubble height

throughout

-Reduce energy

-One to three inches

the year, also

use and CO2

deep soil disturbance is

minimizing

losses from soil

preferable to release

soil

-Increase

less CO2

disturbance

moisture content

(USDA-

of the soil which

NRCS, 2010)

increases the
plant available
water

32

Table 2-1 (cont’d)
Grassed

Establishmen

-Protect erosion

-Areas need

-Can control the peak

Waterway

t of suitable

and flooding in

erosion control

runoff from 10-year 24-

(NRCS Code

vegetation

the terraces and

from

hour rainfall

412)

along the

diversions

concentrated

-Bottom width of the

graded

-Improve water

flow

grass waterway should

channel

quality and

be less than 100 feet

which

reduce soil

-Slide slope should be

prevents

erosion

less than 0.5

erosion

-Freeboard should be

(USDA-

provided above the

NRCS, 2010)

designed depth to avoid
damage

Sediment

Construction

-Capture

-Urban land,

-Basin can at least store

Basin

of a basin by

sediment in the

construction

3600 ft /acre of

(NRCS Code

earthen

runoff and

sites,

350)

embankment,

provide longer

agricultural

or

period of time

lands

excavation,

which allows to

or

settle in the

combination

basin

3

drainage area
- Length to width ratio
of the basin should be 2
to1 or greater
-Construction of porous
baffles in the entire

of both

basin is recommended

(USDA-

to control the

NRCS, 2010)

turbulence in the basin
33

Table 2-1 (cont’d)
Riparian

Areas

-Improve

-Applied in an

-Minimum width of the

Forest Buffer

dominated by

aquatic life

area closer to

riparian forest buffer

(NRCS Code

shrubs and

habitat by

streams, ponds,

should be 35 ft.

391)

trees located

maintaining the

and wetlands

-Sheet flow is preferred

closer to the

water

through the riparian

waterbodies

temperature

buffer

(USDA-

-Decrease

-Native and non-

NRCS, 2010)

amount of

invasive trees and shrub

sediments,

species are in the

nutrients, and

riparian buffer

pesticides in the
runoff
-Increase carbon
storage
-Reduce
pesticide drift
into the
waterbodies
- Help in
stabilizing
stream bank or
shoreline

34

In the United States, different agencies such as United States Department of Agriculture
(USDA), Natural Resources Conservation Service (NRCS), Environmental Protection Agency
(EPA), and other state natural resources and agricultural agencies work with landowners to
implement agricultural BMPs in their fields to reduce NPS pollutants. For examples, USDA
offers different conservation programs to promote BMPs implementation such as Conservation
Reservation Program (CRP), the Environmental Quality Incentive Program (EQIP), and the
Conservation Stewardship Program (CSP) (Sommerlot et al., 2012).

The pollution reduction efficiency of a BMP is site specific (Giri et al., 2012a). BMP efficiency
varies due to topography, soil characteristics, geological formations, climate, crops, and cultural
practices (Cunningham et al., 2003). Implementation of BMPs is difficult due to conflicting
environmental, economic, and institutional interests (Arabi et al., 2007). Certain pollutants
cannot be controlled appropriately by a single method for all situations (USEPA, 1999).
Therefore, selection, design, and implementation of BMPs are required to evaluate potential
impact.

Meanwhile, a broad range of BMPs is available to the farmers, each having their own strength
and limitations. The wide array of BMPs and influential criteria makes selecting a BMP a
daunting task. Apart from selection, placement of BMPs also plays an important role in pollutant
reduction efficiency, as pollutant contribution is disproportionate between areas in a watershed
(Tripathi et al. 2003). In addition, implementation of BMPs randomly throughout a watershed is
time consuming, expensive, and resource intensive. Also, in general monitoring water quality is
expensive, time consuming, and ineffective for larger area. However, developments of powerful

35

watershed/water quality computer models are providing reliable information regarding BMP
effectiveness both at field and watershed levels.

2.12 COMPUTER MODELS
BMP effectiveness in a watershed can be evaluated through computer modeling, since computer
modeling can accurately capture site-specific characteristics (Tuppad et al., 2010) by using
various input data. Several watershed models such as Soil and Water Assessment Tool (SWAT),
Hydrologic Simulation Program FORTRAN (HSPF), Annualized Agriculture Non-Point Source
(AnnAGNPS), Agriculture Policy/Environmental eXtender (APEX), Spreadsheet Tool for
Estimating Pollutant Load (STEPL), GIS Pollutant Load Application (PLOAD), Long Term
Hydrologic Impact Analysis (L-THIA), and High Impact Targeting (HIT) are available. The
following sections represent each model, their components, methods, and finally the application
of the model.

2.12.1 Soil and Water Assessment Tool
SWAT is specifically used to predict the impact of land use management practices on water,
sediments, and agrochemicals in a complex watershed with different soil, land use, and
management scenarios (Parajuli et al., 2009; Boscha et al., 2011; Zhang et al., 2011; Gassman et
al., 2007; Gassman et al., 2010).

SWAT is a physically based and spatially distributed

watershed-scale model developed by USDA-ARS (Arnold et al., 1998; Neitsch et al., 2005;
Gassman et al., 2007). SWAT uses readily available input data and can simulate flow, sediment,
nutrient, and pesticide movement in a watershed. The major components of this model include
weather, hydrology, soil characteristics, plant growth, nutrients, pesticides, and land management

36

practices (Gassman et al., 2007). Surface runoff is calculated either by the Soil Conservation
Service (SCS) curve number or by the Green and Ampt infiltration method. Sediment yield for
hydrologic response units is calculated by the Modified Universal Soil Loss Equation (MUSLE),
and Manning’s equation is used for flow calculation (Neitsch et al., 2005).

2.12.2 Hydrologic Simulation Program –FORTRAN
The HSPF is useful to evaluate the impact of land use change and point and NPS pollution
management scenarios. HSPF model is one of the most widely used water quality model
developed by EPA to simulate hydrology and water quality (Bicknell et al., 1997). This model is
a comprehensive and continuous watershed-scale model, which simulates several water quality
parameters (Bicknell et al., 2000). Pollutant loads are estimated by taking into consideration both
in-stream processes and overland flow. The hydrological response units are considered during
overland flow (Bhaduri et al., 2000). The HSPF model components consist of sediment,
phosphorus, nitrogen, dissolved oxygen, pesticides, zooplankton, temperature, and pH. This
model uses many theoretically and empirically developed relations in order to simulate physical,
chemical, and biological processes. This model is incorporated into EPA-Better Assessment
Science Integrating Point and Non-point (BASIN) and is used for total maximum daily load
analysis.

2.12.3 Annualized Agriculture Non-Point Source Model
AnnAGNPS was developed by the USDA to simulate the complex problems related to NPS
pollution. It estimates NPS pollution on a daily time step through continuous and event-based
simulations (Bosch et al., 2001). This model simulates runoff, sediment, and nutrients from a

37

watershed dominated by agriculture lands using a cell-by-cell basis (Finn et al., 2002). These
cells are formed by dividing the watershed into uniform square areas (Polyakov et al., 2007). The
model requires 22 input parameters, which come from base data elevation, land cover, and soil
(Finn et al., 2002). The model components include hydrology, sediments, nutrients, pesticides,
irrigation, precipitation, and snowmelt (Bosch et al., 2001). The lateral subsurface flow is
calculated by Darcy’s equation. Flow in tile drain is estimated by Hooghoudt’s equation.
Sediment load is calculated by the Revised Universal Soil Loss (RUSLE) equation and sediment
transport in channel is estimated by the revised Einstein equation (Bosch et al., 2001).

2.12.4 Agriculture Policy/Environmental eXtender
APEX is used to determine land management strategies, erosion, soil quality, and plant
competition for small watersheds or whole farm managements (Williams et al., 2008). APEX can
be used to model field- scale management practices such as furrow diking, buffer strips, terraces,
waterways, tillage, grazing, pesticides, crop rotation, and manure management, (Williams and
Izaurralde, 2005). The model components comprise of hydrology, sediments, nutrient cycling,
pesticides, crop growth, tillage, and weather simulation (Williams et al., 2008). In APEX, surface
runoff is calculated by SCS curve number. Peak runoff is estimated by TR-55 and soil erosion is
calculated by different methods such as Universal Soil Loss Equation (USLE), Revised
Universal Soil Loss Equation (RUSLE), and MUSLE (Williams et al., 2008).

2.12.5 Spreadsheet Tool for Estimating Pollutant Load
STEPL is a watershed-scale model used to estimate pollutant load in the stream. This model
calculates the impact of BMPs and low impact development (LIDs) on sediments and nutrients

38

load from various land uses (Tetra Tech, 2006). The outputs are surface runoff, nutrients load,
and five-day biological oxygen demand. The model components consist of runoff, ground water,
all types of erosion, and pollutant transport (Tetra Tech, 2006). The annual sediment load is
estimated by universal soil loss equation and sediment reduction is calculated by predefined
BMP efficiencies (Nejadhashemi et al., 2011).

2.12.6 GIS Pollutant Load Application
PLOAD is designed to calculate the impact of BMPs on pollutant loads on an annual basis
(USEPA, 2001). The input data for this model are land use, watershed physiographical
characteristics, BMP site, pollutant loading rate, impervious terrain factor, point source location,
and loads (USEPA, 2001). This model is generally recommended when a high level of
uncertainty with BMP effectiveness is acceptable. Several studies have used PLOAD model to
evaluate water quality due to NPS pollution (Nejadhashemi et al., 2011; Endreny and Wood,
2003).

2.12.7 Long Term Hydrologic Impact Analysis
L-THIA was developed as a spreadsheet tool but it was incorporated into geographic information
systems (GIS) by Perdue University (Bhaduri et al., 1997). The input data for this model are
climate data, soil, and land use. The model outputs include runoff and NPS pollution and are
presented in the form of tables and graphs (Bhaduri et al., 2000). This is a lumped parameter
model where curve number method is used to calculate the annual runoff (Bhaduri et al., 2000).
Several studies have used L-THIA to predict NPS pollution in different watersheds (Bhaduri et
al., 2000; Muthukrishnan, 2002; Yang et al., 2006; Nejadhashemi et al., 2011).

39

2.12.8 High Impact Targeting
HIT is the combination of Revised Universal Soil Loss Equation (RUSLE) and Spatially Explicit
Delivery Model (SEDMOD), which estimates annual sediment loading for waterbodies (IWR,
2012). The HIT model identifies critical areas for sediment loading both at the field level and the
watershed- scale. The input data required is to identify the critical erosive areas are land cover,
soil clay content, digital elevation model, land use/ tillage, soil erodibility, rainfall, and support
practices (IWR, 2012).

2.13 TARGETING APPROACH
Effective control of agricultural NPS pollution requires BMP implementation in the areas of
watersheds that produce more pollution. These areas, called CSAs, produce a disproportionate
amount of pollution in the watershed (White et al., 2009). Prioritizing BMP implementation
based on CSAs is called a targeting approach, which can resulted in greater pollution reduction
(Giri et al., 2012). This prioritization is based on their pollution load generation. The CSAs of
NPS pollution can be further divided based on land resources and water quality prospective
(Maas et al., 1985). Based on the land resources, CSAs are the areas where soil erosion rates
exceed the soil tolerance value (maximum annual soil loss without hampering current crop
production level). From the water quality prospective, CSAs are the areas where the greatest
improvement in water quality can be achieved with minimum BMP implementation cost
(Tripathi et al., 2003). The identification of CSAs and the selection of appropriate BMPs are
performed by watershed and water quality models and tools such as SWAT and GIS.

40

3.

INTRODUCTION TO METHODOLOGY AND RESULTS

This thesis consists of three research papers out of which two are already published and the third
one is submitted. In the first paper, the CSAs were identified by using different targeting
methods and BMPs were implemented using in order to evaluate the effectiveness of BMPs
using these targeting methods. In the second paper, the additional benefit of BMP
implementation was considered which is reduction of non-targeting pollutant while reduction of
targeting pollutants. Also, the effect of multiyear BMP implementation plan in high priority area
was examined in order to determine the order of BMP implementation in the watershed in order
to achieve maximum BMP efficiency. Additionally, a correlation between two targeting methods
was identified to reduce the number of targeting methods to simplify. In the third paper, BMPs
and placement location were identified by considering environmental, economic, and social
factors simultaneously using the targeting methods.

The title of the first paper is “Evaluation of targeting methods for implementation of best
management practices in the Saginaw River Watershed”. The objectives were to identify priority
areas (high, medium, and low) using four different targeting methods: concentration impact
index(CII), load impact index (LII), load per subbasin area index (LPSAI), and load per unit area
index (LPUAI); and to evaluate the effectiveness of BMPs in the priority areas. SWAT was used
to simulate ten BMPs in the identified priority areas. Results suggested that the LPSAI targeting
method is able to identify the highest priority areas for sediment, whereas the CII targeting
method was able to identify the highest priority areas for TN and TP. Additionally, out of the ten
BMPs, terraces and native grass were the most effective whereas conservation tillage and no-till
were least effective both for subbasin and watershed outlet analysis.
41

The second paper is titled “Analysis of best management practice effectiveness and
spatiotemporal variability based on different targeting strategies”. The objectives of this study
were to assess the spatiotemporal variability of critical source areas (CSAs) and to evaluate the
spatial correlation among the targeting methods in the categorization of priority areas based on
different pollutants. The study area was same as the previous study and SWAT was used to
simulate BMPs in the high priority areas of all four targeting methods. For spatiotemporal
analysis of priority areas, native grass and contour farming were implemented for two
consecutive years. Kappa, weighted kappa coefficient, and agreement plus linear-by-linear
association model were used to determine the correlation among the targeting method in
categorizing priority areas. Results suggest that a distinct change in high priority areas of native
grass was observed by the end of second year whereas a minimal change in high priority area
was found in case of contour farming due to greater pollution reduction capacity of native grass
compared to contour farming. A strong agreement was found between LPSAI and LPUAI
targeting methods in categorization of priority areas for sediment and TN targeting as these
methods are both based on pollutant load targeting.

The title of the third paper is “Application of analytical hierarchy process for effective selection
of agricultural best management practices”. The objective of this study was to identify the best
BMP and implementation site using the analytical hierarchy process (AHP) while considering
social, economic, and environmental issues under different spatial targeting methods. Five
BMPS, strip cropping, residue management, conservation tillage, native grass, and no till, were
implemented in the agricultural lands of Saginaw River Watershed using SAWT based on the

42

high priority areas of three targeting methods (CII, LPSAI, and LPUAI). Based on environmental
factors, native grass was selected most over the CSAs for subbasin level analysis while strip
cropping, residue management, and native grass were selected in the CSAs for watershed outlet
analysis. However, native grass was replaced by strip cropping in CSAs of both subbasin level
and watershed scale analysis, when BMP selection was based on environmental and economic
factors. When the BMP selection was based on environmental, economic, and social issues, strip
cropping was preferred in all the CSAs based at the subbasin level analysis while strip cropping
and residue management were selected in the CSAs for the watershed outlet analysis.

43

4. EVALUATION OF TARGETING METHODS FOR IMPLEMENTATION OF BEST
MANAGEMENT PRACTICES IN THE SAGINAW RIVER WATERSHED

4.1ABSTRACT
Increasing concerns regarding water quality in the Great Lakes region are mainly due to changes
in urban and agricultural landscapes. Both point and nonpoint sources contribute pollution to
Great Lakes surface waters. BMPs are a common tool used to reduce both point and nonpointsource pollution and improve water quality. Meanwhile, identification of critical source areas of
pollution and placement of BMPs plays an important role in pollution reduction. The goal of this
study is to evaluate the performance of different targeting methods in 1) identifying priority areas
(high, medium, and low) based on various factors such as pollutant concentration, load, and
yield, 2) comparing pollutant (sediment, total nitrogen-TN, and total phosphorus-TP) reduction
in priority areas defined by all targeting methods, and 3) determining the BMP relative
sensitivity index among all targeting methods. Ten BMPs were modeled in the Saginaw River
Watershed using the Soil and Water Assessment Tool (SWAT) model following identification of
priority areas. Each targeting method selected distinct high priority areas based on the
methodology of implementation. The concentration based targeting method was most effective at
the reduction of TN and TP, likely because it selected the greatest area of high priority for BMP
implementation. The subbasin load targeting method was most effective at reducing sediment
because it tended to select large, highly agricultural subbasins for BMP implementation. When
implementing BMPs, native grass and terraces were generally the most effective, while
conservation tillage and residue management had limited effectiveness. The BMP relative
sensitivity index revealed that most combinations of targeting methods and priority areas resulted
44

in a proportional decrease in pollutant load from the subbasin level and watershed outlet.
However, the concentration and yield methods were more effective at subbasin reduction while
the stream load method was more effective at reducing pollutants at the watershed outlet. The
results of this study indicate that emphasis should be placed on selection of the proper targeting
method and BMP to meet the needs and goals of a BMP implementation project, because
different targeting methods produce varying results.

4.2 INTRODUCTION
Maintaining proper water quality conditions is important to protect human, animal, and plant
health and is an ongoing concern in water resources (Pejman et al., 2009). Overall, the quality of
water bodies depends on natural processes such as precipitation rate, infiltration, and weathering
processes. Meanwhile, human activities such as urbanization and agricultural practices disturb
natural processes and ultimately affect water quality (Nouri et al., 2008). Operations such as
improper application of fertilizer, pesticides, animal wastes, irrigation water, and frequent
plowing can elevate the concentration of sediments, nutrients, and fecal bacteria in receiving
waters. Increased nutrient concentration leads to eutrophication in water bodies, which is harmful
to aquatic organisms (TNRCC, 1999). Activities such as nutrient enrichment cause adverse
effects on water quality and lead to increases in toxic substances, reduction of available aquatic
habitat, and decreases in overall values of human uses (Wang et al., 2007). Therefore, it is
important to restore and protect water quality through mitigation of negative impacts of human
disturbances.

45

In 2009, the USEPA National Water Quality Inventory Report determined that 44% of rivers and
streams and 64% of lakes and reservoirs in the United States are impaired; these water bodies did
not meet water quality standards for designated uses such as fishing and swimming. Water
quality impairment in rivers and lakes can be attributed to point and non-point sources NPS of
pollution. NPS pollution is particularly difficult to control due to its diffusive nature.
Agricultural activities are a prominent source of NPS pollution and, therefore, are a major
contributor to the degradation of water quality (USEPA 2009).

The management of NPS pollution requires a strategic combination of practices to prevent their
entry into receiving water bodies. BMPs are widely accepted methods that minimize the impact
of agricultural activities on both surface water and groundwater (Arabi et al. 2007). However,
pollutant reduction efficiencies of BMPs fluctuate due to varying design methods,
implementation, and maintenance frequency. Consequently, a thorough understanding of BMP
mechanisms in pollution mitigation and uncertainty in BMP effectiveness are needed during the
BMP selection process. Apart from BMP selection, placement in the watershed also plays a vital
role in the pollution reduction as the contribution of pollutants is disproportionate in the
watershed (Maringanti et al. 2009). This means that potential BMP effectiveness is site specific.
Therefore, an effective BMP implementation strategy for one site may or may not be useful in
reducing and/or controlling pollution for other sites in a watershed (Tuppad and Srinivasan,
2008).

Measuring pollution loads from all fields in a watershed and evaluation of BMP effectiveness
through actual implementation at the field level is time consuming, expensive, resource

46

intensive, and impractical. However, watershed/water quality models are efficient and provide
accurate information needed for evaluating pollution loads and BMP implementation strategies
at the field and watershed levels. Using watershed/water quality models allows for the
identification of CSAs, which are locations that contribute significantly high pollution load per
unit area. Using CSAs to prioritize placement of BMPs is called the targeting approach, which
provides greater reduction of pollutants. Targeting CSAs in the watershed is a well-known
procedure for implementing BMPs to control NSP and to improve environmental quality (Qiu
2009; Gitau et al. 2004; Srinivasan et al. 2005; Tripathi et al. 2003; Yang et al. 2005). However,
the comparison of different targeting techniques in identifying CSAs and the overall impacts of
these techniques to reduce NPS pollution at both the field and watershed levels are yet to be
determined.

Among existing watershed/water quality models, SWAT has been widely used to evaluate the
water quality impacts of different land use changes at watershed scale (Arnold et al. 1998;
Gassman et al. 2007, Arabi et al. 2007). The SWAT model is capable of simulating various
agricultural management practices such as tillage operations, fertilizer and pesticide applications,
vegetative filter strips, and crop rotations which makes it an ideal model for evaluation of
agricultural watersheds. For this reason, several studies have used SWAT to develop BMPs
implementation strategies in conjunction with various targeting methods (Jha et al. 2010;
Maringanti et al. 2009; Parajuli et al. 2008; Schilling and Wolter 2009; Tuppad et al. 2010;
White et al. 2009). Srinivasan et al. (2005) used SWAT to identify critical source runoff areas for
phosphorus transport and compared the results with the Soil Moisture Distribution and Routing
(SMDR) physically based model. Overall, it was determined that SWAT performed better than

47

SMDR. Jha et al. (2010) studied the impacts of land use restoration to 1990 conditions and land
use conversion in the CSAs (defined as highly erodible land areas, floodplain areas, and upper
subbasin areas) to native grass in order to assess the effect of nitrate load reduction strategies in
an Iowa agricultural watershed. Nitrate load reduction was determined to be 7% for the land use
restoration and 47%, 16%, and 8% for the land use conversions in the highly erodible lands,
upper subbasin areas, and floodplain areas, respectively. Tuppad et al. (2010) implemented
various BMPs (reduced tillage, edge of field vegetative filter strips, and contoured terraced) on
10%, 26%, 52%, and 100% of total targeted cropland and compared the pollutant reduction
efficiency at the outlet of the watershed using targeting and random placement. The results
demonstrated that the targeting method is more effective than the random placement method. In
both the Jha et al. (2010) and Tuppad et al. (2010) studies, CSAs were identified based on a total
load per unit area at the subbasin basis. White et al. (2009) used SWAT to identify CSAs and
quantify sediment and total phosphorus loads generated from five watersheds in Oklahoma. The
identification of CSAs was based on the threshold unit area load at each hydrologic response unit
(HRU). The HRUs were ranked based on sediment and phosphorus yields and the highest
ranking fractions were defined as CSAs. They found that only 5% of agricultural land produced
approximately 22% of sediment and phosphorus load. Schilling and Wolter (2009) used SWAT
to evaluate nitrate load reduction in Des Moines River in Iowa using four targeting methods. All
targeting methods were based on CSAs that have the potential to generate greater than 15 kg/ha
nitrate annually. Four different configurations were identified: all subbasins with the above
criteria, only CSA subbasins within the Boon River basin, targeting CSA subbasins closer to the
Des Moines Water Works, and targeting CSAs subbasins away from the Des Moines Water
Works. Results showed that 95% of total nitrate originated from nonpoint sources and the

48

greatest nitrate reduction was found when fertilizer application was reduced in subbasins closer
to the watershed outlet. However, in all of the targeting strategies the fertilizer application rate
was reduced assuming that the difference of fertilizer application rate compared to the base
scenario would be compensated by different BMPs. Diebel et al. (2008) introduced four
allocation

approaches

(aggregated/targeted,

aggregated/random,

dispersed/targeted,

and

dispersed/random) for implementing BMPs at both field and watershed scales. In the
aggregated/targeted allocation approach fields were selected according to the descending order
of phosphorus load contribution to the watershed, whereas in the aggregated/random approach
fields producing highest phosphorus load were selected randomly until the desired application
area was reached. In the dispersed/random approach, fields were selected randomly without
having any criteria while in dispersed/targeted approach fields were selected based on the
descending order of phosphorus load production without consideration of watershed
membership. The allocation approaches were evaluated by two methods: modeled pollutant
reduction index and water quality change index. The modeled pollutant reduction index was the
proportion of phosphorus load reduced after BMP application, while the water quality change
index was the proportion of the watershed observing significant reduction of stream phosphorus
concentration. For both methods, the targeted approach performed better than the random
approach.

As it was discussed above, some studies exist that relate the effectiveness of targeting methods
and BMP implementation strategies to environmental health and water quality improvement.
However, these methods have not been comprehensively evaluated and compared for multiple
pollutants. The objectives of this research are to (1) identify CSAs using multiple targeting

49

techniques and pollutants, (2) assess the sensitivity of BMPs to different targeting methods using
SWAT, and (3) evaluate the impact of BMP application in CSAs at subbasin and watershed
scales modeled using SWAT. The results of this study will aid policymakers and stakeholders in
making informed decisions regarding BMP placement while maximizing the environmental
benefits at a lower cost than current approaches.

4.3 MATERIALS AND METHODS
4.3.1 Study Area
The Saginaw River Watershed (SRW) (hydrologic unit code-HUC 040802) located in eastcentral Michigan was selected for this study. The SRW consists of six subwatersheds:
Tittabawassee (HUC 04080201), Pine (HUC 04080202), Shiawassee (HUC 04080203), Flint
(HUC 04080204), Cass (HUC 04080205), and Saginaw (HUC 04080206) (Figure 4-1). The
2

Saginaw River flows north towards Lake Huron. The total watershed area covers 22,260 km , of
which 42% is forest, 23% is agriculture, 17% is pasture, 11% is wetlands, and the remaining is
urban. Dominant agricultural crops in the watershed are corn and soybean. Expansive wetland
areas provide habitat to large populations of wildlife species. Average watershed elevation is 242
m above mean sea level, while the minimum elevation is 177 m and the maximum elevation is
457 m.

50

Tittabawassee

Pine

Saginaw
Cass
Shiawassee
Flint

N

0 10

20 40 60

Kilometers

Figure 4-1. Saginaw River Watershed. For interpretation of the references to color in this
and all other figures, the reader is referred to the electronic version of this thesis.

51

4.3.2 Model Description
Watershed/water quality models are useful tools to assess the effectiveness of BMPs on the
watershed scale (Woznicki et al. 2011). SWAT was selected to evaluate CSAs for sediment, TN,
and TP. SWAT is a physically based, spatially distributed watershed scale model developed by
the USDA-ARS (Arnold et al. 1998; Neitsch et al. 2005; Gassman et al. 2007). In SWAT, a
watershed is divided into subbasins and further divided into hydrologic response units (HRUs)
based on homogeneous land use, soil, slope, and management practices. The major components
of the model consist of weather, hydrology, soil characteristics, plant growth, nutrients,
pesticides, and land management practices (Gassman et al. 2007). Runoff volume in SWAT is
calculated either by the SCS curve number or Green and Ampt infiltration method (Neitsch et al.
2005).

Soil erosion comprises of three processes (detachment, transport, and deposition/degradation)
and is caused by two forces: raindrop impact and surface runoff. SWAT uses the Modified
Universal Soil Loss Equation (MUSLE) to calculate erosion and sediment yield for each
hydrologic response unit (HRU) within the watershed. In MUSLE, the average annual gross
erosion is calculated as a function of runoff (where runoff is the function antecedent moisture
condition and rainfall energy). Sediment yield in MUSLE is also function of peak runoff rate,
HRU area, soil erodibility, land cover, topography, and percent coarse fragments of soil (Neitsch
et al., 2005). Sediment transport in the channel depends on deposition and degradation. When
sediment concentration in the reach is exceeded the maximum sediment carrying concentration
of the reach, deposition occurs, while when the reverse is true degradation occurs. The maximum

52

amount of sediment that can be transported from a reach segment is a function of peak channel
velocity (Neitsch et al. 2005).

Nitrogen transport from overland areas into the stream is estimated in different ways in SWAT.
The nitrogen in the main channel is transported by surface runoff and lateral subsurface flow.
SWAT calculates nitrogen in the soil profile and the shallow aquifer. Two forms of nitrogen
(organic and inorganic) in the soil are computed by SWAT. The organic forms of nitrogen
consist of fresh organic nitrogen (crop residue and microbial mass), active organic nitrogen, and
stable organic nitrogen. The active and stable organic nitrogen are related to soil humus. The
+

inorganic form of nitrogen comprises of NH4

-

and NO3 . In SWAT, the organic nitrogen

associated with sediment is calculated as a function of concentration of organic nitrogen in the
top 10 mm of soil, sediment yield on a given day, and nitrogen enrichment ratio. The nitrate in
runoff, lateral flow, and percolation is calculated as the product of runoff volume and nitrate
concentration in the soil layer.

SWAT tracks both mineral and organic forms of phosphorus. Organic phosphorus consists of
fresh organic phosphorus and active and stable organic phosphorus. Fresh organic phosphorus is
related to crop residue and microbial biomass whereas active and stable organic phosphorus
connected to soil humus. Organic and mineral phosphorus attached to sediment are estimated as
a function of concentration of phosphorus attached to sediment in top 10 mm of soil, sediment
yield on a given day, and phosphorus enrichment ratio. The amount of soluble phosphorus is
calculated as a function of solution phosphorous concentration, runoff volume, and a partitioning
factor. The movement of phosphorus in soil is primarily driven by diffusion, which is based on
53

concentration gradient. Nutrient routing (nitrogen and phosphorus) in SWAT is calculated
through the use of the nitrogen and phosphorus cycle.

4.3.3 Data Sources
The SWAT model requires different types of physiographic data such as topography, land use,
soil, and stream network. Topography data was obtained from the Better Assessment Science
Integrating point and nonpoint Sources (BASINS) program in the form of a digital elevation
model (DEM). For land use representation the 2008 Cropland Data Layer (CDL) was obtained
from the USDA National Agricultural Statistics Service (NASS, 2008). Watershed soil
characteristics were defined using the State Soil Geographic Database (STATSGO).

The

STATSGO dataset was developed by the National Cooperative Soil Survey at a scale of
1:250,000 and was linked to tabular data containing soil chemical and physical properties
(Muttiah and Wurbs, 2002). Stream network was defined using the United States Geological
Survey (USGS) National Hydrography Dataset (NHD). The NHD dataset was used to improve
hydrologic segmentation and sub-watershed boundary delineation (Winchell et al. 2007).

Daily streamflow data from USGS gauging station 04157000 was used for streamflow
calibration and validation. Water quality calibration and validation was performed using data
from Michigan Department of Environmental Quality (MDEQ) station 090177. The USGS Load
Estimator (LOADEST) was used to convert the daily observed water quality (sediment, TN, and
TP) data to monthly loads. Twenty years (1990-2009) of observed daily precipitation and
temperature data was obtained from the National Climatic Data Center (NCDC). In this study 19
precipitation stations and 11 temperature stations were represented within the SRW. The
54

remaining required meteorological data (wind speed, relative humidity, and solar radiation) were
estimated using the SWAT weather generator program.

The SRW land use is 23% agricultural, with corn and soybean being the predominant crops. To
more accurately assess the fate and transport of sediment and nutrients in the watershed,
management operations were developed based on common agricultural practices in the region,
including tillage and fertilizer applications. Regarding crop rotations, the continuous corn
rotation is six years in length, where corn is planted for five years and soybeans are planted in
the final year. The continuous soybean rotation is three years in length and contains two years of
soybean planting and a final year of corn planting. A detailed description of tillage operations
and fertilizer applications are provided in tables 4-1 and 4-2.

55

Table 4-1. Continuous corn conventional tillage management operations.
Date

Practice

SWAT Practice

1-May

Soil Finish
Nitrogen Application
(Urea)
Soil Finish

Urea

Year

Field Cultivator > 15 ft

4-May

Application
Rate

4-May
5-May
5-May

Phosphorus Application
(P2O5)
Plant Corn Seed

5-May

1-5
194 kg/ha

1-5

Field Cultivator > 15 ft
Elemental Phosphorus

1-5
59 kg/ha

1-5

Bicep II Magnum (PRE)

Plant/Begin Growing
Season
Atrazine

1.39 kg/ha

1-5

5-May

Bicep II Magnum (PRE)

Metolachlor

1 kg/ha

1-5

1-Nov

Combine Harvest Corn
Grain
Fall Chisel

Harvest and kill

1-5

Coulter-Chisel Plow

1-5

Soil Finish

Field Cultivator > 15 ft

Phosphorus Application
(P2O5)
Soil Finish

Elemental Phosphorus

15Nov
14May
14May
14May
15May
7-June
1-Oct
30-Oct

Plant Soybean Seed
Spay Roundup
Weathermax
Combine Harvest Soybean
Grain
Fall Chisel

1-5

6
45 kg/ha

6

Field Cultivator > 15 ft

6

Plant/Begin Growing
Season
Glyphosate Amine

6
0.87 kg/ha

6

Harvest and kill

6

Coulter-Chisel Plow

6

56

Table 4-2. Continuous soybean conventional tillage management operations.
Date

Practice

SWAT Practice

14-May
14-May

Soil Finish
Phosphorus Application
(P2O5)
Soil Finish

Field Cultivator >15 ft
Elemental Phosphorus

14-May

Application
Rate

Year
1-2

45 kg/ha

1
1-2

Field Cultivator >15 ft

1
1-2

15-May

Plant Soybean Seed

7-June

30-Oct

Spray Roundup
Weathermax
Combine Harvest
Soybean Grain
Fall Chisel

1-May

Soil Finish

1-Oct

4-May
4-May
5-May
5-May
5-May
5-May
1-Nov
15-Nov

Nitrogen Application
(Urea)
Soil Finish
Phosphorus
Application (P2O5)
Plant Corn Seed
Bicep II
Magnum (PRE)
Bicep II
Magnum (PRE)
Combine Harvest
Corn Grain
Fall Chisel

Plant/Begin Growing
Season
Glyphosate Amine

1
1-2
0.87 kg/ha

1
1-2

Harvest and kill

1
1-2

Coulter-Chisel
Plow
Field Cultivator
>15 ft
Urea

1
1-2
3
3
194 kg/ha

3
3

Field Cultivator >
15 ft
Elemental
Phosphorus
Plant/Begin
Growing Season
Atrazine

3
3
59 kg/ha

3
3
3
3

1.39 kg/ha

3
3

Metolachlor

1kg/ha

3
3

Harvest and kill

3
3

Coulter-Chisel
Plow

57

3
3

4.3.4 Sensitivity Analysis and Calibration Process
Sensitivity analysis is conducted to determine the most influential parameters on model output to
be used in model calibration by providing a rank of parameters sensitive to model outputs. The
SWAT model uses Latin Hypercube One factor-at-a-time (LH-OAT) sampling to perform the
sensitivity analysis (van Griensven et al. 2006). A sensitivity analysis was performed in this
study to evaluate the model parameters sensitive to streamflow, sediment, TN, and TP.

Calibration is an iterative process of adjusting model input parameters that compares simulated
and observed data of interest. It plays an important role in watershed modeling through reducing
uncertainty in model prediction. This process consists of sensitivity analysis followed by manual
and automatic calibration In automatic calibration, model selects the parameter for the
calibration process whereas in manual calibration, user select the parameter for calibration
process. The most sensitive parameters are used to perform model calibration. In order to
determine model reliability, validation is performed for the time period following the calibration
period without implementation of BMPs. (Woznicki et al. 2011).

Calibration and validation were performed on monthly a time step for streamflow, sediment, TN,
and TP. A warm up period of two years was used to initialize model parameters. Ideal calibration
and validation consists of three to five years of data (Moriasi et al. 2007). In this study, the
calibration period was from 2002-2003 and the validation period was from 2004-2005. When a
calibrated model is used to simulate multiple processes such as streamflow, sediment, and
nutrients, two or more model evaluation statistics are required to assure model reliability in
addressing different processes (Balascio et al. 1998). Three statistical methods were used to
58

evaluate the prediction of calibrated model: Nash-Sutcliffe efficiency (NSE) (Nash and Sutcliffe,
2

1970), coefficient of determination (R ), and percent bias (PBIAS). On a monthly time step, the
model performance is classified as satisfactory if the NSE value greater than 0.50 and the PBIAS
value for streamflow remains between ±25, sediment between ±55, and TN and TP between ±70
2

(Moriasi et al. 2007). Values of R greater than 0.5 are considered acceptable for a monthly time
step (Santhi et al. 2001).

NSE describes the fitting of observed and simulated data in 1:1 line (Moriasi et al. 2007). The
range of NSE varies between negative infinity to 1, where 1 is the optimal value. The calculation
of NSE is presented in equation 4-1:
n
∑ Y obs
i
i =1

(

NSE = 1

n
∑ Y obs
i
i =1

(

obs

Where, Yi

Yisim

)2
(4-1)

)

2
Y obs,mean
sim

is the ith observed value of the constituent, Yi
obs,mean

constituent, and Y

is the ith simulated value of the

is the mean observed data of the constituent. The NSE value is

calculated as the sum of squared values of the difference between observed and predicted values,
resulting in strong overestimation for the larger values while neglecting smaller values. This
leads to an overestimation of model performance during peak flows and underestimation during
low flow periods (Krause et al. 2005).

59

The coefficient of determination describes the degree of collinearity between observed and
2

predicted data (Moriasi et al. 2007). The range of R varies from 0 to 1 with 1 being the optimal
2

value. R is calculated using equation 4-2.

n
∑ Y obs
i
i =1

(

R2 =

n
∑ Y obs
i
i =1

(

)(

Y obs, mean Yisim

Y sim, mean

n
∑ Y sim
i
i =1

)

(

2
Y obs, mean

)
(4-2)

)

2
Y sim, mean

PBIAS depicts the tendency of the simulated data to be larger or smaller compared to the
observed data (Moriasi et al. 2007). A positive value of PBIAS indicates model underestimation
and a negative value reflects overestimation, while zero is the optimal value (Moriasi et al.
2007). PBIAS is calculated using equation 4-3.

n
∑ Y obs Ysim
i
i
PBIAS = i =1
n
∑ Y obs
i
i =1

(

)

100

(3)

4.3.5 Best Management Practices in SWAT
In order to assess the impact of BMP effectiveness on pollution reduction, ten BMPs were
modeld using SWAT on agricultural lands in the SRW. Each BMP was evaluated compared to a
base scenario in which no BMPs are implemented. The ten BMPs selected were: contour farming
(CF), terraces (T), recharge structures (RS), conservation tillage (CT), no tillage (NT), native
60

grass (NG), residue management (0 kg/ha) (RM 0), residue management (1000 kg/ha) (RM
1000), residue management (2000 kg/ha) (RM 2000), and strip cropping (SC). Implementation
procedures for BMPs were collected from various published literatures and implemented in
SWAT.

4.3.5.1 Contour Farming
Contour farming consists of ridges and furrows constructed by tillage, planting, and other
operations, which creates numerous small dams (USDA-NRCS, 2005). This BMP slows down
runoff, increases infiltration, and thereby reduces the soil erosion. To implement this practice in
SWAT, curve number (CN2) was reduced by three and the Universal Soil Loss Equation
Practice factor (USLE_P) was adjusted to 0.6, 0.5, and 0.55 for slopes of 0-2%, 2-5%, and 510%, respectively (Arabi et al.2007; Tuppad and Srinivasan, 2008).

4.3.5.2 Terraces
Terraces are earth embankments or a combination of ridge and channel constructed across the
slope (USDS_NRCS, 2005). This BMP reduces long slopes and serves as a small dam, guiding
water to an outlet. Terraces reduce surface runoff by holding water in small depressions and
decreases peak runoff by reducing the hillside slope (Arabi et al. 2007). Incorporation in SWAT
is accomplished by reducing CN2 by five and adjusting USLE_P to 0.12, 0.1, and 0.11 for slopes
of 0-2%, 2-5%, and 5-10%, respectively (Arabi et al.2007; Tuppad et al. 2010).

4.3.5.3 Recharge Structures
Recharge structures are small dams in the channel designed to capture a portion of flowing water
(Tuppad and Srinivasan, 2008). This BMP increases infiltration and percolation of water while
61

reducing stream energy, resulting in less sediment carrying capacity. To simulate recharge
structures in SWAT the effective hydraulic conductivity (CH_K1) value is replaced by 25 mm/hr
(Tuppad and Srinivasan, 2008).

4.3.5.4 Conservation Tillage
Conservation tillage involves reducing tillage operations and soil disturbances when compared to
conventional tillage (Tuppad and Srinivasan, 2008). Crop residue on the surface is left in place,
which acts as ground cover and prevents soil erosion. Detailed operations of continuous soybean
conservation tillage schedules are provided in Table 4-3.
Table 4-3. Continuous soybean conservation tillage operations.
Date

Practice

SWAT Practice

14-May

Phosphorus Application
(P2O5)
Soil Finish
No Till Planting
Plant Soybean Seed

Elemental Phosphorus

14-May
15-May
15-May
7-Jun
1-Oct
4-May
4-May
5-May
5-May
5-May
5-May
5-May
1-Nov

Spay Roundup
Weathermax
Combine Harvest
Soybean Grain
Nitrogen
Application(Urea)
Soil Finish
Phosphorus Application
(P2O5)
No Till Planting
Plant Corn Seed

Field Cultivator > 15 ft
Generic no Till Mixing
Plant/Begin Growing
Season
Glyphosate Amine

Application
Rate
45 kg/ha

Year
1-2
1-2
1-2
1-2

0.87 kg/ha

Harvest and kill

1-2
1-2

Urea

194 kg/ha

3

Field Cultivator > 15 ft
Elemental Phosphorus

59 kg/ha

3
3

Generic no Till Mixing
Plant/Begin Growing
Season
Bicep II Magnum (PRE)
Atrazine
Bicep II Magnum (PRE)
Metolachlor
Combine Harvest Corn
Harvest and kill
Grain
62

3
3
1.39 kg/ha
1 kg/ha

3
3
3

4.3.5.5 No Tillage
No tillage operations leave crop residue on the field and the soil is kept undisturbed from the
time of harvest until planting (USDA, 2010). Minimum soil disturbances are achieved during
nutrient application and planting of crops. The objective of no tillage farming is to increase soil
moisture while reducing soil erosion. Detailed operations of continuous soybean no tillage
schedules are provided in Table 4-4.
Table 4-4. Continuous soybean no tillage operations.
Date

Practice

SWAT Practice

14-May

Elemental Phosphorus

15-May

Phosphorus Application
(P2O5)
No Till Planting

15-May

Plant Soybean Seed

7-Jun
1-Oct

Spay Roundup Weathermax
Combine Harvest Soybean
Grain
Nitrogen Application (Urea)
Phosphorus Application
(P2O5)
No Till Planting

4-May
5-May
5-May
5-May
5-May
5-May
1-Nov

Generic no Till
Mixing
Plant/Begin Growing
Season
Glyphosate Amine
Harvest and kill
Urea
Elemental Phosphorus

Generic no Till
Mixing
Plant Corn Seed
Plant/Begin Growing
Season
Bicep II Magnum (PRE)
Atrazine
Bicep II Magnum (PRE)
Metolachlor
Combine Harvest Corn Grain Harvest and kill

Application
Rate
45 kg/ha

Year
1-2
1-2
1-2

0.87 kg/ha

1-2
1-2

194 kg/ha
59 kg/ha

3
3
3
3

1.39 kg/ha
1 kg/ha

3
3
3

4.3.5.6 Native Grass
Native grass planting involves replacing agricultural row crops with native grasses such as
Indian switchgrass and big bluestem (Nejadhashemi and Mankin, 2007). Sediment and nutrient
transport in runoff are reduced by the dense vegetative cover combined with the elimination of
63

tillage practices and fertilizer applications. Native grass was implemented in SWAT by
converting all agricultural row cropland to range grass (Woznicki et al. 2011).

4.3.5.7 Residue Management - 0 kg/ha
Residue management involves controlling the amount and distribution of crop residue on the soil
surface (USDA, 1996). This management practice reduces sheet and rill erosion by providing
soil cover. Representation of residue management within SWAT is accomplished by applying
no till operations on agricultural land while adjusting CN2, USLE_P, Universal Soil Loss
Equation Cover factor (USLE_C), and Manning’s n value for overland flow (OV_N) (Arabi et
al. 2007). These values vary depending on the amount of residue left on the field in kg/ha (0,
1000, and 2000), and are presented in Table 4-5.
Table 4-5. SWAT Inputs for Residue Management.
Residue (kg/ha)
0
1000
2000

CN2
-2
-2
-2

USLE_P
1
0.39
0.29

USLE_C0
0.2
0.2
0.2

OV_N
0.14
0.2
0.3

4.3.5.8 Strip Cropping
Strip cropping involves growing planned rotations of row crops in a systematic arrangement of
equal width strips across a field (USDA-NRCS, 2011). This BMP serves as a vegetative barrier
and reduces soil erosion due to wind and water. Strip cropping is represented in SWAT by
reducing the CN2 by three and adjusting USLE_P to 0.3, 0.25, and 0.27 for slopes of 0-2 %, 25%, and 5-10%, respectively (Arabi et al. 2007).

64

4.3.6 Spatial Targeting Methods
Establishing BMPs throughout the watershed is impractical and expensive. Therefore,
identifying CSAs of pollutants in the watershed should be the primary step before BMP
implementation. Four targeting techniques, namely, Concentration Impact Index (CII), Load
Impact Index (LII), Load per Subbasin Area Index (LPSAI), and Load per Unit Area Index
(LPUAI) were analyzed to prioritize BMP placement within the watershed. Each targeting
method was applied for sediment, TN, and TP, creating three sub-targeting methods (sedimentbased, TN-based, and TP-based) for a total of 12 targeting scenarios. Then, subwatersheds were
categorized into high, medium, and low priority areas based on the natural breaks method of
classification. Based on natural statistical groupings in the dataset, different classes are formed in
natural breaks classification. Data having similar values are put together into groups while trying
to maximize the difference between the groups using geographical information system.
Therefore, a relatively substantial difference is found between the data values of any two groups.
In addition, although each targeting strategy is always based on one pollutant (e.g. sediment), the
impacts of BMPs implementation on other contaminants (e.g. TN and TP) were also analyzed.

4.3.6.1 Concentration Impact Index
The CII method identifies high priority areas based on the pollutant concentration level in the
subwatershed reaches (Tuppad and Srinivasan, 2008). This method considers pollutants from the
adjacent subwatershed as well as the upstream. It is effective in addressing localized pollution in
low and high flows, especially concerning aquatic health.

65

4.3.6.2 Load Impact Index
The LII method identifies high priority areas based on the total pollutant load from the reach
(Tuppad and Srinivasan, 2008). This method considers pollutant load from both the
subwatershed and the upstream subwatersheds. LII represents the cumulative pollutant load up to
a point of interest, such as a water treatment plant.

4.3.6.3 Load per Subbasin Area Index
The LPSAI index method identifies high priority areas based on the pollution load for each
subbasin. Subbasin area was multiplied by load per unit area to calculate total pollutant load per
subbasin. This method is also effective in identifying local concerns by identifying the subbasins
with the largest amount of pollutant discharge. This can lead to aquatic health concerns since
large amounts of pollution load may enter small streams.

4.3.6.3 Load per Unit Area Index (LPUAI)
The LPUAI method identifies the high priority areas based on average pollutant load per unit
area from each subbasin (Tuppad and Srinivasan, 2008). As this method portrays load within
individual subwatershed, it is applicable to identify local concerns within the subbasin.

4.3.7 BMP Relative Sensitivity Index
Relative sensitivity index (RSI) was calculated for each targeting method using equation 4-4.
RSI =

S base
R base

S BMP R base
×
R BMP S base

(4-4)

66

where, Sbase is the total subbasin load for the no BMP scenario, SBMP is the total subbasin loads
after BMPs application to CSAs (high, medium, or low priorities), Rbase is the reach load at the
outlet of the watershed for the no BMP scenario, and RBMP is the reach load at the outlet after
application of BMPs to the targeted areas (high, medium, or low priorities). The RSI may vary
from -∞ to + ∞. Positive relative sensitivity index indicates that the BMP implementation
strategy was effective in pollution reduction both at the outlet and subbasin level while negative
relative sensitivity index indicates that BMPs implementation strategy was unsuccessful in
pollution mitigation. Meanwhile, RSI = 0 represents no impact due to BMPs implementation and
RSI = 1 demonstrates an equal pollution reduction rate at subbasins and the outlet. If 0 < RSI <1
then the BMP implementation strategy is more effective at pollution reduction at the watershed
outlet than at the subbasin level, while for RSI > 1 the reverse is true. In this study, the relative
sensitivity index was calculated for sediment, TN, and TP under each targeting method.

4.4 RESULTS AND DISCUSSIONS
4.4.1 Sensitivity Analysis
Sensitivity analysis was performed for flow, sediment, TN, and TP of the SRW. The ten most
sensitive parameters for flow, sediment, TN, and TP are presented in Table 4-6. In general, CN2
and Alpha_Bf (baseflow recession constant) are highly sensitive for all constituents. Parameters
such as Spcon (linear re-entrainment parameter for channel sediment routing) and Spexp
(exponential re-entrainment parameter for channel sediment routing) are highly sensitive for
sediment but are not sensitive for the other constituents. For flow, Cn2, Alpha_Bf, Rchrg_Dp
(fraction of percolation from root zone that recharges deep aquifer), Esco (soil evaporation
67

compensation factor), and Timp (snow pack temperature lag factor) were the most sensitive
whereas for sediment Spcon, Ch_N2 (Manning’s n value for main channel), Cn2, Spexp, and
Usle_P were the most sensitive. Sensitive parameters for TN and TP were similar: Cn2,
Alpha_Bf, and Timp. Sensitivity analysis results were taken into consideration when performing
model calibration.
Table 4-6. Sensitivity analysis results for flow, sediment, TN, and TP.
Rank
1
2
3
4
5
6
7
8
9
10

Flow
Cn2
Alpha_Bf
Rchrg_Dp
Esco
Timp
Gwqmn
Sol_Awc
Canmx
Sol_Z
Ch_K2

Sediment
Spcon
Ch_N2
Cn2
Spexp
Usle_P
Alpha_Bf
Ch_K2
Timp
Esco
Surlag

TN
Cn2
Alpha_Bf
Canmx
Timp
Sol_Awc
Nperco
Blai
Ch_K2
Rchrg_Dp
USLE_P

TP
Cn2
Alpha_Bf
Ch_K2
Timp
USLE_P
Sol_Awc
Canmx
Surlag
Blai
Smtmp

4.4.2 Model Calibration
Calibration of the model was performed for flow, sediment, TN, and TP on a monthly basis in
the SRW. The calibration, validation, and overall combined calibration/validation results are
presented in Table 4-7. According to guidelines developed by Moriasi et al. (2007) the flow,
sediment, and TP calibrations were considered ‘good’, while the TN calibration was considered
‘satisfactory’ for this study. The flow calibration/validation hydrograph is presented in Figure 42. Peaks and baseflow are generally well represented by the calibrated model, although the
model under-predicts for one peak in the validation period.

68

Table 4-7. SRW calibration and validation results.
Constituent

Statistic

Sediment

TP

TN

Validation

Overall

NSE
2
R
PBAIS (%)
NSE
2
R
PBAIS (%)
NSE
2
R
PBAIS (%)
NSE
2
R
PBAIS (%)

Flow

Calibration
0.77
0.78

0.71
0.71

0.74
0.74

7.33
0.55
0.69

0.73
0.74
0.79

3.47
0.72
0.77

-47.81
0.57
0.65

-30.68
0.78
0.80

-36.66
0.76
0.78

-15.22
0.76
0.82

-18.12
0.41
0.81

-16.99
0.53
0.719

24.32

50.53

39.36

600
Calibrated

Flow (cms)

500

Observed

400

300
200
100
0
11-2001

03-2003

08-2004
Date

Figure 4-2. Flow calibration and validation hydrograph.

69

12-2005

04-2007

4.4.3 Spatial Targeting Methods
Four spatial targeting methods (CII, LII, LPSAI, and LPUAI) were compared in the SRW to
prioritize the subbasins into three classes (high, medium, and low priority area) according to the
pollution of interest (sediment, TN, and TP). A detailed description of the results of each method
is presented below.

4.4.3.1 CII
The location prioritization in the CII method is based on the pollution concentration in the reach
(Figure 3). The watershed was divided into high, medium, and low priority areas and the
sediment concentration range for high, medium, and low priority area was 0.0 to 86.4 mg/l, 86.4
to 259.0 mg/l, and 259.0 to 515.0 mg/l, respectively. For the sediment targeting scenario (Figure
4-3a), the high priority areas were located on the upstream sections of the watershed and the
main source of pollution was from agricultural land. The TN concentration range was 0.0 to 1.7
mg/l, 1.7 to 4.8 mg/l, 4.8 to 10.8 mg/l for high, medium, and low priority area whereas the TP
concentration range was 0.0 to 0.3 mg/l, 0.3 to 0.8 mg/l, and 0.8 to 1.5 mg/l for high, medium,
and low priority areas, respectively. Both TN and TP high priority area were located upstream,
center, and also near the outlet of the watershed (Figure 4-3b) and (Figure 4-3c), respectively.
Similar to the sediment high priority areas, the main source of pollution in high priority areas of
TN and TP was due to the agricultural lands. A greater number of subbasins were identified as
high priority area according to TN and TP targeting. Additionally, a greater number of subbasins
were categorized into medium priority areas for TN targeting, followed by sediment and TP
targeting methods, respectively. In all three cases, most subbasins were selected as low priority.

70

(a)

N

Legend
Sediment concentration (mg/l)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-3. (a) CII targeting method priority areas based on sediment concentration.

71

(b)

N

Legend
Total nitrogen concentration (mg/l)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-3. (b) CII targeting method priority areas based on TN concentration.

72

(c)

N

Legend
Total phosphorus concentration (mg/l)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-3. (c) CII targeting method priority areas based on TP concentration.

73

4.4.3.2 LII
Priority areas in this method were determined according to the pollutant load for each reach
(Figure 4-4). The sediment load range for high, medium, and low priority areas was 0 to 18300
tons, 18300 to 98500 tons, 98500 to 206000 tons, respectively for the sediment targeting
scenario. For the TN targeting scenario, the load ranges for high, medium, and low priority area
were 0 to 1000000 kg, 1000000 to 4190000 kg, and 4190000 to 8050000 kg, respectively
whereas for the TP targeting scenario, the load ranges were 0 to 46900 kg, 46900 to 223100 kg,
and 223100 to 585700 kg for high, medium, and low priority areas, respectively. An approximate
equal number of subbasins were categorized as high priority for all pollutants. High priority
areas were generally located near the outlet of the SRW, while medium priority subbasins were
also located within the general proximity of the watershed outlet. The large concentration of high
and medium priority areas near the outlet is likely due to the targeting methodology because
pollutant load from both the subwatershed and the entire upstream watershed are accumulated
near the outlet. Low priority areas are predominant for all three pollutants using the LII targeting
method, which is likely due to a large number of small tributaries in the SRW. Tributaries
generally have smaller flows when compared to the main channel, which leads to smaller load
carrying capacity and therefore low priority under this targeting method.

The LII targeting maps are very similar when performing a comparison between sediment, and
TP, which is likely due to the strong correlation between sediment and phosphorus transport as
phosphorus attaches to sediment. Conversely, there are a considerably smaller number of
medium priority areas for TN, indicating that most nitrogen is generated near the outlet of the
watershed. While this method has the advantage of considering pollution from both the subbasin

74

and upstream, it seems that the resulting priority areas are skewed toward being placed near the
watershed outlet.
(a)

N

Legend
Sediment (metric ton)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-4. (a) LII targeting method priority areas for sediment.

75

(b)

N

Legend
Total nitrogen (kg)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-4. (b) LII targeting method priority areas for TN.

76

(c)

N

Legend
Total phosphorus (kg)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-4. (c) LII targeting method priority areas for TP.

77

4.4.3.3 LPSAI
In this method the priority areas were identified based on the pollutant load from each subbasin
(yield multiplied by area), which primarily identifies the origin of a pollutant from its local
source (Figure 4-5). Based on the sediment targeting scenario, the sediment load varied from 0.0
to 8210 tons, 8210 to 25624 tons, and 25624 to 86539 tons for low, medium, and high priority
areas, respectively. In the TN targeting scenario, the TN load varied from 0 to 69200 kg, 69200
to 257000 kg, and 257000 to 1000000 kg for low, medium, and high priority areas, respectively,
whereas based on TP targeting scenario, the TP load varied from 0 to 8170 kg, 8170 to 26291 kg,
26291 to 108680 kg for low, medium, and high priority areas, respectively. The LPSAI method
generally identifies high and medium priority areas as subbasins that generate a relatively large
amount of pollution and are large in area because of the inherent areal dependence. Therefore,
most high and medium priority areas defined using this method are subbasins that predominantly
contain agricultural row crops and are large in size. Priority area identification using the LPSAI
method can be considered subjective, because it is highly dependent on watershed delineation
and DEM resolution for creation of subbasins in the model. One way to limit the subjectivity is
to ensure that all subbasins are of similar size when performing watershed delineation.

Targeting maps using the LPSAI method are very similar for each pollutant. High and medium
priority areas are generally scattered across the watershed for all three pollutants. The subbasins
that fell under high priority areas were almost the same for all targeting pollutants. A greater
number of subbasins were identified as medium priority for the TN targeting method, whereas a
similar number of subbasins were categorized as medium priority for both sediment and TP.

78

(a)

N

Legend
Sediment (metric ton)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-5. (a) LPSAI targeting method priority areas for sediment.

79

(b)

N

Legend
Total nitrogen (kg)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-5. (b) LPSAI targeting method priority areas for TN.

80

(c)

N

Legend
Total phosphorus (kg)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-5. (c) LPSAI targeting method priority areas for TP.

81

4.4.3.4 LPUAI
This targeting method is based on pollutant load per area, which normalizes each subbasin for
comparison (Figure 4-6). Based on sediment targeting scenario, the sediment yield varied from
0.0 to 0.8 (tons/ha), 0.8 to 2.8 tons/ha, and 2.8 to 6.9 tons/ha for low, medium, and high priority
areas, respectively. According to the TN targeting scenario, the TN yield varied from 0.0 to 8.9
kg/ha, 8.9 to 25.5 kg/ha, 25.5 to 73.4 kg/ha for low, medium, and high priority areas,
respectively, while based on the TP targeting scenario the TP yield varied from 0.0 to 1.1 kg/ha,
1.1 to 3.4 kg/ha, and 3.4 to 10.4 kg/ha for low, medium, and high priority areas, respectively.
High and medium priority areas identified with this method are strongly correlated with
agricultural land in the SRW. Subbasins that are predominantly agricultural were identified as
high and medium priority with this method. Compared to the LPSAI method, there are less high
and medium priority subbasins because some large subbasins were eliminated due to the areal
normalization.
Targeting maps were similar for all pollutants in this method. More subbasins were categorized
into high priority area for the sediment targeting method, followed by TN and TP. An equal
number of subbasins were identified as medium priority area for sediment, TN, and TP. Similar
to the other targeting methods, a majority of the SRW subbasins were characterized as low
priority.

82

(a)

N

Legend
Sediment yield (metric ton /ha)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-6. (a) LPUAI targeting method priority areas for sediment.

83

(b)

N

Legend
Total nitrogen yield (kg /ha)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-6. (b) LPUAI targeting method priority areas for TN.

84

(c)

N

Legend
Total phosphorus yield (kg /ha)
Low
Medium
High

0 10 20 40 60 80
Kilometers

Figure 4-6. (c) LPUAI targeting method priority areas for TP.

85

4.4.4 Comparison of Agricultural Lands in Sub-targeting Scenarios for Sediment, TN, and TP

The goal of this section is to quantify the capability of each targeting method to identify
agricultural lands as source of NPS. In addition, comparing the distribution of priority areas for
the entire SRW and its agricultural lands provides insight into what watershed characteristics
influence each method of identifying CSAs.

4.4.4.1 Sediment Targeting Scenario
A different proportion of high, medium, and low priority areas is obtained for the sediment
targeting scenario in each of the four targeting methods as presented in Figure 4-7.
2000000
Low
Medium
High

Area (ha)

1600000
1200000
800000
400000
0
CII Area

CII
LII Area
LII
LPSAI LPSAI LPUAI LPUAI
Ag.Land
Ag.Land Area Ag.Land Area Ag.Land

Figure 4-7. Distribution of high, medium, and low priority areas for sediment targeting
scenario.

For the whole watershed, the LPSAI method identified the largest high priority area (143349 ha),
while the LII method identified the smallest (9543 ha). The CII method identified the largest area
(448083 ha) under medium priority, while the LPSAI method had the smallest amount of

86

medium priority area (160973 ha). The CII method had the smallest low priority area (1050596
ha), while the other three targeting methods had approximately equal areas of low priority.

A similar trend was observed for agricultural lands as was observed for the whole SRW. The
LPSAI method had the greatest amount of agricultural lands (143318 ha) identified as high
priority, while the LII method did not identify any agricultural land as high priority. Therefore,
no BMPs were implemented for the LII high priority area in this study. The LPUAI method had
the highest amount of medium priority area (249368 ha) on agricultural land, while LPSAI had
the least (160939 ha). For low priority areas on agricultural land, the CII and LII targeting
methods had the most while LPUAI had the least (11077 ha).

The targeting methods based primarily on field generated pollutants (LPSAI and LPUAI) had
most high and medium priority areas attributed to agricultural lands. This is likely because
agricultural practices have a large impact on sediment generation. In addition, the LPSAI method
has a higher total high priority area because it does not normalize for subbasin area, unlike the
LPUAI method. Conversely, the targeting methods based primarily on the reach (CII and LII)
identified some high and medium priority areas not located in agricultural subbasins. In the case
of LII, this is likely because the high priority areas were located at the outlet (where
accumulation of upstream sediment occurs). For the CII method, tributaries may have high
sediment concentrations because of relatively low flows, creating medium and high priority areas
in these locations, regardless of land use.

87

4.4.4.2 TN Targeting Scenario
Proportions of high, medium, and low priority areas under the TN scenario are different for each
sub-targeting method for both the entire SRW and agricultural land in the SRW. Figure 4-8
presents the proportions of priority areas for each sub-targeting method.
2000000
Low
Medium
High

1600000

Area (ha)

1200000
800000
400000
0
CII Area

CII
LII Area
LII
LPSAI LPSAI LPUAI LPUAI
Ag.Land
Ag.Land Area Ag.Land Area Ag.Land

Figure 4-8. Distribution of high, medium, and low priority areas for TN targeting scenario.
Based on the TN targeting scenario for high priority areas in the whole watershed the CII method
had the largest area (289,157 ha), while the LII method had the smallest (9,543 ha). For medium
priority areas, the CII, LPSAI, and LPUAI methods had approximately equal areas, whereas the
LII targeting method had the least amount of area (157,566 ha). Meanwhile, the LII method had
the largest (1,359,231 ha) area of low priority, while CII had the smallest area (612,177 ha) of
low priority.

For agricultural lands, the CII method had the greatest amount of high priority areas (242,318
ha), while LII had none. The LPUAI method had the most agricultural land (2694656 ha) under

88

medium priority and LII had the least (33,865 ha) of all targeting methods. Under low priority,
LII had the greatest area (311,558 ha) while LPUAI had the smallest area (24,227 ha).

Similar to sediment targeting, the field based methods’ (LPSAI and LPUAI) high and medium
priority areas were generally all classified as agriculture. Meanwhile, the reach methods (CII and
LII) identified high and medium priority areas that were both agricultural and non-agricultural
lands. In addition, for CII, LPSAI, and LPUAI methods, 50% or less of the land was classified as
low priority, indicating that to target TN for medium and high priority areas would still be
resource and cost intensive. Once again, the LII method was mostly low priority, because
medium and high priority areas were located only near the outlet of the SRW.

4.4.4.3 TP Targeting Scenario
Similar to sediment and TN, quantity of priority areas are different using each of the four subtargeting methods. Proportions of high, medium, and low priority areas for the entire SRW and
agricultural lands of the SRW are presented in Figure 4-9.
2000000

Low
Medium
High

Area (ha)

1600000
1200000
800000
400000
0
CII Area

CII
LII Area
LII
LPSAI LPSAI LPUAI LPUAI
Ag.Land
Ag.Land Area Ag.Land Area Ag.Land

Figure 4-9. Distribution of high, medium, and low priority areas for TP targeting scenario.
89

When considering the entire SRW, CII has the greatest area (232,636 ha) of high priority, while
LII has the smallest area (9,853 ha) for TP. In the case of medium priority areas, LPUAI had the
greatest area (296,666 ha), while CII had the least (95,150 ha) of all targeting methods. The low
priority area was greatest (1,267,596 ha) for the LII method, while the other targeting methods
had approximately equal areas of low priority.
The area of agricultural lands attributed to high, medium, and low priorities for the four targeting
methods varies based on technique. The CII method had the largest (221,198 ha) amount of high
priority, while LII had no high priority for agricultural lands. The greatest amount of area
(291,776 ha) to fall under medium priority was LPUAI and the least (46,558 ha) was CII. In the
case of low priority the LII targeting method was had the greatest (219,175 ha), while LPUAI
had the least area (11,521 ha).

The TP targeting depicts similar trends to both sediment and TN. For the field based methods
(LPSAI and LPUAI), the high and medium priority areas are completely characterized as
agricultural lands. For the LII scenario, there is no high priority area classified as agriculture
because the high priority areas are located near the SRW outlet. The CII method reveals that
agricultural land is a prominent contributor to high and medium priority areas.

4.4.5 BMP Pollutant Reduction
The effectiveness of BMPs was determined in terms of percent reduction compared to the no
BMP scenario. The results are presented according to the reduction of pollutants (sediment, TN,
and TP) at the watershed outlet and subbasin by four targeting methods. In addition, BMP

90

effectiveness was compared to the effectiveness of the BMP after it is normalized by its
application area. The total BMPs application area of each targeting scenario was divided by the
amount of pollution reduction from the base scenario of each BMP in each targeting method in
order to normalize reduction by area. Normalization allows for an accurate comparison of the
targeting methods, because each method identifies different priority areas.

4.4.5.1 Sediment Reduction

4.4.5.1.1 Sediment reduction without normalizing BMPs application area
Sediment reduction at the watershed outlet for each BMP without areal normalization is
presented in Figure 4-10 (a, b, and c). Sediment reduction varies between the BMP applied, the
priority area the BMP was applied on, and the method of priority area identification. Overall, the
greatest percentage of sediment reduction for all priority areas was native grass, while the lowest
percent reduction was residue management (0 kg/ha) for all targeting methods and priority areas.
Comparing BMPs, native grass likely has the highest reduction efficiency because it is the most
intensive, while residue management (0 kg/ha) is much less. Overall, the BMPs applied under
LPSAI for high priority are the most effective in sediment reduction at the outlet because the
priority areas were identified based on the pollutant load from each subbasin and this method
identified the greatest amount of high priority area. For medium priority, the case is similar with
LII, where the BMPs are most effective because they are applied on more area. Comparing,
medium priority generally had the greatest reduction because the greatest proportion of area
(agricultural land) was selected as medium priority.
Similar sediment reduction trends as the watershed outlet were observed at the subbasin level
(Figure 4-10 d, e, and f). Native grass and terraces were generally the most effective, regardless
91

of priority and targeting method, likely because of the intensive nature of the BMP
implementation. Recharge structures were the only BMP to have a prominent decrease in percent
sediment reduction at the subbasin when compared to the watershed outlet across all targeting
methods and priorities. This is likely because recharge structures are implemented in the channel
rather than on the field. Once again it is observed that overall reduction is greatest when the area
of application is large. Therefore, LPSAI has the overall greatest percent sediment reduction for
high priority, while LII and CII have the greatest for medium and low priority, respectively.

High Priority Area
80
(a)

70

CII
LPSAI

% Sediment Reduction

60

LPUAI

50
40
30
20
10

NG

T

RS

R M 2000

SC

CF

NT

R M 1000

-20

CT

-10

RM0

0

Figure 4-10. (a) Sediment reduction by BMPs in reach for different targeting methods
without normalizing the BMPs application area.

92

Medium Priority Area

80

(b)
CII

% Sediment reduction

60

LII
LPSAI

40

LPUAI
20

NG

T

RS

R M 2000

SC

CF

NT

CT

R M 1000

-20

RM0

0

Figure 4-10. (b) Sediment reduction by BMPs in reach for different targeting methods
without normalizing the BMPs application area.

93

Low Priority Area

35

(c)

% Sediment reduction

30

CII

25

LII
LPSAI

20

LPUAI

15
10
5
NG

T

RS

R M 2000

SC

R M 1000

CF

CT

NT

-5

RM0

0

Figure 4-10. (c) Sediment reduction by BMPs in reach for different targeting methods
without normalizing the BMPs application area.

94

(d)

CII
LPSAI

NG

T

RS

R M 2000

SC

R M 1000

CF

NT

CT

LPUAI

RM0

% Sediment reduction

50
45
40
35
30
25
20
15
10
5
0
-5

Figure 4-10. (d) Sediment reduction by BMPs in high priority area in subbasin for
different targeting methods without normalizing the BMPs application area.

95

70

CII

(e)

LII

50
% Sediment reduction

60

LPSAI

40

LPUAI

30
20
10

-20

NG

T

RS

R M 2000

SC

CF

NT

R M 1000

-10

RM 0

CT

0

Figure 4-10. (e) Sediment reduction by BMPs in medium priority area in subbasin for
different targeting methods without normalizing the BMPs application area.

96

(f)

CII
LII
LPSAI

NG

T

RS

RM
2000

SC

RM
1000

CF

NT

CT

LPUAI

RM0

% Sediment reduction

80
70
60
50
40
30
20
10
0
-10
-20

Figure 4-10. (f) Sediment reduction by BMPs in low priority area in subbasin for different
targeting methods without normalizing the BMPs application area.

4.4.5.1.2 Sediment reduction after normalizing BMPs application area
The total BMP application area of each targeting scenario was divided by the amount of
sediment reduction of each BMP in each targeting method in order to normalize the area of BMP
application. The new percentage of sediment reduction in four targeting methods for the
watershed outlet and the subbasin is presented in Figure 4-11.

97

High Priority Area
0.9

(a)
CII

Sediment Yield (tons/ha)

0.7

LPSAI
LPUAI

0.5

0.3

NG

T

RS

RM 2000

SC

RM 1000

CF

NT

CT

-0.1

RM 0

0.1

Figure 4-11. (a) Sediment reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area.

98

Medium Priority Area

CII

0.9

(b)

Sediment Yield (tons/ha)

LII
0.7

LPSAI
LPUAI

0.5
0.3

NG

T

RS

RM 2000

SC

RM 1000

CF

NT

CT

-0.1

RM 0

0.1

Figure 4-11. (b) Sediment reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area.

99

Low Priority Area

0.9

(c)
CII
LII

Sediment Yield (tons/ha)

0.7

LPSAI
0.5

LPUAI

0.3

NG

T

RS

RM 2000

SC

RM 1000

CF

NT

CT

-0.1

RM 0

0.1

Figure 4-11. (c) Sediment reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area.

100

4
(d)

CII
3.5
LPSAI
Sediment Yield (tons/ha)

3
LPUAI
2.5
2
1.5
1
0.5
NG

T

RS

RM 2000

SC

RM 1000

CF

CT

NT

-0.5

RM 0

0

Figure 4-11. (d) Sediment reduction by BMPs in high priority area in subbasin for different
targeting methods after normalizing the BMPs application area.

101

2.5

(e)

CII

Sediment Yield (tons/ha)

2

LII
LPSAI

1.5
LPUAI
1
0.5

NG

T

RS

RM 2000

SC

CF

NT

RM 1000

-0.5

RM 0

CT

0

Figure 4-11. (e) Sediment reduction by BMPs in medium priority area in subbasin for
different targeting methods after normalizing the BMPs application area.

102

CII
(f)

LII
LPSAI

NG

T

RS

RM 2000

SC

RM 1000

CF

NT

CT

LPUAI

RM 0

Sediment Yield (tons/ha)

9
8
7
6
5
4
3
2
1
0
-1

Figure 4-11. (f) Sediment reduction by BMPs in low priority area in subbasin for different
targeting methods after normalizing the BMPs application area.

When normalizing for application area, sediment reduction at the watershed outlet had similar
trends to the non-normalized reduction. Applying BMPs on the medium priority areas has the
greatest impact when normalizing for application area (Figure 4-11a, b, and c). In addition, all
targeting methods have relatively similar performance in sediment reduction for all three
priorities. In the low priority area, recharge structures have the highest reduction in LPUAI
because of the relatively low implementation area coupled with the fact that this BMP is
implemented in the reach rather than the field. This indicates that recharge structures are highly
effective if sediment reduction at the watershed outlet is a concern.

Normalized sediment reduction at the subbasin level also exhibits a similar trend to the nonnormalized reduction (Figure 4-11d, e, and f). Similar to the reach results, the targeting methods
had similar normalized percent sediment reduction. The BMP to note from high and medium
103

priority (for all targeting methods) is recharge structures, which is much less effective at the
subbasin than at the reach because of the manner in which it is implemented. Under low priority
conservation tillage and no tillage have considerably higher normalized sediment reduction for
the LPUAI method, which is also true for the non-normalized case. This is likely due to the
limited area of agricultural land attributed to low priority in this method coupled with the
relatively high sediment reduction under low priority for these BMPs.

4.4.5.2 TN Reduction

4.4.5.2.1 TN reduction without normalizing BMPs application area
At the watershed outlet, varying percent TN reduction was observed based on targeting method,
priority area, and BMP (Figure 4-12a, b, and c). Overall, native grass and terraces have the
highest reduction efficiencies, while no tillage and conservation tillage have the lowest. BMP
application on high priority areas generally has the highest percent TN reduction at the watershed
outlet, while low priority is the lowest. This indicates that targeting TN high priority areas is a
viable solution to reducing TN loads at the watershed outlet, regardless of the targeting method.
Meanwhile, there are two methods (LPUAI-medium priority and LII-low priority) that have
much higher reduction efficiencies than other methods in the same priority. In the case of LPUAI
this is because the method identified a large area of medium priority, therefore the BMPs were
applied on a larger scale. For LII under low priority, the reduction efficiency is much higher
because most agricultural land in this scenario was identified to be low priority, increasing BMP
application area. This also indicates that LII may be misleading, because application on medium

104

priority areas is much less efficient. Therefore, it may not always be optimal to place BMPs near
the watershed outlet just because the TN load is higher at that location.

High Priority Area
30
25

LPSAI

20

LPUAI

15
10
5

CT

% TN reduction

(a)

CII

NG

T

SC

RS

CF

RM 2000

RM 1000

-10

NT

-5

RM 0

0

Figure 4-12. (a) TN reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area.

105

Medium Priority Area

25
20
15

5

RM 0

10
CT

% TN reduction

(b)

CII
LII
LPSAI
LPUAI

NG

T

SC

RS

CF

RM
1000
RM
2000

-5

NT

0

Figure 4-12. (b) TN reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area.

106

Low Priority Area

30

CII
LII
LPSAI
LPUAI

25

15
10
5

RM 0

CT

-10

NT

NG

T

SC

RS

CF

-5

R M 2000

0

R M 1000

% TN reduction

20

(c)

Figure 4-12. (c) TN reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area.

107

40
(d)

35

% TN reduction

30

CII

25

LPSAI

20
LPUAI

15
10
5

NG

T

SC

RS

CF

R M 2000

RM 0

CT

-10

NT

-5

R M 1000

0

Figure 4-12. (d) TN reduction by BMPs in high priority area in subbasin for different
targeting methods without normalizing the BMPs application area.

108

35

(e)

30

CII
LII
LPSAI
LPUAI

20
15
10

RM 0

% TN reduction

25

5

NG

T

SC

RS

CF

R M 2000

R M 1000

-10

CT

-5

NT

0

Figure 4-12. (e) TN reduction by BMPs in medium priority area in subbasin for different
targeting methods without normalizing the BMPs application area.

109

50
(f)

CII
LII
LPSAI

30

LPUAI

20

NG

T

SC

CF

RM
2000

RM
1000

-10

RM 0

0

CT

10
NT

% TN reduction

40

Figure 4-12. (f) TN reduction by BMPs in low priority area in subbasin for different
targeting methods without normalizing the BMPs application area.

The subbasin level generally exhibits the same reduction patterns that are observed at the
watershed outlet. Native grass and terraces have the highest percent TN reduction for any
targeting method and priority, while no till, conservation tillage, and residue management (0
kg/ha) are generally the least effective. The contrast in efficiencies is likely due to the manner of
intensive implementation for native grass and terraces. Overall, BMP application on high priority
areas has the greatest percent TN reduction, while low priority has the lowest (except for the LII
method). The LII method on low priority areas has higher reduction at the subbasin level
primarily because this method identified most agricultural area as low priority. In most cases,
there is a positive correlation between area of BMP application based on the targeting method
and that method’s percent TN reduction. However, recharge structures have a negative reduction

110

of TN at the subbasin level, while a positive reduction was observed in the reach. This may be
due to the fact that this BMP is implemented in the main channel (reach) rather than in the field.

4.4.5.2.2 TN reduction after normalizing BMPs application area
The BMPs application area for TN targeting scenario for four targeting methods were variable.
Therefore, the percentage of TN reduction before normalization was divided by the total BMP
application area in order to obtain the normalized percentage of TN reduction.

At the watershed outlet, the normalized percent TN reduction follows trends that are similar to
the non-normalized reduction, although some trends are more pronounced (Figure 4-13a, b, and
c). For example, the LII method on medium priority becomes the most effective because
normalization reveals that the BMPs under this targeting scenario have the greatest reduction
because of limited application area. Similarly, LPUAI on high priority becomes the most
efficient after normalization because this method only identifies a small area of the watershed
that is high priority. Therefore, it can be said that the LPUAI method requires less investment in
application (due to the limited high priority area) while still producing large percent TN
reduction at the watershed outlet. Low priority areas for the CII, LII, and LPSAI methods have
percent TN reductions close
to zero for all BMPs, indicating their limited effectiveness at TN reduction at the watershed
outlet when applied across broad low priority areas. Conversely, the LPUAI method on low
priority identified a limited number of low priority areas defined as agriculture and observed
increased reduction for contour farming and recharge structures.

111

High Priority Area

18

(a)

14

LPSAI

12

LPUAI

10
8
6

2

CT

4

NT

Total Nitrogen Yield (kg/ha)

16

CII

NG

T

SC

RS

CF

RM 2000

-4

RM 1000

-2

RM 0

0

Figure 4-13. (a) TN reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area.

112

Medium Priority Area

25
(b)

LII

15

LPSAI
LPUAI

10
5

NG

T

SC

RS

CF

RM 2000

RM 1000

RM 0

-5

CT

0
NT

Total Nitrogen Yield(kg/ha)

CII
20

Figure 4-13. (b) TN reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area.

113

Low Priority Area
250
(c)

CII
LII
150

LPSAI

100

LPUAI

50

NG

T

SC

RS

CF

RM 2000

RM 1000

RM 0

-50

CT

0
NT

Total Nitrogen Yield (kg/ha)

200

-100
-150
-200

Figure 4-13. (c) TN reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area.

114

30
(d)

Total Nitrogen Yield (kg/ha)

CII
25

LPSAI

20

LPUAI

15
10
5

NG

T

SC

RS

CF

RM 2000

RM 1000

RM 0

CT

-5

NT

0

Figure 4-13. (d) TN reduction by BMPs in high priority area in subbasin for different
targeting methods after normalizing the BMPs application area.

115

30
CII
Total Nitrogen Yield (kg/ha)

25

(e)

LII
LPSAI

20

LPUAI
15

10
5

NG

T

SC

RS

CF

RM 2000

RM 1000

RM 0

CT

-5

NT

0

Figure 4-13. (e) TN reduction by BMPs in medium priority area in subbasin for different
targeting methods after normalizing the BMPs application area.

116

CII
(f)

LII

200

LPSAI
150

LPUAI

100
50

NG

T

SC
RS

CF

RM 2000

RM 1000

-50

RM 0

CT

0
NT

Total Nitrogen Yield (kg/ha)

250

Figure 4-13. (f) TN reduction by BMPs in low priority area in subbasin for different
targeting methods after normalizing the BMPs application area.

At the subbasin, trends observed in the non-normalized percent TN reduction are amplified in the
areal normalization (Figure 4-13d, e, and f). Consistent with sediment, native grass and terraces
are generally the most efficient. Under high priority, the LPUAI targeting becomes the most
efficient because of the limited agricultural area identified as high priority. The situation is
similar for the medium priority LII targeting method. This indicates that on a reduction per area
basis, these two targeting methods have the highest efficiency. For low priority areas, the LPUAI
method has predominantly higher reduction efficiencies because most land applicable for BMPs
was identified as high or medium priority, therefore what is applied on low priority will have a
large impact in this case.

117

4.4.5.3 TP Reduction

4.4.5.3.1 TP reduction without normalizing BMPs application area
Percent reduction of TP at the watershed outlet varies widely between targeting method and
priority area (Figure 4-14a, b, and c). As with sediment and TN, native grass and terraces have
the greatest percent reduction overall, while residue management, no tillage, and conservation
tillage have the smallest percent reduction. Under high priority areas, CII and LPSAI are
generally the most effective at TP reduction at the watershed outlet because the agricultural high
priority areas identified are largest. Medium priority areas generally have less than 10% TP
reduction for all BMPs and targeting methods. This indicates that regardless of targeting method,
BMP placement on medium priority areas has limited impact at the watershed outlet. Most
BMPs and targeting methods have close to 0% TP reduction, except for the LII method. The
impact of BMP placement under LII is more substantial in this case because there is a large
amount of agricultural land identified as low priority. A negative reduction of TP was observed
by residue management (0 kg/ha), no till, and conservation tillage at both reach and subbasin
level. This may be due to higher concentration of dissolved phosphorus in the runoff. A similar
result was observed by Bundy et al. (2001) in which they found increased dissolve phosphorus
concentration from no till corn fields.

Similar trends are observed at the subbasin level as those at the watershed outlet (Figure 4-14d,
e, and f). BMP application on high priority areas generally has the greatest impact on percent TP
reduction. The targeting methods have the most variable reductions under medium priority,
which is likely because each method selected very different areas of medium priority agricultural
land. For low priority, the reach based methods (CII and LII) have the greatest percent TP

118

reduction, which is likely because these methods selected the most low priority area. This
indicates that the reach based methods may not be applicable for TP reduction at the subbasin,
because it is more beneficial to have greater reduction on areas identified as high priority
because implementation is more likely to occur there.

High Priority Area
60

(a)
CII

40
% TP reduction

50

LPSAI

30

LPUAI

20
10

-30

NG

T

RS

SC

CF

R M 2000

R M 1000

CT

-20

NT

-10

RM0

0

-40

Figure 4-14. (a) TP reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area.

119

Medium Priority Area
25

(b)
CII

20

LII

LPUAI

10

5

RM 0

% TP reduction

LPSAI
15

NG

T

RS

SC

CF

R M 2000

R M 1000

CT

-5

NT

0

Figure 4-14. (b) TP reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area.

120

Low Priority Area
40

CII
LII

(c)

LPSAI

20

LPUAI
CT

10

NG

T

RS

SC

CF

R M 2000

-20

R M 1000

-10

RM 0

0

NT

% TP reduction

30

-30

Figure 4-14. (c) TP reduction by BMPs in reach for different targeting methods without
normalizing the BMPs application area.

121

60
50
40

LPSAI

30

LPUAI

20
10

NT

-20

-40

NG

T

RS

SC

RM 0

-30

R M 2000

CT

-10

CF

0

R M 1000

% TP reduction

(d)

CII

Figure 4-14. (d) TP reduction by BMPs in high priority area in subbasin for different
targeting methods without normalizing the BMPs application area.

122

80

(e)

CII
60

LII
LPSAI
LPUAI

20

NG

T

SC

RS

CF

RM 0

-60

CT

-40

NT

-20

R M 2000

0

R M 1000

% TP reduction

40

Figure 4-14. (e) TP reduction by BMPs in medium priority area in subbasin for different
targeting methods without normalizing the BMPs application area.

123

60
50
40
30
20
10

-30
-40

NG

T

SC

RS

CF

R M 2000

RM 0

-20

CT

-10

R M 1000

0

NT

% TP reduction

(f)

CII
LII
LPSAI
LPUAI

Figure 4-14. (f) TP reduction by BMPs in low priority area in subbasin for different
targeting methods without normalizing the BMPs application area.

4.4.5.3.2 TP reduction after normalizing BMPs application area
Percent TP reduction at the watershed outlet normalized by area exhibits uniformity across
targeting methods, while high, medium, and low priorities have similar efficiencies (Figure 415a, b, and c). Native grass and terraces are still the most efficient BMPs, although other BMPs
have similar efficiencies. High and medium priority areas have similar results for all targeting
methods, indicating that no BMP or targeting method is superior when examining percent TP
reduction per area. In the case of low priority areas, recharge structures for LPUAI have
significantly greater reduction efficiency, which is likely because recharge structures are
implemented in stream and the amount of low priority area in this method is minimal. However,
124

for some BMPs, such as no till, conservation tillage, residue management 0 kg/ha, residue
management 1000 kg/ha, and residue management 2000 kg/ha, produced negative TP reduction
at reach (Figure 4-15 c) based on LPUAI targeting method in low priority areas.

At the subbasin level, trends are similar to the reach for the normalized percent TP reduction
(Figure 4-15d, e, and f). All priority areas have similar normalized percent TP reduction,
although high priority areas have the greatest reduction. In addition, all targeting methods
produce similar results when BMPs are applied. Therefore, on a per area basis, any targeting
method or BMP is likely to have similar normalized percent TP reduction. The exceptions to this
are no tillage and conservation tillage on low priority for the LPSAI and LPUAI methods, which
is likely due to the limited area identified as low priority. Consequently, when targeting with the
LPSAI or LPUAI it is advisable to implement conservation or no tillage practices to get the most
percent TP reduction at the subbasin level.

125

High Priority Area

2

(a)

1.5

LPSAI
LPUAI

1
0.5
CT

NG

T

RS

SC

CF

RM 2000

RM 0

-0.5

RM 1000

0
NT

Total Phosphorus Yield (kg/ha)

CII

-1

Figure 4-15. (a) TP reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area.

126

Medium Priority Area
1.5

CII
(b)
LII
LPSAI
LPUAI

0.5

NG

T

RS

SC

CF

RM 2000

RM 1000

CT

-0.5

NT

0
RM 0

Total Phosphorus Yield (kg/ha)

1

-1

-1.5

Figure 4-15. (b) TP reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area.

127

Low Priority Area

7
CII

(c)

LII
5

LPSAI

4

LPUAI

3
2
1

RM 0

-3

NG

T

RS

SC

RM 2000

-2

CT

-1

RM 1000

CF

0

NT

Total Phosphorus Yield (kg/ha)

6

Figure 4-15. (c) TP reduction by BMPs in reach for different targeting methods after
normalizing the BMPs application area.

128

5

(d)

CII
LPSAI
LPUAI

3
2
1

NG

T

RS

SC

CF

RM 2000

RM 1000

CT

-1

NT

0
RM 0

Total Phosphorus Yield (kg/ha)

4

-2

Figure 4-15. (d) TP reduction by BMPs in high priority area in subbasin for different
targeting methods after normalizing the BMPs application area.

129

5
(e)

4

LII
LPSAI

3

LPUAI
2
1

RM 0

CT

NT

-1

NG

T

RS

SC

CF

RM 2000

0
RM 1000

Total Phosphorus Yield (kg/ha)

CII

-2

Figure 4-15. (e) TP reduction by BMPs in medium priority area in subbasin for different
targeting methods after normalizing the BMPs application area.

130

10.00

(f)

CII
LII

6.00

LPSAI
LPUAI

4.00
2.00

NG

T

RS

SC

CF

RM 2000

RM 1000

-2.00

NT

CT

0.00

RM 0

Total Phosphorus Yield (kg/ha)

8.00

-4.00

Figure 4-15. (f) TP reduction by BMPs in low priority area in subbasin for different
targeting methods after normalizing the BMPs application area.

4.4.6 Evaluation of Relative Sensitivity Index among Targeting Methods
Evaluation of relative sensitivity index allows for comparison of the pollution reduction
tendencies of the targeting methods combined with priority level. Eleven scenarios were plotted
(all possible combinations of targeting method and priority area) in Figure 4-16. The median and
th

th

the 25 and 75 percentile sensitivity index across all BMPs for each sensitivity scenario were
included to identify variability in subbasin versus watershed outlet pollution reduction. The
median relative sensitivity index for most sensitivity scenarios is close to one with little
variability, indicating that reducing pollution at the subbasin level causes a proportional
reduction at the watershed outlet. Meanwhile, CII-high and LPUAI-high have a high relative

131

sensitivity index for all three pollutants, while the LII-medium scenario is highly variable for
nutrients.

4.4.6.1 Sediment Targeting Scenario
For sediment relative sensitivity index, most scenarios had a median close to one with little
variability (Figure 4-16a). The CII-high and LPUAI-high were the only sensitivity scenarios with
large variability and median relative sensitivities greater than one. A high relative sensitivity
index indicates that there is a much greater proportion of reduction occurring at the subbasin
level rather than at the watershed outlet.

6
(a)

Relative Sensitivity Index

5

4

3
2

1

Figure 4-16. (a) BMP relative sensitivity index sediment.
132

LPUAI-L

LPUAI-M

LPUAI-H

LPSAI-L

LPSAI-M

LPSAI-H

LII-L

LII-M

LII-H

CII-L

CII-M

CII-H

0

CII-high has a high relative sensitivity index because of the targeting methodology; CII-high
generally targets streams with low flows, so even relatively small loads will increase pollution
concentration. This can be a very effective approach to protect sensitive aquatic ecosystems. As
BMPs are implemented on the high priority areas identified by CII, reduction is high at the
subbasin but there is not as great of an impact at the watershed outlet because only small
upstream areas are targeted. This results in a higher proportion of reduction at the subbasin and
consequently, a high relative sensitivity index. Variability of CII-high is also large, which is due
to the difference in sediment reduction between BMPs at the subbasin level and at the watershed
outlet. Percent reduction is similar among all BMPs at the watershed outlet, while native grass
and terraces have much higher reduction efficiency at the subbasin than residue management (0
kg/ha) and conservation tillage (not shown in Figure 16). These trends create variability in the
relative sensitivity index.

LPUAI-high has a large relative sensitivity index because it targets pollution generation at the
subbasin level while ignoring pollution in the reach. Therefore, percent sediment reduction for
all BMPs is larger at the subbasin level than at the watershed outlet, which creates a large
relative sensitivity index. In addition, the variability is fairly large compared to other sensitivity
scenarios because percent sediment reduction at the watershed outlet is relatively the same across
all BMPs, while at the subbasin level native grass, tillage, and residue management (2000 kg/ha)
have much higher percent sediment reduction than other BMPs.

133

4.4.6.2 TN Targeting Scenario
Overall, the TN scenario has less extreme sensitivity than the sediment scenario, while the
variability is generally greater than the TP scenario (Figure 4-16b).

1.6
(b)

Relative Sensitivity Index

1.4
1.2
1.0
0.8
0.6
0.4
0.2

Figure 4-16. (b) BMP relative sensitivity index TN.

134

LPUAI-L

LPUAI-M

LPUAI-H

LPSAI-L

LPSAI-M

LPSAI-H

LII-L

LII-M

LII-H

CII-L

CII-M

CII-H

0.0

The median relative sensitivity index is close to one for all sensitivity scenarios, although the
CII-high and LII-medium are greater than one and less than one, respectively. The CII-high
sensitivity is generally higher for similar reasons as in the sediment sensitivity scenario.

LII-medium has a median relative sensitivity lower than one due to the targeting methodology.
This method targets high stream loads while ignoring what is happening at the subbasin level.
Therefore, it is likely that the proportion of reduction at the watershed outlet is greater than at the
subbasin level. High range of sensitivity in the LII-medium scenario is due to variable percent
TN reduction at the watershed outlet between BMPs, while at the subbasin level, there is less
difference in reduction between BMPs (indicating that the most efficient BMPs have a greater
impact at the watershed outlet).

4.4.6.3 TP Targeting Scenario
For TP, the relative sensitivity index for most sensitivity scenarios is close to one with little
variability (Figure 4-16c). This indicates that percent TP reduction of BMPs is similar at the
subbasin level and the watershed outlet for most of the sensitivity scenarios. Overall, sediment
and TP are similar for most sensitivity scenarios because of the transport relationship between
the two pollutants.

135

(c)

1.4

Relative Sensitivity Index

1.2
1.0

0.8
0.6
0.4
0.2

LPUAI-L

LPUAI-M

LPUAI-H

LPSAI-L

LPSAI-M

LPSAI-H

LII-L

LII-M

LII-H

CII-L

CII-M

CII-H

0.0

Figure 4-16. (c) BMP relative sensitivity index TP.

LPUAI-high is the only sensitivity scenario with a median sensitivity substantially larger than
one, which is similar to the sediment case. This method targets pollution at the subbasin level
rather than in the reach, therefore it is likely that the proportion of percent TP reduction is higher
at the subbasin than at the watershed outlet. The variability of LPUAI-high is larger than other
sensitivity scenarios because percent TP reduction is more variable at the subbasin level than at
the watershed outlet. This indicates that BMP selection plays a more important role when the
goal of implementation is TP reduction from the field than at the watershed outlet.

136

4.5 CONCLUSION
BMPs are one of the primary options in maintaining surface water quality. Informed placement
of BMPs plays an important role in achieving maximum pollution reduction while minimizing
costs. The objectives of this research were to (1) identify CSAs using multiple targeting
techniques and pollutants, (2) assess the sensitivity of BMPs to different targeting methods, and
(3) evaluate the impact of BMP application in CSAs at subbasin and watershed scales. In the
SRW, four targeting methods (CII, LII, LPSAI, and LPUAI) were used in SWAT to prioritize
BMPs for sediment, TN, and TP reduction. Ten BMPs (RM 0 kg/ha, CT, NT, CF, RM 1000
kg/ha, RM 2000 kg/ha, SC, RS, T, and NG) were implemented for each targeting method.

Initially, priority area selection between targeting methods for the entire SRW was compared for
sediment, TN, and TP. In the sediment targeting scenario, the LPSAI targeting methods had the
largest high priority areas, while LII had the least. For TN and TP, CII had the greatest high
priority area, while LII had the least. The CII targeting method characterized large high priority
area due to numerous smaller tributaries within the SRW. Low flow in the tributaries combines
with pollutant load from agricultural lands, thereby causing high pollutant concentrations.
Conversely, the LII targeting method consistently identified the smallest area of high priority
area because of less load carrying capacity of smaller tributaries compared to the main channel in
the watershed. Therefore, only subbasins near the SRW outlet were identified as high priority.
The LPSAI targeting method categorized the greatest high priority area for sediment targeting
scenario because of the combination of selecting large subbasin areas with pollutant loads from
agricultural lands. The characterization of priority areas in LPUAI targeting method primarily
depends on the pollutant load from agricultural areas, but is normalized by subbasin area.

137

Therefore, this method produced less high priority areas when compared to the LPSAI targeting
method.

The targeting methods were also compared for sediment, TN, and TP on agricultural lands
(available land for BMP implementation). This allows us to quantify the capability of each
targeting method to identify agricultural lands as a NPS source. Trends in high priority area
selection were similar for the entire SRW and agricultural lands. In the sediment targeting
scenario, the LPSAI targeting method had the greatest high priority area for agricultural lands in
the watershed, while the LII had the least. Under the TN and TP targeting scenarios, the CII
targeting method had the highest high priority area for agricultural land whereas the LII targeting
method had the least. The LPSAI targeting method produced larger high priority area because
the method is based on field generated pollutant load across an area. Therefore, large subbasins
primarily composed of agricultural lands are likely to be high priority. In contrast, the LII
targeting method had the least high priority area due to the method itself, as it targets the
pollutant load on the reach, without considering the field. The CII targeting method had the
highest amount of agricultural high priority area both for TN and TP targeting scenario likely
because of low flow conditions receiving pollutants from agricultural lands, creating higher than
average concentrations.

When applying BMPs to the priority areas based on the four targeting methods, varying results
were observed. Sediment, TN, and TP reduction was compared at the subbasin level and at the
watershed outlet for each BMP/targeting method combination. Comparing BMPs across all
targeting scenarios, native grass and terraces were generally the most effective at pollution

138

reduction, where residue management (0 kg/ha), conservation tillage, and no tillage were
consistently the least effective at both the subbasin and watershed outlet. This result is likely due
to the manner of BMP implementation, where native grass and terraces are more intensive
compared to adjustment of tillage practices. The LPSAI method was most effective in reducing
sediment at the watershed outlet when applying BMPs on high priority areas compared to other
targeting methods due to a combine effect of the method itself and the greatest amount of high
priority areas. A similar trend was observed after normalization of BMP application area both at
outlet and subbasin level. The CII targeting method was most effective in reducing TN at the
subbasin level and watershed outlet by applying BMPs in the high priority area because of
greater BMP application area. However, after normalization of BMP application area the LPUAI
targeting method reduced a greater amount of TN both at the outlet and subbasin level, indicating
that the most reduction per BMP implementation will occur under this scenario. Like the TN
targeting scenario, the CII targeting method was most effective in reducing TP compared to other
targeting methods because of greater agricultural high priority areas, and therefore more area in
which BMPs were implemented. After normalization of BMP application area no targeting
method was superior when comparing the TP reduction per area at the watershed outlet. The
results of the BMP implementation based on targeting method indicate that selection of targeting
method and BMP must be done with care and will depend on the goal of policymakers and
watershed managers.

BMP relative sensitivity index was compared among all targeting methods and priority areas.
Understanding the relative sensitivity index indicates whether the targeting method and priority
with applied BMPs is more sufficient at pollution reduction at the subbasin or watershed outlet,

139

and to what proportion. Except for two scenarios (CII-high and LPUAI-high), the overall relative
sensitivity index for the sediment targeting scenarios was approximately one, indicating that
reduction at the reach and subbasin is proportional. This result was due to the targeting
methodology itself. CII-high targets low flow streams with high concentration (not necessarily
load), which results in implementation in many areas, reducing sediment production from the
field, although the impact is not strongly observed at the watershed outlet. LPUAI targets the
load from the subbasin rather than the watershed outlet; therefore the effects of BMP
implementation are less likely to be observed at the watershed outlet. Relative sensitivity index
was similar for the TN scenario, although LII-medium relative sensitivity index was less than
one, indicating that reduction is more apparent at the watershed outlet than it is at the subbasin
level. The LII method targets stream loads at the watershed outlet rather than load from the
subbasin, causing sensitivity less than one. Overall, relative sensitivity index was approximately
one for the TP targeting scenario, except for LPUAI-high. The higher relative sensitivity index in
LPUAI-high was due to a greater proportion of TP reduction in subbasin level compared to the
watershed outlet because the targeting method aims at reducing pollutant at subbasin rather than
the outlet.

This study compared targeting methods under different priority areas based on targeting
pollution type (sediment, TN, and TP) to develop possible BMP implementation decisions. No
single method was found to be significantly better or worse for priority area selection and
pollutant type, although each method produced different high and medium priority areas. In
addition, utilization of a specific targeting method should be based on the goals stated in a BMP
implementation project. For example, when the goal of project is to protect aquatic health, it may

140

be useful to use a method that focuses in-stream pollutant concentration. Conversely, if
preventing sedimentation of a reservoir is the goal of a BMP implementation project, selecting
the targeting method that focuses on load reduction in the stream may be more appropriate.
Further studies need to be completed to address spatial variability of priority areas with respect
to time, climate change, and land use change.

141

5. ANALYSIS OF BEST MANAGEMENT PRACTICE EFFECTIVENESS AND
SPATIOTEMPORAL VARIABILITY BASED ON DIFFERENT TARGETING
STRATEGIES.

5.1 ABSTARCT
In this study, ten best management practices (BMPs) were modeled for agricultural areas in the
Saginaw River Watershed using the Soil and Water Assessment Tool based on four targeting
methods (Load per Subbasin Area Index (LPSAI), Load per Unit Area Index (LPUAI),
Concentration Impact Index (CII), and Load Impact Index (LII). The effective BMPs both for
targeting and non-targeting pollutants were contour farming (except total nitrogen reduction
during total phosphorus targeting scenario), residue management 1000 kg/ha and 2000 kg/ha,
strip cropping, recharge structures, terracing, and native grass. In contrast, conservation tillage
and no tillage did not reduce significant amount of pollutants for any combination of targeting
methods and priority areas. In regard to spatial correlation between targeting methods, a strong
relationship was found between the LPSAI and LPUAI methods both for the sediment and total
nitrogen targeting scenarios. In addition, a similar result was found between the CII and LPSAI
targeting methods. Regarding the spatiotemporal variability of BMP implementation plan, a
distinct change in priority area was observed in the case of native grass implementation by the
end of the second year; however, this impact was minimal for contour farming due to less
pollutant reduction efficiency compared to native grass.

142

5.2 INTRODUCTION
Degradation of water quality in recent decades is a major concern for society, which is further
compounded by land use change and intensified agricultural practices. According to the US
Environmental Protection Agency (2009), 44% out of 3.5 million miles of the nation’s rivers and
streams were impaired. Excessive nutrient loading into waterbodies originates from different
sources such as improper application of fertilizer, animal wastes, irrigation water, frequent
plowing, forestry, and urban development. This leads to eutrophication, resulting in diminished
water quality (TNRCC, 1999). Accumulation of surplus nutrients, primarily phosphorus and
nitrogen, enhances excessive algal growth. Algal bloom in waterbodies decreases light
availability and increases organic matter production, resulting in degradation of habitat,
decreased fishery production, and substantial economic impact. Additionally, eutrophication
decreases property value, disrupts recreation and tourism, and creates taste and odor problems
resulting in increased drinking water treatment costs (Dodds et al., 2009).

Improving water quality through implementing best management practices (BMPs) on
agricultural lands is currently receiving increase interest. BMPs are a widely accepted method to
control pollution from agricultural activities within a watershed (Arabi et al., 2007). However,
the effectiveness of BMPs varies from site to site and with the type of BMP applied (Giri et al.,
2012a). Considering resource and time constraints, it is impractical to develop best management
strategies through field studies. Therefore, watershed managers prefer modeling for development
of watershed management plans. Meanwhile, BMP implementation on all agricultural lands is
not required as application of BMPs at critical source areas (CSAs) may reduce the pollutants to
acceptable levels (Maringanti et al., 2009). CSAs generate a disproportionate amount of

143

pollutants in the watershed, which is a combination of land use, soil type, slope, and proximity to
the waterbodies. Targeting CSAs is more effective and resource efficient than randomly
implementing BMPs throughout a watershed. Therefore, identification of CSAs is essential to
optimize nonpoint source (NPS) pollution reduction. Preferential implementation of BMPs on
CSAs is known as the targeting approach.

Models such as the Soil and Water Assessment Tool (SWAT), Hydrological Simulation Program
Fortran (HSPF), Annualized Agricultural Nonpoint Source (AnnAGNPS), Areal Nonpoint
Source

Watershed

Environment

Response

Simulation

(ANSWERS-2000),

Watershed

Assessment Model (WAM), and Water Erosion Prediction Project (WEPP) have been used for
developing BMP implementation plans. Among these models, SWAT is preferred because its
built-in equations describe various agricultural components such as tillage operations, fertilizer
and pesticide applications, vegetative filter strips, and crop rotations in more detail than other
models. SWAT has been widely used to assess the effectiveness of BMPs on the watershed scale
(Arnold et al., 1998; Gassman et al., 2007, Arabi et al., 2007, Arabi ei al., 2008, Panagopoulos et
al., 2011). Several studies have used the SWAT model in targeting watershed CSAs for BMP
implementation (Panagopoulos et al., 2011, Ghebremichael et al.,2010, Jha et al., 2010;
Maringanti et al., 2009; Parajuli et al., 2008; Schilling and Wolter, 2009; Tuppad et al., 2010;
White et al., 2009). However, these studies did not address the spatiotemporal impacts (change in
space and time) of BMP implementation plans or explore BMP effectiveness on both targeted
pollutants and non-targeted pollutants.

144

5.2.1 CSA Identification
Ghebremichael et al. (2010) used the SWAT model to identify and quantify phosphorus CSAs in
the Rock River watershed, which is the primary phosphorus contributor to Lake Champlain.
They found that more than 50 % of the sediment and TP originated from corn fields, and another
20-25% of TP was coming from other row crops. Overall, 24% of the watershed area contributed
80% of TP. Tuppad et al. (2010) used random and targeting methods for implementation of
BMPs (reduced tillage, edge of field vegetative filter strips, and contoured terraces) in the Smoky
Hill watershed in Kansas. The BMPs were implemented on 10%, 26%, 52%, and 100% of the
total targeted cropland and the pollutant reduction efficiencies were compared at the outlet of the
watershed. CSAs were defined based on the total pollution load per unit area. They observed that
the targeting method was more effective compared to the random method. White et al. (2009)
identified CSAs and quantified sediment and TP loads at the watershed scale in Oklahoma using
the SWAT model. Within each hydrologic response unit (HRU), CSAs were identified based on
threshold unit area load and then HRUs were ranked from highest to lowest based on the
sediment and phosphorus yields. They observed that approximately 22% of the sediment and
phosphorus load originated from only 5% of the agricultural land. Diebel et al. (2008)
implemented BMPs using four allocation approaches: aggregated/targeted, aggregated/random,
dispersed/targeted, and dispersed/random. These allocation approaches were evaluated by two
methods: modeled pollutant reduction index and water quality change index. The proportion of
phosphorus load reduction after BMP implementation is known as the modeled pollutant
reduction index, while the proportion of the watershed having significant reduction of stream
phosphorus concentration is called the water quality change index. For modeled pollutant

145

reduction index and water quality change index, the targeted approach performed better
compared to the random approach for both methods.

Few studies have looked at the performance of BMPs based on different CSA identification
methods. This study is unique because 1) the effects of targeting for a particular pollutant on
non-targeted pollutants are considered; 2) no studies have addressed the impact of gradual BMP
implementation on spatial variability of CSAs despite the fact that in reality, it is very common
that BMP implementation plans are applied in a gradual manner over the course of multiple
years; and 3) spatial correlation (similarity in identifying location) was calculated to determine
the relationship between priority area and targeting methods. In this study, a statistical model
was used to remove time series autocorrelation among the dataset, while comparing the base
with the BMP scenario in order to determine the BMP effectiveness. The objectives of this study
were to 1) determine the most effective BMPs both for targeting and non-targeting pollutants
based on different targeting methods while minimizing the area devoted to BMP implementation;
2) evaluate the spatial correlation among the targeting methods in categorization of priority area
(high, medium, and low) based on targeting pollutant; and 3) assess the spatiotemporal
variability of CSAs.

146

5.3 MATERIALS AND METHODS
5.3.1 Study Area

N

Legend
Saginaw
Michigan
0 10 20 40 60 80
Kilometer
Figure 5-1. Location of Saginaw River Watershed.

147

The study was conducted on the predominantly agricultural lands of the Saginaw River
Watershed (SRW). The SRW (Figure 5-1) is located in east central Michigan, and is one of the
most diverse areas of Michigan having agriculture, manufacturing, tourism, outdoor recreation,
and wildlife habitat. This is the largest six digit hydrologic unit code (HUC)- watershed in
Michigan (040802), and is the nation’s largest contiguous freshwater coastal wetland system
(USEPA, 2009). The SRW flows northward and finally drains into Lake Huron. The total
watershed area in this study is 15,263 km2 with approximately 42% forest, 23% agriculture, 17%
pasture, 11% wetland, while the remaining is high, medium, and low density urban. Corn and
soybeans are the most predominant crops in the watershed.

5.3.2 Model Description
In this study, SWAT was used to identify the critical source areas of sediment, total nitrogen
(TN), and total phosphorus (TP). The SWAT model is developed by the USDA-ARS, and is a
physically-based, spatially distributed watershed scale model (Arnold et al., 1998; Neitsch et al.,
2005; Gassman et al., 2007). The SWAT model estimates flow, sediment, nutrients, and
pesticides on a watershed scale based on land use, soil type, and management operations. A
watershed in SWAT is divided into subbasins and further divided into HRUs based on
homogeneous land use, soil, slope, and management practices. Hydrology, soil characteristics,
plant growth, weather, nutrients, pesticides, and land management practices are major model
components (Gassman et al., 2007).

148

5.3.3 Data Sources
Different types of physiographic data such as topography, land use, soil, and stream network are
required by the SWAT model. The Better Assessment Science Integrating point and Nonpoint
Sources (BASINS) program was used to attain the topography data in the form of the digital
elevation model (DEM). The 2008 Cropland Data Layer was obtained from the USDA's National
Agricultural Statistics Service (NASS, 2008) in order to represent the land use in the watershed.
The State Soil Geographic Database (STATSGO) was used to represent the watershed soil
characteristics. The STATSGO dataset was developed by the National Cooperative Soil Survey
and was linked to tabular data containing soil chemical and physical properties at a scale of
1:250,000. The United States Geological Survey (USGS) National Hydrography Dataset (NHD)
was used to define the stream network in the watershed in order to improve the hydrologic
segmentation and subwatershed boundary delineation.

Daily streamflow data were downloaded for USGS gauging station 04157000 for streamflow
calibration and validation. The water quality data were obtained from the Michigan Department
of Environmental Quality station 090177. In this study, 19 precipitation stations and 11
temperature stations were used to represent the precipitation and temperature data required by
the SWAT. Twenty years (1990-2009) of observed daily precipitation (19 stations) and
temperature (11 stations) data were obtained from the National Climatic Data Center. The
SWAT weather generator program was used to generate the remaining required meteorological
data (wind speed, relative humidity, and solar radiation).

149

Agricultural management operations for corn and soybeans were developed based on common
practices in the region in order to assess the fate and transport of sediment and nutrients in the
watershed. The common practices include tillage and fertilizer applications. The continuous corn
rotation period is six years, out of which corn is planted for five years and soybean is planted in
the final year. The continuous soybean rotation period is three years in length, and consists of
two years soybean planting and a final year of corn planting.

5.3.4 Sensitivity Analysis and Calibration
The most influential parameters on model output (obtained in sensitivity analysis) are required
for use during the model calibration process. Sensitivity analysis ranks parameters in terms of the
sensitivity of model outputs to the input parameters. Latin Hypercube One factor-at-a-time (LHOAT) sampling is used to perform the sensitivity analysis in the SWAT model (van Griensven et
al. 2006). In this study, a sensitivity analysis was performed to estimate the model parameters
sensitive to streamflow, sediment, TN, and TP. Complete sensitivity analysis, calibration, and
validation results are presented in Giri et al. (2012a). Calibration and validation were performed
for streamflow, sediment, TN, and TP on a monthly time step with a two-year model warm-up.

5.3.5 Best Management Practices in SWAT
The ten BMPs modeled on agricultural lands were contour farming (CF), terraces (T), recharge
structure (RS), conservation tillage (CT), no tillage (NT), native grass (NG), residue
management (0 kg/ha) (RM 0), residue management (1000 kg/ha) (RM 1000), residue
management (2000 kg/ha) (RM 2000), and strip cropping (SC)One BMP was used at a time to
assess the impact of BMP effectiveness on pollution reduction. After modeling BMPs in the

150

SRW, the results were compared to the base scenario (no BMPs implemented) to evaluate the
pollution reduction effectiveness of each. The BMPs were implemented in SWAT according to
procedures in various published literatures (USDA, 1996; USDA-NRCS, 2005; Arabi et al.,
2007; Nejadhashemi and Mankin, 2007; Tuppad and Srinivasan, 2008; Tuppad et al., 2010;
USDA, 2010; USDA-NRCS, 2011; Woznicki et al., 2011).

5.3.6 Spatial targeting Methods
Implementation of BMPs throughout the watershed is impractical, expensive, and time
consuming. Proper utilization of limited resources is achieved through the right selection criteria.
Hence, the primary step before BMP implementation is to identify CSAs of pollutants in the
watershed. In this study, four targeting methods – CII, LII, LPSAI, and LPUAI were evaluated in
prioritization of the SRW for three targeting pollutants (sediment, TN, and TP). A total of 12
targeting scenarios were studied (all combinations of targeting methods and targeting pollutants).
Using these four targeting methods, the watershed was categorized into high (H), medium (M),
and low (L) priority areas for all pollutants. The categorization was based on the natural breaks
method of data classification, where different classes are formed based on natural groupings of a
dataset. Similar values of data form a group that attempts to maximize the differences between
the groups, resulting into a relatively substantial difference between the data values of any two
groups.

In addition to the effect of each targeting method on the targeted pollutant, the effectiveness of
BMPs on non-targeted pollutants was also evaluated. For example, a BMP such as no tillage was
implemented in the priority areas aiming at maximum sediment reduction (primary pollutant).

151

Meanwhile, the effectiveness of the BMPs on TN and TP reduction (secondary pollutants) was
also estimated.

5.3.6.1 Concentration Impact Index
High priority areas are identified based on the pollutant concentration level in the reach (Tuppad
and Srinivasan, 2008). This method considers pollutant load from both the subwatershed as well
as the upstream subwatersheds. This method addresses aquatic health localized pollution in low
and high flows.

5.3.6.2 Load Impact Index
The LII method identifies high priority areas based on the total pollutant load from the reach
(Tuppad and Srinivasan, 2008). Like the CII targeting method, this method also considers
pollutants from the subwatershed as well as the upstream watershed. This method represents
cumulative load up to the point of interest, and is useful in establishing waste water treatment
plants and water withdrawal processes.

5.3.6.3 Load Per Subbasin Area Index
High priority areas are identified based on total pollutant load from each subbasin. The total
pollutant load of each subbasin is calculated by multiplying subbasin area with pollutant load per
unit area of that subbasin. This method identifies the subbasins with the largest amount of
pollutant discharge into the stream, resulting in identifying local concerns. Hence, this method
can be used to identify hot spots for excessive loads, which endanger aquatic species.

152

5.3.6.4 Load Per Unit Area Index
The LPUAI method identifies the high priority areas based on average pollutant load per unit
area from each subbasin (Tuppad and Srinivasan, 2008; Giri et al., 2012a). This method is useful
to identify local concerns within an area.

5.3.7 Statistical Analysis
5.3.7.1 Temporal Analysis
A statistical analysis was conducted to detect temporal autocorrelation when comparing BMP
effectiveness on both targeting and non-targeting pollutants with respect to base calibrated
scenarios. Therefore, autocorrelation functions (ACFs) for sediment, TN, and TP with and
without BMPs were created. However, we are only presenting the sediment function here (Figure
5- 2). A similar procedure was performed for TN and TP. Figure 5-2 a, b show that sediment
load datasets without and with BMP scenarios have a strong time series autocorrelation at lag 1
(the time step where the data failed to keep up the trend), which prevents the usage of a general
two-sample paired t-tests for the comparison of significant differences between two datasets
since the independent assumption is violated. Therefore, a modified paired t-test (Napier-Munn
and Meyer, 1999) was used to compare monthly sediment, TN, and TP loads at the watershed
outlet. In addition, comparisons were conducted among the priority areas (H vs. H + M, H+M vs.
H+M+L) for the same time period.

153

high

base
1.0

(a)

0.8

Autocorrelation function

Autocorrelation function

1.0

0.6
0.4
0.2
0.0

(b)

0.8
0.6

0.4
0.2
0.0

1

6

12

16

1

6

12

16

Lag time (months)

Figure 5-2. (a) ACF-sediment base scenario and (b) ACF-high priority area . Bar beyond
the confidence band (dashed horizontal line) show significance at that time lag.

154

Autocorrelation function

1.0

1.0

(c)

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.0

(d)

0.0

1

6

11

16

1

6

11

16

Lag time (months)

Figure 5-2. (c) ACF-pooled series of sediment base and high priority area and (d) Partial
ACF-pooled series. Bar beyond the confidence band (dashed horizontal line) show
significance at that time lag.

In order to hold the normality assumption, a log transformation was performed on each dataset
before the analysis. In addition, an autocorrelation coefficient (Ф) was introduced in the
modified paired t-test autoregressive model of order one AR (1) to nullify the effect of temporal
autocorrelation among the pooled series. According to the AR (1) model, the time series data can
be modeled using the following equation:
X t = ΦX t 1 + e t

(5-1)
155

Where Xt is the mean at time t, Xt-1 is the mean at time t-1, and et is the white noise.
In this modified t-test, a special sampling technique for pooling two series was followed, where
the pollution load output of the current month of base scenario was compared with the load
output of the next month of BMP implementation scenario, and so on (Napier-Munn and Meyer,
1999). For example, the base load output of January was compared with the load output of
February of the no-till BMP implementation scenario on high priority areas. Figure 5-3
represents the pooled time series for pollution generated under base (black circle) and high
priority area (white circle) scenarios. The ACF and partial autocorrelation function (PACF) or
pooled series are shown in Figure 5- 2 c, d. Figure 5-2d demonstrates that the strongest partial
autocorrelation in the time series occurs at lag 1 and is significant for each cycle, which suggests
that the AR (1) model for the pooled series is suitable. After the special sampling, the modified ttest was performed in R using the following equations:

⁄
√

[
Where

(5-2)

√

(

(

)

)]

(5-3)

is the mean difference for all pairs, Sd is the standard deviation of the differences, and n

is the number of pairs

156

Again, normality of the sample dataset was verified. If normality was satisfied, the modified ttest value was taken into account, else the outliers were removed from the datasets and the
modified t-test was performed again.

5.3.8 Spatial Correlation
The spatial correlation analysis was performed to determine the similarities of targeting methods
in identifying priority areas. Each targeting method was used to categorize SRW subbasins into
high, medium, and low priority areas. The possible combinations of two targeting methods were
compared against their categorization of priority area for each pollutant. The total number of
common subbasins was calculated between two targeting methods for all possible combinations
of priority areas and pollutions. A squared contingency table with r = 3 categories was prepared
between two targeting methods. This was repeated for all targeting methods and pollutants. An
example of a contingency table with rounded proportions and marginal sums is presented Table
5-1.
Table 5-1. Contingency table between CII and LPSAI targeting methods based on the TN
targeting scenario. The number in the parenthesis represents the ratio of counts in that
category (e.g. number of subbasins identified as high priority) to the total number of
counts (total number of subbasins).
CII/LPSAI Counts (%)

High

Medium

Low

Sum

High

7 (0.028)

10 (0.039)

16 (0.063)

33 (0.130)

Medium

1 (0.004)

15 (0.059)

48 (0.189)

64 (0.252)

Low

0 (0.000)

5 (0.020)

152 (0.598)

157 (0.618)

Sum

8 (0.031)

30 (0.118)

216 (0.850)

254 (1.000)

157

Table 5-1 was prepared to investigate the agreement (spatial similarity) between two targeting
methods (CII vs. LPSAI). Perfect agreement occurs when all nonzero counts fall in the main
diagonal cells (the targeting methods match exactly). The sum of the proportion in the maindiagonal cell was calculated as the estimated probability of agreement. However, most
contingency tables had a large number of observations for low-low category. Therefore, a high
proportion of main-diagonal entries in the contingency table may be misleading in determining
the actual agreement beyond chance, because it is difficult to create a sufficient predictor model
due to a high concentration of data falling in the low-low category. The Cohen’s kappa
coefficient (parameter to measure agreement between two methods in categorical data) was
calculated as an agreement measure assuming independence between two methods (Cohen,
1960). The large-sample asymptotic variance (variance when sample size approaches infinity)
was also calculated to obtain the 95% confidence interval, as the sample size n is sufficiently
large (Fleiss et al., 1969). Having the Cohen’s kappa coefficient closer to one represents a
stronger agreement between two methods. A zero-value kappa indicates that the agreement
occurs by chance, and a negative value depicts that the agreement is even weaker than expected
by chance. Moreover, since the three categories have ordinal responses from high to low levels,
disagreements of high vs. low levels are considered more severe than that for medium vs. low
levels. The weighted kappa with squared weights can be considered as an intra-class correlation
coefficient (Fleiss and Cohen, 1973). The weighted kappa behaves similarly to kappa, ranging

158

from negative values to 1 with perfect agreement. The 95% confidence intervals from a largesample standard error of weighted kappa were also calculated (Fleiss et al., 1969).

However, summarizing a contingency table with one single index can lead to a significant loss of
information. Therefore, an agreement plus linear-by-linear association model (Agresti, 1988)
was fitted to each contingency table to quantitatively investigate the class agreement and
association between two methods, which is presented in Equation 5-4.

( )

(5-4)

Where i, j values 1, 2, 3 correspond to high, medium, and low priority areas, respectively, for
A

two methods, mij is the expected counts falling cell (i,j), μ is the grand mean, λi is the effect
due to the i

th

B

level of method 1, λj

is the effect due to the j

th

level of method 2, β is the

expected baseline association and ui, uj are the fixed scores of the ordered categories from high
to low level. The uniform association score {ui = i} was considered, which incorporates no extra
assumptions beyond equal intervals between categories; δ(i,j) = δ for i = j and 0 otherwise, with
δ the agreement parameter. This model is a special case of log-linear quasi-symmetric models. It
involves the effects of individual method (A and B), which are nuisance parameters in our case,
and an agreement term which takes values δ only for main-diagonal cells, plus a linear-by-linear
association term with effect β. The interested parameters are β and δ. A non-significant β
indicates that there is no extra association beyond that due to exact agreement, while a nonsignificant δ suggests no extra agreement beyond that due to the baseline association. Both nonsignificances indicate independence of two methods.
159

The model was fitted in the statistical software R (version 2.13.2). The likelihood ratio statistic,
2

G , as a measure of model goodness-of-fit was calculated using R and reported. In general, a
2

smaller G indicates that the fitted values from the model are not deviated from the observed
2

counts. The residual degree of freedom was calculated as 2 using the equation r -(r-1)×2-3 for r
= 3. The interested parameters β and δ with standard errors were estimated and are significant if
the associated p-values are less than 0.1. Furthermore, the local indistinguishable parameter
(Equation 5-5) under uniform association was calculated for each contingency table, which is
(β+2δ)

equal to e

.

(5-5)

5.3.9 Spatiotemporal Variability of Priority Area
We hypothesize watershed priority areas change with time, especially after the establishment of
BMPs. In order to analyze the variability of the priority area after BMP implementation in the
SRW watershed, two BMPs (contour farming and native grass) were selected because the
manner in which they are implemented are quite different. The BMPs were implemented only in
the high priority areas of the SRW for two consecutive years for all targeting methods and
pollutions. For example, native grass is planted on high priority areas before year one (base
scenario), which leads to new high priority areas at the beginning of year two, on which native
grass is subsequently applied. The base scenario was the same for both BMP implementation
scenarios.
160

5.4 RESULTS AND DISCUSSIONS
5.4.1 Determining the Most Effective BMPs for Targeting and Non-targeting Pollutants
The objective of this section was to determine the most effective BMPs for primary and
secondary pollutants when applied on agricultural lands. The results presented in this study are
based on pollutant reduction at the watershed outlet.

5.4.1.1 BMP Pollutant Removal Efficiency (Sediment Targeting Pollutant)
Sediment reduction varies between targeting methods, type of BMP applied, and the priority area
the BMP was applied on. Table 5-2 presents significant differences in sediment reduction for
each BMP between various priority area targets and methods. Native grass, terraces, recharge
structures, residue management (2000 kg/ha), strip cropping, residue management (1000 kg/ha),
and contour farming were effective in sediment reduction; whereas conservation tillage and no
tillage were not effective even when the BMP implementation area increased from H to H+M+L
priority areas (Table 5-2). The insignificant sediment reduction in the case of conservation tillage
and no tillage may be due to the fact that fewer soil disturbances (conservation tillage) and
minimum soil disturbances (no tillage) did not create a larger impact of sediment delivery into
the stream. Residue management (0 kg/ha) showed a mixed effectiveness in sediment reduction
based on the priority areas. Native grass, terraces, and recharge structure reduced significant
amounts of sediment when the base scenario was compared to priority areas as well as among the
priority areas, most likely due to the intensive nature of the application of these BMPs. When the
sediment reduction in priority areas are compared among the targeting methods, no BMPs

161

reduced a significant amount of sediment in the high priority area of the CII targeting methods
while some BMPs reduced significant amount of sediment in the high priority area of other
targeting methods (Table 5-2). This result was due to less area of high priority implementation in
CII (Table 5-3) than the other targeting methods and using a concentration basis whereas other
targeting methods are based on a pollutant load basis. When priority areas (H+M vs. H+M+L)
were compared, only two and three BMPs reduced significant amounts of sediment compared to
the H+M priority area in the LPUAI and LPSAI targeting methods, respectively (Table 5-2). It
indicates that BMP application on H+M would be the wise decision for both the LPUAI and
LPSAI targeting methods, rather than additionally including implementation on low priority
areas. Overall, LPSAI had higher significant sediment reduction counts (highlighted cells in
Table 5-2), indicating that this method is the most effective.

162

Table 5-2. P-values for BMPs based on the AR (1) model for different targeting methods based on sediment targeting
(sediment).
Scenario
CT
NT
RM 0
CF
RM
SC
RM
RS
T
NG
1000
2000
CII Targeting Method
B vs. H
2.3E-01 2.3E-01 1.7E-01 0.14
1.5E-01
1.3E-01 1.4E-01 1.5E-01 1.0E-01 1.7E-01
B vs. H+M
1.4E-01 1.4E-01 6.3E-02 5.9E-04 5.2E-05
1.7E-06 2.9E-06 8.0E-08 6.4E-15 3.3E-13
B vs. H+M+L
1.1E-01 1.0E-01 2.8E-02 1.2E-05 2.0E-07
3.1E-10 8.0E-10 1.1E-12 0.0E+00 0.0E+00
H vs. H+M
1.4E-01 1.4E-01 9.3E-02 1.4E-03 1.3E-04
6.6E-06 8.5E-06 2.1E-07 5.2E-10 9.5E-13
H+M vs. H+M+L 2.0E-01 2.0E-01 1.2E-01 1.5E-02 7.6E-03
7.2E-04 1.4E-03 9.7E-05 3.4E-09 0.0E+00
LII Targeting Method
B vs. H
B vs. H+M
1.2E-01 1.2E-01 3.8E-02 7.9E-05 2.8E-06
3.0E-08 4.5E-08 7.2E-10 0.0E+00 0.0E+00
B vs. H+M+L
1.1E-01 1.0E-01 2.8E-02 1.2E-05 2.0E-07
3.1E-10 8.0E-10 1.1E-12 0.0E+00 0.0E+00
H vs. H+M
H+M vs. H+M+L 2.3E-01 2.4E-01 1.9E-01 7.6E-02 6.5E-02
2.1E-02 3.4E-02 2.3E-01 3.1E-05 4.1E-13
LPSAI Targeting Method
B vs. H
1.6E-01 1.5E-01 5.5E-02 1.6E-03 4.4E-04
4.5E-05 7.7E-05 2.7E-05 1.4E-07 5.6E-08
B vs. H+M
1.2E-01 1.2E-01 3.1E-02 2.8E-05 9.1E-07
3.1E-09 5.3E-09 1.4E-10 0.0E+00 0.0E+00
B vs. H+M+L
1.1E-01 1.0E-01 2.8E-02 1.2E-05 2.0E-07
3.1E-10 8.0E-10 1.1E-12 0.0E+00 0.0E+00
H vs. H+M
1.9E-01 2.0E-01 1.4E-01 1.4E-02 4.9E-03
4.7E-04 6.1E-04 1.0E-04 4.4E-08 7.6E-13
H+M vs. H+M+L 2.4E-01 2.5E-01 2.2E-01 1.2E-01 1.1E-01
5.9E-02 9.6E-02 1.3E-02 1.0E-03 2.9E-10
LPUAI Targeting Method
3.
B vs. H
2.1E-01 2.1E-01 9.8E-02 3.8E-02
1.8E-02 2.7E-02 3.3E-02 5.4E-03 7.8E-03
66E-02
B vs. H+M
1.1E-01 1.1E-01 2.8E-02 1.4E-05 2.8E-02
6.0E-10 1.7E-09 2.5E-11 0.0E+00 0.0E+00
B vs. H+M+L
1.1E-01 1.0E-01 2.8E-02 1.2E-05 2.8E-02
3.1E-10 8.0E-10 1.1E-12 0.0E+00 0.0E+00
H vs. H+M
1.3E-01 1.3E-01 8.6E-02 4.0E-04 1.6E-05
1.4E-07 2.5E-07 2.2E-09 1.5E-15 0.0E+00
2.08EH+M vs. H+M+L 2.5E-01 2.6E-01 2.5E-01 1.7E-01 1.9E-01
1.3E-01 1.6E-01 4.4E-02 3.1E-02
01

163

*Cells

having darker color are significantly different at p <0.05 level of significance
based on the AR (1) model. Abbreviations: conservation tillage (CT), no tillage (NT),
residue management 0 kg/ha (RM 0), contour farming (CF), residue management 1000
kg/ha (RM 1000), strip cropping (SC), residue management 2000 kg/ha (RM 2000),
recharge structures (RS), terraces (T), and native grass (NG).
Table 5-3. BMP implementation area for sediment targeting scenario.
Targeting method
CII
LII
LPSAI
LPUAI

H (ha)
27,648
143,319
84,977

H+M (ha)
230,043
232,351
304,258
334,345

H+M+L(ha)
345,423
345,423
345,423
345,423

5.4.1.2.1 BMP Removal Efficiency for TN
The TN reduction trend (when targeting sediment) among the BMPs was similar to sediment
reduction, varying between the BMPs, priority areas, and the targeting methods (Table A-1in the
Appendix). Native grass, terraces, recharge structure, residue management (1000 and 2000
kg/ha), strip cropping, and contour farming reduced significant amounts of TN. Conservation
tillage and no tillage were not effective in TN reduction in any combination of priority area and
targeting method. The statistically insignificant TN reduction in these BMPs may be due to an
increase in organic matter (nutrients) because of diminished soil disturbances, while the nutrients
in organic matter can still be released into the soil.

Significant differences in TN reduction while targeting sediment between various priority areas
and methods are presented in Table 5-4. TN reduction was insignificant in high priority areas vs.
base for CII, while significant TN reduction was observed in the high priority areas of LPSAI
and LPUAI for most BMPs. This result was due to the considerably smaller BMP
implementation area in CII compared to the LPSAI and LPUAI (Table 5-3), and because CII
164

targets high concentration rather than load, leading to less load reduction at the outlet. When
comparing to the priority areas (H+M vs. H+M+L) among the targeting methods, most of the
BMPs in CII reduced significant amounts of TN, whereas the BMPs did not reduce significant
amounts of TN both in the LPSAI and LPUAI targeting methods. This depicts BMP
implementation in H+M+L of the LPSAI and LPUAI targeting methods does not significantly
reduce TN compared to H+M application. Therefore, we can optimize TN reduction by applying
BMPs only to H+M priority areas of LPSAI and LPUAI and still achieve similar results as
applying BMPs on all areas.

165

Table 5-4. P-values for BMPs based on the AR (1) model for different targeting methods based on sediment targeting for nontargeted TN.
Scenario

CT

NT

RM 0

CF

RM
1000

SC

RM
2000

RS

T

CII Targeting Method
1.1E-01 1.4E-01 9.3E-02 1.4E-01 4.7E-01 6.1E-02
3.1E-04 7.2E-03 2.5E-05 9.5E-03 2.3E-03 5.7E-08
3.1E-08 6.1E-04 3.8E-10 3.9E-04 1.1E-04 6.1E-12
2.7E-03 2.7E-02 4.1E-04 1.7E-02 8.4E-03 3.3E-06
2.8E-02 8.4E-02 1.9E-02 1.0E-01 5.7E-02 2.9E-03
LII Targeting Method
B vs. H
B vs. H+M
5.9E-01 5.9E-01 3.1E-02 5.8E-06 1.4E-03 2.3E-06 9.2E-04 6.5E-03 1.7E-09
B vs. H+M+L
8.4E-01 8.7E-01 3.3E-02 3.1E-08 6.1E-04 5.8E-02 3.9E-04 1.1E-04 6.1E-12
H vs. H+M
H+M vs. H+M+L 2.6E-01 3.9E-01 2.5E-01 1.0E-01 1.7E-01 5.8E-02 1.8E-01 2.2E-02 1.6E-02
LPSAI Targeting Method
B vs. H
5.0E-01 5.1E-01 4.4E-02 5.6E-04 6.3E-03 1.2E-04 4.7E-03 4.3E-02 2.7E-06
B vs. H+M
6.2E-01 6.4E-01 3.2E-02 1.9E-07 8.8E-04 3.2E-07 2.6E-06 2.9E-03 7.4E-11
B vs. H+M+L
8.4E-01 8.7E-01 3.3E-02 3.1E-08 6.1E-04 3.8E-10 3.9E-04 1.1E-04 6.1E-12
H vs. H+M
3.6E-01 3.3E-01 2.0E-01 3.5E-02 9.0E-02 8.1E-03 3.6E-03 4.5E-02 8.1E-04
H+M vs. H+M+L 4.2E-01 3.6E-01 4.4E-01 2.1E-01 4.5E-01 3.3E-01 5.3E-01 5.2E-02 1.8E-01
LPUAI Targeting Method
B vs. H
4.1E-01 4.1E-01 9.2E-02 1.4E-02 4.9E-02 9.4E-03 4.9E-02 1.0E-01 2.6E-03
B vs. H+M
6.3E-01 6.5E-01 2.9E-02 8.1E-08 6.0E-04 1.4E-09 6.0E-04 2.3E-03 5.5E-08
B vs. H+M+L
8.4E-01 8.7E-01 3.3E-02 3.1E-08 6.1E-04 3.8E-10 3.9E-04 1.1E-04 6.1E-12
H vs. H+M
5.0E-01 5.0E-01 4.6E-01 7.7E-04 1.2E-02 8.4E-05 1.2E-02 1.3E-02 1.8E-07
H+M vs. H+M+L 4.0E-01 3.5E-01 4.8E-01 4.3E-01 5.1E-01 4.4E-01 2.8E-01 3.1E-02 3.1E-01
Cells having darker color are significantly different at p <0.05 level of significance based on AR (1) model.
B vs. H
B vs. H+M
B vs. H+M+L
H vs. H+M
H+M vs. H+M+L

3.4E-01
1.4E-01
1.1E-01
5.3E-01
4.4E-01

3.4E-01
1.4E-01
1.0E-01
5.4E-01
3.9E-01

1.9E-01
6.3E-02
2.8E-02
1.9E-01
6.3E-02

166

NG

5.9E-02
2.9E-08
1.2E-08
5.9E-07
1.6E-02
9.5E-09
1.2E-08
4.3E-02
8.4E-07
9.8E-10
1.2E-08
4.0E-04
1.2E-01
3.3E-03
6.3E-10
1.2E-08
4.8E-07
1.8E-01

5.4.1.2.2 BMP Pollutant Removal Efficiency for TP (Targeting Sediment)
TP reduction also varies among the BMPs, types of priority area, and targeting methods. Table 55 represents the p-value of TP reduction by BMPs in different priority areas and targeting
methods based on the AR (1) model. BMPs such as native grass, terraces, recharge structure,
residue management (1000 and 2000 kg/ha), strip cropping, and contour farming reduced
significant amounts of TP, whereas conservation tillage, no till, and residue management (0
kg/ha) did not reduce when targeting sediment. The insignificant TP reduction by these BMPs
associates with insignificant sediment reduction, likely due to their related transport mechanisms
(Table 5-2). Similar results were demonstrated by Bundy et al. (2001), where dissolved
phosphorus concentration from a corn field increased when implementing no tillage practices.

When TP reduction was compared among the priority areas with base scenarios, a significant TP
reduction was observed by BMPs in all the priority areas among all targeting methods (Table 55). Among the targeting methods CII was most effective in counts (combination of significant
TP reductions for priority area/BMPs), indicating its overall superior TP reduction abilities
among targeting methods (when targeting sediment). TP reduction compared among H+M vs.
H+M+L of the LPUAI targeting method, an insignificant TP reduction was found in most of the
BMPs due to the smallest increase in BMP implementation area from H+M to H+M+L (Table 53).

167

Table 5-5. P-values for BMPs based on the AR (1) model for different targeting methods based on sediment targeting for nontargeted TP.
Scenario

CT

NT

RM 0

CF

RM
1000

SC

RM
2000

RS

T

CII Targeting Method
1.3E-02 2.0E-02 6.9E-03 1.8E-02 5.0E-03 3.2E-03
2.5E-06 1.7E-03 1.5E-09 8.9E-05 1.0E-10 0.0E+00
2.0E-08 2.8E-04 1.8E-14 9.7E-06 0.0E+00 0.0E+00
2.4E-05 5.3E-03 5.3E-08 4.7E-04 6.3E-09 1.8E-13
1.3E-02 5.9E-02 1.1E-03 5.1E-02 4.7E-06 5.1E-06
LII Targeting Method
B vs. H
B vs. H+M
2.8E-01 9.4E-01 6.3E-01 5.4E-07 4.4E-04 4.3E-11 2.7E-05 3.2E-11 0.0E+00
B vs. H+M+L
3.4E-01 6.3E-01 8.9E-01 2.0E-08 2.8E-04 1.8E-14 9.7E-06 0.0E+00 0.0E+00
H vs. H+M
H+M vs. H+M+L 8.6E-02 1.4E-01 4.2E-01 2.5E-02 1.1E-01 2.9E-03 8.3E-02 8.6E-02 3.8E-05
LPSAI Targeting Method
B vs. H
1.8E-01 6.8E-01 4.9E-01 1.1E-05 1.6E-03 4.2E-08 2.5E-04 1.5E-05 4.1E-11
B vs. H+M
3.1E-01 7.7E-01 8.7E-01 1.0E-07 2.9E-04 5.2E-13 1.2E-05 3.7E-13 0.0E+00
B vs. H+M+L
3.4E-01 6.3E-01 8.9E-01 5.8E-06 2.8E-04 1.8E-14 9.7E-06 0.0E+00 0.0E+00
H vs. H+M
1.2E-01 2.4E-01 3.2E-01 1.3E-02 1.4E-02 4.1E-04 5.7E-03 1.3E-05 1.3E-07
H+M vs. H+M+L 7.7E-02 1.0E-01 3.5E-01 7.8E-02 1.8E-01 3.3E-02 2.2E-01 7.2E-05 4.8E-03
LPUAI Targeting Method
B vs. H
9.9E-02 2.4E-01 1.4E-01 4.1E-04 6.4E-03 5.5E-05 3.5E-03 4.0E-04 2.7E-05
B vs. H+M
3.2E-01 7.1E-01 9.4E-01 1.5E-06 2.0E-04 7.7E-14 7.5E-06 8.5E-14 0.0E+00
B vs. H+M+L
3.4E-01 6.3E-01 8.9E-01 2.0E-08 2.8E-04 1.8E-14 9.7E-06 0.0E+00 0.0E+00
H vs. H+M
2.3E-01 7.7E-01 9.4E-01 3.7E-05 3.4E-03 5.0E-09 4.1E-04 8.7E-08 6.6E-16
9.15EH+M vs. H+M+L 7.5E-02 9.0E-02 3.3E-01 1.1E-01
7.9E-02 2.3E-01 2.3E-04 3.1E-01
02
Cells having darker color are significantly different at p <0.05 level of significance based on AR (1) model.
B vs. H
B vs. H+M
B vs. H+M+L
H vs. H+M
H+M vs. H+M+L

6.9E-02
2.6E-01
3.4E-01
2.2E-01
7.7E-02

1.1E-01
9.3E-01
6.3E-01
8.0E-01
1.2E-01

8.1E-02
9.2E-01
8.9E-01
9.4E-01
1.1E-01

168

NG

7.7E-15
0.0E+00
0.0E+00
4.0E-04
3.3E-01
2.2E-16
0.0E+00
1.0E-04
1.2E-10
2.2E-16
0.0E+00
6.5E-07
1.6E-02
5.4E-05
0.0E+00
0.0E+00
1.0E-12
7.4E-02

This indicates that the implementation of a BMP on only H+M priority areas of the LPUAI
targeting method is enough to make a significant TP reduction, so additional resources do not
need to be invested on low priority areas to get achieve significant reduction.

5.4.1.2 Targeting Component (TN)
In targeting TN, similar trends were observed for the non-targeted components as in the case for
sediment targeting (and its effects on non-targeted components). Table 5-6 describes the
summary of the targeting component (TN) and non-targeting components (sediment and TP) for
each BMP, targeting method, and priority comparison. In addition, the total number of
significant reductions for each targeting method and component combination are presented,
where the method with the highest number is the most effective for the pooled set of BMPs and
priority area combinations. This summarizes results obtained from the supplementary material
Table A-1, Table A-2, and Table A-3 in the appendix. All BMPs excluding conservation tillage
and no tillage reduced a significant amount of TN, sediment, and TP. Residue management 0
(kg/ha) exhibited mixed effect (significant reduction in some priority areas/targeting methods
and insignificant reduction in others) in sediment and TN reduction.

169

Table 5-6. Summary of targeting component (TN) and non-targeting components (sediment
and TP).
Targeting
component
TN
CF,RM 1000,
SC, RM 2000,
RS, T, NG
CT, NT

Significant
BMPs
Insignificant
BMPs
Mixed effects
Significant
reductions

RM 0
B vs. H
B vs. H+M
B vs. H+M+L
H vs. H+M
H+M vs.
H+M+L

Total
significant
reductions

CII
LII
LPSAI
LPUAI

CII, LPSAI,
LPUAI
CII, LII,
LPSAI, LPUAI
CII, LII,
LPSAI, LPUAI
CII, LPSAI,
LPUAI
LII

24/30
11/30
23/30
23/30

Non-targeting components
Sediment
CF, RM 1000,
SC, RM 2000,
RS, T, NG
CT, NT

TP
CF, RM 1000,
SC, RM 2000,
RS, T, NG
CT, NT, RM 0

RM 0
CII, LPSAI,
LPUAI
CII, LII,
LPSAI, LPUAI
CII, LII,
LPSAI, LPUAI
CII, LPSAI,
LPUAI
CII, LII,
LPSAI,
LPUAI
23/30
9/30
23/30
23/30

CII,LPSAI,
LPUAI
CII, LII,
LPSAI, LPUAI
CII, LII,
LPSAI, LPUAI
CII, LPSAI,
LPUAI
CII, LII,
LPSAI, LPUAI
21/30
14/30
21/30
21/30

Table 5-7. BMP implementation area for TN targeting scenario.
Targeting method
CII
LII
LPSAI
LPUAI

H(ha)
242318
150276
73943

H+M(ha)
319972
33865
321195
343408

170

H+M+L(ha)
345423
345423
345423
345423

Examining significant reduction when comparing priority areas, a significant TN reduction was
observed between H+M and H+M+L based on the LII targeting method, which indicates that
including BMP implementation on low priority areas significantly impacts reduction. However,
all BMPs (except recharge structure in CII and LPSAI) did not show a significant TN reduction
based on the rest of the three targeting methods due to a smaller increase of the BMP application
area from H+M to H+M+L priority area (Table 5-7). This suggests that the application of BMPs
up to the H+M priority area is effective in TN reduction based on the CII, LPSAI, and LPUAI
targeting methods. Comparing all targeting methods, CII was most effective in reducing targeted
TN (24/30) and non-targeting sediment (23/30) and TP (21/30), where the counts indicate the
combinations of BMP and targeted areas that reduced significant amounts of pollutant from the
no-BMP scenario. For example, for CII reducing non-targeting sediment TN (23/30), seven
BMPs had significant reductions in base vs. H, eight BMPs had significant reductions in both
base vs. H+M and base vs. H+M+L, which adds to 23 total significant reductions.

5.4.1.3 Targeting Component (TP)
The results of TP targeting and non-targeting components (sediment and TN) are presented in
Table 5-8. This table summarized the results obtained from the supplementary material Table A4, Table A-5, and Table A-6 in the Appendix. Overall, the majority of BMPs were effective in
reduction of TP and the non-targeted components, while conservation tillage and no tillage did
not exhibit significant reductions. Contour farming had mixed results when reducing TN, while
residue management and contour farming reduced significant amounts of TP and sediment.

171

Table 5-8. Summary of targeting component (TP) and non-targeting components (sediment
and TN).
Targeting
Non-targeting components
component
TP
Sediment
TN
Significant
CF, RM 1000,
CF, RM 1000,
RM 1000, SC,
BMPs
SC, RM 2000,
SC, RM 2000,
RM 2000, RS,
RS, T, NG
RS, T, NG
T, NG
Insignificant
CT, NT, RM 0
CT, NT
CT, NT
BMPs
Mixed effect

_

RM 0

CF, RM 0

CII, LPSAI,
LPUAI
CII, LII, LPSAI,
LPUAI

CII, LPSAI,
LPUAI
CII, LII,
LPSAI, LPUAI

CII, LPSAI,
LPUAI
CII, LII, LPSAI,
LPUAI

CII, LII,
LPSAI, LPUAI
CII, LPSAI,
LPUAI
CII, LII,
LPSAI, LPUAI

CII, LII, LPSAI,
LPUAI
LPSAI, LPUAI

H+M vs.
H+M+L

CII, LII, LPSAI,
LPUAI
CII, LPSAI,
LPUAI
CII, LII, LPSAI,
LPUAI

CII

21/30

21/30

20/30

LII

14/30

14/30

15/30

LPSAI
LPUAI

21/30
21/30

22/30
18/30

20/30
20/30

B vs. H
B vs. H+M
Significant
reduction

B vs. H+M+L
H vs. H+M

Total
significant
reductions

CII, LII, LPSAI,
LPUAI

However, it showed a mixed result among targeting methods in reducing TN. Residue
management (0 kg/ha) did not reduce significant amounts of TP but had mixed results among
targeting methods in the reduction of sediment and TN. Overall, CII and LPSAI targeting
methods were most effective in reducing TP and the non-targeting components in this scenario.
None of the BMPs reduced significant amounts of sediment and TN in the CII high priority area;
however, they reduced significant amounts of TP, suggesting that the BMP implementation for
targeting TP does not have the added benefits or reduction of non-targeting pollutants. Except for
172

the LII targeting method, no other targeting methods reduced significant amounts of TN between
the H+M and H+M+L priority areas. This demonstrates that BMP implementation up to the
H+M priority area of these targeting methods is adequate for significant TN reduction. Out of all
targeting methods, LPSAI had the highest number of significant reduction options for sediment
and TP targeting scenarios, while CII, LPSAI, and LPUAI had highest number of options for TN
targeting scenario.

5.4.2 Spatial Correlation among the Priority Methods
The objective of this section was to determine the spatial correlation between priority areas of
targeting methods based on the targeting pollutant. In this manner it can be determined if there
are statistically significant differences between the priority areas that a targeting method
identifies. The agreement between two targeting methods is determined through kappa and
weighted kappa coefficients and an agreement plus linear-by-linear association model. The
kappa and weighted kappa coefficients being closer to one indicate stronger agreement between
two methods, while zero value represents that agreement occurs by chance, and a negative value
depicts the agreement as weaker than expected by chance. A stronger agreement was found
between CII and LPSAI, and LPSAI and LPUAI for sediment and TN targeting scenarios.
However, there is no agreement between most of the targeting methods for the TP targeting
scenario.

5.4.2.1 Sediment Targeting Scenario
Parameters required for the spatial correlation analysis between all targeting methods while
considering sediment as the targeting pollutant are presented in A-7 (Appendix). When the CII

173

2

and LPSAI targeting methods were compared, the G was relatively small (9.65), indicating that
the fitted value from the model is nearly equal to the actual value. This is demonstrated in Table
5-9, which compares the common subbasins selected for high/medium/low priority between CII
and LPSAI from SWAT (actual) and by the developed spatial model. For example, the fitted
value for low-low is 221.9, whereas the actual value is 223.
Table 5-9. Comparison of actual vs. modeled counts based on the agreement plus linear-bylinear association model: CII and LPSAI targeting methods (sediment).
CII / LPSAI High counts:
Medium counts:
Low counts:
Actual (Model)
Actual (Model)
Actual (Model)
High

2 (3.052)

0 (0.461)

2 (0.488)

Medium

4 (1.436)

8 (8.00)

2 (4.564)

Low

2 (3.512)

11 (10.539)

223 (221.948)

The estimated β is 0.79 with a moderately significant p-value (0.09) under the 90% confidence
interval (Table A-7 in the Appendix). The estimated δ is 1.41 with a highly significant p-value
(0). Therefore, the calculated local indistinguishable parameter (τ) i+1 = 36.91 (Table A-7) can
be interpreted as: if LPSAI gives a result in category i+1 rather than i (say medium rather than
high, or low rather than medium), CII is about 37 times more likely to give a result in the
category i+1 rather than i, which suggests a strong agreement. This result is further supported by
the results provided by the kappa coefficient 0.5 (0.33, 0.67) and weighted kappa coefficient 0.56
(0.36, 0.76) (Table A-7). Both the kappa and weighted kappa confidence intervals do not contain
zero, indicating that there is a significant agreement beyond that expected by chance between the
categorization of the watershed into H, M, and L priority areas by CII and LPSAI. Therefore, it
can be concluded that these two targeting methods select similar subbasins to be H, M, and L
priority.
174

Excluding CII vs. LPSAI and LPSAI vs. LPUAI, all other comparisons between targeting
methods in the sediment scenario did not show significant spatial agreement because p-values of
either β or δ were not significant (< 0.10). Similarity between LPSAI and LPUAI exists likely
because they both target sediment loads from the subbasin. In the majority of the comparisons,
the targeting methods behave differently when targeting sediment by selecting diverse priority
areas.

5.4.2.2 TN Targeting Scenario
Like the sediment targeting scenario, the CII and LPSAI targeting methods are presented. The
2

G was 0.33 (Table A-7 in the Appendix), indicating the model fitting value is approximately
equal to the actual values, which can be observed in Table 5-10 comparing the actual and model
priority counts for the targeting methods. The estimated β is highly significant, whereas the
estimated δ is moderately significant under the 90% CI. Hence, the chance of CII giving
selecting a subbasin in category i+1 rather than i (medium rather than high, or low rather than
medium) is 10 times higher when LPSAI selects a subbasin in category i+1 rather than i,
suggesting a strong agreement between the methods.

Similarly, LPSAI selecting a subbasin in category i+1 rather than i is 21 times higher when
LPUAI selects a subbasin in category i+1 rather than i. As was the case with sediment targeting,
the CII vs. LPSAI and LPSAI vs. LPUAI are the only methods with significant spatial agreement
for TN targeting. Therefore, all other combinations of targeting methods are likely to produce

175

differing H, M, and L priority areas, illustrating the necessity of selecting a targeting method that
aligns with stakeholders’ specific goals in mind.
Table 5-8. Comparison of actual versus modeled counts based on agreement plus linear-bylinear association model: CII and LPSAI targeting methods (TN).
CII / LPSAI
High counts:
Medium counts:
Low counts:
Actual (Model)
Actual (Model)
Actual (Model)
High

7 (6.86)

10 (10.29)

16 (15.85)

Medium

1 (0.991)

15 (15)

48 (48.009)

Low

0 (0.150)

5 (4.71)

152 (152.140)

5.4.2.3 TP Targeting Scenario
2

Unlike the sediment and TN targeting scenario, the G is zero between all targeting methods
except the CII vs. LII and CII vs. LPUAI targeting methods (Table A-7 in the Appendix). This
result suggests that the model is saturated (model predictions are similar to actual values) as G

2

is zero due to data sparseness (zeroes in non-diagonal elements of the contingency tables). The
sparseness of the data causes computational problems, which leads into problems in model
fitting. Hence, the parameters (β, δ, and τ) estimated for all pair wise comparisons except CII vs.
LII and CII vs. LPUAI are unreliable. Only, the kappa and weighted kappa coefficients were
used to measure the agreement for such pairs. However, when the CII and LII targeting methods
2

were compared, the G was found to be 1.53 (Table A-7), suggesting a good model fit, which is
further supported from the observation in Table 5-11.

176

Table 5-9. Comparison of actual versus modeled counts based on agreement plus linear-bylinear association model for CII and LII targeting methods (TP).
C2 / LII
High counts:
Medium counts:
Low counts:
Actual (Model)
Actual (Model)
Actual (Model)
High
0 (0.305)
4 (3.101)
21 (21.594)
Medium

0 (0.288)

0 (0)

14 (13.712)

Low

6 (5.406)

36 (36.899)

173 (172.695)

However, both β and δ are highly non-significant (Table A-7 in the Appendix). This is further
supported by the negative kappa and weighted kappa coefficients, indicating a strong
disagreement between the categorization of priority areas between CII and LII targeting
methods. The disagreement is likely due to the CII targeting method being based on pollutant
concentration, while LII is based on the pollutant load. Overall, no conclusions can be drawn for
the spatial correlation of most targeting methods for TP. The exceptions are CII vs. LII and CII
vs. LPUAI, which were spatially dissimilar in selecting H, M, and L areas.

5.4.3 Spatiotemporal Variability of Priority Areas
The objective of this section was to understand the variability of priority areas with respect to the
time of BMP implementation. Native grass and contour farming were selected because these
BMPs demonstrated the highest and lowest pollutant reduction efficiencies among studied BMPs
with significant reductions. The process begins by implementing BMPs in the high priority areas
based on the CII, LPSAI, and LPUAI targeting methods for two successive years. The LII
targeting method was not used in this analysis due to lack of agricultural lands identified as high
priority. Overall, the high priority areas of both sediment and TP scenarios changed faster over
implementation years than in the TN scenario. A greater variability of high priority areas was
observed in native grass compared to contour farming in all scenarios and targeting methods.

177

The findings of this section will help policymakers and stakeholders to develop better
implementation strategies based on time and location of BMPs installment. This scenario reflects
the manner in which implementation projects are completed over a number of years, which is
closer to reality.

5.4.3.1 Sediment Targeting Scenario
Subbasin priority area designations were altered in each year of implementation for contour
farming and native grass for all applicable targeting methods, as demonstrated in the Appendix
Figure A-8, Figure A-9, and Figure A-10. Considering year zero, BMPs are only implemented on
the high priority areas. When native grass is implemented, most high priority areas are
reclassified to medium or low priority areas. However, in order to achieve the highest rate of
pollution reduction while targeting smaller areas, the area of study was reclassified to high,
medium, and low again for the next round of BMP implementation. Meanwhile, the specific
pollutant concentration or load intervals of each priority change between each year of
implementation. Contour farming is not as effective as overall, as most high priority areas did
not change until year two, when they are converted to medium priority. In addition, some
medium priority areas become high priority because the definition of high priority changes with
time.

For CII, the high and medium priority areas of both BMPs were nearly equal after the first year
of BMP implementation (Figure A-8). However, different priority areas were observed for both
BMPs after the second year of implementation. A greater high priority area was found in contour
farming compared to the native grass applied scenario. In fact, the high priority area of native

178

grass is not visible in the map (year two) due to the very small area selected as high priority in
year zero. This result was due to greater sediment reduction efficiency of native grass compared
to contour farming. Variability of the priority area changes rapidly for most effective BMPs (e.g.
native grass); while it changes slowly for less effective BMPs (e.g. contour farming).

Medium and high priority areas of both BMPs were different after BMP implementation under
LPSAI (Figure A-9). A comparatively smaller high priority area was observed in the native grass
implementation scenario compared to contour farming due to greater sediment reduction in case
of native grass. Also, a distinct change in the high priority area of native grass was observed
between the first and second year of BMP implementation. However, a minimal change in the
high priority area was observed in contour farming between the first and second year of BMP
implementation.

Like the CII and LPSAI targeting methods, less high priority areas were found in native grass
compared to contour farming after one year of BMP implementation under LPUAI (Figure A10). However, after the second year of BMP implementation, the high priority areas of contour
farming were smaller than the native grass. This was due to the conversion of more high priority
areas (after year one) of contour farming implementation into medium priority areas and
alteration of medium priority areas (after year one implementation) of native grass into high
priority areas.

179

5.4.3.2 TN Targeting Scenario
The pattern of priority areas in this scenario was similar to the sediment targeting scenario, as
shown in the Appendix Figure A-11, Figure A-12, and Figure A-13. Based on the CII targeting
method, the high priority area of contour farming was greater than the high priority area of native
grass for both the first and second years of implementation (Figure A-11). This was due to
comparatively less TN reduction by contour farming than native grass (Table A-2). In the LPSAI
targeting method, smaller high priority areas were observed in native grass compared to the
contour farming applied scenario after year one of implementation (Figure A-12). However, a
comparatively greater high priority area was found in the native grass applied scenario compared
to the contour farming scenario after the second year due to presence of greater medium priority
area after one year of BMP implementation and conversion of those medium priority areas to
high priority areas after the second year. Like the CII targeting method, a similar pattern of high
priority areas was found among the BMPs based on the LPUAI targeting method (Figure A-13).
The conversion of a high priority area to a medium priority area by BMPs was slower than the
sediment targeting scenario for all targeting methods.

5.4.3.3 TP Targeting Scenario
CII high priority areas of native grass (year one and two) were smaller compared to the contour
farming priority area (Figure A-14 in the Appendix) after the first year due to comparatively
more significant TP reduction under native grass. However, the medium priority area of native
grass was greater than contour farming for both of the years due to the conversion of low to
medium priority during both years of BMP implementation. In the LPSAI (Figure A-15 in the
Appendix) and LPUAI (Figure A-16 in the Appendix) targeting methods, native grass had less

180

high priority areas compared to contour farming in both years due comparatively greater TP
reduction than for native grass. In most of the native grass scenarios, high priority areas
decreased gradually to the point where less high priority areas were observed by the end of
second year. However, the high priority areas of contour farming did not show a significant
change by the end of second year due to less effective TP reduction. Like the sediment targeting
scenario, a faster conversion of a high priority area to medium priority area was observed in this
scenario by native grass.

5.5 CONCLUSION
The objectives of this research were to 1) determine the most effective BMPs both for targeting
and non-targeting pollutants based on different targeting methods while minimizing area devoted
to BMP implementation; 2) evaluate the spatial correlation among the targeting methods in
categorization of priority area (high, medium, and low) based on targeting pollutant; and 3)
assess the spatiotemporal variability of CSAs. Four targeting methods (CII, LII, LPSAI, and
LPUAI) were used in SWAT to prioritize BMP implementation in the Saginaw River Watershed
for sediment, TN, and TP reduction at the watershed outlet. Ten BMPs, namely conservation
tillage, no till, residue management (0 kg/ha), contour farming, residue management (1000
kg/ha), residue management (2000 kg/ha), strip cropping, recharge structures, terraces, and
native grass were implemented for each targeting method.

BMP effectiveness among targeting methods and priority areas for targeted and non-targeted
pollutants was compared using autocorrelation functions and modified paired t-tests that account
for temporal autocorrelation. Most BMPs (excluding no tillage and conservation tillage)
181

exhibited at least some significant pollutant reduction (targeted and non-targeted) for all
targeting methods. As BMP implementation areas increased from high+medium to
high+medium+low priority areas, significant pollutant reduction was not present. This indicates
that BMP implementation up to the high+medium priority area is enough to achieve significant
pollutant reduction. Overall, LPSAI had the most significant reductions for sediment and TP,
while CII was most effective for targeting TN.

Spatial correlation among the targeting methods was determined by kappa and weighted kappa
coefficients with an agreement plus linear-by-linear association model. A strong agreement was
found between LPSAI and LPUAI targeting methods when categorizing the priority areas into
high, medium, and low in the SRW for both sediment and TN targeting scenarios, likely because
these methods target pollutant loads. However, a similar result was found between the CII and
LPSAI targeting methods even though both methods are based on different principles (CII is
based on pollutant concentration in the reach, LPSAI is based on pollutant load in the subbasin)
for both sediment and TN targeting scenarios. In the TP targeting scenario, strong disagreement
between CII and LII was observed, indicating that these methods target TP differently.
Examining spatial correlation between targeting methods highlights their similarities and
differences. By considering spatial correlation, policymakers and stakeholders can better
understand which critical source area targeting method will achieve their specific watershed
management plan goals.

The spatiotemporal variability of priority areas primarily depends on the effectiveness of a BMP.
By implementing BMPs on high priority areas in consecutive years, realistic BMP

182

implementation scenarios were achieved. In all targeting methods, the defining factors of priority
area changed with each implementation year. A distinct change in the high priority area of native
grass was observed by the end second year, while a minimal change in high priority areas was
found in the case of contour farming due to greater pollutant reduction efficiency of native grass.
Therefore, the high priority area of native grass changes rapidly compared to contour farming.
Examining spatiotemporal variability emulates realistic BMP implementation plans and allows
for the understanding of how critical source areas will be altered as BMPs are put into practice.

183

6. APPLICATION OF ANALYTICAL HIERARCHY PROCESS FOR EFFECTIVE
SELECTION OF AGRICULTURAL BEST MANAGEMENT PRACTICES.

6.1 ABSTARACT
In this study an analytical hierarchy process (AHP) was used for ranking best management
practices (BMPs) in the Saginaw river watershed based on environmental, economic, and social
factors. Three spatial targeting methods were used for placement of BMPs on critical source
areas (CSAs). The environment factors include sediment, total nitrogen, and total phosphorus
reductions at the subbasin level and the watershed outlet. Economic factors were based on total
BMP cost, including installation, maintenance, and opportunity costs. Social factors were
divided into three favorability rankings (most favorable, moderately favorable, and least
favorable) based on area allocated to each BMP. Equal weights (1/3) were considered for the
three main factors while calculating the BMP rank by AHP. In this study three scenarios were
compared. A comprehensive approach in which environmental, economic, and social aspects are
simultaneously considered (Scenario 1) versus more traditional approaches in which both
environmental and economic aspects were considered (Scenario 2) or only environmental
aspects (sediment, TN, and TP) were considered (Scenario 3). In Scenario 1, only stripcropping
(moderately favorable) was selected on all CSAs at the subbasin level, whereas stripcropping (49
to 69 % of CSAs) and residue management (most favorable, 31 to 51% of CSAs) were selected
by AHP based on the watershed outlet and three spatial targeting methods. In Scenario 2, native
grass was eliminated by moderately preferable BMPs (stripcropping) both at the subbasin and
watershed outlet levels due the lower BMP implementations cost compared to native grass.
Finally, in Scenario 3, at subbasin level, the least socially preferable BMP (native grass) was

184

selected in 100% of CSAs due to greater pollution reduction capacity compared to other BMPs.
At watershed level, nearly 50% the CSAs selected stripcropping, and the remaining 50% of
CSAs selected native grass and residue management equally.

6.2 INTRODUCTION
NPS pollution is the primary source of water quality problems in the United States (USEPA,
2003). In the past few decades, NPS pollution generated from agricultural activities have become
the primary contributor to water quality impairments in rivers and lakes (USEPA, 2005). Higher
agricultural yields obtained by increasing nutrient application have resulted in environmental
concerns such as eutrophication (Shen et al., 2013). Additionally, in order to meet energy
security needs, the rapid growth of bioenergy crop production will likely jeopardize aquatic
ecosystems (Love et al., 2011; Yousefpour, 2013).

Implementing BMPs on agricultural lands to improve water quality is a well-known method
(Giri et al., 2012a). However, BMP performance depends on placement, timing, and selection
procedures (Giri et al., 2012b). Effective BMP implementation strategies cannot be achieved
without simultaneous consideration of economic and social aspects of these strategies. To
address these concerns, watershed management decision making plans should consist of
evaluating, balancing, and making trade-offs between these components and available alternative
management practices (Kaplowitz and Lupi, 2012). Multi-criteria decision analysis (MCDA) is a
widely accepted method to address these challenges (Yatsalo et al., 2007). The multi-attribute
utility theory (MAUT), multi-attribute value theory (MAVT), and analytical hierarchy process
(AHP) are examples of MCDA methods, which use optimization algorithms to solve complex

185

decision making problems (Linkov and Steevens, 2013). In particular, AHP uses systematic
evaluation criteria based on pairwise comparison and expert knowledge (Young et al., 2009).

Several studies in water resources have used AHP to support decision making. Young et al.
(2009) introduced AHP for selection of BMPs to reduce pollutant loadings downstream from a
small parking lot in a residential/commercial development area. The selection of BMP ranking
was obtained through pairwise comparison of selection criteria, BMPs among themselves, and
BMPs against selection criteria (aesthetic benefit, limiting the BMP installation site to less than
one acre, total suspended solid removal, total phosphorus removal, and total nitrogen removal).
The pairwise comparison of selection criteria generated a criteria priority vector (weight of
individual criteria), while the pairwise comparison of BMPs produced a BMP decision matrix.
Finally, the BMP decision matrix was multiplied by the criteria priority vector to generate the
priority BMP ranking. The final ranking of BMPs suggested bioretention, porous pavement, and
storm water filtering systems were the most effective BMPs in descending order. Calizaya et al.
(2010) used AHP to solve MCDA and to identify a sustainable water resources management plan
in the Lake Poopo basin, Bolivia. The MCDA structure consisted of three major objectives
(economic, social, and environmental issues), 10 conflicts (lower level objectives and subcriteria), seven instruments to solve the conflicts (alternatives), and implementing actors
(organizations). They evaluated the solutions from the MCDA based on the active participation
of stakeholders. Forty five pairwise comparisons were included in the MCDA structure. The
weights used in this study for environmental, social, and economic criteria were 0.62, 0.33, and
0.06, respectively, and were obtained by stakeholder participation. The most effective
instruments of this MCDA structure were educational training program, formation of local water

186

management organizations, and stakeholder involvement; whereas the most effective
implementing actor was local government. Garfi et al. (2011) used AHP in multi-criteria analysis
(MCA) to improve strategic environmental assessment of water programs in developing
countries and validated for a semi-arid region in Brazil. Both general and specific criteria were
selected to determine the best alternative among the One Million Cisterns Project and the Spring
Assessment Program for water management. The goal of the study was to improve drinking
water supplies to communities living in a semi-arid region. The final criteria were further divided
into 11 general sub-criteria for human development and 12 technical sub-criteria for water
supply. The relative weights were determined by pairwise comparison among the sub-criteria of
each respective group. The results of this study showed that the Cisterns Project were more
effective compared to the Spring Assessment Program considering economic, social, political,
and environmental aspects.

A number of studies have applied AHP for decision support in water resources. A few of those
studies have used AHP to determine the most effective BMP implementation, primarily in urban
areas. It shows that AHP can be used multiobjective BMP implementation plan effectively.
However, this study is unique because it focuses on evaluating suitable application of BMPs on
agricultural lands on a large scale, which to the best of our knowledge has not been done. The
specific objectives for this study were to: (1) evaluate the cost of pollution reduction associated
with BMP installation both at subbasin level and the watershed outlet and (2) identify the best
BMP type and implementation site using AHP while considering social, economic, and
environmental issues based on different spatial targeting methods.

187

6.3 MATERIALS AND METHODS
6.3.1 Study Area
This study was conducted on the SRW, which is located in east central Michigan (Figure 6-1). It
consists of six subwatersheds: Tittabawassee, Shiawassee, Pine, Flint, Cass, and Saginaw. This
watershed is one of the most diverse watersheds in Michigan, consisting of agriculture,
manufacturing, tourism, wildlife habitat, and outdoor recreation (Giri et al., 2012a). The Saginaw
River flows towards north direction and finally drains into Lake Huron. This watershed contains
nation’s largest contiguous freshwater coastal wetland (USEPA, 2009). The mean, minimum,
and maximum watershed elevations are 242 m, 177 m, and 457 m, respectively. The total
watershed area is 15,263 km2, of which 42% forest, 23% agriculture, 17% pasture, 11% wetland,
and 7% urban. It is one of the predominant agricultural-based watersheds in Michigan, with
predominantly corn and soybean cropping rotations.

188

N

Legend
Saginaw
Michigan
0

20

40

80
Kilometer
Figure 6-1. Location of Saginaw River Watershed.

189

6.3.2 Model Description
In order to evaluate the BMP effectiveness in reducing NPS pollution in the SRW, a physically
based, spatially distributed, watershed-scale model (Arnold et al., 1998; Neitsch et al., 1998;
Neitsch et al., 2005) known as SWAT was used. Primary model components include hydrology,
soil, landuse, plant growth, nutrients, pesticides, management practices, and weather (Gassman
et al., 2007). SWAT calculates flow, sediment, nutrients, and pesticides transport both over land
and in-stream based on the physiographic, meteorological, and land-management characteristics
of the watershed. The watershed is divided into subbasins and further divided into hydrologic
response units (HRUs) based on the homogeneous landuse, soil type, slope, and management
practices.

6.3.2.1 Data Sources
The SWAT model requires input data such as topography, land use, soil, and stream network.
Topography data in the form of digital elevation model (90m×90m) was obtained through the
Better Assessment Science Integrating point and Nonpoint Sources (BASINS) software.
Landuse data at 56m resolution 2008 Cropland Data Layer) for the watershed was obtained from
USDA’s National Agricultural Statistics Service (NASS, 2008). To represent the soil
characteristics in the watershed, the State Soil Geographic Database (STATSGO) was used,
which was developed by the National Cooperative Soil Survey. The stream network in the form
of a National Hydrography Dataset was obtained from the United States Geological Survey
(USGS) to improve hydrologic segmentation and subwatershed boundary delineation in the
SRW.

190

For stream flow calibration and validation in the watershed, daily stream flow data was obtained
from USGS gauging station 04157000. Water quality data (sediment, TN, and TP) were obtained
from Michigan Department of Environmental Quality (MDEQ) for station 090177. Climatic data
for 19 precipitation and 11 temperature stations were downloaded from the National Climatic
Data Center (NCDC) for twenty years (1990-2009). The data for remaining climatic parameters
(wind speed, relative humidity, and solar radiation) were generated from the SWAT weather
generator.

In order to assess the fate and transport of sediment and nutrients in the watershed, agricultural
management operations were developed based on common practices by local farmers in the
watershed. A detail description of timing of tillage and type and amount of fertilizer applied is
provided in the Giri et al. (2012a).

6.3.2.2 Sensitivity Analysis and Calibration
Model calibration is performed by adjusting the most sensitive model parameters, which are
obtained through a process called as sensitivity analysis. This analysis provides a rank of the
most influential parameters on model output. To evaluate the most influential parameters during
sensitivity analysis, Latin-Hypercube One-factor-At-a-Time (LH-OAT) parameter sampling
technique is used in the SWAT framework (van Griensven et al., 2006). After determining the
sensitive parameters, calibration and validation was performed for stream flow, sediment, and
nutrients on monthly time step with a two-year model warm-up period. A detail description of
sensitivity analysis, calibration, and validation is described in Giri et al. (2012a).

191

6.3.2.3 Best Management Practices in SWAT
Five BMPs: stripcropping (SC), residue management (RM), conservation tillage (CT), native
grass (NG), and no till (NT) were implemented separately on agricultural lands in the RW. The
pre- and post- BMP implementation results were compared with each other to evaluate BMP
effectiveness both at subbasin and watershed scales.

6.3.2.4 Spatial Targeting Methods
To optimize resource allocation for a BMP implementation plan, it is important to identify
locations that contribute the most to overall pollution load either at the edge of a field or at the
watershed outlet. These areas are known as CSAs. Four targeting methods; Concentration
Impact Index (CII), Load Impact Index (LII), Load per Subbasin Area Index (LPSAI), and Load
per Unit Area Index (LPUAI) were used to identify CSAs for sediment, total nitrogen, and total
phosphorus. These four targeting methods were used to categorize the watershed into high,
medium, and low priority areas. The categorization was performed based on the Jenks natural
break optimization technique, where similar values of data form a group while maximizing the
difference between groups (Jenks, 1967). The priority areas in CII targeting method is
determined based on the pollutant concentration in the reach subbasin (Tuppad and Srinivasan,
2008) while LII targeting method categorized priority area based on cumulative pollutant load in
the reach (Tuppad and Srinivasan, 2008, Giri et al., 2012a). The LPSAI targeting method
classifies priority areas based on total pollutant load produced by each subbasin (Giri et al.,
2012a) whereas the LPUAI targeting method determines priority areas based on the pollutant
load per unit area of each subbasin (Tuppad and Srinivasan , 2008; Giri et al., 2012a). A detailed
description of these targeting methods is provided in Giri et al. (2012a).

192

6.3.3 Environmental, Economic, and Social Aspects of BMP Implementation Plan
Before evaluating the overall impacts of BMP implementation plan using the AHP method, the
environmental, economic, and social aspects of these implementation scenarios need to be
assessed separately. The following sections describe the evaluation procedure.

6.3.3.1 Environmental Aspects of BMP Implementation Plan
The five BMPs were implemented one at a time based on the targeting scenarios (sediment, total
nitrogen, and total phosphorus) by four targeting methods (CII, LII, LPSAI, and LPUAI).
Overall, 180 different environmental scenarios (5 BMPs × 3 priority areas × 3 targeting scenarios
×

4

targeting

methods)

were

evaluated.

The

SWAT

model

outputs

were

aggregated/disaggregated to 56 m grids. The grid size is equal to the 2008 Cropland Data Layer
land use resolution. The same resolution was used for the economic and social analyses to
preserve spatial resolution consistency. The sediment, TN, and TP reduction for each BMP were
then calculated at the field and watershed scales to determine BMP effectiveness.

6.3.3.2 Economic Aspects of BMP Implementation Plan
The BMPs used in this study were assumed to be implemented for five years (equal to the
longest design life span among the BMPs – stripcropping and native grass). Therefore, the total
BMP cost was calculated for a five-year period. Installation of BMPs is the one-time cost
required (although this cost is multiplied by the five-year period if the lifespan is one year). The
annual maintenance cost of BMP is the money spent to maintain effective pollution reduction
throughout the year. Opportunity cost is the interest rate (3.9%) of BMP implementation cost if
the life span of BMP is more than one year (NRCS, 2011). The total cost column in Table 6-1 is
193

the summation of installation, annual maintenance, and opportunity costs. The BMP application
area of each targeting scenario for each targeting method was multiplied with the respective total
BMP cost to determine the total cost spent for that BMP. Then the total BMP cost was divided
by the amount of pollution reduction (both at subbasin and watershed outlet), which provides
dollar per mass of pollution reduction. The BMP cost and design lifespan presented in Table 6-1
are obtained from the United State Department of Agriculture - Natural Resources Conservation
Service (NRCS, 2011). Overall, five different economic scenarios (five BMPs) were evaluated at
56 m resolution.

Table 6-1. Five-year itemized cost for different BMPs used in this study (NRCS, 2011).
Annual
Installation
Total fiveMaintenance Opportunity
*
BMPs
Cost
Design
year cost
Cost
cost
(dollar/per
lifespan(year) (dollar/per
(dollar/per
(dollar/year)
hectare)
hectare
hectare-year)
SC
RM
CT
NG
NT

64.22
29.64
49.40
687.57
67.26

0.64
0.00
0.00
20.63
0.00

2.5
26.8
-

5
1
1
5
1

79.95
148.20
247.00
924.79
336.29

6.3.3.3 Social Aspects of BMP Implementation Plan
The BMP application area is considered as the social component in this study. The preference of
BMP selection by farmers depends on the BMP application area in the cropland. For example,
BMPs that require a small implementation area are preferred by most of the farmers, whereas
BMPs that need a large implementation area are preferred the least. Larger implementation areas
require farmers to remove crops from production to accommodate the implementation. The BMP

194

implementation area was obtained from local and national NRCS datasets (NRCS, 2008).
Overall, five different social scenarios (five BMPs) were evaluated at 56 m resolution.
Table 6-2. The BMP allocation area and associated social preferences of different BMPs.
BMPs
Allocation area to BMP
Social Preferences
Implementation (%)
SC
50
Moderate
RM
0
Most Favorable
CT
0
Most Favorable
NG
100
Least Favorable
NT
0
Most Favorable

6.3.4 Identify the best BMP type and implementation site using Analytic Hierarchy Process
The AHP method was introduced by Saaty in 1980 to solve complex decision making problems.
It can be applied to a wide array of fields such as project management, strategic planning, and
alternative selection processes. The AHP consists of four step processes: 1) construction of
pairwise comparison matrix, 2) computation of priority vector, 3) calculation of consistency
ratio, and 4) ranking of alternatives (Young et al., 2009).

1) Construction of pairwise comparison matrix: In this step, a pairwise matrix is constructed
separately among the alternatives for each relevant criterion. An additional pairwise
matrix is formed only among the relevant criteria. In each comparison matrix, each row
entry (alternatives/criteria) is compared to each column entry (alternatives/criteria) by
using a scale (1-9) of relative importance, where one is equally important and nine is
absolutely more important.

195

2) Computation of priority vector: the priority vector of the pairwise comparison matrix
(constructed in the previous step) is created by computing a normalized principal
eigenvector (Saaty, 1980). To calculate the principal eigenvector, each column entry in
the pairwise comparison matrix is divided by the sum of its respective column, generating
a new entry for that column. The sum of the new column entry should equal one. Then
the average of each row is calculated using the new entry in the matrix, which provides
the priority vector.
3) Calculation of consistency ratio: Calculation confirms the consistency of importance of
one entry over another in the matrix. The first step in calculating the consistency ratio is
multiplying the original pairwise comparison matrix (step 1) with the priority vector
calculated in step 2 to create a new matrix. Each entry of this new matrix is divided by
the priority vector (step 2) to create another a new matrix. The average of all the entries
in the new matrix provides the maximum eigenvalue (λmax). The closer the value of
maximum eigenvalue is to the number of rows/columns in the pairwise comparison
matrix the better the consistency among the entries. The consistency index is calculated
by Saaty (1980) using the following equation:

( )

(

)
(

)

(6-1)

n is the number of rows/columns in the pairwise comparison matrix. The consistency
ratio is calculated by using the following equation developed by Saaty (1980):
(

196

)

(6-2)

The random index varies based on n value and was obtained from Saaty (1980). If the
consistency ratio is 0.1 or less, then the pairwise comparison matrix formed during first
step is consistent. Otherwise, rearrangement of entries in the pairwise comparison matrix
is performed to ensure the logic between the alternatives/criteria.
4) Ranking of alternatives: the priority vectors calculated among the alternatives for each
criterion are extracted and placed into a new single matrix. The new matrix is multiplied
by the priority vector obtained from relevant criteria which provides the rank of all the
competing alternatives.
An AHP extension (ext_ahp.dll) in ArcGIS developed by Marinoni (2009) was used to calculate
the BMP ranking considering environmental, social, and economic criteria. A flow chart
representing the steps required during the application of AHP is provided in the Figure 6-2.

197

Identify
alternative
s

Identify
criteria

Formation of
comparision matrix
among alternatives

Rearrange
matrix

Formation of
comparision matrix
among criteria

Computation of
priority vector

Computation of
priority vector
Rearrange
matrix

Consistency
ratio check

Consistency
ratio check

CR > 0.1
CR ≤ 0.1

CR > 0.1

CR ≤ 0.1
Multiply priority vector
of alternatives with
criteria

Enter priority
vector into
single matrix

Rank of
alternative
s
Figure 6-2. AHP flowchart to determine the rank of competing alternatives.

BMP implementation strategies can be developed based on pollution reduction at the subbasin or
watershed level. Therefore, three scenarios were evaluated at each level to determine the
applicability of AHP to optimize overall outcomes and compare the new approach in which
environmental, economic, and social aspects are simultaneously considered (Scenario 1) versus
more traditional approaches in which both environmental and economic aspects were considered

198

(Scenario 2) or only environmental aspects (sediment, TN, and TP) were considered (Scenario
3).

6.3.4.1 Scenario 1 -Using AHP to Optimize the Environmental, Economic, and Social
Outcomes
BMPs are usually first implemented on agricultural fields within the high priority subwatersheds.
However, regardless of targeting scenarios, no agricultural land was identified as high priority
for the LII targeting method. Therefore, the comparisons were only performed on the high
priority areas identified by for CII, LPSAI, and LPUAI methods for all targeting scenarios
(sediment, TN, and TP). The following section describes the steps that were taken to develop the
BMP implementation plan using AHP.

Step 1) Construction of pairwise comparison matrix: The BMP effectiveness was calculated
based on the sediment, TN, and TP reductions at both subbasin and the watershed outlet. This
was performed by calculating the total pollutants reduction before and after BMP implication.
The process is presented here for one subbasin (43), which was identified as a high priority area
by all targeting methods (CII, LPSAI, and LPUAI) (Figure B-1 in the Appendix). The pairwise
comparison matrix are presented for sediment (Table 6-3), TN (Table 6-4), and TP (Table 6-5)
reductions at the subbasin level. Similar tables for all subbasins were developed based on
pollutions reductions at the watershed outlet that are presented in Tables B-2 through B-4 in the
Appendix.

199

Table 6-3. Pairwise comparison matrix developed for subbasin 43 based on sediment
reduction at the subbasin level.
BMPs
SC
RM
CT
NG
NT

SC
1.00
0.95
0.26
1.40
0.29

RM
1.05
1.00
0.27
1.47
0.30

CT
3.85
3.68
1.00
5.40
1.11

NG
0.71
0.68
0.19
1.00
0.20

NT
3.48
3.32
0.90
4.88
1.00

Table 6-4. Pairwise comparison matrix developed for subbasin 43 based on TN reduction
at the subbasin level.
BMPs
SC
RM
CT
NG
NT

SC
1.00
0.50
0.11
1.78
0.11

RM
2.00
1.00
0.11
3.56
0.11

CT
9.00
9.00
1.00
9.00
0.89

NG
0.56
0.28
0.11
1.00
0.11

NT
9.00
9.00
1.12
9.00
1.00

Table 6-5. Pairwise comparison matrix developed for subbasin 43 based on TP reduction at
the subbasin level.
BMPs
SC
RM
CT
NG
NT

SC
1.00
0.45
0.11
1.60
0.11

RM
2.24
1.00
0.11
3.59
0.11

CT
9.00
9.00
1.00
9.00
1.00

NG
0.62
0.28
9.00
1.00
0.11

NT
9.00
9.00
1.00
9.00
1.00

Steps 2) Computation of priority vector: After developing the pairwise comparison matrices, the
weight vector was calculated by using the AHP extension (ext_ahp.dll) in ArcGIS developed by
Marinoni (2009). The weight vector for sediment, TN, and TP reduction matrix at the subbasin
level is presented in Table 6-6 and at the watershed outlet is presented in Table B-5.

200

Table 6-6. Weight vector calculation of BMPs for sediment, TN, and TP reduction for
subbasin 43 at subbasin level.
BMPs
SC
RM
CT
NG
NT

Sediment Weight
0.2522
0.2449
0.0672
0.3620
0.0737

TN Weight
0.2942
0.2033
0.0333
0.4377
0.0316

TP Weight
0.3057
0.1990
0.0324
0.4305
0.0324

Step 3) Calculation of consistency ratio: The consistency ratios were calculated for subbasin (43)
using the AHP extension (Marinoni, 2009). The consistency ratios for sediment, TN and TP at
subbasin level are 0.0076, 0.0363, and 0.0302, and at watershed level are .0000, 0.0000, and
0.0000, respectively. All of the consistency ratios for all subbasins were acceptable as they were
less than 0.1.

Similar to the pollutions (sediment, TN, and TP), total BMP cost (economic component) and
BMP application area (social component) pairwise comparison matrices were prepared.
However, contrary to the pollution pairwise comparison matrices, only one set was developed for
all subbasins because BMP installation cost and farmer preference for applying those BMPs are
consistent throughout the watershed. For this section, the consistency ratios and weight vectors
were calculated using the AHP extension. The weight vector for total BMP implementation cost
is provided in Table 6-7 and the weight vector for BMP application area is shown in Table 6-8.

201

Table 6-7. Weight vector calculation of BMPs for total BMP cost.
BMPs
SC
RM
CT
NG
NT

Table 6-8. Weight vector calculation of BMP application area.
BMPs
SC
RM
CT
NG
NT

Weight
0.4440
0.2513
0.1511
0.0424
0.1112

Weight
0.0770
0.2984
0.2984
0.0278
0.2984

Regarding the social component, the weight vector of conservation tillage, no-till, and residue
management were the greatest (0.2984) whereas the weight vector of native grass was smallest
(0.0278) (Table 6-8) which is justified, because the native grass takes 100% of agricultural land
out of production, while this number is 0% in conservation tillage, no-till, and residue
management (Table 6-2).

The weights for the criteria priority vector were divided equally among the three criteria
(pollution reduction, total BMP cost, and BMP application area). The pollution reduction
criterion priority vector was further divided into three equal weights as it consists of three
components (sediment, TN, and TP). The weights of the criteria priority vector are presented in
Table 6-9.

202

Table 6-9. Weight vector of criteria used in this study.
BMPs
Total BMP cost
Sediment reduction
TN reduction
TP reduction
BMP Application Area

Weight
0.33
0.11
0.11
0.11
0.33

Step 4) Ranking of alternatives: Finally, the weight vectors for each criterion among the BMPs
were entered into a single matrix (Table 6-10 at the subbasin level and Table B-6 in the
Appendix at watershed level) and this matrix was multiplied by the criteria priority vector (Table
6-9) to generate the final weight for individual BMPs. The final weight for individual BMP is
presented in Table 6-11.
Table 6-10. Decision matrix of BMPs for all criteria developed for subbasin level analysis
(subbasin 43).
BMPs
Total BMP
Sediment
TN reduction
TP reduction BMP Application
cost
reduction
Area
SC
0.4440 0.2522
0.2942
0.3057
0.0770
RM
0.2513 0.2449
0.2033
0.1990
0.2984
CT
0.1511 0.0672
0.0333
0.0324
0.2984
NG
0.0424 0.3620
0.4377
0.4305
0.0278
NT
0.1112 0.0737
0.0316
0.0324
0.2984

Table 6-11. Final weight vector of individual BMPs for subbasin 43.
BMPs

Weight
(subbasin level)
0.2657
0.2526
0.1630
0.1585
0.1503

SC
RM
CT
NG
NT

Stripcropping and residue management are the most favorable BMPs, while native grass is most
effective in reduction of sediment, TN, and TP (Tables 6-3 to 6-5). This result is due to the
203

higher total BMP cost (Table 6-1) associated with higher BMP application area (Table 6-2) in
the case of native grass compared to stripcropping and residue management. This result is only
valid for the BMP ranking in subbasin 43 after simultaneously considering environmental,
economic, and social factors. Similarly, the BMP ranking was calculated for all subbasins
identified as high priority area by the three targeting methods (CII, LPSAI, and LPUAI) and all
targeting scenarios (sediment, TN, and TP). The final ranking of BMPs for all subbasins
identified as high priority area is presented in Table B-8 at the subbasin level and Table B-9 at
the watershed level in the Appendix.

6.3.4.2 Scenario 2 -Using AHP Method to Optimize the Environmental and Economic
Outcomes

The BMP selection was calculated based on sediment, TN, and TP reductions at the watershed
outlet and total BMP cost. The pollution reductions were calculated for all BMPs for each
subbasin in the previous scenario. The BMP cost for each subbasin was calculated by
multiplying the total BMP cost (Table 6-1) for each BMP by the BMP application area in each
subbasin. The weights for the criteria priority vector were divided equally among two criteria
(pollution reduction and total BMP cost). The pollution reduction criterion priority vector is
further divided into three equal weights (0.1667) as it consists of three components (sediment,
TN, and TP). The BMP cost received the weight of 0.5. Similar procedures to Scenario 1 were
followed to produce the most favorable location for BMP placement for each targeting method.
The final ranking of BMPs for all subbasins identified as high priority area is presented in Table
B-10 at the subbasin level and Table B-11at the watershed level in the Appendix.

204

6.3.4.3 Scenario 3 -Using AHP Method to Optimize the Environmental Outcomes
The BMP selection was calculated based on the sediment, TN, and TP reduction. The amount of
pollution generated by an individual subbasin was calculated after BMP application and this
amount was subtracted from the pollution generated by the subbasin during the no BMP applied
condition to determine the pollution reduction for each subbasin by individual BMP. Based on
the amount of sediment reduction, a BMP rank was prepared for each subbasin. The weight for
criteria priority vector was divided equally among three criteria in which each criterion received
the weight of 0.3333. The final ranking of BMPs for all subbasin identified as high priority area
is presented in Table B-12 at subbasin level and Table B-13 at watershed level in the Appendix.

6.4 RESULTS AND DISCUSSION

6.4.1 Determining the Cost of Pollution Reduction Associated with BMP Installation Both at
Subbasin Level and the Watershed Outlet
The objective of this section was to determine the most effective BMPs by considering both
pollutant reduction and total BMP cost. Pollutant reduction varies among the BMPs, priority
areas and targeting methods. Figure 6-4 represents the sediment reduction in dollars per ton
among BMPs, priority areas, and targeting methods. Overall, stripcropping and residue
management were most effective compared to other BMPs both at the reach and subbasin levels
(Figure 6-4) due to higher sediment reduction and lower BMP cost (Table 6-1). However, when
BMP effectiveness is evaluated only based on environmental factors such as sediment reduction,
native grass was the most effective, which agrees with Woznicki et al. (2011) and Giri et al.
205

(2012a). In this study, native grass demonstrated moderate effectiveness due to having the
highest total BMP cost (Table 6-1). Conservation tillage and no-till are the least effective BMPs,
both at the reach and subbasin levels, which is attributed to significantly lower sediment
reduction compared to other BMPs.

The reduction cost per ton of sediment at the watershed outlet in the high priority areas ranges
from $230 dollars to approximately $31,000, in the medium priority areas the cost ranges from
$250 to nearly $11,000, and in the low priority areas the cost ranges from $390 to $15,000
(Figure 6-4a, 6-4b, and 6-4c). However, the sediment reduction cost at the subbasin level is
lower in all priority areas compared to sediment reduction cost at the watershed outlet. The
reduction cost per ton of sediment for subbasins in the high priority areas ranges from $50 to
nearly $750, in the medium priority areas the cost ranges from $60 dollars to approximately
$1150 dollars, and in the low priority areas the cost ranges from $50 to $1100 dollars (Figure 64d, 6-4e, and 6-4f). The sediment reduction cost of native grass is always less than the cost of
conservation tillage and no-till at both the watershed outlet and the subbasins in all priority areas
and targeting methods except for the low priority area at the subbasin level (Figure 6-4f). This
result is due to combination of the highest total BMP cost of native grass (Table 6-1) and a
slightly higher sediment reduction of native grass compared to conservation tillage and no-till on
the low-priority area.
The effectiveness of targeting methods at the watershed outlet also varies among the priority
areas. For example, the most cost effective method in the high priority areas is LPSAI (Figure 64a). However, this method is the least cost effective in the medium priority areas due to a smaller
amount of available land for BMP implementation compared to other targeting methods (Table

206

6-12). Like the watershed outlet, the effectiveness of targeting methods also varies between the
priority areas at the subbasin level. For example, the most effective method in the high priority
areas is LPUAI (Figure 6-4d). However, this method is least cost-effective in the low priority
areas due to less available land for BMP implementation, resulting in less sediment reduction
compared to other targeting methods (Table 6-12).

207

39000
(a)

26000

39000

dollar/ton

13000

(b)

13000

0

39000
CII
LII
LPSAI
LPUAI

0

(c)

26000
dollar/ton

dollar/ton

26000

SC RM NG CT NT

Low Priority Area

Medium Priority Area

High Priority Area

13000

0
SC RM NG CT NT

SC RM NG CT NT

Figure 6-4. (a, b, and c) BMPs sediment reduction at the watershed outlet. by different targeting methods.

208

Medium Priority Area

High Priority Area
1200

(d)

1200

(e)

e

1200

(f)

800

800

dollar/ton

400

400

400

0

0

dollar/ton

dollar/ton

800

Low Priority Area

SC RM NG CT NT

0
SC RM NG CT NT

SC RM NG CT NT

Figure 6-4. (d, e, and f) BMPs sediment reduction at the subbasin by different targeting methods.

209

Table 6-12. BMP implementation area for different priority area based on sediment
targeting scenarios.
Targeting Method
CII

LII

LPSAI

LPUAI

Priority area
H
M
L
H
M
L
H
M
L
H
M
L

BMP implementation area (ha)
27647.6
202395.7
115379.6
0.0
232351.0
113071.9
143318.7
160939.2
41165.0
84977.4
249368.0
11077.5

In regard to other pollutants, a similar pattern was observed for TN and TP reduction by BMPs at
the reach and subbasin levels (Figures 6-5 and 6-6). However, the overall cost per kilogram of
pollutant is higher. In addition, residue management was identified as the least effective BMP on
the low priority areas. Unlike sediment, an increase of TN and TP was observed for both
conservation tillage and no-till at the reach and subbasin levels (Figures 6-5 and 6-6). This
indicates that the project sponsor will lose money by implementing these BMPs to control
nutrients generated by agricultural fields. The increased TN and TP may be due to an increase in
organic matter content, which may be released into soil due to implementation of these practices
(Giri et al., 2012a).

210

High Priority Area
4000

4000

(a)

Low Priority Area

Medium Priority Area

4000

(b)

(c)

CII

2000

LPSAI

2000

dollar/kg

dollar/kg

LPUAI

dollar/kg

LII

2000

-2000

NT

SC RM NG -2000

(d)

0

-800

-2000
800

0

-1600

NT CT SC RM NG

(f)

0
-800

-1600

-2400

-2400

NT CT SC RM NG

(e)

-800

-1600

0

NT CT SC RM NG

800
dollar/kg

dollar/kg

800

CT

0

dollar/kg

0

NT CT SC RM NG

-2400
NT CT SC RM NG

Figure 6-5. BMPs TN reduction at outlet (a, b, and c) and subbasin (d, e, and f) by different targeting methods.

211

High Priority Area
3000

Low Priority Area

Medium Priority Area

3000

(a)

3200

(b)

2200

1000

1000

1200

0

-1000

0

-1000

-2000

-2000

-3000

CII
LII
LPSAI
LPUAI
CT NT SC RM NG

-3000
CT NT SC RM NG

4000

(d)

4000

1000

0

-1000

CT NT SC RM NG

-1800
-2800
CT NT SC RM NG
4000
(f)

2000
1000

0

0

-1000

-1000

-2000

-800

dollar/kg

dollar/kg

2000

1000

200

3000

3000

2000

dollar/kg

3000

(e)

dollar/kg

2000
dollar/kg

dollar/kg

2000

(c)

-2000

-2000
CT NT SC RM NG

CT NT SC RM NG

Figure 6-6. BMPs TP reduction at outlet (a, b, and c) and subbasin (d, e, and f) by different targeting methods.

212

6.4.2 Identify the Best BMP Type and Implementation Site Using Analytic Hierarchy Process
The objective of this section is to evaluate the rank and placement of BMPs based on different
influential factors (environmental, economic, and social aspects) to achieve pollutant reduction.
The findings of this section provide decision makers and stakeholders a wide array of solutions
to improve water quality at the lowest cost and highest chance at being accepted by producers.

Figure 6-7 represents the placement of BMPs that ranked highest by different targeting methods
at the subbasin level. In Scenario 1, where environmental, economic, and social factors have
equal weights (33% each) considering equal importance of three pollutants, strip cropping was
preferred in all identified CSAs by all targeting methods (Figure 6-7g, h, and i). This was due to
the low BMP cost (Table 6-1) combined with high pollution reduction, and the moderate
acceptance of this BMP by producers (Table 6-2). This finding suggests that policy makers and
stakeholders should consider this BMP over the others during the BMP implementation strategy
as it can satisfy the desires of both policy makers and stakeholders by reducing pollution in over
a smaller BMP application area. The results were the same for Scenario 2, in which social
impacts of BMP implementation were ignored even though environmental and economic impacts
are considered. Stripcropping was preferred over all CSAs for all targeting methods. When
Scenario 2 was compared with Scenario 3, native grass (preferred BMP in scenario 3) was
replaced by stripcropping in all CSAs (Figure 6-7 d, e and f). This result was due to the low cost
of stripcropping ($80/hectare) compared to the high cost of native grass ($925/hectare) combined
with a high pollution reduction capacity. When considering only pollutant reduction (Scenario

213

3), native grass was selected in all of the CSAs by all targeting methods (Figure 6-7a, b, and c).
This result was due to greater pollution reduction capacity of native grass and the fact that the
whole field is converted to grassland (Woznicki et al., 2011; Giri et al., 2012a). Overall, the
percentage of BMP placement areas was the highest for the CII targeting method compared to
LPSAI and LPUAI targeting methods as CII method is based on pollutant concentration (Giri et
al., 2012a). One BMP (native grass- Scenario 3, stripcropping- Scenario 2 and Scenario 3) was
preferred in all the subbasins if the BMP implementation strategy is focused at the subbasin. The
lack of variation in preferred BMP for the CSAs may be due to the smaller distance between the
BMP application area and measured effect at the subbasin level compared to the watershed
outlet.

The placement of the highest ranked BMPs, dependent only on pollution reduction at the
watershed outlet based on different scenarios and targeting methods, is provided in Figure 6-8.
When the BMP selection criteria was based on combined environmental, economic, and social
factors (Scenario 1); approximately, 59% of CSAs identified by the CII targeting method
selected the moderately preferable BMP (stripcropping) and 41% selected the most favorable
BMP (residue management) (Table 6-13). The replacement of some stripcropping from the
CSAs in scenario 2 with residue management (Figure 6-8 g) is due to producers’ preference of
residue management over stripcropping, as the BMP application area for stripcropping shall at
least consist of 50% erosion resistant crops or sediment trapping cover (NRCS,2008). This
suggests that selecting residue management saves 50% of the crop land that stripcropping would
keep fallow for BMP application. When both environmental and economic factors are considered
for BMP selection (Scenario 2), all native grass and some residue management (preferred BMP

214

in Scenario 3) were eliminated by stripcropping (Figure 6-8d) due to its low total BMP cost
(Table 6-1). Nearly 83% of CSAs selected stripcropping (Table 6-13) and the remaining 17% of
CSAs selected residue management. Stripcropping was selected more compared to residue
management because of the lower BMP cost of stripcropping compared to residue management
(Table 6-1). For example, when only environmental factors are considered for BMP selection
under the CII targeting method (Scenario 3); approximately 53% of CSAs preferred
stripcropping, 24% of CSAs preferred native grass, and the remaining 23% of CSAs preferred
residue management (Table 6-13). The number of preferred BMPs chosen for the CSAs
increased from just native grass (based on pollution reduction at subbasin level) to stripcropping,
native grass, and residue management (based on pollution reduction at watershed outlet). The
change in preferred BMPs, based on pollution reduction at the watershed outlet, was due to instream processes and distance of individual BMP application areas to the watershed outlet.
Similar BMPs were selected in the CSAs for different combinations of targeting methods and
scenarios.

A summary of the AHP rank one BMPs’ effectiveness, which includes social, environmental,
and economic factors both for subbasin level and watershed outlet for different CSA targeting
methods, is presented in Table 6-13. At the watershed outlet, the pollution reduction decreases
from Scenario 3 to Scenario 1, as BMPs that reduce less pollution are more preferable over
BMPs reducing more pollution by stakeholders due to the less required area for BMP application
and less expensive total implementation cost. Among the targeting methods at the watershed
outlet, the LPUAI targeting method selected residue management (most favorable) BMP as
having the lowest percentage of social acceptance compared to other targeting methods. This

215

suggests that this method is less capable of identifying socially acceptable BMPs compared to
other targeting methods. However, in the subbasin level, all targeting methods identified BMPs
were categorized into same social acceptance level (Table 6-13) due to the selection of same
type of BMP over all their respective CSAs.

Overall, the greater pollution reduction, higher expense, and least socially acceptable BMP
(native grass) was eliminated in favor of the selection of the lesser pollution reduction, least
expensive, and moderately socially acceptable stripcropping. Furthermore, some of the
stripcropping-identified subbasins were replaced by the most socially acceptable BMP, residue
management. More variety of BMPs is preferred in the CSAs based on pollution reduction at the
watershed outlet than at the subbasin level due to stream morphology and in stream processes.

216

(a)

N

Legend
B BMP
NO
NG
0

10

20 Kilometer

Figure 6-7. (a) Placement of BMP rank one in subbasin considering only environmental
factor based on CII targeting methods.

217

(b)

N

Legend
B
NO BMP
NG
0

10

20 Kilometer

Figure 6-7. (b) Placement of BMP rank one in subbasin considering only environmental
factor based on LPSAI targeting methods.

218

(c)

N

Legend
B
NO BMP
NG
0

10

20 Kilometer

Figure 6-7. (c) Placement of BMP rank one in subbasin considering only environmental
factor based on LPUAI targeting methods.

219

(d)

N

Legend
B
NO BMP
SC
0

10

20 Kilometer

Figure 6-7. (d) Placement of BMP rank one in subbasin considering environmentaleconomic factors based on CII targeting methods.

220

(e)

N

Legend

0

B
NO BMP
SC
10
20 Kilometer

Figure 6-7. (e) Placement of BMP rank one in subbasin considering environmentaleconomic factors based on LPSAI targeting methods.

221

(f)

N

Legend
NO BMP
B
SC
0

10

20 Kilometer

Figure 6-7. (f) Placement of BMP rank one in subbasin considering environmentaleconomic factors based on LPUAI targeting methods.
222

(g)

N

Legend
B BMP
NO
SC
0

10

0

10

20 Kilometer

20 Kilometer

Figure 6-7. (g) Placement of BMP rank one in subbasin considering environmentaleconomic-social factors based on CII targeting methods.

223

(h)

N

Legend
B
NO BMP
SC
0

10

20 Kilometer

Figure 6-7. (h) Placement of BMP rank one in subbasin considering environmentaleconomic-social factors based on LPSAI targeting methods.

224

(i)

N

Legend
NO BMP
B
SC
0

10

20 Kilometer

Figure 6-7. (i) Placement of BMP rank one in subbasin considering environmentaleconomic-social factors based on LPUAI targeting methods.

225

(a)

N

Legend
Legend
B
B
NO BMP
NG
RM
SC
0

10

20 Kilometer

Figure 6-8. (a) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering only environmental factor by CII targeting method.

226

(b)

N

Legend
B
NO BMP
NG
RM
SC
0

10

20 Kilometer

Figure 6-8. (b) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering only environmental factor by LPSAI targeting method.

227

(c)

N

Legend
B BMP
NO
NG
RM
SC
0

10

20 Kilometer

Figure 6-8. (c) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering only environmental factor by LPUAI targeting method.

228

(d)

N

Legend
B
NO BMP
RM
SC
0

10

20 Kilometer

Figure 6-8. (d) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering environmental- economic factors by CII targeting method.

229

(e)

N

Legend
B BMP
NO
RM
SC
0

10

20 Kilometer

Figure 6-8. (e) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering environmental- economic factors by LPSAI targeting method.

230

(f)

N

Legend
B BMP
NO
RM
SC
0

10

20 Kilometer

Figure 6-8. (f) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering environmental- economic factors by LPUAI targeting
method.

231

(g)

N

Legend

0

10

B
NO BMP
RM
SC
20 Kilometer

Figure 6-8. (g) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering environmental-economic-social factors by CII targeting
method.

232

(h)

N

Legend
B
NO BMP
RM
SC
0

10

20 Kilometer

Figure 6-8. (h) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering environmental-economic-social factors by LPSAI targeting
method.

233

(i)

N

Legend
B BMP
NO
RM
SC
0

10

20 Kilometer

Figure 6-8. (i) Placement of BMP rank one in subbasin based on pollution reduction at
watershed outlet considering environmental-economic-social factors by LPUAI targeting
method.

234

Table 6-13. Summary of AHP identified rank one BMP effectiveness in all aspects (social,
environmental, and economic) both subbasin level and watershed outlet.

AHP model

Parameter

Environme
nt

Social (% of
total BMP
application
area)

Environme
nt Economic

Environme
ntEconomicSocial

Environmenta
l (pollution
reduction in
ton)
Economic
(total BMP
cost in dollar)
Social (% of
total BMP
application
area)
Environmenta
l (pollution
reduction in
ton)
Economic
(total BMP
cost in dollar)
Social (% of
total BMP
application
area)
Environmenta
l (pollution
reduction in
ton)

CII

LPSAI

LPUAI

23

23

12

Most
favorable
Moderate
Least
favorable
Sediment

53
24

39
38

37
51

80,700

64,400

36,700

TN
TP

1650
214

1452
170

1382
155

75,282,680

65,612,48
6

60,823,681

Most
favorable
Moderate

17

13

5

83

87

95

Least
favorable
Sediment

-

-

-

66,600

51,500

28,000

TN
TP

1337
186
23,686,458

1138
141
14,085,58
7

1024
118
9,800,501

Most
favorable
Moderate

41

51

31

59

49

69

Least
favorable
Sediment

-

-

-

63,400

48,200

24,900

TN

1062

867

769

TP

138

94

76

235

Table 6-13 (cont’d)
Economic
(total BMP
cost in dollar)

Environme
nt

Environme
nt Economic

Environme
ntEconomicSocial

Social (% of
total BMP
application
area)
Environmenta
l (pollution
reduction in
ton)
Economic
(total BMP
cost in dollar)
Social (% of
total BMP
application
area)
Environmenta
l (pollution
reduction in
ton)
Economic
(total BMP
cost in dollar)
Social (% of
total BMP
application
area)
Environmenta
l (pollution
reduction in
ton)
Economic
(total BMP
cost in dollar)

27,771,921

18,101,72
7

11,816,275

-

-

-

100

100

100

470,533

394,623

384,232

TN
TP

3737
571
238,604,221

2999
436
146,099,7
73

2697
390
108,365,621

Most
favorable
Moderate
Least
favorable
Sediment

-

-

-

100
-

100
-

100
-

323,841

279,725

276,471

TN
TP

2,229
346
20,629,205

1,812
272
12,631,47
0

1,653
247
9,369,057

Most
favorable
Moderate
Least
favorable
Sediment

-

-

-

100
-

100
-

100
-

323,841

279,725

276,471

2,229
346
20,629,205

1,812
272
12,631,47
0

1,653
247
9,369,057

Most
favorable
Moderate
Least
favorable
Sediment

TN
TP

236

6.5 CONCLUSION
The objectives of this study were to (1) evaluate the cost of pollution reduction associated with
BMP installation at both the subbasin level and the watershed outlet and (2) identify the best
BMP and implementation site using AHP while considering social, economic, and
environmental issues based on different spatial targeting methods. Four targeting methods were
used to identify cost of pollution reduction both at subbasin level and watershed outlet whereas
only three targeting methods were used to identify best BMP and implementation site using
AHP, as the LII targeting method identified no agricultural land in its CSAs. The targeting
methods were used to identify CSAs of sediment, TN, and TP and five BMPs (SC, RM, CT, NG,
and NT) were implemented in the identified CSAs to evaluate BMP effectiveness. The AHP
method was used to rank the BMPs for three different scenarios, which were based on different
combinations of environmental, environmental-economic, and environmental-economic-social
factors. The environmental factor consisted of sediment, TN, and TP reduction, the economic
factor consisted of total BMP cost, and the social factor consisted of farmer preference in BMP
implementation. In Scenario 1 environmental, economic, and social factors had equal weight
(0.33). Scenario 2 was based on both environmental and economic factors where an equal
weight. Scenario 3 was only based on environmental factors where an equal weight of 0.33 was
assigned to each component (reduction of sediment, TN, and TP) during the ranking of BMPs.

The BMP effectiveness was compared among targeting methods and priority area by considering
both pollution reduction and BMP cost. Results suggest that stripcropping and residue
management were the most cost-effective BMPs having lesser pollution reduction cost both at
the subbasin level and watershed outlet, whereas conservation tillage and no-till were the least

237

cost-effective. Increases of TN and TP associated with increased organic matter content were
observed for conservation tillage and no-till at the watershed scale. Meanwhile, native grass
demonstrated a moderate pollution reduction cost even though having the highest total cost
among all the BMPs.

Among the five BMPs native grass, stripcropping, and residue management were selected as
rank one by AHP for various combinations of targeting methods (CII, LPSAI, and LPUAI) and
scenarios. In Scenario 1, at the subbasin level, stripcropping was placed all over the CSAs in all
targeting methods, suggesting that it can satisfy from both a pollution reduction stand point
(policy makers objective), BMP implementation cost, and social acceptance (from stakeholders
objective). However, at watershed outlet, some of the stripcropping was replaced by the socially
most acceptable BMP (residue management). In Scenario 2 and at subbasin level, stripcropping
was selected for all CSAs by all targeting methods due to higher pollution reduction capacity and
the lowest total BMP cost. Meanwhile, more residue management was selected at the watershed
outlet in Scenario 2 compared to Scenario 1. At the subbasin level, Scenario 3, the least socially
preferable BMP (native grass) was selected at the subbasin level due to a greater pollution
reduction capacity of native grass compared to other BMPs. In fact, at the subbasin level all
CSAs selected native grass for all targeting methods. However, in the case of pollution reduction
at watershed outlet, the percentage of placement of native grass in CSAs varied from 24 to 51%
between the targeting methods.

Overall, no single BMP can satisfy all environmental, economic, and social issues. The results
from different scenarios provide a wide variety of solution for different conditions and should be
selected based on the watershed management plan requirements. The result of this study can help
238

policy makers and stakeholders determine the placement of suitable BMPs in suitable locations
for strengthening science-based decision making.

239

7.CONCLUSIONS
BMP pollutant reduction efficiency depends on several parameters such as type, placement,
implementation plan, design procedure, and maintenance frequency. This study primarily
focuses on the placement, selection, and implementation plan of BMPs. The findings help create
a useful BMP implementation strategy, which can fulfill the objectives of both policy makers
and stakeholders with the intent of improving water quality on a watershed scale. In this study,
four targeting methods, concentration impact index (CII), load impact index (LII), load per
subbasin area index (LPSAI), and load per unit area index (LPUAI), were used to identify the
critical source areas based on sediment, TN, and TP in the watershed using Soil and Water
Assessment Tool (SWAT) resulting following conclusions.



When sediment control is the objective, LPSAI should be considered (Figure 7-1).
Similarly, when TN and TP control is the goal, the pollutant CII should be considered
(Figure 7-1).



The LPSAI targeting method identifies the maximum amount of agricultural lands in the
high priority areas based on sediment targeting (Figure 7-1) whereas CII targeting
method identifies the maximum amount of agricultural lands in the high priority area
based on TN and TP targeting scenarios (Figure 7-1).



The LII targeting method, based on pollutant load in the stream, was incapable of
identifying agricultural lands in its high priority areas, indicating this method should not
be used to identifying suitable BMP placement location.
240



A strong agreement was found between LPSAI and LPUAI for the categorization of
priority areas based on sediment and TN targeting scenarios (Figure 7-2). This provides
insight to the method of categorization for priority areas by targeting methods.

After understanding the different targeting methods, 10 BMPs namely: contour farming, terraces,
recharge structures, conservation tillage, no-till, native grass, residue management (0 kg/ha),
residue management (1000 kg/ha), residue management (2000 kg/ha), and strip cropping were
modeled in the agricultural lands of SEW based on the priority areas identified by each targeting
method. A modified paired t-test was used to evaluate BMP pollutant reduction effectiveness and
the conclusions are provided below:



All the BMPs, except conservation tillage and no-till, showed some significant pollutant
reduction (targeted and non-targeted) for all targeting methods (Figure 7-3).



An insignificant pollution reduction was observed by BMPs when the BMP
implementation areas went from high plus medium priority areas to high plus medium
plus low priority areas; indicating that BMP implementation in high plus medium priority
areas is sufficient at achieving significant pollution reduction.

In the previous study, BMPs were evaluated based on the pollution reduction only. In this study,
AHP was used to determine the best BMP type using environmental (reduction of sediment, TN,

241

and TP), economic (total BMP implementation cost), and social factors (preference of BMP by
stakeholders) simultaneously. The conclusion of this study is provided below.



When only environmental factors are considered for BMP selection, native grass is
preferred in all the CSAs to achieve maximum pollution reduction at the subbasin level
(Figure 7-4).



When both environmental and economic factors are considered for BMP selection, strip
cropping is preferred in all CSAs to achieve maximum pollution reduction at the subbasin
level (Figure 7-4).



Strip cropping was also preferred in all CSAs (to attain maximum pollution reduction at
the subbasin level) when the BMP selection criteria was based on environmental,
economic, and social factors (Figure 7-4).



Using environmental factors as BMP selection criteria, strip cropping, residue
management, and native grass was selected in the CSAs to obtain maximum pollution
reduction at the watershed outlet (Figure 7-4).



When the BMP selection was based on environmental and economic factors, only strip
cropping and residue management were selected in the CSAs to achieve maximum
pollution reduction at the watershed outlet (Figure 7-4).

242



Strip cropping and residue management were also selected when the BMP selection
criteria was based on environmental, economic, and social factors to attain maximum
pollution reduction at the watershed outlet (Figure 7-4).

Understanding the spatiotemporal variability of critical source areas is important for constructing
a successful BMP implementation plan. In this study, native grass and contour farming was
selected as to examine the variability of priority areas, these BMPs were chosen because they
showed the highest and lowest significant pollution reduction among all BMPs respectively.
These BMPs were implemented only in the high priority areas of the watershed for two
consecutive years for all targeting pollutants (sediment, TN, and TP). The conclusions of this
study are provided below.



A distinct change in high priority areas was observed by the end of the second year in the
case of native grass due to higher pollution reduction efficiency of native grass compared
to contour farming.



A minimal change in high priority areas was observed for contour farming by the end of
second year due to lesser pollution reduction capacity compared to native grass.

243

Targeting method

CII

Maximum
agricultural
lands in high
priority area
of TN and TP

TN and TP
control

More
sensitive
to field than
outlet

LPSAI

LII

Maximum
agricultural
lands in high
priority area
of sediment

Average
agricultural
lands in
high priority
area

Sediment control

Average pollution
Sediment control
control

No agricultural
land in high
priority area
of TN and TP
scenario

Not recommended

Equal sensitive
to both field and
outlet

LPUAI

Equal sensitive to
both field and
outlet

More sensitive to
field than outlet

Figure 7-1. Targeting methods recommendation for different pollutants.

244

LII

CII

LPSAI

LPUAI

Identification of similar priority area based on
sediment and TN targeting

Identification of different priority
area based on TP targeting

Figure 7-2. Spatial correlation between targeting methods in identifying priority areas.

245

CF

RM

SC

RS

T

CT

NG

NT

Significant both for targeting and non-targeting
pollutants

Insignificant both for targeting
and non-targeting pollutants

Figure 7-3. Effectiveness of different BMPs in this study.

246

Scenario

EnvironmentEconomic

EnvironmentEconomic-Social

Subbasin

Strip
cropping

Outlet

Strip
cropping
Residue
manage
ment

Outlet

Subbasin

Strip
cropping

Strip cropping
Residue
management

Environment

Subbasin

Native
grass

Outlet

Strip cropping
Residue
management
Native grass

Figure 7-4. Most effective BMPs in CSAs both for subbasin and outlet for all scenarios.

247

8. RECOMMENDATION FOR FUTURE RESEARCH
Based on the finding of this study, the following future studies are recommended.



Targeting methods were used to categorized priority area into high, medium, and low
based on individual pollutants (sediment, TN, and TP). The next step of this study
would be combining the priority area of sediment, TN, and TP simultaneously to obtain
the overall priority area for all pollutants and targeting methods using geographical
information system.



The BMP selection criteria in AHP were based on only three factors environmental,
economic, and social. However, other factors such as BMPs serving as habitat to
wildlife, frequency of maintenance, efficiency in removing pathogen and bacteria,
aesthetic value of BMP, and ability of BMPs to make a value added product can be
added as BMP selection criteria during construction of pairwise comparison matrix.



During BMP selection process in AHP, weights are assigned to individual factors and
AHP tries to meet the criteria while satisfying the given weights for individual factors.
However, in the attempt to satisfy all the criteria, some criteria may not reach the optimal
values resulting in a non-ideal solution. Therefore, more than one technique should be
used during the decision making process just like the BMP selection process.



During the spatiotemporal variability analysis of priority areas, only one BMP from the
highest pollution reduction category and one from the lowest pollution reduction
248

category were used for the study. However, more BMPs (such as strip cropping and
conservation tillage) should be used from each category (highest and lowest pollution
reduction) to reinforce the results found in this study.



Spatiotemporal variability of the selected BMPs was studied only in the high priority
areas of the watershed. However, it would be interesting to incorporate these BMPs in
the medium and low priority areas as well to determine the spatiotemporal variability of
the BMPs in medium and low priority areas.

249

APPENDICES

250

(APPENDIX A – ADDITIONAL MATERIAL TO SECTION 5 TITTLED “ANALYSIS
OF BEST MANAGEMENT PRACTICE EFFECTIVENESS AND SPATIOTEMPORAL
VARIABILITY BASED ON DIFFERENT TARGETING STRATEGIES”)

251

APPENDIX A
Table A-1. P-values for BMPs based on AR (1) model for different targeting methods based on TN targeting(Sediment).
p-value
CT
NT
RM 0
CF
RM 1000 SC
RM 2000 RS
T
NG

B vs. H

1.32E01
B vs. H+M 1.12E01
B vs.
1.10EH+M+L
01
H vs. H+M 2.26E01
H+M vs.
2.70EH+M+L
01

2.36E01
1.40E01
1.05E01
1.43E01
2.06E01

6.37E02
2.90E02
2.80E02
1.29E01
2.38E01

B vs. H
B vs. H+M 2.31E01
B vs.
1.10EH+M+L
01
H vs. H+M H+M vs.
1.16EH+M+L
01

2.31E01
1.05E01
1.11E01

1.36E01
2.80E02
5.71E02

CII Targeting Method
4.48E2.56E-05 5.87E04
07
1.85E3.67E-07 8.02E05
10
1.24E2.02E-07 3.12E05
10
2.68E1.98E-02 3.47E02
03
1.55E1.78E-01 1.16E01
01
LII Targeting Method
9.45E9.48E-02 7.40E02
02
1.24E2.02E-07 3.12E05
10
8.48E1.88E-06 6.39E05
09

7.81E-07
1.98E-09
8.05E-10
7.31E-03
1.51E-01

8.17E-02
8.05E-10
1.36E-08

1.21E09
1.74E13
1.11E12
6.92E04
2.85E01

4.43E12
0.00E+
00
0.00E+
00
7.48E07
2.40E02

2.22E16
0.00E+
00
0.00E+
00
9.82E13
6.26E05

5.27E02
1.11E12
3.74E11

4.47E02
0.00E+
00
0.00E+
00

8.58E02
0.00E+
00
0.00E+
00

1.18E05
5.44E11
1.11E12
7.56E05

1.06E07
0.00E+
00
0.00E+
00
3.07E08

2.85E09
0.00E+
00
0.00E+
00
1.22E11

LPSAI Targeting Method

1.63E01
1.23EB vs. H+M
01
B vs.
1.10EH+M+L
01
1.96EH vs. H+M
01
B vs. H

1.61E01
1.19E01
1.05E01
1.99E01

6.32E02
3.27E02
2.80E02
1.38E01

1.93E03
3.14E05
1.24E05
1.39E02

4.85E-04
8.83E-07
2.02E-07
4.58E-03

252

5.08E05
3.02E09
3.12E10
4.31E04

7.53E-05
8.12E-09
8.05E-10
8.09E-04

Table A-1 (cont’d)
H+M vs.
H+M+L

2.41E01

2.51E01

2.11E01

5.37E5.37E1.82E1.04E6.04E-01
7.21E-02
02
02
02
03
LPUAI Targeting Method
2.13E2.11E5.63E1.88E8.45E6.14E1.65EB vs. H
1.88E-02
1.21E-02
01
01
02
02
03
03
03
1.17E1.92E4.61E4.23E1.01E1.06E0.00E+
B vs. H+M
2.33E-07
01
01
02
04
09
1.01E-09 11
00
B vs.
1.10E1.05E2.80E1.24E8.05E1.11E0.00E+
H+M+L
01
01
02
05
2.02E-07 10
8.05E-10 12
00
1.37E2.23E1.35E1.70E7.65E2.26E5.55EH vs. H+M
4.15E-05
7.65E-07
01
01
01
02
07
08
15
H+M vs.
2.53E1.41E2.50E2.72E2.06E7.31E6.55E2.20E-01
2.06E-01
H+M+L
01
01
01
02
01
02
02
Cell having darker color are significantly different at p<0.05 level of significance based on AR(1) model.

253

1.44E10
2.23E03
0.00E+
00
0.00E+
00
0.00E+
00
3.64E02

Table A-2. P-values for BMPs based on AR (1) model for different targeting methods based on TN targeting (TN).
p-value

B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L

CT

NT

RM 0

7.47E01
8.47E01
8.44E01
3.10E01

7.85E01
8.78E01
8.72E01
2.72E01

4.56E02
3.84E02
3.39E02
1.64E01

CII Targeting Method
4.52E1.51E5.89E-03
04
05
5.28E7.15E7.69E-04
08
08
3.16E3.81E6.18E-04
08
10
3.42E4.37E1.21E-01
02
02

2.47E01

1.10E01

2.62E01

2.89E01

B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L

3.65E01
8.44E01

3.57E01
8.72E01

-

CF

1.52E01
3.39E02

-

-

7.02E01

7.32E01

5.89E02

4.97E01
6.65E01
8.44E01

5.16E01
6.88E01
8.72E01

5.97E02
1.32E02
3.39E02

RM 1000

SC

3.07E4.97E-01 01
LII Targeting Method
5.76E4.13E8.86E-02
02
02
3.16E3.81E6.18E-04
08
10
6.13E3.43E05
3.60E-03
06
LPSAI Targeting Method
6.06E1.10E6.23E-03
04
04
8.74E1.31E8.22E-04
08
07
3.16E3.81E6.18E-04
08
10

254

RM2000

3.70E-03
4.92E-04
3.94E-04
1.34E-01

5.36E-01

RS

T

NG

3.01E03
7.97E04
1.11E04
3.78E02

1.11E
-03
2.97E
-06
1.99E
-06
3.40E
-01

4.47E
-09
0.00E
+00
0.00E
+00
3.64E
-02

3.62E02

3.30E
-01

3.59E
-01

7.98E-02
3.94E-04

1.04E01
1.11E04

8.00E-04
4.32E-03
5.60E-04
3.94E-04

2.38E
-02
6.12E
-12

-

1.56E
-02
1.21E
-08

-

-

1.02E03

2.44E
-09

1.89E
-06

1.62E02
1.54E03
1.11E04

1.83E
-06
1.23E
-11
6.12E
-12

6.58E
-07
5.14E
-11
1.21E
-08

Table A-2 (cont’d)
H vs. H+M
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L

3.91E01
3.86E01

3.66E01
3.29E01

1.63E01
2.38E01

4.08E01
6.96E01
8.44E01
4.97E01
3.57E01

4.16E01
5.48E01
8.72E01
3.56E01
2.84E01

3.96E02
3.54E02
3.39E02
1.27E01
2.70E01

1.10E9.18E7.98E-02
02
03
2.41E2.26E4.37E-01
01
01
LPUAI Targeting Method
2.40E1.22E1.36E-02
03
03
1.32E4.71E6.61E-04
04
08
3.16E3.81E6.18E-04
08
10
7.34E5.38E1.97E-02
02
04
1.43E3.45E5.16E-01
01
01

8.60E-02
2.81E-01

1.20E-02
4.20E-04
3.94E-04
1.70E-02
5.57E-01

6.86E02
4.12E02

3.39E
-04
2.83E
-01

4.56E
-03
2.47E
-01

8.70E02
1.32E03
1.11E04
1.20E02
5.05E02

1.35E
-04
2.99E
-12
6.12E
-12
2.30E
-06
4.85E
-01

5.75E
-05
0.00E
+00
0.00E
+00
1.70E
-04
3.96E
-01

Cell having darker color are significantly different at p<0.05 level of significance based on AR(1) model

255

Table A-3. P-values for BMPs based on AR (1) model for different targeting methods based on TN targeting (TP).
p-value

CT

NT

RM 0

CF

RM 1000

SC

RM2000

RS

T

NG

8.19E05
1.12E05
9.75E06
5.75E02
2.11E01

6.02E12
2.22E16
0.00E+0
0
1.03E03
8.80E03

4.44E16
0.00E+0
0
0.00E+0
0
3.66E05
1.46E04

7.77E15
0.00E
+00
0.00E
+00
4.01E04
3.32E01

1.38E02
9.75E06
6.39E05

2.30E03
0.00E+0
0
6.66E16

1.69E03
0.00E+0
0
0.00E+0
0

1.74E04
1.52E05
9.75E06
6.33E03

1.11E07
3.18E13
0.00E+0
0
7.15E05

8.30E11
0.00E+0
0
0.00E+0
0
3.12E08

CII Targeting Method
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M

2.50E01
3.42E01
3.46E01
6.96E02
5.50E02
8.83E02
3.46E01
2.55E01
1.82E01
3.42E01
3.46E01
1.25E01

9.35E01
6.50E01
6.33E01
1.04E01
5.77E02

8.85E01
9.09E01
8.94E01
1.05E01
2.26E01
-

1.33E01
8.94E01
7.77E- 9.20E01
01
8.41E02
6.33E01

7.11E01
6.73E01
6.33E01
2.67E01

6.02E01
9.77E01
8.94E01
2.78E01

1.48E06
2.78E08
2.06E08
2.49E02
1.35E01

3.86E10
3.06E3.17E-04
14
1.82E2.84E-04
14
4.08E7.94E-02
03
1.29E2.04E-01
01
LII Targeting Method
1.06E4.09E2.19E-02
02
03
1.16E0.00E+0
2.84E-04
02
0
4.03E0.00E+0
07
8.10E-04
0
LPSAI Targeting Method
1.04E1.47E1.40E-03
05
07
4.94E1.10E1.01E-04
08
13
2.06E1.82E2.84E-04
08
14
6.81E1.40E1.42E-02
03
04
1.26E-03

256

1.27E03
0.00E
+00
1.33E15
3.90E11
0.00E
+00
0.00E
+00
4.71E08

Table A-3 (cont’d)
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L

6.04E02

6.54E02

2.28E01

1.02E01
3.77E01
3.46E01
2.30E01
7.38E02

3.16E01
6.93E01
6.33E01
1.73E01
2.50E01

1.55E01
9.00E01
8.94E01
8.67E01
2.22E01

9.63E5.97E1.58E-01
02
02
LPUAI Targeting Method
2.67E8.47E4.87E-03
03
05
1.57E1.91E1.19E-03
06
14
2.06E1.82E9.37E-04
08
14
1.64E1.05E4.60E-03
03
08
3.48E1.53E2.10E-01
02
01

1.54E01

3.76E04

2.05E02

1.01E01

2.02E03
9.93E06
9.75E06
7.27E04
2.22E01

1.75E04
1.01E13
0.00E+0
0
7.55E08
1.00E03

8.15E06
0.00E+0
0
0.00E+0
0
0.00E+0
0
1.30E01

9.02E06
0.00E
+00
0.00E
+00
8.08E13
4.04E01

Cell having darker color are significantly different at p<0.05 level of significance based on AR(1) model

257

Table A-4. P-values for BMPs based on AR (1) model for different targeting methods based on TP targeting (sediment).
p-value
CT
NT
RM 0 CF
RM 1000 SC
RM2000 RS
T
NG

B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.

CII Targeting Method
2.17E4.91E5.35E-04
03
05
2.66E3.21E1.55E-05
04
06
1.24E3.12E2.02E-07
05
10
5.50E4.55E3.05E-02
02
02
2.76E5.24E1.79E-02
02
04
LII Targeting Method
1.56E- 1.52E- 8.12E- 2.92E8.26E5.41E-04
01
01
02
03
05
1.10E- 1.05E- 2.80E- 1.24E3.12E2.02E-07
01
01
02
05
10
1.87E- 1.91E- 1.08E- 6.10E7.71E1.94E-03
01
01
01
03
05
LPSAI Targeting Method
1.62E- 1.59E- 5.20E- 1.41E7.49E3.90E-04
01
01
02
03
03
1.23E- 1.19E- 3.27E- 3.14E2.96E8.83E-07
01
01
02
05
09
1.10E- 1.05E- 2.80E- 1.24E3.12E2.02E-07
01
01
02
05
10
1.97E- 2.01E- 1.63E- 1.79E1.74E5.55E-03
01
01
01
02
06
2.41E- 2.51E- 2.11E- 1.11E- 1.08E-01 5.41E1.64E01
1.37E01
1.10E01
2.13E01
2.09E01

1.62E01
1.34E01
1.05E01
2.20E01
2.16E01

6.77E02
5.57E02
2.80E02
2.04E01
1.40E01

258

7.77E05
1.34E06
8.05E10
1.36E02
5.55E03
9.25E05
8.05E10
1.90E04
5.21E05
8.12E09
8.05E10
1.13E03
7.21E-

8.52E06
5.88E09
1.11E12
3.93E03
5.97E04

4.45E08
0.00E
+00
0.00E
+00
2.69E04
2.00E07
-

2.79E05
1.11E12

4.16E10
4.16E10
0.00E
+00
1.78E01
0.00E
+00
-

5.63E07
0.00E
+00
-

1.50E06
0.00E
+00

-

-

1.18E06

1.34E11

0.00E
+00

8.52E06
1.11E09
1.11E12
1.15E04
1.69E-

3.91E08
0.00E
+00
0.00E
+00
8.33E08
1.04E-

7.52E10
0.00E
+00
0.00E
+00
7.10E12
4.49E-

Table A-4 (cont’d)
H+M+L

01

01

01

01
02
02
02
03
LPUAI Targeting Method
2.22E- 2.20E- 1.18E- 6.98E5.02E- 6.24E- 7.12E- 2.88EB vs. H
7.09E-02
01
01
01
02
02
02
02
02
1.21E- 1.17E- 2.78E- 1.52E6.81E- 1.91E- 2.73E- 0.00E
B vs. H+M
3.48E-07
01
01
02
05
10
09
11
+00
B vs.
1.10E- 1.05E- 2.80E- 1.24E8.05E- 8.05E- 1.11E- 0.00E
2.02E-07
H+M+L
01
01
02
05
10
10
12
+00
1.33E- 1.30E- 6.87E- 1.68E2.82E- 5.38E- 5.45E- 0.00E
H vs. H+M
5.50E-06
01
01
02
04
08
08
10
+00
H+M vs.
2.48E- 2.59E- 2.51E- 1.73E1.28E- 1.56E- 4.07E- 2.54E1.83E-01
H+M+L
01
01
01
01
01
01
02
02
Cell having darker color are significantly different at p<0.05 level of significance based on AR(1) model

259

10
4.77E02
0.00E
+00
0.00E
+00
0.00E
+00
1.89E03

Table A-5. P-values for BMPs based on AR (1) model for different targeting methods based on TP targeting (TN).
p-value
CT
NT
RM 0
CF
RM 1000
SC
RM2000
RS
T
NG

B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.

CII Targeting Method
5.26E9.32E9.51E-03
01
05
6.08E3.81E4.53E-03
01
08
2.95E3.81E6.18E-04
01
10
3.26E1.93E2.63E-01
01
01
4.29E3.66E1.34E-01
01
02
LII Targeting Method
4.71E- 4.63E- 7.05E- 2.95E1.41E1.62E-02
01
01
02
04
03
8.44E- 8.72E- 3.39E- 3.16E3.81E6.18E-04
01
01
02
08
10
5.63E- 5.49E- 5.29E- 8.02E2.97E1.95E-02
01
01
02
04
04
LPSAI Targeting Method
5.01E- 5.20E- 4.68E- 4.13E8.54E5.30E-03
01
01
02
04
02
6.65E- 6.88E- 1.32E- 8.74E1.24E8.22E-04
01
01
02
08
07
8.44E- 8.72E- 3.39E- 3.16E3.81E6.18E-04
01
01
02
08
10
3.86E- 3.65E- 2.01E- 4.48E4.68E9.37E-02
01
01
01
02
07
3.86E- 3.29E- 2.38E- 2.41E4.37E-01 2.32E5.26E01
6.08E01
8.44E01
3.26E01
4.29E01

5.45E01
6.26E01
8.72E01
2.93E01
3.71E01

1.07E01
3.85E02
3.39E02
2.95E01
1.77E01

260

6.38E03
2.85E03
3.94E04
2.67E01
1.46E01
1.32E02
3.94E04
1.86E02
3.49E03
5.60E04
3.94E04
1.04E01
2.81E-

5.74E03
1.40E03
1.11E04
1.58E01
5.92E02

9.92E07
1.13E08
6.12E12
5.48E02
1.20E02
-

4.33E02
1.11E04

1.74E06
2.41E09
1.21E08
1.14E01
3.46E02
-

6.12E12
6.12E12
-

2.27E04
1.21E08

-

-

3.66E03

2.76E06

1.47E04

1.60E02
1.61E03
1.11E04
7.71E02
8.76E-

5.69E07
1.23E11
6.12E12
6.39E04
2.83E-

3.87E07
4.43E11
1.21E08
5.72E03
2.60E-

Table A-5(cont’d)
H+M+L

01

01

01

01
01
01
02
01
LPUAI Targeting Method
3.52E- 3.51E- 1.15E- 4.33E3.62E- 8.32E- 1.75E- 1.94EB vs. H
8.30E-02
01
01
01
02
02
02
01
02
6.43E- 6.61E- 3.12E- 8.70E1.33E- 3.87E- 2.34E- 2.34EB vs. H+M
6.21E-04
01
01
02
08
07
04
03
11
B vs.
8.44E- 8.72E- 3.39E- 3.16E3.81E- 3.94E- 1.11E- 6.12E6.18E-04
H+M+L
01
01
02
08
10
04
04
12
5.46E- 5.58E- 1.04E- 2.14E1.37E- 4.34E- 6.37E- 1.39EH vs. H+M
6.14E-03
01
01
01
04
05
03
03
08
H+M vs.
4.00E- 3.47E- 4.81E- 4.26E2.61E- 5.57E- 3.23E- 3.18E5.18E-01
H+M+L
01
01
01
01
01
01
02
01
Cell having darker color are significantly different at p <0.05 level of significance based on AR (1) model.

261

01
2.52E02
7.22E10
1.21E08
8.47E08
1.88E01

Table A-6. P-values for BMPs based on AR (1) model for different targeting methods based on TP targeting (TP).
p-value
CT
NT
RM 0
CF
RM 1000
SC RM2000
RS
T
NG

B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M
H+M vs.
H+M+L
B vs. H
B vs. H+M
B vs.
H+M+L
H vs. H+M

CII Targeting Method
4.85E9.36
1.35E1.38E-03
06
E-07
04
7.33E1.41
5.63E9.34E-04
07
E-10
05
2.06E1.82
9.75E2.84E-04
08
E-14
06
5.98E4.67
1.14E1.48E-01
02
E-02
01
3.43E1.92
7.54E1.01E-01
02
E-03
02
LII Targeting Method
1.62E- 4.75E- 3.73E- 6.38E2.40
1.45E4.93E-03
01
01
01
04
E-05
03
0.00
3.46E- 6.33E- 8.94E- 2.06E9.75E2.84E-04 E+0
01
01
01
08
06
0
1.62E- 4.79E- 4.95E- 7.23E8.01
1.10E4.93E-03
01
01
01
05
E-02
03
LPSAI Targeting Method
1.85E- 7.32E- 5.35E- 6.63E8.56
1.31E1.62E-03
01
01
01
06
E-05
04
3.42E- 6.73E- 9.77E- 4.94E8.10
1.52E1.01E-04
01
01
01
08
E-14
05
3.46E- 6.33E- 8.94E- 2.06E1.82
9.75E2.84E-04
01
01
01
08
E-14
06
1.27E- 2.67E- 3.36E- 1.03E2.68
9.21E1.54E-02
01
01
01
02
E-09
03
2.15E01
2.88E01
3.46E01
8.57E02
7.40E02

8.58E01
8.68E01
6.33E01
1.29E01
1.05E01

8.24E01
8.89E01
8.94E01
1.59E01
3.17E01

262

3.36E09
1.15E11
0.00E+0
0
6.84E03
4.63E05

1.62E14
0.00E
+00
0.00E
+00
2.15E05
1.46E04

4.86E13
2.22E15
0.00E
+00
1.65E03
6.68E04

1.52E05

-

-

5.35E07

1.25E06

0.00E+0
0

0.00E
+00

0.00E
+00

3.11E10

3.35E01

4.49E11

7.46E08
4.06E13
0.00E+0
0
1.44E04

3.06E12
0.00E
+00
0.00E
+00
9.15E08

1.55E11
0.00E
+00
0.00E
+00
8.32E09

-

-

Table A-6 (cont’d)
H+M vs.
H+M+L

6.04E02

6.54E02

2.28E01

9.63E6.57
1.54E3.59E- 2.05E- 1.16E1.58E-01
02
E-02
01
04
02
01
LPUAI Targeting Method
6.97E- 1.28E- 6.95E- 6.40E3.19
9.87E4.45E- 1.46E- 2.76EB vs. H
1.36E-02
02
01
02
03
E-03
03
03
03
03
3.26E- 7.19E- 9.43E- 3.67E8.79
6.76E8.75E0.00E 0.00E
B vs. H+M
1.89E-04
01
01
01
08
E-14
06
14
+00
+00
B vs.
3.46E- 6.33E- 8.94E- 2.06E1.82
9.75E- 0.00E+0 0.00E 0.00E
9.37E-04
H+M+L
01
01
01
08
E-14
06
0
+00
+00
2.66E- 9.57E- 9.17E- 5.75E1.53
8.40E7.13E0.00E 1.40EH vs. H+M
1.12E-03
01
01
01
06
E-10
05
11
+00
14
H+M vs.
7.39E- 9.01E- 3.32E- 1.17E7.98
2.34E2.40E- 3.23E- 1.40E2.36E-01
H+M+L
02
02
01
01
E-02
01
04
02
14
Cell having darker color are significantly different at p <0.05 level of significance based on AR (1) model.

263

Table A-7. Spatial correlation among the targeting methods for different targeting scenarios.
2
Parameter
Kappa (95% CI)
Weighted Kappa
β (p-value)
G
(95% CI)
(df=2)
CII vs. LII
0.24 (0.09,0.38)
0.20 (0.05,0.35)
2.60
-0.15 (0.77)
CII vs. LPSAI
0.50 (0.33,0.67)
0.56 (0.36,0.76)
9.65
0.79 (0.09)
CII vs. LPUAI
0.37 (0.22,0.51)
0.30 (0.14,0.46)
0.26
-0.12 (0.76)
LII vs. LPSAI
0.32 (0.18,0.45)
0.37 (0.22,0.52)
14.89
0.48 (0.20)
LII vs. LPUAI
0.27 (0.15,0.39)
0.31 (0.17,0.44)
16.80
0.47 (0.12)
LPSAI vs.
LPUAI
0.56 (0.43,0.68)
0.63 (0.49,0.77)
12.78
1.10 (0.02)
CII vs. LII
0.15 (0.08,0.23)
0.10 (0.04,0.17)
12.13
-0.42 (0.28)
CII vs. LPSAI
0.29 (0.19,0.38)
0.39 (0.27,0.51)
0.33
1.26 (0.00)
CII vs. LPUAI
0.29 (0.20,0.38)
0.48 (0.39,0.58)
28.39
1.58 (0.00)
LII vs. LPSAI
-0.06 (-0.10,-0.02) -0.05 (-0.14,0.04)
2.20
6.42 (1.00)
LII vs. LPUAI
-0.02 (-0.09,0.06) -0.03 (-0.12,0.06)
2.55
-0.34 (0.63)
LPSAI vs.
LPUAI
0.51 (0.39,0.63)
0.59 (0.45,0.72)
6.74
0.89 (0.03)
TP targeting scenario
CII vs. LII
-0.08 (-0.13,-0.02) -0.07 (-0.16,0.03)
1.53
6.53 (1.00)
CII vs. LPSAI
-0.10 (-0.13,-0.08) -0.13 (-0.16,-0.09)
0.00
-7.28 (1.00)
CII vs. LPUAI
0.24 (0.15,0.33)
0.43 (0.30,0.55)
29.05
1.13 (0.00)
LII vs. LPSAI
0.70 (0.58,0.82)
0.81 (0.72,0.90)
0.00
8.94 (1.00)
LII vs. LPUAI
-0.14 (-0.21,-0.07) -0.17 (-0.23,-0.11)
0.00
-15.49 (1.00)
LPSAI vs.
LPUAI
-0.13 (-0.17,-0.10) -0.15 (-0.19,-0.11)
0.00
-7.07 (1.00)

264

δ (p-value)

τ

1.24 (0.00)
1.41 (0.00)
1.99 (0.00)
0.91 (0.01)
0.61 (0.04)

10.20
36.91
47.39
9.90
5.42

1.57 (0.00)
1.75 (0.00)
0.53 (0.06)
-0.02 (0.94)
-13.13(1.00)
0.12 (0.85)

69.61
21.79
10.09
4.65
0.00
0.90

1.08 (0.00)

21.19

-13.45(1.00)
-8.72 (1.00)
-0.21 (0.65)
9.99 (1.00)
7.00 (1.00)

0.00
0.00
2.05
3.6E+1
0.22

-9.87 (1.00)

0.00

(a)

(b)

(d)

(c)

(f)

(e)

N

0

10

Legend
Sediment (mg/l)
Low
Medium
High

20

Figure A-8. CII priority areas – sediment targeting scenario. (a) base scenario, (b) year one contour farming, (c) year two
contour farming, (d) base scenario, (e) year one native grass, (f) year two native grass.

265

(a)

(b)

(c)

(d)

(e)

(f)

N

0

10

Legend
Sediment
(metric tons)
Low
Medium
High

20

Figure A-9. LPSAI priority areas – sediment targeting scenario. (a) base scenario, (b) year one contour farming, (c) year two
contour farming, (d) base scenario, (e) year one native grass, (f) year two native grass.
266

(a)

(b)

(c)

(d)

(e)

(f)

N

0

10

Legend
Sediment
(ton/ha)
Low
Medium
High

20

Figure A-10. LPUAI priority areas – sediment targeting scenario. (a) base scenario, (b) year one contour farming, (c) year two
contour farming, (d) base scenario, (e) year one native grass, (f) year two native grass.
267

(a)

(b)

(e)

(d)

(c)

(f)

N

0

10

Legend
Total Nitrogen
(mg/l)
Low
Medium
High

20

Figure A-11. CII priority areas – TN targeting scenario. (a) base scenario, (b) year one contour farming, (c) year two contour
farming, (d) base scenario, (e) year one native grass, (f) year two native grass.

268

(a)

(b)

(c)

(d)

(e)

(f)

N

0

10

Legend
Total Nitrogen
(kg)
Low
Medium
High

20

Figure A-12. LPSAI priority areas – TN targeting scenario. (a) base scenario, (b) year one contour farming, (c) year two
contour farming, (d) base scenario, (e) year one native grass, (f) year two native grass.
269

(a)

(c)

(b)

(f)

(e)

(d)

N

0

10

Legend
Total Nitrogen
(kg/ha)
20

Low
Medium
High

Figure A-13. LPUAI priority areas – TN targeting scenario. (a) base scenario, (b) year one contour farming, (c) year two
contour farming, (d) base scenario, (e) year one native grass, (f) year two native grass.
270

(a)

(b)

(e)

(d)

(c)

(f)

N

0

10

Legend
Total Phosphorus
(mg/l)
Low
Medium
High

20

Figure A-14. CII priority areas – TP targeting scenario. (a) base scenario, (b) year one contour farming, (c) year two contour
farming, (d) base scenario, (e) year one native grass, (f) year two native grass.

271

(c)

(b)

(a)

(d)

(f)

(e)

N

0

10

Legend
Total
Phosphorus (kg)
Low
Medium
High

20

Figure A-15. LPSAI priority areas – TP targeting scenario. (a) base scenario, (b) year one contour farming, (c) year two
contour farming, (d) base scenario, (e) year one native grass, (f) year two native grass.
272

(b)

(d)

(c)

(e)

(a)

(f)

N

0

10

Legend
Total Phosphorus
(kg/ha)
Low
Medium
High

20

Figure A-16. LPUAI priority areas – TP targeting scenario. (a) base scenario, (b) year one contour farming, (c) year two
contour farming, (d) base scenario, (e) year one native grass, (f) year two native grass.
273

(APPENDIX B – ADDITIONAL MATERIAL TO SECTION 6 TITTLED
“APPLICATION OF ANALYTICAL HIERARCHY PROCESS FOR EFFECTIVE
SELECTION OF AGRICULTURAL BEST MANAGEMENT PRACTICES.”)

274

APPENDIX B

N

Legend
Saginaw
Subbasin 43

Figure B-1. Subbasins boundaries within the Saginaw River Watershed.

275

Table B-2. Pairwise comparison matrix developed for subbasin 43 based on sediment
reduction at the watershed outlet.
BMPs
SC
RM
CT
NG
NT

SC
1.00
0.88
0.95
0.95
0.92

RM
1.13
1.00
1.08
1.08
1.04

CT
1.05
0.93
1.00
1.00
0.96

NG
1.05
0.93
1.00
1.00
0.97

NT
1.09
0.96
1.04
1.03
1.00

Table B-3. Pairwise comparison matrix developed for subbasin 43 based on TN reduction
at the watershed outlet.
BMs
SC
RM
CT
NG
NT
SC
1.00
1.43
1.61
0.72
2.31
RM
0.70
1.00
1.13
0.50
1.61
CT
0.62
0.89
1.00
0.45
1.43
NG
1.39
2.00
2.25
1.00
3.22
NT
0.43
0.62
0.70
0.31
1.00

Table B-4. Pairwise comparison matrix developed for subbasin 43 based on TP reduction
at the watershed outlet.
BMs
SC
RM
CT
NG
NT
SC
1.00
1.26
1.63
0.66
2.13
RM
0.79
1.00
1.29
0.52
1.69
CT
0.61
0.77
1.00
0.41
1.30
NG
1.51
1.91
2.47
1.00
3.22
NT
0.47
0.59
0.77
0.31
1.00

Table B-5. Weight vector calculation of BMPs for sediment, TN, and TP reduction for
subbasin 43 at the watershed level.
BMPs
SC
RM
CT
NG
NT

Sediment-Weight
0.2117
0.1875
0.2028
0.2020
0.1961

TN-Weight
0.2405
0.1681
0.1493
0.3377
0.1045

276

TP-Weight
0.2271
0.1801
0.1394
0.3461
0.1073

Table B-6. Decision matrix of BMPs for all criteria developed for watershed level analysis.
BMPs
Total BMP
Sediment
TN
TP
BMP
cost
reduction
reduction
reduction
Application
Area
SC
0.4440
0.2117
0.2405
0.2271
0.0770
RM
0.2513
0.1875
0.1681
0.1801
0.2984
CT
0.1511
0.2028
0.1493
0.1394
0.2984
NG
0.0424
0.2020
0.3377
0.3461
0.0278
NT
0.1112
0.1961
0.1045
0.1073
0.2984

Table B-7. Final weight vector of individual BMPs for subbasin 43.
BMPs
Weight
(watershed outlet)
SC
0.2467
RM
0.2403
CT
0.2024
NG
0.1206
NT
0.1800

277

Table B-8. BMP ranking at subbasin level based on environmental, economic, and social
factors.
Subbasin
SC
RM
CT
NG
NT
43
1
2
3
4
5
46
1
2
3
4
5
59
1
2
3
4
5
62
1
2
4
3
5
83
1
2
4
3
5
86
1
2
4
3
5
88
1
2
4
3
5
89
1
2
4
3
5
90
1
2
3
4
5
91
1
2
3
5
4
94
1
2
3
4
5
95
1
2
3
4
5
96
1
2
3
5
4
99
1
2
3
5
4
103
1
2
3
5
4
106
1
2
4
3
5
109
1
2
3
5
4
115
1
2
3
4
5
116
1
2
3
4
5
127
1
2
3
4
5
129
1
2
3
4
5
135
1
2
3
4
5
136
1
2
3
4
5
140
1
2
3
4
5
142
1
2
3
5
4
147
1
2
3
4
5
148
1
2
4
3
5
152
1
2
3
4
5
159
1
2
3
4
5
181
1
2
3
4
5
184
1
2
3
4
5
187
1
2
3
4
5
188
1
2
3
4
5
205
1
2
3
4
5
214
1
2
3
4
5
215
1
2
3
4
5
219
1
2
3
5
4

278

Table B-9. BMP ranking at watershed level based on environmental, economic, and social
factors.
Subbasin
SC
RM
CT
NT
NG
43
1
2
3
4
5
46
1
2
3
4
5
59
1
2
3
4
5
62
1
2
3
4
5
83
1
2
3
4
5
86
1
2
3
4
5
88
1
2
3
4
5
89
1
2
3
4
5
90
1
2
3
4
5
91
1
2
3
4
5
94
1
2
3
4
5
95
1
2
3
4
5
96
1
2
3
4
5
99
1
2
3
4
5
103
1
2
3
4
5
106
1
2
3
4
5
109
1
2
3
4
5
115
1
2
3
4
5
116
4
1
3
2
5
127
4
1
3
2
5
129
4
1
3
2
5
135
4
1
3
2
5
136
4
1
3
2
5
140
4
1
3
2
5
142
4
1
3
2
5
147
4
1
3
2
5
148
3
1
4
2
5
152
3
1
4
2
5
159
3
1
4
2
5
181
3
1
4
2
5
184
3
1
4
2
5
187
2
1
4
3
5
188
2
1
4
3
5
205
1
2
4
3
5
214
1
2
4
3
5
215
1
2
4
3
5
219
1
2
4
3
5

279

Table B-10. BMP ranking at subbasin level based on environmental and economic factors.
Subbasin
SC
RM
NG
CT
NT
43
1
2
3
4
5
46
1
2
3
4
5
59
1
2
3
4
5
62
1
3
2
4
5
83
1
3
2
4
5
86
1
3
2
4
5
88
1
3
2
4
5
89
1
3
2
4
5
90
1
2
3
4
5
91
1
2
3
4
5
94
1
3
2
4
5
95
1
3
2
4
5
96
1
2
3
4
5
99
1
2
3
4
5
103
1
2
3
4
5
106
1
3
2
4
5
109
1
2
3
4
5
115
1
3
2
4
5
116
1
3
2
4
5
127
1
3
2
4
5
129
1
3
2
4
5
135
1
2
3
4
5
136
1
2
3
4
5
140
1
2
3
4
5
142
1
3
2
4
5
147
1
2
3
4
5
148
1
3
2
4
5
152
1
3
2
4
5
159
1
2
3
4
5
181
1
2
3
4
5
184
1
2
3
4
5
187
1
2
3
4
5
188
1
2
3
4
5
205
1
2
3
4
5
214
1
2
3
4
5
215
1
2
3
4
5
219
1
2
3
4
5

280

Table B-11. BMP ranking at watershed level based on environmental and economic factors.
Subbasin
SC
RM
NG
CT
NT
43
1
2
3
4
5
46
1
2
4
3
5
59
1
2
5
3
4
62
1
2
5
3
4
83
1
2
5
3
4
86
1
2
5
3
4
88
1
2
4
3
5
89
1
2
3
4
5
90
1
2
5
3
4
91
1
2
5
3
4
94
1
2
3
4
5
95
1
2
3
4
5
96
1
2
3
4
5
99
1
2
5
3
4
103
1
2
3
4
5
106
1
2
5
3
4
109
1
2
3
4
5
115
1
2
3
4
5
116
2
1
5
4
3
127
2
1
5
4
3
129
2
1
5
4
3
135
2
1
5
4
3
136
2
1
5
4
3
140
2
1
5
4
3
142
1
3
2
4
5
147
2
1
5
4
3
148
2
1
5
4
3
152
1
2
5
4
3
159
2
1
5
4
3
181
1
2
5
4
3
184
1
2
5
4
3
187
1
2
3
5
4
188
1
2
4
5
3
205
1
2
4
5
3
214
1
2
4
5
3
215
1
2
4
5
3
219
1
2
4
5
3

281

Table B-12. BMP ranking at subbasin level based on environmental factors.
Subbasin
NG
SC
RM
NT
43
1
2
3
4
46
1
2
3
4
59
1
2
3
5
62
1
2
3
5
83
1
2
3
5
86
1
2
3
4
88
1
2
3
4
89
1
2
3
4
90
1
2
3
4
91
1
2
3
5
94
1
2
3
4
95
1
2
3
4
96
1
2
3
4
99
1
2
3
5
103
1
2
3
5
106
1
2
3
4
109
1
2
3
5
115
1
2
3
5
116
1
2
3
4
127
1
2
3
5
129
1
2
3
5
135
1
2
3
5
136
1
2
3
4
140
1
2
3
5
142
1
2
3
5
147
1
2
3
5
148
1
2
3
4
152
1
2
3
5
159
1
2
3
4
181
1
2
3
5
184
1
2
3
4
187
1
2
3
5
188
1
2
3
5
205
1
2
3
5
214
1
2
3
4
215
1
2
3
4
219
1
2
3
4

282

CT
5
5
4
4
4
5
5
5
5
4
5
5
5
4
4
5
4
4
5
4
4
4
5
4
4
4
5
4
5
4
5
4
4
4
5
5
5

Table B-13. BMP ranking at watershed level based on environmental factors.
Subbasin
NG
SC
RM
CT
43
1
2
3
4
46
2
1
3
4
59
3
1
2
4
62
5
1
2
3
83
5
1
2
3
86
2
1
3
5
88
2
1
3
4
89
2
1
3
4
90
5
1
2
3
91
5
1
2
3
94
2
1
3
4
95
2
1
3
4
96
2
1
3
4
99
5
1
2
3
103
2
1
3
4
106
4
1
2
3
109
2
1
3
4
115
2
1
3
4
116
3
5
1
4
127
4
5
1
3
129
4
5
1
3
135
4
5
1
3
136
4
5
1
3
140
3
5
1
4
142
1
2
3
4
147
3
5
1
4
148
3
5
1
4
152
3
5
1
4
159
4
5
1
3
181
3
4
1
5
184
5
3
1
4
187
1
4
3
5
188
1
2
4
5
205
2
1
4
5
214
2
1
4
5
215
3
1
4
5
219
2
1
4
5

283

NT
5
5
5
4
4
4
5
5
4
4
5
5
5
4
5
5
5
5
2
2
2
2
2
2
5
2
2
2
2
2
2
2
3
3
3
2
3

REFERENCES

284

REFERENCES

Akerblom, S., Meili, M., Bringmark, L., Johansson, K., Kleja, D.B., Bergkvist, B., 2008.
Partitioning of Hg between solid and dissolved organic matter in humus layer of boreal
forests. Water Air and Soil Pollution 189, 239-252.
Alabaster, J.S., 1972. Suspended solids and fisheries. Proceeding of the Royal Society, London,
B 180, 395-406.
Alcamo, J., Vorosmarty, C., Naiman, R. J., Lettenmaier, D., Pahl-Wostl, C., 2008. A grand
challenge for freshwater research: understanding the global water system. Environmental
Research Letters 3, 1–6.
Agresti, A., 1988. A model for agreement between ratings on an ordinal scale. International
Journal of Biometeorology 44, 539-548.
Arabi, M., Frankenberger, J.R., Engel, B.A., Arnold, J.G., 2007. Representation of agricultural
conservation practices with SWAT. Hydrologic Processes 22, 3042-3055.
Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. . Large area hydrologic model
development and assessment part 1: Model development. Journal of the American Water
Resources Association 34, 73–89.
ASWP, 2012. Pollution assessment and prevention in Alabama. Available at:
http://www.aces.edu/waterquality/themes/pollution.htm. Accessed on September 25,
2012.
Auld, H., MacIver, D., Klaassen, J., 2004. Heavy rainfall and waterborne disease outbreaks: The
Walkerton example. Journal of Toxicology and Environmental Health Part A 67, 1879–
1887.
Babu, V., Unnikrishnan, P., Anu, G., Nair, S. M., 2011. Distribution of Organophosphorus
Pesticides in the Bed Sediments of a Backwater System Located in an Agricultural
Watershed: Influence of Seasonal Intrusion of Seawater. Archives of Environmental
Contamination and Toxicology 60, 597–609.
Balascio, C. C., Palmeri, D. J., Gao, H., 1998. Use of a genetic algorithm and multiprogramming for calibration of a hydrologic model. Transaction of the ASAE 41, 615619.
Benham, B.L., Baffaut, C., Zeckoski, R.W., Mankin, K.R.,Pachepsky, Y.A., Sadeghi, A.M.,
Brannan, K.M., Soupir, M.L.,Habersack, M.J., 2006. Modeling bacteria fate and transport
in watersheds to support TMDLs. Transection of ASABE 49 (4), 987-1002.

285

Benstead, J.P., Rosemond, A.D., Cross, W.F., Wallace, J.B., Eggert, S.L., Suberkropp, K., Gulis,
V., Greenwood, J.L., Tant, C, J., 2009. Nutrient enrichment alters storage and fluxes of
detritus in a headwater stream ecosystem. Ecology 90, 2556–2566.
Bhaduri , B., Harbor, J., Bengel, B., Grove, M., 2000. Assessing watershed scale, long term
hydrologic impacts of landuse change using a GIS-NPS model. Environmental
Management 26, 643-658.
Bhaduri, B., Grove, M., Lowry, C., Harbor, J., 1997. Assessment of long term hydrologic impact
from landuse change in the Cuppy-McClure watershed, Indiana. Journal of the American
Water Works Association, 89(11), 94-106.
Bicknell, B.R., Imhoff, J.C., Kittle, J.C., Jr, J.L., Donigian, A.S., Johanson, R.C., 1997.
Hydrological Simulation Program- Fortran, User’s manual for version 11: U.S.
Environmental Protection Agency, National Exposure research Laboratory, Athens, G.,
EPA/600/R-97/080, 755 p. Biogeochemical model pnET_N_DNDC, Journal of
Geophysical research 106: 34155-34165.
Bicknell, B.R., Imhoff, J.C., Kittle, J.C., Jobes, T.H., Donigian, A.S., 2000. Hydrological
Simulation Program- Fortran, User’s manual for release 12. Athens, Ga.: USEPA.
Bilotta, G.S., Brazier, R.E., 2008. Understanding the influence of suspended solids on water
quality and aquatic biota. Water Research 42 (12), 2849-2861.
Bilotta, G.S., Burnside, N.G., Cheek, L., Dunbar, M.J., Grove, M.K., Harrison, C., Joyce, C.,
Peacock, C., Davy-Bowker, J., 2012. Developing environment-specific water quality
guidelines for suspended particulate matter. Water Research 46(7), 2324-2332.
Ebionews, 2010. Cause of Dead Zone in Gulf: Tile Drainage Directly Related to Nitrate Loss
(2010) Available from: http://www.ebionews.com/news-center/research-frontiers/ag-bioa-bio-agriculture/26675-cause-of-dead-zone-in-gulf-tile-drainage-directly-related-tonitrate-loss.html. Accessed on November 25, 2012.
Borah, D.K. 2002. Watershed scale nonpoint source pollution models: Mathematical bases.
ASABE Paper No. 022091. ASABE Annual International Meeting / CIGR World
Congress. St. Joseph, Mich.: ASABE.
Bosch, D., Theurer, F., Bingner, R., Felton, G., Chaubey, I., 2001. Evaluation of the AnnAGNPS
water quality model. Agricultural Non-point Source Water Quality Models: Their use and
Application, Southern Cooperative Series Bulletin # 398, 45-62.
Boscha, N.S., Allanb, J.D., Dolanc, D.M., Hand, H., Richardse, R.P., 2011. Application of the
Soil and Water Assessment Tool for six watersheds of Lake Erie: Model parameterization
and calibration. Journal of Great lakes Research 37,263-271.
Brigham, M.E., Wentz, D.A., Aiken, G.R., Krabbenhoft, D.P., 2009. Mercury cycling in stream
ecosystems : water column chemistry and transport. Environmental Science and
technology 43, 2720-2725.
286

Brown, T.C., Froemke, P., 2012. Nationwide assessment of nonpoint source threats to water
quality. Bioscience 62(2), 136-146.
Bundy, L.G., Andraski, T.W., powell, J.M., 2001. Management practice effects on
phosphorus losses in runoff in corn production systems. Journal of Environmental
Quality 30, 1822-1828.
Buyuksonmez, F., Yang, L., Beighley, R.E., 2012. Evaluation of Best Management Practice
Products in Preventing Discharge of Metals: A Laboratory Evaluation. Journal of
Environmental Quality 41, 800–806.
Calizaya, A., Meixner, O., Bengtsson, L., Berndtsson, R., 2010. Multi-criteria decision analysis
for integrated water resources management in the Lake Poopo Basin, Bolivia. Water
Resource Management 24, 2267-2289.
Capel, P.D., Larson, S.J., Winterstein, T.A., 2001. The behavior of thirty-nine pesticides in
surface waters as a function of scale. Hydrological Process 15, 1251–1269.
Carey, C.M., Lee, H., Trevors, J.T., 2004. Biology, persistence and detection of 432
Cryptosporidium parvum and Cryptosporidium hominis oocyst. Water Resource 38, 818862.
Carpenter, S.R., 2008. Phosphorus control is critical to mitigating eutrophication. Proceeding of
National Academy of Science. USA 105, 11039–11040.
Carpenter, S.R., Caraco, N.E., Correll, D.L., Howarth, W., Sharpley, A.N., Smith, V.H., 1998.
Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecological
Application 8(3), 559-568.
Carter, A.D., 2000. Herbicide movement in soils: principles, pathways and processes. Weed
Research 40, 113-122.
CCC. 1998. State of Texas coastal nonpoint source pollution control program. Available
at:http://www.glo.texas.gov/what-we-do/caring-for-thecoast/_documents/environmental-protection/antipollution/Volume1_1998_TexasCoastalNPSPollutionProgram.pdf. Assesses on July 7,
2011.
Chang, M. 2006. Forest hydrology: An introduction to water and forests. 2nd edition. Boca Raton,
Florida. Taylor & Franscis Group.
Chiwa, M., Onikura, N., Ide, J. K, A., 2012. Impact of N-Saturated Upland Forests on
Downstream N Pollution in the Tatara River Basin, Japan. Ecosystems 15, 230–241.
Cohen, J., 1960. A coefficient of agreement for nominal scales. Educational and Psychological
Measurment 20, 37-46.

287

Covino, T., McGlynn, B., McNamara, R., 2012. Land use/land cover and scale influences on instream nitrogen uptake kinetics. Journal of Geophysical Research 117,
doi:10.1029/2011JG001874.
Cunningham, J. H., Benham, B. L., Brannan, K. M., Mostaghimi, S., Dillaha, T. A., Pease,J. W.,
2003. An assessment of the quality of the agricultural best management practices
implemented in the James River basin of Virginia. Conference paper, page,281-190.
Das, K., Gazi, N. H., 2010. Random excitations in modelling of algal blooms in estuarine
systems. Ecological Modelling 222(14), 2495-2501.
Das, S.K., Grewal, A.S., Banerjee, M., 2011. A brief review: Heavy metal and their analysis.
Internation Journal of Pharmaceutical Sciences Review and Research 11, 13–18.
Devlin, D.L., McVay, K.A., 2001. Suspended solids: a water quality concern for Kansas.
Available at: http://www.ksre.ksu.edu/library/h20ql2/mf2501.pdf. Accessed on
September 25, 2012.
Diebel, M.W; Maxted, J.T., Nowak, P.J., Zanden, M.J.V., 2008. Landscape planning for
agricultural nonpoint source pollution reduction I: a geographical allocation framework.
Environmental Management 42,789-802.
Ding, X. W., Shen, Z. Y., Hong, Q., Yang, Z. F., Wu, X., Liu, R. M., 2010. Development and
test of the export coefficient model in the upper reach of the Yangtze River. Journal of
Hydrology 383, 233–244.
Dodds, W.K., Bouska, W.W., Eitzmann, J.L., Pilger, T.J., Pitts, K.L., Riley, A.J., Schloesser,
J.T., Thornbrugh, D.J., 2009. Eutrophication of U.S. freshwaters: Analysis of potential
economic damages. Environmental Science Technology 43(1),12-19.
Driscoll, C.T., Whitall, D., Aber, J., Boyer, E., Castro, M., Cronan, C., Goodale, C.L., Groffman,
P., Hopkinson, C., Lambert, K., Lawrence, G., Ollinger, S., 2003. Nitrogen pollution in
the northeastern United States: Sources, effects, and management options. BioScience 53,
357–374.
Dubrovsky, N.M., Hamilton, P.A., 2010. The quality of our nation’s waters nutrients in the
nation’s streams and groundwater: National findings and implications. US Geological
Survey. Fact sheet 2010-3078.
Dudgeon, D. 2010. Prospects for sustaining freshwater biodiversity in the 21st century: linking
ecosystem structure and function. Current Opinion in Environmental Sustainability 2(56), 422-430.
Dudgeon, D., Arthington, A. H., Gessner, M. O., Kawabata, Z., Knowler, D., Leveque, C.,
Naiman, R. J., Prieur-Richard, A-H., Soto, D., Stiassny, M. L. J., Sullivan, C .A., 2006.
Freshwater biodiversity: importance, threats, status and conservation challenges.
Biological Review 81, 163-182.
288

Eatherall, A.,Warwick, M. S.,Tolchard, S., 2000. Identifying sources of dissolved organic carbon
on the river Swale, Yorkshire. Science of Total Environment 251-252, 173-190.
Edge, T. A., El-Shaarawi, A., Gannon, V., Jokinen, C., Kent, R., Khan, I. U. H., Koning, W.,
Lapen, D., Miller, J., Neumann, N., Phillips, R., Robertson, W., Schreier, H., Scott, A.,
Shtepani, I., Topp, E., Wilkes, G., Van Bochove, E., 2012. Investigation of an
Escherichia coli Environmental Benchmark for Waterborne Pathogens in Agricultural
Watersheds in Canada. Journal of Environmental Quality 41, 21–30.
Fleiss, J.L., Cohen, J., 1973. The equivalence of weighted kappa and the intraclass correlation
coefficient as measurements of reliability. Educational and Psychological Measurement
33, 613-619.
Enderlein,U.E., Enderlein, R.E., Williams, W.P., 2012. Water quality requirements. World
Health Organization. Available at:
http://www.who.int/water_sanitation_health/resourcesquality/wpcchap2.pdf. Accessed on
September 25, 2012.
Endreny, T.A., Wood, E.F., 2003. Watershed weighting of export coefficients to map critical
phosphorus loading areas. Journal of the American Water Resources Association 36,
1053-1062.
Environment Canada. 2010. Finding a balance between agriculture and the environment. En4135/2-2010E. National Agri-Environmental Standards Initiative (NAESI).
ERS-USDA. 1996. Production management. Available at:
http://www.ers.usda.gov/publications/arei/ah712/AH7124-2.PDF. Assessed on July 8,
2011.
Espejo, L., Kretschmer, N., Oyarzun, J., Meza, F., Nunez, J., Maturana, H., Soto, G., Oyarzo, P.,
Garrido, M., Suckel, F., Amezaga, J., Oyarzun, R., 2012. Application of water quality
indices and analysis of the surface water quality monitoring network in semiarid NorthCentral Chile.Environmental Monitoring and Assessment 184(9), 5571–5588.
Everard, M., Moggridge, H.L., 2012. Rediscovering the value of urban rivers. Urban Ecosystems
15, 293-314.
Falkenmark, F., Folke, C., 2003.Freshwater and welfare fragility: syndromes, vulnerabilities and
challenges—introduction. Philosophical Transactions of the Royal Society of London.
Series B, Biological Sciences 358, 1917–1920.
Fayer, R., 2004. Cryptosporidium: a water-borne zoonotic parasite. Veterinary Parasitology 126,
37-56.
Finn, M.P., Scheidt, D. J., Jaromack, G.M., 2002. User’s guide for the Agricultural Non-Point
Source (AGNPS) pollution model data generator. United States Geological Survey.
Report # 03-130.
289

Fitzhugh, T.W., Richter, B.D., 2004. Quenching urban thirst: growing cities and their impacts on
freshwater ecosystems. Bioscience 54, 741–754.
Garfi, M., Ferrer-Marti, L., Bonoli, A., Tondelli, S., 2011. Multi-criteria analysis for improving
strategic environmental assessment of water programmes, A case study in semiarid
region of Brazil. Journal of Environmental Management 92, 665-675.
Garrido, A., Dinar, A., 2008. Managing water resources in a time of global change. Routledge,
Oxon, p. 298.
Gartner, J.D., Renshaw, C.E., Dade, W.B., Magilligan, F.J., 2012. Time and depth scales of fine
sediment delivery into gravel stream beds: constraints from fallout radionuclides on fine
sediment residence time and delivery. Geomorphology 151-152, 39-49.
Gassman, P.W., Arnold, J.G., Srinivasan, R., Reyes, M., 2010. The worldwide use of the SWAT
model: Technological drivers, networking impacts and simulation trends. Watershed
Technology: Improving Water Quality and Environment. Universidad Earth, Costa Rica:
American Society of Agricultural Engineering.
Gassman, P.W., Reyes, M.R., Green, C.H., Arnold, J.G., 2007. The Soil and Water Assessment
Tool: Historical development, applications and future research directions. Transaction of
the ASABE 50, 1211–1250.
Ghebremichael, L.T., Veith, T.L., Watzin, M.C., 2010. Determination of critical source areas for
phosphorus loss: lake Champlain basin, Vermont. Transaction of ASABE 53, 1595-1604.
Gilliom, R.J., 2007. Pesticides in U.S. streams and ground water. Environmental Science &
Technology 41(10), 3407-3413.

Giorgi, F., Bi, X., Pal, J., 2004. Mean, interannual variability and trends in a regional climate
change experiment over Europe II: climate change scenarios(2071-2100). Climate
Dynamics 23, 839-858.
Giri, S., Mukhtar, S., Wittie, R., 2010. Vegetative covers for sediment control and phosphorus
control from dairy waste application fields. Transaction of the ASABE 53(3), 803- 811.
Giri, S., Nejadhashemi, A.P., Woznicki, S.A., 2012a. Evaluation of targeting methods for
implementation of best management practices in the Saginaw River Watershed. Journal
of Environmental Management 103, 24-40.
Giri, S., Nejadhashemi, A.P., Woznicki, S.A., Zhang, Z., 2012b. Analysis of best management
practice effectiveness and spatiotemporal variability based on different targeting
strategies. Hydrological Processes, 10. 1002/hyp.9577.
Gitau, M.W., Veith, T.L., Gburek, W.J., 2004. Farm-level optimization of BMP placement for
cost-eff ective pollution reduction. Transaction of the ASABE 47, 1923–1931.

290

Glozier, N.E., Struger, J., Cessna, A.J., Gledhill, M., Rondeau, M., Ernst, W.R., Sekela, M.A.,
Cagampan, S.J., Sverko, E., Murphy, C., Murray, J.L., Donald, D.B., 2012. Occurrence
of glyphosate and acidic herbicides in select urban rivers and streams in Canada, 2007.
Environmental Science and Pollution Ressearch19, 821–834.
Gormley, K.L., Teather, K.L., Guignion, D.L., 2005. Changes in salmonid communities
associated with pesticides runoff events. Ecotoxicology 14(7), 671-678.
Govers, G., Gimenez, R., Oost, K. V., 2007. Rill erosion: Exploring the relationship between
experiments, modeling and field observations. Earth Science Reviews 84, 87-102.
Grynkiewicz, M., Polkowska, Z., Gorecki, T., Namiesnik, J., 2003.Pesticides in precipitation
from an urban region in Poland (Gdansk-Sopot-Gdynia Tricity) between 1998 and 2000.
Water Air and Soil Pollution 149, 3-16.
Haan, C.T., Barfield, B.J., Hayes, J.C., 1994. Design hydrology and sedimentology for small
catchments. 2nd ed. Sandiego, California:Elsevier.
Hangsleben, M., Suh, D., 2006. Sediment Pollution. TSM 424 Semester Project. Available at:
http://www3.abe.iastate.edu/tsm424/TSM424TermProj2006/HangslebenSuhFinalPaper.p
df. Accessed on May 15, 2012.
Hanselman, T.A., Graetz, D.A., Wilkie, A.C., 2003. Manure-borne estrogens as potential
environmental contaminants: a review. Environmental Science and Technology 37,
5471–5468.
Hansen, N.C., Daniel, T.C., Sharpley, A.N., Lemunyon, J.L., 2002. The fate and transport of
phosphorus in agricultural systems. Journal of Soil and Water Conservation 57, 409–416.
Hassan, R., Scholes, R., Ash, N., 2005. Ecosystems and human well-being; current state and
trends. 1st edition. Washington, D.C. : Coghill Composition, Inc.
Hathaway, J.M., Hunt, W.F., 2012. Indicator Bacteria Performance of Storm Water Control
Measures in Wilmington, North Carolina. Journal of Irrigation and Drainage Engineering
138 (2), 185-17.
Haynes, H., Vignaga, E., Holmes, W.M., 2009. Using magnetic resonance imaging for
experimental analysis of fine sediment infiltration into gravel beds. Sedimentology 56,
1961-1975.
Helfrich, L.A., Weigmann, D.L., Hipkins, P., Stinson, E.R., 2009. Pesticides and aquatic
animals: a guide to reducing impacts on aquatic systems. Available at:
http://pubs.ext.vt.edu/420/420-013/420-013_pdf.pdf. Accessed November 5, 2012.
Hoffman, R.S., Capel, P.D., Larson, S.J., 2000. Comparison of pesticides in eight U.S. urban
streams. Environmental Toxicology & Chemistry 19, 2249–2258.

291

Hogan, J. N., Daniels, M. E., Watson, F. G., Conrad, P. A., Oates, S. C., Miller, M. A., Hardin,
D., Byrne, B. A., Dominik, C., Melli, A., Jessup, D. A., Miller, W. A., 2012.
Longitudinal poisson regression to evaluate the epidemiology of cryptosporidium,
giardia, and fecal indicator bacteria in coastal California wetlands. Application of
Environmental Microbiology 78 (10), 3606.
Hutchison, M.L., Walters, L.D., Avery, S.M., Synge, B.A., Moore, A., 2004.Levels of zoonotic
agents in British livestock manures. Letters in Applied Microbiology 39,207–214.
Ibekwe, A.M., Murinda, S.E., Graves, A.K., 2011. Microbiological evaluation of water quality
from urban watersheds for domestic water supply improvement. International Journal of
Environmental Research and Public Health 8, 4460-4476.
Ice, G., 2004. History of innovative best management practice development and its role.
Addressing of Environmental Engineering 130, 684-689.
IWR, 2012. High Impact Targeting, A tool for optimizing sedimentation reduction efforts in the
Great Lakes Basins. Available at: http://www.iwr.msu.edu/hit2/. Accessed on November
14, 2012.
Jeffrey, J.S., Kirk, J.H., Atwill, E.R., Cullor, J.S., 1998. Prevalence of selected microbial
pathogens in processed poultry waste used as dairy cattle feed. Poultry Science 77, 808–
819.
Jenkins, M.B., Endale, D.M., Schomberg, H. H., Sharpe, R.R., 2006. Fecal bacteria and sex
hormones in soil and runoff from cropped watersheds amended with poultry litter.
Science of the Total Environment 358, 164– 177.
Jenkins,, M. B., Endale, D.M., Schomberg, H. H., Hartel, P. G., Cabrera, M. L., 2009. 17bEstradiol and testosterone in drainage and runoff from poultry litter applications to tilled
and no-till crop land under irrigation. Journal of Environmental Management 90, 2659–
2664.
Jenks, G. F., 1967. The Data Model Concept in Statistical Mapping, International Yearbook of
Cartography 7: 186–190.
Jha, M.K., Schilling, K.E., Gassman, P.W.,Wolter, C.F., 2010. Targeting land use change for
nitrate nitrogen load reductions in an agricultural watershed. Journal of Soil Water
Conservation 65, 342-352.
Johnson, A.C., Williams, R.J., Matthiessen, P., 2006. The potential steroid hormone contribution
of farm animals to freshwaters, the United Kingdom as a case study. Science of Total
Environment 362,166–78.
Johri, N., Jacquillet, G., Unwin, R., 2010. Heavy metal poisoning: The effects of cadmium on the
kidney. BioMetals 23,783–792.

292

Jokinen, C. C., Edge, T. A., Koning, W., Laing, C. R., Lapen, D. R., Miller, J., Mutschall, S.,
Scott, A., Taboada, E. N., Thomas, J. E., Topp, E., Wilkes, G., Gannon, V. P. J., 2012.
Spatial and Temporal Drivers of Zoonotic Pathogen Contamination of an Agricultural
Watershed. Journal of Environmental Quality 41,242–252.
Jones, K.B., Neale, A.C., Nash, M.S., van Remortel, R.D., Wickham, J.D., Riitters, K.H.,
O’Neill, R.V., 2001. Predicting nutrient and sediment loadings to streams from landscape
metrics: A multiple watershed study from the United States Mid-Atlantic region.
Landscape Ecology 16, 301–312.
Jordan, P., Haygarth, P.M., Shortle, G., Harris, R.C., 2012. Catchment science and policy for
agriculture and water quality. Environmental Science &Policy 24, 1-3.
Juniori R.S., Silva, J.D., 2011. Sensibility analysis of the pearl model for pesticide leaching in
the state of Mato Grosso Do Sul, Brazil. Engineering Agricultural Jaboticabal 31(5), 965973.
Kaplowitza, M. D., Lupi, F., 2012. Stakeholder preferences for best management practices for
non-point source pollution and stormwater control. Landscape and Urban Planning 104,
364– 372.
Katrijn, M.A., Seuntjens, P., Vanrolleghema, P.A., 2007. Monitoring and modeling pesticide fate
in surface waters at the catchment scale. Ecological Modeling 2 0 9, 53–64.
Kazi, T.G., Arain, M.B., Jamali, M.K., Jalbani, N., Afridi, H.I., Sarfraz, R.A., Baig, J.A., Shah,
A.Q., 2009. Assessment of water quality of polluted lake using multivariate statistical
techniques: A case study. Ecotoxicology and Environmental Safety 72(2), 301-309.
Kellog, R.L., Lander, C.H., Moffitt, D.C., Gollhon, N., 2000. Manure nutrients relative to the
capacity of cropland and pastureland to assimilate nutrients: spatial and temporal trends
for the United States. USDA Natural Resources Conservation Service, Editor, Fort
Worth, TX. Available at: http://www.ncrcs.usda.gov/technical/nri/pubs/manntr.pdf.
Accessed on November 5, 2012.
Kjaer, J., Olsen, P., Bach, K., Barlebo, H.C., Ingerslev, F., Hansen, M., 2007. Leaching of
estrogenic hormones from manure-treated structured soils. Environmental Science and
Technology 41, 3911–3927.
Kolodziej, E.P., Sedlak, D.L., 2007. Rangeland grazing as a source of steroid hormones to
surface waters. Environmental Science & Technology 41, 3514–20.
Kolpin, D.W., Thurman, E.M., Lee, E.A., Meyer, M.T., Furlong, E.T., Glassmeyer, S.T., 2006.
Urban contributions of glyphosate and its degradate AMPA to streams in the United
States. Science of Total Environment 354, 191–197.
Kotti, M.E., Vlessidis, A.G., Thanasoulias, N.C., Evmiridis, N.P., 2005.Assessment of river
water quality in Northwestern Greece. Water Resources Management 19, 77–94.
293

Kovacsa, A., Hontib, M., Zessnera, M., Ederc, A., Clementd, A., Bloschl, G., 2012.
Identification of phosphorus emission hotspots in agricultural catchments. Science of The
Total Environment 433, 74-88.
Krause, P., Boyle, D. P., Base, F., 2005. Comparison of different efficiency criteria for
hydrological model assessment. Advances in Geoscience 5, 89–97.
Kreuger, J., 1996.Pesticides in stream water within an agricultural catchment in southern
Sweden, 1990. The Science of the Total Environment 216, 227-251.
Kreuger, J., Brink, N., 1998. Losses of pesticides from agriculture. In: Pesticides: Food and
Environmental Implications. International Atomic Energy Agency, Vienna, pp. 101–112.
Laia, Y.C., Yangb, C.P., Hsiehc, C.Y., Wua, C.Y., Kao, C.M., 2011. Evaluation of non-point
source pollution and river water quality using a multimedia two-model system. Journal of
Hydrology 409(3-4), 583-595.
Lebo, M.E., Paerl, H. W., Peierls, B., 2012. Evaluation of Progress in Achieving TMDL
Mandated Nitrogen Reductions in the Neuse River Basin, North Carolina. Environmental
Management 49, 253–266.
Lee, J.G., Selvakumar, A., Alvi, K., Riverson, J., Zhen, J.X., Shoemaker, L., Lai, F., 2012. A
watershed-scale design optimization model for stormwater best management practices. .
Environmental Modelling & Software 37, 6-18.
Lee, M.D., Newell, D.G., 2006. Campylobacter in poultry: Filling an ecological niche. Avian
Diseases 50,1–9.
Lerch, R.N., Sadler, E.J., Sudduth, K.A., Baffaut, C., Kitchen, N.R., 2011. Herbicide transport in
Goodwater Creek experimental watershed: I. longterm research on atrazine. Journal of
the American Water Resources Association 47(2), 209-223.
Leung, A.O.W., Duzgoren-Aydin, N.S., Cheung, K.C., Wong, M.H., 2008. Heavy metals
concentrations of surface dust from e-waste recycling and its human health implications
in Southeast China. Environmental Science and Technology 42, 2674–2680.
Linkov, I., Steevens, J., 2013. A detailed analysis of the theoretical foundation of different
MCDA. Chapter 35 Appendix A Multi-criteria decision analysis. Available at:
http://www.epa.gov/cyano_habs_symposium/monograph/Ch35_AppA.pdf. Accessed on
May 1, 2013.
Lintelmann, J., Katayama, A., Kurihara, N., Shore, L., Wenzel, A., 2003. Endocrine disruptors in
the environment. Pure and Applied Chemistry 75, 631–681.
Lorenz, E., 2009. Potential health effects of pesticides. Agricultural Communication and
Marketing 18,1-8.

294

Love, B. J., Einheuser, M.D., Nejadhashemi, A. P., 2011. Effects on aquatic and human health
due to large scale bioenergy crop expansion, Science of The Total Environment, 409(17),
3215-3229.
LRRB, 2010. Improved automatic sampling for suspended solids. Available at:
http://www.dot.state.mn.us/research/TS/2010/201038TS.pdf. Accessed on September 25,
2012.
Maas, R. P., Smolen, M. D., Dressing, S. A., 1985. Selecting critical areasfor nonpoint source
pollution control. Journal of Soil and Water Conservation 40(1), 68–71.
Maciej, D., 2000. Activities in nonpoint pollution control in rural areas of Poland. Ecological
Engineering 14, 429–434.
Malcolm, R. L., 1985, Geochemistry of stream fulvic and humic substances, in Aiken, G. R.,
McKnight, D. M., Wershaw, R. L., MacCarthy, P., Humic Substances in Soil, Sediment
andWater,Geochemistry, Isolation, and Characterization, Wiley-Interscience, New York,
USA, pp. 181–209.
Mallin, M.A., 2000. Impacts of industrial animal production on rivers and estuaries. American
Scientist 88, 26–37.
Mallin, M.A., Cahoon, L.B., 2003. Industrialized animal production — a major source of
nutrient and microbial pollution to aquatic ecosystems. Population and Environment 24,
369–379.
Malone, S.J., 2009. Agricultural Nonpoint Source Pollution Management: Water Quality Impacts
of Balm Road Treatment Marsh, Hillsborough County, Florida. Master of Science
Thesis. University of South Florida, Tampa, Florida.
MARC, 2012. Mid-America Regional Council. What is sediment pollution. Available at :
http://cfpub.epa.gov/npstbx/files/KSMO_Sediment.pdf. Accessed on May 23, 2012.
Maringanti, C., Chaubey, I., Popp, J., 2009. Development of a multiobjective optimization tool
for the selection and placement of best management practices for nonpoint source
pollution control. Water Resources Research 45, W06406.
Marinoni, O., 2009. Macro ext_ahp.dll: Available at:
http://arcscripts.esri.com/details.aspdbid=13764. Acessed on January 15, 2013.
Matthiessen, P., Arnold, D., Johnson, A.C., Pepper, T.J., Pottinger, T.G., Pulman, K.G.T., 2006.
Contamination of headwater streams in the United Kingdom by oestrogenic hormones
from livestock farms. Science of the Total Environment 367, 616–630.
Meybeck, M., 2004. The global change of continental aquatic systems: dominant impacts of
human activities. Water Science and Technology 49, 73-83.

295

Michel, P., Wilson, J.B., Martin, S.W., Clarke, R.C., McEwen, S.A., Gyles C.L., 1999. Temporal
and geographical distributions of reported cases of Escherichia coli O157:H7 infection in
Ontario. Epidemiology & Infection 122,193–200.
Millennium Assessment. 2005. Ecosystems and Human Well- Being: Synthesis. Island Press:
Washington, DC.
Milner, M.J., Kochian, L.V., 2008. Investigating heavy-metal hyperaccumulation using Th laspi
caerulescens as a model system. Annals of Botany102, 3–13.
Moore, P.A., Daniel, T.C., Sharpley, A.N., Wood, C.W.,1995. Poultry manure management:
environmentally sound options. Journal of Soil Water Conservation 50, 321–337.
Moriasi, D. N., Arnold, J. G., Van Liew, M. W., Binger, R. L., Harmel, R. D., Veith, T. L., 2007.
Model evaluation guidelines for systematic quantification of accuracy in watershed
simulations. Transaction of the ASABE 50, 885‐900.
Morrison, J., Gleick, P., 2004. Fresh water resources: Managing the risks facing the private
sector. Pacific Institute, Okland, California.
Mulholland, P. J., Webster, J. R., 2010. Nutrient dynamics in streams and the role of J-NABS,
Journal of the North American Benthological Society 29(1), 100-117.
Muthukrishnan, S., 2002. Development and testing of an integrated water pollution assessment
tool for analyzing impacts of urban sprawl. Ph.D. Purdue University.
Muttiah, R. S., Wurbs, R. A., 2002. Scale‐dependent soil and climate variability effects on
watershed water balance of the SWAT model. Journal of Hydrology 256, 264‐285.
Napier-Munn, T.J., Meyer, D.H., 1999. A modified paired t-test for the analysis of plant trials
with data auto-correlated in time. Minerals Engineering 12, 1093-1100.
Nash, J. E., Sutcliffe, J. V., 1970. River flow forecasting through conceptual models: Part 1. A
discussion of principles. Journal of Hydrology 10, 282-290.
Neitsch, S. L., Arnold, J. G., J. R. Kiniry., Williams, J. R., 2005. Soil and Water Assessment
Tool : Theoretical documentation. Temple, Tex.: USDA-ARS Grassland, Soil and Water
Research Laboratory.
Nejadhashemi, A. P., Mankin, K. R., 2007. Recommended practices for modeling BMPs
using SWAT model. Topeka, Kans.: Kansas Water Office.
Nejadhashemi, A.P., Woznicki, S.A., Mankin, K.R., 2011. Comparision of four models(STEPL,
PLOAD, L-LHIA, ans SWAT) in simulating sediment, nitrogen, and phosphorus loads
and pollutant source areas. Transections of the ASABE 54, 875-890.
Newcombe, C.P., MacDonald, D.D., 1991. Effects of suspended sediments on aquatic
ecosystems. North American Journal of Fisheries Management 11, 72-82.
296

Neitsch, S. L., Arnold, J. G., Kiniry, J. R., Williams, J. R., 2005. Soil and Water Assessment
Tool: Theoretical documentation. Temple, Tex.: USDA‐ARS Grassland, Soil and Water
Research Laboratory.
Nouri, J; Karbassi, A.R., Mirkia, S., 2008. Environmental management of coastal regions in the
Caspian sea. International Journal of Environmental Science and Technology 5, 43-52.
NRCS, 2008. Conservation practice standard, stripcropping. Available at: ftp://ftpfc.sc.egov.usda.gov/NHQ/practice-standards/standards/585.pdf. Assessed on May 20,
2013.
NRCS, 2011. Field office technical guide, Statewide conservation practice typical installation
cost information.
NWP 2011. Nutrient and pesticide management extension outreach for producers. Available at:
http://www.usawaterquality.org/themes/npm/extension/producers.html. Assessed on July
7, 2011.
NSW-DPI, 2012. Department of Primary Industries, New South Wales, Australia. Soil erosion
solutions. Available at:
http://www.dpi.nsw.gov.au/__data/assets/pdf_file/0003/255153/fact-sheet-1-types-oferosion.pdf. Accessed on May 25, 2012.
Olson, M.E., Thorlakson, C.L., Deselliers, L., Morck, D.W., McAllister.T.A.,1997. Giardia and
Cryptosporidium in Canadian farm animals. Veterinary Parasitology 68,375–381.
Ondrusek, M., E. Stengel, C. S. kinkade, R. L. Vogel, P. Keegstra, C. Hunter, C. Kim. 2012. The
development of a new optical total suspended matter algorithm for the Chesapeake Bay.
Remote Sensing of Environment 119, 243-254.
Opporto, B., Estaban, J.I., Aduriz, G., Juste, R.A., Hurtado, A., 2007. Prevalence and strain
diversity of thermophilic campylobacters in cattle, sheep and swine farms. Journal of
Applied Microbiology 99,1–8.
Osmond, D., Meals, D., Hoag, D., Arabi, M., Luloff, A., Jennings, G., McFarland, M., Spooner,
J., Sharpley, A., Line, D., 2012. Improving conservation practices programming to
protect water quality in agricultural watersheds: lessons learned from the national
institute of food and agriculture –conservation effects assessment project. Journal of Soil
and Water Conservation 67 (5), 122-127.
Ou, Y., Wang, X.Y., 2008. Identification of critical source areas for nonpoint source pollution in
Miyun reservoir watershed near Beijing China. Water Science and Technology 58, 2235–
2241.
Ouyang, W., Wang, X., Hao, F., Srinivasan, R., 2009. Temporal-spatial dynamics of vegetation
variation on non-point source nutrient pollution. Ecological Modeling 220, 2702–2713.

297

Palmer, M.A., Bernhardt, E.S., Chornesky, E.A., Collins, S.L., Dobson, A.P., Duke, C.S., Gold,
B.D., Jackson, R.B., Kingsland, S.E., Kranz, R.H., Mappin, M.J., Martinez, M.L.,
Micheli, F., Morse, J.L., Pace, M.L., Pascual, M., Palumbi, S.S., Reichman, O.J.,
Townsend, A.R., Turner, M.G., 2005. Ecological science and sustainability for the 21st
century. Frontiers in Ecology and the Environment 3, 4–11.
Panagopoulos, Y., Makropoulos, C., Mimikou, M., 2011. Reducing surface water pollution
through the assessment of the cost effectiveness of BMPs at different spatial scales.
Journal of Environmental Management 92, 2823-2835.
Parajuli, P.B; Mankin, K.R., Barnes, P.L., 2008. Applicability of targeting vegetative filter strips
to abate fecal bacteria and sediment yield using SWAT. Agricultural Water Management
95, 1189-1200.
Parajuli, P.B., Nelson, N.O., Frees, L.D., Mankin, K.R., 2009. Comparison of AnnAGNPS and
SWAT model simulation results in USDA-CEAP agricultural watersheds in south-central
Kansas. Hydrological Processes 23, 748-763.
Paul, S., 2003. Bacteria total maximum daily load(TMDL): development and evaluation of a new
classification scheme for impaired waterbodies of Texas. Doctoral dissertation. Texas A
& M University, College Station, Texas.
Paukert, C. P., Pitts, K.L., Whittier, J.B., Olden, J. D., 2010. Development and assessment of a
landscape-scale ecological threat index for the Lower Colorado River Basin. Ecological
Indicators, 11(2), 304-310.
Pejman, A.H; Nabi Bidhendi, G.R., Karbassi, A.R., Mehrdadi, N., Esmaeili Bidhendi, M.,2009.
Evaluation of spatial and seasonal variations in surface water quality using multivariate
statistical techniques. International Journal of Environmental Science and Technology
6(3), 467-476.
Persyn, R. A., Glanville, T. D., Richard, T. L., Laflen, J. M., Dixon, P. M., 2005. Environmental
effects of applying composted organics to new highway embankments: Part III.
Transection of the ASABE 48(5), 1765-1772.
Pimentel, D., Houser, J., Preiss, E., White, O., Fang, H., Mesnick, L., Barsky, T., Tariche, S.,
Schreck, J., Alpert, S., 1997. Water resources: agriculture, the environment, and society.
American Institute of Biological Sciences 47, 97-106.
Polyakov, V., Fares, A., Kubo, D., Jacobi, J., Smith, C., 2007. Evaluation of a non-point source
pollution model, AnnAGNPS, in a tropical watershed. Environmental Modeling and
Software 22, 1617-1627.
Postel, S., 2000. Entering an era of water scarcity: The challenges ahead. Ecological
Applications 10(4), 941-948.
Postel, S., Richter, B., 2003. Rivers for Life : Managing Water for People and Nature. Island
Press, Washington D.C., U.S.A.
298

Qiu, Z., 2009. Assessing critical source areas in watersheds for conservation buffer planning and
riparian restoration. Environmental Management 44, 968-980.
Rabalais, N. N., Diaz, R. J., Levin, L. A., Turner, R. E., Gilbert, D., Zhang, J., 2010. Dynamics
and distribution of natural and human-caused hypoxia. Biogeosciences 7(2), 585–619.
Raisanen, J., Hansson, U., Ullerstig, A., Doscher, R., Graham, L.P., Jones, C., Meier, H.E.M.,
Samuelsson, P., Willen, U., 2004. European climate in the late twenty-first century:
regional simulations with two driving global models and two forcing. Climate Dynamics
22 (1), 13-31.
RAMP. 1996. Water quality parameters. River Assessment Monitoring Project, Kentucky Water
Watch. Available at: http://www.state.ky.us/nrepc/water/ramp/rmtss.htm. Accessed on
June 6, 2012.
Rao, N.S., Easton, Z.M., Schneiderman, E.M., Zion, M.S., Lee, D.R., 2009. Modeling
watershed-scale effectiveness of agricultural best management practices to reduce
phosphorus loading. Journal of Environmental Engineering 90, 1385-1395.
Rasmussena, J.J., Baattrup-Pedersena,A., Wiberg-Larsena, P., McKnightb, U.S., Kronvanga B.,
2011. Buffer strip width and agricultural pesticide contamination in Danish lowland
streams: Implications for stream and riparian management. Ecological Engineering
37(12), 1990–1997.
Revenga, C., Campbell, I., Abell, R., Devilliers, P., Bryer, M., 2005. Prospects for monitoring
freshwater ecosystems towards the 2010 targets. Philosophical Transactions of the Royal
Society B 360, 397–413.
Riebschleager, K.J., 2008. Development and application of the spatially explicit load enrichment
calculation tool (SELECT) to determine potential E. Coli loads in watersheds. Master
Thesis. Texas A & M University, College Station, Texas.
Ritter, W.F., Shirmohammadi, Adel., 2001. Agricultural nonpoint source pollution: Watershed
management and hydrology. Boca Raton, Florida: CRC Press.
Rouse, J.D., Bishop, C.A., Struger, J., 1999. Nitrogen pollution: an assessment of its threat to
amphibian survival. Environmental Health Perspectives 107 (10), 799–803.
Saaty, T.L., 1980. The analytic hierarchy process, Mcgraw-Hill, New York.
Sala, O.E., Chapin, F.S., Armesto, J. J., Berlow, R., Bloomfield, J., Dirzo, J., Huber-Sanwald, R.,
Huenneke, E., Jackson, L.F., Kinzig, R.B., Leemans, A., Lodge, R., Mooney, D.,
Oesterheld, H.A., Poff, M., Sykes, N.L., Walker, M.T., Walker, B.H., M& Wall, D.H.,
2000. Global biodiversity scenarios for the year 2100. Science 287, 1770–1774.
Santhi, C., Arnold, J.G., Williams, J.R., Dugas, W.A., Srinivasan,R., Hauck, L.M., 2001b.
Validation of the SWAT Model on a Large River Basin With Point and Nonpoint
Sources. Journal of the American Water Resources Association 37, 1169-1187.
299

Santhi, C., Srinivasan, R., Arnold, J.G., Williams, J.R., 2006. A modeling approach to evaluate
the impacts of water quality management plans implemented in a watershed in Texas.
Environmental Modelling & Software 21, 1141-1157.
Scalley, H., Scatena, F. N., Moya, S., Lugo, A. E., 2012. Long-term dynamics of organic matter
and elements exported as coarse particulates from two Caribbean montane watersheds.
Journal of Tropical Ecology 28, 127-139.
Schilling, K.E., Wolter, C.F., 2009. Modeling nitrate-nitrogen load reduction strategies for the
Des Moines river, Iowa using SWAT. Environmental Management 44, 671-682.
Sharpley, A.N., Daniel, T.C., Edwards, D.R., 1993. Phosphorus movement in the landscape.
Journal of Production Agriculture 6, 492–500.
Sharpley, A.N., Daniel, T., Sim, T., Lemunyon, J., Stevens, R., Parry, R., 2003.Agricultural
phosphorus and eutrophication , second edition. USDA-ARS. Publication no. ARS 149.
Shen, J., Cui, Z., Miao, Y., Mi, G., Zhang, H., Fan, M., Zhang, C., Jian, R., Zhang, W., Li, H.,
Chen, X., Li, X., Zhang, F., 2013. Transforming agriculture in China:From solely high
yield to both high yield and high resource use efficiency. Global Food Security 2, 1-8.
Shi, Q., Deng, X., Wu, F., Zhan, J., Xu, L., 2012. Best management practices for agricultural
non-point source pollution control using PLOAD in Wuliangsuhai watershed. Journal of
Food, Agriculture & Environment 10 (2), 1 3 8 9 - 1393.
Shore, L.S., Correll, D.L., Chakraborty, P.K., 1995. Fertilization of fields with chicken manure
as a source of estrogens in small streams. In: Steele K, editor. Animal Waste and the
Land-Water Interface. Boca Raton, FL: Lewis Publishers. p. 155–62.
Shortle, J. S., Horan, R. D., Blandford, D., 2012. Reforming Agricultural Nonpoint Pollution
Policy in an Increasingly Budget-Constrained Environment. Environmental Science and
Technology 46, 1316-1325.
Sin, S.N., Chua, H., Lo,W., Ng, L.M., 2001. Assessment of heavy metal cations in sediments of
Shing Mun River , HongKong. Environment International 26, 297-301.
Sommerlot, A.R., 2012. Developing new modeling techniques to evaluate the environmental and
economic impacts of individual management practices at the field and watershed scale.
Master Thesis. Michigan State University, East Lansing, Michigan.
Soupir, M. L., Mostaghimi, S., Yagow, E. R., 2006. Nutrient Transport from Livestock Manure
Applied to Pastureland Using Phosphorus-Based Management Strategies. Journal of
Environmental Quality 35, 1269–1278.
Srebotnjak, T., Carr, G., Sherbinin, A.d., Rickwood, C., 2012. A global water quality index and
hot deck imputation of missing data. Ecological Indicators 17, 108-119.
300

Srinivasan, M.S., Gerard-Marchant, P., Veith, T.L., Gburek, W.J., Steenhuis, T.S., 2005.
Watershed scale modeling of critical source areas of runoff generation and phosphorus
transport. Journal of the American Water Resources Association 41, 361–377.
Staddon, C., 2010. Managing Europe’s Water Resources: Twentyfirst Century Challenges.
Ashgate Publishing Ltd, Farnham, p. 290.
Teixeiraa, M.R., Nunesb, L.M., 2011. The impact of natural organic matter seasonal variations in
drinking water Quality. Desalination and Water Treatment 36,344–353.
Tetra Tech, 2006. Spreadsheet Tools for the Estimation of Pollutant Load (STEPL), User’s
guide, Version 4.0. Available on line at: http://it.tetratechffx.com/stepl/STEPLmain_files/STEPLGuide400.pdf. Accessed on November 14, 2012.
Thomas, K.M., Charron, D.F., Waltner-Toews, D., Schuster, C., Maarouf, A.R., Holt, J.D., 2006.
A role of high impact weather events in waterborne disease outbreaks in Canada. 1975–
2001. International Journal of Environmental Health Research 16,167–180.
TNRCC. 1999. Texas nonpoint-source pollution assessment report and management program.
Publ. No. SFR-68/99. Austin, Tex.: Texas Natural Resource Conservation Commission.
Trenberth, K.E., Dai, A., Rasmussen, R.M., Parsons, D.B., 2003. The changing character of
precipitation. Bulletin of the American Meteorological Society 84, 1205-1217.
Tripathi, M.P., R.K.Panda., N.S. Raghuwanshi. 2003. Identification and prioritisation of critical
subwatersheds for soil conservation management using the SWAT model.Biosystems
Engineering 85(3), 365-379.
Tsakovski, S., Kudlak, B., Simeonov, V., Wolska, L., Gracia, G., Namiesnik, J., 2012.
Relationship between heavy metal distribution in sediment samples and their ecotoxicity
by the use of the Hasse diagram technique. Analytica Chemica Acta 719, 16-23.
Tuppad, P., Srinivasan, R., 2008. Bosque River environmental infrastructure improvement plan:
Phase II BMP modeling report. Publ. No. TR-313. College Station, Tex.: Texas A&M
Univ., Texas AgriLife Research.
Tuppad, P., Mankin, K.R., Mcvay, K.A., 2010. Strategic targeting of cropland management
using watershed modeling. Agricultural Engineering International: CIGR Journal 12, 3-4.
Tuppad, P., N. Kannan., R. Srinivasan., C.G. Rossi., J.G. Arnold. 2010. Simulation of
agricultural management alternatives for watershed protection. Water Resource
Management 24, 3115-3144.
Ullrich, A., M. Volk., 2009. Application of the Soil and Water Assessment Tool (SWAT) to
predict the impact of alternative management practices on water quality and quantity.
Agricultural Water Management 96, 1207-1217.

301

UNEP, 2005. Water: facts and trends. Available at:
http://www.unwater.org/downloads/Water_facts_and_trends.pdf. Accessed on November
25, 2012.
UNEP,2005. Challenges of water
scarcity.http://www.siwi.org/documents/Resources/Reports/Challenges_water_scarcity_b
usiness_case_study_2005.pdf.
UNEP GEM Water, 2006. Water quality for ecosystem and human health. Available at:
http://www.unep.org/gemswater/Portals/24154/publications/pdfs/water_quality_human_h
ealth.pdf. Accessed on September, 2012.
UNEP GEM Water, 2008. Water quality for ecosystem and human health. Available at:
http://www.unep.org/gemswater/Portals/24154/publications/pdfs/water_quality_human_h
ealth.pdf. Accessed on September, 2012.
UNEP, 1998. Best management practices for agricultural nonpoint sources of pollution. CEP
technical report no. 41. Available at: http://www.cep.unep.org/publications-andresources/technical-reports/tr41en.pdf. Accessed September 25, 2012.
UN-Water, 2013. Water use statistics. Available at: http://www.unwater.org/statistics_use.html.
Accessed on August 8, 2013.
USDA, 2012. Agricultural resources and environmental indicators. Available at:
http://www.ers.usda.gov/media/873632/waterquality.pdf. Assessed on October 15, 2012.
USDA- NRCS, 2010. Conservation practice. Available at:
http://www.nrcs.usda.gov/wps/portal/nrcs/detailfull/?ss=16&navtype=SUBNAVIGATIO
N&cid=nrcs143_026849&navid=120160340000000&pnavid=120160000000000&positi
on=Not%20Yet%20Determined.Html&ttype=detailfull&pname=Conservation%20Practi
ces%20|%20NRCS. Accessed on October 15, 2012.
USDA, 1992. Agricultural Waste Management Handbook. Chapter 3: Agricultural wastes and
water, air, and animal resources. USDA Natural Resources Conservation Service.
USDA, 1992. Chapter 3: Agricultural wastes and water, air, and animal resources. USDA
Natural Resources Conservation Service, editor. Agricultural Waste Management
Handbook.
USDA, 1997. Water quality and agriculture: Status, conditions, and trends. Working Paper #16.
United States Department of Agriculture, Natural Resources Conservation Service.
USDA, 2012. Water quality impacts of agriculture. Available at:
http://www.ers.usda.gov/media/873632/waterquality.pdf. Accessed on September 25,
2012.
USDA,2005. About Rangeland. Avaialble at:
http://www. fs.fed.us/rangelands/whoweare/index.shtml. Accessed on September 25, 2012.
302

USDA-ERS, 1997. Agricultural Resources and Environmental Indicators, Agricultural
Handbook No. 712.
USDA 2010. No-till farming is a growing practice . Economic Information Bulletin No. 70.
Available at: http://www.ers.usda.gov/Publications/EIB70/EIB70.pdf. Assessed on July
8, 2011.
USDA-NRCS, 2005. National conservation practice standards -NHCP. Washington, D.C.
Available at: www.nrcs.usda.gov/Technical/Standards/nhcp.html . Accessed 25 July
2011.
USDA-NRCS, 2011. Strip Cropping. Available at:
http://www.sera17.ext.vt.edu/Documents/BMP_strip_cropping.pdf. Accessed 25 July
2011.
USDA 1996. AREI/ Production management. Avaiable at:
http://www.ers.usda.gov/publications/arei/ah712/AH7124-2.PDF. Assessed on July 9,
2011.
USDA-NRCS. 2005. National conservation practice standards ‐NHCP. Washington, D.C.:USDA
National
Resources
Conservation
Service.
Available
at:
www.nrcs.usda.gov/Technical/Standards/nhcp.html.
USDA-NRCS, 2011. Phosphorus best management practices : protecting water quality.Little
Rock,
Arkansas.
Available
at:
http://www.sera17.ext.vt.edu/Documents/BMP_strip_cropping.pdf. Assessed on July 8,
2011.
USEPA .2005. Non point source pollution: The nation’s largest water quality problem.
Publication no. EPA841-F-96-004A. Available at:
http://water.epa.gov/polwaste/nps/outreach/point1.cfm. Accessed on September 25, 2012.
USEPA 2003. National management measures for the control of non-point pollution from
agriculture. EPA-841-B-03–004. Washington, DC: US Environmental Protection
Agency, Office of Water; p. 2–8.
USEPA, 2003. Agricultural nonpoint source pollution. Available at:
http://water.epa.gov/polwaste/nps/agriculture/agmm_index.cfm. Accessed on September,
25, 2012.
USEPA, 2005. Non point source pollution: The nation’s largest water quality problem.
Publication no. EPA841-F-96-004A. Available at:
http://water.epa.gov/polwaste/nps/outreach/point1.cfm. Accessed on September 25, 2012.
USEPA, 2009. Saginaw River and Bay. Available at: http://www.epa.gov/glnpo/aoc/saginawriver/index.html. Accessed on September 25, 2012.
303

USEPA, 2005. Total maximum daily loads: national section 303(d) List Fact Sheet: top 100
impairments. USEPA office of water, Washington, DC.
USEPA, 1992. The quality of our nation’s water. Available at:
http://www.epa.gov/owow/305b/water.pdf. Accessed on September 25, 2012.
USEPA, 2002. National Water Quality Inventory - 2000 Report to Congress. EPA 841-F-02-003.
Office of Water (4503-T). U.S. Environmental Protection Agency, Washington, D.C.
USEPA, 2002. The Twenty Needs Report: How research can improve the TMDL program. EPA
841-B-02-002. USEPA Office of Water, Washington, DC.
USEPA, 2003. Agricultural nonpoint source pollution. Available at:
http://water.epa.gov/polwaste/nps/agriculture/agmm_index.cfm. Accessed on September,
25, 2012.
USEPA, 2007. Total maximum daily loads with storm water sources: a summary of 17 TNDLs.
EPA 841-R-07-002. Available at:
http://water.epa.gov/lawsregs/lawsguidance/cwa/tmdl/upload/17_TMDLs_Stormwater_S
ources.pdf. Accessed on September 25, 2012.
USEPA, 2008. Total Maximum Daily Loads: National Section 303(d) List Fact Sheet:Top 100
Impairments. U.S. Environmental Protection Agency: Washington, D.C.
USEPA, 2012. National management measures for the control of nonpoint pollution from
agriculture. EPA-841-B-03-004. Washington , DC.
USEPA, 2012. National Management Measures to Control Nonpoint Pollution from Agriculture.
Available at:
http://water.epa.gov/polwaste/nps/agriculture/upload/2003_09_24_NPS_agmm_chap1.pd
f. Accessed on September 25, 2012.
USEPA,1993. Guidance Specifying Management Measures for Sources of Nonpoint Pollution in
Coastal Waters. EPA-840-B-92-002. U.S.Environmental Protection Agency, Washington,
DC.
USEPA. 1999. Protocol for developing nutrient TMDLs. EPA-841-B-99-007. Washington DC:
United States Environmental Protection Agency, Office of Water.
USEPA. 2001c. Protocol for developing pathogen TMDLs (First Edition). EPA/841/R-00/002.
Washington D.C.: Office of Water, US Environmental Protection Agency.
USEPA. 2009. National water quality inventory: Report to congress, 2004 reporting cycle.
USEPA Doc. 841-R-08-001. U.S. EPA, Washington, DC.
USEPA, 2001. PLOAD version 3.0. An arcview GIS tool to calculate nonpoint source of
pollution in watershed and stormwater projects. User’s manual. Available on line at:

304

http://water.epa.gov/scitech/datait/models/basins/upload/2002_05_10_BASINS_b3docs_
PLOAD_v3.pdf. Accessed on November 14, 2012.
USEPA, 2012. Point source pollution. Available at: http://www.riverproject.org/point-sourcepollution.html. Accessed on September 25, 2012.
USEPA, 2012. What is nonpoint source pollution? Available at:
http://www.epa.gov/owow/NPS/forestrymgmt/pdf/ch1.pdf. Accessed on September 25,
2012.
USEPA , 2009. National water quality inventory: Report to congress, 2004 reporting cycle.
USEPA Doc. 841-R-08-001. U.S. EPA, Washington, DC. Available at:
http://www.epa.gov/owow/305b/2004report/2004_305Breport.pdf. Accessed 20 July
2011.
USGS, 2006. Nutrients in streams and rivers across the nation, 1992–2001. Available at:
http://pubs.usgs.gov/sir/2006/5107/pdf/SIR06-5107_508.pdf. Accessed on September 25,
2012.
USGS, 2012. Where is earth’s water located. Available at:
http://ga.water.usgs.gov/edu/earthwherewater.html. Accessed on May15, 2012.
USGS, 2012. Nitrogen and water. Problems with excess levels of nitrogen in the environment.
Available at: http://ga.water.usgs.gov/edu/nitrogen.html. Accessed on August 8, 2013.
Van Griensven, A., Meixner, T., Grunwald, S., Bishop,T., Diluzio, M., Srinivasan,R., 2006. A
global sensitivity analysis tool for the parameters of multi‐variable catchment models.
Journal of Hydrology 324, 10-23.
Valcour, J.E., Michel, P., McEwen, S.A., Wilson, J.B., 2002. Associations between indicators of
livestock farming intensity and incidence of human Shiga toxin-producing Escherichia
coli infection. Emerging Infectious Diseases 8, 252–257.
Varol, M., Sen, B., 2012. Assessment of nutrient and heavy metal contamination in surface water
and sediments of the upper Tigris River, Turkey. CATENA 92, 1-10.
Verma, S., Markus, M., Cooke, R.A., 2012. Development of error correction techniques for
nitrate-N load estimation methods. Journal of Hydrology 433, 12-25.
Vitousek, P. M., Naylor, R., Crews, T., David, M. B., Drinkwater, L. E., Holland, E., Johnes, P.
J., Katzenberger, J., Martinelli, L. A., Matson, P. A., Nziguheba, G., Ojima, D., Palm, C.
A., Robertson, G. P., Sanchez, P. A., Townsend, A. R., Zhang,F. S., 2009. Nutrient
imbalance in agricultural development. Science 324, 23-33.
Wali, U.G., Nhapi, I., Ngombwa, A., Banadda, N., Nsengimana, H., Kimwaga, R. J., Nansubuga,
I., 2011. Modelling of nonpoint source pollution in Akagera Transboundary River in
Rwanda. The Open Environmental Engineering Journal 4, 124-132.
305

Wang, L; Robertson, D.M., Garrison,P.J., 2007. Linkages between nutrients and assembles of
macroinvertebrates and fish in wadeable streams: Implication to nutrient criteria
development 39, 194-212.
Wanga, X., Wanga, Q., Wua, C., Liang, T., Zhengc, D., Weid, X., 2012. A method coupled with
remote sensing data to evaluate non-point source pollution in the Xin'anjiang catchment
of China. Science of the Total Environment 430, 132-143.
Weston, D.P., Holmes, R.W., Lydy, M.J., 2009. Residential runoff as a source of pyrethroid
pesticides to urban creeks. Environmental Pollution 157, 287–294.
White, E.M., Kieber, D.J., Sherrard, J., 2010. Carbon dioxide and carbon monoxide
photoproduction quantum yields in the Delaware Estuary. Marine Chemistry 118, 11–21.
White, M.J., Storm, D.E., Busteed, P.R., Stoodley, S.H., Phillips, S.J., 2009. Evaluating nonpoint
source critical source area contributions at the watershed scale. Journal of Environmental
Quality 38, 1654-1663.
Wilkes, G., Edge, T.A., Gannon, V.P.J., Jokinen, C., Lyautey, E., Neumann, N.F., Ruecker, N.,
Scott, A., Sunohara, M., Topp, E., Lapen, D.R.,2011. Associations among pathogenic
bacteria, parasites, and environmental and land use factors in multiple mixed-use
watersheds. Water Research 45(18), 5807-5825.
Wilkes, G., Edge, T., Gannon, V., Jokinen, C., Lyautey, E., Medeiros, D., Neumann,N., Ruecker,
N., Topp, E., Lapen, D.R., 2009. Seasonal relationships among indicator bacteria,
pathogenic bacteria, Cryptosporidium oocysts, Giardia cysts, and hydrological indices
for surface waters within an agricultural landscape. Water Resource 43,2209–2223.
Williams, J.R., Izaurralde, R.C., 2005. The APEX model. In: Joint global change research
institute, B.A., suite 201, College park, MD 20740-2496. 720 East Blackland Rd., temple,
TX 76502-9622: Texas agricultural experimentation station.
Williams, J.W., Izaurralde, R.C., Steglich, E.M., 2008. Agricultural Policy/Environmental
Extender Model, Theoretical documentation. BREC report # 2008-17.
Winchell, M., Srinivasan, R., Di Luzio, M., Arnold, J. G., 2007. ArcSWAT interface for SWAT
2005: User’s guide. Grassland Soil and Water Research Lab, available at:
http://www.geology.wmich.edu/sultan/5350/Labs/ArcSWAT Documentation.pdf.
Woznicki, S.A; Nejadhashemi, A.P., Smith, C.M., 2011. Assessing best management practice
implementation strategies under climate change scenarios. Transaction of the ASABE 54,
171-190.
Wu, L., Long, T., Liu, X., Mmereki, D., 2012. Simulation of soil loss processes based on rainfall
runoff and the time factor of governance in the Jialing River Watershed, China.
Environmental Monitoring and Assessement 184, 3731-3748.

306

Yang, W; Sheng, C., Voroney, P., 2005. Spatial targeting of conservation tillage to improve
water quality and carbon retention benefits. Canadian Journal of Agricultural Economics
53, 477-500.
Yang, L., Ma, K., Guo, Q., Zhao, J., Luo, Y., 2006. Zoning planning in non-point source
pollution control in in Hanyangdistrict. Huan Jing ke Xue 27(1), 31-6.
Yatsalo, B.I., Kiker, G.A., Kim, J., Bridges, T.S., Seager, T.P., Gardner, K., Satterstrom, F.K.,
Linkov, I., 2007. Application of multi-criteria decision analysis tool to two contaminated
sediment case studies. Intergrated Environmental Assessment and Management 3(2),
223-233.
Yonkos, L.T., Fisher, D.J., Van-veld, P.A., Kane, A.S., McGee, B.L., Staver, K.W., 2010.
Poultry-litter induced endocrine disruption in fathead minnow, sheepshead minnow, and
mummichog laboratory exposures. Environmental Toxicology and Chemistry 29, 2328–
2340.
Young, K. D., Thackston, E. L., 1999. Housing density and bacterial loading in urban streams.
Journal of Environmental Engineering 1177-1180.
Young, W.F., Whitehouse, P., Johnson, I., Sorokin, N., 2004. Proposed predicted-no-effect
concentrations (PNECs) for natural and synthetic steroid estrogens in surface waters.
Technical Report, P2-T04/1.
Young, K.D., Kibler, D.F., Benham, B.L., Loganathan, G.V., 2009. Application of the analytical
hierarchial process for improved selection of stormwater BMPs. Journal of Water
Resources Planning and Management 135(4), 264-275.
Yousefpour, R., 2013. Bioenergy production and soil conservation from Norway spruce (Picea
abies L. Karst) plantation in Denmark. Agroforest Systems 87, 287-294.
Zanardo, S., Basu, N. B., Botter, G., Rinaldo, A., Rao, P. S. C., 2012. Dominant controls on
pesticide transport from tile to catchment scale: Lessons from a minimalist model. Water
Resources Research 48, W04525.
Zhang, X., Srinivasan, R., Arnold, J., Izaurralde, R.C., Boasch, D., 2011. Simultaneous
calibration of surface flow and baseflow simulations: a revisit of the SWAT model
calibration framework. Hydrological Processes 25, 2313-2320.
Zhoua, T., Wub, J., Penga, S., 2012. Assessing the effects of landscape pattern on river water
quality at multiple scales: A case study of the Dongjiang River watershed, China.
Ecological Indiactors 23, 166-175.
Zhu, Q., Schmidt, J.P., Bryant, R.B., 2012. Hot moments and hot spots of nutrient losses from a
mixed land use watershed. Journal of Hydrology 415, 393-404.
T

307