35,1 ‘

"IE!" x 's I I
511%? 3 7"" ~

I
'8 iti
0 ‘<

   
 
   
  
  

 
 

   

:4 Lft‘EBlfifiifi
W .!;a§‘2‘-‘.1'.5~1
wig 7-: 'Fw

. .‘
‘5‘“: "Qt
glhjiiu“!

. 9g . .
V . "X
‘ .31! R’sjgt'lmz

    
     

A. .-
.l v"'

-u‘

9-:“i'.'§'$-
”.3, 1 . .—

    

27' Y'

n“ "
“1"".
c.“ .
‘ -l..‘:...; .h_
< ’ jo', ‘ I.

v
.—

-5-

w

.

   

€235

I" ‘
. 5"}.

m: :

  
 
 

.V.‘ ! f‘;.
‘2',“‘331-3
‘35:
I

 
 
  
 
    

  

    
   

  

     
  
  

      
               

 

  
  
 

 

         
 
  
 
 
  

.‘ 3;,
u‘g ' {L‘gfgw‘fi‘i

‘I
c r‘ . h ' " ’

4 ‘ t_ MMM'“ .

'. ;., itn‘ L.1,.". ‘ u H“.
V “V." ”W“ . I 3'“. “

.‘ .“ ; . ‘ I‘m'. ‘1 ,{iz‘z‘ .‘ft ,. taw'it‘ V9.33. $13565“. ‘
‘ "f R ‘2‘.\‘. I" 1' t ‘u '| El ll 5 \ '1' .‘A
‘ "My. ' ' ’- n‘HVs‘,

Mr. 1

     
 
   
  

‘H‘ | u. ‘| ' ‘1‘“.“1‘il '2'!“ I. ‘ .‘» “II.
‘5‘} EY"“',:"I‘¢“" ‘ I} 'ls) . I “‘1

 

 
 
  

  
 
  

  

  

19‘ -.>‘
‘ \"‘~1~| V, I'M“, _' ‘4 III’): ‘
‘ ‘. 1‘ l 13.31,. WV“ ,1: 1': W T1? "'Et"|!~t.:ri"l‘ "Wm
. ‘  .  " “-“'.“'Jf‘ .‘I'L‘i‘ . I.‘ ‘I’IH‘ ":1 'w'd ‘-. (h "n . ‘ «"x'
x . ‘l‘lv "‘ ' " " "'1" ‘ ‘ -| ‘L 'i H" ‘ .1“wa 3"“ Iql'lv‘v' "331.131le .i‘...‘,"H\1“f“I‘ "is ”'5‘! “1'“?
it 12‘5")", ‘ .| - ' ‘ . ‘ .vl ‘ ‘ ' "\1\"r\'|":."". "IRAN "\ .3: '1." ['M

‘2‘.)1 " ‘.‘ ",,:’.:‘u.1‘ a" . " ‘
.;. 2:1... mm'rJ‘A‘ m , g '

   

v . 0‘ . _ I "‘l,_”““‘>gnul uni-Ifldh
lull ‘

  

     
  
  

IHES'S

S|TY LIBRARIES

\\l\\\\\\\l\\\ol\1\\ l l} lllll Gilli

 

ll \\

1293

 

 

 

\

 

 

l

 

 

This is to certify that the

dissertation entitled

A WATERSHED BASED OPTIMIZATION APPROACH
FOR AGRICULTURAL NON-POINT SOURCE POLLUTION
MANAGEMENT

presented by

Yung-Tsung Kang

has been accepted towards fulfillment

of the requirements for 7.
A

Eh. D . __degree in RasmLcaJeveIOpment

V 4%
/

Major professor

Date May 13, 1993

MSU is an Affirmative Action/Equal Opportunity Institution 0-12771

' LIBRARY

Mlchlgan sum
Unlvonlty

 

 

 

PLACE IN RETURN BOX
to remove this checkout from your record.
TO AVOID FINES return on or before date due.

 

MTE DUE DATE DUE DATE DUE

ml E 05%

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1193 MM.“

 

 

 

A WATERSHED BASED OPTIMIZATION APPROACH
FOR AGRICULTURAL NON -POINT SOURCE POLLUTION MANAGEMENT

By

Yung-Tsung Kang

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Resource Development

1998

ABSTRACT

A WATERSHED BASED OPTIMIZATION APPROACH
FOR AGRICULTURAL NON -POINT SOURCE POLLUTION MANAGEMENT

By

Yung-Tsun g Kang

Agricultural non-point source pollution was identified as the major threat to the
Nation’s water quality goal in the past Clean Water Act and its Amendments. Current
non-point source pollution management strategy requires a three-step procedure: critical
area identification, Best Management Practices selection, and area-wide comprehensive
planning. However, several difficulties exist in applying this three-step procedure for
non-point source pollution control. Presently, non-point source pollution control utilizes
a field-based approach. This approach does not take into account the diffuse, dynamic,
and stochastic nature of non-point source pollution. Decision-makers will need a better
approach to more effectively address these difficulties and problems.

The objective of this research is to overcome the difficulties of the three-step
procedure of non-point source pollution control. This research used a two-tier, aquatic
and terrestrial, systematic perspective within a watershed based optimization framework
to examine these problems. The aquatic component defines non-point source pollution
problems in the streams or lakes as a set of goals in a Min-Max Optimization strategy.
The Min-Max strategy utilizes a watershed based water quality model, AGNPS, for

estimation of non-point source pollution on a watershed basis. The strategy finds the

maximum deviation of a non-point source pollution problem from the goal in the channel
segment. Then, the upstream areas of the identified channel segment are delineated, that
is, the sub-watershed of the terrestrial system. A source accounting technique is applied
to identify “hot spots”, i.e., high-risk fields that contribute most to the identified problem
in the sub-watershed. Finally, alternative Best Management Practices are evaluated in the
optimization process to optimally reduce (minimize) non-point source pollution problem.

This optimization procedure continues to find better and better solutions through
iteration processes. The final result is a set of non-commensurable solutions, called
Pareto optimal solutions. Each Pareto solution provides not only the achievement
accomplished relative to the goal but also a unique scenario on which alternative field
management practices were determined. Decision-makers can then evaluate these
solutions with social-economical considerations to fine-tune a comprehensive
management plan for non-point source pollution control. An expert focus group was
used to evaluate the research approach, methods, and findings.

The evaluation results are general due to the nature of the focus group process.
Participants of the focus group agreed that the research approach seems to be valid and
the findings are encouraging. This research effort demonstrated a well-defined process
that can potentially be applied to existing water quality programs, such as Environment
Quality Incentive Program, designated river uses, and Total Maximum Daily Load. The
established framework represents a scientifically based and environmentally sound
approach for decision-makers to assist with economically viable and socially acceptable

solutions for non-point source pollution management.

TABLE OF CONTENTS

LIST OF TABLES ......................................................................................................... vii
LIST OF FIGURES ...................................................................................................... viii
CHAPTER 1 INTRODUCTION .................................................................................... 1
1.1 Overview .................................................................................................................. 1
1.2 Problem Setting ........................................................................................................ 3
1.3 Research Questions .................................................................................................. 5
1.4 Research Approach ................................................................................................... 7
1.5 Dissertation Organization ......................................................................................... 8
CHAPTER 2 LITERATURE REVIEW ........................................................................ 10
2.1 Agricultural Non-Point Source Pollution Control .................................................. 10
2.2.1 NPS Policy Options ..................................................................................... 10
2.2.2 Current NPS Policy Setting ......................................................................... 14
2.2 Critical Area Identification and NPS Pollution Models ......................................... 15
2.2.1 Empirical (Lumped) Modeling Approach .................................................... 16
2.2.2 Physical-Process (Causal) Simulation Modeling Approach ........................ 19
2.2.3 Distributed Modeling Approach .................................................................. 22
2.2.4 Model Selection ........................................................................................... 24
2.3 Best Management Practices .................................................................................... 25
2.3.1 Paradigm Shift ............................................................................................. 27
2.3.2 A System Perspective .................................................................................. 28
2.3.3 An Ecological Perspective ........................................................................... 30
2.4 Area-wide Comprehensive Planning and System Integration ................................ 31
2.4.1 Approaches for BMPs Selection .................................................................. 31
2.4.2 618 Integration ............................................................................................. 34
2.4.3 Information Systems and What Ifs ............................................................... 37
2.4.4 Decision Support System (DDS) ................................................................. 38
2.4.5 Social Consideration .................................................................................... 41
2.5 Optimization ........................................................................................................... 42
2.5.1 Optimization Problems ................................................................................ 42
2.5.2 The Pareto Concept ...................................................................................... 44
2.5.3 Optimization Issues ...................................................................................... 45
2.6 Focus Group ........................................................................................................... 49
2.6.1 Overview ...................................................................................................... 49

iv

2.6.2 Processes ...................................................................................................... 50

2.6.3 Issues and Concerns ..................................................................................... 52
CHAPTER 3 RESEARCH METHODS ...................................................................... 54
3.1 Research Design ..................................................................................................... 54

3.1.1 Overview ...................................................................................................... 54

3.1.2 The AGNPS Model ...................................................................................... 57

3.1.3 Scenario Analysis and Optimization ............................................................ 59
3.2 Study Area .............................................................................................................. 62

3.2.1 Geographical Setting .................................................................................... 62

3.2.2 Watershed Baseline Conditions ................................................................... 74
3.3 Data Collection ....................................................................................................... 75

3.3.1 Base Data and Model Data .......................................................................... 75

3.3.2 Watershed .................................................................................................... 77

3.3.3 Topography .................................................................................................. 80

3.3.4 Channel ........................................................................................................ 83

3.3.5 Soils ............................................................................................................. 83

3.3.6 Land Use/Cover and Field Management ..................................................... 85
3.4 Running the AGNPS Model ................................................................................... 91
3.5 Scenario Analysis and Optimization ...................................................................... 92

3.5.1 The Framework ............................................................................................ 92

3.5.2 Objective Functions and AGNPS Model Results ........................................ 92

3.5.3 Decision Variables and the Min-Max Strategy ............................................ 97
CHAPTER 4 RESEARCH RESULTS AND DISCUSSION .................................... 102
4.1 Optimization Process ............................................................................................ 102

4.1.1 Processing Limitations ............................................................................... 102

4.1.2 Adjustment ................................................................................................. 104

4.1.3 Results Interpretation ................................................................................. 105
4.2 Watershed Results ................................................................................................ 107

4.2.1 Pareto Optimal Solutions ........................................................................... 107

4.2.2 The Iterations Processes ............................................................................. 116
4.3 Sub-watershed Results ......................................................................................... 118

4.3.1 The Two Sub-watersheds ........................................................................... 118

4.3.2 Pareto Optimal Solutions ........................................................................... 137

4.3.3 The Iteration Processes .............................................................................. 137
4.4 Discussion of the Relevance to the Research Questions ...................................... 142

4.4.1 Critical Area Identification ........................................................................ 143

4.4.2 Best Management Practices ....................................................................... 146

4.4.3 Area-wide Comprehensive Planning ......................................................... 149

4.4.4 Water Quality Programs and NPS Pollution Policy ................................... 152
CHAPTER 5 FOCUS GROUP EVALUATION ....................................................... 157
5.1 Overview .............................................................................................................. 157
5.2 Procedures ............................................................................................................ 158

5.3 Focus Group Session Report ................................................................................ 161

5.3.1 Model Input ................................................................................................ 161
5.3.2 The AGNPS Model .................................................................................... 162
5.3.3 Goals and Objective Functions .................................................................. 163
5.3.4 Best Management Practices (BMPs) Selection .......................................... 165
5.3.5 Research Findings ...................................................................................... 166
5.3.6 Social-Economic and Planning (Implementation) ..................................... 168
5.4 Evaluation Conclusions ........................................................................................ 169
5.4.1 Research Approach and Method ................................................................ 170
5.4.2 Research Questions and Results ................................................................ 171
CHAPTER 6 CONCLUSIONS .................................................................................. 173
Appendix A - Basic Crop Rotation Information ......................................................... 17 8
Appendix B - Basic Field and Land Use/Cover Data ................................................. 181

Appendix C - Land Use/Cover and Field Management Data for the AGNPS Model 193

Appendix D - Listings of Field Management Operations of Example Pareto Optimal

Solutions .............................................................................................. 196
Appendix E - Cover Letter and Attachments for Focus Group Evaluation ................ 205
Bibliography ................................................................................................................ 211

LIST OF TABLES

Table 3-1 Major Land Use/Cover Distribution of the Sycamore Creek Watershed... 63
Table 3-2 Major Land UselCover Distribution of the Study Area .............................. 64
Table 3-3 Slope Distribution of the Study Area. ......................................................... 71
Table 3-4 Summaries of AGNPS Model Variable Derivation .................................... 77
Table 3-5 The Field Slope Length Based on the Land Slope ...................................... 82
Table 4-1 Pareto Optimal Solutions for the Study Area (Watershed) ....................... 110
Table 4-2 Average Objective Function Values per Channel Grid for the Study

Area (Watershed) ..................................................................................... 1 11
Table 4-3 A Simple Count Statistics of Pareto Solutions Based on Their Values 115
Table 4-4 Summaries of Identified Channel Grids that Have NPS Pollution

Problem for the Watershed ....................................................................... 121
Table 4-5 Summaries of Identified Fields that Have NPS Pollution Problem for

The Watershed .......................................................................................... 122
Table 4—6 Pareto Solutions for Sub-watershed with Outlet at Grid No. 178 ............. 125
Table 4-7 Pareto Solutions for Sub-watershed with Outlet at Grid No. 231 ............. 126
Table 4-8 Average Objective Function Values per Channel Grid for

Sub-watershed 178 ................................................................................... 127
Table 49 Average Objective Function Values per Channel Grid for

Sub-watershed 231 .................................................................................... 128
Table 4-10 A Simple Count Statistics of Pareto Solutions Based on Their Values

for Sub-watershed 178 .............................................................................. 138
Table 4-11 A Simple Count Statistics of Pareto Solutions Based on Their Values

for Sub-watershed 231 .............................................................................. 139
Table 4-12 Summaries of Identified Channel Grids that Have NPS Pollution

Problem for the Sub-watershed 178 .......................................................... 140
Table 4—13 Summaries of Identified Fields that Have NPS Pollution Problem for

The Sub-watershed 178 ............................................................................ 140
Table 4-14 Summaries of Identified Channel Grids that Have NPS Pollution

Problem for the Sub-watershed 231 .......................................................... 141
Table 4-15 Summaries of Identified Fields that Have NPS Pollution Problem for

The Sub-watershed 231 ............................................................................ 141
Table 4-16 Management Practices Changes of the Sub-watershed 178 ...................... 142
Table 5-1 A List of Identified Experts for Focus Group Evaluation ......................... 159

vii

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

Figure 3-10

Figure 3-11
Figure 3-12

Figure 4—1
Figure 4—2

Figure 4.3

LIST OF FIGURES

The Study Area - A Sub-watershed of the Sycamore Creek Watershed... 66

The Water System of the Study Area ........................................................ 67
The Land Use/Cover of the Study Area ..................................................... 68
Digital Elevation Model (DEM) of the Study Area ................................... 69
Aspect, Watershed, and Stream Network of the Study Area ..................... 70
Slope of the Study Area ............................................................................. 72
Soils of the Study Area .............................................................................. 73
The AGNPS Model Grid Layout ............................................................... 81
Field Map by Land Use/Cover and Field Management Identification

Code .......................................................................................................... 87
The Schema of the Linkage between Land Use/Cover and Field
Management Database and Map ............................................................... 88
The Optimization Framework for the NPS Pollution Problem ................. 93
The Processing Flow of the Optimization Procedure ................................ 99
39 Pareto Solutions for 3 NPS Pollution Problems for the Watershed

(4 Years) .................................................................................................. 112
39 Pareto Solutions for 3 NPS Pollution Problems for the Watershed

(Year 0) .................................................................................................... 1 12

39 Pareto Solutions for 3 NPS Pollution Problems for the Watershed

 

(Year 1). ..... - -- - . - - -_ - -- - - -- - - 112
Figure 4-4 39 Pareto Solutions for 3 NPS' Pollution Problems for the Watershed

(Year 2) .................................................................................................... 1 12
Figure 4-5 39 Pareto Solutions for 3 NPS Pollution Problems for the Watershed

(Year 3) .................................................................................................... 112
Figure 4—6 39 Pareto Solutions for Sediment for the Watershed (Year 0 - 3) ........... 113
Figure 4-7 39 Pareto Solutions for Nitrogen for the Watershed (Year 0 - 3) ............ 113
Figure 4-8 39 Pareto Solutions for Phosphorus for the Watershed (Year 0 - 3) ....... 113
Figure 4—9 An Example of Pareto Solutions ............................................................. 114
Figure 4—10 The 14 Channel Grids Identified as NPS Pollution Problem for the

Watershed ................................................................................................ 1 19
Figure 4—11 The 19 Fields Identified as Fields that Have Been Changed for the

Watershed ................................................................................................ 120
Figure 4-12 Sub-watershed of the Study Area with Outlet at Grid 178 ...................... 123
Figure 4-13 Sub-watershed of the Study Area with Outlet at Grid 231 ...................... 124

Figure 4-14 33 Pareto Solutions for 3 NPS Pollution Problems for the

Sub-watershed 178 (4 Years) .................................................................. 129
Figure 4-15 33 Pareto Solutions for 3 NPS Pollution Problems for the

Sub-watershed 178 (Year 0) .................................................................... 129
Figure 4—16 33 Pareto Solutions for 3 NPS Pollution Problems for the

Subawatershed 178 (Year 0) .................................................................... 129
Figure 4-17 33 Pareto Solutions for 3 NPS Pollution Problems for the

Sub-watershed 178 (Year 1) .................................................................... 129
Figure 4-18 33 Pareto Solutions for 3 NPS Pollution Problems for the

Sub-watershed 178 (Year 2) ................................................................... 129
Figure 4-19 33 Pareto Solutions for Sediment for the Sub-watershed 178

(Year 0 - 3) .............................................................................................. 130
Figure 4-20 33 Pareto Solutions for Nitrogen for the Sub-watershed 178

(Year 0 - 3) .............................................................................................. 130
Figure 4—21 33 Pareto Solutions for Phosphorus for the Sub-watershed 178

(Year 0 - 3) .............................................................................................. 130
Figure 4-22 22 Pareto Solutions for 3 NPS Pollution Problems for the

Sub-watershed 231 (4 Years) .................................................................. 131
Figure 4—23 22 Pareto Solutions for 3 NPS Pollution Problems for the

Sub-watershed 231 (Year 0) .................................................................... 131
Figure 4-24 22 Pareto Solutions for 3 NPS Pollution Problems for the

Sub-watershed 231 (Year 1) .................................................................... 131
Figure 4—25 22 Pareto Solutions for 3 NPS Pollution Problems for the

Sub-watershed 231 (Year 2) .................................................................... 131
Figure 4—26 22 Pareto Solutions for 3 NPS Pollution Problems for the

Sub-watershed 231 (Year 3) .................................................................... 131
Figure 4-27 22 Pareto Solutions for Sediment for the Sub-watershed 231

(Year 0 - 3) .............................................................................................. 132
Figure 4-28 22 Pareto Solutions for Nitrogen for the Sub-watershed 231

(Year 0 - 3) .............................................................................................. 132
Figure 4—29 22 Pareto Solutions for Phosphorus for the Sub-watershed 231

(Year 0 - 3) .............................................................................................. 132
Figure 4-30 The 7 Channel Grids Identified as NPS Pollution Problem for

Sub—watershed 178 .................................................................................. 133
Figure 4-31 The 15 Fields Identified as Fields that Have Been Changed for

Sub-watershed 178 .................................................................................. 134
Figure 4-32 The 10 Channel Grids Identified as NPS Pollution Problem for

Sub—watershed 231 .................................................................................. 135
Figure 4—33 The 15 Fields Identified as Fields that Have Been Changed for

Sub-watershed 231 .................................................................................. 136
Figure 4—34 Examples of Pareto Optimal Solutions for Field Management

Operations ............................................................................................... 153

ix

CHAPTER] ,

INTRODUCTION

1.1 Overview

Water is one of the elementary substances for human’s survival. We used to have
water for drinking, for irrigation, for industrial usage, for recreation, and even for dilution of
pollutants. Our vast water resource has led us to treat water as an inexhaustible resource in
the history of resource development. However, due to the conflicting needs and increasing
demands of water and the degradation and contamination of water, it has become a scarce
good during the second half of this century.

The maintenance of water quality has become one of the major environmental issues
in the past two decades. Since the 197 2 Federal Water Pollution Control Act, water quality
control methods have been based on the type of pollution source. That is, the regulatory
approach (the National Pollution Discharge Elimination System - NPDES) for point sources
and Best Management Practices (BMPs) of area-wide comprehensive planning for non-
point sources (NPS). In the past 20 years, several Clean Water Act Amendments and the
Safe Drinking Water Act have resulted in increasingly stricter regulations on point source
pollution in the NPDES in order to reach the national water quality management goal "to

restore and maintain the chemical, physical, and biological integrity of the Nation’s waters".

2
Simultaneously, several reports (Goudy, 1986, and Liebesman and Laws, 1988) have shown

that non-point source pollution has become the major threat to the Nation’s water quality.

NPS pollution refers to pollutants that are not from a specific location. Among the
activities related to the NPS pollution, agricultural production activities have long been
identified as a primary contributor to the NPS pollution. Several reports have called
attention to the seriousness of the NPS pollution from the agricultural sector. According to
the 1977 National Water Quality Inventory Report (U .8. EPA, 197 8), it stated that:

“. .. agricultural NPS of pollution significantly affected water quality in 68 percent

of all drainage basins in the United States.”
Later in 1988, the same National Water Quality Inventory Report (U .8. EPA, 1990) still
documented that:

". . . agricultural runoff is by far the most extensive source of pollution, affecting 55

percent of impaired river miles . . . "

In mid 1990, US. General Accounting Office (GAO) published a special study
(U .8. GAO, 1990) on the water quality efforts of the US. Department of Agriculture
(USDA). One of the principal findings indicated that USDA’s water quality
responsibilities are not focused. It recommended that USDA should: (1) establish a
Department-wide focal point or coordinating body with full-time staff support and
responsibility and accountability for all the USDA’s water quality activities; and (2)
develop a comprehensive water quality policy that addresses all agricultural water quality
concerns. Since then, USDA has launched a five-year Water Quality Initiative. The
Initiative (USDA, 1995) established three types of programs: (1) research and

development, (2) education, technical, and financial assistance, and (3) data base and

3
evaluation. These three program areas have initiated hundreds of water quality related

projects, which involve numerous federal, state, and local agencies. The US. GAO (U .8.
GAO, 1995) reported that 618 watershed-based projects aimed at agricultural sources of
pollution were being planned or carried out through early 1995. These projects had
received an estimated $514 million in federal funds.

Despite all the efforts that have been started in agricultural NPS pollution
prevention, the US. GAO (U .8. GAO, 1995) stated that:

“Recent federal, state, and local studies on water quality have identified

agricultural as the United States’ greatest source of nonpoint pollution - that is,

pollution that cannot be traced to a specific point of origin. Agriculture

contributes more than half the pollutants entering the nation’s rivers and lakes.

The thread to water quality posed by nonpoint sources of pollution has prompted

renewed interest in watershed-based approaches to reducing such pollution.”
Recently, EPA also reported (U .8. EPA, 1996) that:

“Presently, states and tribes identify nonpoint source pollution from cropland and

livestock, urban runoff, and storm sewers as the greatest water quality threat to the

Nation’s surface waters.”

1.2 Problem Setting

NPS pollutants generated by agricultural activities usually are classified into three
categories: soil sediments (erosion), nutrients, and pesticides. These pollutants are diffuse,
stochastic, and dynamic in nature (Bailey and Swank, 1983). Therefore, current NPS

management strategy requires a three-step procedure: (1) identification of pollution areas

4
and pollutants, (2) detemrination of BMPs for identified critical areas, and (3) construction

of an area-wide comprehensive pollution control plan. Based upon this three-step
procedure, standards should be established and the NPS problem can be systematically
addressed.

Unfortunately, several sets of difficulties emerged when trying to apply the three-
step procedure for NPS pollution control. First, the interactions between the spatial and
temporal dimensions of the NPS pollution process makes it extremely difficult to correctly
identify the pollution areas and pollutants. NPS pollutants are widely spread across the
land. They can occur naturally. Human activities on land are the major origin of the NPS
pollutants. The landscape setting, such as soil properties, channel morphology, topographic
characteristics, rainfall pattern, and climate conditions, also affects the NPS pollution
process. In addition, NPS pollutants interact with soils and water while travelling through
the landscape. The NPS pollution process has become difficult to measure and highly
variable. Current knowledge still can not accurately simulate the complexity of the NPS
pollution process.

Secondly, the causal relationships between NPS pollution and BMPs are not fully
understood. Many alternative BMPs have been identified or created for different types of
NPS pollution problems. However, most of them are based on a field management
approach. We may have established the causal relationship between a single type of NPS
pollutants (e.g. soil erosion) and a set of BMPs (e.g. contouring practice and residual
management) at the qualitative or nominal level on a field basis. The quantitative scale
measurements are still not adequate and there is no agreement in the NPS communities.

When putting the spatial dimension into consideration, the field based causal relationships

5
have to be reexamined. When it comes to multiple types of NPS pollutants and multiple

sets of BMPs, scientists are still searching for a systematic approach for making the
analysis.

Finally, in the decision-making process, the construction of an area-wide
comprehensive NPS pollution control plan is impossible without resolving the problems in
the first two steps. Without clear identification of pollution areas and associated pollutants,
decision-makers can not plan for solutions. Without established causal relationship
between NPS pollution and alternative BMPs in an area-wide perspective, decision-makers
can not implement solutions. Other factors, such as available resources and the level of
pollution control desired, also create difficulties for decision-makers in comprehensive
planning.

With all of the above difficulties, the problem domain of the NPS pollution, like
other environmental issues, is either unstructured or, at most, semi-structured. Well-
structured problem domains are those that can be solved by application of logical or
algorithmic processes. The NPS pollution control is based on the three-step prowdure. The
context of the three-step procedure is still challenging because of unspecified analytical
methods, undetermined causal relationships, and undefined decision-making rules. When
one faces such an ill-structured problem, data are either not reliable or relative, information

becomes tentative and changeable, and knowledge can hardly be generalized.

1.3 Research Questions
This study will focus on the challenge of the three-step management procedure of

the ill-structured NPS pollution problem. The overall research question is that a watershed

6
based optimization approach can be crafted and can be used for the three-step NPS

management procedure, i.e., by using a watershed based Optimization approach in
place of the current field based approach, can this methodolm provide a better

solution to the difficulties of the three-step NPS management procedure? Several

issues around each step of the NPS pollution management procedure are investigated.

First is critical area identification. The traditional approach used by Natural
Resources Conservation Services (NRCS) is usually a field by field approach. For
example, a field based model, Universal Soil Loss Equation (USLE), is used to estimate
field erosion for targeting critical erosion fields. This traditional field based approach
lacks consideration of spatial variability and distributed attribute characterization. The
diffuse nature of the NPS pollution is not taken into account. However, use of a
watershed based distributed physical process model puts individual fields within a
broader spatial context. Critical area identification or ranking will be based on the
hydrological process in relation to the watercourses. The focus is on what the pollutants
are and where the pollutants are from.

Secondly, can the BMPs be selected? Once critical areas are identified, the next
step is BMPs selection for these critical areas. Although the causal relationship between
NPS pollution problems and BMP selection are known at the qualitative level, their actually
effectiveness still needs to be investigated. A field-based approach usually evaluates the
effectiveness of alternative BMPs at the edge of a field, that is, the reduction of pollution
problems at the edge of a field. A watershed-based approach will need to consider the

effectiveness of alternative BMPs on the watercourse. This research will try to establish a

7
framework that considers the water transport process of NPS pollutants within a watershed

and to address the effectiveness of alternative BMPs from a watershed perspective.

Thirdly, the comprehensive area-wide NPS pollution control plan needs to be
developed that adequately incorporates stakeholders’ values and priorities. Different
stakeholders may have conflicting interests and values, which indirectly affect the level and
types of pollution control desired. A watershed community, includes not only stakeholders
within the geographical watershed boundary but also interested parties outside the boundary
(such as downstream areas or larger ecosystem). They need to work together or battle with
each other for establishing a common agenda in the comprehensive area-wide NPS
pollution control plan. These issues imply the needs to investigate the components of
optimal goals and multi-dimensional interests of the NPS pollution control plan within a

watershed community.

1.4 Research Approach

The approach uses watershed-based optimization for agricultural NPS pollution
management. The structure incorporates the three-step NPS pollution management
procedure. The selected study area is a sub-watershed of the Sycamore Creek watershed
in Ingham County, Michigan. The approach will focus on: (1) agricultural non-point
source pollution, (2) a watershed area, (3) optimization, and (4) management (local plan).
An expert focus group approach will be used to evaluate the process and findings.

This research investigates NPS pollution from a watershed perspective. A
watershed based simulation. model is used as the major tool for analysis. A Min-Max

optimization approach aids in identifying critical areas. Alternative BMPs investigation

8

and selection are based on a watershed context. With a goal setting optimization method,
the area-wide comprehensive management plan can be constructed through a set of Pareto
optimal solutions. These analytical processes incorporated a Geographical Information
System (GIS) and Database Management System (DBMS) as integrating tools to facilitate
processing.

Research findings address the process used for the three-step NPS pollution
management procedure. These results are discussed in terms of each step of the NPS
pollution management procedure. An expert focus group evaluation approach is used to
evaluate the acceptability of research approach and results. The evaluation results are

summarized and discussed.

1.5 Dissertation Organization

Chapter 1 presented the problem setting and research questions. Research
questions are grouped into the three-step NPS management procedure. Chapter 2 reviews
relevant literature for this research. First the policy setting of the NPS pollution control is
discussed. Then, NPS pollution problem regarding the three-step management procedure
is reviewed. The next section presented the topic of optimization technologies. The last
section looks into the use of focus groups.

Chapter 3 describes the detailed research approach in five sections: research
design, study area, data collection, the model, and scenario analysis and optimization.
Chapter 4 presented the research findings for each of the three-steps in the NPS pollution

management procedure. Chapter 5 documents the focus group evaluation processes and

9

results. The final chapter concludes this research and discusses the relevance of the

9

research findings to NSP pollution management.

CHAPTER 2

LITERATURE REVIEW

2.1 Agricultural N on-Point Source Pollution Control

The 1970 Clean Water Act (CW A) and its subsequent Amendments clearly
considered NPS pollution as one of the most serious water quality problems. However,
there is almost no regulatory control on the agricultural NPS pollution problems. One
critical reason for so little control is the lack of mandatory policy for the powerful
agricultural sector. Historically, agricultural policies related activities involved
voluntary/incentive/educational programs. Another major reason is that NPS pollution
problems are mostly observable through off-site impacts. When most forms of
environmental pollution are extemalities, voluntary action has not proved very successful
in clean-up efforts (Epp and Shortle, 1985). Cook (1985) presented a strong view in
favor of using the regulatory approach. He argued that regulation is part of doing

business, and agricultural today is a business as much as a way of life.

2.1.1 NPS Policy Options

One simple cost-effective management approach for agricultural NPS pollution
control is to target hot spots causing priority water quality problems. In the 1985 Farm
Bill, new initiatives that target highly erodible lands were introduced. These initiative

programs are the conservation reserve, sodbuster, swampbuster, and conservation

10

11

compliance provisions. These provisions disqualify farmers from receiving USDA farm
program benefits if they produce agricultural commodities on certain lands - unless they
have and follow a conservation plan approved by their conservation district (Myers,
1986). Such policy focuses on controlling soil loss and associated pollution problems on
the most erodible cropland. The underlying assumption is that large reductions can be
effectively achieved by targeting the most erodible land. Price support and other
commodity programs will provide assistance only for cover practices and croplands that
“constitute an off-farm environmental threat” (Ogg and Pionke, 1986). This national
policy direction indicated the linkage of conservation and price support, which deals with
agricultural NPS pollution problems.

However, there are other policy options for cost-effective targeting of agricultural
NPS pollution controls. Moore and others (1979) reviewed early financial incentives
programs for agricultural NPS pollution control from both farmer’s and public’s
perspectives. Harrington, Krupnick, and Pcskin (1985) reviewed several policy options:
(1) voluntarism (the traditional favorite), (2) “command and contro ” policies, (3)
economic incentives, (4) adjustment of other government policies, and (5) cross-
compliance with other federal programs.

The voluntary approach mainly involves dissemination of information. Through
' moral suasion, education, and technical assistance, farmers are made aware of the
harmful effects of their framing practices and take actions in their self-interest with
provided information. Command and control policies are often enforced through a
permitting process. Two types of standards, design or performance, are the major

regulatory approaches. Design standards, such as BMPs, are easy to implement.

12

However, the outcomes may not be as expected. Performance standards, such as
pollutant discharges, are more efficient but require extensive resources. Both standards
will be hard to monitor and would likely include provisions for cost-share.

Economic incentives here referred to trading or subsidy policies. Pollution
trading is based on economic theory. There are several implementation issues with
economic incentives. First, the target group has to include all potential polluters.
Second, the trading market must have a good transaction mechanism that will minimize
cost. Third, the knowledge of transport and fate of NPS pollutants is needed for better
justification. Finally, there must exist a fair, consensus credit-exchange and monitoring
system for enforcement. Subsidy programs are government payments to farmers or
companies to encourage individual behavior that advances specific policy goal.
Subsidies are quite common in the agricultural sector. For example, cost-differential
payments for adoption of conservation tillage and payments for each unit of pollution
reduced are two typical proposals for agricultural NPS pollution controls. The primary
advantage is its political attractiveness but the major drawback is the inefficiency.

Adjusting government policies include aid for the agricultural sector, tax policies
concerning depreciation, zoning ordinances, and various governmental land and water
programs. Such policies may support, constrain, or negate the potential effectiveness of
policies for reducing NPS pollution. Cross-compliance induces polluters to implement
NPS pollution control under a condition of eligibility for a program that they find
attractive. Under the cross-compliance policy, two approaches, bonus and requirements,
can be used (Christensen and Norris, 1983). The “bonus” approach allows farmers to

receive extra benefit if they also participate in soil conservation program. The

13

“requirements” approach enforces farmers to meet certain rules in order to receive
benefit. The major disadvantage of the cross-compliance policy is that it relies on
another program, in which the polluters may not be interested. The host programs must
be attractive to potential polluters.

Another review (Duda and Johnson, 1985) classified alternative policies similarly.
They discussed the advantages and disadvantages of four types of NPS pollution control
policies: (1) continuing the existing voluntary/educational programs, (2) expanding cost-
sharing/educational efforts, (3) using a system of incentives and disincentives (including
cross-compliance), and (4) a national regulatory program for implementation of
agricultural BMPs. The review concluded that existing voluntary/education or extensive
cost-sharing programs are not likely to achieve the Clean Water Act goals. Only the later
two targeting policy options have any chance of successfulness. It implied that the
incentives and disincentives (including cross-compliance) system will be a popular
framework and the regulatory policy would be difficult to administer and enforce.

Many are debating if the regulatory approach is better than the voluntary approach
or vice verse. Some researchers (Braden and Uchtrnann, 1985) supported a mixture of
assistance and regulatory approach. Assistance will serve as the basic mechanism, but
landowners can be required to act. Moreover, Libby (1985) discussed agricultural NPS
pollution control based on different groups of stakeholders and described two different
approaches, the benefits-received approach and the ability-to-pay approach, related to the
problem of policy efficiency. One of his conclusions is that farmers will be asked to bear

an increasing part of the NPS pollution control bill.

14
2.1.2 Current NPS Policy Setting

All of these policy options reflected different schools of rationale. In the most
recent 1996 Farm Bill, the conservation provisions highlighted a voluntary, cost-share
oriented policy statement for NPS pollution management. The Conservation Reserve
Program (CRP) is the largest and most effective environmental improvement program
(USDA FSA, 1997). CRP Operates through voluntary partnerships between individuals
and Government. The program provides incentives and assistance to farmers and
ranchers for establishing valuable conservation practices that have a beneficial impact on
resources both on and off the farm. The new CRP emphasized four areas: (1) an
improved environmental benefits index, (2) cropped wetlands restoration, (3)
environmental improvement and economic growth, and (4) continuous sign-up.

Another NPS related program under the conservation provisions is the
Environmental Quality Incentives Program (EQIP). The EQIP works primarily in locally
identified conservation priority areas where there are significant problems with natural
resources. The program offers contracts that provide incentive payments and cost sharing
for conservation practices, such as manure management systems, pest management,
erosion control, and other practices to improve and maintain the health of natural
resources. High priority is given to areas where State or local governments offer
financial, technical, or educational assistance, and to areas where agricultural
improvements will help meet water quality objectives. The major operating mechanism
is conservation planning. The plan must protect the soil, water, or related natural
resources in a way that meets the purpose of the program. It is another tool for producers

to use on their farm to help protect natural resources. The underlying philos0phy is to

15

provide an opportunity to recognize and solve the known soil and water problems in the
community by offering incentives and assistance.

The agricultural sector has avoided the regulatory threat for now. However, until
the Nation’s water quality goal is met or significant water quality improvement can be
attributed to the current conservation programs, the voluntary, cost-share, incentive-

oriented NPS policy setting is still under challenge.

2.2 Critical Area Identification and NPS Pollution Models

NPS pollution critical area identification needs to address two basic questions: (1)
what are the pollutants and (2) where are the pollution sources. Mathematical models may
be the most commonly used tools in answering these two questions. Modeling uses
scientific knowledge to represent the real world phenomena in mathematical formulation.
The modeling process often involves simplification and reduction when the real world
phenomena are very complex. A systematic analysis aids in better understand the
components of the phenomena and their interrelationships, that is, the function and structure
of the components. Then, components can be connected to reproduce the phenomena in a
simplified manner. Moreover, modeling is also an art of simplicity.

NPS pollution processes are very complex phenomena that involve hydrological and
nutrient cycling in both spatial and temporal dimensions. Full analysis of the complete
ago-environmental system is required. Bailey and Swank (1983) presented a soil-aquatic
systematic framework for the NPS modeling research. The framework described pesticide
and nutrient (phosphorus and nitrogen) transport and transformation within soil and aquatic

subsystem, and connected through the hydrological cycle within a watershed.

16
Many researchers have developed models to aid in understanding parts of the NPS

pollution system. These models range from simple screening models to complex physical-
process simulation models. There are also many ways to categorize these NPS models,
such as parametric vs. non-parametric, empirical vs. simulation, causal vs. cost-effect,
qualitative vs. quantitative, distributed vs. lumped, field vs. basin, and predictive vs.
descriptive. The following sections will discuss NPS pollution related models in three

groups: empirical, physical-process, and distributed.

2.2.1 Empirical (Lumped) Modeling Approach

Early NPS pollution modeling efforts were limited by the knowledge and available
technologies. Therefore, most models used the empirical approach with a focus on past
experiences. The empirical modeling approach generally results in cause-and-effect
models in which a mathematical expression transforms a set of input variables into a
description of the output results without trying to describe the internal physical processes
taking place. It is also called a “black box” approach, which focuses on the input and
output only. Statistical and probability analyses are the most commonly used
methodology to bridge the causes and effects. The results are often a regression formula,
a simple mathematical expression, or a probability chart.

One very simple empirical model is the Soil Conservation Service (SCS) curve
number method (USDA SCS, 1972 and 1986). The curve number method estimates
direct runoff from storm rainfall for a small watershed. The runoff equation has a very

simple form:

_ (P - 1,)2 Q = (P - 0.2.9)2

Q— or
(P—Ia)+S (P+0.8S)

(when I. = 0.28)

17

where Q = runoff (in)
P = rainfall (in)
S = potential maximum retention after runoff begins (in), and
I. = initial abstraction (in).
Initial abstraction (1.) was found to be approximated as 0.28 for small agricultural
watersheds through many studies. S is related to soil and cover condition of the
watershed through CN (Curve Number) in the following formula:

=£'.99_10
CN

Determination of CN are based on hydrologic soil group, cover type, treatment,
hydrologic condition, and antecedent runoff condition. SCS has developed a set of look-
up tables for average CN based on these factors. Rallison (1980) reviewed the origin and
evolution of this curve number method. He pointed out that the procedures are primarily
for establishing safe limits in design, and for comparing the effectiveness of alternative
systems of measures within a watershed project. They are not used to recreate specific
features of an actual storm. Their use for estimating increments of runoff within a storm
is questionable. He concluded that the procedure is not the state-of-art. However, the
procedure provides satisfactory and valuable results when used appropriately for
answering the types of problems it was designed to solve. The procedure is also very
useful when only meager field data are available.

The Universal Soil Loss Equation (U SLE) is another well-known empirical model
for predicting soil erosion from agricultural fields (W ischmeier and Srrrith, 1978). The
USIE method uses a simple mathematical expression to calculate annual soil loss from a

field:

18
A = RKLSPC

where A = annual soil loss
R = rainfall factor
K = soil erodibility factor
L = field slope length factor
S = field slope factor
P = supporting practice factor, and
C = cover and management factor.

Although the above soil loss formula is simply proportional to six parameters,
each factor is based on a separate set of variables analyzed by statistical or experimental
methods and experiences. For example, the rainfall factor is based on not only the
precipitation amount but also storm type, duration, and intensity. Soil erodibility is an
independent scalar factor derived from experiments. Field slope and slope length factors
represent the topographic setting of the landscape. Supporting practices refers to
contouring and terracing in contrast to the up—and-down slop culture. Cover and
management factor considers crop rotation, tillage and planting methods, and cover
residues. In general, the USLE is factor-evaluation based empirical modeling approach.
Most factors have tables and/or charts for recommended average values (USDA SCS,
1987). While SCS has chosen the USLE as a default tool for evaluating field soil
erosion, many other countries also adopted the USLE with calibration for similar
purposes.

The major advantages of these empirical models are their simplicity and cost-

effectiveness. The simplicity characteristic requires less input data than other types of

19

modeling approach. The result can be easily calculated when limited resources are
available. Once the model is established, it is really the most cost-effective approach for
modeling watershed processes. However, there are also disadvantages. Empirical
models usually can not be extended beyond the range of the data and/or the geographic
unit used in their development. Although calibration can make reuse the models when
they are not directly applicable, the process will require new data collection that may take
years and more efforts. Empirical models can be easily misapplied and/or misinterpreted.
The black box based causal relationship make it hard to explain the causes and effects in
a more precise, quantitative manner. Users will need a process-oriented modeling

approach to better understand the causal relationship.

2.2.2 Physical-Process (Causal) Simulation Modeling Approach

The physical-process modeling approach attempts to describe the physical,
chemical, and biological processes in as much detail as possible without requiring
excessive or unavailable input data. This approach first constructs a conceptual model to
mimic the essential processes, which are important to the phenomenon under study.
Then, each part of the conceptual model is analyzed in detail to gain the functional and
structural interrelationship. Followed by selection of key input and output parameters.
Then model developers can construct a set of complex mathematical formulation based
on best-known scientific knowledge. Novotny (1986) proposed a hydrological
component-based structure for NPS pollution model. The basic components are: surface
runoff generation, soil and groundwater, runoff and sediment routing, erosion, pollutant

accumulation and washoff from impervious areas, and soil adsorption/desorption.

20
The Chemicals, Runoff, and Erosion from Agricultural Management Systems

(CREAMS) model is a physical-process based model for field-size areas (USDA SEA,
1980). It has three major components: hydrology, erosion/sedimentation, and chemistry.
The hydrology component estimates runoff volume and peak rate, infiltration,
evapotranspiration, soil water content, and percolation on a daily basis. The erosion
component estimates erosion and sediment yield including particle-size distribution at the
edge of the field on a daily basis. The chemistry component includes elements for plant
nutrients and pesticides. Storrnloads and average concentrations of adsorbed and
dissolved chemicals in the runoff, sediment, and percolated fractions are estimated.
Simulated values compared favorably with observed data, but the CREAMS model is not
a predictive (absolute quantity) model (Leonard and Krrisel, 1986). CREAMS model
developers recommended that its main usefulness be for determining relative differences
between alternative management systems and chemicals on a given field. By comparing
the differences among alternatives, users can select BMPs on a field. It is especially
useful in planning a complete resource management system, including fertilizer and
pesticide programs.

The Water Erosion Prediction Project (WEPP), supported by the USDA
Agricultural Research Service (ARS), is another example of physical-based simulation
model. Currently, WEPP has developed a hillslope profile model for soil erosion. This
physical-process model predicts soil erosion based on six fundamental components:
climate, infiltration, runoff, soil, plant, hydraulics, and erosion mechanics. The climate
component generates rainfall amount, duration, maximum intensity, time to peak

intensity, maximum and minimum temperature, and solar radiation for the on-site

21
location. The hydrology component calculates infiltration, the daily water balance

including runoff, evapotranspiration, and deep percolation. The plant growth component
calculates growth, senescence and decomposition of plant material. The soil component
makes adjustments to soil properties on a daily time step. Moreover, many soil
parameters are also used in other components. The irrigation component accommodates
solid set, sideroll, and handmove systems. However, spatial variations in application rate
and depth within the irrigation area are assumed negligible for the hillslope slope model.
The erosion component uses the steady state sediment continuity equation as a basis for
the erosion computations. Basically, each component has its knowledge domain that
interconnects with each other through simulation equations. Application of WEPP is
limited to field-sized areas or conservation treatment units and not applicable to areas
having permanent channels. Although the WEPP model provides spatial and temporal
detail of on-site and off-site soil erosion, WEPP model developers emphasized that the
model may serve as a design tool in which soil conservation measures may be spatially
and temporally arranged to meet conservation objectives.

Physical-process simulation models can predict absolute responses and assess
relative effects of environmental changes (DeCoursey, 1985). They serve as a tool for
problem solving and improve our understanding of systems being modeled. However,
the required volumes of data and resources needed for the computation complexity can
offset these advantages. Moreover, none of the physical-process based NPS simulation
models claimed to be able to predict the absolute values of modeling results. Most of
these models may take into consideration of the time scale, but they are still field-based

models, which still do not address the spatial dimension of the NPS pollution problem.

22
2.2.3 Distributed Modeling Approach

The distributed watershed modeling approach adds the spatial and temporal
domains to both empirical and physical process modeling approaches. The modeling
efforts not only consider the connection of components at a specific point time, but also
extend the connection to continuous time and space scales. The conceptual model really
takes into account all water transportation processes of the hydrologic and nutrient cycle
within a watershed. Distributed models usually interrnix empirical and physical process
methods since some components may be better depicted by one approach and others may
be more suitably represented by another. It takes the simplicity from the empirical
models combined with the simulated physical-process models to describe the watershed
processes from a spatial perspective.

Areal Nonpoint Source Watershed Environment Response Simulation
(ANSWERS), Agricultural Nonpoint Source pollution model (AGNPS), Simulator for
Water Resources in Rural Basins - Water Quality (SWRRBWQ), and Soil and Water
Assessment Tool (SWAT) are examples of distributed watershed models. These models
require that users first define small homogenous units (grids or sub-basins) in the
watershed. Each unit needs a set of data for input parameters. Distributed models
themselves usually consist of sub-components that are based on either empirical or
physical-process model. All sub-models are operating at the same time for each unit.
Units are spatially interacting and interconnecting through pre—defined water
transportation pathway. Continuous time-series runs will keep track of precedent unit
conditions and progressively update each unit. The results are not limited to a specific

location or time, but available at all units and time for the watershed.

23
ANSWERS (Beasley et. al., 1980 and Beasley, 1986) is based on one of the

earliest distributed hydrological model developed by Huggins and Monke (1966). It
improved the original runoff modeling framework with the addition of channel and
subsurface drainage components and sediment production/transport relationships.

AGNPS (Young et. al., 1987 and 1989) has three basic components: hydrology,
soil erosion/sedimentation, and nutrients. The hydrology component uses the SCS Curve
Number method to compute the runoff for each grid or cell. The erosion and sediment
component adopted the USLE model for calculating on-site soil erosion. The nutrients
component is based on the CREAMS model but with an add-on manure management
system. A simple additive runoff routing algorithm was developed to spatially
interconnect these three components and to serve as the pollutant transport and
transformation mechanism.

SWRRBWQ (Arnold et. al., 1991) is a basin scale water quality model developed
within the framework of a hydrology model — SWRRB (Williams et. al., 1985 and
Arnold et. al., 1990). It added sediment routing, pesticide movement, nutrient cycling,
and nutrient transport. SWRRBWQ has three major components: weather, hydrology,
and sedimentation. Processes considered include surface runoff, return flow, percolation,
evapotranspiration, transmission losses, pond and reservoir storage, sedimentation, and
crop growth.

SWAT ( Srinivasan and Arnold, 1993 and Arnold et. al., 1996) is a continuous
time model that operates on a daily time step. It incorporated eight major components:
hydrology, weather, sedimentation, soil temperature, crop growth, nutrients, pesticides,

and agricultural management. Most of these components are based on three models:

24
CREAMS, Modified USLE, and GLEAMS (Knisel, 1980). It is very similar to the

AGNPS model with the exception of daily steps simulation and crop growth.

Distributed watershed models are more complex, require increased input data and
more computing time (Vieux et. al., 1989). Thus, the distributed modeling approach
suffers an even greater data volume problem than had the physical-process modeling
approach already encountered. Further, model parameters or the distribution throughout
the watershed may not be known, resulting in the distributed process model’s capability
exceeding the availability of data. Unfortunately, output results from this type of models
contain at least equal amount of information to be analyzed. For some users, it is a great
advantage because knowing more means better control. For others, it is an endless
frustration since they are still in the data world. But this time, it is output instead of
input. The easiness of usage is at the other extreme compared with that of empirical
models. In spite of the above disadvantages, distributed watershed models move closer
and closer to the real world watershed processes and they are powerful tools for resource

managers if required data are available.

2.2.4 Model Selection

Regardless of all types of models, DeCoursey (1985) proposed that model selection
should consider four dimensions: the objectives, time scale, space scale, and most important
of all, data availability. Leonard and Knisel (1986) considered model purpose,
representation, data requirements and availability, ease of parameter estimation, and both
case and cost of simulation as the major factors for model selection. Novotny (1986)

pointed out that a good model should be able to represent the spatial variability of the area

25
and to simulate the distributed physical-process of water pollution. Vieux et. a1. (1989)

reviewed a set of models for GIS-based water quality modeling for agricultural resource
management systems and revealed three criteria for model selection.

(1) The model must be a distributed parameter, watershed scale model that
computes both surface and subsurface water quality and quantity for fertilizers,
pesticides and sediment on a grid or polygon representation of the watershed and
on an event or continuous simulation time scale.

(2) The model must be a disuibuted parameter watershed scale model that computes
at least one of the followings: subsurface water quality or quantity for fertilizers,
pesticides or sediment using either a grid or polygon representation of the
watershed and on an even or continuous simulation time scale.

(3) The model must be a distributed parameter watershed scale model that computes
at least one the followings: subsurface water quality or quantity for fertilizers,
pesticides or sediment using a grid representation on an event or continuous
simulation time scale.

None of the models reviewed met the first most restrictive criterion. Only three models,

ANSWERS, AGNPS, and WEPP, met the third moderately restrictive criterion.

2.3 Best Management Practices

The traditional agricultural BMPs usually refer to practices used for conservation
farming. These alternative BMPs focus on practice methods and field scale management.
Unfortunately, there are two potential dilemmas in the traditional BMPs approach. First, a.

single practice method can not solve the problem of multiple types of NPS pollutants due to

26
the conflicting nature of the NPS pollution mechanism. For example, a practice method

found to be effective in reducing soil erosion usually will increase the nutrient concentration
(Roka et. al., 1990). Secondly, field scale management can hardly address the spatial
dimension of the NPS pollution problem. For example, a watershed without any identified
critical fields may show the NPS pollution problem in the water course due to the
accumulative effect from all or parts of the none critical fields. On the other hand, a
watershed with identified critical fields may not actually have the NPS pollution problem
since pollutants generated from identified critical fields deposit somewhere in the watershed
before reaching the water course.

While many researchers from educational institutions, governmental units, or private
sectors have published on agricultural management practices, most can only be classified as
“management practices” studies. By definition, BMPs must be the “best” management
practices, which are subjected to the purpose or goal in judgement. Without comparing
among different management practices for specific situations and conditions, the “best” is
not really considered. In order to be qualified as BMPs research, the goal must fist be
defined. Then, alternative management practices can be compared relative to the goal. This
implies two major difficulties in BMPs research. First is the goal setting. Goal setting
involves not only the measurement of results but also the broadness, i.e., spatial and
temporal scales, of the goal definition. For example, if a defined goal is to reduce NPS
pollution from a field, it may not resolve the NPS problem in the downstream watercourse.
The second difficulty is the resource and research design issue. A piece of land can not have
different management practices at the same time, that is, the same or similar setting can

hardly be duplicated in the same time and the same place. Although experimental field

27
controls are possible, they are very resource intensive and are generally limited to field

scale.

2.3.1 Paradigm Shift

Due to lack of the “best” component in the BMPs, alternative terms, which better
describe the nature of management practices, have been used. For example, the USDA’s
Water Quality Initiative (USDA 1995) uses agricultural management systems for not only
improved crop and livestock production management practices but also economic and
environmental considerations within a Management System Evaluation Area (MSEA). A
MSEA is usually defined as a watershed. Another term, Resource Management Systems
(RMSs), was used by the Nonpoint Source Unit of the Michigan Department of
Environmental Quality to describe comprehensive management practices for NPS pollution
control. A RMS is a combination of best management practices, which must be applied in
conjunction to prevent or reduce degradation of the water. EPA uses a broader term,
Restoration Practices, from ecological point of view. Ecological restoration for water
quality is the re-establishment of chemical, physical, and biological components of an
aquatic ecosystem that have been compromised by stressors.

The traditional conservation farming concept has a focus on productivity. For
example, the purpose of soil loss reduction is to maintain the long-term soil productivity
(Duttweiler and Nicholson, 1983). It emphasizes on-site impact instead of off-site
extemalities. Until recent years, environmental issues related to agricultural NPS pollution
have been more seriously examined (Troeh et. al., 1980). With the productivity mind set,

the primary goal of any management practices is still based on economic. The agricultural

28

industry used to think of conservation practices in terms of efficiency, that is, economical.
For example, a book entitled “Conservation Farming” published by John Deer (1980) stated
that:

“The basic goal of a fertilizer conservation program should be to use fertilizer more

efficiently - not just use less fertilizer. (p. 94)”
However, the same publication has provided a very thorough examination of conservation
practices for agricultural related energy, water, soil, fertilizer, and pesticides management.

In the meanwhile, the management concept has also progressed toward a more
watershed-based approach. Several studies investigated the effectiveness of alternative
BMPs at the watershed scale. Donigian (1986) demonstrated using the Hydrological
Simulation Program - FORTRAN (HSPF) to evaluate BMPs in a small Iowa watershed.
Klaine et. al., (1988) tried to characterize the nutrient loss and erosion in a single-field
watershed in Tennessee. Prato et. al. (1989) applied the AGNPS model to examine the
economic efficiency (optimal) for reducing soil erosion among alternative erosion control
practices in a Idaho watershed. The size of these watersheds is quite small and there are
very few farms (less than 20) within their study areas. These early studies represent a
gradual paradigm shift in the way BMPs were defined. Recent studies have integrated

BMPs with the planning process and will be discussed in the later section.

2.3.2 A System Perspective
With the advance and change of BMPs concept, i.e., from field to watershed and
from individual practice to systematic practices, many materials on BMPs design and

implementation are now available. In 1992, Michigan Department of Natural Resources

29
(MDNR), Surface Water Quality Division (SWQD) published guidance and specifications

of BMPs (MDNR SWQD, 1992), which provide users with information to help design and
implement the BMP. The document classified BMPs into eight categories: construction site
preparation, housekeeping, managerial, runoff conveyance and outlets, runoff storage,
sedimentation control structures, vegetation establishment, and wetland. Each category
covers several related practices with focuses on planning considerations, design (site
preparation), construction (application), and maintenance (after construction).

In 1993, MDNR SWQD N onpoint Source Unit (NSU) published the “Agricultural
Best Management Practice Manual” to provide the conservation planners a technical
resource to address NPS water quality problems. This manual discusses NPS and impacts,
and includes a step-by-step process for developing watershed plans. It integrated water
quality concepts into the conservation planning process on a watershed basis. Many federal,
state, and local agencies have contributed to this documentation. It emphasized using the
Resource Management Practice Systems approach for NPS pollution control. The RMS
approach has several distinguish characteristics. First, it uses a systematic approach for NPS
pollution problems. RMS identifies specific best management practices that must be applied
in conjunction with each other. Therefore, it considers the interaction of multiple NPS
pollution problems with combination of BMPs. Secondly, the RMS approach evaluates
water quality effectiveness of specific BMPs based on both long-term and downstream
impacts, that is, the spatial and temporal dimensions of the NPS pollution. Thirdly, RMS
tried to integrate existing water quality programs by providing consistency through its
systematic analytical view. There are 19 identified RMSs. Each RMS has a set of BMPs.

The combination of BMPs will not compete with each other, but complement each other in

30
existing programs, which will help accelerate the implementation of BMPs. Finally, the

RMS can easily become part of a watershed project. The recommended BMPs for each
RMS also integrates the conservation of soil, air, plants, and animals as well as improve or
protect the level of water quality. RMS will fit well into different phases (assessment,

planing, implementation, and evaluation) of a watershed project.

2.3.3 An Ecological Perspective

USEPA (1996) took a different perspective, the ecological perspective, on managing
water quality. The ecological restoration approach emphasizes and endorses the use of
restoration techniques. Natural restoration techniques use materials indigenous to the
ecosystem and are linked or incorporated into the dynamics of a river system in an attempt
to create conditions in which ecosystem processes can withstand and diminish the impact of
stressors. These techniques are classified into three categories: instrearn, riparian, and
upland, or surrounding watershed The upland restoration techniques are generally related
to the control of NPS inputs from the watershed. Another key focus is using watershed
protection approach, the Total Maximum Daily Load (TMDL) process, to provide an
impetus for restoration activities. Through water quality standards defined as designated
uses, numeric values or narrative statements, or an antidegradation provision, ecological
restoration techniques can effectively address water quality impairments. When evaluating
the cost-effectiveness of restoration techniques, the approach offers two primary economic
reasons that (l) restoration often has lower marginal costs and (2) restoration provides a

wider range of ecological benefits.

31
2.4 Area-wide Comprehensive Planning and System Integration

Area-wide comprehensive planning is the final decision-making process for NSP
pollution control. At the critical area identification step, decision-makers seek to find what
the pollutants are and where they originate. Then, related agricultural activities are
identified and alternative management practices can be investigated. Finally, decision-
makers will need to construct a plan for the problem. The planning process involves two
major management components: selection and measurement. The selection component
typically refers to BMPs selection, that is, where to install what. Information derived from
the first two steps provides the foundation for BMPs selection. The measurement
component is the goal or achievement of the plan, that is, the level of NPS pollution control
desired. Moreover, the planning process quite often requires an integration approach, which
uses different tools/systems to facilitate the planning process. The following sections
reviewed relevant literatures regarding these two planning components as well as integrated

system approach for NPS pollution control.

2.4.1 Approaches for BMPs Selection

Setia (1987) conducted an innovative research project on selection of conservation
farming practices from a risk perspective. In this study, maximizing the expected utility
from net returns is the primary economic goal. Safety-first criterion is used as constraints.
These two factors represent farmers’ risk attitude in choosing crop rotation and conservation
practices. A Monte Carlo simulation approach was used to predict the stochastic nature of
crOp prices, crop yields, and weather for both short-term and long-term real-world

situations. However, the research only focused on soil productivity and no environmental

32

extemalities were considered. Setia found that farmers’ attitude toward risk will influence
the selection of a management system. Therefore, information on farmers’ attitude about
risk can aid in designing effective conservation recommendation for achieving established
conservation goals.

The risk concept can be interpreted as the probability of damage. It can be
represented as the set of “triples” (Kaplan and Garrick, 1981):

R=(<Si,P,,Xi>)

where Si = a scenario identification or description,

P, = the probability of that scenario, and

X; = the consequence or measure of damage.
Mulkey et. al. (1986) applied this risk triples approach with different pesticide leaching
models to estimate the probability of pesticide exposure given sources defined as field-scale
units. The research found that all models responded similarly in terms of the relative
impacts for different chemicals regardless of the different structures of these models. The
researchers concluded that estimates of exposure probabilities (risk assessment), such as the
magnitude, frequency, and duration of pesticide concentrations, can be used to measure
threats to water quality and provide guidance for BMPs selection.

Recently, USDA Natural Resources Conservation Service (NRCS) has developed
the National Agricultural Pesticide Risk Analysis (NAPRA) tool based on the risk
assessment concept. The NAPRA approach combined pesticide runoff/leaching models
with stochastic weather probability to estimate risks (vulnerability index) of pesticides
application for both surface (edge of the field) and groundwater (bottom of the root zone)

contamination. The goal of NAPRA is to help decision-makers or farmers to make

33
informed pesticide management decisions, i.e., BMPs to prevent environmental

extemalities.

Another school of selecting agricultural NPS management approach is input
management. The basic idea is that consequences of higher inputs will result in higher NPS
pollution (Weinberg, 1990). Attention now must focus on the input side of the production
system while the crop yields or economic returns per input spending has graduated leveled
off (Odum, 1987). The call for “reduced-input agriculture” simply is to manage more
efficiently the inputs in order to reduce the threat of NPS pollution.

Based on this low-input agriculture concept, emphases are on the source controls of
the agricultural production system. Land management practices, such as irrigation
scheduling to rrrinirnize water use, surface runoff, and leaching, conservation tillage can
minimize loss of pollutants; and nutrient and pesticide management to reduce chemicals
application are classified as low-input management practices. Flach (1990) argued that low-
input agriculture is compatible and supplementary to existing soil conservation. He stated
that low-input farming means going back to managing the soil-crop system as an integrated
system and we should seek to improve rather than just to conserve soil.

In early 1980’s, Haitlr (1982) published a systematic approach for farm-level
management of NPS pollution control. He constructed a linear optimization model, which
maximize net farm income with environmental goals as constraints. The study unit is a
dairy farm including pasture lands. Soil and nutrient (nitrogen and phosphorus) loss are
expressed as constraints in the model. Different levels of environmental constrains were
tested against different management practices. Results (maximized net income under

allowable pollutant losses) are compared among these scenarios. One of the conclusions is

34
that reductions in pollutant losses are very expensive for the study farm, since major

decreases in cropped land were necessary. The effects of reductions in pollutant losses can
be measured as trade-off of associated reduction of net income. Two major implications of
this research are that:
(1) enviromnental goals, i.e., pollutant losses, can be analyzed at the same time
without using estimated dollars of pollution damages; and
(2) the approach provides planning agency with information useful in developing a

management program for dairy farm NPS pollution control.

2.4.2 GIS Integration

While different tools and technologies were introduced to facilitate NPS pollution
management, system integration has gradually evolved to be a popular approach to provide
decision-makers to make scientific sound, informed decision. There has been extensive
research on integrating models with Geographic Information System (GIS) to evaluate NPS
pollution. This research ranges from field, to watershed, to statewide.

Ventura et. al. (1988) described a land information system for soil erosion control
planning. The system uses USLE and a GIS to facilitate the soil erosion analysis process.
The system stores four types of data, soil properties, topography, land ownership, and land
cover, in a GIS. A FORTRAN program is used to combine these base data with the USLE
model for soil erosion analysis. Additional information about conservation practices and
crop rotations was incorporated into the analysis to create different scenarios. Results (soil
loss) from different scenarios served as the basis for comparison with current farming

practices and crop rotation. Then, farmers can work with conservation staff to determine the

35
additional efforts necessary to achieve erosion control goals. The study also identified

cropland eligible for the Conservation Reserve Program (CRP) based on three eordibility
criteria defined locally. Policy-makers were impressed when shown the extent and location
of land newly eligible for the CRP withing a week of a rule change on the erodibility
criteria. The automated process to conservation planning provides equal treatment for all
landowners in a County. Uniform quality data are used in consistent analyses to generate a
neutral equity plan, which seemed to be the strongest advantage concluded by the
researchers.

He et. al., (1993) also applied GIS with a NPS water quality model (AGNPS) on a
large watershed. Six different combinations of hypothetical management practices and
tillage methods were applied to all crop lands in the watershed. The study incorporated two
time periods, fallow and mid-cropping season (establishment), into consideration, which
results in twelve scenarios. The focus of the research is on the comparative impacts of
alternative agricultural practices at the watershed scale. One of the interesting findings is
that the institution of conservation tillage resulted in decreased soil loss but increased
nutrient loading. This implied a conflicting effectiveness exist between different types (soil
erosion and nutrient) of NPS pollution control. However, good management practices, such
as contour farming and crOp residue cover, reduced both soil and nutrient nlnoff.

Needlram (1994) conducted a thorough examination on the use of GIS for a complex
time-serious water quality model (Ann-AGNPS). The study compared GIS-based spatial
manipulation methods with conventional map interpretation and reconnaissance field
surveys method in derivation of model input parameters. He concluded that simulation

results are comparable between these two different sets of input method. The errors

36
introduced by GIS-based methods are insignificant relative to the model results.

Furthermore, Needham also analyzed the usefulness of the model results through validation
of monitoring data. He found that the Ann—AGNPS was only capable of providing
identification and ranking description.

Petersen et. al. (1991) published a report on using GIS to evaluate statewide NPS
pollution potential on a watershed basis in Pennsylvania. The study used a ranking system
on a watershed basis. The ranking index, Agricultural Pollution Potential Index (APPI), is
based on four factors: Sediment Production Index (SP1), Runoff Index (R1), Animal Loading
Index (ALI), and Chemical Use Index (CUI). Each index is adjustable by a weighting factor
(WF) and can be represented in the following formula:

APPI=RI*WF,+SPI*WF2+ALI*WF3+CUI*WF4
Most base data, such as land use, topography, and soils, are stored in map format. A GIS
was used to compute the above formula on a watershed basis. The results (APPI)
represented the relative NPS pollution potential for all 104 watersheds in Pennsylvania.
This research demonstrated that GIS can be an efficient tool for handling the large amount
of data nwded in NPS pollution analysis.

Arnold et. al. (1990) described different approaches for GIS integration with natural
resource modeling and site selection. First, the study investigated an intelligent GIS based
water and sediment yield model. It embedded a rule-based system in GIS to determine
model variable values. Secondly, a distributed model (ANSWERS) was integrated with
GIS. The GIS provided a diverse set of functions in model data needs and simulation result
presentation. Lastly, the GIS was integrated with a neural network package for site

selection. The study concluded that GIS will likely supply much of the data needed by

37
agricultural NPS pollution models in the future, which make it possible to predict soil

erosion or water pollution with distributed parameter models.

2.4.3 Information Systems and What Ifs

One of the important features that an integrated system approach can provide is the
ability to handle “what if” questions. Decision-makers will be able to obtain results in a
short time period for different management strategies if the information processing is
automated. USDA NRCS (1997) has recently developed a prototype Hydrologic Unit
Water Quality (HUIW Q) tool (USDA NRCS, 1997). The current version included two
watershed-based water quality models, AGNPS and SWRRBWQ. The HU/W Q tool
integrated a GIS, Geographic Resources Analysis Support System (GRASS), a relational
database, Informix, with these two water quality models. Users can create different
management scenarios and quickly derive the result on a watershed basis. However, no
further scenario comparison analysis is available.

Another integrated system, Better Assessment Science and Integrating Point and
Nonpoint Sources (BASINS), developed by USEPA (1996a) combines point and non-point
sources analysis together. BASINS has become a popular TMDL assessment and analysis
tool. The new version (2.0) contains a set of EPA facility monitoring database, such as
Permit Compliance System (PCS), Toxic Release Inventory (TRI), Superfund National
Priority List (NPL's, Industrial Facility Dischargers), and Resource Conservation and
Recovery Information System (RCRIS), as well as a set of natural resource data layers, such
as weather station data, Reach File (version 3), SSURGO soils, landuse, and elevation

(DEM). It incorporated a water quality model, Hydrological Simulation Program —

38
FORTRAN, a powerful external GIS/Database package, ESRI ArcView, and other utilities,

such as watershed delineation, reporting, and reclassification. These new datasets and
functionalities provided users great flexibility in assessing both point and nonpoint sources
poHudon.

Ironically, one of the difficulties with the use of information system for decision-
making is the volume of information generated. Since information system provides great
efficiency in creating different scenarios, the side effects are that there are too many
alternatives. Decision-makers now face another problem in evaluate these alternatives. This
type of problems involves judgement in ethical choices, trade-off between cost and services,
conflicts of preferences, and even “political” influences (Keen, 1986). In general, it is an
issue about “how to make a decision”. Many researchers have investigated different
decision support approaches for NPS pollution control. One of the most common
approaches is the use of knowledge based system. Knowledge based systems represent
knowledge symbolically and include a means to selectively apply parts of that knowledge to
a specific problem (Ferrnanian, 1989). When symbolic knowledge can be generalized into a
set of rules representing human expert on specific problems, it is called expert systems or

rule-based systems.

2.4.4 Decision Support System (DDS)

Guariso and Werthner (1989) published a book on Environmental Decision Support
Systems. It listed the basic characteristics of a decision support system, which are:

(1) to assist managers in their decision making process for unstructured or semi-

structured tasks,

39
(2) to support and enhance rather that replace managerial judgement,

(3) to improve the effectiveness of decision making rather than the efficiency,

(4) to combine the use of models or analytical techniques with data access functions,

(5) to emphasize flexibility and adaptability to respect changes in the context of the

decision process, and

(6) to focus on features which make them easy to be used interactively by non-

experienced users.

Furthermore, the decision making process usually has four phases: intelligence,
design, choice, and implementation. Intelligence has to do with the search in the decision
context, i.e. looking up necessary data, isolating the raw material of the decision process,
and identifying more precisely the problem domain and its structure. The design phase
analyzes different possible abstract models of the problem. The choice phase performs
the selection of a plan from the set of possible alternatives. The last phase,
implementation, deals with the execution of a selected plan. The authors proposed an
extended decision support system architecture based on function and tools. The
architecture contains five modules: knowledge base, database, modelbase, system
management and control unit, and user interface. Each module was described in details
in theory and through an example.

Jones (1989) discussed the integration of simulation models, database management
systems, and expert systems in aids of the decision-making process. The study revealed the
efficiency of using database management systems for storing model input and output data
and the effectiveness of using expert system for alternative selection of model simulation

and model output interpretation. XGRCWP (Zhao et. al., 1994), Expert GIS Rural Clean

40
Water Program, is an example of integrated decision support system for NPS pollution

management. It integrates GIS, relational databases, simulation models, and Hyper Text
Mark Language (HTML) documents to form an advisory system for the selection, siting,
design, and evaluation of nonpoint source pollution control practices in agricultural
watersheds. A rule-based expert system was built for selecting feasible control practices.
Another example is a rule-based expert system, developed by Go (1996), for evaluating the
costs and benefits of soil conservation. The research focused on crop rotation and
supporting practices, i.e., up-down hill tillage, contouring, and terrace, for soil erosion. The
expert system included 158 rules in adjusting practice recommendation. Environmental
extemalities (soil loss) were converted into monetary measures and integrated into farm
economic analysis. The results (recommendation) showed close match with
recommendation by NRCS.

Mohite and others (1993) constructed an integrated decision support system for
agricultural NPS pollution control. The approach combined visualization, expert system,
GIS, and a NPS model (AGNPS) for BMPs selection. It first defined critical areas based on
three criteria:

(1) absolute soil loss (a expert suggested threshold value relative to 5

tons/acrelyear),

(2) soluble nitrogen > 10 ppm, and

(3) soluble phosphorus > 0.05 ppm.

Then, four different farm management practices, conventional tillage, conservation tillage,
no—till, meadow and pasture, were evaluated to find out the most suitable combination both

in the context of cost effectiveness and increased productivity. The study concluded that an

41
artificial intelligence approach to GIS-modeling can be used to target NPS hot spots and

select appropriate soil conservation practices. However, the researchers also stated that
evaluation of effectiveness of selected BMPs would require collection of field data for

validation of results.

2.4.5 Social Considerations

There is also the social aspect for NPS management control planning. Sorenson
(1985) described the Wisconsin experience of NPS pollution management and summarized
that the information goals of a plan are to create awareness, to aid in understanding
problems and their solutions, and to improve communication between the public and project
staff. Gensemer and Yamaguchi (1985) identified five basic objectives for a successful
area-wide NPS pollution management project. These objectives are:

(1) emphasis on implementation in the initial project concept;

(2) local endorsement of the planning process from the beginning and dedication to
seeing the problem resolved;

(3) formation of oversight, technical, and/or executive committees made up of the
various interested parties to assume shared planning and implementation
responsibilities;

(4) a regional agency acting as mediator between local governments and between
local and state authorities; and

(5) public participation that emphasizes promoting problem solutions rather than

merely dissemination information.

42
This social aspect of NSP pollution control is a very important element in both planning and

implementation processes. Since the policy setting of agricultural NPS pollution control is
still primarily based on a voluntary approach, the social process is the key to successful
agricultural NPS pollution management. Moreover, recent watershed management
strategies also emphasize on the social dimension in addition to the traditional biological,

physical, and chemical dimensions. Although this research does not involve this social

aspect, its importance is recognized.

2.5 Optimization

Optimization is a branch of applied mathematics known as optimization dreary,
which studies the search of extreme (minimum or maximum) values among feasible
solutions. It is a science that studies the best (Pike, 1986). In the optimization
terminology, there are three basic terms: objective functions, decision variables, and
constraints. Objective functions can be as simple as a linear equation or complex
simulation models. These functions are composed of variables, that is, decision
variables. A decision variable may have upper and/or lower limits, that is, constraints.
Then, the optimization process involves changing decision variables, which must satisfy

any given constraints, to find the extreme values of the objective functions.

2.5.1 Optimization Problems
There are two types of optimization problems: (1) single (scalar) optimizationand
(2) multicriteria optimization. The principle of optimization is to select the best possible

solution for a given set of circumstances without having to enumerate all the possibilities.

43

The search for extreme values usually is very time-consuming. Many theories and
techniques have been developed to improve the computation efficiency and to ensure the
correctness of the results.

Scalar optimization begins with a set of independent variables or parameters, and
often includes conditions or restrictions (constraints) that define acceptable values of the
variables and a single measure of “goodness”, termed the objective function, which
depends in some way on the variables (Gill, 1981). The solution is a set of allowed
values of the variables for which the objective function assumes an “optimal” value. In
mathematical terms, the optimal (extreme) value usually involves maximizing or
minimizing.

Multicriteria optimization (also called multi-perforrnance, multi-objective, or
vector optimization) can be simply defined as optimization problems with multiple
objectives (Eschenauer et al., 1990). More specifically, multicriteria optimization tried to
find the “best” solution (optimal value) simultaneously among several non-
commensurable criteria. In the abstract, a multiple-criteria decision problem involves
four important elements (Y n, 1985). These are as follows:

1. The Set of Alternatives from which we will choose our decision.

2. The Set of Criteria with which we are concerned for making a good decision.

3. The Outcome of Each Choice measured in terms of the criteria.

4. The Preference Structures of the Decision Maker.

Multicriteria optimization also has a key characteristic feature, objective conflict,
i.e., none of the feasible solutions allows the simultaneous maximization or minimization

of all objectives, or the individual solutions of each single objection differ (Eschenauer et

44
al., 1990). In addition, multicriteria optimization problem solving is an organized and

constructive approach to explore a broader range Of alternative solutions, which provides
a basis for explicit trade-Off between conflicting objectives or interests.

All multicriteria optimization problem solving looks for an Optimum solution.
However, a single solution point for Optimum does not usually exist due to the Objective
conflict nature but rather a “functional-efficient” (Pareto-Optimal) solution set, i.e., the
decision makers select the most efficient compromise solution out Of such a set
(Eschenauer et al., 1990). Therefore, multicriteria Optimization solutions are usually a set
of solutions called Pareto Optimal solutions, which simply means that one objective can

not be enhanced without diminishing another Objective.

2.5.2 The Pareto Concept
The concept of Pareto Optimality was first introduced by Edgeworth in 1881
(Stadler, 1988). In economics, the criteria are generally the utilities of individual
consumers. Edgeworth first successfully defined an Optimum for such a multiutility
problem in the context of two consumers, P and it:
It is required to find a point (x, y) such that in whatever direction we take an
infinitely small step, P and it do not increase together but that, while one
increases, the other decreases.
Some years later, in 1906, Pareto took the more direct approach of ordering the decision
set directly and subsequently defining an Optimum for n consumers in the following

manner:

45

We will say that the members of a collectivity enjoy maximum Ophelimity in a
certain position when it is impossible to find a way of moving from that position
very slightly in such a manner that the Ophelimity enjoyed by each Of the
individuals of the collectivity increases or decreases. That is to say, any small
displacement in parting from that position necessarily has the effect of increasing
the Ophelimity that certain individuals enjoy, of being agreeable to some and
disagreeable to others.

Consequently, Pareto optimality has become the most important part of the
multicriteria Optimization problem solving and it is usually referred to a Pareto Optimal
solution set (also called efficient, noninferior, nondominated, or admissible solutions).

A Pareto optimal solution is a feasible solution for which an increase in the value
Of one criterion can only be achieved by degrading the value of at least one other criteria
(Romero and Rehman, 1989). Alternatively, it can be defined so that no performance
measure can be improved without worsening at least one other performance measure
(Osyczka, 1984). Furthermore, a set of Pareto Optimal solutions are feasible solutions
such that no other feasible solution can achieve the same or better performance for all the
criteria under consideration and strictly better for at least one criteria (Romero and
Rehman, 1989) or no member of the set is inferior to any other member of the set with

respect to all performance measures (Osyczka, 1984).

2.5.3 Optimization Issues
Searching for optimal solutions and objective function formulation are two major

topics in multicriteria Optimization. Searching involves two basic issues, efficiency and

46
validation. There can be potentially hundreds of criteria in a complex Optimization

problem. In addition, when the Optimization process includes simulation models, the
computation time can be very long if the search algorithm is not very efficient.
Furthermore, derived Optimal solutions may not be a global Optimal solution. .It can be a
local Optimal solution when the search algorithm failed to explore all potential solutions.
The simplest search algorithm is a random method. Basically, solutions were
generated by randomly assign values to decision variables within the specified
constraints. These solutions were compared to each other to find the extreme values. It
is Obvious that the random method will not guarantee a global solution. A more
advanced algorithm is called SIMPLEX. The SIMPLEX method is an algebraic
procedure used to systematically examine extreme points efficiently (Haitlr, 1982). It
iteratively searches for extreme points that improve the value of the objective function.
The SIMPLEX method works very well with linear (scalar) optimization problem, but
also failed to ensure that the final solution is a global optimal solution in non-linear
objective function cases. Another more sophescated algorithm is called COMPLEX.
The COMPLEX first requires generation of a set of random decision variable points, also
called decision variable space. Each point of decision variables can be mapped to a
objective function point, that is, the Objection function space. If it is a two-dimensional
problem, these spaces are similar to the topography of the landscape. The maximum
value is the highest elevation of the terrain and the minimum value is the lowest elevation
of the terrain. These two spaces provided the base setting for searching the Optimal
solution. The searching algorithm uses certain ratio (relative to the other points in the

space) to control the “speed” of movement in the spaces. In this way, it can quickly

 

47
approach the optimal points (extreme values) in the space. Unfortunately, it also can not

guarantee that the solution found is a global solution. Manetsch (1989) modified the
COMPLEX with an adaptive approach. The moving ratio will not only change relative to
the other points in the space, but also the progress (number Of iterations) of the
Optimization process. A large ratio value is used in the beginning and users are allowed
to adjust it based on the “speed” of convergence. Furthermore, it also uses multiple
random initialization to avoid local optimal solution. This adaptive COMPLEX method
adds flexible searching capabilities with human intervention and multiple random
initialization. However, it is still subject to the same local optimal solution limitation but
with a greater chance of finding a global optimal solution.

Since no searching algorithm can guarantee that a derived solution is the global
Optimal solution especially in multicriteria non-linear optimization problems, another
approach is to transform the objective functions into a linear equation and try to retain the
same goals in the new single objective function. Osyczka (1984) reviewed several
scalarization methods on Objective functions. The simplest method is a weighting
method, which adds all the Objective functions together using different coefficients for
each. The second is a hierarchical method based on the importance of the Objective
functions. Each Objective function is minimized separatedly starting from the most
important one. Then, adding in at each step a new constraint, which limits the assumed
increase or decrease Of the previously considered functions. The third method is a trade-
off method. Trade-off means a giving up of one benefit in order to gain another regarded
as more desirable. The trading concept is used to determine the next step in seeking the

preferred solution. The fourth method is using a vector of decision variables which

48

minimizes some global criterion. The global criterion method has a function which
measures “how close the decision maker can get to the ideal vector”. The last
scalarization method is called goal programming. It requires the decision maker to
specify goals for each objective that he wishes to attain. The goals are considered as
additional constraints in which new variables are added to represent deviations from the
predetermined goals. These methods more or less introduce distortions of the original
objective function but the advantage lies on their simplicity.

Other than the function scalarization approach, another alternative is called min-
max approach. It tries to keep the original object function as much intact as possible.
Each Objective function may be subjected to a scalar change relative to itself. It
normalized each objective function by itself. The purpose of the normalization process is
to put different types of objective functions into comparable basis. Then, the
optimization process will minimize the maximum deviation of each normalized objective
function. In this way, each iteration needs to find the maximum deviation among all
Objectives and minimize this particular objective function alone. At the same time,
solutions from each iteration are compared with previous derived solutions by individual
objective function. Objective functions that are worse than previous solutions are
discarded. The final solutions set is a set of Pareto optimal solutions.

In practice, multicriteria optimization has been applied in a diversity set of issues for
resource planning and management. It is a mean of arriving at Optimal decisions in very
complex systems. The emphasis should be on the Optimal decision rather than on the
extreme value of some criteria functions which only served as an artifice for arriving at

optimal decisions (Stadler, 1988).

49
2.6 Focus Groups

2.6.1 Overview

The purpose of focus groups is to obtain data, information, or perceptions on a
defined area of interest in a permissive environment. The focus group technique has been
used as a process of evaluating social programs for years. It is one type of qualitative
research procedures that concentrates on words and Observations to express reality and
attempts to describe impressions from people in a natural situation. Focus groups are
distinct from other kinds of group in several ways (Krueger, 1994):

(1) Focus groups involve a homogeneous set of people that interact in a series of

discussions.

(2) The purpose of focus groups is to collect qualitative data from a focused

discussion.

(3) Focus groups are a qualitative approach to gathering information.

Focus groups typically have several features. First, the size of the group is small.
A focus group typically has 6 to 12 participants, but can range as few as 4 to as many as
12. It must be small so that everyone in the group will have an opportunity to share their
thoughts and yet large enough to provide a diversity of opinions.

Second, the characteristics Of people involved should be homogeneous. Members
in the focus group are similar to each other. Homogeneity can be defined based on the
focused issue to be discussed. The selection of participants is determined by the purpose
of the study. It emphasizes the commonality, instead of diversity, of a group of people.

Third, the environment of a focus group is permissive. The questions asked in the

focus group are simple, clear, and open-ended. People may need to listen to opinions of

50

others before they form their own personal views. Group members influence each other
by responding to ideas and comments in the discussion. Participants are not pressured to
vote or reach consensus.

Fourth, the discussion of a focus group must be focused. Researchers have to
provide consistent and sufficient background information to each participant before the
focus group session. The discussion must be relevant to the context of the purpose. The
focus group discussion should expect relevance, practicality, and utility. Planned
questions are usually used to stimulate the discussion

Fifth, focus groups provide a qualitative data collection opportunity. Focus
groups produce qualitative data that provide insights into the attitudes, preceptions, and
opinions of participants on a specific issue. The results are not based on numbers that
represents Opinions or concepts. They are explanatory and exploratory embedded in the

nature of the discussion.

2.6.2 Processes

A focus group study can be divided into three phases: planning the study,
conducting the meeting, analyzing and reporting the results. The planning phase is
critical for a successful focus group study. Planning processes start with detemrination of
the purposes and the nature of the problem. This will help clarifying the goal to be
achieved and reveal the information needed to address the problem. Then, the target
audience can be identified. The audience must be able to provide answersflnsights to the
problem. The next step is the focus group session design. The design needs to consider

when, where, and how the focus group will be conducted. Tinting may be important

51

relative to other events. The location must be convenient to all participants and the
environment must be comfortable. One of the major concerns is the questions to be
asked since quality answers are directly related to quality questions. The last step will be
to develop a plan of action and estimate resources needed.

The second phase is conducting the focus group session. This phase begins with
participant selection and invitation. Several methods can be used to get people to attend
the focus group. Incentives to participants is the most common method. The most
common type of incentives is money but the amount usually can not compensate the
actual time and effort that participants will provide. However, symbolic incentives can
be even more effective. One type of symbolic incentives is to relate the value of the
problem to participants, which they are concerned about or will be proud of their
participation. Background information related to the discussion issue also need to be
provided beforehand. During the interviewing session, the moderator plays the key role
in facilitating the discussion. The major tasks for the moderator are to direct the
discussion on pro—defined purposes, control discussion progress, handle unexpected
events, and facilitate diverse communication. During the focus group session,
documentation is another important factor. Video or tape recording can provide complete
and truthful evidence but may threaten the discussion environment. Alternatively,
written records may miss some fast conversation. Both methods should be used if
possible. Before the end of the focus group session,.any follow-up actions need to be
announced. A written summary of the discussion has to be delivered for verification as

soon as possible.

52
The last phase is analyzing and reporting the results. Since focus group studies

are a qualitative research approach, the data analysis process will need to consider words,
context, consistency, frequency or extensiveness, specific responses, and ideas in
comments. In general, data analysis is examining the evidence to address the initial
propositions of a study (Yin, 1984). The analysis process typically requires systematic,
verifiable, focused, and practical dimensions. It should also seek enlightenment and
alternative explanations, to lift the level of understanding of the defined area of interest.
The last process is reporting. The report communicates results of the focus group study.
It provides a historic record of findings through a logical description of the investigation.
The report can be in oral, written, or in another media format. It needs to focus on the

purpose of the study and consider the interests of target audiences.

2.6.3 Issues and Concerns

Focus groups research offers several advantages. The technique is a socially
oriented research procedure. It places people in natural, real-life situations to capture the
data, information, or perceptions. Focus groups research has flexibility to explore and
exarrrine issues through interactions among a group of people. It has high validity
through face-to-face participation. Moreover, the cost for conducting a focus group
session is relative low comparing to other interviewing method or survey. The results are
speedy and the volume of information collected in the session is tremendous. On the
other hand, focus groups research also possesses several limitations. Researchers have
less control on individuals in such a group discussion environment. Due to the large

amount of data gathered, it is more difficult to analyze and report. Focus groups also

53
require special skills of moderators to control the progress of the session. Lastly, focus

groups are difficult to assemble.

Focus group interviews have been successfully used in a variety of situations.
The method has also been applied to a group of collaborative activities, such as strategy
development, logistics determination, plan formation, and alternative exploration. When
adapting focus groups to other types of situations, one has to bear in mind what the focus
group can do and what it can not do. Researchers have to consider the purpose of the
effort, the people involved in the process, the nature of the discussion, and the nature of
the environment.

In summary, the purpose of using the focus group approach is not to teach, to
provide therapy, to resolve differences, or to achieve a consensus but to obtain
information in a systematic and verifiable manner (Krueger, 1994). Focus group research
has the potential to be applied in other area of science. It really depends on the
researchers to provide justification for its usage and ensure the validity of using the
method. Linda (1982) stated that:

The (focus group) technique is robust, hardy, and can be twisted a bit and still

yield useful and significant results. This is not an argument for laxity in group

design, nor is it an apology for inadequate moderators. Rather, the point here is

that flexibility rather than rigor ought to characterize the use of focus groups.

CHAPTER 3

RESEARCH METHODS

3.1 Research Design
3.1.1 Overview

This research uses a watershed based optimization approach for agricultural non-
point source pollution management. There are four major components in this study: (1)
agricultural non-point source pollution, (2) watershed, (3) optimization, and (4)
management. In addition, this research uses an expert focus group approach to evaluate
the research findings. The following sections describe the concept and procedures of
these four components and the focus group evaluation.

Agricultural non-point source pollution is the central subject of this study. This
research focuses on agricultural activities, which result in non-point source pollution
problems only. N on-agricultural sources, such as construction and urban, are not
considered. Point source pollution problems are excluded. Only surface water quality is
evaluated and groundwater contamination issues are not addressed. Non-point source
pollution problems Often are classified based on the types of pollutants, such as soil
erosion, nutrients (phosphorus and nitrogen), and pesticides. However, pesticides are not
considered in this research due to lack of data.

The second component is the watershed. A watershed perspective differs from a

field perspective in several ways. A watershed perspective considers the spatial

54

55
interactions within a large defined geographic area. It has diverse and inhomogeneous

characteristics. The bonding mechanism among different landscape objects within a
watershed is water, specifically, water flow path. Implementation of any watershed-wide
management will be time-consuming and resource-intensive. A field perspective has a
focus on a piece of land parcel. Within the parcel boundary, its characteristics are usually
singular and homogeneous. Controls can be applied easier and the required cost and
resources are much less. Because non-point source pollution has a diffuse and dynamic
nature, a field approach may not be able to address such issues. Managers have to take
an area-wide (watershed) perspective to find the sources of non-point source pollution
problems.

Optimization, the third component, is the main method used in the "decision-
making" process for "scenario" analysis. Optimization is the search for extreme
(minimum or maximum) values among feasible solutions. A feasible solution is the
result of a scenario. A scenario can be created by changing the values of decision
variables, i.e., the characteristics (management practices) of a piece of land in a
watershed. There can be many combinations of different land use types for all fields in a
watershed. The emphasis of scenario analysis is on comparing solutions among the
feasible solutions. All of these feasible solutions (scenarios) make up a “space” (set of
solutions) from which one searches for the best solution. While searching in the feasible
solution space for extremes, a better guidance or strategy is needed to efficiently find
better solutions, that is, the decision to chose which direction to move or where to go in

the solution space.

56

The last component, management, looks into several issues regarding the
efficiency and effectiveness of resource usage and watershed management of non-point
source pollution. First, Optimization results quite often are not unique especially when
multiple objectives are involved. When enhancing all objectives can not be reached
simultaneously, the Optimization result is a set of Pareto optimal solutions. These Pareto
optimal solutions simply mean that each objective is equally effective or important, that
is, one can not enhance one objective without diminishing another. Managers will have
to use other criteria to evaluate these given Pareto Optimal solutions. The most common
method is economics, i.e., how much does it cost. However, there are other social factors
that need to be considered from a watershed management perspective. These social
factors, such as values and cultural setting, sometimes are the keys to successful
implementation. A comprehensive management plan must take all these factors into
account.

There are several difficulties in evaluating the research findings by traditional
quantitative or statistic measures. First, non-point source pollution is stochastic; that is, it
is driven by storm events. If there are not enough “big” storm events during the study
period, there will not be enough data for evaluation. There may be lots of historical
storm event data but no corresponding detail agricultural activity data available.
Secondly, no mathematical model has been able to accurately simulate the non-point
source pollution problem accurately at the quantitative level. They are generally applied
in a qualitative manner or used for relative comparison. Finally, scenario analysis is
based on assumptions that can not be actually implemented and tested in a control

environment for a watershed. It will take a long time to test every scenario resulting in

57
very high costs. Due to these difficulties, an expert focus group approach can be used to

evaluate the research finding.

In summary, this research used a physical process based distributed model,
AGNPS, to address the diffuse, dynamic, and stochastic nature of the non-point source
pollution and the spatial diversity and interactions from a watershed perspective. Model
results were applied to evaluate the water quality objectives. These objectives are
embedded into an optimization routine for scenario analysis, that is, to generate feasible
alternatives and find “best” solutions. Then, managers can evaluate these solutions from
a broader watershed management point of view. Finally, research findings were

presented to a team of experts (focus group) for evaluation.

3.1.2 The AGNPS Model

The Agricultural Non-Point Source (AGNPS version 5) pollution model is a
single storm event based model. It divides a watershed into small grids. Each grid has a
set of variables to represent the topography, soil, channel, land use/cover, and field
management characteristics of a grid. These variables are described in detail in the Data
Collection section (section 3.2).

The AGNPS model has four modules: hydrology, sediment, nutrient, and
pesticide. The hydrology module is based on the SCS TR55 method. AGNPS uses TR55
to calculate the overland runoff and peak flow for each grid first. Then, a routing
algorithm, based on a grid width and path factor (flow direction), is applied to calculate
the accumulated effect. Strictly speaking, the AGNPS model is not a true distributed

model. The sediment module uses the Universal Soil Loss Equation (U SLE) to calculate

58
the upland erosion for each grid. AGNPS model added a slope shape factor to adjust the

USLE result. The sediment is further broken down by particle size based on the soil
texture information. Then, overland information and a routing algorithm are incorporated
to calculate the transport capacities and sediment discharge for each particle size.
AGNPS model divides the nutrient (nitrogen and phosphorus) module into two parts:
water soluble and sediment attached. For each grid, the “within grid” water soluble
nutrients are calculated based on the CREAMS model. This “within grid” nutrient
amount for each grid was added gradually through the flow path to give the total amount
of nutrient for a grid. The total amount decays as it runs through the channel to
determine the amount accumulated at the grid outlet. Sediment attached nutrient yields in
the AGNPS model can be calculated based on the following formula:
total nutrient yield =
overland nutrient yield + gully nutrient yield
overland nutrient yield =
soil nutrient concentration * sediment yield * nutrient enrichment ratio
gully nutrient yield =
gully nutrient concentration * gully sediment * nutrient enrichment ratio
(where nutrient is either nitrogen or phosphorus)
Nutrient enrichment ratio is affected by nutrient coefficient and soil texture. The finer
the particle size is, the greater the enrichment ratio.
Outputs from the AGNPS model contain both summary information at the
watershed outlet and detailed information for each grid. They are grouped into the same

categories as the four modules described above. Major outputs used in this research are

59

runoff volume, sediment yield, and nutrient (nitrogen and phosphorus) concentration.
These results will serve as the basis for scenario analysis and optimization.

There is also a new component, source accounting, in the AGNPS model version
5. The source accounting program allows users to evaluate one output variable for one
grid at a time, i.e., sediment yield of grid number 5. It can report the contributed amount
and percent of pollutant for each upstream grid. This new feature is very useful in
identifying the pollutant sources of a particular problem anywhere within the watershed.
It provides both “what and where” information in the optimization process discussed in

the next section.

3.1.3 Scenario Analysis and Optimization

A watershed based approach for non-point source pollution issues can also be
analyzed separately from a terrestrial (land) and aquatic (water) systematic view. The
terrestrial system represents the landscape diversity shaped by nature and human
activities. As water is the major agent that links these two systems together, the aquatic
system reflects the results of human activities on the terrestrial system. In non-point
source pollution management, one addresses the seriousness of the problem in the aquatic
system and searches for the sources of the problem in the terrestrial system. In other
words, land is “where the causes are” and water is “where the results will be”. With this
land and water causal relationship in mind, an optimization routine for scenario analysis
can be built.

Scenario analysis investigates potential solutions. Optimization searches for

“best” solutions within feasible solutions. Both are common techniques used for

60

evaluating a set of solutions. However, optimization focuses on search, and scenario
analysis emphasizes the comparison of solutions based on the causal-effect relationship.
They complement each other in this study.

Combining these two techniques, modeling and optimization, first requires the
creation of alternatives. In this research, an alternative is a combination of the land use
and management practices of all fields in the watershed. The current land use and
management condition is one default alternative. By changing the characteristics of a
field, i.e., from conventional tillage to conservation tillage, it becomes another alternative
but it might not be a better one. In a standard scenario analysis procedure, we will use
our best knowledge to play the “what if . . . then...” game. For example, we can
hypothetically change all conventional tillage fields into conservation tillage fields and
compare the before and after results. The analysis is done per scenario. However, the
biggest challenge is not running out of scenarios, but how to compare the results.
Another potential problem is that there might be conflicting interests in the results when
multiple objectives are involved.

The optimization technique takes a little different approach. It also tries to create
a set of alternatives first, but generates them randomly instead of based on the known
causal-effect logic. There are two key factors, decision variables and Objective functions,
in the optimization process. Decision variables should meet certain constraints, explicitly
or implicitly, so that solutions will be in the feasible range. This means that fields are
randomly assigned with different land use and management practices as an alternative.
This land use and management part corresponds to the terrestrial system within the land

and water context. Objective functions represent the goal or goals that we wish to

61

achieve. For example, one goal can be that the streams are swimmable and another can
be to restore the fish habitat in the rivers and lakes. These types of goals usually can be
converted into a more measurable factor, i.e., the nitrate or phosphorus concentration is
less than a specific amount. This measure part corresponds to the aquatic system in the
land and water context.

Once a set of alternatives is generated, the AGNPS model is used to compute the
results for different scenarios. A scenario contains a set of land use and management
practices for all fields within a watershed. One scenario can map to one set of AGNPS
model results. In the optimization terminology, the former one (land use and field
management) is called the “variable space” and the latter one (AGNPS results) is the
“function space”. The next step is to search for the “best” alternatives, that is, trying to
reach the goals. As mentioned in the previous section, objective functions are equivalent
to goals. When there are multiple criteria to be considered, the goals have to be
normalized so that they can be compared with other. Many searching techniques have
been developed in the Optimization theory. Most techniques focus on minimizing the
amount of search time needed to reach the extremes. This study uses intuition in the
scenario analysis, i.e., the known causal-effect logic, as the searching mechanism to find
the “best” solutions. The optimization routine used a Min-Max approach to derive a set
of Pareto optimal solutions.

The Min-Max approach is to minimize the maximum deviation of a goal. It is
developed specifically for multi-criteria analysis. In the multi-criteria optimization
processes, there are multiple goals that need to be compared. In order to compare one

goal with another, these goals have to be on a common basis. This is usually done

62
through normalization. Optimization iteration process begins with finding the worst goal

value (maximum deviation) among all normalized goals, applies certain rules to change
decision variables, and computes new objective functions. Theoretically, the iteration
process will repeat until all goals are reached simultaneously. However, this may not be
possible due to conflicting interests among goals. Users have to monitor the iteration
process and make a “stOp” decision when necessary. The final result is a set of Pareto
optimal solutions, which are equally important or effective among all criteria.

Managers will now examine these solutions and evaluate them in a broader
watershed context. Although each Pareto optimal solution contains the level of goals
reached and the corresponding land use and management decision variables, they have to
be used in a relative comparison manner. Managers should also look into the recurrence,
magnitude, and order of a particular pollution problem to understand the trends and
directions. Further analysis can investigate the generalized relationship between problem
fields and their landscape characteristics. Most important of all, these results will serve
as a solid base for further evaluation on the social and economic aspects of watershed

management.

3.2 Study Area
3.2.1 Geographical Setting

The study area is a sub-watershed of the Sycamore Creek watershed in lngharn
County located at south-central lower Michigan (Figure 3-1). The Sycamore Creek

watershed is part of the Grand River watershed, which flows from central Michigan

63
westward into Lake Michigan. The Sycamore Creek flows northward into Red Cedar River

at City Lansing and the Red Cedar River flows northwestward into the Grand River.

The drainage area of the Sycamore Creek watershed is approximately 68,000 acres.
It is located at the west center of Ingharrr County. The northern half of the Sycamore Creek
watershed covers parts of the cities of Holt and Lansing. It has a mixed land use pattern of
mainly urban and agriculture. The southern half of the watershed is primarily farmland.
The land use/cover distribution is as in Table 3-1 (SCS, 1990).

Table 3-1 Land Use/Cover Distribution of the Creek Watershed

Land Use/Cover
and

     

Residential
Forest
Idle
Commercial and industrial
Wetlands
and utilities

land recreation and
Gravel and wells
Water and
Other
otal

The geology of the watershed consists of mainly till plains, moraines, and eskers.
There are also organic bogs in the depression areas and drainage ways. The soil textures are
loamy sand, sand loam, loam, and muck, which are well, somewhat poorly, and poorly
drained. The slopes for most the watershed are flat with hills occurring in association with
the eskers. Studies have indicated that the esker and associated loamy sand and sandy loam

soil areas are major groundwater recharge areas for the 355,000 people in Ingham County

64
(SCS, 1991). Both Tri-County Regional Planning Commission and the Ingham County

Health Department have designated these areas as groundwater recharge areas.

The selected sub-watershed is located at the southwest part of the Sycamore Creek
watershed. The outlet of the sub-watershed is at the north end of City Mason. The sub—
watershed includes the main stream of Sycamore Creek and a small tributary, Willow
Creek. There are a few artificial drainage systems within the sub—watershed. In general, the
water system (Figure 3-2) of the sub-watershed is not very complex.

The US. Highway 127 passes from south to north through the center of the sub-
watershed. Most county roads follow the one-mile square section lines with shallow or no
ditches. The major land use of the sub-watershed is agriculture. Based on the Michigan
Resource Information System (MIRIS) and field data obtained from NRCS, the land
use/cover conditions of the sub-watershed are summarized in Table 3-2 and mapped in

Figure 3-3.

Table 3.2 Land Use/Cover Distribution of the Area
Land Use/Cover

 

From the US. Geological Survey (USGS) l:24,000 scale Digital Elevation Model

(DEM), the topography of the selected sub-watershed is also relatively level. The

65

maximum and minimum elevations of the sub-watershed are 256 and 323 meter
respectively. Figure 3-4 presents the elevation map of the sub-watershed in gray tones.
The lighter color areas represent highland areas and the darker color areas are lowland
areas. From the elevation map, one can derive an aspect map (Figure 3-5), representing
the direction of the slope in one of the 365 degree directions (relative to east). The sub-
watershed boundary (outlined in dashed white color) and the stream network (in thick
white color) were overlaid to show how well the aspect map reflects the topography from
visual perspective. This elevation layer is one of the key information for deriving the
drainage pattern in this study.

Another by-product derived from the elevation map is slope. Slopes of the sub-
watershed range from 0 to 25 percent and the average slope (weighted by area) is 2.547
percent. Figure 3-6 showed the slope map of the sub-watershed in gray tones. The
lighter color areas are mostly flat or depression areas and the darker color areas represent
steep slope areas. The slope-acreage distribution of the sub—watershed is summarized in
Table 3-3. Approximately 80 percent of the study area have slope less or equal than 3
percent. The slope map also showed higher slope values along some stream segments Of
the main Sycamore Creek.

The soil of the sub-watershed has approximately 40 major soil map units plus
urban, water, and gravel pits. From the soils map (Figure 3-7), the diversity of soil types
are not evenly distributed across the sub-watershed. The southern portion is more
homogeneous than other areas. Most soils are silt and clay plus a few patches of peat.
These finer soils are usually poorly drained soils, which require tiled drainage system if

used as row crOp agricultural land.

66
Figure 3-1 The Study Area - A Sub-watershed of the Sycamore Creek Watershed

Ingham County
in Michigan

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Mm ”P It.--“

1 ix
Sycamore Creek Watershed 12 =

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

inlngham County g —WT "M-
_ 1L 2.. l —
m ll
i
i I
StudyArea(subwatershed)

in Sycamore Creek Watershed

 

67

Figure 3-2 The Water System oftlre Study Area

I!

D'-
\ .0-
IIIIIIII

D
\D
\\
’D

9"
'O
- -
I
I
I
O
o

.--.------------------T

" ‘DEI'I-

 

Streams and Rivers

Lakes and Ponds
. Intermittent Streams and Drains

/\/ Subwatershed Boundary

N
' \
O
’ \I
I U

68

Figure 3-3 The Land Use/Cover of the Study Area

 

land Use/Cover

Agriculture cropland, orchards, and pasture
Residential Forest

[:1 Forest

  

 

-Transportation.communications.endutililies
-0penland(recreelienandcemetery)
-smeipitsmduens
-Weter(streamsandlakes)

Earlier

69

Figure 34 Digital Elevation Model (DEM) Of the Study Area

    

Digital Elevation Model (DEM - in Meters)
- 256 - 260
- 261 - 265
- 266 - 270
- 271 - 275
- 276 - 280
281 - 285
= 286 - 290
291 - 295
296 - 300
301 - 305
306 - 310
311 - 315
316 - 320
321 - 325

70

Figure 3-5 Aspect, Watershed, and Stream Network of the Study Area

. ’ r
A‘ .
1

rats-.- i: as 7‘“??? .. .r
._ «A .. it .5

wit; )3. ‘5.

:1} : .. . 4". l'...
r: din-3*»! ' if. Marti-l ’
.. --v,.‘ . l, - ~
. ‘ .AA ,1. ‘ ,I. . _ .
r'r'. l 9. g .r.,r

'3'. q. ‘

' . ‘.~ :33}. may,
._ e. 1‘. -. ,3, If, f’
.‘ ~ kg, 2. _5

he: .1. - .
r z ;JIWUP
fig}; if V
fish, ‘3

..;

 

71

Table 3-3 Distribution of the Area

Area acres
2131 .608
3944.531
8134.804
331 1 .165
950.360
405.950
565.760
410.310
269.314
100.298
1 1 0.083
41 .142
.582

25
8.673
9.785
1 6 4.003
1 7 5.1 15
18 3.336
1 9 rcent 2.446
20 1 .334
21 0.222
22 0.667
23 0.890
4

(OQNODOI-th-‘O

O.

 

72

Figure 3-6 Slope ofthe StudyArea

 

73

Figure 3-7 Soils ofthe StudyArea

 

74
3.2.2 Watershed Baseline Conditions

The Sycamore Creek watershed has been selected as one of the two Hydrologic Unit
Area Projects in Michigan under the US. Department of Agriculture (USDA) Water Quality
program. The objective of the five-year water quality program is to promote the adoption of
best management practices to reduce the potential for non-point source pollution of our
water resources from agricultural production.

Several problems have been identified in the Sycamore Creek watershed. In the

1991 progress report (SCS, 1991) for the water quality program by Soil Conservation
Service (SCS), now renamed as Natural Resources Conservation Service (NCRS), it stated
that:

“The major types of pollutants to be controlled are sediment from soil erosion,

phosphorus fertilizers, nitrate fertilizers, and agricultural pesticides. These

pollutants cause sedimentation and turbidity problems, nuisance alga growth, and
groundwater contamination. There are approximately 1,900 acres of cropland that
have a very severe erosion problem with soil losses exceeding 12 tons per acre per

year. There are approximately 13,000 acres of cropland with soil losses of 8 to 10

tons per acre per year. Critical soil erosion areas deliver sediment, nutrients, and

pesticides, which cause water impairment in the stream.”

Several monitoring and sampling projects have been conducted in the beginning of
the water quality program. These early studies showed that the main problem for the
Sycamore Creek watershed is sediment fiom soil erosion. Further analysis also identified
the potential sources. These have been documented in early progress reports (SCS, 1991)

that:

75
“Seven of the eight sites monitored did not meet the Dissolved Oxygen (DO)

standards of 5 mg/l. Severe sedimentation in the stream channel has occurred
throughout the watershed. Much of the loose substrate is organic and exerts a high
oxygen demand.”

“. .. sediment demand is the major consumer of oxygen in the stream. Sediment
sources to the stream include storm water, construction site erosion, agricultural
land, and stream banks.”

“Analysis of water samples from the stream indicates that the concentration of
suspended solids, total phosphorus, and nitrate increase dramatically during runoff

events. This is due primarily to runoff from agricultural lan .”

All of these preliminary findings concluded that:

3.3

3.3.1

(1) sediment is the major problem;

(2) the primary sediment sources are from soil erosion, such as storm water,
construction sites, agricultural land, and stream banks;

(3) the magnitude of the problems increases dramatically during runoff events;
and

(4) the sediment problem has diminished the suitability of the stream for

recreation and other purposes.

Data Collection
Base Data and Model Data

There is a rich set of base data available for the Sycamore Creek watershed. Data

used in this study are either from existing sources or derived from references and

76
calculation. These data can be classified into two categories: (1) base data and (2) model

data. The base data often are available in database and/or map format. They are usually
called primary data. In contrast, most model data require further manipulation on base
data. These derived data can be called secondary data.

This study combined the GIS (GRASS) and database (PostgreSQL) technologies
to facilitate model data computation. GRASS was developed by US. Army Corp
Engineer and SCS. It is a raster based GIS with limited vector data processing
capabilities. PostgreSQL is a relational database developed, by the University of
California at Berkeley. It has most standard SQL (Structured Query Language)
functions. GRASS provides a set of functions to manipulate spatial data in map format.
PostgreSQL stored all soil, crop, and field base data. It is also used to host both model
input and output. These two packages were combined together in scripting environment
in a UNIX operating system. The optimization procedures were implemented in the
scripting language, which can retrieve data from the PostgreSQL database. Model
outputs and optimization results were presented in the GRASS GIS for further analysis.
These tools are applied together to manipulate base data to derive mOdel data (both input
and output), perform optimization problem solving, and final analysis.

Both data types, base (primary) and model (secondary), are classified into five
categories for discussion purposes. These categories are: (1) watershed, (2) topography,
(3) channel, (4) soils, and (5) land use/cover and field management. There are a total of
22 parameters needed for each grid/cell in the AGNPS model. These input data were

derived through different procedures and methods. Table 3-4 shows summaries of the

77

methods used to determine each AGNPS model variable. The corresponding procedures

are described in the following sections based on each category.

Table 3-4 Summaries of AGNPS Model Variable Derivation

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

AGNPS Model Variables Method Fixed or Changeable
Cell number Program Fixed
Receiving cell number Program and manual Fixed
SCS curve number Area-weighted average Chameable
Land slope Area-weighted average Fixed
Slope shape factor Default (uniform for all) Fixed
Field slope length Area-weighted average Fixed
Channel slope Area-weighted average Fixed
Channel side slope Area-weighted average Fixed
Manning’s roughness coefficient Area-weighted average Changeable
Soil erodibility factor Area-weighted average Fixed
Cover and management factor Area-weighted average Changeable
Support practice factor Default (1.0 for all) Fixed
Surface condition constant Area-weighted average Changeable
Soil texture Area-value dominant Fixed
Fertilization level (N and P amount) Area-weighted average Chameable
Fertilizer availability factor Area-weighted average Changeable
Point source indicator Default (0 for all) Fixed
Gully source level Default (0 for all) Fixed
Chemical Oxygen Demand (COD) Area-weighted average Chagnqeable
lmpoundment factor Default (0 for all) Fixed
Channel indicator Manual Fixed
Soil hydrological mup" Area-value dominant Changeable

 

 

number.

1

" This variable is not an AGNPS model variable but is used indirectly for calculating SCS curve

 

3.3.2 Watershed

The Sycamore Creek watershed boundary, obtained from NRCS, has been divided

into three small sub-watersheds. One is delineated based on the upstream Sycamore

Creek and Willow Creek, which is the selected study area. The cast and southeast

portions of the Sycamore Creek watershed form another sub-watershed based on the Mud

Creek. The third sub-watershed is the downstream portion of the Sycamore Creek

watershed.

 

78
To run the AGNPS model, first requires that the watershed be divided into small

square grids. A watershed can have only one outlet grid. Theoretically, the size of the
AGNPS watershed grids should be as small as possible to obtain better simulation results.
In reality, the grid size depends on two factors: (1) the computer capacity and (2) the
homogeneity of the landscape. The physical capacity of a computer, i.e., the amount of
RAM that a computer has, sets the upper limit of the total number of grids that a
watershed can have. If the watershed is very large, users have to increase the amount of
RAM in order to have smaller grid size setting. Otherwise, a larger AGNPS model grid
has to be used.

Landscape homogeneity also affects the decision of the grid size. In the AGNPS
model, each grid requires a set of variables to represent different landscape characteristics
of a grid. The ideal situation is that a grid has only one land use/cover type. However,
this rarely happens in the real world situation. When there are multiple land use/cover
types in a grid, aggregation processing has to be performed. Two types of rules are used
for the aggregation processing. If a variable represents the rating or category of certain
property, the area-value dominant rule is used. If a variable can be measured
quantitatively, the area-weighted average method is applied. An area-value dominant
rule determines the aggregated value by comparing the areas that have the same rating or
category. The rating or category with the largest areas within an AGNPS model grid is
the representative value for the grid. The area-weighted average method uses the

following formula to calculate the representative value for an AGNPS model grid:

area-weighted average value = 2 x, * A, /A,

where

79
i: the number of types in an AGNPS grid

x.-: the value of a variable for type i
Ag: the area of the corresponding variable x.- in the AGNPS grid
A,: the total area of an AGNPS grid

The grid size used in this study is approximately 40 acres. Since this research
used a GIS and database approach to facilitate model data derivation, there is an inherited
limitation on the minimum grid size. It is the base map resolution. The AGNPS model
grid size should not be less than the minimum map resolution. All maps used in this
study have the same resolution, 30 meters, which is equivalent to 0.222 acres. The major
factor for the 40-acre grid size decision is due to extensive computer computation time
needed in the optimization process. This will be explained in later section. The actual
AGNPS model grid size used is 37.58 acres, which is equivalent to 13 by 13 (169) 30-
meter small map grids. If an exact 40-acre AGNPS model grid is used, it will result in
aggregation of less than one map grid situation. This will likely reduce the accuracy of
the final representative variable values and increase the computation complexity of the
aggregation process.

Once the grid size is determined, an algorithm has to be written to process the
sub-watershed map to derive an AGNPS watershed grid map (Figure 3-8). Most
watershed grids are within the sub-watershed except those at the sub-watershed
boundaries. These grids are partial within the sub-watershed. A 50% dominant rule was
used to determine whether these border grids should be included or not. If at least 50%
of the areas of an AGNPS model grid are within the sub-watershed boundary, the grid is

included. Finally, each grid was assigned a number starting from northwest, eastward

80
and southward, to southeast. There is a total of 594 AGNPS model grids for the study

area.

3.3.3 Topography

The base topographic information is elevation. There are two major sources of
elevation: (1) Digital Elevation Model (DEM) maps and (2) contour maps. Both are
available from US. Geological Survey (USGS). The DEM maps are available in 30-
meter grids (rasters) in digital format by USGS 71/2 minutes topographic quadrangles.
Each 30-meter square grid has an elevation value (in meter) and the minimum elevation
difference between grids is one meter. Contour maps (hard c0pies) are also available by
USGS 71/2 minutes topographic quadrangles. The contour interval of contour maps
varied from 10 to 20 feet depending on the complexity of the topography in each
quadrangle. There are four quadrangles that cover the selected sub-watershed. These
elevation layers are used to derive aspect, receiving cell number, and slope related
variables of the AGNPS model.

The meaning of the aspect variable in the AGNPS model is actually the same as
the receiving cell number. The aspect of a grid is the direction of water flow in one of the
eight (1 through 8) directions plus 0 for depression. The receiving cell number is the grid
number to which the water of a grid drains. If the aspect of an AGNPS model grid is
known, the receiving cell number can be derived easily and vise verses. By using the
“r.watershed” program in GRASS, it takes an elevation map as input and gives results of
aspect (water flow direction) for each grid. Since the resolution of the elevation (30

meters) is much smaller than the AGNPS model grid (390 meters), the elevation map was

81

Figure 3-8 The AGNPS Model Grid Layout

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

«a. .m
”mumaaanum
”mumuaammm:
] mummmma.amm
m mumammmuaaamu
mmmmmammmmnmmaamammm.
mammmmumaanmmmaa.mmma..
.mmmmmmmmmmmmmummamamaum.
.a.mmmmmmmmmmmmuumaamamuwmwm
.2.mmmmmmmmnumumumaaaammmumm.
no“.mmmmm.mmmmmmmummumammmmumuum
unu.u.mm.u.mmamummmmnuamamamawmmmmam
"an...“mmmmmmmammmmmuaaaaaumamummmu
«a.«a..mmmm.mu.mmummmmamaummmmmamm
.na.«n.mmmmummmammmmmaam.nmmmmmmam
.ua..nnmmmm.mun.mmmm.¢am..mmu~mn
,nuaanammmmumuuammumaaaamu
.«uauuummmmummmwummummu.am
.uuagnummmmmmnmaamamaa.
«uuuu.mmmm.mmn.mmmmmaa
.unnan.mmmmmmmmaummmaaa:
.unummmmmmmmmammmama
unn.mmmmmmm.ummmmamu
unnummmmmmmmmummmmm
u...mmmmmmm...um
unummm mmmnmmm
mm
mm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Hm

ma;

mum

82
aggregated into the AGNPS model grid size first by the area-weighted average method.

The aggregated map was input to the r.watershed program and the results were plotted on
the screen for further examination.

Due to the aggregated process, the derived aspect map was not very accurate.
Human intervention is necessary to correct the drainage path with the aid of the contour
maps. Once the drainage path of each grid is defined, the aspect and receiving cell
number can be computed.

The slope information is also calculated from the elevation map by the GRASS
“r.slope.aspect” program. The area-weighted average method, then, was applied to
aggregate the slope map for the AGNPS model grids. There are two other slope related
variables, slope shape and field slope length. All AGNPS grids were given the same
value of slope shape, i.e., l for uniform. Field slope length is one of the USLE
parameters. It usually varies with the steepness of the land slope. The rule of thumb for
this factor is the steeper the slope the shorter the field slope length should be. Tables 3-5

showed the field s10pe length for different slope range used in this research.

Table 3—5 The Field Slope Legth Based on the Land Slope

 

 

 

 

 

 

Slope Range Field SIOpe Lem (feet)
0 to 2 percent 300
3 to 6 percent 250
7 to 12 percent 200

> 12 percent 150

 

 

 

All the above five variables are treated as fixed variables that are not likely to be

changed by human activities in this study.

83
3.3.4 Channel

Channel related AGNPS model variables include: (1) channel slope, (2) channel
side slope, and (3) channel indicator. Based on the channel types, streams, drains, or
intermittent streams, each channel map cell was given a value for each channel variables.
Then, they are aggregated into the AGNPS model grid. These variables are all fixed
variables that are not usually related to human activities.

Channel slope was first derived from the elevation map with the GRASS
“r.slope.aspect” program. Only channel map cells were used in the aggregation process.
The rule used is the area-weighted average method. According to the AGNPS User’s
Guide, all non-channel AGNPS grids were given a value of half of the grid land slope,
assuming a series of small channels within the grid. Water or marsh has a default value
of 0. Different channel side slope values were assigned to different types of water system
and aggregated into AGNPS model grids by the area-weighted average method. Non-
channel AGNPS model grids all have a default value of 10 percent based on the AGNPS
User’s Guide. If an AGNPS model grid is mainly water or marsh, it has a value of 0.
The last variable, channel indicator was not really used in the AGNPS model. It is

simply identification of whether a grid is a channel grid or non-channel grid.

3.3.5 Soils

There are two soil related variables in the AGNPS model: soil texture and soil
erodibility. All of these data were based on the NRCS Soil Survey Geographical
(SSURGO) database. SSURGO database is a complex relational database. There are

many variables in the SSURGO database. The basic unit in the SSURGO is "map unit”,

84

which represents a soil type. Each map unit may have several soil series (components).
Each component can have up to 9 layers of soil profile. The soil texture refers to the
particle size of soils. Based on the percentage of three major soil texture classes, clay,
silt, and sand, soil texture can be further divided intolZ sub-categories and each sub-
category can have a modifier to represent subtle differences. However, the AGNPS
model only requires information at the major soil texture classes plus peat and water.

The soil texture classes were first extracted and generalized into major soil texture
class from the “layer” table in the SSURGO database for each soil series. Only the top
layer information was used. Then, the soil texture class was calculated for each map unit
by the area-value dominant rule from its components. Finally, the same area-value
dominant rule is used to determine the soil texture value for the AGNPS model grid.

Soil erodibility is another USLE variables, K factor. It represents the potential of
the soils that can be eroded away. Its value should range from O to 1. It is an interpreted
variable. This variable is also available in the “layer" table of the SSURGO database.
The calculation and aggregation process is similar to the soil texture but the area-
weighted average method was used.

In addition to these two variables, there is another soil variable, soil hydrological
group, that is required by the AGNPS model indirectly. Soil hydrological group is a
rating of the drainage condition of the soil. There are four classes (A to D) where A
representing easily drained soils and D being difficult to drain soils. The soil
hydrological group determines the runoff curve number, which is one of the major

variables of the SCS TR-55 hydrological model. The soil hydrological group can be

85
found in the “comp” table of the SSURGO database. The aggregation process is the

same as the process used for soil texture, i.e., area-value dominant rule.
These soil related variables and all variables discussed so far are fixed variables.
They only need to be calculated once for each grid. It is assumed that neither human
activities nor any other natural factors will change their values. The next section outlines

a set of changeable variables affected by agricultural activities.

3.3.6 Land Use/Cover and Field Management

The land use/cover variables in the AGNPS model are those affected by human
activities, specifically agriculture in this research. Field management variables are those
controlled by farmers. The land use/cover variables are results of the field management
variables. These variables are:

l. SCS Curve Number (CN)

2. Manning’s Roughness Coefficient

3. Cover and Management Factor (C factor in the USLE)

4. Support Practice Factor (P factor in the USLE)

5. Surface Condition Constant

6. Fertilization Level

7. Fertilizer Availability

8. Chemical Oxygen Demand (COD) Factor

These variables are based on a field database and a field map obtained from
NRCS Michigan State Office. The field database contains detailed information on

applied crop rotation of each field in the study area. Each field was given a land

86

use/cover identification code, which represents a four-year crop rotation management
operation. By combining the field database with the field map, one is able to process the
data in the AGNPS model. Figure 3-9 showed the field map classified by the land
use/cover and field management code. If one overlays the AGNPS model grid layout
(Figure 3-8) on top of the field boundary map, it is clear that almost none of the AGNPS
model grids contain only one field. There are always several fields within an AGNPS
model grid. The aggregation process has to consider the spatial aggregation and data
manipulation. This explains why the GIS and database techniques are so helpful in the
aggregation and manipulation processes.

There are 24 land use/cover and field management identification codes used in
this research. Eight of them are specific for crop land (COIDOl to COIDO8) and the
others are for non-crop land (NCIDOl to NCID18). The eight crop specific codes can be
further grouped into three categories based on the tillage method plus the grass or alfalfa-
hay land (COIDO8):

l. COIDOl to COIDO4 are cropland using conventional tillage method;

2. COIDOS and COID06 are cropland using conservation tillage method;

3. COIDO7 is cropland applying no till method.

For each crop specific code, there is a planted crop for each year of a four-year rotation.
Within each year, there is information on three crop stages (stage 1 to 3) plus one (stage
0) before any operation occurred. These three stages usually match with the fallow,
seedbed, and establishment stages. Appendix A contains detailed information on these

eight crop rotation codes.

87

Figure 3-9 Field Map by Land Use/Cover and Field Management Identification Code

  
  
  
    

  
   
 
    

   
  
  

 
 
 
      
    
 
 

\

l.
“in
egg

  

Militia)

L...“ will}: ._
“will; 3- M; .
: - 5:121; : .-

       

   
  
    

      
 

I
\ IQ °

    

  
 

oorooz
001003
9.3 coroo4
338: “”005
m minors
E coroor
mince

NCIDOl
\\\ "0002
m NODO3
"0°“
uaoos
"0°05

 

noon

m NCIDIZ

Mama

NClDH
NCIDIS
Home
Harm
Hams

88

In Figure 3-9, each field or land use/cover type also has a unique number to serve

as the spatial key to link to the field database. The field database contains the same

spatial key, land use/cover and field management identification code, and the number of

year offset in the rotation system. This basic initial condition for each field can be found

in Appendix B.

There is another database that contains land use/cover and field management data

for each land use/cover and field management identification code (Appendix C). This

database pre-defined all variable values used for all land use/cover codes. In this way,

when a field changes its crop Operation, i.e., from conventional tillage to conservation

tillage, related land use/cover and field management variables could be quickly retrieved

for the AGNPS model. Figure 3-10 diagrarned the schema built for database and map

linkages.

Figure 3-10 The Schema of the Linkage between Land Use/Cover and Field

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Management Database and Map
Land Use/Cover and
Management Database
Field Map With Spatial L 1 ID
u C
Field 38 H————' Field ID Year Offset
Field 20 Lulc ID Cm? 338°
Year Variable 1
Field 44 Variable 2
S atial to .
Field 22 Database “nab“ 3

 

 

 

 

 

 

 

 

 

 

89
The key field management variables in the AGNPS model are cover and

management, supporting practices, and the amount of fertilizer applied. Cover and
management factor usually refers to the residues part of a tillage method. Each field will
have a cover and management variable value extracted from Table 3-7. The area-
weighted average method is then applied to derive an aggregated value for each AGNPS
model grid. The fertilizer application level is set to 4, i.e., user input, for all AGNPS
model grids. Then, the amounts of nitrogen (N) and phosphorus (P) applied were
extracted from the Table 3-7. The same aggregation process, area-weighted average, is
used for aggregation.

Supporting practice factor is more specific to applied planting method, such as
contouring or terrace system. This research only considers the cover and management
factor. The supporting practices were all given a default value of 1. There are two major
reasons. First, the study area is really relative level. Contouring or terrace system was
rarely considered as a necessary alternative in the study area. Secondly, this study will
use only “big” storm events for analysis. When a big storm event occurred, it is more
likely that detached soils will be eroded away across the study area. Since the study area
is fairly flat in general, it is assumed that the supporting practice is all the same across the
watershed. This may cause the AGNPS model to produce higher soil erosion and
sediment yields.

The other set of variables is hydrological related variables, which are related to
field management variables. Manning’s roughness coefficient, surface condition
constants, and runoff curve number are all key factors for the hydrological module in the

AGNPS model. They represent the land cover characteristics when the storm event

90
occurred. For the first two variables, the processing is exactly the same as the cover and

management variable. Field data were extracted from the look-up table (Table 3-7) and
the area-weighted average method is used to aggregate field data into the AGNPS grids.
Runoff curve number (RCN) is different in that the RCN variable has to take the soil
hydrological group of the underlying soil into account. Each land cover condition will
have a different RCN value for each hydrological soil group. The GIS was used to
spatially combine both field and soil hydrological group to compute the RCN value for
each map cells first. Then, the same aggregation process, area-weighted average, can be

applied.
The last two variables are the fertilizer availability and COD (Chemical Oxygen

Demand) factors. Fertilizer availability is based on the plowing method of tillage
systems and the COD factor is based on land use situations. The AGNPS model User’s
Guide contains a look-up table of the corresponding values for both variables. Again,
each field will obtain a value from Table 3-7 and the area-weighted average method is
used for the aggregation process. There are three other variables, gully source level,

Point source factor, and impoundment factor, which were not considered. All AGNPS

model grids have 0 for these variables.
In the AGNPS model version 5, users can also change several default values, such

as base N and P in the soil, the amount of N and P extracted (leaching) out of the soil, and

buffer area condition, etc. For all of these variables, the research uses the default values

Provided by the AGNPS model.

91

3.4 Running the AGNPS Model
Once all model input variables are set, the data can be quickly assembled, and the

model ready to run. However, users must first provide the storm event information. The
basic data needed by the AGNPS model are the type of storm, the amount of
Pmcipitation, the duration of rainfall, and the rainfall intensity. Since the AGNPS model
uses the SCS TR-55 and USLE models, many references and look-up tables or charts can
be found to set the value. Moreover, non-point source pollution occurred only when

“big” storm event happened, this study chose a 24-hour, 25-year frequency storm event

for all runs.
The AGNPS model is not capable of evaluating long-term changes. It focuses on

a single storm event. In one way, it is acceptable because the non-point source pollution
is stochastic. However, long-term effects and accumulated results will not be revealed
With the single storm event AGNPS model. In order to compensate for lack of temporal
Consideration in the model, the AGNPS model will run on three crop stages for all four
years in the later optimization and scenario analysis processes. This will be explained
move in the next section.
The AGNPS model produces at least an equal or greater amount of output

information to be processed or examined. Only selected output results related to
seclitnent and nutrient were stored into a database. Post-processing was performed to

131‘er for Optimization. This processing is discussed in the next section also.

 

92
3.5 Scenario Analysis and Optimization

3.5.1 The Framework
In order to use the optimization technique, one must know the objective functions
which will be used for comparison. By changing decision variables, alternatives can be
generated. Each new alternative will be compared with old or existing ones. This is a
iterative searching process that gradually reaches or approaches the defined goals in the
objective functions. This is the underlying framework logic of the optinrization.
Figure 3-11 showed a simple optimization framework diagram for the non-point
source pollution problem solving. There are three components plus the optimization
procedure itself. The three components are: objective functions, decision variables, and
simulation model. The simulation model has been discussed in previous section (Section
3 - l -2). The following sections will focus on the other two components, goals and

decision variables, and their roles and relationship in the optimization procedures.

35.2 Objective Functions and AGNPS Model Results

The first step of optimization is to determine what it will be searching for or
comparing with, i.e., the objective functions. In the beginning of this chapter, the
oVet'View section (3.1) covered the rationale of using the optimization technique. This
Section will explain the details.

Non-point source pollution problems can only be examined by what will result in
the aquatic system -streams and lakes. Current water quality standards worldwide use
concentration of pollutants in the water as the measure. The ultimate goal is to meet the

Water quality standards. Among the three pollutants (N, P, and sediment) that this

 

93
Figure 3-11 The Optimization Framework for the NPS Pollution Problem

GOALS \

AQUATIC - STREAMS
(l) P - the total phosphorus in any stream
segment should be less than 1 ppm.
(2) N - the total nitrogen in any stream
segment should be less than 10 ppm.
(3) Sediment - the total sediment in any
stream segment should be less than

K 1000 mg/l. J

 

 

 

 

[Objective Functions]

 

 

 

 

 

 

 

 

 

 

[ Simulation Model] [ Decision Variables ]
K AGNPS \ [ FIELD \
Agricultural NPS Model CROP TYPE AND ROTATION

(l) A storm-event, watershed based (1) Each agricultural field was
water quality simulation model assigned a crop rotation type.
for agricultural non-point source (2) There are 8 crop rotation types
pollution. plus the idle agricultural land.

(2) Hydrology (TR55) - prediction of (3) Each rotation is a four-year span
runoff volume (inches) and peak (1992 - 1995) and each year has 3
flow (cfs). crop stages - fallow, seedbed, and

(3) Soil Erosion (USLE) - estimates establishment.
of upland erosion, channel (4) Each stage has a set of variables,
erosion, and sediment yeild. such as cover and management
(4) Nutrients (CREAMS) - factor, Manning’s roughness
measurements of phosphorus, coefficient, runoff curve number,
nutrients, and Chemical Oxygen fertilizer (N and P) applied, and

 

 

 

Demand in concentration and fertilization availability.
m... j x J

94
research is trying to investigate, EPA regulates only nitrogen. Current primary drinking

water quality standards specify that the Maximum Contaminant Level (MCL) of nitrate
concentration in tap water should not exceed 10 ppm. Sediments were partially regulated
by the secondary water quality standard of 5 NTU upper limit of turbidity. Phosphorus is
not regulated directly but it is known to be the primary cause for eutrophication in lakes.
There are many other water quality related standards set by state or local government for
other purposes, such as recreation, irrigation, industrial, and even wildlife. The Nation’s
water quality goal, as described in the Clean Water Act, not only emphasizes drinking
water goals, but also calls for swimmable and fishable purposes in the streams and lakes.

By examining other sources of water quality standards for different usage and
purposes, this study used the following standards as the ultimate goals:

(1) the total nitrogen in any stream segments should be less than 10 ppm;

(2) the total phosphorus in any stream segments should be less than 1 ppm; and

(3) the total sediment of any stream segments should be less than 1,000 mg/l.

The goal for nitrogen, 10ppm, is considered pretty high for surface water in the
watercourse. However, surface water may be extracted for drinking directly. These three
goals need to be embedded in the objective functions for optimization. Unfortunately,
objective functions quite often are not equivalent to the goal in the optimization problem
solving. When the purpose is to attain a certain measurable level, the goal will not be the

minimum or maximum value that the optimization is trying to reach. Instead, it is the
deviation from the goal that the optimization procedure will try to minimize. If there is
only a single goal, the optimization procedure only needs to consider the deviation from

the goal. However, when multiple criteria are involved, we have to evaluate multiple

95

goals on a common basis. This procedure is called normalization. The following

equation is used to compute the normalized objective function:

(the value of a problem - the goal of a problem)
the goal of a problem

 

normalized objective fimction =

In this way, each problem will be evaluated by the magnitude relative to its goal, so that,
all goals can be compared relative to the magnitude to the problems. Once multiple
criteria are at the same standard, several techniques can be applied for optimization. One
set of techniques use trade-off or weighting method to modify multiple criteria
optimization to be a single criteria optimization problem. Another set builds on the
concept of Pareto optimality, which remains a multi-criteria optimization solver.

In the non-point source pollution case, the goal statements defined above also
included the “any stream segments” clause to incorporate spatial consideration. This
means that each channel grid or segment will have three objective functions - one for
each pollutant. The number of objective functions will be in proportion to the complexity
of the stream network in a watershed. Moreover, the study also proposed to consider
three crop stages for a four-year crop rotation period to address the temporal issue. Both
spatial and temporal dimensions implied that there will be potentially many objective
functions, which suggests that some modification will be necessary in order to reduce the
complexity of the optimization process.

Another way to look at the non-point source pollution problems is to evaluate

them for a watershed as a whole for a certain time period. Instead of emphasizing the
results at each single grid or segment at a particular crop stage of a specific year, an
aggregated approach is used. However, the study still needs to address the spatial and

temporal issues if the aggregated approach is used.

 

96
Due to all of these pending difficulties and unresolved issues, it was decided that

two sets of objective functions be used for the optimization processes. First, a combined
objective function approach will be applied to investigate the Pareto optimal solutions.
This approach has the following combined objective functions:
1. the sum of all normalized objective functions for the same goal of all stream grids
and all three crop stages during a year (12 objective functions):

' i=goali(lt03)
objective flmctionfi = 2 goal where j = rotation year j (lto 4)
k k =channel gridk(ltok)

2. the sum of all normalized objective functions for the same goal for all stream
grids, all three crop stages, and all four crop rotation years (3 objective functions):

l I i = goali (l to 3)
objective fitnction, = 22 goal where j = rotation year j (l to 4)
k j k=channel gridk(ltok)

There will be 15 objective functions in total. These 15 combined objective functions are
used in the Pareto optimal solution comparison. They represent the best solution sets
found through the optimization operation. Since these are combined objective functions
for a watershed, the results should be used to interpret the non-point source pollution
problem at the watershed scale.
The second set of objective functions is the original set of uncombined objective
functions of each goal for each channel grid at a specific crop stage for a particular crop
rotation year:

i=goali(lt03)
j = rotation year (1 to 4)
k = channel grid k (1 to 183)
l = crop stagel (l to 3)

objective fitnctionw where

97
This means that each channel grid will have 36 objective functions to be evaluated.

There are 183 channel grids in the sub-watershed (594 grids in total). This summed up to
6,648 individual normalized objective functions. This set of objective functions still has
the spatial and temporal details and is used in the Min-Max optimization procedure as

discussed in the next section.

3.5.3 Decision Variables and the Min-Max Strategy

The second component of the framework (Figure 3-11) is decision variables.
Contrary to the objective function component that addresses problems in the aquatic
system, the decision variable component looks back to the terrestrial system to find which
areas in the watershed that contributes most to the pollution. The primary focuses are to
trace the type and source of a problem and the changes needed to diminish the problem.
Figure 3-12 presented a processing flow diagram to show the steps and task in each
procedure. Details are described in this section.

The first step of this analysis is to create a decision variable space. The current
land use conditions and management practices of all fields in the sub-watershed account
for only one point in the space. Theoretically, any field can have one of the eight types of
crop rotation. There are approximately 500 fields in the study area. There can be many
different combinations and each combination is one point in the decision variable space.

For one combination, it can result in a set of objective functions, which maps to one point
in the function space. In this study, it first randomly generated a set of combinations or

POints to form the decision variable space. Then, the corresponding objection functions

98

were calculated to form the function space. These two spaces constitute the basis in
which the Min-Max optimization procedure will be processed.

Once the initial function space has been constructed, the Min-Max optimization
procedure starts to compare the value of each normalized objective function. A
normalized objection function represents one type of problems or goals (N, P, or
sediment) of a channel grid at specific crop stage of a particular year. The task is to find
the maximum deviation value (of a goal) out of the 6,648 normalized objective functions.
Then, a strategy is needed to minimize that particular objective function. With this
approach, we can identify:

(1) what are the most serious pollution problem (in terms of the pollutant);

(2) where does it occur (in terms of a channel grid); and

(3) when does it happen (at a specific crop stage of a particular rotation year.

The purpose of using all normalized objective functions is to address the spatial
(183 channel grids) and temporal (3 crop stages for a 4-years crop rotation cycle)
considerations. Then, the next task is to find better solutions through change of the
decision variable, that is, change the management operation of the upstream fields.
Changing values of decision variables of a field is a three-step process once a channel
problem grid is identified as having the worst objective function. The first step is to
evaluate which upstream grid contributed most to the problem. The new source

accounting module of the AGNPS model is used to examine the source contribution from
all upstream grids. The result is an AGNPS model grid that contributes most to the
identified problem.

The second step is to evaluate which field in the worst grid can be changed and
how much improvement will the change achieve. Almost any AGNPS model grid

99
Figure 3-12 The Processing Flow Of the Optimization Procedure

 

F Initial Setup O
Randomly Create Initial
1 Combination of Crop Type

 

Base Data
Field Data and .' and Management

Crop/Management DataJ Operation as Points in
\Decision Variable SpaceJ

 

 

 

 

 

 

F Simulation Model \
Run the AGNPS Model on
Every Point
for Each Crop Stage of
Each Year to Derive
Output for Objective

r Pareto Optimality N \F‘mcuon Computation J r Apply Changes N
Compare Each Solution Apply the Alternative to
with Existing Pareto the Field by Changing
Optimal Solutions to Corresponding Land
Determine whether the use/cover and Field
Solution is a New Pareto Management Decision
Solution or not Variables
J , \ J
l

A

 

 

 

 

 

 

 

 

 

 

 

 

f Normalized' Objective ‘
Functions
Calculate 6,648 Objective
Functions for Each Channel Grids
K and 15 Aggregated Objective J

 

 

 

 

 

 

7

Maximum Deviation from Goal Minimize Deviation from Goal

Find the Worst Objection Function of Compare Results at the Identified AGNPS
the AGNPS Channel Grid and its Channel Grid to Determine which

Associated Upstream Sub-watershed Alternative Gives Best Achievement

l T

Source Accounting Alternatives
Find the AGNPS Land Grid in the Sub- Change the Management of Each Field in
Watershed that Contributes Most tO the the Identified Land Grid and Run the
Problem Of the Worst Object Function AGNPS Model on Each Alternative

 

 

 

 

 

100
contains more than one field within the 37.58-acre grid area. Each agricultural field will

be temporarily changed. The principle Of change is based on our knowledge Of the
causal-effect relationship between the problem and the source.

Since current scientific knowledge still can not quantitatively address the “how
much” questions accurately, a trial and error approach is used. The strategy is simply to
change the field management operation into a more resource conservation direction and
compare the results from these temporary changes. The procedures begin with finding
out what Operation the field currently is, then, temporarily changing it to a more
environmental friendly Operation, repeating the process for all fields in the grid, and
determining which Operation on which field will reduce the problem most. The overall
guidance is changing from conventional tillage to conservation tillage, conservation
tillage to no tillage, no tillage to alfalfa-hay, and retire (idle) it if it is already an alfalfa-
hay field. The AGNPS model has to be run for the sub-watershed Of the problem grid for
each temporary change and store the results for comparison. Results are compared
among all changing fields in that grid so that one can determine a new management
operation for “the field”. This iterative targeting procedure will ensure that the selected
changes will yield best achievement among all fields considered in the problem grid.
This step reflects the management decision in the spatial context, that is, which Best
Management Practices (BMPs) will be most effective at which field for a specific
problem.

The last step is to actually apply the changes. The values of all the 8 land
use/cover and field management model variables of the identified field have to be

Changed based on the new field operation chosen. Then, the whole watershed has to be

101
re-run with the AGNPS model with new change and ready for the next iteration. For

each iteration, the result of the new “scenario” will be compared with the existing Pareto
optimal solution set. The Pareto Optimal solutions are the first set Of 15 objective
functions. If any solutions in the existing Pareto Optimal solution set is worst than the
solutions of the new “scenario” in all 15 Objective functions separately, they will be
replaced by the new solutions and vice verse. Otherwise, the new solutions will be added
to the existing Pareto Optimal solution set.
The above procedures can be summarized as a five-step iteration:
(1) find a particular stream segment or grid that has the maximum difference
between a goal (problem) and the AGNPS model output (normalized);
(2) use the new source accounting feature in the AGNPS model to evaluate the
pollutant sources for the problem;
(3) find the field that contributes most tO the problem by applying the know
causal-effect logic, that is, minimize the maximum deviation;
(4) change the field (create a new and better scenario) and rerun the AGNPS
model for the new alternative; and
(5) compare each goal Of the new alternative to the existing Pareto Optimal

solutions and replace inferior ones.

CHAPTER 4

RESEARCH RESULTS AND DISCUSSION

4.1 Optimization Process
4.1.1 Processing Limitations

This research has encountered two major computing problems - time and space.
A Pentium Pro 180' personal computer with 64MB Of RAM running a UNIX Operating
system (LINUX) was used in the research. The computer has a ZGB hard drive partition
designated to store programs and data for this study. These two problems resulted from
three causes, complexity of framework, data volume, and search method, directly or
indirectly related the Optimization issues.

The first issue is complexity. The study uses a simulation model, AGNPS, to
compute information needed for calculating the normalized Objective functions. For each
Optimization iteration, the model input data have to be re-calculated for each AGNPS grid
since there are changes (field management practices). This process involves both tabular
data manipulation and spatial data integration utilizing database and GIS. Then, the
AGNPS model can be re-run with the new set of conditions. Model outputs also need to
be stored in the database so normalized Objective functions can be subsequently
calculated. These procedures have been automated through a set Of programs but the

processes are complex, tedious, and very time consurrring.

102

103

The second issue is the volume Of data. There are 12 basic sets of data
corresponding to 3 crop stages for four years. Each set contains input and output data for
each AGNPS grid. In addition, there are more than 6500 individual Objective functions.
These objective functions have to be normalized and aggregated. Once the Optimization
process starts to generate results, these results need to be stored for later evaluation. The
data volume increases dramatically with each Optimization iteration.

The third issue is related to the search method used for finding Optimal solution.
The optimization process searches for better solution through the change Of decision
variables, i.e., field management practices. The study considers the spatial dimension,
i.e., water flow path and field location, in BMPs selection. A program will first find the
maximum deviation of a specific normalized Objective function (Min-Max strategy), that
is, it identifies the most serious NPS pollution problem on a channel grid (or segment) Of
the watershed. An identified channel grid is an outlet of a sub-watershed within the
watershed. Then, the upstream sub-watershed can be delineated based on the water flow
path. With the new source accounting feature in the AGNPS model, the highest
contributing grid to a specific NPS pollution problem at the identified channel grid can be
found. However, there may be several fields within the target grid. Since field is the
management unit, change of management practices must be based on a field instead of a
grid. Change (a better management practice) Of each field will be calculated through the
AGNPS model that takes the water transport into account. Finally, results are compared
at the identified channel grid to determine the BMPs that should be installed on a field.
This search method involves multiple processes and is performed for each Optimization

iteration.

104
4.1.2 Adjustment

Once the optimization processing started, it was found that approximately 10
minutes were required to complete each Optimization iteration. Also, disk space was
consumed very rapidly. Through close monitoring Of the Optimization processes and
careful examination Of intermittent results over a three week period and several hundreds
runs, a decision was made to stop the Optimization process after 1500 iterations.

The Optimization processing converges to the Optimal solution very slowly. This
is probably due to the gradual search approach used, i.e., the change Of management
practice on a field uses a better method instead Of the most effective method. The reason
of using a gradual change approach is tO have a chance tO evaluate the effectiveness Of
different management practices in the optimization iteration process. Due to the slowness
Of approaching the Optimal solution, more iteration runs after 1500 iterations are not
likely to provide much more information. Another problem encountered is the disk space
available.

By reviewing results from the 1500 iterations, some answers to the research
questions have been Obtained with a certain degree Of confidence. Then, two sub-
watersheds were selected (see section 4.3 for details). These two sub—watersheds contain
all identified channel grids that have NPS pollution problems. Then, the data were split
into two sets based on each sub-watershed. The same Optimization process was
performed for each subowatershed on two separate machines. The optimization process
was run for another 500 iterations. A week to complete the setup and 500 iteration runs

was required.

105
4.1.3 Result Interpretation

The next section (section 4.2) presents two sets Of results. The first is for the
whole study area. The other is for two sub-watersheds. Each set contains three pieces Of
information: the Pareto optimal solutions, the identified channel grids that have NPS
pollution problems, and the fields that have changed management practices in the
Optimization process. This section intends tO provide some more background information
on how to interpret these results in general.

The Pareto Optimal solutions are normalized Objective function values. There are
15 normalized Objective functions in each solution set. These 15 normalized Objective
functions are aggregated spatially and temporally. They are summed values Of each
pollutant (sediment, nitrogen, and phosphorus) for all channel grids based on certain time
periods. Three of the 15 normalized Objective functions are over for the four-year study
period. The Others are for each individual year Of the four years. These solutions
represent the “best results” achieved when the Optimization process was terminated.

By definition, one can not enhance any Objective function without diminishing
another Objective function among these Pareto Optimal solutions. They also represent the
normalized pollution control level of each NPS pollutant (sediment, nitrogen, and
phosphorus) during the time period covered. Due to the spatial and temporal aggregation
Of these normalized Objective functions, they represent the study area or a sub-watershed
as a whole. Since they are normalized Objective functions, i.e., normalized relative to the
goal Of each individual NPS pollutant, negative values represent less than the goal limit

desired and positive values represent higher than the goal limit desired. A second table

106

representing the average value of each Objective function was also provided. The average
objective function values can be represented in the following formula:

AOF=goal+goal * NOF/(grids * stages * years)

Where
AOF = average Objective function value per channel grid
(it is the difference between the goal and the value)
goal = the goal set for each type of NPS pollution problem
(sediment = 1000 mgll, nitrogen = 10 ppm, and phosphorus = 1
mm)

N OF = normalized Objective function value

grids = total number of channel grids in the watershed or sub-watershed

stages = 3 (crop stage)

years = total number of years (1 or 4)
By using the above formula, the average Objective function values convert the normalized
Objective function values back into the same units as the. goals. In this way, it is easier to
interpret the solutions relative to the magnitude Of the goals.

The identified channel grids that have NPS pollution problems and the fields that
have changed management practices were presented in both map and table formats. Maps
represent the “hot spots” of the aquatic and terrestrial systems in the spatial dimension.
Tables were summarized by time (crop stage and rotation year) based on the NPS
pollution type. They can be treated as summaries of the Optimization process since both

were derived in the Optimization process.

107
The identified channel grids that have NPS pollution problems were derived from

comparing all Objective functions of all channel grids. These Objective functions are the
second set of Objective functions (6656 in total) described in chapter 3. Each Objective
function represents a specific channel grid Of a specific NPS pollutant at a specific crop
stage Of a year. It retains three characteristics: where (grid number), when (crop stage and
rotation year), and what (NPS pollution type).

The identified fields that have changed management practices are derived from
three criteria in the Optimization process. First, they are fields that contributed most to a
particular NPS pollution problem Of a hypothetical scenario based on the AGNPS source
accounting feature. Secondly, the rule of management practice changes is a gradual
approach, that is, from conventional tillage to conservation tillage, cOnservation tillage to
no till, no till to pasture or grassland, and pasture or grass land to idle. Thirdly, they were
determined by evaluating the AGNPS model outputs at the corresponding identified
channel grid.

With the above understanding Of what results represent, the following sections

further describe in details the research findings.

4.2 Watershed Results
4.2.1 Pareto Optimal Solutions

There were 39 Pareto Optimal solutions generated before the stop call. The
associated Objective function values for each Pareto solution are presented in Table 4-1.
These solutions are “best” results among all scenarios that have been examined. These

15 normalized Objective functions can be divided into 5 sets based on the time period

108

each covered. That is, one set for the four-year period and four sets for each year within
the four years. Each set contains three NPS pollution types, sediment, nitrogen, and
phosphorus. Figure 4-1 to 4-5 showed these five sets of Pareto solutions. Furthermore,
the normalized objective function values were converted into their corresponding goal
units as average values (per channel grid) in Table 4-2. Table 4-3 also provides a simple
count statistics Of normalized Objective function values and average Objective function
values grouped by 3 types of NPS pollution problems and the time periods. Several
conclusions can be drawn from these tables and figures.

From Table 4—1 (and the left portion of Table 4-3), there are three general
findings. First, all normalized Objective functions Of nitrogen for the 4—year period and
each individual year are negative. This means that the watershed in general will not have
a nitrogen problem. Secondly, all normalized Objective functions of phosphorus for the
4-year period and each individual year are positive. This means that the watershed in all
will have a phosphorus problem. Thirdly, the watershed may have sediment problem for
certain scenarios since there is a mixture of positive and negative normalized Objective
function values Of these 39 Pareto solutions.

In general, we can conclude that the watershed as a whole will have a phosphorus
problem and may have sediment problem. These conclusions are based on the goals set
for each NPS pollution problem and also the number of iterations (1500) that the
Optimization process has been run. The conclusions will be different if the goals are
changed and the Optimization process continues to run for more iterations. By examining
the average Objective function values (Table 4—2) relative to the goal set, the magnitude Of

each NPS pollution problem can be further explained. The right portion of Table 4-3

109
divided the magnitude into 5 ranges. For sediment pollution problem, the magnitudes of

most “best scenarios” are less than 1.1, i.e., less than 1100 mg/l per channel grid. The
magnitudes Of most “best scenarios” for the phosphorus pollution problems are above
1.24, i.e., 1.25 ppm per channel grid. These magnitude figures showed the watershed in
average.

Since these solutions were derived from randomly generated initial scenarios and
changed through each Optimization iteration, different spatial combinations of field
management operation were implicitly embedded. It implies that these solutions also take
into account the landscape characteristics and still reflects potential sediment and
phosphorus NPS pollution problems.

Figure 4-1 to 4-8 also presented the magnitude of normalized Objective function
values for all 39 Pareto solutions. Figure 4-1 to 4-5 grouped normalized Objective
functions by the time period each covered. These figures show clearly a phenomenon that
three types of NPS pollution problems have similar trends with each other for different
time of period covered. If a Pareto solution has higher value Of sediment than the
sediment values of other Pareto solutions, the other two NPS pollutants (nitrogen and
phosphorus) will also have similar situations. Furthermore, the scale (slope) Of changes
of these Pareto solutions between sediment and phosphorus are very similar. This can be
attributed to phosphorus pollution mainly coming from sediment and nitrogen mainly in
water soluble form. Figure 4-6 to 4-8 grouped normalized objective functions based on
NPS pollution types. Each figure showed a stack bar, containing 4 individual year parts,
for each Pareto solution. The purpose is to find whether a particular time period (year 0

to 4) will consistently have a particular NPS pollution problem across these 39 Pareto

110

 

lll

3?.

05

2.5

mos—.5

.oz

 

N4. 035.

112

Figure 4-1. 39 Pareto Solutions for 3 NPS Pollution Problems for the Watershed (4 Years)
2000

 

d

Sum of Norm. Obj.
Function Values

 

 

 

 

+ Sediment + Nitrogen + Phosphorus

Figure 4—2. 39 Pareto Solutions for 3 NPS Pollution Problems for the Watershed (Year 0)

 

 

 

 

 

 

—o— Sediment + Nitrogen + Phosphorus

Figure 43. 39 Pareto Solutions for 3 NPS Pollution Problems for the Watershed (Year 1)

 

 

 

 

 

 

 

 

 

 

 

 

 

1 13
Figure 4—6. 39 Pareto Solutions for Sediment for the Watershed (Year 0 - 3)

 

Normalized Objective
Function Values
its..§s§§

 

 

13579111315171921232527293133353739
llYearO IYear-l BYear2 DYear3

Figure 4-7. 39 Pareto Solutions for Nitrogen for the Watershed (Year 0 - 3)

55%..

  
  

§§§§

Normalized Objective
Function Values

I
d

ii

79111315171921232527293133353739
[1on Ile BYeerZ DYeer3

.a
(a?
CI

Figure 4-8. 39 Pareto Solutions for Phosphorus for the Watershed (Year 0 - 3)

 

         

woo: ----------------------------------- "i
" ------ ——————— ----
> ----- 3;i? “a? - 5: 55
g; 400 ' I a! .EEBI :;:.I .35." IE Egg
2 20° null Illlllllllllll llllllnnl lulllll llllllllln ll“

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 79111315171921232527293133353739
5[lYearO IYearl BYearZ DYear3

l 14
Figure 4-9. An Exapmle of Pareto Solutions

lONamlizedijeefivannefiomquefimeflmdPhosphmsmiMNiuogen)

 

 

 

o‘ o‘ ‘ S 0‘
9569 @° \6‘00 @e gee
9 Q Q Q Q Q
l m.m I I I J I I I I I
800.00
600.“)

 

 

 

WWWVdm

 

-200.00 -

 

 

 

 

400.00
—0— Solution No. 36 —'+- Solution No. 37 - - -I - - Solution No. 38 —9— Solution No. 39
New

1. Solution No. 36. 38, and 39 are Pareto Solutions.
2. Solution No. 37 is not a Pareto Solution since all values are higher than solution No. 39.

115

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4-3 A Sim ale Count Statistics of Pareto Solutions Based on Their Values
NPS Problem No. of Norm. Objective No. of Average Values Relative to the
And Function Values that are Magnitude of the Goal that are
m” ”was Neggive Positive <= 1.0 < 1.1 < 1.25 < 1.5 >= 1.5
.. 4 years 19 20 19 9 4 7 o
8 YearO 14 25 14 1o 4 11 o
.13. Year 1 23 16 23 6 3 7 o
'3 Year 2 26 13 26 3 o o 10
Year 3 22 17 22 7 10 o o
4 Years 39 o 39 o o o o
a YearO 39 o 39 o o o o
g Year 1 39 o 39 o o o o
g Year2 39 o 39 o o o 0
Year 3 39 o 39 o o o o
a 4 Years 0 39 o o 14 23 2
o YearO o 39 o o 9 19 11
'8. Year 1 o 39 o 9 12 12 6
E Year2 o 39 o 3 16 6 1o
9* Year3 o 39 o o 17 22 o

 

 

 

 

 

 

 

 

 

 

 

solutions or not. It seems that there is no such phenomenon.

The goal setting greatly affects the derived 39 Pareto solutions in the optimization
process. Since all nitrogen normalized objective functions are negative and much below
the goal value (see the left portion of Table 4—3), this nitrogen goal can potentially be
dropped out of the optimization process. This will yield different Pareto solutions.
However, we can drop out all normalized objective functions related to nitrogen from
Table 4—1 and re-apply the Pareto concept to find the Pareto Solutions. There will only be
3 Pareto solutions left, solution number 36, 38, 39 (backgron highlighted in light gray
in Table 41). These three solutions plus solution number 37 are plotted in Figure 4—9 to
show the Pareto concept. There are a total of 10 normalized objective functions for

sediment and phosphorus. For all the 10 normalized objective functions, their values of

116

solution number 37 are all higher than the corresponding values of solution number 39, so

that, solution number 37 is not a Pareto solution.

4.2.2 The Iteration Processes

Another source of results is from the optimization iteration processes. This
iteration process tried to find the most serious source contributing fields for a particular
NPS pollution problem in a particular channel grid at a specific point in the optimization
process. Each time the optimization changed a field management operation, the
information was saved into the database. The saved information includes:

1. the current optimization iteration (it)

2. which initial “scenario” (cp)

3. the problem channel grid number (cellno)

4. the field that contributed most to the problem grid (fdid)

5. the current land use/cover and field management code (lulc_old)

6. the new land use/cover and field management code (lulc_new)

7. the problem type (ptype)

There are only 14 distinct channel grids recorded out of 1,500 iteration runs. This
means that the NPS pollution problem found always happened at 14 channel grids with
all the 1500 changes. These 14 channel grids were identified through the Min-Max
strategy, which finds the maximum deviation of individual objective functions. There are
also only 19 distinct fields that have been changed. Changes occurred 967 times due to
sediment load and the rest (733) are because of a phosphorus problem. The land

use/cover and field management of all changed fields are all conventional tillage

l 17
(COIDOl to COIDO4), so that, all new land use/cover and field management operation

used for all changes are conservation tillage (COIDOS and COIDO6). Figure 4-10 showed
the spatial location for these 14 grids. The corresponding 19 fields that have changed are

presented in Figure 4—11.

Most of these channel grids are located at the upstream area of the channel
network in the study area as shown in Figure 4—10. None of them occurred at the
downstream area. The landscape characteristics, i.e., fixed AGNPS model input, and
some iteration information of these channel grids is presented in Table 4-4. Six AGNPS
model input variables, slope, slope length, channel slope, channel sideslope, soil
erodibility, and soil texture, were presented in Table 44. Most of these data values are
slightly higher than their corresponding watershed average values. There is not enough
discrepancy for any of these grids to be much different than the average of all channel
grids.

Most fields that have changed management practices are adjacent to the channel
network as shown in Figure 4-11. Field information related to the AGNPS model and
optimization iteration processes were summarized in Table 4-5. The AGNPS model
input data plus elevation of these identified fields does not differ too much with the
corresponding watershed average values. One can not attribute any particular landscape
setting to the result, except the distance to the stream network. Field ID. 137, 180, 952,
and 1050 have been identified and changed over 200 times each. These fields which are
typically called “hot spot” most likely contribute more to either sediment or phosphorus

pollution problem in the watershed.

l 18
4.3 Sub-watershed Results

4.3.1 The Two Sub-watersheds

Based on the watershed results, two sub-watersheds were selected to repeat the
same optimization process. Since all 14 identified channel grids that have NSP pollution
problems are located at the upstream half of the watershed (Figure 4-10), two channel
grids, grid number 178 and 231, were identified as the outlets for each sub—watershed.
Their corresponding sub-watershed areas are shown in Figure 412 and 4-13.

Then, the optimization processing data were split into half, one for each sub-
watershed. In this way, the optimization process can continue to use the result from the
first 1500 iterations, i.e., each will inherit the data, information, and results from the
previous watershed runs. Each sub-watershed was run on a separate machine for another
500 optimization iterations. The purpose is to obtain results in a shorter time period and
examine whether more iterations will provide more information or not.

After 500 iterations, tables, figures, and graphs similar to the previous watershed
results were created. Table 4-6 and 4-7 present the Pareto solutions for each sub-
watershed. The average objective function values are shown in Table 4~8 and 4-9. The
same analyses were performed to create trend figures (Figure 4-14 to Figure 4-29) for the
Pareto solutions. Identified channel grids that have NPS pollution problems and
associated fields that have changed management practices were shown in Figure 4-30 and

4-31 for sub-watershed 178 and Figure 4-32 to 4-33 for sub-watershed 231.

119

Figure 4-10 The 14 Chimnel Grids Identified as NPS Pollution Problem for the Watershed

 

Water Features
AGNPS Model Grid

120

Figure 4-11 The 19 Fields Identified as Fields that Have Been Changed forthe Watershed

 

121

Table 4-4 Summaries of Identified Channel Grids that Have NPS Pollution Problem for

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

the Watershed
AGNPS Model Input Variables
Grid NO- $10 $16 Channel Channel . . Soil
(96")e Lengtbeat) Slope (%) SideSlope 1%) 8°“ Emm" Texture

195 1.7 239 1.75 29 0.19 4
263 5.2 244 250 33 0.22 1
275 2.7 279 2.33 29 0.24 2
309 3.0 272 3.00 33 0.17 4
321 3.1 275 6.25 29 0.26 2
356 3.3 261 2.50 33 0.27 2
331 3.0 270 2.45 33 0.23 2
431 2.7 275 11.00 50 0.30 2
449 2.4 231 2.33 29 0.23 2
471 . 2.4 279 2.00 33 0.30 2
507 3.1 271 3.50 29 0.31 2
522 2.9 276 2.00 29 0.32 2
523 2.9 275 6.00 29 0.32 2
534 3.4 263 2.00 29 0.16 4

Average' 3.0 273 3.54 32 0.26 2'

Avera e2 3.0 273 2.39 31 0.23 2r

 

 

 

 

 

 

 

 

*-'I‘hevalueisthedominantsoiltexture,where l =sand,2=
l - Average value of 14 identified channel grids.
2 - Average value of all channel grids (183).

silt, 3 = clay, and 4 = peat/muck.

 

 

122

Table 4-5 Summaries of Identified Fields that Have NPS Pollution Problem for the
Watershed

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

AGNPS Model Input Variables Iteration Information
IE" Elev Slope 33:; Soil 8"“ Tm". Problem Times
(m) (‘5) if!) Erodibility 1 2 4 Type Identified
88 279 4.0 259 0.24 11.6 47.8 1.3 P 1 17
90 279 5.2 244 0.22 8.0 P 4
1 14 292 3.7 263 0.30 63.8 S 13
120 294 3.6 263 0.27 49.6 28.2 S 2
124 287 3.9 260 0.26 20.7 20.9 S 84
125 292 3.6 263 0.26 19.6 18.0 S 4
137 277 2.5 279 0.21 12.9 13.6 P 226
162 302 2.5 278 0.28 14.2 29.4 S 99
166 297 2.2 283 0.29 21.3 90.5 S 2
180 297 2.6 277 0.30 132.8 S 227
952 301 2.4 281 0.32 1 19.4 P 208
953 299 2.8 278 0.32 17.3 S 2
1044 277 2.5 282 0.24 1.8 34.9 S 84
1046 278 3.0 273 0.25 36.0 S 18
1050 281 2.6 279 0.27 18.7 1 16.8 S 232
1379 275 1.5 294 0.16 32.0 P 102
1648 289 2.7 275 0.22 8.9 10.0 3.1 P 1
1716 290 3.6 263 0.17 6.9 6.0 P 36
1719 287 3.6 270 0.17 5.1 P 39
Average' 233 3.1 272 0.25 131 779 61 3 p 733 P
Averagez 234 2.6 279 0.25 8606 12102 1353 11 S 767 S
I"-‘I'hevalueistheacreageofeachsoiltextureinthefield,where1=sarld,2=silt,3=clny,and4=
peat/muck.
“' - The NSP problem type, where S = sediment and P = phosphorus.
1 - Average value of 19 identified fields.
‘Lm_e_mgevalueofmewatershed,exceptthatsoiltexhueisthesum.

 

 

123

Figure 4-12 Sub-watershed of the Study Area with Outlet at Grid 178

 

124

Figure 4-13 Sub-watershed of the Study Area with Outlet at Grid 231

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

125

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.8. 8.. .8... .8... ...8. .8 8.... .88. .8... .88.. _.i..8. «N8 ..~.8 8.88. ...8. 8
.88. .88 8.8 8.8. 8.« 8.3 .88. .88 .mmnz «..8. «..8 8.8. 8.88. 8.8. 8
.88. 8.8 8.«.. 8.8. 8... .88 .88. .81.. .818 «..3. . 8
...8. 8.8 8.3 8.8. 3.. .88. .88. .8. .«..8 8.8. 8
.88. 8.8 8.8 8.8. 8.. .38 .88. «8.. .38 8.8. 8
.88. 8... 8.8. 8.8. 8. 8 .38. 8.8. ...8 .88. .88. 8|:
.88. ..... .8.«. «...8.. ...8.. .8...» 88.. 8... .88 .88. .«
8.8.. 8.8. 8.8. .88. .88 .88. 3“». .m8 8.8. 8...... 8
8.8. ...« 8.8. .83. EM .88. 8.18. 8%. .3. 8.8. 8
...8. 8.« 81.8. .23. .«... .88. .38. .8.«. 8.... 8.8. .«
8.8. 8% .88 W8. 8.8 .«.8 8.8. .88. 8.... 8.8. 8
...8. o..«. .8... 8.8. .8. 8.8 8.3. .mm... #88 ...3. 8|.
«..8. .8. _8.8 8.8.. 8.8. 3.8 1«$8.. .8... 3.8 8.8.. 8.8. 8.8 .88.. ...8. .«
.848. ob.“ 8... 8.8. 8.. .88 8.8. .«m.. 8.8. «..8. 8B $8 .ol888. 8.3 8
8.8. «N8 ...8 8.8. a... .88. 8.8.8. .8. 1.1.1.... 8.8. .8. 8.88 «...8.... ...8. 8.
...8. 8.. 3.8 8.8.. 8mm... .88 8.8. 8.... 8.8. 8.8. 8... 8.88 ...8.... 8.8. a.
8.8. 8.8 8.8 2.8.. .88.. .88 ...8. 8.«. .8... .88. .8... 8.88 8.88. 8.... ..
8.8. ...8 .88 8.8. .8... .8... ...8. 8.1 R... #8. .8. o...mm 8.88. 8.8 o.
...8. an. 8.8 8.8.. .8... .38 8.3.. ...8. 8.13 88.. .88W 8.88 8.8.. 38. m.
8.8. 8... 8.8 8.8. 8...- 8.1.8 3.8. .... 8.8 8.8. .om«« 8.8 8.88. 8... ..
.m8. 8... 8.8 8.8. 8..« .«8 .88. 8m 8... 8.8. .mm. .888 8M8. 8.8 a.
8.8. 8.8 cm 8. .88. .88 8.3 8.3. o..«« .88 8.3. .881. .888 .8. .8. 8.8 «.
8.8. 8.8 8.8 W3. 8.8 8. 8 85.... 8.8. 8.8 .88. .88 .«.88 8.8.8. .«.«. ..
.88. 8... 8.8 «..8. 8.8. ...8 ...8.. 8.8. 8.8 ...8. 8... 8.8« 8.8. .8.« o.
.88. .8... 8.8 .88.. 8.... 8.8 ...8. 8... ...8. ...8. 8... 81.88 8.38. ...8. 8
8. 8.. 8...... .88. .88. «.8 8.8. 8.3. .88 8w... 8.8. 8.8 an... 8.88. ...8.. a
8.8. 8.8 .88. .88. .«...8 «mm. .38. 8.. .818 $8. 8.8. .188 8. .8. 3.3. .
8.8. .8... .88 3.8. .88 .88 .8m8.. 8.». .88 8.8. 8.8. .888 £89 ...8. o
8.8. .«... .8... 8.8.. 8.... .3. 8 ...N 8.. 8.... .8... .88. 8.8 8.8 .38.. ...8.. o
.8. 8.. .8... .8... 8.8. 8.8 .88.. 8.8. 3.8 .88.. .88. 8.8 8.8. .888. .88. .
.88. ...8 88 8.8. 8.8. .8... 8.8. .8... .88 .88. 8.8. .388 8.8. 8.8 8
...8. .818 8P8 8m... 8.... .8L..8 .8. 8.. 8.... _8.3 .88. .38. .888 8.88. ...8 «
.«.8. 8.8 ...8 8.8. .88 .818 ...8. 8... 8... 8.8. 8.8 .8..8 8.88. 8.8 .
z . ...8 a z r .8 .. . z .98 a z ...8 _ .— z ...8.
81.8» «8|...» ...8.. ...8.. . ..am .8... ..z
.32.... 3...... a .8...» 5.3.5. 8.38.5.2

 

 

 

 

 

 

 

 

a: .02 2.5 8 85:0 5.3 confinibsm 8.. 8.0.3.9..“ 20.8.. c... 03....

 

126

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.8 .88.- 38 8.8 8...... 488 +848 .8... .. cm... ...8 8...... E .388 .88... 8.8. «ml
«8 .88... 8.8 8.8 ...8...- o.... al....» 88.. 8.... _88 8.8.. 8..« .38. .88.- 88 .«
.8 .88.. 8.8 3.8 «mm-..- 8.... .88 81......- 38 mm... ...«... 8.8 8.8 .888. 81.8.8. 8
.3. Wow..- .8... 8.1.... cu...- _3..l«. ..3 .8... 3.... 8... 3.8... .m8 ...8« 8. .8. 8.... 8.
.8 88.. .88 8.8 «.8... 8!... ..1... 8......- 8... ...8 mm...- 88 8.8 8.8.. 8.3.. a.
8.8 .88.. 8.8 8.8 8.8.. .88 8.8 .88... 8... 8. 8 .8..... 8.8 8.8 ...8.. W8. ..
..8 M8.- 83 .8. 8 8.8.. ...8 8.... ...8... 8.... E. 8.... 8.18 _8..8 8.88. 81... o.
.3 8.8..- .«18 mm... 3.8..- .8.o 8.8 8.8.. 8.... 3.. 8.8... .8... ...8 «..8.. .818 m.
.8 8.8.. .88 8.8 .88.- .38 ...8 .88..- ..81 .38. ...8.. .88 .888 8.83. 8...... ..
...3. .88.. .«h8 ..... 8.8... .8... ...8 819m..- 88 ...8 .88.. .88 .8.«8 8.83. ...8. 8.
8.8 .88..- .88 .8... 8.8... m... 8.8 8.8.. ..L... .3... .fifi... .88 .888 .88.. «..m. «.
8.. .88.. .88. 8.3 .88.. 8... .88. 8.«... .2... .88.. .88.. .83. ...8. .888. 8.88 ..
8.8 ...8.. .8. 8 «.8. .88... 8.8 8. .. 8...- ..«18 .38 ...8.. ...8 8.8 .888. ...8« o.
3.8 ...8.. 818 .88 .88... 8M8 ...8 8.... 8M8 .88 .88.- .83 «...8 8.8. 8.8. 8
.8 8.8.. ««.8 8.8 .8...- 38 .88 «.8... .8. 3.8 8.8.- .88 ...38 8.08. «..8. a
...8.. «.8... 3.... 8.8. .83.. 8.3 88 8.8.. B8 ...8. 8.8.. L88. 8&8 _8..8... 8.8. .
.8 3.8.. 8.8 ...8 .88..- .88 8.1.... .88... «m: .88 _38... .38 .388 .888. ...8. 3
.8. .88... 8.8 8.8 .88... .881. 8.8 .88... .8... $.18 8.3.18.8 «...8« ...8... .8... m
.8 .88... «.18 3.8 .88... .88 8.3 8.8.. .88 .88 _8..8.- .m.« 8.88 8!.«8. 8.8. .
.8 .88.. _8..8 ...8 .88... 8.8 ...8 ...8... .1313 ..m8 fig. 83 888 W883. 8.8. m
«8.. 8.8.. .83 ...8. Q8? 8... 5...... ...8.. 88!. alum- 83... 8.8 8.88 8.8. 8.8« «
.8 8...... 8.8 8.8 «N...- 8..« J88... .3108... .8... «N8 3......- 88 8.«.« 3.8.. ...8 .
. .— z ...8 .. m1 ...8 a . z . .8 .— z ...8 .. z ...8
. 9.1.8.. ...8.. . mm... ...8.. 38.8... ..z
F In...- - a... .8 .5 .8....E.z

 

 

.8 .oz 26 a .85 5.; 8.3.3.88 3.. 2828 32... S. 0......

 

127

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.. 8m .9... .8. _..h« .8. .. find 8.. 8.. 88 .88.. _8. 88 W8... .3...
. 88 ..8 _8. mam 88. 8. _88 mm 8.8... ...8 ‘MN... 8.. SM _8«. 8
. 88 ..v. Eh. _8.« .8. ch. _88 _8«. _vm. 88 .8. v... _..8 8«. «8
. _va8 88. 48.. SM ...8. .me .88 .88. .8. H88 .8. «h. 8B8 ..«. .8
.... 988 _8«. 8. .88 88. 8h ...8 .8. .m. _88. .88.. BM. .88 «8. 8
8. 88 _8: Wm... #8 «.o. 8:. 88 «8. .8. _8m« «8.. E... 88 88. 8
8. _8.« 68. 8.8 2% 8.8. g .3 88. oi.« _88 88. .8. ..8 «8. 8
m. «m« «8.. .84. «88 L88 _8. _8.« .3 on. _88 So. .4. 88 88 ..ml
Tm. .mm .«8 8.« _88 3.8. E _88 «.8. .8“ _«.8 .8. 8.. .88 «m. o«
.8. 8.8 8.8. .88 8.8 88. «ha am... o8. ..o.« EH8 8.8. «W. .88 8... 8|.
.5. 8.18 .«.... ..8 ..8 «cm. EM «...8 W8. _ab.« _88 .8. 3.. ow... .8. m«ll
.. 6N8 ...«. «W. 8|...8 «8. an. «88 8.0. Wm. 88 8°. ..oM. 88 8.. 8«
.. _8.« 08. _mm. 8% 88. Mom. .88 8.. .8. 88 _8: _«m. 88 .... ««
.8. 48m 888. 8. cm... .8 «N. W« _«8 8h. _8.« 7.... _am. 38 .I8 .ml.
8... 8.8 .8. 8.. 88 «8. 8.. 88 .88. «UN. _88 _88. _8. SM 8.. 8
mm. 88 8.. an. 8m ..8 a”. EM «.8. 5%. 8m 8 _8. EN. .8. a.
.. _8.« .8. _8. .cm 88 _8. 8m 88 .M. ...8 8o. 8.. .8.i« _88 8.
.8. _88 ..8. _8«. 8“ 08 8m... am .88 8.. «88 .8. on. ..m _88 ..
8.. 88. .8: .8. .88 8.. _8..r. _8.|« _88. Tm. 8m 8. 8.. 8m _88. 8.
. .88 _88. «8. _u8..|« 88 EN. ...8« =8 _8. .88 .8. .8. 8m _88 8.
.. _8« 88.. .8. Ltd .88. L8. .88 .88. EN. .38 8.. 8.. 88 _88. c.
8. .88 «8. «... 8.8 _8.. _8 . 88 8.. _8. EN 88. .... .88 8.. 8.
. :88 «8. .8... 8.8 _8.«. _o... g 8.. _8. .88 .8. L8... «.8 8.. «.
... EN... 3«. 8.. 88 _8«.. _8. 8.8. ..8 _8. 8m 8.. _«8. 8m 8.. ..
.8. _88 .8. .v. .88 _8..8 _S. .84..« 88 ...... 88 .88. 8. am 808. o.
8. EM .2. 8.. 88 .88 EN. .88 ...8. WWW. 8m _8.. WM. _8.« omo. o
8. _88 88 8.. 8.8 8v. ...8... ..8 8c. 8.. 8.8 _08. «8.. 88 ..«8. 8
.. «cl8 .8. .8. cm." 98. _m. 348 .8. _8. 88 _8«.. «... «.8 3.. .
.q. .88 38. v... ...8 _8«. 8.. .mw .88 8.. .88 8F: 8.. .mm" .... 8
83 8.8 82 .8. 8M .88 8 . cm 98 W. «.8 «3. .... 8M ...Ia m
.. ..H8 ...in .8. 88 8.8. mm... 88 ..8. _8. .m8 888. 8... .N8 .8. v
... «N8 _8«. 8.. 8w ..8 .8. 8.8 ..o. .8. 88 8.. 8.. 8.|« 8o. 8Il.
.. 8.8 .... _8. 38 ..o. 3 . _.v« «.8 8.. «88 8.8 .8. 8.8 .88. «
.. _88 .«8. 8. «.8 «8.. .8. _88 :2. 8.. 488 _8«.: .8. «8.8 ...... .
_ _ z ...8 a _ z ...8 a r 2 L ...8 .— _ z _ ...8 z ...8
_ 8.4m» «mu... .8. time. 58.8» v ..z
_

a§fl§>gta8q3lfifl

 

 

8: Ragga 8. 26 .0520 E 8..; 888... 288.30 088.22 8... 2%..

 

 

 

 

 

 

«:7

.8... 8.. 88 8.. 8.. 8.8 _8«. .8. 8.8 La... ”Hi.
«8.

 

 

 

 

 

 

 

 

 

 

 

 

V

 

 

 

128

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

."-

 

«....
8... .8... _va E 8.. 88 _8«. ow. «.8 8..
.8.

88. 3.. _88 «8. 8.. .88 8.8. .m8 ...8.

 

 

 

 

 

 

 

 

 

 

8

I .

.2. «N. .88 8o. 8. «88 9.8. 8.8 .... 8
8

e

 

«M: .m. .88 .88.. _wfl. 8w 8... _848 53.

 

7"2393'

8.

«I8. :8. 7.88 888. _8«. 88 8.. .W. 88 _8...
8|...
8.

00 w _. _QNJ _aod _o: p mu. . mod ov— . _ and m... w p n

 

 

 

8.8. .88.". .98 8m. 3m 88. _8«. 88

 

 

.

. .I«.. .8. 8m . 88. 88 _8... .q. 88 _8«..

. «8. 85 33.8 .88 cm. 88 8.. .W. 8m +8.8. WI
8.. 82 _nm. 8.8 78.. .m. 8...... 8... 5. EM «8.. 8

. 8.. _8«. 88 .88. oh. .88 _m... cm. 8W« 8.. 8.
.q. ...8. sh. 8m .3. 8w. 8.8 .8... _8m. .88 _8«.... 8.
. .. 88. .8. 8m .8. . .... in... W. 88. ..
8. _8«. a... 8M 88 .... SM «i8 8.. 8.

. 38. E ...:«W ..8 8... .W. _88 8.8. 8.
. ... 88. .8. 8H _8I8. 8.. 8m. ....m 8.. c.
_ .. .88. WIN. .8.« 8.. ...«. Fm. _8..« .8: 8.
. . ..o. E .8I88 2.8 8.. _8«. meH 8.8. «.
. . _88.| ch. 88 a... ...... 8H. SH. 88. ..
. . .8: oh. «.8 _«mw. 88. EM. 88 g. o.

. 8.. «n. 8.8 _8«. film. «.8 7m. ...8 88. a

.

.

.

.

.

.

.

.

. _8«. I
«8. .... 38 88. on. _88 8.. «N. _88 _88.

 

flu

 

 

 

 

 

3._z_3._m|888_z_88

 

. .42. 8.8» 58.8: ..z

 

 

 

 

 

 

fine 3.flg>§flaolmflq

 

 

.8 82.58.88 8. 8.8 .855 :8 8..?» 888... 288.30 889... a... 3.3.

 

129

Figure 4-14. 33 Pareto Solutions for 3 NPS Pollution Problems for Sub-watershed 178 (4 Years)

 

 
   

 

1000 ‘
“2‘.“
ag: .. ------------------------------------------------
Egg 0 .‘ .A‘. ....x- -. _A A—
g . ,
tn 2 a

m _IJJ—l—Ll

+Sediment +Nitrogen +Phospborus

Figure 4-15. 33 Pareto Solutions for 3 NPS Pollution Problems for Sub-watershed 178 (Year 0)
200

 

 

Sum (4 Years) of
Normalized Obj.
Function Values

 

 

 

+ Sediment + Nitrogen + Phosphorus

Figure 4—16. 33 Pareto Solutions for 3 NPS Pollution Problems for Sub-watershed 178 (Year 1)
200

 

 

Sum (4 Years) of
Normalized Obj.
Function Valuea

 

 

 

+Sediment +Nitrogen +Pbosphorus

Figure 4-17. 33 Pareto Solutions for 3 NPS Pollution Problems for Sub-watershed 178 (Year 2)
200

 

 

Sum (4 Years) of
Normalized Obj.
Function Val

 

 

 

+Sediment +Nitrogen +Phosphorus

Figure 4-18. 33 Pareto Solutions for 3 NPS Pollution Problems for Sub—watershed 178 (Year 3)
200

 

'é

 

Function Values
0

u

Sum (4 Years) of
Normalized Obj.

 

 

 

130

Figure 4-19. 33 Pareto Solutions for Sediment for the Sub~watershed 178 (Year 0 - 3)
8 300 >
=

 

  

Normalized Objective Function Val
8
I
I

 

13579111315171921232527293133
IlYearO IYearl flYearZ DYear3

Figure 4-20. 33 Pareto Solutions for Nitrogen for the Sub-watershed 178 (Year 0 - 3)

 

Normalized Objective Function Values

1 2 3 4 5 6 7 8 9101112131415181718192021222324252827282931313233

 

IIIYearO IYearl BYearZ DYear3
Figure 4-21. 33 Pareto Solutions for Phosphorus for the Sub-watershed 178 (Year 0 - 3)
600 *
= u
E l

  
 

 

   
 

  

       
 

 

‘°° """" i! ““““ E! “““““““““““““““ 3|!
300 a-- iii—~- . —— FE“§§~?§“E

is is 255;: -';? ' #523515 I H i? H 2;: 552 -s
u II II II II III n u I: II n H u II III

12 3 4 5 8 7 8 9101112131415161718192021222324252627282933313233
I'lYearO IYearl EYearZ DYear3

   

   

 

 

 

0

131

Figure 4-22. 22 Pareto Solutions for 3 NPS Pollution Problems for Sub-watershed 231 (4 Years)

 

 

 

 

 

 

Sum (4 Years) of
Normalized Obj.
Function Values

 

+ Sediment +Nitrogen +Phosphorus

)

Figure 4-23. 22 Pareto Solutions for 3 NPS Pollution Problems for Sub-watershed 231 (Year 0)
200

 

 

 

Sum (4 Years) of
Normalized Obj.
Function Values

 

 

 

+ Sediment + Nitrogen + Phosphorus

Figure 4-24. 22 Pareto Solutions for 3 NPS Pollution Problems for Sub-watershed 231 (Year 1)

 

 

 

 

zoo-
§§§ 100 .....................................................
§ 3 0 AW ._ _A.
gag 400 1 ------------ 5 ............ 1 .1 ........... 16_ ___________ 2 _1___
(15:2 .200 ......-

Figure 4-25. 22 Pareto Solutions for 3 NPS Pollution Problems for Sub—watershed 231 (Year 2)
200

 

 

 

 

Sum (4 Years) of
Normalized Obj.
Function Values

 

 

 

+ Sediment + Nitrogen +Phosphorus

 

zoo

 

 

 

Sum (4 Years) of
Normalized Obj.
Function Values

 

 

132

Figure 4-27. 22 Pareto Solutions for Sediment for the Sub-watershed 231 (Year 0 - 3)

 

 

 

 

 

Normalized Objective Function Values

12 3 4 5 6 7 8 910111213141516171819202122
llYearO IYearl flYearZ DYear3

Figure 4-28. 22 Pareto Solutions for Nitrogen for the Sub-watershed 231 (Year 0 - 3)

 

o
g -100
g -200
E -300
g -4oo
E -soo
3 4300
a -700

-aoo
z

12 3 4 5 6 7 8 910111213141516171819202122
llYearO IYearl flYear2 DYear3

Figure 4—29. 22 Pareto Solutions for Phosphorus for the Sub-watershed 231 (Year 0 - 3)
60° ,

 

 

Normalized Objective Function Values

12345 678 9101112131415161718192021a

DYearO IYearl flYear2 DYear3

 

133

Figure 4-30 The 7 Channel Grids Identified as NPS Pollution Problem for Sub-watershed 178

 

Water Features
Subwatershed 178
[:3 AGNPS Grids

  

134

Figure 4-31 The 15 Fields Identified as Fields that Have Been Changed for Sub-watershed 178

 

135

Figure 4-32 The 10 Channel Grids Identified as NPS Pollution Problem for Sub-watershed 231

 

Water Features
Subwatershed 231
|:] AGNPS Grids

136

Figure 4-33 The 15 Fields Identified as Fields that Have Been Changed for Sub-watershed 231

 

137
4.3.2 Pareto Solutions

There are 33 and 22 Pareto solutions for each sub-watershed with outlet at grid
number 178 and 231 respectively. The same conclusions on the type of NPS pollution
problems can be drawn. Both sets of Pareto solution have similar trends to the set from at
the watershed as a whole. That is, both sub-watersheds have phosphorus and sediment
NPS pollution problem but no nitrogen NPS pollution problem. Moreover, the
magnitude of the phosphorus and sediment seems to be greater than the overall
watershed. The average objective function values for both sub-watersheds (Table 4-8 and
4—9) are higher than those in the watershed results (Table 4-2). The same count statistics
were summarized in Table 4-10 and 4—11 for both sub-watersheds. Phosphorus-sediment
proportion relationship also can be found in Figure 4-14 to 4—18 and Figure 4~22 to 4-26.
There are no extra information that can be derived from the Pareto solutions of these two
sub-watersheds. We only confirmed that these two sub-watersheds have sediment and

phosphorus NPS pollution problems.

4.3.3 The Iteration Processes

With 500 extra runs on both sub-watersheds, 7 and 10 channel grids were
identified as NPS pollution problem grids (or segments) as shown in Figure 4—30 and 4»
32 for each sub-watershed. Each sub-watershed also identified 15 fields that have
changed management practices (Figure 4-31 and 4-32). The corresponding summaries of
the landscape characteristics are presented in Table 4—12 to Table 4-15.

In sub-watershed 178, a new channel grid, grid number 357, was identified (Table

4—12). There are 7 more fields, field id 89, 91, 121, 122, 154, 1375 and 1380, identified

138

Table 4-10 A Simple Count Statistics of Pareto Solutions Based on Their Values for
Sub-watershed 178

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

NPS Problem No. of Normlized Objective No. of Average Values Relative to the
And Function Values that are Magnitude of the Goal that are
'fime Periods Negative Positive <= 1.0 < 1.1 < 1.25 < 1.5 >= 1.5
.. 4 years 6 27 6 9 11 7 o
8 YearO 3 3o 3 12 10 5 3
.§ Year 1 10 23 10 11 5 3 4
3 Year2 10 23 10 7 7 5 4
Year 3 4 29 4 10 11 8 o
4 Years 33 o 33 o o o o
a YearO 33 o 33 o o o o
g Year 1 33 o 33 o o 1 o o
'2 Year2 33 o 33 o o o 0
Year 3 33 o 33 o o o o
m 4 Years 0 33 o o o 4 29
E YearO o 33 o o o 19 11
g. Year 1 o 33 o o o 12 11
g Year2 o 33 o o o 12 11
a. Year3 o 33 o o o s 28

 

 

 

 

 

 

 

 

 

 

as fields that have changed management practices (Table 4—13). Fields id 88, 90, 124,
and 1379 have been changed more than 50 times for these 500 extra runs. These fields
are areas that contributed most to the potential NPS pollution problems. Phosphorus
related problem occurred 321 times, which may indicated that sub-watershed 178 is more
vulnerable to phosphorus related problems.

In sub-watershed 231, 4 new channel grids, grid number 316, 505, 537, and 568,
were identified (Table 4-14). There are 7 more fields, field id 169, 172, 958, 959, 1033,
1035, and 1725, identified as fields that have changed management practices (Table 4-
15). Field id 174, 1044, 1725 are fields potentially will contribute most to the NSP
pollution problem since they have been changed more than 50 times for these 500 extra
runs. 395 of the 500 extra runs occurred as sediment related problems. This indicated

that sub-watershed 231 is more vulnerable to sediment related problems.

139

Table 4—11 A Simple Count Statistics of Pareto Solutions Based on Their Values for
Sub-watershed 231

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

NPS Problem No. of Normalized Objective No. of Average Values Relative to the
And Function Values that are Magnitude of the Goal that are

“me Periods Negative Positive <= 1.0 < 1.1 < 1.25 < 1.5 >= 1.5
.. 4 years 0 22 o 4 16 2 o
5 Yearo o 22 o 1 17 4 0
.§ Year 1 s 17 5 8 s o o
3 Year 2 o 22 o 9 11 2 0
Year 3 o 22 o o 13 9 o
4 Years 22 o 22 o o o o
a Year 0 22 o 22 o o o o
g Year 1 22 o 22 o o o o
52 Year 2 22 o 22 o o o 0
Year 3 22 o 22 o o o o
m 4 Years 0 22 o o 4 18 o
E YearO o 22 o o 1 20 1
'g. Year 1 o 22 o o 14 8 o
_2 Year 2 o 22 o o 9 13 o
‘3‘ Year 3 o 22 o o 2 18 2

 

 

 

 

 

 

 

The management practices also have some changes. In sub-watershed 178, no till
is required for some fields (field id 88, 137, 162, and 1379) in order to achieve a better
solution. However, all management practices changes in sub-watershed 231 are still from
conventional tillage to conservation tillage. For the 4 fields in sub-watershed 178, the
times applied for no till are shown in Table 4—16. The no till change requirement means
that applying no till on these four fields will be more effective in reducing related NPS
pollution problems than applying conservation tillage on other fields that still uses
conventional tillage. All changes (52) occurred on field 137 are from conventional tillage
to no till. This field probably is the “hottest spot” in sub-watershed 178.

The 500 extra runs provided similar information, i.e., phosphorus and sediment

pollution problem for both sub-watersheds, with additional identified channel grids and

 

140

Table 4-12 Summaries of Identified Channel Grids that Have NPS Pollution Problem for

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

the Sub-watershed 178
AGNPS Model Input Variables
Grid No. Slope Slope Channel Channel Soil Soil
(%) Length (ft) Slope (96) 811163113” (96) Erodibility Texture
195 1.7 289 1.75 29 0.19 4
268 5.2 244 2.50 33 0.22 1
309 3.0 272 3.00 33 0.17 4
356 3.8 261 2.50 33 0.27 2
357 3.9 260 4.00 33 0.26 l
381 3.0 270 2.45 33 0.28 2
449 2.4 281 2.33 29 0.28 2
Average] 3.3 268 2.65 32 0.24 2‘
, Average! 3.0 273 2.89 31 0.23 2‘
‘-'I‘hevalueisthedorninant soil texture. where 1 =sand,2=silt.3=clay,and4=peetlmuck.
‘ 1 - Average value of 7 identified channel grids. 2 - Average value of all channel Eds 5183).

 

 

Table 4—13 Summaries of Identified Fields that Have NPS Pollution Problem for the

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sub-watershed 178
AGNPS Model Input Variables Iteration Information
FE)“ Elev. Slope :25; Soil . Soil Texture. Problem Times
(m) (‘5) (ft) Erodibilrty 1 2 4 Type Identrfied
88 279 3.95 259 0.24 1 1.6 47.8 1.3 P 53
89 288 4.52 253 0.24 5.3 2.0 P 13
90 286 5.2 244 0.22 8.0 P 72
91 283 4.05 259 0.25 5.8 4.9 P 15
120 294 3.57 263 0.27 49.6 28.2 SIP 28
121 287 3.45 265 0.26 22.5 13.8 S 9
122 298 3.23 268 0.26 37.8 42.5 SIP 15
124 287 3.87 260 0.26 20.7 20.9 S 60
125 292 3.6 263 0.26 19.6 18.0 S 3
137 277 2.51 279 0.21 12.9 13.6 P 52
154 296 2.62 277 0.27 13.8 8.2 S 1
162 302 2.5 278 0.28 14.2 29.4 S 67
1359 277 2.45 277 0.25 22.2 9.6 8.0 P 2
1379 275 1.53 294 0.16 32.0 P 86
1380 274 2.05 284 0.16 22.7 35.1 P 24
Average' 286 3.27 268 0.24 253 238 90 mp 3211)
Average2 284 2.6 279 0.25 8606 12102 1353 7 S 179 S
* - 'lhe value is the acreage of each soil texture in the field, where l = sand, 2 = silt, 3 = clay, and 4 = peat/muck.
“ - The NSP problem type, where S = sediment and P = phosphorus.
l - Aver—age value of 15 identified fields. 2 - Average value of the watershed, except that soil texture is the sum.

 

 

 

 

141

Table 4-14 Summaries of Identified Channel Grids that Have NPS Pollution Problem for

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

the Sub-watershed 231
AGNPS Model Input Variables
Grid No. Slope ( %) Slope Channel Channel Soil Soil
Legth (ft) Slope (%) SideSlope 1%) Erodibility Texture
275 1.7 279 2.33 29 0.24 2
316 5.2 297 1.43 29 0.28 2
431 2.7 275 1 1.00 50 0.30 2
471 2.4 279 2.00 33 0.30 2
505 1.7 294 2.80 33 0.30 2
522 2.9 276 2.00 29 0.32 2
523 2.9 275 6.00 29 0.32 2
534 3.4 268 2.00 29 0.16 4
537 1.8 290 1.83 29 0.31 2
568 2.3 279 1.36 29 0.27 2
Average‘ 2.43 231 3.3 32 0.28 2‘
Average: 3.0 273 2.89 31 0.23 2'
‘-The value isthedominant soil texture, where 1 asand, 2=si1t, 3 =c1ay, and4=peatlmuck.
‘ 1 - Ame value of 10 identified channel Es. 2 — Average value of all channel Eds (1832.

 

Table 4-15 Summaries of Identified Fields that Have NPS Pollution Problem for the
Sub-watershed 231

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

AGNPS Model Input Variables Iteration Information
”Ed Elev Slope SM” Soil 5°“ Tm" Problem 132-fie
. (m) (96) Lug)“ Erodibility r 2 4 Tyre d
1 14 292 3.72 263 0.30 63.8 S 85
166 297 2.15 283 0.29 21.3 90.5 S 42
169 302 1.95 289 0.31 34.9 S 23
172 291 2.35 280 0.30 14.9 S 4
952 301 2.38 281 0.32 1 19.4 S 9
953 299 2.76 278 0.32 17.3 S 35
958 304 2.20 289 0.31 28.2 S 18
959 304 2.90 289 0.32 9.6 S 12
1033 280 1.50 298 0.28 23.8 S
1035 281 1.43 298 0.28 2.7 18.0 S 1
1044 277 2.49 282 0.24 1.8 34.9 S 55
1648 289 2.70 275 0.22 8.9 10.0 3.1 P 8
1716 290 3.64 263 0.17 6.9 6.0 P 48
1719 287 3.55 270 0.17 5.1 P 49
1725 294 2.1 1 282 0.26 37.1 0.9 S 109
Avera el 293 2.52 281 0.27 41 502 15 2 P 105 P
Average2 284 2.6 279 0.25 8606 12102 1353 13 S 395 S
“ - The value is the acreage of each soil texture in the field, where 1 = sand, 2 = silt, 3 = clay, and 4 = peat/muck.
” - The NSP problem type, where S = sediment and P = phosphorus.
‘ 1 - Average value of 15 identified fields. 2 - Average value of the watershed, except that soil texture is the sum.

 

142

Table 4-16 Management Practices Changes of the Sub-watershed 178.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Field Management Practices Changes

ID. Conventional to Conservation Conservation to No till Times Changed ‘
88 50 3 53
89 13 0 13
90 72 0 72
91 15 0 15
120 28 0 28
121 9 0 9
122 15 0 15
124 60 0 60
125 3 0 3
137 0 52 52
154 l 0 l
162 53 14 67
1359 2 0 2
1379 54 32 86
1380 24 0 24
'Tohfl 399 101 500

 

watershed 178 seems to have more serious phosphorus pollution problems than sub-

associated fields that have NPS pollution problems. The results showed that sub-

watershed 231. Moreover, more effective management practices method, i.e., no till,

started to be applied to certain fields in sub-watershed 178 in order to obtain better

solutions.

4.4

Discussion of the Relevance to the Research Questions

The research uses a watershed based water quality model, AGNPS, for evaluation

of agricultural NSP pollution problems. In addition, a goal based Min-Max strategy is

embedded in a Pareto multicriteria optimization framework. By combining these two

approaches, the three-step NSP pollution management difficulties were examined. The

 

 

143

following sections discuss the relevance of the research approach, method, and findings
to the research questions under investigation. The discussion emphasizes the contrasts in

the field-based approach vs. the watershed-based approach.

4.4.1 Critical Area Identification

Critical areas can be simply defined as areas that contributed pollution most to the NPS
pollution. They (the critical areas) are defined relative to the water quality problem. A
NPS pollution problem is defined by the pollutants contributed to the watercourses or
water bodies. The overall water quality goal is to restore and maintain the chemical,
physical, and biological integrity of the Nation’s waters. The successfulness of meeting
this water quality goal is measured in the streams or lakes, not on the land. When it
comes to defining NPS pollution problems, one has to focus on impacts to the aquatic
system.

Moreover, NPS pollution has a diffuse, dynamic, and stochastic nature. Pollutants
are mostly transported through water movement on the land or in the water. Water flow
is typically dictated by the landscape topography and a watershed is usually used as a
management unit for water related issues. This terrestrial boundary draws the spatial
scope of the NPS pollution problem.

This research starts by setting three NPS pollution goals in the streams. The NPS
pollution transport is calculated by a watershed based distributed model. The model
requires a set of natural landscape and human management variables together with a
water transport algorithm to predict NPS pollutant concentrations in overland flow to

watercourses. By comparing results predicted by models with the water quality goals

 

144
desired in the stream, one can find the stream segment most impacted by the terrestrial

NPS pollution problem. Then, the upstream areas, where the water comes from, for the
stream segment can be delineated. Only natural resources and human activities within the
corresponding upstream areas will contribute to the seriousness of the problem. Such
rationale was evaluated in this study through delineating water flow paths in the
watershed.

A source accounting program is used to evaluate the contribution from different
spots (grids) for the upstream source area. The “spots” that contributed most can be
identified as the critical areas relative to the identified NPS pollution problem in the
stream segment. These rationale and procedures were implemented in the research
modeling methodology. Once the critical areas are identified, the activities on the critical
areas can be changed. Since the geographical unit for management change is typically a
farm field, the identified critical “spots” were recorded as fields. In addition, the process
continued and was repeated after each change to find other “spots” in the study area. The
NPS pollution problem identified may differ in type and location each time and also the
identified critical area.

In this way, critical areas within the whole watershed are identified. The same
approach was applied to two upstream sub-watersheds. Each critical area is always
attached to a specific N SP pollution problem. Through multiple runs, the occurrence of
identified channel segments and fields also serve as another information set for ranking
identified critical areas.

However, the results are better interpreted as “potential” critical areas for two

reasons. First, multiple runs are based on a set of “hypothetical initial conditions and

145
applied changes”. The initial setup contains hypothetical “scenarios” that were randomly

created, i.e., randomly assigned crop system and management practices to each field
constitute one scenario. This approach has the advantage of allowing the exploration and
discovery of alternative sets when combined with the optimization iteration process.
Secondly, each change applied represents a new scenario with a slight modification to the
old scenario. The effectiveness of applicable changes (better management practices) is
quantified through comparisons of model results, i.e., the relationship between new and
old scenarios. Although the results measure quantitatively different values, it is more
meaningful and reliable in terms of the “qualitative direction of goodness”. The
implication is that the results (identified critical areas) reflect landscape responses to a set
of hypothetical conditions, which are not the current existing situations, for the problem
under investigation.

Critical areas identified in this research have one common landscape characteristic
- they are very close to the watercourses. This finding matches with our knowledge on
the importance of the “buffer” zone to the NPS pollution control. The average values of
slope and soil erodibility of these critical fields are slighter higher than the average values
of the watershed and the slope length are shorter than the average. No other common
landscape characteristics are evident. However, there are several interesting phenomena
found in Table 4-5, 4-13, and 4—15.

First, identified fields with phosphorus problems seem to be located on finer soil
texture, i.e., peat or muck. For example, the last four fields as well as field id 88 and 137
in Table 4—5 are identified as having phosphorus problems. These fields all contain

certain amounts of peat/muck soils. When fields identified as having phosphorus

 

146
problems but contain no or little peat/muck soils (the first four fields in Table 4-13), their

slope values (4.0 to 5.2) are much higher than the average slope value of the watershed.
Secondly, fields identified as having sediment problems seem to have either higher values
of soil erodibility (near or over 3.0 in Table 4-15) or steep slope (around 3.5 in Table 4—
13). These two phenomena also demonstrated a match with our knowledge on sediment
and phosphorus pollution issues.

In all, the identified critical areas (fields) in this study represent potential hazard
or high risk areas that may not be an accurate representation of the current existing
situation. The critical areas represent the natural landscape risk setting plus hypothetical
field management scenarios. These results can be used as a filter in comparison with the
current existing situation. For example, if the existing management practices of these
fields are using conventional tillage method, these fields are most likely causing the NPS
pollution problems and will need further conservation treatment. There are also fields
that applied conservation treatment in the hypothetical scenarios but they may not be the
case in the current existing situation. These fields also need special attention. In
addition, the results can also serve as a good basis for investigation of “what it” scenarios.
Managers can now have a more focused and manageable set of fields that are identified as
“hot spots” with related NPS pollution problems. Thus, decision makers can work with

the land owners and start to explore alternatives for these critical areas.

4.4.2 Best Management Practices
The BMPs concept has undergone a revolutionary change in recent years. The old

BMP concept focused on management practices that are most effective to N SP pollution

 

147
reduction at the edge of a field. Under the old BMPs concept, the “best” is defined

relative to the management unit, i.e., a farm field, where management practices are
applied. However, the best NPS pollution control for a field does not guarantee that the
NPS pollution problems can be solved in the watercourses. The diffuse nature of NPS
pollution is not considered.

The new BMP concept has shifted toward a more systematic and ecological
perspective. The “best” component of BMPs is now defined in terms of the NPS
pollution problems, that is, the management unit (a watershed) of water pollution issues.
To be qualified as a BMP, the impact of the BMP on multiple types of NPS pollution
problems with spatial and temporal dimensions must be considered. Moreover, the
effectiveness of BMPs is evaluated based on the reduction of NPS pollution problems in
the watercourse. This new BMP concept has led to the need for vastly improved
approaches for BMPs selection.

This research uses a two-step procedure for BMP selection. First, BMP selection
is based on the type and seriousness of NPS pollution problems. Each identified critical
area has information on the NPS pollution problems that exist. The Min-Max
optimization strategy used also considers the seriousness of the NPS pollution problem
relative to the watercourse goal desired. Then, a gradual approach of management
practice changes is used to recommend BMPs for an identified critical area. The gradual
approach basically changes management practices from conventional tillage to
conservation tillage, conservation tillage to no till, no till to pasture or grassland, and

pasture or grassland to idle.

148
Secondly, the effectiveness of BMPs is compared for fields within a critical area

(an AGNPS grid) identified by the AGNPS model. For example, an AGNPS grid may
contain several fields, each has a crop rotation system and associated management
practices. One field may be in conventional tillage and another is conservation tillage.
Based on the gradual approach, the conventional tillage field will be changed into
conservation tillage and the conservation tillage will be changed into no till. Then, the
results are compared at the channel grid, where the NPS pollution problem occurred. If a
field management change from the conventional tillage to conservation tillage can reduce
the NPS pollution problem more than another field management change from the
conservation tillage to no till, the former one is recommended and implemented because
it is more effective to the identified NPS pollution problem.

This two-step procedure is applied whenever a critical area is identified in the
optimization iteration process. In the first 1500 iterations for the whole watershed, all
recommendations are changes from conventional tillage to conservation tillage. This
indirectly implies that field management change from conventional tillage to conservation
tillage is more effective than other field management changes (such as from conservation
tillage to no till) for these fields than other fields. For example, an identified critical area
AGNPS grid may contain more than one field. If the hypothetical field management
change of one field in the AGNPS grid is from conventional tillage to conservation tillage
and another field (in the same AGNPS grid) is from conservation tillage to no till. Both
field management changes were tested and compared in terms of their effectiveness in the
identified channel grid, i.e., from a watershed perspective. A field-based approach will

measure the effectiveness of both field management changes at the edge of the fields and

 

149
may not yield the same results as the research method used. It is the spatial location of

fields relative to the landscape setting that should also be used to determine what BMPs
should be installed and where.

The 500 extra runs for the two sub-watersheds recommended that some fields be
changed from conservation tillage to no till. Since these 500 extra runs for the sub-
watershed are continued from the previous 1500 runs for the whole watershed, their
results can actually be merged. This further analysis indicated that no till treatment for
some fields is necessary in order to obtain better reduction of the NPS pollution problem.
It also shows that the approach is capable of selecting more effective management
practices to achieve better results.

The research approach demonstrated that BMPs selection can be based on a
watershed perspective. Also, BMPs selection can take the spatial location into
consideration when the effectiveness of alternative BMPs is measured relative to the

watercourses NPS pollution problems.

4.4.3 Area-wide Comprehensive Planning

Planning is a complicated process. It is an art of handling the unknown and
uncertainty in the future. NPS pollution planning has a great degree of unknowns and
uncertainty. One of the major difficulties is that it is a semi-structured problem solving
process. Managers or planners have to borrow the best known tools and technologies
currently available to address these issues. In such cases when the causal relationship of a
problem is not clear, the decision making process is extremely difficult and can be easily

jeopardized or criticized.

 

150

NPS pollution planning also requires area-wide and comprehensive
considerations. The area-wide component needs to take into account the spatial
dimension of the diffuse nature of NPS pollution. The logical way is to manage the
problem from a watershed perspective. The comprehensive component involves the
holistic dimension of the NPS pollution problem. A holistic view of the NPS pollution
problem will require a systematic approach to investigate the parts of the system.

When a field based approach (in general use today) is used for NPS pollution
control, there are several problems. First, the goal or measure is not appropriately
matched with the problem. For example, the sediment and soil erosion problem is
measured as the amount of soil loss from a field. Secondly, the critical areas are
identified without spatial consideration by field based models, such as USLE. There
often is little connection between the Source (a field) and the destination (watercourses).
Thirdly, when the BMPs are oriented toward reduction of NPS pollution at the edge of
the field, their overall effectiveness is questionable. If NPS pollution planning is based
on such misunderstanding of the measures and misperception of the goal, we will be
inefficient and ineffective in using our time and resources.

This research used a totally different approach to support the NPS pollution
control. The goal is defined in the watercourses where the NPS pollution problem can be
measured. A watershed model is used as the base to compute the amount of NPS
pollution through water transport. Critical areas are identified relative to the difference
between the goal and the problem in the watercourses. BMPs selection is based on the

effectiveness of NPS pollution reduction in the watercourse. This information will

151

provide decision makers with a more appropriate and clearer setting for planning of NPS
pollution control.

The research results contain three important pieces of information vital for
planning. First, the critical fields identified provide what the related NPS pollution
problems are and where they might occur in the streams and on the land. Secondly,
recommended BMPs are associated with the level of NPS pollution control that can be
achieved relative to the goal desired. Thirdly, and most important, a set of hypothetical
scenarios is obtained based on the landscape and management setting. .

The hypothetical scenarios can be treated as hypothetical solutions, i.e., a set of
hypothetical plans for NPS pollution control. These plans can be presented in a very
simplest format — maps. Figure 4—34 presented four maps, one for the current existing
situation (same as Figure 3-9) and the others are example scenarios derived in the Pareto
optimal solutions for the corresponding watershed and sub-watersheds. These maps are
classified by the C011), which contains the crop rotation and management practices for
each field. The corresponding tabular data are presented in Appendix D. With the goal
achieved, that is, the Pareto optimal solutions (presented in Table 4-1, 4-2, and 4—6 to 4—
9), decision makers can start to add “values” with these solutions.

The most straight forward “added value” is a monetary measurement. For
example, the economic return can be compared with the management cost among these
scenarios. Then, a ranking of these scenarios can be based on the cost-effectiveness as
well as environmental goals achieved. In addition, there are many other uses for the
results. For example, the results can also be used as a discussion medium to foster

community watershed committee agreement upon a plan. They can also be used in a

152

reverse way to plan based on the available resources. Furthermore, managers can put
these plans into the social setting of the watershed community to examine the
acceptability of the plans. In summary, the results represent a scientifically based and
environmentally sound basis for decision-makers to assist with economically viable and

socially acceptable solutions for NPS pollution control.

4.4.4 Water Quality Programs and NPS Pollution Policy

This research has provided a unique approach in addressing the three-step
procedure of NPS pollution management from a watershed perspective. It also provides a
new direction for solving the NPS pollution problems. Many existing water quality
programs may benefit from using this research approach. This section first discusses
three programs, Total Maximum Daily Load (TMDL), Conservation Reserve Program
(CRP), and Environmental Quality Incentives Program (EQIP), in relation to this research
followed by potential changes in future NPS policy setting.

CRP has been a major USDA conservation policy for reducing NPS pollution
from the agricultural sector. The criteria used in the current CRP concern mostly the soil
erosion problem. It has been a voluntary program since its inception. Stricter rules have
been imposed for qualification by the Farm Bills in the past decade. The program has
claimed to be successful in protecting the Nation’s water quality. However, the measure
is the amount of soil loss reduced from farm fields that have participated in the program.

Another program, EQIP - managed by NRCS USDA, focuses on implementation

of BMPs for conservation. The EQIP was introduced in the latest Farm Bill (1996). The

153

Figure 4—34 Examples of Pareto Optimal Solutions for Field Management Operations

":""""“II

I! 'F'in .
'..v4. .

(9'
: In?

272$. ‘

3.
3.’

. F
ll , ..I
I}.

II'l.

-
a;
J?

v' I

 

'I'lrcdim'hltionafiu’SOOextnnmsformb-mshed 178. ThedistributionafiuSOchtnnmsforsub—wmhedfll.

154

basic idea is to apply appropriate management practices on target critical areas to prevent
negative environmental extemalities. It is also a voluntary program with cost-share.
When farmers apply for the program, a detail conservation plan has to be created based
on defined problems by the Soil Conservation Districts and NRCS. Then, a set of
selection criteria based on the effectiveness of the proposed plan in ameliorating with the
identified problem is develOped. Proposed plans have to compete with each other in
order to obtain the financial aids. In addition, monitoring of implementation and
measurement of achievements are required.

TMDL has been used for point source regulation. EPA and other State
governments are now investigating or applying the TMDL as a regulatory tool for NPS
pollution control. Although the exact rules are not clear, the idea is moving toward a
performance, watershed-based management approach. Specific limits of NPS pollutants
at a specific location in the watercourses will be used as the goal that needs to be
accomplished for a geographic unit - watershed. Communities within the watershed will
need to work together in order to achieve the goal. One potential mechanism is water
quality trading between point source and non-point source sectors. However, there are
still many issues to be resolved before this approach can be applied. Issues include
trading credit exchange, market mechanism, standards desired, monitoring, and
measurement.

By applying this research approach to these programs, implementation of EQIP
can be better evaluated with scientific basis from critical area identification and BMPs
selection. The TMDL concept can be reinforced with a spatially based management

strategy for comprehensive planning. The CRP can better target the problem areas by

155
knowing their associated problems. Furthermore, the goal setting oriented approach also

provides a common ground for stakeholders to communicate and/or resolve
discrepancies. Then, the added value can be merged into the planning process to ensure
the soci-economical consideration for the NPS pollution management.

These programs reflect the current NPS pollution management strategy set by the
policy makers. They are based on our knowledge of the NPS pollution problem and
society’s value of the importance of the problem. This research approach has explored
and demonstrated a new method for dealing with the NPS management procedure. The
research findings also improve our knowledge in this area. With the new knowledge and
experience we gain from science, there is and has been a gradual change in the “value
system” of the NPS pollution control domain. When the NPS pollution problem is
evaluated with the watershed perspective, critical areas and management practices are
defined differently than those under the field-based point of view. Our value on the NPS
pollution is transforming into a spatial oriented perspective.

Our perception of the agriculture sector’s responsibility for NPS pollution control
is changing. Agriculture’s exemption from many environmental responsibilities is
changing. The agriculture industry has undergone a dramatic structure change from
small, individual family farms to big, business-oriented enterprises. The change will
affect value in our society as a whole, which will reshape and redefine the next generation
NPS pollution policy. Any potential NPS pollution source contributor, including farmers,
will be held responsible for reducing NPS pollution. Society may eventually force the

agricultural sector to accept a regulatory management policy for NPS pollution control.

 

156
The research findings can help reinforce the importance of the spatial (watershed)

concept on NPS pollution issues while society is in a value change transition. The
research approach demonstrated a workable method to facilitate planning and

management for the new era of NPS pollution control.

CHAPTER 5

FOCUS GROUP EVALUATION

5.1 Overview

Focus groups research has been one of the effective methods used to evaluate
social programs for years. It also has been applied to other areas of sciences. Focus
group research is one type of qualitative research that gathers data, information, and
perceptions of a specific area of interest through a small group of people in a permissive
environment.

This research has several difficulties in evaluating research questions with the
traditional method. First, the research approach studies land use changes (mainly farm
management practices) that can not be implemented in the real world. The control
variables in the study are actually uncontrollable in the real world. Instead, they can only
be hypothetically pr0posed or changed in the study. Secondly, these hypothetical changes
were evaluated through a mathematical water quality model - AGNPS. The AGNPS
model can only estimate quantitatively accurate results of the NPS pollution, so are any
other existing water quality models. However, model developers and some practitioners
believed that model results are useful and provide important information in relative
comparison of alternative scenarios. Thirdly, NPS pollution has a stochastic nature. NPS

pollution is especially important in “big” storm events, which may not happen during the

157

158
study period. This means that evaluation data may not be available even with good

monitoring efforts. The above difficulties have led this research to use an expert focus
group approach to evaluate the research approach, methods, and results.

The study will assemble a small group of experts in water quality related issues.
This group will be provided with basic and detail information on this research. Then, a
special session will be held to allow discussion of the research approach, method, and
results among the group members. Comments will be documented and reported in the
dissertation. Finally, these comments will be analyzed relative to the objectives of the

proposed research.

5.2 Procedures

The first step of focus group evaluation is selection of participants. Since this
study is a fundamental research, participants need a certain level of experience or
knowledge in related research topics. Also, the participants must have some practical
experience in conservation practices, their application and adoption. These participants
can be classified as “experts” in water quality related issues. The major criterion is
whether a candidate possesses such knowledge. After consulting with several people, a
list (Table 5-1) of 10 potential experts was identified.

A cover letter and three attachments (Appendix B) were sent to invite identified
experts to attend the focus group evaluation session. The cover letter described the
purpose. A four-page attachment contains basic research information, which briefly
described the research topic, research question, research approach, research methods,

research results, and potential research contribution. The attachment provides a general

159

description of the research subject and was designed to get their attention and interests.
The second attachment is an introduction about focus group evaluation. The purpose is to
give people background information for the focus group evaluation. It emphasizes the
objectives and the process of the focus group evaluation since they may not be familiar
with the focus group process. The third attachment is a schedule form to identify the

potential date and time for the focus group session.

Table 5-1 A List of Identified Experts for Focus Group Evaluation

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Name Organization Attended
Jon Suppnick MDEQ, Surface Water Quality Division, NPS Unit Yes
Ruth Shaffer NRCS/MDEQ Water Quality Liasian No
Mark Hansen MSUE Ingham County Agent Yes
Maurice Vitosh MSU Dept. of Crop and Soil Science Yes
Eckhart Dersch MSU Dept. of Resources Development No
Pat Lindemann Ingham County Drain Commissioner No
. . Farmer forrnerl worked on the S camore Creek
T‘m D‘c‘z Water Q(uality Pioject) y Y“
Lois Wolfson MSU Institute of Water Research Yes
Joe Ervin MSU Institute of Water Research Yes
. NRCS forrnerl worked on the S camore Creek
Brran MacMaster Water ((luality grojec t) y No

 

Once the information was sent out, follow-up phone calls and e-mails were made
hmediately to initialize personal contact. Other channels of communication were also
used when a person had not responded. Eventually, all invitees had expressed their
interest in attending the focus group evaluation session. 9 out of 10 persons also returned
the schedule form. Unfortunately, not all the 9 persons had a common date and time for

the session. Another effort was made to contact those that had conflicting schedule trying

160
to increase the participant rate. The final result allowed 7 experts to attend the focus

group session.

Once the date and time was determined, detailed information was sent out to the 7
participants. This information included draft capies of Chapter 3 — Research Methods
and Chapter 4 - Research Results. E-mail was send to all participants before the date of
the focus group session to remind them of the event. 6 people attended the focus group
evaluation session. The individual that did not attend had a conflicting meeting that
developed after agreeing to attend the focus group session. The focus group evaluation
session was held in the conference room of the Institute of Water Research at Michigan
State University. The location was very convenient to all participants. The session lasted
for three hours in the afternoon of late January 27th.

My major advisor, Dr. J on Batholic, was asked to be the moderator of the focus
group evaluation session. The focus group session starts with a presentation on the
research approach and results. The evaluation session was tape recorded with agreement
from all participants. The presentation lasted longer than expected. A lot of time was
spent on clarifying the research approach and communicating the interpretation of
research results. Participants really needed a certain degree of understanding about the
subject before in-depth discussion could occur. The first half of the session involved
trying to answer questions that participants had. The second half was handled by the
moderator in facilitated discussion. At the end of the session, we notified all participants
that there would be a discussion report distributed to them in a few weeks for comments.
We also emphasized that all comments will be made anonymous in the dissertation and

only participant’s names would be listed.

161

A report on the discussion of the focus group evaluation was sent out after the
focus group session. The purpose of the report was to truthfully document what had been
discussed in the focus group session. It needed to be validated by the participants.
Additionally, it allowed participants to add more comments. Corrections were made from
feedback returned. The next section (Section 5.3) contains the focus group evaluation

report including most comments and feedback.

5.3 Focus Group Session Report

The following reports the discussion, comments, and feedback from the focus
group evaluation. The discussion shifted from one t0pic to another randomly, the report
however tries to group all comments into subject categories. The related research
methods or results precede each subject. Then, participant’s comments follow in Italic
font style. Furthermore, these comments do not reflect consensus of all participants, but
they do represent what most people agreed upon. The subsequent section, 5.4, will

further analyze these comments in terms of the research questions.

5.3.1 Model Input

There are 23 variables in the AGNPS model. These 23 variables are divided into
two categories: fixed and variable. The fixed variables are those that are not likely to
change with human activities. The second group of variables will change for each model
run based on field management operation, a crop stage, and a crop type.

The major concern on model input data is the applied amount of phosphorus

fertilizer. The phosphorus amount used in this research is too high compared with the

162
average amount. For example, 90 to 110 lbs/acre of phosphorus fertilizer was used in

this study for corn, but the average amount of phosphorus fertilizer for the state is
approximately 50 lbs/acre or less.

Furthermore, the same amount of phosphorus fertilizer is used for all three crop
stage, which may be not appropriate. The phosphorus fertilizer should only be applied
once for a growing season.

The research should use the soil nutrient (especially phosphorus) level plus the
actual or average fertilizer nutrient application rate for each actual cr0p to calculate the
nutrient loading. It is important to take the existing soil nutrient level into account in the
AGNPS model input.

There are a few big livestock operations in the study area (how many and how
big .7). That was not considered in this study. If this factor was incorporated, these
feedlots are another major source of nitrogen loading and it may change the results.

A lot of croplands have tile drainage system. The drainage direction of these
underground drainage systems may not match with the surface topography. This may
change the drainage path, but probably is a minor factor, which could be ignored. The

drainage path for the surface water would not change as a result of subsurface drainage.

5.3.2 The AGNPS Model

The AGNPS model is run for each crop stage for each of the 4-year rotation years.
The same “big” storm event, 24—hour, 25-year frequency, is used for each run. The
AGNPS model has three modules: SCS TR-SS as the hydrological model, USLE as the

soil erosion model, and CREAMS as the nutrient model.

163
It is not clear how the AGNPS model calculates the phosphorus loading. It

should consider the soil phosphorus level together with the applied fertilizer phosphorus.
Phosphorus loading is not in proportion to the applied phosphorus fertilizer. It is more
likely that when the soil phosphorus level reaches a certain amount ( approximately 300
lbs/acre), the added phosphorus fertilizer will move with the runoff It is more related to
soil phosphorus level than with the fertilizer phosphorus applied.

Participants indicated that it may be better to use the hydrograph of a storm event
to evaluate the problem. Moreover, it will be interesting to look into storm events in
other seasons, such as late spring (March or April) and late summer (September).
Problems may not have occurred only during the three-stages that the research used,
although these three stages are probably the high risk periods during a crop growing
season or a year.

The model developers often ask users not to believe or use the actual predicted
values, but use the difi'erences between two predicted values instead. These
“suggestions " are not quite acceptable.

The nitrogen is a wealatess of the analysis due to the tile drainage system that
delays the peak nitrogen concentration in the streams. The approach is also heavily
based on the AGNPS model, which can be criticized. Furthermore, the model mechanism
for nitrogen transport via surface runofi' only does not match what actually occurs in

Sycamore Creek watershed, subsurface delivery of nitrogen.

5.3.3 Goals and Objective Functions

The three general goals set in this research are:

 

164
l. the total nitrogen in any stream segments should be less than 10 ppm;

2. the total phosphorus in any stream segments should be less than 1 ppm; and

3. the total sediment of any stream segments should be less than 1,000 mg/l.

The basic spatial unit of objective function is a stream segment, i.e., a AGNPS model
grid. There are two sets of objective functions. The first set contains all channel grids for
each stage for a 4-year crop rotation year. There are 6,648 objective functions in this set.
This set of objective functions is used to find the worst channel grids of any problems in
the Min-Max Optimization procedure. The second set combines stages for a year and all
channel grids together. There are 15 combined objective functions, one for each of the 4_
year rotation and one of the three problems (sediment, nitrogen, and phosphorus) plus one
for each of the three problems for the sum of four year.

Most participants indicated that 10 ppm for N and 1 ppm for P seemed to be very
high standards for streams or lakes. For nitrogen, 10 ppm nitrate concentration is the
standard for drinking water, and 1 ppm phosphorus concentration is generally used as
eflluent or discharge standards. The receiving waters, i.e., streams and lakes, could use
lower standards for both pollutants since these pollutants will be diluted in the receiving
waters. The 1,000 mg/l of suspended solids goal for sediment is high too, but it probably
is fine for a “big " storm event.

EPA is currently working on setting water quality standards for suspended solids
in receiving waters based on protecting fish. The State is undergoing similar efl‘orts to
set standards for difi'erent usage and purposes in the receiving waters. These standards

should provide better guidance for such goal setting studies in the future.

165
The research also set the above three goals for every channel grid or segment in

the watershed. This means that all channel grids were given equal weight no matter if it
is at the head water or the downstream area. One attendee expressed that this may not
be true from a resource management point of view. For example, the headwaters may
not be suitable for fish habitat but the downstream channel segments are more likely to
have a good fisheries. Therefore, setting goals for upstream channel segments will be
unnecessary from the resource management perspective.

Another comment on the goal setting is that it is more meaningful to use a storm
event average to evaluate the seriousness of a problem. Most fish or plants are not going

to be afiected by only a few hours of peak flow during a storm event.

5.3.4. Best Management Practices (BMPs) Selection

The BMPs selection uses a uniform, gradual approach, that is, from conventional
tillage to conservation tillage, no tillage, alfalfa hay, and idles, because each BMP, in
general, can reduce all three problems that the research is trying to investigate. The
gradually approach also provides a means to evaluate the effectiveness of each BMP.

The research approach used the same prescription for all problems. However, if
a problem is nutrient related, it could have applied nutrient management on the field
instead of a no tillage, which may not be the most efiective practice for the problem.

BMP selection is generally known and straightforward if we know where the

problems are. To apply a BMP to a problem should not be that difi‘icult.

 

166
The actual prescribed BMPs may have already been in place. The results may not

be able to provide any guidance to the non-point source pollution management in the

watershed.

5.3.5 Research Findings

Overall, this research concluded that the study area has sediment and phosphorus
non-point source pollution problems from the Pareto optimal solution set. Detailed
optimization records showed that these problems only happened on approximately a
dozen channel grids. The study also identified approximately 20 fields that contributed
most to the problems after thousands of optimization runs. These fields are not scattered
everywhere within the study area. Most of these fields are nearby the streams or drains.
The similarity of the landscape characteristics of the study area seems to be the reason for
the results.

The research approach seems to be valid for analyzing the non-point source
pollution problem. However, the input data can be improved to reflect the “real ”
situation of the fields and watershed. It must take into account the tile drainage,
livestock operation, soil test of nutrient level, and the actual fertilizer application rate for
difi'erent crops. Furthermore, only with ground truth checking, can one evaluate the
accuracy of whether the “hot spots ” identified in this research are “real ” sources of the
problems.

For ground truth checking, one also needs to investigate areas other than the
identified fields. Without comparing the identified fields with other fields, it is still

impossible to evaluate whether the identified fields are the problem sources.

167
The results can be checked by running the real storm events in the past few years.

Then, one can compare the AGNPS model results with the monitoring data to get a better

sense on the AGNPS model outputs.

 

Most of the identified problem areas (fields) are located nearby the streams
and/or drains. Such results are encouraging because the findings match with our
knowledge on the non-point pollution issues. Obviously, the farther away from the
streams the less chance or importance the pollutant will get to the streams.

The magnitude of the phosphorus problem does not match with the magnitude of
the sediment problem. This seems to be in conflict with our knowledge of the relationship
between phosphorus and sediment. Usually, these two problems are in proportion to
each other. That is, the higher the sediment yields or concentration, the greater the
phosphorus problem will be. The conflict of research results may be an artifact of the
standards chosen, i. e., 1.0 ppm for phosphorus and 1,000 ppm for sediments.

The tile drainage system is the major cause for the delay of the peak nitrate
concentration during a storm event. From the monitoring data, the peak amount of
nitrate concentration in the streams lags behind the peak flow. This may also be another
reason why nitrogen pollution was not a problem in the research results.

It will be interesting to summarize the characteristics of the fields that were
identified and changed, that is, applied with an improved field management operation.
After that, one can compare the similarities and difi'erences of their characteristics.
Further analysis, such as factorial analysis, soil types comparison, and land slope

comparison, can be used to analyze their characteristics to derive general summaries.

168
These summaries should be potentially usefitl for or applicable to other watersheds that

do not have all the required data and/or resources.

5.3.6 Social-Economic and Planning (Implementation)

The results have provided a solid base for managers to evaluate the non-point
source pollution problem from a watershed perspective. By analyzing the characteristics
of the watershed and identified fields, the approach can be used on other watersheds to
provide a quick and simple evaluation of agricultural non-point source pollution problem.
The final plan should also consider the social-economic aspect of watershed management.
These social-economic factors are keys to successful implementation of the plan.

The current water quality program in the Sycamore Creek watershed does not
have any special criteria. The program is basically ofl'ered to anyone. It is up to the
individual grower to sign up to receive a subsidy for implementing BMPs.

The social-economic aspect mentioned in the research is very important. It
afiects whether farmers are going to participate in the actual realization
(implementation). It is also important to get the support from the producer
organizations. If some of the outputs can be simplified in color graphics, they will be
very efl'ective for educational purpose to communicate with producer organizations to
contribute more resources for problem solving.

Farmers may be pretty scared of the BBQ in temrs of being told that they are
responsible for the results. They are being watched because they are the major

contributors. It may be better if this information comes from the producer organizations.

169
Moreover, it may be better to show a map with fields that meet certain criteria

than a map with identified problem fields. Although theremay be more areas that met
the criteria, there are still benefits overall in the watershed if areas not identified in the
results also apply BMPs.

The same strategy can be used in groundwater. With a set of conditions that may
result in contaminated wells, we can go into another area and use the same scenarios to
identity contaminated wells more quickly and easily. In this way, we are not identifying
“people " that may contribute to the problem.

Most farmers are willing to change if we (MS UE) can tell them the “right” things
to do. If their areas are more vulnerable to particular problems or have high risk or
potential for certain issues, most farmers will apply BMPs to reduce the problem,

especially with incentive payment.

5.4 Evaluation Conclusions

To analyze focus group results is not an easy task due to the “discussion” nature
of the focus group research. One has to derive conclusion from the words and interprets
the perception of the discussion. In addition, the focus group session does not force
members to reach consensus of any issues or questions. It is an open-ended result that
must be analyzed with neutral mind. This is especially difficult for evaluating the
research questions in this study. Not all the sections in Chapter 4 were made available to
the focus group for discussion. Mostly, they saw raw or preliminary results that require

group members to use their judgement. The following summarizes the focus group

I70

evaluation with pros and cons or advantages and disadvantages relevant to the research

objectives and questions.

5.4.1 Research Approach and Method

In general, the group seemed to agree that the research approach used is
appropriate or valid. However, there are several flaws in the research method that can be
improved or avoided. First, the model input data, especially nutrient and phosphorus
applied, needs to better reflect the current existing situation. Secondly, the research
excluded feedlot contribution, which is not appropriate if their scale is large. Thirdly,
some participants still do not agree that models can be used for evaluation of NPS
pollution problem because they can not provide accurate estimates of quantitative values
for NPS pollution control. These problems also reflect the data intensive disadvantage of
such a research approach.

The optimization process of the research approach seemed to be hard to
understand by most participants. After more explanation to clarify their questions and
further discussion on this subject, the group gradually accepted the optimization
framework. However, some attendants still do not fully understand the Pareto concept
and how it is related to the research.

Although the three goals set for sediment, nitrogen, and phosphorus in this
research may be too strict and it may not be necessary to set these goals for all stream
segments, most experts liked the idea of using goal setting in NPS pollution evaluation.
Attendants with extensive experiences in monitoring water quality problems especially

advocated such an approach.

 

171

5.4.2. Research Questions and Results

The research questions are based on the three-step NPS pollution management
procedure. Each was discussed loosely in the focus group session. First, some experts
seemed to like the research results in identified critical areas, especially that these areas
are close to the watercourses. However, no experts are willing to conclude that these
areas are the “hot spots” for the watershed. Instead, the discussion shifted to how to
verify the research results with field survey or in-stream monitoring. Additionally,
participants recommended that further investigation can be performed on these identified
critical areas to review their similarities and dissimilarities. Hopefully, some rules can be
generalized and applied to analyze NPS pollution problems in other watersheds.

Secondly, BMPs selection method was criticized during the focus group session
due to the gradual approach used. Participants thought that the most effective
management practices can be applied directly to identified critical area if the related NPS
pollution problem is known. The importance of the spatial consideration of the BMPs
selection process was not received more attention by the group. The purpose of the
gradual approach is to investigate the BMPs not only for their effectiveness but also their
spatial locations relative to the landscape setting. The old BMPs concept, i.e., the field-
based perspective, is still deeply embedded in participants’ mind setting.

The Pareto optimal solutions for use in NPS pollution management was not
discussed widely. Participants are still pondering the meaning of the Pareto concept. In
contrast, the group spent much time on discussing water quality implementation issues

for the Sycamore Creek watershed. Attendants agreed that current voluntary approach for

172
the agricultural NPS pollution reduction is not very effective. The “first come first

served” method is not really an effective way to use available limited resources. It lacks
of problem focus and is not likely to resolve the problem. They all wished to have good
tools and accountable knowledge that can be used to improve BMP implementation of the
Sycamore Creek water quality program.

The economical and social factors were emphasized by some group members
again and again. They believed that these are the keys to successful implementation of
any NPS pollution control plans.

It was also mentioned that the map format results are a very powerful medium.
Map outputs can be used in a variety of ways by the managers. Several participants
started to discuss how to use them to convince farmers that their farms are the problem
areas that need special attention without raising hostility or controversies.

In general, the group members expressed their interests in continuing research in
the direction of using watershed based approach for NPS pollution management. All
participants agreed that NPS pollution is a complex, open-ended problem that still needs

lots of work both in research and practical applications.

 

CHAPTER 6

CONCLUSIONS

Agricultural Non-point source pollution was identified as the major threat to the
Nation’s water quality goals in the past Clean Water Act and its Amendments. Many
efforts have been initiated to reduce agricultural NPS pollution; however, EPA still
reported NPS as the primary source for impaired streams and lakes. Current NPS ,
pollution management strategy requires a three-step procedure: critical area identification,
Best Management Practices (BMPs) selection, and area-wide comprehensive planning.
However, several difficulties exist in applying this three-step procedure for NPS pollution
control. Additionally, NPS pollution control presently utilizes a field-based approach,
which does not take into account the diffuse, dynamic, and stochastic nature of NPS
pollution. Decision-makers will need a different approach to more effectively address
these difficulties and problems.

The objective of this research was to address the difficulties of the three-step NPS
pollution control procedure. A watershed based optimization approach was used to
examine these problems. This study used a watershed based water quality model,
AGNPS, for estimation of NPS pollution on a watershed basis. The model is capable of
predicting NPS pollution in the stream and on the land through its hydrological, nutrient,

and soil erosion modules. This research divided NPS pollution problems into two

173

 

174
systematic perspectives, aquatic and terrestrial. A goal setting idea to define the NPS

pollution problems in the streams and lakes was used for the aquatic system through a
Min-Max optimization strategy. The Min-Max strategy first finds the maximum
deviation from the goal (defined as the limit of a NPS pollution problem) in the channel
segments. Then, the upstream areas of the identified channel segment are delineated, that
is, the sub-watershed of the terrestrial system. A source accounting technique is applied
to examine the contribution from different fields within the sub-watershed, so that, “hot
spots” can be identified. Finally, BMPs is embedded into the optimization process to
reduce (minimize) the identified NPS pollution problem.

The optimization process continues to find better and better solutions by repeating
the process. Each iteration identified four items of information: the most serious
problem, the high risk streams segments, the hot spot field, and the management practices
needed. The result is a set of objective functions that are relative to the goal defined.
These objective functions were compared with each other to obtain a set of Pareto
optimal solutions. A set of Pareto optimal solutions has an equal importance
characteristic; that is, an objective can not be enhanced without diminishing another
objective. At the end of the optimization process, the set of Pareto optimal solutions
represents the achievement accomplished relative to the goal defined. Each Pareto
solution also provides a unique scenario on which alternative field management practices
were determined. Then, decision-makers can further evaluate these solutions with other
social-economical consideration to fine-tune a comprehensive management planning for

NPS pollution control.

175
The research is aimed at providing a workable process for the three-step NPS

pollution management procedure. Preliminary results and detailed methods were
presented to a group of water quality experts. This focus group evaluated the
appropriateness of the research approach, the validity of the research methods, and the
acceptability of the research results. The focus group discussion was reported and
analyzed in terms of the relevance to the research questions under investigation.

The evaluation results form focus group session is general in nature. One of the
reasons is that focus groups do not force consensus agreement, rather focus groups

provide focused discussion. The perceptions from the discussion are summarized below:

(1) participants are not willing to draw conclusions on whether the identified critical
areas are really “hot spots” in the study area but they think that the results are
encouraging because the areas are adjacent to the watercourses;

(2) participants pointed out some flaws in the research method, especially in model data
input part, but agreed that the overall research approach seems valid;

(3) participants want more information on NPS pollution management but are not totally
willing to commit to model results since current models can not accurately estimate
quantitatively reliable pollution values;

(4) participants agreed that a watershed-based approach is better than a field-based
approach, but their mind is not completely accepting of this newer approach - for
example, BMPs are discussed in a field-based manner without taking the spatial

location and water transport mechanism into consideration;

176
(5) participants liked the goal setting investigation approach but don’t know what the

final goal could be or should;

(6) participants admitted that current water quality programs (including the Sycamore
Creek Water Quality Project) lack the “targeting” component because we either don’t
have enough knowledge or available resources for finding the “hot spots”;

(7) participants advocated the importance of social-economical considerations in NPS
pollution management.

This research has built a framework for the three-step NSP pollution management
procedure. This research effort demonstrated not only encouraging results but also a
well-defined process that can be potentially applied to NPS pollution control problem.
Existing water quality programs, such as CRP, EQIP, designated river use, and TMDL,
can all potentially use this research approach to more efficiently manage available
resources and prevent negative environmental extemalities.

However, there are many unanswered puzzles with NSP pollution issues. One
that needs special and urgent attention is watershed-based modeling. The NPS pollution
process can be addressed through distributed water quality models. But, without more
accurate estimation of NPS pollution from the land and to the watercourses, the present
uncertainty may scare away decision-makers from using “better tools. This research
would be benefited with better verification of such model. Since the basis of this research
relies heavily on the water transport mechanism of NPS pollution from terrestrial to
aquatic systems, improvements in modeling accuracy can promote the validity and

applicability of this research method to the real world.

177
Another research need is to validation the research approach used in this study

with monitoring data and/or other reliable measures. Watershed management differs
from field-scale management in the extent of geographic coverage. The resources needed
are much more extensive and often require a longer period of time. However, in order to
increase the degree of confidence in the optimization modeling approach, it will require
more reliable monitoring.

In summary, the watershed concept has been available for several decades but has
not really been associated with NPS pollution problems until the 1990’s. Many managers
still do not understand the concept of a watershed perspective. The old or field-based
mindset is still influencing our perceptions and decisions related to water quality
problems. While the new movement toward watershed management for water related
issues is gradually emerging and may become the default “value” system for the society
in the future, we also need to explore and develop different approaches to accommodate
the new concept. This research has demonstrated the application of an optimization
framework to better understand and redefine the three-step NPS pollution management
procedures. It sets the fundamental basis for practitioners to further discover its
applicability in NPS pollution management. Researchers can continue to nurture related
studies; and ultimately be a part of the movement toward the new era of the NPS

pollution management.

APPENDICES

178
Appendix A

Basic Crop Rotation Information

DATE STAGE OPERATION
~4Il
4116 ~

Sll ~
5/16 ~
4116 ~

Sll ~ 5/1
5/16 ~

~ 411
4/16 --

5/1 ~
5/16 ~
4’16 ~

5!] ~ 5115
5/16 ~

COIDO2 ~ 4115
CCSBdW 4/16 ~
Conventional 5/1 ~
Tillage 5/16 ~
~ 411
4116 ~
5!] ~
5116 ~
~ 471
4116 ~
5!] ~ 5/15
5116 ~
4716 ~
5]! ~
5/16 «-
~ 4/1
4/16 ~
Sll ~
5/16 ~

and
and

and
and
ammonia

and field cultivate

drillmd

4116 -
5/1 ~
5/16 -
~ 411
4716 ~
511 ~
Sll6 ~
~ 4/1
4116 ~
5/1 ~
5116 ~
~ 411
4116 ~

None

Inversion

 

"OUNHOUNF'OUNHOUNV‘OUN-‘OUNHOUN—‘OUNHOUNflONN—‘OUNF'OUNF‘O

 

5l1~5ll
5116-
~4ll
4116~
511 ~5/15
5116 ~
~4115
4Il6~
Sll ~5/l
5/l6-
~4/15
4116 ~
5!] ~ 5/15
5/16 ~

~4/15
4/16~
5/1-5/15
5/l6~
~4/15
4116 ~
5/1-5/15
5/16 ~
~4/15
4116~
Sll ~5115
5ll6~
~4115
4116 ~
5/1~5/15
5115 ~

~4115
4116 ~
5/1 ~5/l
5/l6~
~4115
4116-
5/1 ~5/15
Sll6~
~4115
4ll6~
5/1~5/15
5ll6~
~4115
4116 ~
5/l~5/15
5/l6~
~4115
4116 ~
5!] ~5/15
5/16~
~4/l5
4ll6~
Sll ~5/l
5116 ~
~4ll
4/16 ~
511-

179

N—OwN—OWN-‘OUNflOwN—OUN—OUN—OwN—OWN—‘OUN—OwN—OUN—OWN-‘OUN—‘OUN

None
None
Inversion

field

cultivate and

None
None

None

Row

None

Row

None
None
Inversion

None
None
None
None
None
Inversion

None
Inversion

None
None
Inversion

None
None
None
None

None
None
Row

None

Row

None
None

and
and

and
and

and

and

and drill

and

and

 

 

t0
4

Alfalfa-Hay

5116 ~
~ 411
4116 ~
5/1 ~ 5115
5/15 -
~ 4115
4116 ~
511 ~ 511
5/16 -

180

wN—‘OwN—‘OU

 

Basic Field and Land Use/Cover Data

18 1
Appendix B

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Field Land Use/Cover Land UselCover Crop Year Ofl’set
Identification Description Identification

l ALF/GRASS HAY COIDOS 3
2 ALF/GRASS HAY COID08 3
3 ALF/GRASS HAY COID08 3
4 ALF/GRASS HAY COID08 3
5 CORN, CONV. COID01 0
6 CORN, CONV. COID01 0
7 ALF/GRASS HAY COID08 3
8 ALF/GRASS HAY COID08 3
9 CORN, CONV. COID01 0
10 CORN, CONV. COID01 0
ll SOYS, DRILL, CONV. COID02 2
12 ALF/GRASS HAY COID08 3
l3 CORN, CONV. COID01 0
14 IDLE NCID01 -1
15 CORN, CONV. COID01 0
16 IDLE NCIDOI -1
17 CHISELED COIDOl 0
l9 IDLE NCIDOI -1
20 CORN, CONV. COID01 0
21 WHEAT, CONV. COIDOl 3
22 IDLE NCIDOI -1
25 IDLE NCIDOl -1
26 IDLE NCIDOI -l
27 CORN, CONV. COIDOl 0
28 SOYS, ROW, CONV. COIDOl 2
29 SOYS, ROW, CONV. COID01 2
30 CORN, CONV. COIDOl 0
41 CORN, CONV. COIDOl 0
42 WHEAT, CONV. COIDOl 3
43 SOYS, DRILL, CONV. COID02 2
44 CORN, CONV. COID01 0
49 IDLE NCIDOl -1
56 CORN, CONV. COID01 O
57 CORN, CONV. COIDOl O
58 SOYS, ROW, CONV. COID01 2
73 CORN, CONV. COID01 0
79 CORN, CONV. COID01 0
8O CORN, CONV. COID01 0
81 GRAVEL PITS NCID07 -1
82 ALF/GRASS HAY COID08 3
83 CORN, CONV. COID01 0
84 CORN, CONV. COIDOl O
86 ALF/GRASS HAY COID08 3
87 ALF/GRASS HAY COID08 3
88 CORN, CONV. COIDOl 0

 

 

 

182

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

89 ICORN, CONV. COIDOl 0
9o ICORN, CONV. COIDOl o
91 CORN, CONV. COID01 o
92 IDLE NCIDOl -1
93 IDLE NCIDOl -1
95 ALF/GRASS HAY COID08 3
98 IDLE NCIDor -1
99 ALF/GRASS HAY COID08 3
100 IDLE NCIDOI -1
101 IDLE NCIDOI -1
102 IDLE NCIDOI -1
103 WHEAT, CONV. COID01 3
104 ALF/GRASS HAY COID08 3
105 CORN, CONV. COID01 o
106 CORN, CONV. COIDOl o
107 ALF/GRASS HAY COID08 3
109 ALF/GRASS HAY COIDO8 3
110 ALF/GRASS HAY COIDO8 3
111 DRYBEANS, CONV. COID03 2
112 DRYBEANS, CONV. COID03 . 2
114 DRYBEANS, CONV. COIDO3 2
115 WHEAT, CONV. COIDOl 3
116 WHEAT, CONV. COID01 3
117 CORN, CONV. COIDOl o
118 CORN, CONV. COIDor o
119 ALF/GRASS HAY COID08 3
120 WHEAT, CONV. COIDOl 3
121 WHEAT, CONV. COIDor 3
122 CORN, CONV. comm o
123 SOYS, DRILL, CONS. COIDO6 2
124 ICORN, CONV. comor o
125 ICORN, CONV. COIDOl o
126 ICORN, CONV. COID01 o
127 SOYS, ROW, NT com 2
128 ALF/GRASS HAY COID08 3
129 ALF/GRASS HAY COID08 3
130 CORN, CONV. COID01 o
131 CORN, CONv. COIDOl o
132 WHEAT, CONV. COID01 3
133 WHEAT, CONV. COIDOl 3
134 WHEAT, CONV. COIDOl 3
135 IDLE NCIDor -1
136 CORN, CONV. COID01 o
137 CORN, CONV. COIDOl o
138 IDLE NCIDOl -1
139 WHEAT, CONV. COIDOl 3
140 WHEAT, CONV. COIDOl 3
141 WHEAT, CONV. COIDOl 3
142 WHEAT, CONV. COID01 3
143 WHEAT, CONV. COIDOl 3
144 HERBACEOUS NCIDor -1
145 SOYS, ROW, CONS. cows 2

 

 

 

183

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

146 CLOVER, CONV. COIDO3 3
147 WHEAT, CONV. COID01 3
148 WHEAT, CONV. COID01 3
149 ALF/GRASS HAY COID08 3
150 WHEAT, CONV. COID01 3
151 CORN, CONV. COIDOl 0
152 IDLE NCID01 -1
153 CORN, CONV. COID01 O
154 CORN, CONS. COIDOS O
155 CORN, CONS. COIDOS 0
156 CORN, CONS. COIDOS O
158 CORN, CONV. COID01 0
162 SOYS, DRILL, CONS. COID06 2
163 WHEAT, CONV. COID01 3
165 SINGLE FAMILY NCIDIB -1
166 CORN, CONV. COID01 O
167 CORN, CONV. COID01 0
168 WHEAT, CONV. COID01 3
169 WHEAT, CONV. COID01 3
170 CORN, CONV. COID01 O
171 CORN, CONV. COID01 0
172 CORN, CONV. COID01 0
173 1CORN, CONV. COID01 0
174 lCLOVER, CONV. COID03 3
175 CORN, CONV. COID01 O
176 SOYS, ROW, CONV. COIDOl 2
177 SOYS, ROW, CONV. COIDOl 2
178 SOYS, ROW, CONV. COIDOl 2
179 SOYS, ROW, CONV. COID01 2
180 SOYS, ROW, NT COIDO7 2
181 SOYS, ROW, NT COIDOT 2
182 SOYS, ROW, CONV. COID01 2
183 CORN, CONS. COIDOS 0
184 CORN, CONS. COIDOS 0
185 CORN, CONS. COIDOS 0
186 SOYS, ROW, CONV. COID01 2
187 IDLE NCID01 -l
188 IDLE NCID01 -1
189 IDLE NCID01 -1
190 IDLE NCID01 -l
191 WHEAT, CONV. COID01 3
192 IDLE NCID01 -1
193 IDLE NCID01 -1
946 SOYS, ROW, CONS. COIDOS 2
947 CORN, CONV. COID01 0
948 CORN, CONV. COID01 0
949 WHEAT, CONV. COID01 3
950 CORN, CONV. COID01 0
951 CORN, CONV. COID01 0
952 CORN, CONV. COID01 0
953 CORN, CONV. COIDOl O
954 CORN, CONV. COIDOl 0

 

 

 

 

 

184

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

955 CORN, CONV. COIDOl 0
956 SINGLE FAMILY NCIDl3 -l
957 SOYS, DRILL, CONV. COIDO2 2
958 CORN, CONS. COID05 0
959 CORN, CONS. COID05 0
963 DRYBEANS, CONV. COID03 2
965 WHEAT, CONV. COID01 3
976 SOYS, ROW, CONV. COID01 2
979 ALF/GRASS HAY COID08 3
980 ALF/GRASS HAY COIDOS 3
981 SOYS, ROW, CONV. COID01 2
982 SINGLE FAMILY NCID13 -1
983 CORN, CONV. COIDOl 0
984 CORN, CONV. COIDOl 0
985 CORN, CONV. COIDOl 0
986 CORN, CONV. COID01 0
987 CORN, CONV. COID01 0
988 SINGLE FAMILY NCID13 -l
989 CORN, CONV. COIDOl 0
990 ALF/GRASS HAY COIDOB 3
991 ALF/GRASS HAY COIDOB 3
992 ALF/GRASS HAY COIDOS 3
993 CORN, CONS. COID05 0
994 CORN, CONS. COIDOS 0
995 CORN, CONS. COID05 0
996 CORN, CONS. COIDOS 0
997 SOYS, ROW, CONS. COID05 2
998 CORN, CONV. COID01 0
999 CORN, CONV. COID01 0
1029 NURSERY-TREES NCID03 -l
1030 SOYS, ROW, CONV. COIDOl 2
1031 CORN, CONV. COIDOl 0
1032 CORN, CONV. COIDOl 0
1033 CORN, CONV. COID01 0
1034 CORN, CONV. COID01 0
1035 ALF/GRASS HAY COID08 3
1036 CORN, CONV. COID01 0
1037 WHEAT, CONV. COIDOl 3
1038 CORN, CONV. COIDOl 0
1039 WHEAT, CONV. COID01 3
1040 CORN, CONV. COIDOl 0
1041 WHEAT, CONV. COIDOl 3
1042 WHEAT, CONV. COID01 3
1043 WHEAT, CONV. COIDOl 3
1044 CORN, CONV. COID01 0
1045 WHEAT, CONV. COIDOl 3
1046 SOYS, ROW, CONV. COIDOl 2
1047 PAST -FAIR NCID06 -1
1048 CORN, CONV. COIDOl 0
1049 CORN, CONV. COID01 0
1050 WHEAT, CONV. COID01 3
1051 1CORN, CONV. COIDOl 0

 

 

185

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1052 WHEAT, CONV. COID01 3
1053 SOYS, ROW, CONV. COID01 2
1054 CORN, CONV. COIDOl 0
1055 CORN, CONV. COIDOl 0
1056 CORN, CONV. COID01 0
1057 CORN, CONV. COID01 0
1058 CORN, CONV. COID01 0
1059 CORN, CONV. COID01 0
1060 CORN, CONV. COID01 0
1061 CORN, NT COIDO7 1
1062 CORN, CONV. COID01 0
1063 CORN, CONV. COID01 0
1064 WHEAT, CONV. COIDOl 3
1065 ALF/GRASS HAY COID08 3
1066 SOYS, DRILL, CONV. COID02 2
1067 SOYS, DRILL, CONV. COIDO2 2
1068 CORN, CONV. COID01 0
1069 ALF/GRASS HAY COIDOS 3
1070 SINGLE FAMILY NCID13 -l
1071 SINGLE FAMILY NCIDI 3 -l
1072 ALF/GRASS HAY COID08 3
1073 CORN, CONV. COID01 0
1074 ALF/GRASS HAY COIDO8 3
1075 CORN, CONV. COIDOl 0
1076 CORN, CONV. COID01 0
1077 CORN, CONV. COIDOl 0
1078 WHEAT, CONV. COID01 3
1079 CORN, CONV. COIDOl 0
1080 CORN, CONV. COIDOl 0
1081 CORN, CONV. COID01 0
1082 ALF/GRASS HAY COID08 3
1084 PLOWED COID01 0
1085 SINGLE FAMILY NCIDl3 -1
1086 SINGLE FAMILY NCID13 -1
1087 VEG & FRUIT NCID02 -l
1088 WHEAT, CONV. COID01 3
1089 ALF/GRASS HAY COID08 3
1090 ALF/GRASS HAY COID08 3
1091 ALF/GRASS HAY COID08 3
1092 SINGLE FAMILY NCID13 -1
1093 SINGLE FAMILY NCID13 -1
1094 CORN, CONV. COID01 0
1095 CORN, CONV. COID01 0
1096 CORN, CONV. COID01 0
1097 ALF/GRASS HAY COID08 3
1098 CORN, CONV. COIDOl 0
1099 ALF/GRASS HAY COID08 3
1 100 IDLE NCID01 -1
1101 SOYS, DRILL, CONV. COID02 2
1102 SOYS, DRILL, CONV. COIDO2 2
1103 SOYS, DRILL, CONV. , COID02 2
1 104 IDLE NCID01 -l

 

 

186

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 105 IDLE NCID01 -1
1 106 IDLE NCID01 -1
1 107 IDLE NCID01 -l
l 108 IDLE NCID01 -1
1 109 IDLE NCID01 -l
1 1 10 IDLE NCID01 -1
11 l 1 IDLE NCIDOI -1
l 1 12 IDLE NCID01 -l
l l 13 IDLE NCID01 -1
ll 14 CORN, CONV. COIDOl 0
1 l 15 IDLE NCID01 -l
l l 16 IDLE NCIDOl -l
1117 CORN , CONV. COID01 0
11 18 iCORN, CONV. COIDOl 0
l l 19 PLOWED COID01 0
1120 CORN, CONV. COIDOl 0
1121 ICORN, CONV. COID01 0
1 122 PLOWED COIDOl 0
1 123 PLOWED COID01 0
1124 SOYS, ROW, CONV. COIDOl 2
1125 SOYS, ROW, CONV. COIDOl 2
l 126 SINGLE FAMILY NCID13 -1
1127 SINGLE FAMILY NCID13 -1
1 128 IDLE NCID01 -1
1129 ALF/GRASS HAY COIDOS 3
1 131 CORN, CONV. COID01 0
l 132 ALF/GRASS HAY COIDO8 3
1 133 ALF/GRASS HAY COID08 3
1134 ALF/GRASS HAY COID08 3
1135 ALF/GRASS HAY COIDOS 3
1136 ALF/GRASS HAY COID08 3
1137 SOYS, ROW, CONV. COID01 2
1138 SINGLE FAMILY NCID13 -1
1139 SINGLE FAMILY NCID13 -1
1 140 |CHISELED COIDOl 0
1141 ICORN, CONV. COIDOl 0
1142 SOYS, ROW, CONV. COID01 2
1 143 SINGLE FAMILY NCID13 -1
l 144 WHEAT, CONV. COID01 3
1145 WHEAT, CONV. COID01 3
1146 WHEAT, CONV. COIDOl 3
1149 SOYS, DRILL, CONV. COID02 2
1150 'CORN, CONV. COID01 0
1151 CORN, CONV. COID01 0
1153 FCORN, CONV. COIDOl 0
1 154 IDLE NCID01 -l
1155 iCORN, CONV. COID01 0
1156 ICORN, CONV. COID01 0
1157 ALF/GRASS HAY COID08 3
1158 ALF/GRASS HAY COIDOS 3
1159 ALF/GRASS HAY COID08 3
1160 ALF/GRASS HAY COID08 3

 

 

 

 

187

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1161 ALF/GRASS HAY COID08 3
l 162 IDLE NCID01 -l
1165 WHEAT, CONV. COID01 3
1166 WHEAT, CONV. COID01 3
1167 DRYBEAN S, CONV. COID03 2
1168 DRYBEANS, CONV. COID03 2
1169 CORN, CONS. COIDOS 0
1170 CORN, CONS. COID05 0
1171 iCORN, CONS. COIDOS 0
1172 ALF/GRASS HAY COIDO8 3
1173 SOYS, DRILL, CONS. COID06 2
1174 SOYS, DRILL, CONS. COID06 2
1175 SOYS, DRILL, CONS. COID06 2
1176 ALF/GRASS HAY COID08 3
1177 WHEAT, CONV. COIDOl 3
1178 FCORN, CONS. COID05 0
1179 CORN, CONS. COID05 0
1180 CORN, CONS. COIDOS 0
1181 SINGLE FAMILY NCID13 -l
1182 SINGLE FAMILY NCIDl3 -1
1183 SINGLE FAMILY NCID13 -1
1184 SINGLE FAMILY NCIDIB -1
1186 CORN, CONV. COIDOl 0
1187 CORN, CONV. COID01 0
1188 ALF/GRASS HAY COIDO8 3
1189 OATS, CONV. COIDO4 3
1190 OATS, CONV. COIDO4 3
1191 ICORN, CONV. COID01 0
1192 CORN, CONV. COID01 0
1193 WHEAT, CONV. COID01 3
1194 |CORN, CONV. COIDOl 0
1195 |CORN, CONV. COIDOl 0
1196 |CORN, CONV. COID01 0
1197 |CORN, CONV. COID01 0
1199 |CORN, CONV. COID01 0
1200 |CORN, CONV. COID01 0
1201 lCORN, CONV. COIDOl 0
1202 CORN, CONV. COIDOl 0
1203 ALF/GRASS HAY COIDOS 3
1204 PAST-GOOD NCIDOS -1
1205 ALF/GRASS HAY COIDOS 3
1206 CORN, CONV. COID01 0
1207 CORN, CONV. COID01 0
1208 IDLE NCID01 -1
1209 IDLE NCID01 -l
1210 |CORN, CONV. COIDOl 0
1211 |CORN, CONV. COID01 0
1212 lCORN, CONV. COID01 0
1213 CORN, CONV. COID01 0
1214 WHEAT, CONV. COID01 3
1215 CORN, CONV. COIDOl 0
1216 ALF/GRASS HAY COID08 3

 

 

 

 

 

 

188

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1217 ALF/GRASS HAY COID08 3
1218 SOYS, ROW, CONV. COIDOl 2
1219 ALF/GRASS HAY COID08 3
1220 CORN, CONV. COID01 0
1221 CORN, CONV. COID01 0
1222 WHEAT, CONV. COID01 3
1223 PAST-FAIR NCID06 -1
1224 WHEAT, CONV. COID01 3
1227 ALF/GRASS HAY COID08 3
1228 WHEAT, CONV. COID01 3
1229 CORN, CONV. COID01 0
1230 CORN, CONV. COID01 0
1231 CORN, CONV. COID01 0
1232 CORN, CONV. COIDOl 0
1233 WHEAT, CONV. COIDOl 3
1234 CORN, CONV. COID01 0
1251 WHEAT, CONV. COID01 3
1259 WHEAT, CONV. COID01 3
1260 CORN, CONS. COIDOS 0
1261 CORN, NT COIDOT 1
1262 ALF/GRASS HAY COID08 3
1263 CORN, NT COID07 1
1264 SOYS, DRILL, CONV. COID02 2
1265 IDLE NCID01 -1
1266 IDLE NCID01 -1
1267 CORN, CONV. COIDOl 0
1269 PLOWED COID01 0
1271 PLOWED COID01 0
1272 SOYS, ROW, CONV. COID01 2
1273 SOYS, ROW, CONV. COID01 2
1274 SOYS, ROW, CONS. COID05 2
1275 CORN, CONV. COID01 0
1276 CORN, CONV. COID01 0
1277 SOYS, ROW, CONV. COID01 2
1278 SOYS, ROW, CONV. COID01 2
1279 SOYS, ROW, CONV. COID01 2
1280 CORN, CONV. COID01 0
1281 CORN, CONV. COID01 0
1282 CORN, CONV. COID01 0
1283 WHEAT, CONV. COID01 3
1284 CORN , CONS. COIDOS 0
1285 SOYS, DRILL, CONV. COIDO2 2
1286 CORN, CONV. COIDOl 0
1287 WHEAT, CONV. COID01 3
1288 CORN, CONV. COID01 0
1289 CORN, CONV. COID01 0
1290 ALF/GRASS HAY COID08 3
1291 CORN, CONV. COID01 0
1292 SOYS, ROW, NT COID07 2
1293 SOYS, ROW, NT COID07 2
1294 SOYS, ROW, NT COID07 2
1297 CORN, CONV. COID01 0

 

 

 

189

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1298 CORN, CONV. COIDOl 0
1300 CORN, CONV. COIDOl 0
1301 IDLE NCID01 -1
1307 CORN, CONV. COID01 0
1308 WHEAT, CONV. COID01 3
1323 CORN, CONS. COID05 0
1325 CORN, CONV. COID01 0
1326 CORN, CONV. comm 0
1327 IDLE NCID01 -1
1330 SOYS, ROW, CONS. COID05 2
1331 SOYS, ROW, CONS. COID05 2
1332 IDLE NCID01 -1
1333 IDLE NCID01 -1
1335 IDLE NCID01 -1
1336 CORN, NT COID07 1
1337 CORN, CONS. cows 0
1338 CORN, CONS. COIDos 0
1339 WHEAT. CONV. COIDOl 3
1340 CORN, CONV. COID01 0
1341 CORN, NT COID07 1
1342 CORN, CONV. COID01 0
1343 IDLE NCID01 -1
1345 SOYS, ROW, CONS. COIDos 2
1346 CORN, CONV. COIDOl 0
1347 SOYS, DRILL, CONV. COID02 2
1348 CORN, CONV. COID01 0
1349 CORN, CONV. COID01 0
1350 CORN, CONV. COID01 0
1351 CORN, CONV. COIDOl 0
1352 CORN, CONV. COID01 0
1353 IDLE NCID01 -1
1354 CORN, CONV. COIDOl 0.
1355 SOYS, DRILL, CONV. COID02 2
1356 SOYS, DRILL, CONV. COIDO2 2
1357 CORN, CONV. COID01 0
1358 CORN, CONV. COIDOl 0
1359 CORN, CONV. COID01 0
1360 SOYS, DRILL, CONV. COID02 2
1361 SOYS, DRILL, CONV. comoz 2
1362 IDLE NCID01 -1
1364 CORN, CONV. COID01 0
1367 SOYS, Row, CONV. COID01 2
1368 CORN, CONV. COID01 0
1369 CORN, CONV. COID01 0
1370 CORN, CONV. COID01 0
1371 CORN, CONV. COID01 0
1372 CORN, CONS. COID05 0
1373 IDLE NCID01 -1
1374 SOYS, DRILL, CONV. COIDO2 2
1375 WHEAT, CONV. COID01 3
L 1378 SOYS, ROW, CONV. COID01 2
r 1379 SOYS, ROW, CONV. COID01 2

 

 

 

 

 

190

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1380 CORN, CONS. COID05 0
1381 WHEAT, CONV. COIDOl 3
1382 CORN, CONV. COID01 0
1383 WHEAT, CONV. COID01 3
1538 SINGLE FAMILY NCID13 -1
1554 CORN, CONV. COID01 0
1556 CORN, CONV. COID01 0
1558 CORN, CONV. COIDOl 0
1559 CORN, CONV. COID01 0
1561 CORN, CONV. COID01 0
1562 ALF/GRASS HAY COID08 3
1563 CORN, CONV. COID01 0
1565 CORN, CONV. COID01 0
1566 CORN, CONV. COID01 0
1579 ALF/GRASS HAY COID08 3
1580 ALF/GRASS HAY COID08 3
1582 ALF/GRASS HAY COID08 3
1583 ALF/GRASS HAY COID08 3
1584 CORN, CONV. COID01 0
1586 CORN, CONV. COID01 0
1587 ALF/GRASS HAY COID08 3
1588 ALF/GRASS HAY COID08 3
1589 CORN, CONV. COID01 0
1590 CORN, CONV. COID01 0
1591 CORN, CONV. COID01 0
1592 ALF/GRASS HAY COID08 3
1593 ALF/GRASS HAY COID08 3
1594 CORN, CONV. COID01 0
1595 CORN, CONV. COID01 0
1596 ALF/GRASS HAY COID08 3
1598 CORN, CONV. COID01 0
1637 CORN, CONV. COID01 0
1638 ALF/GRASS HAY COID08 3
1639 CORN, CONV. COID01 0
1640 CORN, NT COID07 1
1641 IDLE NCID01 -1
1642 ALF/GRASS HAY COID08 3
1643 ALF/GRASS HAY COID08 3
1644 IDLE NCID01 -1
1645 ALF/GRASS HAY COID08 3
1646 ALF/GRASS HAY COID08 3
1647 ALF/GRASS HAY COID08 3
1648 SOYS, DRILL, CONV. COID02 2
1649 SOYS, DRILL, CONV. COIDOZ 2
1650 SOYS, DRILL, CONV. COIDOZ 2
1654 ALF/GRASS HAY COID08 3
1655 SOYS, DRILL, CONV. COIDO2 2
1657 CORN, CONV. COID01 0
1670 CORN, CONV. COID01 0
1671 CORN, CONV. COIDOl 0
1672 CORN, CONV. COIDOl 0
1673 SOYS, ROW, CONS. COID05 2

 

 

 

 

 

191

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1674 CORN, CONV. COID01 0
1675 CORN, CONV. COIDOl 0
1676 CORN, CONV. COID01 0
1677 CORN, CONV. COIDOl 0
1678 CORN, CONV. COID01 0
1679 CORN, CONV. COIDOl 0
1680 ALF/GRASS HAY COID08 3
1681 CORN, CONV. COID01 0
1682 CORN, CONV. COID01 0
1683 CORN, CONV. COID01 0
1684 SOYS, DRILL, CONV. COIDOZ 2
1685 WHEAT, CONV. COID01 3
1686 IDLE NCID01 -1
1687 IDLE NCIDOl -1
1688 WHEAT, CONV. COID01 3
1689 SOYS, DRILL, CONV. COID02 2
1690 CORN, CONV. COID01 0
1691 SOYS, DRILL, CONV. COIDO2 2
1692 SOYS, DRILL, CONV. COID02 2
1693 SOYS, ROW, CONV. COID01 2
1694 SOYS, ROW, CONV. COIDOl 2
1695 WHEAT, CONV. COID01 3
1696 OATS, CONV. COIDO4 3
1697 WHEAT, CONV. COIDOl 3
1699 CORN , CONV. COIDOl 0
1700 CORN, CONV. COID01 0
1701 SOYS, ROW, CONS. COID05 2
1702 WHEAT, CONV. COID01 3
1712 CORN, CONV. COID01 0
1713 CORN, CONV. COID01 0
1714 CORN, CONV. COIDOl 0
1715 CORN, CONV. COIDOl 0
1716 HCORN, CONV. COID01 0
1717 CORN, CONV. COID01 0
1718 CORN , CONV. COID01 0
1719 CORN, CONV. COIDOl 0
1720 IDLE NCID01 -1
' 1721 CORN, CONV. COID01 0
1722 CORN, CONV. COIDOl 0
1725 ALF/GRASS HAY COID08 3
1726 CORN, CONV. COIDOl 0
1727 CORN, CONV. COID01 0
1736 SOYS, ROW, CONV. COID01 2
2000 INDUSTRIAL NCIDl l -1
2012 COMM. SERVICES NCIDll -l
2013 INDUSTRIAL NCIDl l -1
2019 OPEN LAND, OTHER NCID01 -1
2022 ORCHARDS NCID04 -1
2024 PERM. PAST. N CIDOS -1
2029 IDLE NCID01 -l
2031 HERB ACEOUS NCID01 -1
2032 SHRUB NCID09 -1

 

 

192

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2052 LAKE NCID10 -1
21 12 MULTI-FAMILY NCID14 -1
2113 SINGLE FAMILY NCID13 -l
2115 MOBILE HOME NCID14 -1
2121 CENT. BUS. DIST. NCIDll -1
2124 SEC. BUSINESS NCIDll -1
2126 IN STITUT. NCID01 -1
2141 AIR TRANS NCID18 -1
2144 ROAD TRANS. NCID12 -1
2145 COMMUN. NCID18 -1
2146 UTILITIES NCID18 -1
2171 OPEN PIT NCID07 -1
2173 WELLS NCID] 1 -1
2193 OUTDR REC. NCID08 -1
2194 CEMETARY NCID18 -1
2412 CENT. HRDWD NCID15 -1
2413 ASPEN/BIRCH NCID15 -1
2414 LOWLAND HW NCID17 -1
2422 OTHER CON. NCID16 -1
2423 LOWL. CONIF. NCID17 -1
2612 SHRUB, SCRUB NCID09 -1
2622 EMERG. NCID10 -1
2999 SHRUB NCID09 -1

 

 

 

193
Appendix C

Land Use/Cover and Field Management Data for the AGNPS Model

Lend UadCover Year C Surface Manning’s Fert. Fert.
Identification Off Stage Factor Cond. Coefficient A B C D COD Fer-t. P N Avail.

0
Com.
81 91

81 91

1
1
1
1
1
1
1
1
1

dflé‘di“

 

194

1
1
1
1
1
1
1
1
1
1
1
1

d‘dddddddddd‘d

 

195

10‘ 0310-4 (010-40)

”dam-4010‘

1
2
3

Theland use/coveridentiflcationandehortdeecrtption.

A number representing the year Offset, 0-1993,1-1994, 2-1995. and 3.1996.
Crop stage (see Table X for details).

Cover and management factor.

Surface Condition Constant.

Manning’s Roughness Coefficient.

Runoff Curve NumbertorHydrOiogicelSOiiGroupA. 8, 0.11116 D respectively.
The amount (lbs/acre) at Phosphorus and Nitrogen applied.

Fertilization

 

 

196

Appendix D

Listings of Field Management Operations of Example Pareto Optimal Solutions

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Field After 1500 Runs for After 500 Extra Runs for After 500 Extra Runs for
ID : I g the Watershed“ Subwatershcd 178" Subwatershed 231‘”
LULC 11) Yr. Off LULC ID Yr. Off LULC ID Yr. Ofl’ LULC ID Yr. Off
1 COID08 3 COID08 3 COID08 3 COID08 l
2 COID08 3 COIDO8 3 COID08 3 COID03 2
3 COIDOB 3 COID08 3 COID08 3 COID06 0
4 COID08 3 COID08 3 COID08 3 COID08 1
5 COID01 0 COID01 0 COID01 0 COID01 2
6 COID01 0 COID01 0 COID01 0 COID07 3
7 COID08 3 COID08 3 COID08 3 COIDO4 2
8 COID08 3 COID08 3 COIDO8 3 COID07 3
9 COIDOl 0 COID01 0 COID01 0 COIDO7 1
10 COIDOl 0 COIDOl 0 COIDOl 0 COIDO4 1
l 1 COID02 2 COID02 2 COIDOZ 2 COIDO2 3
12 COID08 3 COID08 3 COIDO8 3 COIDO3 3
13 COID01 0 COIDOl 0 COIDOl 0 COIDOé 1
15 COID01 0 COIDOl 0 COIDOl 0 COIDOS 2
17 COID01 0 COIDOl 0 COID01 0 COID01 3
20 COID01 0 COID01 0 COID01 0 COIDO4 1
21 COIDOl 3 COID01 3 COID01 3 COID07 3
27 COID01 0 COID01 0 COID01 0 COIDO2 2
28 COID01 2 COID01 2 COID01 2 COID02 2
29 COID01 2 COID01 2 COID01 2 COIDOS 0
41 COID01 0 COIDOl O COIDOl 0 COIDO7 0
42 COID01 3 COID01 3 COID01 3 COIDO7 1
43 COIDOZ 2 COIDO2 2 COIDO2 2 COID01 1
44 COIDOl 0 COID01 0 COIDOl 0 COIDOS 2
56 COID01 0 COID01 0 COID01 0 COID08 3
73 COIDOl 0 COIDOl O COID01 0 COID07 3
79 COID01 0 COIDOl 0 COID01 0 COID07 0
80 COIDOl 0 COID01 O COIDOl 0 COIDO4 0
82 COIDOB 3 COIDO8 3 COID08 3 COIDO7 2
83 COID01 0 COID01 0 COIDOl 0 COID05 0
84 COID01 0 COIDOl O COIDOl 0 COID03 2
86 COID08 3 COID08 3 COID08 3 COID07 0
87 COID08 3 COIDO8 3 COID08 3 COID06 0
88 COID01 0 COIDOS 0 COID05 0 COID05 3
89 COID01 0 COIDOl 0 COID01 O COIDOS 2
90 COID01 0 COID01 0 COID05 0 COID07 3
91 COID01 O COID01 0 COIDOS 0 COIDOZ 3
95 COIDO8 3 COIDOS 3 COID08 3 COIDO2 2
99 COID08 3 COID08 3 COID08 3 COID03 1
103 COIDOl 3 COID01 3 COID01 3 COIDOZ 2
104 COIDO8 3 COID08 3 COID08 3 COID06 1
105 COID01 0 COID01 O COID01 0 COIDO4 3
106 COIDOl 0 COIDOl 0 COID01 0 COIDOl 2

 

 

 

 

 

 

 

 

 

 

197

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

107 COIDOS 3 COID08 3 COIDOB 3 COID02 2
109 COIDOS 3 COID08 3 COID08 3 COIDO7 3
1 10 COID08 3 COID08 3 COID08 3 COIDO6 2
11 1 COIDO3 2 COIDO3 2 COID03 2 COIDOT 0
112 COID03 2 COID03 2 COID03 2 COID02 1
1 14 COID03 2 COIDOS 2 COID05 2 COIDO4 2
1 15 COID01 3 COIDOl 3 COID01 3 COID03 2
l 16 COID01 3 COID01 3 COID01 3 COID06 l
1 l7 COID01 0 COIDOl 0 COID01 0 COIDO3 1
l 18 COIDOl 0 COID01 0 COID01 0 COID08 3
1 19 COID08 3 COID08 3 COID08 3 COID01 1
120 COID01 3 COID01 3 COIDOS 3 COIDO2 1
121 COID01 3 COID01 3 COID05 3 COID05 1
122 COID01 0 COID01 0 COID05 0 COIDOZ 3
123 COID06 2 COID06 2 COID06 2 COIDO4 3
124 COID01 0 COID05 0 COID05 0 COID06 1
125 COID01 0 COID01 0 COID01 0 COID02 2
126 COID01 0 COIDOl 0 COID01 0 COIDO4 2
127 COID07 2 COIDO7 2 COIDO7 2 COID06 2
128 COIDOS 3 COIDO8 3 COIDOS 3 COID03 1
129 COID08 3 COID08 3 COIDO8 3 COIDO6 2
130 COID01 0 COIDOl 0 COID01 0 COID05 2
131 COID01 0 COID01 0 COID01 0 COID02 0
132 COID01 3 COIDOl 3 COID01 3 COIDO3 1
133 COIDOl 3 COID01 3 COID01 3 COIDOZ 1
134 COID01 3 COIDOl 3 COIDOl 3 COID07 l
136 COID01 0 COID01 0 COID01 0 COIDO4 1
137 COID01 0 COID05 0 COIDOS 0 COID05 0
139 COID01 3 COIDOl 3 COID01 3 COID08 0
141 COID01 3 COIDOl 3 COIDOl 3 COIDOl 2
143 COID01 3 COIDOl 3 COID01 3 COIDO7 1
145 COID05 2 COIDOS 2 COIDOS 2 COID01 2
146 COID03 3 COID03 3 COID03 3 COIDO3 3
147 COID01 3 COIDOl 3 COIDOl 3 COID06 1
148 COID01 3 COIDOl 3 COIDOl 3 COIDO6 0
149 COID08 3 COIDO8 3 COIDOS 3 COIDO7 3
150 COIDOl 3 COID01 3 COID01 3 COIDOS 2
151 COID01 0 COIDOl 0 COIDOl 0 COIDOS 1
153 COIDOl 0 COID01 0 COID01 0 COIDO6 2
154 COID05 0 COID05 0 COID05 0 COID01 2
155 COID05 0 COIDOS 0 COID05 0 COIDO8 3
156 COIDOS 0 COIDOS 0 COID05 0 COIDOZ 3
162 COIDO6 2 COID06 2 COID06 2 COID07 1
163 COID01 3 COID01 3 COID01 3 COID05 2
166 COIDOl 0 COID01 0 COID01 0 COID02 3
167 COID01 0 COID01 0 COID01 0 COID03 2
168 COID01 3 COID01 3 COID01 3 COIDO3 3
169 COID01 3 COIDOl 3 COID01 3 COIDO7 l
170 COIDOl 0 COIDOl 0 COID01 0 COID07 2
171 COID01 0 COID01 0 COID01 0 COID06 2
172 COID01 0 COID01 0 COID01 0 COIDO4 2
173 COID01 0 COID01 0 COIDOl 0 COIDOZ 2

 

 

 

 

 

 

 

 

 

 

198

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

175 COIDOl 0 COIDOl 0 COID01 0 COIDO8 3
176 COIDOl 2 COID01 2 COID01 2 COIDO6 2
177 COID01 2 COIDOl 2 COID01 2 COIDO4 2
178 COIDOl 2 COID01 2 COID01 2 COID05 2
179 COID01 2 COID01 2 COIDOl 2 COID05 1
180 COID07 2 COID07 2 COIDO7 2 COID07 2
181 COID07 2 COID07 2 COID07 2 COIDO4 1
182 COID01 2 COID01 2 COIDOl 2 COIDO4 1
183 COID05 0 COID05 0 COID05 0 COID08 2
184 COID05 0 COID05 0 COID05 0 COID05 2
185 COID05 0 COID05 0 COID05 0 COID08 1
186 COID01 2 COIDOl 2 COID01 2 COID07 l
191 COID01 3 COID01 3 COID01 3 COID08 2
946 COIDOS 2 COID05 2 COIDOS 2 COIDO4 2
947 COIDOl 0 COID01 0 COID01 0 COID02 1
948 COID01 0 COID01 0 COID01 0 COIDO4 1
949 COID01 3 COIDOl 3 COID01 3 COIDOZ 0
950 COIDOl 0 COID01 O COID01 0 COIDO6 0
951 COIDOl 0 COID01 0 COIDOl 0 COIDO6 2
952 COID01 0 COID05 0 COID05 0 COID08 l
953 COIDOl 0 COID01 0 COID01 0 COIDOZ 1
954 COID01 0 COID01 0 COIDOl 0 COID07 3
955 COIDOl 0 COIDOl 0 COID01 0 COIDO3 2
957 COIDOZ 2 COID02 2 COID02 2 COID03 1
958 COIDOS 0 COID05 0 COID05 0 COID07 1
959 COID05 0 COID05 0 COID05 0 COID05 3
976 COID01 2 COID01 2 COID01 2 COID03 2
979 COID08 3 COID08 3 COID08 3 COIDOS 1
980 COIDOB 3 COID08 3 COID08 3 COID01 1
981 COIDOl 2 COID01 2 COID01 2 COIDO6 3
983 COIDOl 0 COID01 0 COID01 0 COIDO4 1
984 COID01 0 COIDOl 0 COID01 0 COID03 0
985 COIDOl 0 COID01 0 COIDOl 0 COID02 1
986 COIDOl 0 COIDOl 0 COID01 0 COID08 3
987 COID01 0 COID01 0 COID01 0 COID03 3
989 COID01 0 COID01 0 COIDOl 0 COID03 3
990 COIDOS 3 COIDO8 3 COID08 3 COID07 2
991 COIDO8 3 COID08 3 COID08 3 COID02 0
992 COIDO8 3 COIDO8 3 COID08 3 COID02 0
993 COID05 0 COID05 0 COID05 0 COIDO4 1
994 COID05 0 COIDOS 0 COID05 0 COIDOB 0
995 COID05 0 COIDOS 0 COIDOS 0 COID06 3
996 COIDOS 0 COID05 0 COID05 0 COIDO3 2
997 COIDOS 2 COID05 2 COID05 2 COID08 1
1030 COIDOl 2 COID01 2 COIDOl 2 COID03 1
1031 COIDOl 0 COIDOl 0 COID01 0 COID05 2
1032 COIDOl 0 COID01 0 COID01 0 COIDOT 2
1033 COID01 0 COID01 0 COIDOl 0 COID02 2
1034 COID01 0 COIDOl 0 COIDOl 0 COID06 3
1035 COID08 3 COIDOS 3 COID08 3 COID07 0
1036 COID01 0 COID01 0 COIDOl 0 COID02 0
1037 COIDOl 3 COID01 3 COID01 3 COID06 0

 

 

 

 

 

 

 

 

 

 

199

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1038 COID01 0 COID01 0 COID01 0 COIDO2 3
1039 COID01 3 COID01 3 COID01 3 COID01 3
1040 COID01 0 COID01 0 COID01 0 COID03 1
1041 COID01 3 COID01 3 COID01 3 COIDO3 1
1042 COID01 3 COID01 3 COID01 3 COIDO3 2
1043 COIDOl 3 COID01 3 COID01 3 COID03 0
1044 COID01 0 COID05 0 COIDOS 0 COIDOS 2
1045 COID01 3 COID01 3 COID01 3 COID02 l
1046 COIDOl 2 COID01 2 COIDOl 2 COID08 2
1048 COID01 0 COID01 0 COID01 0 COIDO4 2
1049 COIDOl 0 COID01 0 COID01 0 COIDO4 2
1050 COID01 3 COID05 3 COID05 3 COID07 2
1051 COID01 0 COID01 0 COID01 0 COID06 1
1052 COID01 3 COID01 3 COIDOl 3 COID03 2
1053 COID01 2 COID01 2 COID01 2 COID07 2
1054 COID01 0 COID01 0 COID01 0 COIDOS 0
1055 COID01 0 COID01 0 COID01 0 COIDO7 3
1056 COID01 0 COID01 0 COID01 0 COID03 2
1057 COID01 0 COIDOl 0 COID01 0 COID02 0
1058 COIDOl 0 COID01 0 COID01 0 COID06 2
1059 COID01 0 COID01 0 COID01 0 COIDO7 l
1060 COIDOl 0 COID01 0 COID01 0 COID02 0
1061 COID07 1 COID07 1 COID07 l COID03 1
1062 COID01 0 COIDOl 0 COID01 0 COIDOl 1
1063 COID01 0 COID01 0 COID01 0 COID07 1
1064 COID01 3 COIDOl 3 COID01 3 COID02 1
1065 COID08 3 COID08 3 COIDO8 3 COIDOS l
1066 COID02 2 COID02 2 COIDO2 2 COIDO4 3
1067 COID02 2 COID02 2 COID02 2 COIDOS 2
1068 COID01 0 COIDOl 0 COID01 0 COIDO4 2
1069 COIDOB 3 COID08 3 COID08 3 COID07 l
1072 COID08 3 COIDO8 3 COIDO8 3 COIDO4 3
107 3 COID01 0 COIDOl 0 COID01 0 COID06 2
1074 COID08 3 COID08 3 COIDO8 3 COID07 2
1075 COIDOl 0 COIDOl 0 COID01 0 COIDO4 2
1076 COIDOl 0 COIDOl 0 COIDOl 0 COID02 2
1077 COID01 0 COIDOl 0 COID01 0 COIDO6 1
107 8 COID01 3 COID01 3 COIDOl 3 COIDO4 1
1079 COIDOl 0 COID01 0 COID01 0 COIDO4 3
1080 COID01 0 COID01 0 COID01 0 COID03 2
1081 COIDOl 0 COIDOl 0 COIDOl 0 COID08 1
1082 COID08 3 COID08 3 COID08 3 COIDO4 1
1084 COID01 0 COIDOl 0 COID01 0 COID03 3
1088 COID01 3 COID01 3 COIDOl 3 COID05 3
1089 COID08 3 COIDOS 3 COIDO8 3 COID06 3
1090 COIDOB 3 COIDO8 3 COID08 3 COIDO4 3
1091 COID08 3 COID08 3 COID08 3 COID08 2
1094 COID01 0 COID01 0 COID01 0 COID03 2
1095 COID01 0 COIDOl 0 COIDOl 0 COID05 2
1096 COIDOl 0 COID01 0 COID01 0 COID03 0
1097 COIDOB 3 COID08 3 COIDO8 3 COID06 3
1098 COID01 0 COID01 0 COID01 0 COIDOZ 1

 

 

 

 

 

 

 

 

 

 

200

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1099 COID08 3 COID08 3 COID08 3 COID06 1
1 101 COIDO2 2 COID02 2 COIDO2 2 COID03 2
1 102 COID02 2 COID02 2 COIDO2 2 COID01 0
1103 COID02 2 COID02 2 COID02 2 COIDO3 2
1 1 14 COID01 0 COID01 0 COID01 0 COID01 0
1 1 17 COID01 0 COID01 0 COID01 0 COID08 3
1 1 18 COID01 0 COID01 0 COID01 0 COID07 1
1 l 19 COID01 0 COIDOl 0 COID01 0 COIDO4 2
1 120 COIDOl 0 COIDOl 0 COID01 0 COID05 2
1 121 COIDOl 0 COID01 0 COID01 0 COIDOZ 2
l 122 COID01 0 COID01 0 COID01 0 COIDOS 1
1 123 COIDOl 0 COID01 0 COID01 0 COID01 l
1 124 COIDOl 2 COID01 2 COIDOl 2 COID05 0
1 125 COID01 2 COID01 2 COID01 2 COID08 1
1129 COIDOB 3 COID08 3 COID08 3 COID07 1
1 131 COID01 0 COIDOl 0 COID01 0 COIDOS 0
1132 COIDO8 3 COIDOS 3 COID08 3 COID05 2
l 133 COIDO8 3 COID08 3 COID08 3 COID06 3
1 134 COIDO8 3 COIDO8 3 COID08 3 COIDO4 2
1 135 COID08 3 COID08 3 COID08 3 COID06 0
1136 COIDO8 3 COIDO8 3 COIDOB 3 COID07 2
1 137 COIDOl 2 COID01 2 COID01 2 COIDO8 2
1 140 COID01 0 COID01 0 COIDOl 0 COID03 2
1 141 COIDOl 0 COID01 0 COIDOl 0 COID03 l
1 142 COIDOl 2 COIDOl 2 COID01 2 COID02 2
1 144 COID01 3 COID01 3 COID01 3 COIDOB 1
1 145 COIDOl 3 COIDOl 3 COID01 3 COID07 2
1 146 COID01 3 COID01 3 COID01 3 COIDOG 1
1 149 COID02 2 COIDO2 2 COID02 2 COIDO6 1
1 150 COID01 0 COID01 0 COID01 0 COIDOS 2
1151 COIDOl 0 COIDOl 0 COID01 0 COIDO6 2
1 153 COID01 0 COIDOl 0 COID01 0 COID02 0
1155 COID01 0 COIDOl 0 COID01 0 COID01 1
1156 COIDOl 0 COID01 0 COIDOl 0 COID01 3
1157 COID08 3 COID08 3 COID08 3 COID08 2
l 158 COID08 3 COIDO8 3 COID08 3 COID01 2
1159 COID08 3 COID08 3 COID08 3 COIDO4 3
1160 COIDO8 3 COID08 3 COID08 3 COID05 0
1161 COIDO8 3 COID08 3 COID08 3 COIDOB 2
1165 COID01 3 COID01 3 COIDOl 3 COID05 3
1166 COIDOl 3 COIDOl 3 COIDOl 3 COID03 1
1 167 COID03 2 COIDO3 2 COID03 2 COID02 2
1168 COID03 2 COID03 2 COIDO3 2 COID02 1
1169 COID05 0 COID05 0 COIDOS 0 COID06 2
1170 COID05 0 COID05 0 COID05 0 COIDO4 3
1171 COID05 0 COID05 0 COID05 0 COID05 1
1172 COIDO8 3 COID08 3 COID08 3 COIDO4 1
1173 COIDO6 2 COID06 2 COIDO6 2 COIDOS 2
1 174 COIDO6 2 COID06 2 COIDO6 2 COID05 1
1175 COID06 2 COIDO6 2 COID06 2 COID08 2
1176 COIDOS 3 COID08 3 COID08 3 COID03 2
1 177 COID01 3 COID01 3 COID01 3 COIDO4 0

 

 

 

 

 

 

 

 

 

 

 

 

201

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l 178 COID05 0 COID05 0 COIDOS 0 COIDOl 2
1 179 COID05 0 COIDOS 0 COID05 0 COIDO4 3
1 180 COID05 0 COID05 0 COID05 0 COIDOT 1
1 186 COIDOl 0 COIDOl 0 COID01 0 COID07 0
1 187 COID01 0 COIDOl 0 COIDOl 0 COID07 0
1 188 COID08 3 COIDO8 3 COIDOS 3 COIDO4 3
1 189 COIDO4 3 COIDO4 3 COIDO4 3 COID06 0
1190 COIDO4 3 COIDO4 3 COIDO4 3 COIDO7 0
1191 COID01 0 COID01 0 COID01 0 COIDO4 1
1 192 COID01 0 COID01 0 COID01 0 COIDO4 l
1 193 COID01 3 COID01 3 COIDOl 3 COID02 3
l 194 COID01 0 COID01 0 COID01 0 COIDO4 0
1 195 COID01 0 COID01 0 COIDOl 0 COID05 2
1196 COID01 0 COID01 0 COID01 0 COID03 1
1197 COID01 0 COID01 0 COID01 0 COIDO4 3
1199 COID01 0 COID01 0 COIDOl 0 COIDOS 2
1200 COID01 0 COID01 0 COID01 0 COID05 1
1201 COID01 0 COID01 0 COIDOl 0 COIDO7 0
1202 COID01 0 COIDOl 0 COIDOl 0 COIDOZ l
1203 COID08 3 COIDO8 3 COID08 3 COID05 2
1205 COID08 3 COID08 3 COID08 3 COID06 3
1206 COID01 0 COID01 0 COID01 0 COID08 3
1207 COID01 0 COIDOl 0 COID01 0 COIDOS 1
1210 COID01 0 COID01 0 COID01 0 COIDO6 3
121 1 COID01 0 COID01 0 COID01 0 COIDOB l
1212 COID01 0 COID01 0 COIDOl 0 COID03 1
1213 COIDOl 0 COIDOl 0 COID01 0 COID02 l
1214 COIDOl 3 COIDOl 3 COID01 3 COID02 2
1215 COID01 0 COIDOl 0 COID01 0 COID08 2
1216 COID08 3 COIDOS 3 COID08 3 COID06 1
1217 COID08 3 COIDOS 3 COIDO8 3 COID08 1
1218 COID01 2 COIDOl 2 COID01 2 COID07 1
1219 COIDOB 3 COID08 3 COID08 3 COIDO3 2
1220 COIDOl 0 COIDOl 0 COID01 0 COIDOT 2
1221 COIDOl 0 COID01 0 COIDOl 0 COID05 1
1222 COID01 3 COID01 3 COIDOl 3 COIDO4 1
1224 COID01 3 COID01 3 COIDOl 3 COIDO3 1
1227 COID08 3 COID08 3 COIDO8 3 COID07 1
1228 COID01 3 COID01 3 COID01 3 COID02 1
1229 COID01 0 COIDOl 0 COID01 0 COIDO6 2
1230 COID01 0 COIDOl 0 COIDOl 0 COID06 1
1231 COIDOl 0 COID01 0 COIDOl 0 COID07 3
125 1 COID01 3 COID01 3 COID01 3 COID05 3
1259 COID01 3 COID01 3 COIDOl 3 COID06 3
1260 COID05 0 COID05 0 COID05 0 COID03 2
1261 COIDOT l COID07 1 COID07 l COID07 1
1262 COIDOS 3 COIDOS 3 COIDO8 3 COID06 1
1263 COID07 1 COID07 1 COID07 l COID07 2
1264 COIDOZ 2 COIDOZ 2 COIDOZ 2 COID05 3
1267 COID01 0 COID01 0 COID01 0 COIDO4 2
1269 COID01 0 COID01 0 COID01 0 COID02 l
1272 COID01 2 COIDOl 2 COIDOl 2 COIDOS 2

 

 

 

 

 

 

 

 

 

 

 

202

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1273 COID01 2 COID01 2 COIDOl 2 COID02 1
1274 COID05 2 COID05 2 COID05 2 COIDO6 1
1275 COID01 0 COID01 0 COID01 0 COID07 1
1276 COIDOl 0 COIDOl 0 COID01 0 COID06 1
1277 COIDOl 2 COID01 2 COIDOl 2 COID03 2
1278 COID01 2 COID01 2 COIDOl 2 COID06 2
1279 COID01 2 COID01 2 COIDOl 2 COIDOS 2
1280 COID01 0 COID01 0 COID01 0 COIDOB 1
1281 COID01 0 COID01 0 COID01 0 COID07 3
1282 COID01 0 COID01 0 COID01 0 COID03 l
1283 COID01 3 COID01 3 COID01 3 COID01 1
1284 COID05 0 COID05 0 COIDOS 0 COIDO4 0
1286 COID01 0 COID01 0 COIDOl 0 COID07 2
1287 COID01 3 COID01 3 COID01 3 COID07 3
1288 COID01 0 COID01 0 COID01 0 COIDO6 0
1289 COID01 0 COID01 0 COID01 0 COID05 1
1290 COID08 3 COID08 3 COIDOS 3 COID07 2
1291 COID01 0 COID01 0 COID01 0 COID01 1
1292 COIDO7 2 COID07 2 COID07 2 COID06 2
1293 COID07 2 COIDOT 2 COID07 2 COIDO4 2
1294 COID07 2 COID07 2 COID07 2 COIDO4 3
1297 COID01 0 COID01 0 COID01 0 COID06 0
1298 COID01 0 COID01 0 COID01 0 COID05 0
1300 COIDOl 0 COID01 0 COID01 0 COID01 0
1307 COIDOl 0 COID01 0 COIDOl 0 COID03 1
1308 COIDOl 3 COID01 3 COIDOl 3 COID07 1
1323 COID05 0 COIDOS 0 COID05 0 COIDO4 1
1325 COID01 0 COIDOl 0 COID01 0 COID08 1
1326 COID01 0 COID01 0 COID01 0 COIDO2 1
1330 COID05 2 COID05 2 COID05 2 COIDO4 3
1331 COIDOS 2 COID05 2 COID05 2 COIDO4 2
1336 COIDO7 1 COID07 1 COIDO7 1 COID05 3
1337 COIDOS 0 COIDOS 0 COID05 0 COIDOZ 1
1338 COID05 0 COID05 0 COID05 0 COID08 2
1339 COID01 3 COIDOl 3 COIDOl 3 COIDO4 2
1340 COIDOl 0 COID01 0 COID01 0 COID06 1
1341 COID07 1 COIDO7 1 COID07 1 COID05 2
1342 COID01 0 COID01 0 COID01 0 COID07 1
1345 COIDOS 2 COID05 2 COID05 2 COIDO4 0
1346 COID01 0 COID01 0 COIDOl 0 COIDO4 2
1347 COID02 2 COID02 2 COID02 2 COIDO2 2
1348 COID01 0 COID01 0 COID01 0 COID01 0
1349 COID01 0 COIDOl 0 COID01 0 COIDO4 2
1350 COIDOl 0 COIDOl 0 COIDOl 0 COIDO7 0
1351 COID01 0 COIDOl 0 COID01 0 COIDOT 2
1352 COIDOl 0 COID01 0 COID01 0 COIDO3 1
1354 COID01 0 COID01 0 COID01 0 COID02 1
1355 COID02 2 COID02 2 COID02 2 COIDO4 3
1356 COID02 2 COID02 2 COID02 2 COIDO4 1
1357 COID01 0 COID01 0 COID01 0 COID01 1
1358 COID01 0 COID01 0 COID01 0 COID07 2
1359 COIDOl 0 COIDOl 0 COID01 0 COID05 2

 

 

 

 

 

 

 

 

 

 

 

 

203

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1360 COIDO2 2 COIDOZ 2 COID02 2 COID02 1
1361 COID02 2 COID02 2 COIDO2 2 COIDO4 l
1364 COID01 0 COIDOl 0 COID01 0 COID02 2
1367 COID01 2 COID01 2 COID01 2 COIDO4 2
1368 COID01 0 COID01 0 COID01 0 COID07 1
1369 COIDOl 0 COID01 0 COID01 0 COID07 1
1370 COID01 0 COID01 0 COID01 0 COIDO4 1
1371 COID01 0 COID01 0 COID01 0 COID07 2
1372 COID05 0 COID05 0 COID05 0 COID07 l
1374 COID02 2 COIDO2 2 COID02 2 COID05 0
1375 COIDOl 3 COID01 3 COID01 3 COID01 3
1378 COID01 2 COID01 2 COIDOl 2 COID02 0
1379 COID01 2 COID01 2 COID05 2 COID05 l
1380 COID05 0 COID05 0 COID05 0 COID03 0
1381 COID01 3 COID01 3 COID01 3 COIDO7 1
1382 COID01 0 COID01 0 COIDOl 0 COID06 3
1554 COIDOl 0 COID01 0 COID01 0 COID03 0
1558 COID01 0 COID01 0 COIDOl 0 COID06 0
1559 COIDOl 0 COID01 0 COIDOl 0 COID07 l
1561 COID01 0 COID01 0 COID01 0 COID06 0
1562 COIDOB 3 COID08 3 COID08 3 COID05 2
1563 COIDOl 0 COID01 0 COID01 0 COID08 2
1565 COIDOl 0 COIDOl 0 COID01 0 COID02 3
1566 COIDOl 0 COIDOl 0 COID01 0 COID03 2
1579 COID08 3 COID08 3 COIDO8 3 COID08 3
1580 COID08 3 COID08 3 COIDO8 3 COID07 3
1582 COID08 3 COID08 3 COID08 3 COID01 3
1583 COIDOS 3 COID08 3 COID08 3 COIDO4 3
1584 COIDOl 0 COID01 0 COID01 0 COID02 2
15 86 COID01 0 COID01 0 COID01 0 COID07 2
15 87 COIDO8 3 COID08 3 COIDOS 3 COID05 3
1588 COIDO8 3 COID08 3 COID08 3 COID07 2
1590 COID01 0 COID01 0 COIDOl 0 COIDO6 3
1591 COID01 0 COID01 0 COID01 0 COID03 2
1592 COIDO8 3 COIDO8 3 COIDOS 3 COIDO6 1
1593 COIDO8 3 COID08 3 COIDOS 3 COID02 0
1594 COIDOl 0 COID01 0 COID01 0 COIDOS 1
1595 COID01 0 COID01 0 COIDOl 0 COID02 1
1596 COIDOS 3 COID08 3 COIDO8 3 COID08 2
1598 COID01 0 COID01 0 COID01 0 COIDO6 1
1637 COIDOl 0 COID01 0 COID01 0 COIDO4 3
1638 COIDO8 3 COIDO8 3 COIDOS 3 COID07 2
1639 COID01 0 COIDOl 0 COIDOl 0 COIDO4 1
1640 COID07 1 COID07 1 COID07 l COIDOl 1
1642 COID08 3 COID08 3 COIDO8 3 COID02 2
1643 COID08 3 COID08 3 COID08 3 COIDO4 3
1645 COID08 3 COIDOS 3 COID08 3 COID05 2
1646 COID08 3 COIDOS 3 COIDO8 3 COID05 1
1647 COID08 3 COID08 3 COIDOS 3 COID08 l
1648 COIDOZ 2 COID02 2 COIDOZ 2 COID06 2
1649 COID02 2 COID02 2 COID02 2 COID07 2
1650 COID02 2 COID02 2 COID02 2 COID07 3

 

 

 

 

 

 

 

 

 

 

 

 

204

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1654 COID08 3 COID08 3 COID08 3 COIDOl 1
1670 COID01 0 COID01 0 COID01 0 COID01 3
1671 COID01 0 COID01 O COID01 0 COIDO4 0
1672 COID01 0 COIDOl 0 COID01 0 COIDOS 2
1673 COID05 2 COIDOS 2 COID05 2 COIDO2 3
1674 COID01 0 COID01 0 COIDOl 0 COIDO3 3
1675 COID01 0 COIDOl 0 COIDOl O COID03 2
1676 COID01 0 COID01 0 COID01 0 COIDO3 1
1677 COIDOl 0 COID01 0 COID01 0 COIDO3 1
1678 COID01 0 COID01 0 COID01 0 COIDOS 2
1679 COID01 0 COID01 O COID01 0 COIDOZ 3
1680 COIDO8 3 COIDO8 3 COIDO8 3 COIDOl 2
1681 COID01 O COID01 0 COID01 0 COIDO3 O
1682 COID01 0 COID01 0 COID01 0 COIDO6 2
1683 COID01 0 COID01 0 COIDOl 0 COIDOS 0
1684 COID02 2 COID02 2 COID02 2 COIDO4 2
1685 COID01 3 COID01 3 COID01 3 COID05 2
1688 COID01 3 COID01 3 COID01 3 COIDO6 1
1689 COID02 2 COID02 2 COID02 2 COID02 0
1690 COID01 O COID01 O COIDOl 0 COIDO7 2
1691 COID02 2 COID02 2 COIDO2 2 COIDOS 1
1692 COIDO2 2 COID02 2 COID02 2 COID01 1
1693 COID01 2 COID01 2 COID01 2 COID05 3
1694 COIDOl 2 COID01 2 COID01 2 COIDO2 1
1695 COIDOl 3 COID01 3 COID01 3 COIDOZ 3
1696 COIDO4 3 COIDO4 3 COIDO4 3 COIDOO 0
1697 COID01 3 COID01 3 COID01 3 COID06 1
1699 COID01 0 COID01 O COID01 0 COIDO4 2
1700 COID01 O COID01 0 COID01 O COID05 3
1701 COID05 2 COID05 2 COIDOS 2 COID02 1
1702 COID01 3 COID01 3 COID01 3 COIDO6 0
1712 COID01 0 COID01 0 COIDOl O COIDO7 2
1713 COID01 O COID01 O COID01 0 COIDOl 1
1714 COIDOl 0 COIDOl O COIDOl 0 COIDO4 2
1715 COID01 0 COID01 O COID01 O COIDOZ 0
1716 COID01 0 COIDOS 0 COID05 0 COID05 1
1717 COID01 0 COID01 O COID01 O COIDO4 2
1718 COID01 0 COID01 0 COID01 0 COIDOS 2
1719 COIDOl 0 COID05 O COID05 0 COID07 O
1721 COID01 0 COID01 O COIDOl O COID08 2
1722 COIDOl 0 COID01 0 COID01 0 COIDOT 3
1725 COID08 3 COIDO8 3 COIDOB 3 COID05 2
1726 COID01 0 COID01 0 COID01 0 COIDO4 2
1727 COIDOl 0 COID01 O COID01 0 COID05 2
1736 COIDOl 2 COID01 2 COID01 2 COIDOT 3
Note:

’ - This Pareto Optimal solution is from Pareto Solution NO. 39 Of Table 4-1 for the watershed.

'1' - This Pareto Optimal soultion is from Pareto solution NO. 33 Of Table 4-6 for sub-watershed 178.

"* - This Pareto Optimal soultion is from Pareto solution NO. 33 Of Table 4-7 for sub-watershed 231.

 

 

205
Appendix E

Cover Letter and Attachments for Focus Group Evaluation

Cover Letter

Dear Sir:

I am a Ph.D. student in the Department Of Resource Development, Michigan State
University. I have recently completed my research on using a watershed approach for
agricultural non-point source pollution management for the Sycamore Creek watershed in
Ingham County. In order tO finish my dissertation, I am using an expert focus group
approach tO evaluate my research results.

After consulting with my advisor (Dr. Jon Bartholic), Dr. Lois Wolfson, and Mr. Joe Ervin
at the Institute Of Water Research, your name was recommended as a potential expert. I
would like to invite you tO attend the focus group session to discuss and evaluate my
research approach and results. The session will be sometime during the last two weeks Of
January (20th tO 30th) depending on participant availability. If you are willing tO attend the
focus group session, please review the attached date and time table and return it tO me.
Alternatively, you can contact me by phone or e-mail. I will coordinate everyone’s schedule
to set the final time and date.

Brief information about my research and the focus group session is attached. If you have
any questions, I can be reached by phone (Home: 248-473-5664; Office: 517-355-0170) or
by e-mail (kangyung@pilot.msu.edu).

Thank you very much for your assistance and consideration.

Sincerely,

Yung-Tsung Kang

 

206

Basic Research Information

Research Topic:
A Watershed Based Optimization Approach for Agricultural N on-POint Source (NPS)
Pollution Management

Research Questions (Overall and Specific):
Does the use Of a watershed based approach in place Of the current field based
approach provide better solutions tO the three-step NPS management procedure.
(1) Critical Area Identification
- How could one use a watershed based model for critical area
identification?
- Can the watershed based model improve the accuracy Of critical area
identification?
(2) Best Management Practices (BMPs)
- How could one use a watershed based approach tO select BMPs for
identified critical areas?
- How well (effective) will the selected BMPs ameliorate the target NPS
pollution?
(3) Comprehensive Area-Wide Management Plan
What are the goals and rules that should be used for generating a
comprehensive area-wide pollution control plan?
- How can a plan balance the multi-dimensional interests within a watershed
community?

Research Approach:
The characteristics Of NPS pollution are diffuse, stochastic, and dynamic in
nature. There are three major problems or difficulties in NPS pollution management:
(1) complex spatial and temporal interaction Of the NPS pollution processes make
it very difficult to accurately identify critical areas,
(2) unclear causal relationships (at the quantitative level) between NPS pollution
and BMPs reduce confidence on the effectiveness Of selected treatments, and
(3) the decision-making process have to balance not only different types Of NPS
pollution but also available resources and multi-interests.

In order tO address the above problems, a watershed based approach for NPS
pollution management has to consider the following elements:

(1) spatial and temporal interactions within a watershed context,

(2) causal relationships between multiple types Of pollutants and BMPs, and

(3) goal setting and compromising strategies.

For the NPS in the watershed context, use Of a watershed based physical process
simulation model can help identify the critical areas and associated pollutants. Then,
alternative scenarios with different BMPs can be combined with the model tO better
understand the causal relationships (the effectiveness) between problems and treatments.

 

207

Lastly with both goal programming and multi-criteria Optimization techniques, they can
provide a more solid basis for compromising different interests.

Research Methodologies:
Study Area

The selected study area is a subwatershed Of the Sycamore Creek watershed in
Ingham County. The outlet Of the subwatershed is located at the north end Of City
Mason. The size Of the subwatershed is about 64,000 acres. There are two major
tributaries flowing from south tO north and joined near the south end Of Mason.
Agriculture is the major land use type Of the subwatershed.

Data Sources

The major data sources are from the NRCS Hydrological Unit Water Quality
(HUW Q) database plus Other ancillary data collected or derived by me. These data
includes topography, soils (Soil-S database), land use/cover, field boundaries, crop and
management, etc...

Watershed Model

The Agricultural Non-Point Source pollution model (AGNPS version 5.0) is used
as the watershed model. It is a single-storm based physical process simulation model.
AGNPS divided a watershed into grids and each grid requires a set Of variables, such as
land slope, channel slope, drainage direction, soil texture and erodibility, cover and
management factor, applied fertilizer (N and P) amount, etc. Outputs from the model
contain detailed information on hydrology, soil erosion, and nutrients for each grid and
summaries at the watershed outlet. There is also a new module, source accounting, added
tO the newest version Of AGNPS. With this module, upstream pollution sources for any
grid within the watershed can be analyzed.

Data Processing
The primary data processing is to compute AGNPS variables. A set Of programs

was written tO automate the processes. Different tOOls and technologies, Geographic
Information System and Database Management System, were also used to facilitate
deriving model variable values. There are two set Of data, fixed and variable. The fixed
data set is mostly related tO topography, soil, and channel variables, such as slope. They
are not likely to change or to be modified by human activities. The variable data set is
mostly related to land use or more specifically agriculture practices, such as cover and
management factor. Values Of these variables will change based on the crop and crop
stage. For example, each agricultural field has a four-year Span crop rotation code and
there are three crop stages, fallow, seedbed, and establishment, for each year. Once these
base data are computed or set for each field, they are aggregated into 40-acre grids for the
study watershed and the input data is then ready for the AGNPS model.

Goal Setting

Three major NPS pollution problems, sediment, nitrogen, and phosphorus, were
considered in this research. For each stream grid, they were assigned a maximum (goal)
loading value for each type Of pollution problem, i.e., 1000 mg/l for sediment, 10.0 ppm

 

208

for nitrogen, and 1.0 ppm for phosphorus. These goals are the basis Of the “Objective
function” for Optimization. For each stream grid, a “normalized” goal value for each type
Of pollution problems was computed so that comparisons among different goals
(pollution problems) will not be biased toward a particular Objective.

Compromising Strategies

The basis Of compromising strategies is multi-criteria Optimization. There are
many theories and approaches on this subject matter. A Min-Max plus Pareto
Optimization approach was chosen for my study. The followings outline provides the
simplified procedures (please also see attached figure):

(I) randomly create a set Of “complex points”, i.e., each field is randomly
assigned a crop rotation id and the Objective function values is computed;

(2) determine which point has the worst Objective function (maximum) value, i.e.,
the maximum nitrogen concentration in a stream grid for a particular year;

(3) minimize the maximum value (Min-Max) by changing decision variables, i.e.,
change from conventional tillage to conservation tillage for the field in the
worst grid; and

(4) compare all goals respectively for Pareto Optimal solution set.

Research Results:

The final results are a set Of Pareto Optimal solutions. Pareto Optimal solution
simply means that one can not enhance an Objective without diminishing another
Objective. From these Pareto Optimal solutions and the changing processes, I can analyze
them and derive the following information:

(1) critical area identification

- Which field has been changed?
- What are the NPS pollution problems?
(2) alternative BMPs
- What treatments have been applied?
- How effective is each applied treatment?
(3) comprehensive area-wide management plan
- What are the final goal values?
- How well does the solutions balance among different goals?

Research Contribution:

The research has established a watershed based approach and methodologies for
the three-step NPS pollution management procedures. It also demonstrated an integrated
decision-making process for the ill-structured problem domain Of NPS pollution. The
findings provided a new perspective Of “watershed” management Of NPS pollution
problem, which will improve NPS pollution control to meet the national water quality
goal.

 

209

Focus Group Evaluation

Objectives:

The primary goal Of the focus group session is tO evaluate my research findings.
However, a focus group approach requires experts “talking” tO each Other on the subject
matter. With a gOOd moderator, the discussion can be focused and many interesting
comments and suggestions are Often provided by the attendants in the discussion. Thus,
this will not only be a discussion on my research, but also an experience for you to share
ideas with other experts on the subject with which we are all concerned.

Preparation:

I will send each attendant a package containing: (1) the purposes Of the session,
(2) the draft chapters Of Methodology and Results Of my dissertation, (3) summarized fact
sheets, and (4) other miscellaneous information. It will be very helpful to read them
before coming to the session.

Principles and Format:

The process will last for approximately 3 hours. I will begin with a quick 20
minutes (or less) presentation and answer any questions. The moderator will then take
over and focus on the Objectives Of the session. It will be up to the group and moderator
tO carry on the interactive session. I will only answer questions. It will NOT be my
Opinion against yours. I will be no more than a listener. The group members will discuss
THEIR points with each Other. At the end Of the session, I will ask each Of you to fill
out a simple questionnaire. This will provide me with a more Objective perspective on
what has been said in the discussion.

The evaluation will focus on two things: (1) the accuracy Of the results and (2) the
rationale of the approach. The accuracy will not be judging numbers. It will more likely
be about whether or not one agrees or disagrees on the fields I chose (where), the NPS
pollutants identified (what), and the BMP treatment applied (how). It is your expert
judgement. Evaluation Of the rationale Of the approach will include the overall
procedures, general design, logical processes, and any other matter you think that is
appropriate.

The session will be taped to help with a summarization Of the discussion. This
may make some people uncomfortable and hesitate to express their Opinion. However, I
will ask the group tO make a final decision on whether or not the session should be taped
before it begins. All comments will be anonymous in my dissertation. There will only be
a list Of participants in my dissertation. I will provide all attendants with a copy Of my
summaries. You will review it and make modifications if necessary. This evaluation will
be a chapter Of my dissertation.

Footnote:

If you would like to attend my defense or have a copy Of my dissertation, I would be
more than happy tO notify you and/or prepare it for you. If you have any questions on the
focus group issues, please feel free tO contact me.

 

210

Please fill out the following contact information and check the time period that you will
be available for the focus group session. The session will be held at the Institute Of Water
Research (Manly Miles Building) Michigan State University. It should last no more than

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3 hours.

Your Name:

The best way tO be contacted:

Phone: E-Mail Address:

Januaryl998 8-11AM 9—12AM l—4PM 2-5PM

20 (Tuesday)
21 (Wednesday) Not available NOt available Not available NOt available
22 (Thursday)
23 (Friday) Not available Not available
26 (Monday) Not available Not available Not available Not available
27 (Tuesday) Not available Not available * *
28 (Wednesday) Not available Not available * *
29 (Thursday) * *
30 (Friday) NOt available Not available * *

 

 

 

 

 

 

NOTE: "' - preferred time and date

Please put this sheet into the self-stamped envelope and mail it back tO me. Or you can
send me an e-mail or call me. Here is my contact information:

E-Mail: kangflng@pilot.msu.edu
Phone: (Office) 517-353-0170 (with answering machine)

(Home) 248-473-5664

 

 

Once I sort out everyone’s schedule, I will contact you for the final time and date ASAP.
I will also provide more detail information on my research approach and results for use in
the focus group session. Thank you very much!

BIBLIOGRAPHY

 

BIBLIOGRAPHY

Arnold, J.G., X. Zruang, B.A. Engel, R. Srinivason, R. Muttiah, and C. Rewertz. 1990,
"Intelli gent GIS for Natural Resource Modeling and Site Selection", Proceedings on
Remote Sensing and GIS Applications to Nonpoint Source Planning, US. EPA and
Northeastern Illinois Planning Commission, Chicago, Illinois, pp. 45-54.

Arnold, J.G., J.R. Williams, A.D. Nicks, and NB. Sammons, 1990. SWRRB: A Basin Scale
Simulation Model for Soil and Water Resources Management, Texas A&M Press, 255 pp.

Arnold, J.G., LR. Williams, R.H. Griggs, and NB. Sammons, 1991. SQWRRBWQ: A
Basin Scale Model for Assessing Management Impacts on Water Quality, USDA,
Agricultural Research Service, National Soil Erosion Research Labortary, Purdue
University, West Lafayette.

Arnold, J.G., J.R. Williams, R. Srinivasan, and K.W. King, 1996. SWAT: Soil and Water
Assessment Tool, USDA, Agricultural Research Service, Grassland, Soil and Water
Research Labortary, Temple, Texas.

Bailey, G.W. and RR. Swank JR., 1983. "Modeling Agricultural Nonpoint Source
Pollution: A Research Perspective", Agricultural Management and Water Quality, Edited
by Shaller, F.W. and G.W. Bailey, The Iowa State University Press, Ames, Iowa, pp. 2747.

Beasley, D.B., Huggins, LE, and EN. Monke, 1980. ANSWER: A Model for Watershed
Planning, Transactions of ASAE, Vol. 23, NO. 4, pp. 938-944.

Beasley DB, 1986. "Distributed Parameter Hydrologic and Water Quality Modeling",
Agricultural Nonpoint Source Pollution: Model Selection and Application, Edited by
Giorgini, A. and A. Franco, Elsvier Science Publishing Company Inc., pp. 9-35.

Braden, J.B. and D.L. Uchtrnann, 1985. “Agricultural Nonpoint Pollution Control: An
Assessment”, Journal of Soil and Water Conservation, January-February, pp. 23-26.

Christensen, LA. and RE. Norris, 1983. “Soil Conservation and Water Quality
Improvement: What Farmers Think”, Journal of Soil and Water Conservation, January-

February, pp. 15-20.

Cook, K., 1985. “Agricultural Nonpoint Pollution Control: A Time for Sticks?”, Journal of
Soil and Water Conservation, January-February, pp. 105-106.

211

 

212

DeCoursey, D.G., 1985. "Mathematical Models for Nonpoint Water Pollution Control",
Journal of Soil and Water Conservation, VOl. 40, NO. 5, pp. 408-413.

Donigian, Jr. A.S., 1986. "Integration Of Runoff and Receiving Water Models for
Comprehensive Watershed Simulation and Analysis Of Agricultural Management
Alternatives", Agricultural Nonpoint Source Pollution: Model Selection and Application,
Edited by Giorgini, A. and A. Franco, Elsvier Science Publishing Company Inc., pp. 265-
275.

Duda, A.M. and R]. Johnson, 1985. “Cost-Effective Targeting Of Agricultural Nonpoint-
Source Pollution Controls”, Journal of Soil and Water Conservation, J anuary-February, pp.
108-1 11.

Duttweiler D.W. and HP. Nicholson, 1983. "Environmental Problems and Issues Of
Agricultrual Nonpoint Source Pollution", Agricultural Management and Water Quality,
Edited by Shaller, F.W. and G.W. Bailey, The Iowa State University Press, Ames, Iowa, pp.
3-24.

Eschenauer, H., Koski, J ., and A. Osyczka, 1990. “Multicriteria Optimization -
Funcarnentals and Motivation” in Multicriteria Design Optimization: Procedures and
Applications edited by Eschenauer et al., Springer-Verlag, New York, pp. 1-32.

Epp, DJ. and J .S. Shortle, 1985. “Agricultural Nonpoint Pollution Control: Volunary or
Mandatory?”, Journal of Soil and Water Conservation, January-February, pp. 111-114.

Ferrnanian, T.W., 1989. “Development Of An Expert System for Agriculture”, In
Proceedings Climate and Agriculture: System Approaches tO Decision Making, edited by
Albert Weiss, Charleston, South Carolina, 1-7 March, pp. 180-193.

Flach,K.W.,1990."Low-Input Agriculture and Soil Conservation, Journal of Soil and
Water Conservation, Vol. 45, NO. 1, pp. 42-44.

Gensemer, M.K. and M. Yamaguchi, 1985. “Successful Water Quality Planning: An
Areawide Perspective”, Journal of Soil and Water Conservation, January-February, pp. 76-
68.

Gill, P.E., Murry, W. and M. Wright, 1981. Practical Optimization, Academic Press Inc.,
New York, 401 pp.

GO, V.A., 1996. An Expert System Based Methodology for Evaluating the Costs and
Benefits of Soil Conservation, Ph.D. Dissertation, Michigan State University, E. Lansing,
Michigan, 200 pp.

 

 

213

Goudy, G. W. ed., 1986. Water Quality and Pollution Control in Michigan Report: 1986
Report, Michigan Department Of Natural Resources, Surface Water Quality Division,
Lansing, Michigan, 132 pp.

Guariso, G. and H. Werthner, 1989. Environmental Decision Support Systems, John Wiley
& Sons Inc., 240 pp.

Haith, D.A., 1982. "Chapter 6 - Application: Management Of Agricultural Nonpoint Source
Pollution", Environmental Systems Optimization, John Wiley & Sons Inc., pp. 135-159.

Harrington, B., A.J. Krupnick, and HM. Peskin, 1985. “Policies for Nonpoint-Source
Water Pollution Control”, Journal of Soil and Water Conservation, J anuary-February, pp.
27-32.

He, C., J .F. Riggs, and Y. Kang, 1993. "Integration Of Geographic Information Systems and
a Computer Model tO Evaluate Impacts Of Agricultural Runoff on Water Quality", Water
Resources Bulletin, American Water Resources Association, Vol. 29, NO. 6, pp. 891-900.

Huggins, LP. and EJ. Monke, 1966. "The Mathematical Simulation Of the Hydrology Of
Small Watersheds”, Technical Report NO. 1, Water Resources Research Center, Purdue
University, W. Lafayette, Indiana.

Jones, J .W., 1989. “Integrating Models with Expert Systems and Data Bases for Decision
Making”, In Proceedings Climate and Agriculture: System Approaches tO Decision Making,
edited by Albert Weiss, Charleston, South Carolina, 1-7 March, pp. 180-193.

Kaplan, S. and BJ. Garrick, 1981. “On the Quantitative Definition Of Risk”, Risk Analysis,
Vol. 1, NO. 1., pp. 11-27.

Keen P.G.W., 1986. “Decision Support Systems: The Next Decade”, in Procedings:
Converence on Decision Support Systems: A Decade in Perspective, edit by McLean RE.
and HG. SOl, Noordwijkerhout, The Netherlands, 16—18 June, pp. 221-237.

Klaine, SJ., M.L. Hinman, D.A. Winkelmann, and JR. Martin, 1988. "Characterization of
Agricultural Nonpoint Pollution: Nutrient Loss and Erosion in a West Tennessee
Watershed", Environmental Toxicology and Chemistry, Vol. 7, pp. 601-607.

Knisel, W.G. JR., 1980. CREAMS: A Field Scale Model for Chemicals, Runofl; and
Erosion from Agricultural Management Systems, Conservation Research Report NO. 26,
US. Department Of Agriculture, Washington, DC.

Krueger, R.A., 1994. Focus Groups: A Practical Guide for Applied Research, Second
Edition, Sage Publications, Inc., Thousand Oaks, California.

 

 

214

Lane, LI. and M.A. Nearing, 1989. USDA - Water Erosion Prediction Project: Hillslope
Profile Model Documentation. NSERL Report No. 2, USDA-ARS National Soil Erosion
Research Laboratory, West Lafayette, Indiana.

Leonard, RA. and W.G. Knisel, Jr., 1986. "Selection and Application Of Models for
Nonpoint Source Pollution and Resource Conservation", Agricultural Nonpoint Source
Pollution: Model Selection and Application, Edited by Giorgini, A. and A. Franco, Elsvier
Science Publishing Company Inc., pp. 213-229.

Libby, L.W., 1985. “Paying the N onpoint Pollution Control Bill”, Journal of Soil and
Water Conservation, January-February, pp. 33-36.

Liebesman, LR. and ER Laws, 1988. "Chapter 2 - The Water Quality Act Of 1987: A
Major Step in Assuring the Quality Of the N ation's Water", Clean Water Deskbook,
Environmental Law Institute, pp. 2241.

Linda, 6., 1982. “Focus Groups: A New LOOk at an Old Friend”, Marketing and Media
Decisions, VOl. 17, NO. 10, pp. 96-98.

Michigan Department Of Natural Resources, Surface Water Quality Division, 1992.
Guidebook of Best Management Practices for Michigan Watersheds, Lansing, Michigan.

Michigan Department Of Natural Resources, Surface Water Quality Division, Nonpoint
Source Unit, 1993. Agricultural Best Management Practices Manual for Michigan 's
Nonpoint Source Pollution Program, Lansing, Michigan.

Mohite, M., A.D. Whittaker, and R. Srinivasan, 1993. "A Knowledge Based System
Linked to AGNPS/GRASS Interface", ASAE Paper 93-3041, American Society Of
Agricultural Engineers, St. Joseph, Michigan.

Moore, I.C., BM.H. Sharp, SJ. Berkowitz, and RR. Schneider, 1979. “Financial
Incentives tO Control Agricultural Nonpoint-Source Pollution”, Journal of Soil and Water
Conservation, March-April, pp. 60-64.

Mulkey, L.A., R.F. Carsel, and C.N. Smith, 1986. "Development, Testing, and
Applications Of Nonpoint Source Models for Evaluation Of Pesticides Risk to the
Environment", Agricultural Nonpoint Source Pollution: Model Selection and Application,
Edited by Giorgini, A. and Z. Franco, Elsevier Science Publishing Company Inc., pp. 383-
397.

Myers, P.C., 1986. “Nonpoint-Source Pollution Control: The USDA Position”, Journal of
Soil and Water Conservation, May-June, pp. 156-158.

 

 

215

Needham, SE, 1994. Assessing a Geographic Information System and Ann-AGNPS Water
Quality Model: A Case Study, Ph.D. Dissertation, Michigan State University, E. Lansing,
Michigan, 274 pp.

Novotny, V., 1986. "A Review Of Hydrologic and Water Quality Models Used for
Simulation Of Agricultural Pollution", Agricultural Nonpoint Source Pollution: Model
Selection and Application, Edited by Giorgini, A. and A. Franco, Elsvier Science Publishing
Company Inc., pp. 9-35.

Odum, ER, 1987. "Reduced-Input Agriculture Reduces Nonpoint Pollution", Journal of
Soil and Water Conservation, VOl. 42, NO. 6, pp. 412-414.

Ogg, CW. and RB. Pionke, 1986. “Water Quality and the New Farm Policy Initiatives”,
Journal Of Soil and Water Conservation, March-April, pp. 85-88.

Osteen, C., W.D. Seitz, and LB. Stall, 1981. "Managing Land tO Meet Water Quality
Goals", Journal of Soil and Water Conservation, VOl. 36, NO. 3, pp. 138-141.

Osyczka, A, 1984. Multicriterion Optimization in Engineering: with FORTRAN
Programs, Ellis Horwood Limited, West Sussex, England, 178 pp.

Petersen, G.W. et. al., 1991. Evaluation of Agricultural Nonpoint Pollution Potential in
Pennsylvania Using a Geographic Information System, Environmental Resources Research
Institute, Pennsylvania State University, University Park.

Pelletier, RE, 1985. "Evaluating N onpoint Pollution Using Remotely Sensed Data in Soil
Erosion Models", Journal of Soil and Water Conservation, Vol. 40, No.4, pp. 332-335.

Pike, R.W., 1986. Optimization for Engineering Systems, Van Nostard Reinhold
Company Inc., New York, 417 pp.

Prato, T., H. Shi, R. Rhew, and M. Brusven, 1989. "Soil Erosion and Nonpoint Source
Pollution Control in an Idaho Watershed", Journal of Soil and Water Conservation, Vol. 44,
NO. 4. pp. 323-328.

Rallison, RE, 1980. “Origin and Evolution Of the SCS Runoff Equation”, Symposium on
Watershed Management ’80, American Society Of Civil Engineering, Boise, ID, July 21-23,
pp.912-924.

Romero, C. and T. Rehman, 1989. Multiple Criteria Analysis for Agricultural Decisions,
Elsevier Science Publising Company Inc., New York, 257 pp.

ROka, F.M., R.A. Levins, B.V. Lessley, and W.L. Megette, 1990. "Reducing Field Losses
Of Nitrogen: Is Erosion Control Enough?", Journal of Soil and Water Conservation, Vol.
45, NO. 1, pp. 144—147.

 

 

 

 

 

 

 

216

Setia, RR, 1987. "Consideration Of Risk in Soil Conservation Analysis", Journal of Soil
and Water Conservation, Vol. 42, NO. 6, pp. 435-437.

Sorenson, DD., 1985. “Organizing an Information Program for Nonpoint Pollution
Contro ”, Journal of Soil and Water Conservation, J anuary-Fegruary, pp. 82-84.

Srinivasan, R. and BA Engel, 1991. "GIS: A TOOl for Visualization and analyzation",
ASAE Paper 91-7574, American Society Of Agricultural Engineers, St. Joseph, Michigan.

Srinivasan, R. and BA. Engel, 1991. "A Knowledge Based Approach to Extract Input Data
from GIS", ASAE Paper 91-7045, American Society Of Agricultural Engineers, St. Joseph,
Michigan.

Srinivasan R. and J.B. Arnold, 1993. “Basin Scale Water Quality Modeling Using GIS”,
Proceedings, Application Of Advanced Information Technologies: Effective Management
Of Natural Resources, June 18-19, Spokane, Washington. Pp. 475-484.

Stadler, W., 1988. “Fundamentals Of Multicriteria Optimization” in Multicriteria
Optimization in Engineering and in the Science edited by Stadler, W., Plenum Press,
New York, pp. 1-25.

Thum P.G., S.R. Pickett, B.J. Niemann JR., and SJ. Ventura, 1990. "Integrating Nonpoint
Pollution Assessment with Land Development Planning Using a GIS", Proceedings on
Remote Sensing and GIS Applications to Nonpoint Source Planning, U.S. EPA and
Northeastern Illinois Planning Commission, Chicago, Illinois, Oct. 1-3, pp. 93-102.

Troeh F.R., J.A. Hobbs, and R.L.Donahue, 1980. Soil and Water Conservation for
Productivity and Environmental Protection, Prentice-Hall, Inc., Englewood Cliffs, New
Jersey, 718 pp.

U.S. Department Of Agriculture, 1995. Water Quality: A Report of Progress, Washington,
DC.

U.S. Department Of Agriculture, Farm Service Agency, 1997. The Conservation Reserve
Program, Washington, DC.

U.S. Department Of Agriculture, Natural Resources Conservation Service, 1997.
Hydrologic Unit Water Quality Tool (H U/WQ) - User’s Manual, Washington, DC.

U.S. Department Of Agriculture, Science and Education Administration, 1980. CREAMS -
A F ield-Scale Model for Chemicals, Runofli and Erosion from Agricultural Management
Systems, Conservation Research Report NO. 26, W.G. Knisel ed., 643 pp.

U.S. Department Of Agriculture, Soil Conservation Service, 1972. National Engineering
Handbook, Section 4 — Hydrology (NEH-4), Washington, DC.

 

 

217

U.S. Department Of Agriculture, Soil Conservation Service, 1986. Urban Hydrology for
Small Watersheds, Technical Release 55, Washington, DC.

U.S. Department Of Agriculture, Soil Conservation Service, 1987. Technical Guide,
Section I-C, Water Erosion Control, East Lansing, Michigan.

U.S. Department Of Agriculture, Soil Conservation Service, 1990. Sycamore Creek
Watershed Water Quality Plan, East Lansing, Michigan.

U.S. Department Of Agriculture, Soil Conservation Service, 1991. Annual Progress Report
- Sycamore Creek Water Quality Program, East Lansing, Michigan.

U.S. Environmental Protection Agency, 1978. National Water quality Inventory: 1977
Report To Congress, EPA-440/2-84-006, Washington, DC.

U.S. Environmental Protection Agency, 1990. National Water quality Inventory: I 988
Report To Congress, EPA-44014-90—003, Washington, DC.

U.S. Environmental Protection Agency, 1995. Ecological Restoration: A Tool to
Management Stream Quality, EPA-841-F-95-007, Washington, DC.

U.S. Environmental Protection Agency, 1996. Environmental Indicators of Water Quality
in the United States, EPA-841-R-96-002, Washington, DC.

U.S. Environmental Protection Agency, 1996a. Better Assessment Science Integrating
Point and Nonpoint Sources — BASINS Version 1.0 User ’s Manual, EPA-823-R-96-001,
Washington, DC.

U.S. General Accounting Office, 1990. USDA Needs to Better Focus Its Water Quality
Responsibilities, GAO/RCED-90-162, Washington, DC.

U.S. General Accounting Office, 1990. Information on and Characteristics of Selected
Watershed Projects, GAO/RCED—95-218, Washington, DC.

Ventura S.J., N.R. Chrisman, K.C. Connors, R.F. Gurda, and R.W. Martin, 1988. "A Land
Information System for Soil Erosion Control Planning”, Joumal of Soil and Water
Conservation, Vol. 43, NO. 3, pp.230-233.

Vieux, B.E., Hemdon, LR, and R.L. Liston, 1989. "GIS and Water Quality Modeling for
Agricultural Resource Management Systems", ASAE Paper 89-2184, American Society Of
Agricultural Engineers, St. Joseph, Michigan.

Weinberg, A.C., 1990. "LOW-Input Agriculture Reduces Nonpoint-Source Pollution",
Journal of Soil and Water Conservation, VOl. 45, No.1 , pp. 48-50.

 

 

218

Williams, J .R., A.D. Nicks, and JG. Arnold, 1985. “Simulator for Water Resources in
Rural Basins”, Journal of Hydrological Engineering, ASCE, Vol. 111, NO. 6, pp. 970-986.

Wischmeier, W.H. and D. D. Smith, 1978. Predicting Rainfall Erosion Losses - A Guide to
Conservation Flaming, Agriculture Handbook NO. 537, U.S. Department of Agriculture,
Washington, DC.

Yin, R.K., 1984. Case Study Research, Sage Publications, Inc. Beverly Hills, Calfornia.

Young, R.A., C.A. Onstad, D.D. Bosch, and WP. Anderson, 1987. AGNPS, Agricultural
Non-Point-Source Pollution Model: A Watershed Analysis Tool, Conservation Research
Report 35, U.S. Department Of Agriculture, Agricultural Research Service, Washington,
DC.

Young, R.A., D. A. Onstad, D. D. Bosch, and WP. Anderson, 1989. "AGNPS: A
Nonpoint-Source Pollution Model for Evaluating Agricultural Watersheds", Journal of Soil
and Water Conservation, Vol. 44, NO. 2, pp. 168-173.

Yu, P., 1985. Multiple-Criteria Decision Making, Plenum Press, New York, 388 pp.

ZhaO, R., M.A. Foster, P.D. RObillard, and D.W. Lehning, 1994. “XGRCWP, A
Knowledge and GIS Based System for Selection, Evaluation, and Design Of Water
Quality Control Practices in Agricultural Watersh ”, In National Conference on
Environmental Problem Solving with Geographic Information Systems, sponsored by US
EPA, September 21-23, Cincinnati, Ohio.

_, 1980. Conservation Farming, Deer & Company, MOline, Illinois, 150 pp.