OPTIMIZATION OF ENVIRONMENTAL FLOW TO
PRESERVE/IMPROVE ECOLOGICAL FUNCTION
By
Matthew Ryan Herman

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Biosystems and Agricultural Engineering – Master of Science
2014

ABSTRACT

OPTIMIZATION OF ENVIRONMENTAL FLOW TO
PRESERVE/IMPROVE ECOLOGICAL FUNCTION
By
Matthew Ryan Herman
Freshwater is vital for all life, and with the growth of the human population, the need for this
limited resource has increased. However, human activities have significant impacts on
freshwater ecosystems, leading to their degradation. In order to ensure that freshwater resources
remain sustainable for future generations, it is critical to understand how to evaluate stream
health and mitigate degradation. To address these issues, the following research objectives were
developed: 1) assess current methods used to evaluate stream health, in particular
macroinvertebrate and fish stream health indices and 2) introduces a new strategy to improve
stream health to a desirable condition at the lowest cost by optimizing best management practice
(BMP) implementation plan. Analysis of over 85 macroinvertebrate and fish stream health
indices indicated that the most commonly used macroinvertebrate and fish indices are: Benthic
Index of Biotic Integrity (B-IBI), Ephemeroptera Plechoptera Trichoptera (Index) index,
Hilsenhoff Biotic Index (HBI), and Index of Biological Integrity (IBI). These indices are often
modified to take into account local ecosystem characteristics. In order to address objective two,
several hydrological models including Soil and Water Assessment Tool and Hydrologic Integrity
Tool were integrated and the results were used to develop stream health predictor models. All of
the models were guided by a genetic algorithm to design the watershed-scale management
strategies. The coupled system successfully identified eight BMP implementation plans that were
resulted in excellent stream health conditions according to the IBI score.

Copyright by
MATTHEW RYAN HERMAN
2014

This thesis is dedicated to my family and friends for all the love and support they have given me.

iv

ACKNOWLEDGEMENTS

I would like to thank everyone who has supported me and helped throughout my graduate
studies. To my major advisor Dr. Pouyan Nejadhashemi for his constant guidance and support.
Without which this thesis would not be nearly as complete as it is today. In addition, I would like
to thank my committee, Dr. Timothy Harrigan and Dr. Abdol-Hossein Esfahanian, for their
support throughout my research.
I would like to thank my friends and lab mates; Fariborz Daneshva, Dennis Ross, and Sean
Woznicki, without their incredible assistance this thesis would have never been completed. In
addition, I would also like to thank my friends; Melissa Rojas-Downing, Dr. Umesh Adhikari,
Reza Javid, Elaheh Esfahanian, Dr. Edwin Martinez, Dr. Subhasis Giri, Dr. Yaseen Hamaamin,
Tony Zhong, Kristen Henn, Georgina Sanchez, Ray Chen, and Sarah Steudle for their constant
support throughout my studies. You have made this whole journey more enjoyable.
Finally, I would like to specially thank my family. To my parents, Mark and Christine, for
encouraging me to peruse my graduate studies and always being there no matter the time. To my
brothers, Michael and James, for always being around to make me laugh and enjoy life. Thank
you for all the love you have given me.

v

TABLE OF CONTENTS

LIST OF TABLES ......................................................................................................................... ix
LIST OF FIGURES ....................................................................................................................... xi
KEY TO ABBREVIATIONS ....................................................................................................... xii
1.

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

2.

LITERATURE REVIEW ........................................................................................................ 4
2.1 Stream Health/Function.................................................................................................... 4
2.1.1
Indicators................................................................................................................... 4
2.1.1.1 Stream Health Index .............................................................................................. 5
2.1.1.2 Dundee Hydrologic Regime Assessment Method and Indicators of Hydrology
Alterations ............................................................................................................................ 5
2.1.1.3 Fish ........................................................................................................................ 6
2.1.1.3.1 Index of Biological Integrity ........................................................................... 6
2.1.1.4 Macroinvertebrates ................................................................................................ 7
2.1.1.4.1 EPT .................................................................................................................. 7
2.1.1.4.2 Benthic Index of Biological Integrity .............................................................. 7
2.1.1.5 Biotic Index ....................................................................................................... 8
2.1.1.6 Water Footprint.................................................................................................. 8
2.2 Environmental Flow ......................................................................................................... 9
2.3 Anthropogenic Impacts on Stream Flow ........................................................................ 10
2.3.1
Urbanization ............................................................................................................ 11
2.3.1.1 Runoff.................................................................................................................. 11
2.3.1.2 Drinking water..................................................................................................... 12
2.3.2
Agriculture .............................................................................................................. 12
2.3.2.1 Irrigation .............................................................................................................. 12
2.3.2.2 Runoff.................................................................................................................. 13
2.3.3
Dams ....................................................................................................................... 14
2.4 Conservation Practices/Best Management Practices (BMPs) ........................................ 15
2.4.1
Bioretention............................................................................................................. 15
2.4.2
Constructed Wetlands ............................................................................................. 15
2.4.3
Detention Basins ..................................................................................................... 16
2.4.4
Filter Strips/Riparian buffer .................................................................................... 16
2.4.5
Vegetated Swale...................................................................................................... 16
2.4.6
Native Grasses ........................................................................................................ 17
2.5 Optimization/Modeling .................................................................................................. 17
2.5.1
SWAT ..................................................................................................................... 18
2.5.2
SWAT-CUP ............................................................................................................ 18
vi

2.5.3
R-SWAT-FME ........................................................................................................ 19
2.6 Data Mining.................................................................................................................... 20
2.7 Conclusion...................................................................................................................... 22
3.

INTRODUCTION TO METHODOLOGY AND RESULTS............................................... 23

4.

A Review of Macroinvertebrate and Fish Stream Health Indices ......................................... 25
4.1 Abstract .......................................................................................................................... 25
4.2 Introduction .................................................................................................................... 25
4.3 Metrics ............................................................................................................................ 27
4.3.1
Abundance .............................................................................................................. 28
4.3.2
Species Richness ..................................................................................................... 28
4.3.3
Functional Feeding Groups ..................................................................................... 29
4.4 Collection strategies ....................................................................................................... 30
4.4.1
Wadeable Waterways.............................................................................................. 30
4.4.2
Non-wadeable Waterways ...................................................................................... 31
4.5 Application ..................................................................................................................... 32
4.6
Materials and Methods .................................................................................................. 33
4.6.1
Macroinvertebrate Indices ...................................................................................... 33
4.6.1.1 Benthic Index of Biotic Integrity ........................................................................ 40
4.6.1.2 Hilsenhoff Biotic Index ....................................................................................... 41
4.6.1.3 Ephemeroptera Plecoptera Trichoptera ............................................................... 43
4.6.2
Fish Indices ............................................................................................................. 45
4.6.2.1 Index of Biological Integrity (IBI) ...................................................................... 54
4.7 Conclusions .................................................................................................................... 56
4.7.1
Benefits ................................................................................................................... 57
4.7.2
Limitations .............................................................................................................. 58

5. Optimization of Conservation Practice Implementation Strategies in the Context of Stream
Health ............................................................................................................................................ 59
5.1 Abstract .......................................................................................................................... 59
5.2 Introduction .................................................................................................................... 59
5.3 Materials and Methodology ........................................................................................... 62
5.3.1
Study Area .............................................................................................................. 62
5.3.2
Data Collection ....................................................................................................... 63
5.3.2.1 Physiographic Data ............................................................................................. 63
5.3.2.2 Biological data..................................................................................................... 64
5.3.3
Modeling Process .................................................................................................... 66
5.3.3.1 Development Phase ............................................................................................. 67
5.3.3.1.1 Soil Water Assessment Tool.......................................................................... 67
5.3.3.1.2 Hydrological Index Tool ............................................................................... 68
5.3.3.1.3 Variable Selection.......................................................................................... 69
5.3.3.1.4 Stream Health Model ..................................................................................... 71
5.3.3.2 Scenario Phase..................................................................................................... 72
5.3.3.2.1 Develop the Reference Condition.................................................................. 72
5.3.3.2.2 Best Management Practices ........................................................................... 72
vii

5.3.3.2.3 Optimizing BMP Placement .......................................................................... 74
5.4. Results and Discussion ................................................................................................... 77
5.4.1
SWAT Model Calibration and Validation .............................................................. 77
5.4.2
Variable Selection for the Best Stream Health Model ............................................ 78
5.4.3
The Reference versus Current Conditions .............................................................. 80
5.4.4
Optimizing BMP Placement ................................................................................... 84
5.4.5
Associated cost of BMP Implementation Scenarios ............................................... 88
5.5 Conclusion...................................................................................................................... 91
5.6 Acknowledgments .......................................................................................................... 93
6.

CONCLUSIONS ................................................................................................................... 94

7.

FUTURE RESEARCH RECOMMENDATIONS ................................................................ 96

APPENDIX ................................................................................................................................... 97
REFERENCES ........................................................................................................................... 125

viii

LIST OF TABLES

Table 4.1. List of macroinvertebrate based indices. ..................................................................... 35
Table 4.2. List of fish based indices. ............................................................................................ 48
Table 5.1. IBI stream heath class ranges (adapted from Lyons, 1992). ........................................ 65
Table 5.2. BMP costs per unit area (NRCS, 2014b). .................................................................... 74
Table 5.3. Statistical criteria for the calibrated SWAT model at different USGS gauging stations
(Figure S2). ................................................................................................................................... 78
Table 5.4. Coefficients of determination for each variable selection technique. .......................... 79
Table 5.5. Comparison of pre-settlement and current conditions based on the length of stream for
each health class. ........................................................................................................................... 82
Table 5.6. Wilcoxon Signed-Ranks test and percent changes between the pre-settlement and
current conditions for stream orders 1 through 3. ......................................................................... 82
Table 5.7. Wilcoxon Signed-Ranks test and percent changes between the pre-settlement and
current conditions for stream orders 4 through 5 and for all streams. .......................................... 83
Table 5.8. Percentage of allocated area to BMP implementation within the agricultural lands in
scenarios that overall IBI score is greater than 65. ....................................................................... 85
Table 5.9. Percentage of individual streams with a stream health score greater than 65 for all
BMP scenarios. ............................................................................................................................. 87
Table 5.10. All unique BMP scenarios with an overall stream health score above 65. ................ 88
Table A1. Comparisons between B-IBI and other indices that either used or modified it for
application elsewhere. ................................................................................................................... 97
Table A2. Comparisons between HBI and other indices that either used or modified it for
application elsewhere. ................................................................................................................. 100
Table A3. Comparisons between EPT and other indices that either used or modified it for
application elsewhere. ................................................................................................................. 101
ix

Table A4. Comparisons between IBI and other indices that either used or modified it for
application elsewhere. ................................................................................................................. 103

x

LIST OF FIGURES

Figure 5.1. Saginaw Bay Watershed (HUC 040802) and the Honeyoey Creek-Pine Creek
Watershed (HUC 0408020203). ................................................................................................... 63
Figure 5.2. Flow diagram of the development and scenario phases. ............................................ 67
Figure 5.3. Pre-settlement and current stream health classes. ...................................................... 80
Figure 5.4. Distributions of individual stream health scores for all BMP scenarios. ................... 86
Figure 5.5. BMP placement and individual stream health scores for Scenario 2. ........................ 90
Figure 5.6. Improvements and declines in stream health scores between the pre-settlement and
current conditions.......................................................................................................................... 91
Figure S1. IBI monitoring stations in the Saginaw Bay Watershe ............................................. 114
Figure S2. USGS streamflow gauging stations........................................................................... 115
Figure S3. Improvements and declines in stream health scores between the pre-settlement and
current conditions........................................................................................................................ 116
Figure S4. Distributions of individual stream health scores against stream order (scenarios 1). 117
Figure S5. Distributions of individual stream health scores against stream order (scenarios 2). 118
Figure S6. Distributions of individual stream health scores against stream order (scenarios 3). 119
Figure S7. Distributions of individual stream health scores against stream order (scenarios 4). 120
Figure S8. Distributions of individual stream health scores against stream order (scenarios 5). 121
Figure S9. Distributions of individual stream health scores against stream order (scenarios 6). 122
Figure S10. Distributions of individual stream health scores against stream order (scenarios 7).
..................................................................................................................................................... 123
Figure S11. Distributions of individual stream health scores against stream order (scenarios 8).
..................................................................................................................................................... 124
xi

KEY TO ABBREVIATIONS

AMBI

AZTI Marine Biotic Index

ANFIS

Adaptive Neuro-fuzzy Inference Systems

BI

Biotic Index

B-IBI

Benthic Index of Biotic Integrity

BMPs

Best Management Practices

CCT

Cool-Cold Transition

CDL

Cropland Data Layer

CWA

Clean Water Act of 1972

CWT

Cool-Warm Transition

DHRAM

Dundee Hydrologic Regime Assessment Method

ElPT

Elmidae, Plecoptera, and Trichoptera

EPA

United States Environmental Protection Agency

EPT

Ephemeroptera, Plecoptera, and Trichoptera

EPTC

Ephemeroptera, Plecoptera, Trichoptera and Coleoptera

FBI

Fish Based Index

FBIL

Fish Based Index for Lakes

xii

GA

Genetic Algorithms

GLUE

Generalized Likelihood Uncertainty Estimation

HBI

Hilsenhoff Biotic Index

HIP

Hydroecological Integrity Assessment Process

HIT

Hydrologic Integrity Tool

HUC

Hydrologic Unit Code

IBI

Index of Biological Integrity

IBIi

Individual stream health scores

ICI

Invertebrate Community Index

IHA

Indicators of Hydrology Alterations

ISI

Invertebrate Species Index

Li

Individual stream lengths

MA24

Variability of flow values in January

MCMC

Markov Chain Monte Carlo

MFs

Membership Functions

MH18

Variability across annual maximum flows

MH19

The skewness of the annual maximum flows

MNFI

Michigan Natural Features Inventory

xiii

NBI

Nutrient Biotic Index

NBI-N

Nitrogen Nutrient Biotic Index

NBI-P

Phosphorous Nutrient Biotic Index

NCDC

National Climatic Data Center

NED

National Elevation Data

NHDPlus

National Hydology Dataset plus

NRCS

Natural Resources Conservation Service

NSE

Nash-Sutcliffe model efficiency coefficient

ParaSol

Parameter Solution

PBIAS

Percent Bias

PCA

Principal Component Analysis

PSO

Particle Swarm Optimization

R2

Coefficient of determination

RA5

The number of days where flow increased from the previous day

RA9

Variability in reversals

RMSE

Root-mean-square deviation

RSR

Root-mean-squared error-observations standard deviation ratio

R-SWAT-FME

R program language-Soil and Water Assessment Tool-Flexible Modeling
Environment
xiv

SEMCOG

Southeast Michigan Council of Governments

SGI

Stressor Gradients Index

SHI

Stream Health Index

SOFM

Self-organizing feature map

SSURGO

Soil Survey Geographic Database

SUFI-2

Sequential Uncertainty Fitting

SWAT

Soil and Water Assessment Tool

SWAT-CUP

SWAT Calibration and Uncertainty Analysis

TIVI

Tolerance Indicator Values Index

TN

Total Nitrogen

TP

Total Phosphorous

TSS

Total Suspended Solids

USDA-ARS

United States Department of Agriculture - Agricultural Research Service

USDA-NASS

United States Department of Agriculture-National Agricultural Statistics
Service

USGS

United States Geological Survey

xv

1.

INTRODUCTION
With the continued population growth, the demand for freshwater has also increased in order

to sustain human needs including crop production and drinking water. However, anthropogenic
activities have negatively impacted freshwater ecosystems, resulting in their degradation (Dos
Santos et al., 2011; Pander and Geist, 2013; Walters et al., 2009: Young and Collier, 2009).
Furthermore, changing climates are expected to add additional stress to these already strained
systems (Meyer et al., 1999; Ridoutt and Pfister, 2010). To ensure that freshwater ecosystems
will be available for future generations, evaluation of stream health condition has become vital
(EPA, 2011). In the United States, the focus on protecting freshwater resources began with the
passing of the Clean Water Act of 1972, with the goal of reducing point and non-point source
pollution and improving water quality (EPA, 2012a). Originally, chemical indicators were used
to evaluate stream conditions, leading to noticeable water quality improvements. However,
recent evaluation of the biotic components of freshwater ecosystems revealed that they are still
degraded, indicating that using only chemical indicators for stream health is not effective (EPA,
2011). This led to the introduction of a new type of evaluation called biological assessments or
bioassessment (Jeong, et al., 2012). Bioassessment can be used to assess physical, chemical, and
biological stressors within stream systems, which makes them ideal for evaluating stream
conditions (Brazner, et al., 2007; Pelletier, et al., 2012).
In the first study, different stream health indices were reviewed to aid in the selection of the
most appropriate index in different region, stressor, and species. For example, some indices are
sensitive to specific stressors such as organic pollution (Hilsenhoff, 1987; Johnson et al., 2013)
or nutrients (Smith et al., 2007; Hasse and Nolte, 2008). Other indices are developed for specific
water systems, such as warm water (Lyons, 1992) or cold water (Kanno et al., 2010; Lyons,
1

2012); or regional areas (Wan et al., 2010; Esselman et al., 2013; Krause et al., 2013) All of this
variability makes it challenging to determine which index should be used for different studies.
The goal of this study is to provide a detailed analysis of the different indices to aid in index
selection.
For the first study the overall goal is to provide a review of macroinvertebrate and fish based
stream health indices that can be used for watershed management.
The specific objectives of this study are to:


Assess current methods used to evaluate stream health, in particular macroinvertebrate
and fish stream health indices.

After identifying degraded streams, the next logical step is to develop mitigation strategies.
Best management practices (BMPs) are commonly used to control runoff and filter pollutants,
sustaining water quantity and improving water quality. However, implementation of multiple
BMP scenarios on the ground and monitor them over the years to identify the best option is not
feasible due to the cost and time constraints. Therefore, models are inexpensive and fast
alternative to monitoring and therefore widely used in water resources management (Giri, et al.,
2012). Meanwhile, modeling presents its own set of challenges by producing large volume of
data that is hard to interpret.
For the second study the overall goal is to develop a system that can be used to evaluate
different BMP scenarios to find near-optimal solution(s) for a watershed in Michigan by
maximizing stream health score and minimizing implementation cost. In addition, this study will
explore the relationship between a fish-based stream health index (IBI) and hydroecological

2

variables. This will be done in order to develop a stream health predictor model that can be
applied to the entire study area allowing for evaluation of stream conditions.
The specific objectives of this study are to:


Assess current methods used to evaluate stream health, in particular macroinvertebrate
and fish stream health indices.



Develop a Soil and Water Assessment Tool model that can model to estimate long-term
streamflow data for all stream segments within the study area.



Identify the most influential hydroecological parameters.



Develop a stream health predictor model based on selected hydroecological parameters
with the use of fuzzy logic techniques.



Evaluate the impacts of different best management practice scenarios with the use of
genetic algorithm to maximize stream health and minimize cost.

3

2.

LITERATURE REVIEW

2.1

Stream Health/Function
As water resources become more scarce, the importance of riverine ecosystem and their

condition has become more important to insure that there will be enough water for both human
and natural needs for the future (USGS, 2013a). However, our knowledge about natural system
needs, health, and its interrelations is limited.
Analysis of river systems is being performed to identify the status or health of the
riverine ecosystem. Stream health can be defined as the combined analysis of alterations caused
by anthropogenic activities in aquatic organisms, riparian vegetation, invertebrates, and channel
properties (Jeong et al., 2012). Anthropogenic impacts, often referred to as stressors, are defined
as an abiotic or biotic factors that are varied by human activities to the point where it has a
negative impact on an organism or the environment (Magbanua, 2012). It is important to note
that stressors often compound upon each other to create the environmental degradation
(Magbanua, 2012). This makes it difficult to restore the ecosystem when the actual cause of the
degradation cannot be easily identified. However, there is a solution, biological indicators are
able to represent the complex nature of stream ecosystems and provide information about what is
occurring within the stream system (Jeong et al., 2012).
2.1.1 Indicators
Indicators are aspects of the ecosystem that can be used to identify degradation in the
system; they can include nutrient uptake and denitrification (Young and Collier, 2009); as well
as biological indicators (Bunn et al., 2010), and hydrologic changes (Jeong et al., 2012). These
indicators can describe different functions and interactions within the stream allowing them to be
useful in determining what is impacting the stream or what the condition of the stream is.
4

However, biological indicators are often are used because they are able to represent multiple
layers of interaction within the ecosystem (Jeong et al., 2012), as well as being easier to observe
while still providing detailed information about the condition of the stream (Einheuser, 2011). In
the following sections, more detailed information is provided on a variety of indicators that can
be used to assess stream health.
2.1.1.1 Stream Health Index
The Stream Health Index (SHI) was developed to determine the degree of impairment of
a river or watershed, which is found by observing the pollutant loads within the water (Carlson et
al., 2012). This relatively simple method allows for determining degradation. However, it only
considers the Total Suspended Solids (TSS), Total Nitrogen (TN), and Total Phosphorous (TP)
levels (Carlson et al., 2012), and there are many other stressors that are not accounted for, such
as water quantity. Nevertheless, the SHI model allows for easy comparison between locations
and relates the results in a layman’s perspective useful for communicating with the public about
river and watershed degradation.
2.1.1.2 Dundee Hydrologic Regime Assessment Method and Indicators of Hydrology
Alterations
The Dundee Hydrologic Regime Assessment Method (DHRAM) assesses changes to the
hydrologic cycle and patterns caused by human activities (Jeong et al., 2012). It does this by
using a set of characteristics called the Indicators of Hydrology Alterations (IHA). IHA is a very
comprehensive method used to determine the alteration to the hydrology of the system; it uses a
set of 67 indicators to determine the condition of the stream and uses statistics to display the
results of the alterations (Jeong et al., 2012). At this point DHRAM is used to link the indicators,
from IHA, to what risk they pose to the environment (Jeong et al., 2012). This is useful at
5

identifying which river systems are most threatened, allowing policy makers and stokeholds to
make decisions on how to improve the environment.
2.1.1.3 Fish
Fish are a common, easily observed indicator of stream health. Their long lifespans and
migrations within the river systems (Karr, 1981) allow them to provide long term and largescale results to impairment in the entire system. Also due to their distribution within the tropic
levels (Karr, 1981), they can provide insight to the interactions that occur within the aquatic
ecosystem. Another benefit to using fish is that they tend to have well documented life histories
and are resistant to harsh environmental conditions (Karr, 1981) allowing for easy classification
of disturbances occurring in the system. As a system is degraded, it is expected that more of the
tolerant fish species will be found, and knowing what each species is tolerant to helps identify
what is impacting the river system. A final benefit to using fish as indicators is that very little
training is needed for identification (Karr, 1981), reducing the cost of monitoring them over an
entire watershed.
2.1.1.3.1

Index of Biological Integrity

The Index of Biological Integrity (IBI) is an indicator that utilizes the fish community in
river systems and is often used to monitor the health of the river and shed some light on the
interactions within the system (Jeong et al., 2012). IBI is calculated by observing a variety of
metrics; including species diversity, trophic composition, and abundance and condition
(Einheuser, 2011). Each metric is observed and ranked given a score of 1, 3, or 5 which higher
scores indicating better conditions. These scores can be summed for the calculation of a score
for the river network and be compared to other sites to determine restoration project order. A

6

benefit of this index is that it can be modified to match the species found in the region
(Einheuser, 2011).
2.1.1.4 Macroinvertebrates
Along with fish, invertebrates are a major component to river ecosystems. Having species in all
of the trophic levels and being able to easily identifiable, especially macroinvertebrates, makes
them efficient indicators of stream health and function. However, unlike fish, invertebrates are
not as well traveled and thus tend to show the health of a stream in a localized area (Einheuser,
2011). However, with the vast diversity of macroinvertebrates, several different indicator
systems have been developed and are used to monitor stream health.
2.1.1.4.1
EPT
EPT is an indicator based on the observation of organisms of the Ephemeroptera
(mayflies), Plecoptera (stoneflies), and Trichoptera (caddisflies) families (Goetz and Fiske,
2013). These species tend to be very sensitive to changes in the environment and thus make EPT
as an ideal indictors for early detection of stream degradation (Johnson et al., 2013). However,
since they have shorter lifespans than fish, EPT indicators are not as efficient at looking at large
watershed level disturbances, but do excel at local degradation identification (Einheuser, 2011).
Also like fish, it is relatively easy to identify the EPT species so it is easy to collect data for
analysis; allowing EPT to be an efficient indicator for identifying local degradation before it
becomes a larger problem to solve.
2.1.1.4.2

Benthic Index of Biological Integrity

The Benthic Index of Biotic Integrity (B-IBI) is a multi-metric index developed by
Kerans and Karr (Kerans and Karr, 1994) and is based on the IBI. Like with the IBI, the B-IBI’s
metrics are divided into 3 categories: species diversity, trophic composition, and abundance and
7

condition (Einheuser, 2011). However, the characteristics observed by the metrics are all
characteristics of the invertebrate community in the river system. This allows for a detailed
analysis of the system and its condition. Each metric is given a score based on the observations,
just like in the IBI, and that score is used to evaluate the overall system as well as being used to
compare between different sites (Kerans and Karr, 1994).
2.1.1.5 Biotic Index
The Biotic Index (BI) or HBI developed by Hilsenhoff in the 70’s, was based on the tolerance of
each taxa observed to organic pollutants (Goetz and Fiske, 2013). After recording all of the
tolerances, the river system was ranked on a scale from 0 to 10, with 0 being the best (Goetz and
Fiske, 2013); this value could then be compared to other sites to determine the degradations
across the region.
2.1.1.6 Water Footprint
Another way to look at the health of a stream is to observe how much water is being
removed from the system; this can be done with a water footprint calculation. Water footprints
are similar to carbon footprints, where analysis is preformed to see how much water is being
used by various practices (Ridoutt and Pfister, 2010). This allows for easy identification of the
major hydrologic stressors to river systems and can be used to show how different methods of
irrigation, farming, industry, etc… compared to each other. By showing the comparison between
different practices makes individuals conscious of the water requirements needed to produce
products, and how much can be saved by changing to a more efficient method of production. By
finding the least withdrawing practices, improvements can be made to limit the damages caused
by over taxing the river systems. Unfortunately, there is no one standard to calculating a water
footprint, so depending on the calculation process different water footprints may be calculated
8

(Ridoutt and Pfister, 2010). Therefore, if a comparison between water footprints is planned it
should be verified that all the water footprints being compared were calculated by the same
method to allow for fair comparisons.
2.2

Environmental Flow
Environmental flow describes the patterns and quantity of water flow needed to support

aquatic ecosystems as well as the needs of humans (King et al., 2009; Poff et al., 2010; Chen and
Zhao, 2011). Originally, this idea led to a minimalist strategy, where only a static minimum
amount of water was released so that the environment could survive (Alfredsen et al., 2012).
This insured that we could alter the flow by storing and removing almost as much as we wanted,
only the minimum had to remain to insure the environment did not die out. However, further
studies showed that supplying the environment with just the minimum level of water needed was
flawed because it was actually more damaging to the riverine ecosystems than originally thought
(Poff et al., 2010). In recent years, environmental flow has undergone a change from supplying
the minimum amount of flow to a river system to support the ecosystem to trying to replicate the
natural flow cycles in both timing and volume, to better support aquatic ecosystems (King et al.,
2009; Poff et al., 2010; Alcázar and Palau, 2010; Chen and Zhao, 2011;).
With the demand for fresh water growing so being able to sustain the use of freshwater
systems is vital to insure long-termed benefits (Nel et al., 2011). Based on current research, it
has been well documented that maintaining the flow regime is vital to sustaining ecological
integrity of river systems (Belmar et al., 2001; Poff et al., 2010; Poff and Zimmerman, 2010; Nel
et al., 2011; Pinieski et al., 2011). This means that environmental flow has become a key factor
in the management plans for freshwater systems (McCartney et al., 2009).

9

The process of defining an environmental flow has several steps. First selection of
riverine organisms for which the flow will be established is preformed and a team of specialists
gathered to determine what the organisms need to survive (Piniewski et al., 2011). It makes
sense for this selection to be a species that is more sensitive than others are so that the final
environmental flow will support more than one organism. Next, the sections of a river in which
the environmental flow will be defined must be selected (Piniewski et al., 2011). These sections
are usually riffles, runs and pools. Next specialists need to define what flow characteristics the
organism needs which leads to the final step, the definition of the environmental flow (Piniewski
et al., 2011). After establishing the environmental flow criteria, monitoring should be put in
place to observe whether the desired organism is able to establish, if it fails, the environmental
flow should be revised to insure success of the project.
2.3

Anthropogenic Impacts on Stream Flow
As humans, we rely on the environment for everything, from raw resources to make

houses and tools to food and water, which are needed for survival. And to obtain what we need
to survive we take from the environment and leave behind what we cannot use along with any
destruction or disturbances inflicted on the environment. This leads to degradation of the
environment, loss of habitat, and the destruction of the resources we need for the future, for
example deforestation (Coe et al., 2011), during which we destroy forests that provide lumber,
clean air, and produce when left standing. But are cut down to make farmland which quickly
loses its fertility and thus productivity. Even sometimes when we attempt to reduce the impacts
we have on the environment, the environment is still negatively impacted. For example selective
logging, which has been considered as a compromise between deforestation and preservation
(Putz et al., 2012). However, the impacts of this method result in the loss of carbon from

10

damaged plants for several decades and result in a century long re-growth process for the forest
to return to its pre-logging state (Huang and Asner, 2010). And while these examples describe
the impact on forests, similar outcomes can be seen in river systems. In this review, emphasis
will be put on impacts to the health of streams in terms of water quantity. In general, the
exchanges between humans, water, and the environment can be grouped into two categories,
withdraws, and returns. Both of which have impacts that degrade the environment.
2.3.1 Urbanization
Urbanization is the conversion of land to urban regions to support the increasing human
population (USGS, 2013b). As the population of humans on the earth continues to grow, more
land is needed for homes and more water is need for drinking and cleaning. These new demands
on the environment have several negative impacts on the aquatic environment. Nevertheless,
being aware of sources that degrade the systems allows steps to be taken to reduce the observed
degradation. For this review, urban lands include residential, commercial, and industrial.
2.3.1.1 Runoff
With the increased urbanization, negative impacts to the hydrologic cycle can be seen. In
Goetz and Fiske’s study (2013) the hydrologic impacts of urbanization included reduced
infiltration, increased peak flows, and reduced time to peak discharge. These impacts result in
rapid inflows to nearby river systems, which can cause degradation to the health of the river
system as well as damage the structural stability of the riverbanks. Impervious areas reduce the
amount of infiltration that can occur, and the water that can no longer filter into the ground has to
drain elsewhere, causing stormwater runoff. And as more land is converted into impervious
surfaces, more stormwater runoff can be observed. In one study, urbanization of a region was
shown to cause approximately a 200% increase in average annual flows as well as an increase in
11

the mean daily water flow (Jeong et al., 2012). The increased storm water that flows into the
river system can cause flooding and erosion of the riverbank, destroying habitats and threatening
infrastructure too close to the river.
2.3.1.2 Drinking water
As the population grows, so does the need for freshwater. Over the last century, the
demand for fresh water has more than tripled (Olden and Naiman, 2010). To obtain freshwater,
it has to be obtained from some source, whether that is an underground aquifer, lake, river or
other body of water, it depletes the amount of available water in the ecosystem. And if too much
is withdrawn from the environment, the water source may be reduced to a stream or run dry like
the Colorado River (USGS, 2012). This causes severe destruction to the natural ecosystem
because habitats will be destroyed and riverine organisms’ population will shrink or even die off.
Studying and implementing environmental flows will help reduce this impact, but a compromise
must be found to allow the ecosystem to be sustainable while still providing us with the water
that we need.
2.3.2 Agriculture
As the population of the earth grows there is an increasing demand on the need for food
and fiber. In order to accommodate this, farmers try to increase their yields and provide as much
food as possible to the ever-increasing demand by using more water and agrochemicals.
Transport of sediment and agrochemicals increase the risk of stream health degradation.
2.3.2.1 Irrigation
To support a growing population, more nutrients and water are needed to allow higher
crop yields. Nutrients are obtained through the applications of fertilizers that can be obtained
from local or regional resources. However for the needed water either the region has to have
12

sufficient rainfall to support the growth of the plants or other sources need to be sought out. The
obvious choice is often irrigation, especially in the dryer regions of the United States like Idaho
(USGS, 2013a). Studies have shown that in the United States the largest use of water is
irrigation, being 65% of all the water use between 1950 to 2005 (USGS, 2013a). And while
most of the water currently used for irrigation comes from groundwater, the reservoirs that
supply the groundwater are quickly shrinking or vanishing completely (Scanlon et al., 2012).
For example, it is well known than the Ogallala aquifer has been severely depleted due to more
water being withdrawn than can go through the soil to recharge it (Sophocleous, 2012). This
means that eventually farming will have to find new sources of water for irrigation, and the most
obvious choice is the river systems. And while taking some water from the environment has
little impact, withdrawing larger amounts leads to the sever degradation of the aquatic
ecosystem. Like mentioned above, however, due to the greater water need of irrigation, one
could suspect that the impact from irrigation would be much greater than drinking water if not
kept in check and regulated.
2.3.2.2 Runoff
As crops grow and are harvested the soil is disturbed, tilled, and left bare to withstand the
forces of nature. Most soil systems are held together by a vast root system that holds the soil in
place and allows the runoff to slow, infiltrate, and be used by plants; reducing the environmental
impacts. Farmland lacks this system for periods of the year when crops are not being grown.
This causes greater amounts of runoff and erosion to occur. While not nearly as severe as
stormwater in urban area, the runoff can still cause degradation in river systems, by rapidly
altering the water levels and clogging the streams with sediment from erosion. The settling of

13

this sediment downstream causes alteration of the flow patterns and can lead to habitat
destruction and reduced stream function.
2.3.3 Dams
Dams have been used since ancient times. There are accounts as far back as the 3rd to 4th
millennium BC at Jawa in ancient Jordan where the Jawa Dam was built to hold water for
irrigation of crops (Fahlbusch, 2009). Today dams provide a variety of services, they continue to
provide water for irrigation and drinking, but in addition, they also provide hydroelectric power,
protection from floods, and zones for recreational activities. By blocking the river and
controlling its flow, we are able to harness water for our needs.
While dams are very useful in harnessing resources from water, the alteration of the flow causes
a variety of negative impacts on the river ecosystem. These impacts include disrupting aquatic
organism migrations and habitats, altering water temperature, preventing the transfer of nutrients,
and interrupting the natural flow cycle (International Rivers, 2014). By retaining water and
releasing specific amounts, the structure and function of the river is altered both above and below
the dam. Dams that severely limit the water discharges can rest the order of the river system,
which reduces the usefulness of the river downstream. Above the dam, a pool of water
accumulates; this pool tends to be deep and hold cooler water, and while this pooling may seem
like an ideal place to introduce fish species and draw water from for irrigation and drinking, it
interrupts the natural habitats resets the ecosystem downstream (International Rivers, 2014). The
region downstream for a dam alters water temperatures (warmer or cooler based on the design of
the dam) and reduces nutrient levels making it difficult for aquatic species to survive.

14

2.4

Conservation Practices/Best Management Practices (BMPs)
To reduce the degradation of anthropogenic activities a variety of conservation practices

and best management practices (BMPs) have been introduced. The majority of these can be
implemented on agricultural lands however; several can also be used in urban settings as well to
improve water quantity in riverine ecosystems (SEMCOG, 2008).
2.4.1 Bioretention
Bioretention basins are shallow vegetated structures that can be used to control
stormwater runoff in both urban and agricultural areas. Designed to temporarily hold water and
promote infiltration by allowing the water to seep through the basin and into the groundwater
(SEMCOG, 2008). They can be implemented on large as well as small plots of land, which
makes them very versatile for urban applications where available land is a constraint and
impervious surfaces have increased the amount of runoff present. Also, the use of native
vegetation in the design helps create a sustainable system as well as provide an aesthetically
appealing area (SEMCOG, 2008).
2.4.2 Constructed Wetlands
Constructed wetlands are vegetated aquatic systems that provide flow regulation and
habitats for aquatic and terrestrial organisms. Designed to mimic natural wetlands, constructed
wetlands primarily improve the water quality however they also slow the flow of water through
the system, reducing peak flows (SEMCOG, 2008). Similar to bioretention systems except
where bioretention designs reach unsaturation a few after days after the storm event; constructed
wetlands can be designed to treat a continuous flow and never become unsaturated. They can be
implemented as a standalone treatment system connected to an outlet or can be installed in rivers

15

and lakes to improve the water quality and flow regime. This allows for the restoration of
natural ecosystems while providing improvement to the system.
2.4.3 Detention Basins
Detention basins are vegetated depressions that are used to temporally hold stormwater
runoff. Designed to catch stormwater runoff, promote infiltration, and reduce peak flows and
flooding (SEMCOG, 2008). They can be implemented in a variety of areas, including urban,
residential, and agricultural regions. Due to their ability to hold stormwater runoff, they are very
useful in urban areas were the increased impervious areas result in high peak flows. Here again
the use of native vegetation provides the benefit of sustainability as well as aesthetic appeal
(SEMCOG, 2008).
2.4.4 Filter Strips/Riparian buffer
Filter strips or riparian buffer zones are vegetated zones that reduce the quantity of runoff
before it enters rivers and lakes. They can be used to efficiently reduce water quantity along
bodies of water. Designed to restore or replicate natural systems found along water bodies, they
use native vegetation to slow water flows, promote infiltration, and stimulate plant uptake
(SEMCOG, 2008; Merritt et al., 2010); often implemented along rivers, lakes, and wetlands to
prevent degradation of the natural ecosystem as well as to prevent flood damage. The vegetation
in the BMP includes trees and shrubs as well as grasses and forbes (SEMCOG, 2008), which
provides flow reduction, filtration, and habitat creation. Here again the use of native vegetation
allows for the creation of a sustainable system, like in bioretention basins.
2.4.5 Vegetated Swale
Vegetated swales are shallow vegetated channels used to direct and control flows of
water. Designed to reduce the flow velocities, promote infiltration, and control the flow
16

direction (SEMCOG, 2008), by using the channel as a vegetated gutter. They can be
implemented in a variety of locations however heavily urbanized regions often lack the available
space needed to implement their design. Often used in agricultural lands where they slow and
direct runoff from farm fields to nearby streams. Also like all BMPs that utilize vegetation the
use of native species allows the design to be customized for the region allowing for a more
efficient and sustainable system (SEMCOG, 2008).
2.4.6 Native Grasses
Native grasses are plots of land that are restored to natural prairies and grasslands to help
control the flow of stormwater runoff. Designed to slow the flow of runoff and increase
infiltration by filtering the water through the grasses and into the soil (SEMCOG, 2008), they
utilize native vegetation to reduce maintenance costs and create a more sustainable system
(SEMCOG, 2008). These systems function much like bioretention and vegetated filter strips,
using plants and soil to control water flow, but in this case, the plant selection plays a bigger role
since it is also used for restoration projects. Often their designs tend to take up more space than
is available in heavily urbanized areas and are thus more common in areas where there is
available plots of land.
2.5

Optimization/Modeling
For many applications in environmental sciences, it is expensive, inefficient, and time

consuming to implement every possible solution and then monitor them to determine the best
design to use for the project goal. Also, BMPs’ effectiveness is dependent heavily on the
location, type of pollutant, and pollution concentration so just because a BMP works in one area
does not guarantee that it will work in another location. To solve these problems, modeling can
be used because it is inexpensive, effective, and fast. Models allow us to gather information
17

about how the entire system would respond to stressors as well as provide information on how
BMPs will perform in the region (Giri et al., 2012). The following sections discuses a couple of
the key models that are used to optimize and model the BMPs and the watershed.
2.5.1 SWAT
Soil and Water Assessment Tool (SWAT) is a commonly used watershed model that was
developed by USDA Agricultural Research Service (USDA-ARS) and Texas A&M AgriLife
Research (Texas A&M University System, 2013). It uses data like topography, water levels,
pollution concentrations, and weather events and predicts the effects of different managements
systems and BMPs on the environment (Texas A&M University System, 2013). Some of the
processes that can be simulated by this model include runoff, erosion, and sediment transport.
SWAT has been documented by a variety of studies which use it to predict outcomes in river
systems (Cibin et al., 2010; Lam et al., 2010; Setegn et al., 2010). In Setegn et al. (2010) study
of the Lake Tana Basin, SWAT was used to predict the stream flow based on the topography,
land use, soil, and climate condition. They concluded that the predicted values were very
similar to the observed values validating the use of the SWAT model in prediction of stream
flow. The study done by Lam et al. (2010) reached a similar result of accuracy of the SWAT
model but this time when applied to modeling point and non-point source pollution. However
the study done by Cibin et al. (2010) noted that depending on the location, SWAT was not
always as sensitive to paraemeters as desired. This can lead to the need to calibrate models to
make them more accutate which leads to the next model SWAT-CUP.
2.5.2 SWAT-CUP
When dealing with models as comprehensive as SWAT, it is expected that some
uncertainty would develop; the typical categories of uncertainty are conceptual uncertainty, input
18

uncertainty, and parameter uncertainty (Abbaspour, 2007). There is a variety of causes for these
uncertainties from the model being too simple to errors in the input data but instead of trying to
locate these issues, it is easier to calculate the uncertainty of the results of the model. This
provides insight to how accurate the model actually is. SWAT Calibration and Uncertainty
Analysis (SWAT-CUP) is a model that does just that. SWAT-CUP is used to calibrate and then
validate a SWAT model to insure it can predict known observations and then it preforms an
uncertainty analysis on the model to determine how accurate the predicted values are and what
range of error can be associated with them (Abbaspour, 2007). The uncertainty can be calculated
by using one of the following five methods, Sequential Uncertainty Fitting (SUFI-2),
Generalized Likelihood Uncertainty Estimation (GLUE), Parameter Solution (ParaSol), Mark
Chain Monte Carlo (MCMC), and Particle Swarm Optimization (PSO) (Abbaspour, 2007).
SWAT-CUP was used in a study of a Japanese river catchment and it was found that the
calibration and validation of the model lead to accurate predictions with low uncertainty values
(Luo et al., 2011).
2.5.3 R-SWAT-FME
R program language-Soil and Water Assessment Tool-Flexible Modeling Environment
(R-SWAT-FME) is another method that can be used to calibrate SWAT models and analyze the
uncertainty and sensitivity of the model. This model is still based on SWAT, so all of the
calibration and testing is based on a premade SWAT project. However when the model is run, it
is converted into a Fortran compatible version that can be run by RFortran (Wu and Liu, 2012);
this allows it to run faster, which is beneficial to large watersheds that can be modeled in SWAT.
The use of a Flexible Modeling Environment (FME) allows the SWAT model to undergo the
uncertainty and sensitivity analysis, performed by SWAT-CUP. To test their new model Wu and

19

Liu (2012) ran a case study using R-SWAT-FME. While running the case study on the Cedar
River in the Iowa River Basin, the model successfully converted the SWAT model to the
RFortran platform and was able to perform the calibration, sensitivity, and uncertainty with
satisfactory results (Wu and Liu, 2012).
2.6

Data Mining
When using models such as SWAT were there could be thousands of subbasins and data

being calculated for each, analysis of the results becomes challenging. Without the use of new
technology, the only way to evaluate this data would be to slowly go through the results looking
at all the data points manually to find any patterns that may be present and draw conclusions
about the results of the run. However, this method is a slow and tedious process, which would
have to be repeated if some data was overlooked or misread. A more efficient method would be
to use Data Mining. Data Mining is an approach that automatically analyzes the data and can be
taught to find specific ranges or develop patterns seen among databases (Alcalá-Fdez et al.,
2009). This is very useful and saves time during the process of running a project with lots of
results to analyze. There are a variety of ways to preform Data Mining including artificial neural
networks and genetic algorithms. Artificial neural networks are used to predict values given a
known dataset; it does this by mapping the inputs and outputs of a system a developing a system
that when given a new set of inputs with no outputs it can predict the missing values (Singh et
al., 2009). This makes it useful for predicting water flows in a river system. The model can be
calibrated with a known set of flow values and then applied to a range of years where flow data
is unknown and it will be able to generate the flow data. On the other hand, genetic algorithms
are mainly used for nonlinear and spatial optimization, by initiating a set of possibilities and
finding the best solution from them and then re-calculating based on the results (Arabi et al.,

20

2006). This is useful for identifying the optimal placement of BMPs within a watershed over
several years, by insuring that the application of BMPs will provide maximum reduction of
pollutants in the river system.
There are five phases through which the process of DM occurs. The first phase is called
data understanding; during this phase, data collection occurs and the program identifies the data
and begins the analysis of the data (Terzi, 2011). The second phase is known as data
preparation; here the data is refined and cleaned to allow easier final analysis (Terzi, 2011). The
next phase is called modeling; during this phase, multiple models are applied to the data to
determine the optimal results that are desired (Terzi, 2011). The fourth phase is known as
evaluation; in this step, the program preforms validation and confirms that the selected model is
the most efficient (Terzi, 2011). For the final step know as knowledge; the results and statistics
are displayed as the solution to the problem (Terzi, 2011).
Applications of Data Mining within the realm of stream heath are not well known,
however there are several studies that discuss the use of Data Mining to help improve modeling
and evaluation of stream health. One example of a study that incorporated Data Mining
techniques into a modeling method was performed by Chen and Mynett (2003). In this study, a
neural network technique known as a self-organizing feature map (SOFM), was added to the
steps needed to behind create a fuzzy logic model which was designed to predict algal blooms in
Taihu Lake in China (Chen and Mynett, 2003). The SOFM was used to make clusters of data
that could then be combined with expert knowledge to create the membership functions and
inference rules that are needed for the fuzzy logic model. The final model was tested on 2 sites
and had R2 values of 0.76 and 0.60 (Chen and Mynett, 2003). And Chen and Mynett noted that
including a optimization step and sensitivity analysis would increase the R2 value, the values
21

they obtained indicated that the method developed to create the fuzzy logic model was capable of
predicting the algal blooms. Another application of Data Mining in stream health studies was
done by Beck et al. (2014). In this study, the statistical selection of metrics for stream health
indicators was challenged. Primarily the selection of indicators is based on regressions between
other metrics to insure the most informative metrics are kept (Beck et al., 2014). To improve on
this Beck, et al. applied a feed-forward 3 layer neural network to a set on metrics, in the hope of
selecting the best metrics that corresponded to anthropogenic and natural characteristics. It was
found that the application of the neural networks was capable of determining the connection
between the characteristics and the metrics, however it was only slightly better than other
convention methods like linear regression and decision trees (Beck et al., 2014). The cause for
this is that the neural network used needs to have adequate training data which was not possible
in this study. Both of these applications show how Data Mining can be applied to stream health
studies and so long as it is used correctly can improve the models and techniques currently being
used.
2.7

Conclusion

Stream health is becoming an increasingly important aspect to monitor and regulate as the need
for fresh water increases. To help with this, monitoring environmental flows are a useful way to
regulate the use of water while still maintaining the sustainability of the aquatic ecosystem.
However, to truly use environmental flows, detailed studies are needed to provide adequate
analysis of the systems to determine optimal stream flows. However, running detailed
simulations leads to challenging analysis. Yet with the use of data mining, detailed studies will
become easier to analyze. Allowing for increased model resolution, which in turn increases the
accuracy of the results obtained from them, opening new areas of research.

22

3.

INTRODUCTION TO METHODOLOGY AND RESULTS
This thesis is in the form of two research papers that have been submitted to scientific

journals. The first paper, entitled “A Review of Macroinvertebrate and Fish Stream Health
Indices”, discusses the current uses and developments of macroinvertebrate and fish indices.
Macroinvertebrates and fish are among the most commonly used organisms for evaluating
stream health. Hence, there are many different stream health indices that have been developed
based on these organisms. The overall goal of this study is identify which indices should be used
in the subsequent study. To do this, 85 macroinvertebrate and fish stream health indices were
reviewed and commonly used/modified indices were identified and described. Furthermore,
individual components, collection strategies, and applications of stream health indices were also
discussed.
The second paper, entitled “Optimization of Conservation Practice Implementation
Strategies in the Context of Stream Health”, utilizes one of the stream health indices identified in
the first paper to develop a stream health model based on hydroecological variables. This model
was then used to evaluate BMP scenarios in order to maximize the watershed-level stream health
while minimizing the cost. To accomplish this, a biophysical model was built to estimate daily
streamflows within the study region. This model was calibrated and validated using long term
observed streamflows data from nine monitoring sites within the Saginaw Bay Watershed in
Michigan. Daily streamflows data from biophysical model were used to calculate 171
hydroecological indices for each stream segment within the Honeyoey Creek-Pine Creek
Watershed, which is a subbasin in the Saginaw Bay Watershed. Three dimensionality reduction
techniques (Spearman’s Rank Coefficients, Principal Component Analysis, and Bayesian
variable selection) were used to select a limited number of hydroecological indices that best
23

represented stream health. Selected variables were then incorporated in adaptive neuro-fuzzy
inference systems (ANFIS) to develop stream health predictor models. After the models were
developed, they were coupled with a genetic algorithm that generated and analyzed BMP
scenarios. This identified a near-optimum solution that maximized the stream health score while
minimizing the BMP implementation cost.

24

4.

A Review of Macroinvertebrate and Fish Stream Health Indices

4.1

Abstract
The focus of this review is to discuss the historical and current uses and developments of

macroinvertebrate and fish indicators. Macroinvertebrates and fish are commonly used
indicators of stream heath, due to their ability to represent degradation occurring at site specific
or within the entire river system, respectively. A total of 85 macroinvertebrate and fish indices
were reviewed, and the frequently used macroinvertebrate and fish indices are discussed in detail
in the context of aquatic ecosystem health evaluation. This review also discusses several types of
common components, or metrics, used in the creation of indices. Following this, the review will
focus on the different methods used for macroinvertebrate and fish collection, in both wadeable
and non-wadeable aquatic ecosystems. With the basics of macroinvertebrate and fish indices
discussed, emphasis will be placed on the application of indices and the different regions for
which they are developed. The final section will provide a brief summary of the benefits and
limitations of macroinvertebrate and fish indices.
4.2

Introduction
As the human population continues to grow, it can be expected that anthropogenic

activities will have impacts on the environment (Walters et al., 2009; Dos Santos et al., 2011;
Pander and Geist, 2013). This in combination with changing climates will only cause greater
impacts to the stream ecosystems (Meyer et al., 1999). To determine how climate change and
anthropogenic activities impact aquatic ecosystems, it has been recognized that monitoring the
health of streams is required to insure systems are able to function and will be able to provide
ecosystem services for future generations (USGS, 2013c). Stream health can be defined as the
combined analysis of impacts caused by anthropogenic activities on aquatic organisms, riparian
25

vegetation, and channel properties (Jeong et al., 2012). This definition describes aspects of a
very complex system, in which organisms interact with their surrounding and vice versa.
To evaluate stream health three components are often used, these three components are
the chemical, physical, and biological integrity of the surface water (Butcher et al., 2003a).
Traditionally of these three, chemical is the most commonly used to evaluate stream health;
however, recently it has be recognized that the use of biological integrity can be lead to a better
understanding of what is occurring in the ecosystem as well as identify the cause of degradations
(EPA, 2011). And with the high diversity found within aquatic ecosystems (Pander and Geist,
2013), there are many organisms that can be included into the decision making process to
evaluate the quality of the stream health. Another benefit to using biological indicators for
evaluating stream health is that they are not only take into account biological factors but also the
physical and chemical characteristics of the system (Brazner et al., 2007; Pelletier et al., 2012).
This is because biological factors are influenced by the physical and chemical characteristics of
the ecosystem. By using indicators to evaluate the biotic integrity, environmental resource
managers are able to identify degradation areas and can allocate resources to restore the
ecosystem’s with the greatest needs (Butcher et al., 2003a; Walters et al., 2009; Einheuser et al.,
2012; Pelletier et al., 2012), in the most cost-effective way (Neumann et al., 2003a). The overall
goal of this study is to provide a comprehensive review of macroinvertebrate and fish based
stream health indices. This will be done by first reviewing the individual components, collection
strategies, and applications of stream health indices. And then by exploring the
macroinvertebrate and fish based indices that have been developed as well as more detailed
reviews of the major indices being used in the field.

26

4.3

Metrics
The complexity of stream systems makes it difficult to create an index that is applicable

in multiple regions. To account for local characteristics when determining stream health an
index is often developed or modified; this insures that the analysis of the system accurately
describes what is occurring at the site. To create these personalized indices, individual
characteristics of the ecosystem are measured (Butcher et al., 2003a). These different
measurements are known as metrics. The information that metrics represent provide insight to
the condition of the ecosystem; from identifying the species richness (Butcher et al., 2003a;
Walters et al., 2009; Couceiro et al., 2012) to the number of trophic levels or functional groups
present in the ecosystem (Butcher et al., 2003a; Monaghan and Soares 2010; Oliveira et al.,
2011; Couceiro et al., 2012). The observations of each metric also allows for the calculation of
the index value, which then allows for comparison within and among (when possible) streams.
As the desire for sustainable water resource management grows so has the amount of data
collected from stream monitoring, this additional data has allowed for the creation of multimetric indices. These indices are able to provide a better understanding of what is actually
occurring in the environment since they have different level of sensitivity to different pollutants.
To create a multi-metric index the first step is evaluating a variety of metrics and preforming
statistical analysis to find unique responses to degradation (Butcher et al., 2003a). For example,
Butcher et al. (2003a) study initially included 42 candidate metrics and ended with 10 metrics
that were incorporated into their index for stream health. By using a three-step validation
process, they were able to select the metrics that best described the system. The sections below
describe some of the larger categories metrics can be split into: abundance, species richness, and
functional groups.
27

4.3.1 Abundance
Metrics that fall under the category of abundance are used to describe the number of each
species found in the rivers. This includes looking at the number of individual species collected,
like the number of Ephemeroptera collected per sample (Butcher et al., 2003a), or determining
the percentage of a species in a sample, like the percentage of Oecetis within a sample (Butcher
et al., 2003a; Brazner et al., 2007). In many multi-metric indices, the use of abundance metrics
is common (Houston et al., 2002; Boyle and Fraleigh, 2003; Butcher et al., 2003a; Couceiro et
al., 2012). Often abundance indicators are used to evaluate key or sensitive macroinvertebrate
and fish families, like in the EPT index, to provide information about the condition in the stream.
In general, streams with more organisms that are sensitive to stressors are less impacted by
anthropogenic degradation and vice versa (Johnson et al., 2013).
4.3.2 Species Richness
Metrics that fall under the category of species richness or number of taxa are used to
describe the biodiversity found in the ecosystem. This not only gives an overview of what is
found in the stream but it can also indicate the health condition of the stream. It has been shown
that regions with high biodiversity are in better condition and show less degradation while the
opposite condition, of low biodiversity, indicates a region with more degradation (Boyle and
Fraleigh, 2003). These are calculated by recording the number of different taxa taken from a
stream sample. In many multi-metric indices, including the Index of Biotic Integrity, the Benthic
Community Index, and government indices like the Alabama Department of Environmental
Management Index, include the use of species richness metrics (Houston et al., 2002; Boyle and
Fraleigh, 2003; Butcher et al., 2003a; Couceiro et al., 2012).

28

4.3.3 Functional Feeding Groups
Metrics that fall under the functional feeding groups category are used to study the
transfer of energy through the system. Benthic macroinvertebrates can be classified in one or
more of the following functional groups collectors, scrapers, shredders, and predators (Couceiro
et al., 2012). Meanwhile fish can be classified as omnivores, herbivores, insectivores,
planktivores, and piscivores (Karr, 1981). Each functional group has a specific role in the
ecosystem; collectors either filter or gather nutrients from the water, scrapers live on the rocks on
the streambed and scrap off organic material to eat, shredders break down biomass like leaves,
and predators actively hunt other organisms for a food supply. Similarly herbivores feed off
plant life within the streams, insectivores feed off the macroinvertebrates, planktivores feed off
microscopic organisms, and piscivores feed off other fish. Since macroinvertebrates and fish can
be found in every functional level (Karr, 1981; Barbour et al., 1999), they can be used to develop
an overall picture of the ecosystem. To use these metrics, the functional group of each organism
taxa is determined and then the distribution of functional groups within the system is used to
evaluate the status of the stream. Often changes in the functional feeding groups are driven by
nutrient changes (Smith et al., 2007), which means that the use of these metrics can provide
information about the chemical composition of the river system. Like with the species richness
metrics, many multi-metric indices, including the Index of Biotic Integrity, Benthic Community
Index, and government indices like the Florida Department of Environmental Protection Index,
use function feeding group metrics (Karr, 1981; Houston et al., 2002; Boyle and Fraleigh, 2003;
Butcher et al., 2003a; Couceiro et al., 2012).

29

4.4

Collection strategies
Since the majority of metrics used for indices are based on observations of

macroinvertebrate and fish communities found in rivers, strategies needed to be developed to
collect samples for analysis. And while individual strategies may change from study to study,
like number of samples and equipment used for sampling, all require the use individuals, either
volunteers or trained workers, to go out and take samples (Butcher et al., 2003a). Often times
this includes taking samples at different times of the year to determine the general condition year
round (Neumann et al., 2003b). However, the actual process of collecting the samples is not
uniform across all regions; this brings up the issue of the river size and the availability of
resources to take samples from larger bodies of water. To make a distinction about these
differences, the monitoring sites have been categorized as either wadeable or non-wadeable.
4.4.1 Wadeable Waterways
Streams are classified wadeable by the Environmental Protection Agency (EPA) when
they are shallow enough to take samples in without the use of a boat (EPA, 2006). It was
determined by the EPA that the major focus of the analysis of US waterways would be these
small wadeable streams since they represent about 90% of the perennial streams and river miles
in the United States (EPA, 2006). For macroinvertebrate sampling of these sites, the most often
used method is a collection net that is dragged along the bottom of the river to catch displaced
macroinvertebrates as the upstream environment is disturbed by collectors (Butcher et al., 2003b;
Couceiro et al., 2012). The organisms collected in the nets are then transferred to containers
(Barbour et al., 1999), which are then sent to the labs for analysis and identification. Since this is
easily preformed and the equipment is also relatively easy to obtain and use, the majority of
macroinvertebrate studies are performed in regions that are deemed wadeable (Butcher et al.,
30

2003a; Iliopoulou-Georgudaki et al., 2003; Justus et al., 2010; Couceiro et al., 2012; Li et al.,
2012). As for sampling fish communities in wadeable streams, both nets and electrofishing are
used (Terra et al., 2013). And while this is good for regions that have lots of lower order
streams, 1st order through 5th order (EPA, 2006); regions with lowland rivers and lakes cannot
benefit from the use of stream health indices if the only collection method was using wading
nets.
4.4.2 Non-wadeable Waterways
All other sources of aquatic ecosystems that do not fall into the wadeable regions are
classified as non-wadeable. These sites are too large for an individual to take samples without
the use of a boat (EPA, 2006). Nevertheless, understanding all of the waterways is important to
gain an understanding of the whole ecosystem, especially since larger rivers and lakes contain
the combined flows from many smaller streams and rivers potentially causing an increase in the
concentration of pollutants. Some effort has been put into creating indices that can be used on
non-wadeable water bodies. These water bodies include coastal regions (Muxika et al., 2005),
estuaries (Puente et al., 2008), large rivers (Angradi and Jicha, 2010), and lakes (Rossaro et al.,
2007; Launois et al., 2011). The sampling methods for these types of studies often included the
use of a boat sampling technique (Rossaro et al., 2007), and sometimes the use of a combination
of both wading and boat sampling techniques (Couceiro et al., 2012). And while these studies
provide insight to the impacts of anthropogenic activities on the health in these non-wadeable
regions, there is still much that is unknown about how macroinvertebrates and fish respond to
different anthropogenic stressors in these ecosystems (Rossaro et al., 2007), providing fields of
research for future studies.

31

4.5

Application
Studies involving macroinvertebrate and fish communities often focus on either defining

stream health in a region through the development of a new index (Butcher et al., 2003a) or use a
previously created index (Butcher et al., 2003b), testing an index to see if it can identify a known
stressor (Compin and Cereghino, 2003), comparing the results of different indices in one region
(Justus et al., 2010), or testing to see if a previously created index can be applied to a new region
(Muxika et al., 2005). The first type of study is preformed to provide an index that can be used
for streams in the region; stakeholders and governments to implement projects to improve the
regions that most require it can then use this. Testing already know indices is preformed to see if
the current index can be extended to include more results about the ecosystem. If the results of
the study are positive, this shows that the index can be applied to more regions and provide a
more complete understanding of the environment (Compin and Cereghino, 2003). The
comparison studies between different indices are very useful on several levels. First, it identifies
the best index to use for stream health evaluation in the region; secondly, it allows
generalizations to be drawn about indices and what they can determine. This was the case in the
study by Justus et al. (2010), where macroinvertebrates were not as capable as algae at detecting
low concentration changes in nutrients levels. However, the macroinvertebrates were able to
respond to the low nutrient concentrations better than the fish community. The final type of
study was to determine if an index can be applied to a new region. This is important because it
can expand the use of new indices to provide information about the region without having to
create a new index. This was found in the study of the AZTI Marine Biotic Index (AMBI) by
Muxika et al. (2005). The AMBI was applied to 6 different costal sites throughout Europe with
the goal of determining the suitability of the index for evaluating the health of the ecosystems

32

found there. These sites ranged from the Baltic Sea to the Mediterranean. After evaluating the
ABMI at all the sites with was decided that the AMBI was suitable for all European coastal
ecosystems. At the same time these studies have the chance of showing that the index in
question cannot be applied to the region without modifications.
4.6

Materials and Methods
Indices are evaluation systems used to assess conditions within an aquatic ecosystem and

rank them to allow comparison and identification of the regions of greatest degradation. They
can be designed for individual streams (Hu, et al., 2007) or can be used to analyze entire
ecoregions (Butcher, et al., 2003a). Below, we will discuss the frequently used macroinvertebrate
and fish indices in the context of aquatic ecosystem health evaluation.
4.6.1 Macroinvertebrate Indices
Since there are so many characteristics that can be observed in water bodies, from water
quality to presence of species indices, several components are often used to access stream health
or to understand how a certain stressor will impacts the ecosystem. One group of often-used
organisms for determining stream health are macroinvertebrates. They are useful at determining
local sources of degradation due their limited mobility with in the stream channel (Kerans and
Karr, 1994). Also, macroinvertebrates are sensitive to low levels of pollutants allowing for early
detection of stream degradation (Compin and Cereghino, 2003). Due to the frequent use of
macroinvertebrates (Flinders et al., 2008; Sharma and Rawat, 2009; Pelletier et al., 2012),
several indices have been developed and are used to monitor stream health. Table 1 presents 35
of the macroinvertebrate indices that were reviewed in this study. The first column indicates the
name of the index followed by the reference. The 3rd column indicates the index that it was
based on. The 4th column presents specifics about the index such as the number of metrics,
33

score trends, or aspect that is evaluated. And the final column indicates changes or
modifications made from the based index to create the new index. However, these indices are
generally originated from three common indices, which include Benthic Index of Biotic Integrity
(B-IBI), Hilsenhoff Biotic Index (HBI), and Ephemeroptera Plecoptera Trichoptera (EPT).
These indices can be either multi-metric, looking at many aspects of the ecosystem like B-IBI, or
focused on one particular characteristic of the environment like EPT. Out of the 35
macroinvertebrate indices listed in Table 1, 12 used EPT as their base index. This made EPT the
most often used base index. Of the modifications made to the EPT index, the most common was
the addition of metrics that evaluated other aspects of the streams, such as the presence of other
organisms or other functional feeding groups; this allowed the new index to provide a better
picture of the conditions within the stream as well as take into account local characteristics. The
following sections describe the three main macroinvertebrate indices.

34

Table 4.1. List of macroinvertebrate based indices.
Index Name
Reference
Base Index
Nutrient Biotic Index (Smith et. al.,
Hilsenhoff
2007)
Biotic Index

Tolerance Indicator
Values

(Meador et. al.,
2008)

Hilsenhoff
Biotic Index

Multimetric Index for
Castilla-La Mancha
Benthic Community
Index

(Navarro-Llácer
et al., 2010)
(Butcher et. al.,
2003a)

Original

Benthic Quality Index
Modified

(Rossaro et. al.,
2007)

Benthic
Quality
Index

Non-wadeable
Macroinvertebrate
Assemblage
Condition Index

(Blocksom and
Johnson, 2009)

Includes
EPT

Includes
EPT and
HBI

Specific Characteristics
Used to determine nutrient
tolerances of organisms for
evaluation of nutrient loading
in river systems
Used organism tolerances of
dissolved oxygen, nitrite plus
nitrate
(nitrate), total phosphorus, and
water temperature to evaluate
stream conditions
Uses 3 metrics to evaluate
conditions within streams
Uses 10 metrics, from 3
categories (Structural,
Functional, Conditional)
describing the
Macroinvertebrate community
to evaluate stream health
Scores organisms based on
indicator values, and sums the
scores of all present organisms
to determine the water quality.
Higher scores represent regions
with lower nutrient loads.
Uses 9 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum
of the metrics allow for
comparison, higher scores
indicate less degradation
35

Changes from Base Index
Uses nutrient tolerances instead of
organic pollutant tolerances

Uses dissolved oxygen, nitrite plus
nitrate
(nitrate), total phosphorus, and water
temperature instead of organic
pollutants
Looked at both fish and
macroinvertebrates
No changes
No changes

Looks are more than just
chironomids

No changes

Table 4.1. (cont’d)
Macroinvertebrate
Index of Biotic
Integrity

(Griffith et. al.,
2005)

B-IBI, HBI
and EPT

Macroinvertebrate
Multimetric Index

(Couceiro et. al.,
2012)

EPT

EPTC

(Compin and
Céréghino, 2003)

EPT

Ephemeroptera
Plecoptera
Trichoptera

(Walters et. al.,
2009)

EPT

ICI

(Walters et. al.,
2009)

B-IBI

Guapiacu-Macau
Multimetric Index

(Oliveira et. al.,
2011)

EPT

Uses 9 metrics to determine the
conditions within the stream,
with higher scores indicating
less degradation
Uses 7 metrics to evaluate
conditions within streams.
Higher scores indicate less
degradation
Uses metrics describing
Ephemeroptera, Plecoptera,
Trichoptera and
Coleoptera populations to
determine stream health.
Higher scores indicate healthier
streams
Uses metrics describing
Ephemeroptera, Plecoptera,
and Trichoptera populations to
determine stream health.
Higher scores indicate healthier
streams.

Based on macroinvertebrate
communities instead of fish

ICI used 9 metrics to evaluate
the conditions within the
stream, Higher scores indicates
less degradation
9 metrics used to evaluate the
conditions within streams.
Higher scores indicate healthier
scores

One metric of ICI was dropped due
to the fact that it was not
contributing to the analysis

36

No changes

Added the Coleoptera family

No changes

Table 4.1. (cont’d)
Invertebrate Species
Index

(Haase and Nolte,
2008)

HBI

AZTI Marine Biotic
Index

(Muxika et. al.,
2005)

Original

Abundance Biomass
Comparison

(Monaghan and
Soares, 2010)

Original

Family-level Biotic
Index (FBI/HBI)

(Hu et. al., 2007)

HBI

Chesapeake Bay IBI

(Weisberg et. al.,
1997; Pelletier et.
al., 2012)

B-IBI

Ranks macroinvertebrates on
their tolerances to nutrient
levels. Uses these scores to
determine the conditions within
streams. Higher scores indicate
greater sensitivity and thus
better stream conditions
Ranks organisms based on
sensitivy to pollutants and uses
the composite scores of each
site describe the conditions at
the site
Looks at the distribution of
individuals and biomass within
the region to evaluate pollutioninduced disturbances
Ranks macroinvertebrates on
their tolerances to organic
pollutants. Uses these scores to
determine the conditions within
streams. Higher scores indicate
greater sensitivity and thus
better stream conditions
Uses 15 metrics ranked 1, 3, or
5 and then summed to
determine the condition of the
system. Higher values indicate
better conditions

37

Looks at nutrient tolerances instead
of organic pollutant tolerances

No changes

No changes

No changes

Used to evaluate the Chesapeake
Bay region instead of the Tennessee
Valley
Used 15 metrics instead of 11

Table 4.1. (cont’d)
SIGNAL

(Besley and
Chessman, 2008)

Based on
rapid
biological
assessment
sampling
IBI

Benthic Index of
Biotic Integrity

(Kerans and Karr,
1994)

Alabama Department
of Environmental
Management Index of
Stream Health
Florida Department of
Environmental
Protection Index of
Stream Health
Mississippi
Department of
Environmental
Quality Index of
Stream Health
North Carolina
Division of Water
Quality Index of
Stream Health
South Carolina
Department of Health
and Environmental
Control Index of
Stream Health

(Houston et. al.,
2002)

Original

(Houston et. al.,
2002)

Original

(Houston et. al.,
2002)

Original

(Houston et. al.,
2002)

Original

(Houston et. al.,
2002)

Original

Uses sensitivity values
assigned to organisms found in
the stream to evaluate stream
health

No changes

Uses 13 metrics to evaluate the
condition of the system.
Higher values indicate better
conditions.
Uses 7 metrics to evaluate the
condition of the system.
Higher values indicate better
conditions, includes EPT.
Uses 7 metrics to evaluate the
condition of the system.
Higher values indicate better
conditions, includes EPT.
Uses 8 metrics to evaluate the
condition of the system.
Higher values indicate better
conditions, includes EPT.

Based on Macroinvertebrate
communities instead of fish
communities

Uses 3 metrics to evaluate the
condition of the system.
Higher values indicate better
conditions, includes EPT.
Uses 2 metrics to evaluate the
condition of the system.
Higher values indicate better
conditions, includes EPT.

No changes

38

No changes

No changes

No changes

No changes

Table 4.1. (cont’d)
B-IBI modified

(Roy et. al., 2003)

B-IBI

ICI modified

(Roy et. al., 2003)

B-IBI

NFAM

(SanchezMontoya et. al.,
2010)

Original

Yungas Biotic Index
based on 4 taxa

(Dos Santos et al.,
2011)

Original

ElPT

(Dos Santos et al.,
2011)

EPT

Uses 11 metrics to evaluate the
condition of the stream.
Higher values indicate better
conditions
Uses 10 metrics to evaluate the
condition of the stream.
Higher values indicate better
conditions
Uses the total number of
families to evaluate stream
health

Metrics used modified to fit
conditions in Georgia.

Metrics used modified to fit
conditions in Georgia.

No changes

Uses 4 metrics to evaluate the
No changes
condition of the stream.
Higher scores indicate better
conditions
Uses metrics describing
Looks at Elmidae taxa instead of
Elmidae, Plecoptera, and
Ephemeroptera taxa
Trichoptera populations to
determine stream health.
Higher scores indicate healthier
streams.

39

4.6.1.1 Benthic Index of Biotic Integrity
The Benthic Index of Biotic Integrity (B-IBI) is a multi-metric index developed by
Kerans and Karr (1994) and is based on the Index of Biotic Integrity (IBI) developed by Karr in
1981, which looked at the fish communities found in streams to determine the overall system
health (Karr, 1981). The B-IBI functions just like the IBI by the fact that is looks at organism
communities to evaluate stream health; however, the major change is that the B-IBI considers
macroinvertebrates instead of fish. The metrics used in the B-IBI are divided into three
categories taxa richness, taxa composition, and biological processes of the invertebrate
community in the aquatic ecosystem (Kerans and Karr, 1994). This allows for a detailed
analysis of the system and its condition. The thirteen metrics included in this index are total taxa
richness, intolerant snail and mussel species richness, mayfly richness, caddisfly richness,
stonefly richness, relative abundance of Corbicula, oligochaetes, omnivores, filterers, grazers,
and predators, proportion of individuals in two most abundant taxa, and total abundance. Each
metric is given a score from 1 to 5 based on the observations of the stream region in comparison
to a reference site that had no ecosystem degradation (Kerans and Karr, 1994). A higher score
indicates that the metric is closer to the reference site conditions. All of the metric scores are
then summed to provide the overall B-IBI score for that region, which can then be used to
evaluate the impacts of watershed management scenarios. Based on this analysis, sites that are
given lower scores exhibit greater degradation and thus can be selected for restoration projects.
For example, the original metric score ranged from 0 to 65 with a score of 65 representing a nonimpacted ecosystem and a score of 0 representing a heavily degraded ecosystem (Kerans and
Karr, 1994). Kerans and Karr (1994) showed that this index is effective of detecting industrial
degradations by taking samples above and below the industrial effluents. However, a universal

40

B-IBI does not exist and the B-IBI components need to be adjusted for different regions to better
describe the ecosystem. This was done in the study by Roy et al. (2003), where the B-IBI was
modified to better represent the local condition using 11 metrics instead of original 13 metrics.
Table A1 presents the metrics used in the B-IBI as well as what was added or removed in other
indices that are originated from the B-IBI. Of the indices listed, the most commonly removed
metrics were % Grazers and intolerant snail and mussel species richness; however, no commonly
metric were added. Overall, these changes were made to better represent the local conditions and
the ecosystem.
4.6.1.2 Hilsenhoff Biotic Index
The HBI is a commonly used (Butcher et al., 2003a) index developed by Hilsenhoff in
the 70’s (Hilsenhoff, 1987). It was based on the tolerances to organic pollutants of each
observed taxa in the river system (Goetz and Fiske, 2013). Therefore, HBI is used as an
indicator for chemical degradation within the river system. To use this index, samples are taken
from the river and used to determine the average tolerance value for the system (Hilsenhoff,
1987). After recording all of the tolerances the river system was ranked on a scaled from 0 to 10,
with 0 being the best (Goetz and Fiske, 2013). This value could then be compared to other sites
to determine the degradations across the region. To allow for a faster analysis of the system
Hilsenhoff provided a table describing the HBI values and their corresponding stream health
classification. The scores were grouped into water quality categories of Excellent, Very Good,
Good, Fair, Fairly Poor, Poor, and Very Poor. Each water quality score represented a different
level of organic pollution, for example an Excellent water quality category corresponds to no
apparent organic pollution and a score range of 0.00-3.50, while a Very Poor water quality
category corresponds to severe organic pollution and a score range of 8.51-10.00. Continued use
41

of the HBI has also led to the discovery that this index can also be used to identify regions with
low dissolved oxygen as well as other pollutants (Butcher et al., 2003a). This has become a very
useful measurement of stream heath to the point where is has been included as a metric in other
multi-metric indices (Butcher et al., 2003a) to provide information about the condition of the
stream with respect to organic pollutants.
Other studies have taken the concept used for the HBI and applied it to other stressors to
make new indices. One example of a new index that is based on the HBI, is the Nutrient Biotic
Index (NBI), which instead of considering the impacts of organic pollutants, it was developed to
assess the tolerances of organisms to nutrient loading within aquatic ecosystems and in particular
wadeable streams (Smith et al., 2007). To do this, two different indices were created, one for
nitrogen (NBI-N) and one of phosphorous (NBI-P). To calculate these indices, samples were
taken from the streams and used to determine average nitrogen and phosphorous tolerance scores
(Smith et al., 2007). These values were then used to compare between different streams and
locate the optimal concentration of each nutrient for the organisms (Smith et al., 2007). Smith,
et. al.(2007) identified the tolerances of the 164 collected taxa and ranked them from a 0 to 10
scale where 10 indicated high tolerance and 0 low tolerance (Smith et al., 2007). This allowed
for comparisons between different streams and evaluation of the nutrient loading in the study
region. Using the concept of HBI to evaluate nutrient loading was also used in Haase and Nolte
(2008). The Invertebrate Species Index (ISI) was developed to determine stream health and in
particular the impacts of eutrophication in Queensland, Australia (Haase and Nolte, 2008). They
scaled the sensitivity of macroinvertebrate species from 1 to 10, where a score of 10 means the
species is very sensitive to pollution and a score of 1 means the species is very hardy (Haase and
Nolte, 2008), just like the HBI and NBI. Once all the sensitivity scores were determined an
42

average score is calculated to represent the conditions within the stream (Haase and Nolte, 2008).
In Haase and Nolte (2008), the ISI were calculated for 203 species of macroinvertebrates, which
were used for comparison and evaluation of the upland streams in southeast Queensland,
Australia. However, they were noted that ISI species related scores that were calculated for the
stream classifications may not be accurate in other regions (Hasse and Nolte, 2008). This is
because certain species that were never present in a stream should not be included in the
calculation for the sensitivity score for that stream (Haase and Nolte, 2008). But if reference
conditions are rescored, this index would be useful for identifying nutrient based degradations
within stream systems. In addition to NBI and ISI, other stressors were developed for calculating
nutrient tolerances. A study by Meador et al. (2008) looked at organism tolerances to dissolved
oxygen, nitrite plus nitrate, total phosphorus, and water temperature. This shows how versatile
the concept of organism tolerances is, and the need for studies to explore tolerances of organisms
to other stressors. Table A2 presents the metrics used in HBI as well as what was added or
removed in other indices that are either based on or use HBI for analysis. Of the indices listed in
Table A2, the most common adjustment to the HBI was to change the stressor being evaluated.
The HBI looks at organism tolerances of organic pollutants, while the indices based on the HBI
look at organism tolerances to other stressors like nutrients or temperature.
4.6.1.3 Ephemeroptera Plecoptera Trichoptera
EPT is an indicator based on the observation of organisms of the Ephemeroptera
(mayflies), Plecoptera (stoneflies), and Trichoptera (caddisflies) families (Goetz and Fiske,
2013). These families are used because they are particularly sensitive to pollution levels within
the ecosystem (Compin and Cereghino, 2003); they have been used to identify local regions
impacted by pollution (Compin and Cereghino, 2003) and low dissolved oxygen (Butcher et al.,
43

2003a) as well as to provide an overall view of the conditions in a stream (Butcher et al., 2003a).
Their sensitivity to pollutants allows for early indication of problems in the ecosystem and
subsequent actions to be taken to repair the ecosystem before more degradation can occur
(Johnson et al., 2013). To use this index, the EPT richness and percent abundance is calculated
for each sample taken from the waterbody (Couceiro et al., 2012), and the overall conclusion
about the condition of the river can be made based on the results from all samples. In Couceiro
et al. (2012) study, the use of the EPT index was initially considered and preformed as expected
with higher scores representing less degraded sites. However, the range of scores obtained from
the sites was only 0 to 8, this was considered too small to be useful and was eliminated for
further analysis (Couceiro et al., 2012). In contrast, Oliveira et al. used EPT as one of the final 9
metrics for their multi-metric index with a range from 0.27 to 65.90 (Oliveira et al., 2011). EPT
was also part of the final list of metrics for the benthic community index developed by Butcher,
et al. (2003a). EPT can also be used as a standalone index. However, in the last two examples
EPT was used in multi-metric framework, which can then lead to a better understanding of the
system and what is affecting it (Butcher et al., 2003a; Oliveira et al., 2011;). In addition, the
EPT index has been modified by including invertebrates from the Coleoptera family, which is
known as EPTC (Compin and Cereghino, 2003). By adding an additional species to the index,
the sensitivity of the index to pollution is increased, and helps provide a better view of what is
happening in the ecosystem. EPTC index were used to evaluate both streams and large rivers
conditions (Compin and Cereghino, 2003). The scores from the index we grouped into 5
different classes, Excellent, Good, Good-fair, Fair, and Poor. The score ranges for each class
depended on the type of ecosystem evaluate; for example the 50 or more scores were considered
as “Excellent” for streams while for the large rivers, the scores more than 35 considered as

44

“Excellent”. Meanwhile, EPTC score less than 24 considers as a poor stream condition while
EPTC score less than 2 is poor for the large rivers. Distinction between streams and large rivers
in the EPTC method makes it more realistic because the ecosystems found in each are generally
quite different. However, EPTC is more recommended for evaluation of small bodies of water
like streams than large bodies of water like rivers.
Overall, it can be concluded that while EPT has been successfully used to evaluate stream
health conditions, it can be site specific and may not always be applicable to every system. Table
A3 C presents the metrics used in EPT as well as what was added or removed in other indices
that are either based on or use EPT for analysis. Of the indices listed, the most common change
to the EPT was the removal of the % abundance metric. In the cases when EPT % abundance
was removed additional organisms were added such as Diptera taxa richness, % Coleoptera taxa,
and % Oligochaete and leech taxa (Blocksom and Johnson, 2009). Another common addition to
the EPT index was functional feeding group metrics, like % Collector-filterer individuals,
Predator taxa richness, # Scrapers/# gatherers, # Shredders/total # collected, and Filterers* (%)
(Houston, et. al., 2002; Blocksom and Johnson, 2009). The addition of these metrics increases
the index’s ability to determine what is occurring within the ecosystem. For example the
addition of the functional feeding group metrics helps determine energy and nutrient flows while
the abundance EPT metrics identify pollution levels within the stream.
4.6.2 Fish Indices
Another group of organisms that is often used to evaluate stream heath are fish (Mack
2007; Zhu and Chang 2008; Krause et al., 2013). Karr (1981) listed seven advantages for using
fish for evaluating the stream conditions, which included (1) well known life-history, (2) species
found in many trophic levels (omnivores, herbivores, insectivores planktivores, and piscivores),
45

(3) easy identification, (4) understood by general public, (5) can be used to identify a variety
stresses, (6) are present in most water bodies, (7) can be easily connected with regulations.
Points 1, 2, 5, and 6 show the usefulness of fish as indicators to determine what is occurring
within the ecosystem; while points 3, 4, and 7 show that data collection and presentation is
relatively easy when compared to other types of organisms. Also unlike macroinvertebrates, fish
move throughout entire river systems, which allows for representation of the conditions within
an entire water system over a longer period of time (Karr, 1981). Another benefit to fish is that
they are impacted by changes in flow regime (Navarro-Llácer et al., 2010), which means that
they can be used to evaluate the impacts of flow altering structures, like dams, on the ecosystem.
All of these factors make fish based indices very useful for stream health monitoring.
Nevertheless, the system is not without flaws. Using fish communities for indices has its fair
share of limitations as well. Limitations include sampling selectivity, fish seasonal migrations,
and the cost of sampling. Table 2 shows 28 of the fish indices reviewed in this study. The first
column indicates the name of the index used in the study followed by the reference. The 3rd
column indicates the index that it was based on. he 4th column presents specifics about the index
like the number of metrics, score trends, or aspect that is evaluated. And the final column
indicates changes or modifications made from the based index to create the new index. Out of
the 28 fish indices listed in Table 2, 23 were based on the Index of Biological Integrity (IBI).
This made IBI by far the most often used base index as well as the most commonly modified. Of
the modifications made to the IBI index, the most common was the addition or subtraction of
metrics to provide a better picture of the ecosystems by taking into account local characteristics.
An example of this is the Fish Based Index for Lakes (FBIL) developed by Launois et al. (2011).
To take into account the differences for evaluating a lake in France; 3 metrics were added,

46

number of planktivore species, total biomass of strict lithophilic individuals, % total biomass of
tolerant individuals, and 10 of the 12 original metrics used in the IBI were removed (Launois et
al. 2011). By doing this the FBIL was able to identify urban and local pressures, like as the most
prominent sources of degradation for the French lakes. Of the indices listed in Table 2 few are
not based on the IBI, included in this category is the Tolerance Indicator Values Index (TIVI)
and the Stressor Gradients Index (SGI). The TIVI was developed by Meador et al. (2008) and
functions just like the HBI. However, instead of just looking at organic pollutant tolerances, it
looks at the organism tolerances to dissolved oxygen, nitrite plus nitrate, total phosphorus, and
water temperature (Meador et al., 2008). The scores from each river can be used to compare
between different rivers as well as indicate the levels of each component identifying where there
is too much or too little of each. The SGI was used by Angradi et al. (2009) and was used to
correlate stressor gradients, like total nitrogen, sediment toxicity, and water temperature, to
stream health. This was unique in the fact that the stressor gradients were correlated to
biological metrics to determine the conditions within the stream. The use of the SGI was able to
identify the anthropogenic impacts on the river systems of the Upper Mississippi River basin.

47

Table 4.2. List of fish based indices.
Index Name
Reference

Base
Index
HBI

Specific Characteristics

Changes from Base Index

Used organism tolerances of
dissolved oxygen, nitrite plus
nitrate, total phosphorus, and water
temperature to evaluate stream
conditions

Uses 10 metrics to evaluate the
condition of the system. Higher
values indicate better conditions
Uses 6 metrics to evaluate the
condition of the system. Higher
values indicate better conditions

Uses dissolved oxygen, nitrite plus
nitrate
(nitrate), total phosphorus, and water
temperature instead of organic
pollutants
Looked at both fish and
macroinvertebrates
Some metrics changed to match

Tolerance
Indicator Values

(Meador et. al.,
2008)

Mebane IBI

(Mebane, et. al.,
2003; Pelletier, et.
al., 2012)
(Krause et., al.,
2013)

IBI

(Zhu and Chang,
2008)

IBI

(Terra et. al.,
2013)

IBI

(Lyons, 2012)

IBI

Northern
Glaciated Plains
Index of Biotic
Integrity
Yangtze River
Index of Biotic
Integrity
Multi-metric
Index for Atlantic
Rain Forest
Streams
Cool–cold
transition IBI
Similarity Indices
Cool–warm
transition IBI

IBI

(Navarro-Llácer et Original
al., 2010)
(Lyons, 2012)
IBI

Metrics used were changed to match
the conditions in the Northern
Glaciated Plains Ecoregion

Uses 12 metrics to evaluate the
condition of the system. Higher
values indicate better conditions
Uses 6 metrics to evaluate the
condition of the system. Higher
values indicate better conditions

Metrics used were changed to match
the conditions in the Yangtze River

Uses 5 metrics to evaluate the
condition of the system. Higher
values indicate better conditions
Uses 4 metrics to evaluate the
condition of the system.
Uses 5 metrics to evaluate the
condition of the system. Higher
values indicate better conditions

Modified to represent the communities
in cool-cold rivers

48

Metrics used were changed to match
the conditions in the Atlantic Rain
Forest Streams

No changes
Modified to represent the communities
in cool-warm rivers

49

Table 4.2. (cont’d)
Fish Based Index
for Lakes

(Launois et. al.,
2011)

IBI

Fish Based Index
for Reservoirs

(Launois et. al.,
2011)

IBI

Esturine Multimetric Fish Index

(Harrison and
Kelly, 2013)

Original

Stressor Gradients

(Angradi et al.,
2009)

Original

European Fish
Index

(Musil et. al.,
2012)

IBI

Uses 6 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 9 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 14 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses relationship between abiotic
condition stressor gradients and
biological indicators to determine
conditions within the streams
Uses 10 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation

50

Modified for use in lakes instead of
streams

Modified for use in reservoirs instead
of streams

No changes

No changes

Modified to fit the conditions in
European streams

Table 4.2. (cont’d)
Czech Multimetric Index

(Musil et. al.,
2012)

IBI

Minnesota fish
index of biotic
integrity

(Wan et. al., 2010) IBI

Index of Biotic
Integrity

(Karr, 1981)

Original

Fish Community
Index

(Jordan et. al.,
2010)

Original

Multi-metric
Index for the
Coastal Plain
Ecoregion

(Esselman et al.,
2013)

IBI

Uses 4 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 9 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 12 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 3 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 4 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation

51

Modified to fit the conditions in Czech
Republic streams

Modified to fit the conditions in
Minnesota streams

No changes

No changes

Modified to fit the conditions in the
Coastal Plain Ecoregion

Table 4.2. (cont’d)
Multi-metric
Index for the
Northern
Appalachians
Ecoregion

(Esselman et al.,
2013)

IBI

Multi-metric
Index for the
Northern Plains
Ecoregion

(Esselman et al.,
2013)

IBI

Multi-metric
Index for the
Southern
Appalachians
Ecoregion

(Esselman et al.,
2013)

IBI

Multi-metric
Index for the
Southern Plains
Ecoregion

(Esselman et al.,
2013)

IBI

Multi-metric
Index for the
Temperate Plains
Ecoregion

(Esselman et al.,
2013)

IBI

Uses 4 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 5 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 3 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 4 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 4 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation

52

Modified to fit the conditions in the
Northern Appalachians Ecoregion

Modified to fit the conditions in the
Northern Plains Ecoregion

Modified to fit the conditions in the
Southern Appalachians Ecoregion

Modified to fit the conditions in the
Southern Plains Ecoregion

Modified to fit the conditions in the
Temperate Plains Ecoregion

Table 4.2. (cont’d)
Multi-metric
Index for the
Upper Midwest
Ecoregion

(Esselman et al.,
2013)

IBI

Multi-metric
Index for the
Western Mountain
Ecoregion

(Esselman et al.,
2013)

IBI

Multi-metric
Index for the
Xeric West
Ecoregion

(Esselman et al.,
2013)

IBI

Coldwater Multimetric Index

(Kanno et. al.,
2010)

IBI

Mixed-water
Multi-metric
Index

(Kanno et. al.,
2010)

IBI

Uses 4 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 4 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 4 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 5 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation
Uses 7 metrics to evaluate the
conditions within the stream.
Scores obtained from the sum of
the metrics allow for comparison,
higher scores indicate less
degradation

53

Modified to fit the conditions in the
Upper Midwest Ecoregion

Modified to fit the conditions in the
Western Mountain Ecoregion

Modified to fit the conditions in the
Xeric West Ecoregion

Modified to fit the conditions in
Coldwater streams

Modified to fit the conditions in
Mixed-water streams

4.6.2.1 Index of Biological Integrity (IBI)
The Index of Biotic Integrity (IBI) is a multi-metric index introduced by Karr in 1981. It
is based on fish communities and widely used to determine the overall stream health (Karr,
1981). Karr listed three assumptions that are needed for the use of this index; (1) the fish sample
is a balanced representation of the community at the site, (2) the chosen site is representative of
the region in which the IBI is being applied, and (3) the personal charged with analysis of the
collected data are trained (Karr, 1981). If any of these assumptions is violated, the results of this
index can be misleading. Originally, the IBI was composed of 12 metrics, which can be grouped
in one of the three following classifications; (1) species richness and composition, (2) tropic
composition, and (3) fish abundance and condition (Hu, et al., 2007). Each of these metrics is
given a score of 1, 3, or 5 based on undisturbed reference sites where a score of 5 is the best.
After all the scoring the metrics, the individual scores are summed to provide the IBI score for
each site. The IBI score ranged from 0 to 60 and were grouped into 9 stream classes, Excellent,
Excellent-Good, Good, Good-Fair, Fair, Fair-Poor, Poor, Poor-Very Poor, and Very Poor. Under
this class system, and stream scoring a 23 or less would be classified as Very Poor while scores
of 57-60 would be considered Excellent. Even though, the 9 stream classes are applicable in
different regions, caution should be taken when correlating the IBI score from different regions.
This is because the IBI is a region specific index, and the values presented here or in other
studies may not accurately represent different regions. In order to address this issue, Karr (1981)
also provided description of what may be found in the streams for different scoring system. This
helps IBI technique to be more transferable for multiregional studies of stream health evaluation.
The IBI has been applied and modified in a variety of studies (Zhu and Chang, 2008; Smith and
Sklarew, 2012; Krause et al., 2013). In Europe, a commonly used index of stream health is the

54

Fish-Based Index (FBI) (Launois et al., 2011). With 15 metrics and scores ranging from 0 to 100
with 100 being the best, the FBI was successful to identify degraded water bodies, but lacked the
ability to detect the cause, which was believed to be agricultural related activities and stressors
(Launois et al., 2011),. This shows that the selection of metrics for FBI is vital to ensure that the
regional characteristics and stresses are taken into account.
Recently, Lyons (2012) modified the IBI for use in perennial coolwater streams in
Wisconsin. This required the creation of two different IBIs the Cool-Cold Transition (CCT) IBI
and the Cool-Warm Transition (CWT) IBI. Each index uses five metrics to represent the
ecosystems (Lyons, 2012). Then the metric was given a score of 0, 10, or 20 based on the
analysis of the sample. Next, the metric scores are summed to calculate the IBI score giving a
range of scores from 0 to 100 with 100 being the best just like the FBI (Lyons, 2012). Overall,
the results showed that while both indices identified disturbed areas with low scores; the CWT
index performed better than the CCT index. However, due to the wide variation in scores for
similar stream sites, it was recommended to use multiple samples and then a mean or median
score to classify the systems instead of a single (Lyons, 2012).
A different study that utilized the IBI found that rare taxa had major impacts on the results of IBI
scores (Wan et al., 2010). In Wan et al. (2010), the sensitivity of the IBI was tested and it found
that the presence/removal of rare taxa, often considered an indicator of lower degradation, can
lower the IBI score by 38 points. While this was a concern, this result of the study still shows
that the IBI is sensitive to the conditions within the stream, and as long as the metrics are
weighted correctly, the results of the index can provide accurate information about stream
degradation. Table A4 presents the metrics used in IBI as well as what was added or removed in
other indices that are either based on or use IBI for analysis. Of the indices listed in the table, the
55

most common change to the IBI was the removal of most of the original metrics like the species
richness and composition of darters, suckers, and sunfish (except green sunfish), and the
proportion of green sunfish (Karr, 1981). This was done in combination with the addition of
other metrics to represented local characteristics. For example, number of coolwater species,
percentage tolerant species, % invertivore/piscivore individuals, and % native large river taxa
(Kanno et al., 2010; Esselman et al., 2013). By modifying the IBI to such an extent allows for
better understanding of what is occurring within the ecosystems by taking into account local
characteristics.
4.7

Conclusions
Throughout this review a variety of macroinvertebrate and fish indices were discussed,

each had benefits and limitations. In macroinvertebrate indices the B-IBI was capable of
identifying industrial and chemical degradation (Kerans and Karr, 1994) as well as changes
brought about by land use change like urbanization (Roy et al., 2003). However, these indices
are site specific (Kerans and Karr, 1994), which means that to insure accurate evaluation of
stream health the metrics needs to be fitted to the conditions of the site. The HBI, NBI, and ISI
were all able to determine organism tolerances to pollutants whether organic (HBI) (Goetz and
Fiske, 2013) or nutrient (NBI, ISI) (Smith et al., 2007; Haase and Nolte, 2008). The HBI also
has the benefit that it can be used as a metric of other multi-metric indices (Butcher et al.,
2003a), allowing for better understanding of the ecosystems Yet again these indices may not be
applicable to other regions (Haase and Nolte, 2008) because the tolerances of species may
change based on the natural conditions within different habitats. The EPT index is capable of
detecting low levels of degradation due to the sensitivity of the Ephemeroptera (mayflies),
Plecoptera (stoneflies), and Trichoptera (caddisflies) families (Goetz and Fiske, 2013). And like

56

the HBI, the EPT index can all be included in other multi-metric indices (Butcher et al., 2003a).
However if these families do not appear frequently in a river system the index is not very useful
in evaluating stream health (Couceiro et al., 2012). In terms of fish indices, the most commonly
used and modified index is the IBI. This index allows for the evaluation of entire regions (Karr,
1981) while at the same time being easily modified to take into account different climates
(Lyons, 2012). However, the selection of the metrics used in this index is vital for interpretation
of the results (Wan et al., 2010; Launois et al., 2011).
4.7.1 Benefits
There are many reasons that a macroinvertebrate or fish index would be applied to a river
system; whether it is to indicate to presence of pollutants (Karr, 1981; Johnson et al., 2013) or to
determine the optimal nutrient load for the system (Smith et al., 2007), or to compare levels of
degradation between streams (Karr, 1981; Kerans and Karr, 1994). Meanwhile,
macroinvertebrates are sensitive to very low levels of degradation at local levels; therefore they
can used by stakeholders to detect and correct problems before more serious damage occurs
(Barbour et al., 1999; Flinders et al., 2008). On the other hand, fish indices can be used to
evaluate the conditions on a regional scale, due to their mobility and lifespans (Karr, 1981). This
makes them useful for watershed managers, since they can be used to identify problems found
throughout the entire watershed. Another benefit to using macroinvertebrate and fish indicators
are also sensitive to the development of storage structures like dams (Navarro-Llácer et al., 2010;
Marzin et al., 2012) and can be used to monitor the impact of anthropogenic changes to the flow
levels in the rivers. Besides being able to be used for a variety of different stream health indices,
macroinvertebrates and fish can also be used to identify the stressors causing the degradation of a
site, based on the number of sensitive taxa present. And the wide distribution of
57

macroinvertebrates and fish over the trophic levels allows for a better understanding of what is
actually happening within the system and what changes are occurring due to anthropogenic
impacts. When all of this is taken into account, macroinvertebrates and fish can be seen as a
very versatile indicator of stream health and the impacts humans have on the aquatic ecosystems
in which they reside.
4.7.2 Limitations
While macroinvertebrates and fish are useful indicators of stream health (Karr, 1981;
Iliopoulou-Georgudaki et al., 2003) there are still limitations to their application as well as
regions, like lakes and large rivers, that require further research so that actions can be taken to
reduce the levels of degradation found in freshwater ecosystems.. Often indices will be used to
compare streams within regions, and while some regions like the Northern Lakes and Forest
Ecoregion in the United States are relatively uniform (Butcher et al., 2003b) majority of regions
are not. So if indices are applied outside the region the results may be very inaccurate. This
means that streams indices should always be modified to fit the characteristics of the region of
study.
Thorough out this review, different aspects and applications of macroinvertebrate and fish
indices have been discussed. However, majority of these works were performed in wadeable
streams, describing how the ecosystem responds to different stressors. However, non-wadeable
streams are not nearly as studied. Therefore, future studies should be focus on these waterbodies
to better understand how anthropogenic activities impact the overall aquatic ecosystems at the
local and regional levels.

58

5. Optimization of Conservation Practice Implementation Strategies in the Context of
Stream Health
5.1

Abstract
Sustainability of freshwater ecosystems is vital to insuring their continued use. This study

introduces a new strategy to improve stream health to a desirable condition at the lowest cost by
optimizing the best management practice (BMP) implementation plan. To accomplish this,
several hydrological models including the Soil and Water Assessment Tool (SWAT) and
Hydrologic Integrity Tool (HIT) were integrated and the results were used to develop stream
health predictor models. All of the models were guided by a genetic algorithm to design the
watershed-scale management strategies. Five BMPs were considered for use on agricultural land
within the study area: cover crop, forest, native grass, no tillage, and residue management.
Results from the hundreds of simulation identified eight unique BMP implementation scenarios
that resulted in overall excellent stream health scores for the Honeyoey Creek-Pine Creek
Watershed in Michigan. From these scenarios it was found that the most often implemented
BMP was no tillage (20.43%) followed by residue management (17.26%) and forest (15.88%).
Finally, one scenario was selected at the end for having maximized stream health score while
minimizing the implementation cost. The technique introduced here can be successfully adapted
in different regions and used by stakeholders and decision makers to identify the optimal solution
from both environmental and economic points of view.

5.2

Introduction
With the continued growth of the human population, the need for freshwater has

significantly increased. This increase in freshwater demand is mainly attributed to agricultural
59

production, which accounts for 70% of freshwater consumption worldwide (Worldometers,
2014). However, the impacts of anthropogenic activities is not only limited to water quantity but
also quality due to point and non-point source discharges (Walters, et al., 2009; Dos Santos, et
al., 2011; Giri et al., 2014 Pander and Geist, 2013). For example, water withdrawals and dams
alter the flow regime of river systems (International Rivers, 2014), while agricultural production
increases nutrient and sediment loads within these systems (USGS, 2013a). These activities
degrade river systems, which in turn impact the humans that use freshwater resources for
drinking or recreation. To protect the surface water resources in the United States, the Clean
Water Act was passed (CWA, 1972), with the goal of restoring the chemical, physical, and
biological integrity of the Nation’s waterways. In the framework of the CWA, chemical water
quality has greatly improved and point source discharges have largely been eliminated (EPA,
2012b). Despite all of these improvement, recent assessment has revealed that degradation of
aquatic ecosystems continue and even accelerated since the program was started (EPA, 2011).
EPA (2011) report concluded that a central focus on chemical water quality is not enough to
achieve healthy streams due to river system complexity and the effect of compounding stressors
(Magbanua, 2012). This shortcoming led to the introduction of bioassessment in river monitoring
(Jeong, et al., 2012). Bioassessment is the use of a stream’s biological components to evaluate
the conditions within the stream (Barbour et al., 1999). The hope is that bioassessment, with
chemical and physical assessments, provide a more comprehensive view of stream health,
allowing watershed managers to accurately address water quality issues.
Stream health can be defined as the combined quality of chemical, physical, and
biological components of a stream (USGS, 2011). In order to describe and measure stream
health, the concept of biological integrity was introduced (Karr and Dudley, 1981). Biological

60

integrity describes the ability of an ecosystem to support and maintain a balanced, integrated,
adaptive community of diverse organisms in its original stage and before disturbance due to
human intervention (Karr and Dudley, 1981). Bioassessments use indices of biological integrity
(biological indicators) to evaluate the quality of a system by monitoring the organisms living in a
stream (Pander and Geist, 2013). Biological indicators take into account not only the biological
characteristics of the system but the physical and chemical conditions as well (Brazner, et al.,
2007; Pelletier, et al., 2012).
Environmental flow is another element of bioassessment that is critical to monitor
conditions within river systems. Environmental flows describe the patterns and quantity of water
needed to support both the environment and human needs (King, et al., 2009; Poff, et al., 2010;
Chen and Zhao, 2011). Environmental flows initially focused on maintaining the minimum
levels of water needed to sustain the ecosystem (Alfredsen, et al., 2012). However, the scope of
environmental flows were further expanded to replicate the natural flow cycles in both timing
and volume (King, et al., 2009; Alcázar and Palau, 2010; Poff, et al., 2010; Chen and Zhao,
2011; ).
By incorporating both biological indicators and environmental flows, watershed
managers are able to identify degraded streams and can work on implementation plans to restore
the ecosystem (Butcher, et al., 2003a; Neumann, et al., 2003a; Walters, et al., 2009; Pelletier, et
al., 2012). However, there are several challenges with implementation of bioassessment
techniques in large and diverse watersheds. First, it is expensive and impractical to perform
monitoring in every stream segment to evaluate stream health condition. Second, it is not
possible to examine every possible management scenario to effectively improve overall stream
health condition.
61

Modeling provides an inexpensive and effective way to explore stream health conditions
beyond the monitoring sites or examining the impacts of management practices to improve water
quality (Arabi et al., 2006; Einheuser et al., 2012; Giri et al., 2012; and Einheuser et al., 2013b).
However, to the best of our knowledge no work has been done to optimize best management
practice implementation plan in the context of stream health, which is the goal of this study. The
specific objectives of this study were to: (1) predict stream health conditions beyond the
monitoring points based on a biological indicator, (2) develop series of management practice
scenarios that maximize stream health conditions while minimizing the cost in a watershed.
5.3

Materials and Methodology

5.3.1 Study Area
The region used for this study was the Honeyoey Creek-Pine Creek Watershed (Figure
5.1), located in the central eastern region of the Lower Peninsula of Michigan. This is a 10-digit
hydrologic unit code (HUC 0408020203) watershed the Honeyoey Creek-Pine Creek watershed
is part of the Pine 8-digit HUC watershed and flows into the Tittabawassee and Saginaw 8-digit
HUC watersheds. The final outlet for the region discharges into Lake Huron at the mouth of the
Saginaw River. With a total area of 106,131 ha, the region is dominated by agricultural land
(52%), followed by forest and wetland (both 20%), and finally pasture (8%). With such a large
percentage of agricultural land, water flow throughout this region is in high risk to be altered by
water withdrawal for irrigation or degraded by agrochemical nonpoint source pollution.

62

Figure 5.1. Saginaw Bay Watershed (HUC 040802) and the Honeyoey Creek-Pine Creek
Watershed (HUC 0408020203).

5.3.2 Data Collection
5.3.2.1 Physiographic Data
Several spatial and temporal dataset were used to characterize the physiographic features
of the study area for model developments. These datasets included topography, land use, soil
characteristics, climate data, and management practices. The 30 m spatial resolution National
Elevation Data set from the US Geological Survey (USGS) was used to represent the topography
of the region (NED, 2014). The 30 m spatial resolution 2012 Cropland Data Layer (CDL) from
the United States Department of Agriculture-National Agricultural Statistics Service (USDA63

NASS) was used to represent the land use for the region (NASS, 2012). Pre-settlement
vegetation circa1800 maps were obtained from the Michigan Natural Features Inventory (MNFI)
and were used to represent the pre-settlement land use from the mid-1800s (MNFI, 2014). Soil
characteristics data was obtained from the Natural Resources Conservation Service (NRCS) Soil
Survey Geographic (SSURGO) Database at a scale of 1:250,000 (NRCS, 2014a). Climate data
(precipitation and temperature) were obtained from the National Climatic Data Center (NCDC).
Within the Saginaw Bay watershed, 16 precipitation and 13 temperature stations were used to
supply daily climatological information. These datasets spanned from 1990 to 2012. Other
climate data such as relative humidity, solar radiation, and wind speed were obtained by using
the SWAT weather generator (Neitsch et al., 2011). The stream network and subbasins were
created from a 1:24,000 National Hydology Dataset plus (NHDPlus) and obtained from the
Michigan Institute for Fisheries Research. Each of the 553 subbasins from this dataset contains
an individual stream and is considered to be physicochemical, geomorphological, and biological
unique (Einheuser et al., 2013a). Management operations, schedules, and crop rotations were
modified from SWAT default values, as presented by Love and Nejadhashemi (2011) for the
study area.
5.3.2.2 Biological data
Fish are commonly used for stream health assessment. This is due to their wide
distribution and easy identification as well as their sensitivity to a variety of stressors (Karr,
1981; Mack, 2007; Zhu and Chang, 2008; Navarro-Llácer et al., 2010; Krause et al., 2013).
Furthermore, they provide regional evaluation of stream conditions due to their seasonal
migrations (Karr, 1981).

64

For this study, the Index of Biotic Integrity (IBI) was used to evaluate stream health
conditions. The IBI, first introduced by Karr (1981), is a multi-metric index that looks at the
species diversity, trophic composition, and abundance of the fish community to evaluate stream
health. Each metric used in the index is given a score of 1, 3, or 5, with 5 representing nondisturbed conditions within the stream (Karr, 1981; Lyons, 1992). All the metrics scores are
summed to provide the IBI score for the stream, ranging from 0 to 100, which can then be used
to compare between different streams. IBI scores were divided into five stream health classes
(very poor, poor, fair, good, and excellent) based on Lyons’ warmwater stream IBI (1992), the
ranges for each steam health class are presented in Table 5.1.
Table 5.1. IBI stream heath class ranges (adapted from Lyons, 1992).
Stream Health Class

IBI Score Range

Very Poor

0 – 19

Poor

20 – 29

Fair

30 – 49

Good

50 – 64

Excellent

65 – 100

Data for this index was obtained from the Michigan Department of Natural Resources
Fish Collection System and the Michigan River Inventory dataset (Seelbach et al., 1997).
Samples for these indices were collected from June to September from 1996 to 2003. Due to
limited number of biological sampling points within the Honeyoey Creek-Pine Creek Watershed
(18 sites), all sampling locations (193 sites) throughout the Saginaw Bay Watershed was used to
develop stream health predictor models based on the IBI scores (Figure S1).

65

5.3.3 Modeling Process
In order to accomplish the goal of this study, two phases were established, the
development phase that was performed in the Saginaw Bay Watershed and the scenario phase
that was performed in the Honeyoey Creek-Pine Creek Watershed (Figure 5.2). Within the
development phase, steps are taken to develop a stream health model. These include the
calibration/validation of a biophysical Soil and Water Assessment Tool (SWAT) model to obtain
daily flow rates for all stream segments within the Saginaw Bay Watershed; and calculation of
171 hydrological indices based on long-term daily flow rate for all streams using the
Hydrological Index Tool (HIT). Given the large number of variables (171) three dimensionality
reduction techniques were explored to identify the best variables for the stream health models.
Once the models were developed, they were then used in the scenario phase where different
management practices were applied to agricultural lands. The scenario phase utilizes the
evolutionary algorithm technique known as Genetic Algorithms (GA) to find a near optimum
solution by maximizing the stream heath index and minimizing the implementation cost. The
steps in this phase include, estimating long stream flow rate using SWAT, using HIT to calculate
flow variables selected in the development phase for the stream health models, using ANFIS to
estimate IBI scores for each stream segment, calculating the weighted average value for the
study area, sing the GA for selection and placement of best management practices (BMPs) on
agricultural land, and examining the stream health scores and cost. This process is repeated until
the near optimum solution is achieved.

66

Figure 5.2. Flow diagram of the development and scenario phases.
5.3.3.1 Development Phase
5.3.3.1.1

Soil Water Assessment Tool

Daily stream flow data was modeled throughout the Saginaw Bay Watershed using the
physically based Soil Water Assessment Tool. SWAT was developed by the USDA Agricultural
Research Service (USDA-ARS) and Texas A&M AgriLife Research (Texas A&M University
System, 2013). In addition to streamflow, the model is capable of estimating sediment, nutrient,
and pesticide loadings using physiographical and climatological characteristics of the study area
including land use, river network, topography, soils, agricultural rotations and scheduled
management operations, precipitation, and temperature (Gassman et al., 2007; Neitsch et al.,
2011).
67

The model was calibrated and validated against daily stream flow data from nine USGS
gauging stations (Figure S2) from 2001 to 2010. Calibration was preformed from 2001 to 2005
while validation was preformed from 2006 to 2010. Calibration and validation were considered
successful when the following criteria were met: a) Nash-Sutcliffe model efficiency coefficient
(NSE) is larger than 0.5, b) root-mean-squared error-observations standard deviation ratio (RSR)
is smaller than 0.7, c) and percent bias (PBIAS) < +/- 25. After calibration and validation, the
SWAT models were run from 1996 to 2003 to provide the stream flow values needed for the
development step.
5.3.3.1.2

Hydrological Index Tool

Flow has been called the “master variable” (Power et al, 1995) and the most influential
factors affecting stream health index concerning fish (Einheuser et al., 2013a). The Hydrological
Index Tool (HIT) was developed by the USGS as part of the Hydroecological Integrity
Assessment Process (HIP) (EPA, 2006). It is used to calculate 171 biologically relevant stream
flow indices or hydroecological indices originally introduced by Olden and Poff (2003). These
indices can be divided into five main categories: magnitude, frequency, duration, timing, and rate
of change (Henriksen et al., 2006). Magnitude refers to the availability of water within the
system. This availability is used to describe habitat suitability (Richter et al., 1996). Frequency
refers to how often certain events, such as droughts and floods, occur. These events can have
major impacts on stream organism populations allowing these indices to describe population
dynamics (Gupta, 1995). Duration refers to the length of certain events, such as droughts and
floods. This can be used to indicate the impacts of these events on the system or they can be used
to monitor organism growth stages (Poff et al., 1997). Timing refers to when certain events,
such as seasonal flooding, occur within the year. These indices can be used to monitor organism

68

growth or the impacts severe water events (droughts and floods) have on populations (Richter et
al., 1996). Rate of change refers to the speed at which stream conditions change. These indices
can be related to organism access to adequate water as well as impacts on population caused by
extreme water events (Poff et al., 1997).
For this study, HIT indices were calculated for 193 stream segments with biological
monitoring sites. The sites were further divided into two additional sets according to the river
continuum concept. The river continuum concept describes the predictable physical and
biological patterns seen in different regions of rivers (EPA, 2014). Based on the river continuum
concept, three physically based categories are used to describe the ecological regions of a river
system; headwaters (stream orders 1-3), medium-sized streams (stream orders 4-6), and large
rivers (stream orders > 6) (Vannote et al., 1980). However, because not enough monitoring sites
available for stream order 7 (two sites), the Saginaw Bay watershed were divided into two
classes (stream orders 1 – 3 and 4 – 7) for further model development.
5.3.3.1.3

Variable Selection

To eliminate redundant variables in the final stream health models, three dimensionality
reduction techniques (Spearman’s Rank Coefficients, Principal Component Analysis, and
Bayesian variable selection) were used. These techniques were used for three stream grouping
(all, orders 1 – 3, and orders 4 – 7). Up to six variables were selected from each dimensionality
reduction/stream grouping scenarios and used to develop stream health model using ANFIS.
Spearman’s rank correlation (ρ) is a nonparametric technique that is used to identify the
correlation between two sets of variables. This is a useful technique for selecting the most highly
correlated variables from a larger set (Einheuser et al., 2013a; Einheuser et al., 2013b). For this
study, spearman’s rank correlations were calculated between all predicator variables (171 stream
69

flow indices obtained from the HIT model) and the response variable IBI. Variables that were
significantly correlated (ρ < 0.05) to the stream heath indices (IBI) were selected. Selected
variables were analyzed to insure they were not highly correlated (ρ < 0.7) to other selected
variables to reduce variable redundancy (Waite et al., 2010).
Principle component analysis is a multivariate statistical technique that is used to
orthogonally convert observation sets of possibly correlated observations into sets of linearly
uncorrelated variables or principal components (PCs) (Pearson, 1901). The order of PCs is so
that the first PC (PC1) accounts for the most variability within the original dataset (Schölkopf et
al., 1998). All PCs after that (PC2, PC3…) account for nonincreasing levels of variability within
the original dataset. This means that the original dataset can be described by the first few PCs
(Schölkopf et al., 1998). This was used to find which HIT indices best described the stream heath
scores from the IBI. All of the final variables were selected from PC1.
Bayesian variable selection is a technique that utilizes reversible jump Markov Chain
Monte Carlo (MCMC) methods to sample all possible combinations of variables to find the best
fitting model (Carlin and Chib, 1995). This is done by selecting different samples sets of the
variables (MCMC chains) and randomly adding or removing variables and comparing the
previous sets model likelihood to the modified sets model likelihoods. The model with the better
likelihood is kept and then modified again. This process continues until all variable sets
converge to a model likelihood, at which point the best variables were determined. Five MCMC
chains, with different initial variable sets, were used to insure that the best variables were
selected.

70

5.3.3.1.4

Stream Health Model

The final step in the model development phase is creation of stream health predictor
models capable of estimating IBI score for each stream segment. The artificial neural network
technique known as ANFIS or adaptive neuro-fuzzy inference system was used for development
of stream health models. ANFIS is a multi-layer network that utilizes artificial neural networks
and fuzzy logic to create membership functions (MFs) while minimizing the output errors (Jang,
1993). MFs describe the degree of belonging for each element in a set from full exclusion (0) to
full inclusion (1) (Hamaamin, 2014). In this study, ANFIS was used because is well-suited to
capture uncertainty and complexity of ecological and environmental systems and their data
(Metternicht, 2001; Chen and Mynett, 2003; Adriaenssens et al., 2004; Einheuser et al., 2013a).
The variables selected by Spearman’s rank coefficient, PCA, and Bayesian variable
selection techniques were used for ANFIS model developments. This was done using the Fuzzy
Logic Toolbox in MATLAB R2013b (MathWorks, 2014). Five MFs (triangular, trapezoidal,
generalized bell, Gaussian, and Gaussian composite) were used to represent the variable sets. All
possible combinations of MFs, and stream grouping were evaluated to determine the best model.
At this point, the different variables obtained from three variable selection techniques sets were
used to develop ANFIS models. The criteria for selecting the best variable set were the highest
coefficient of determination (R2) and the lowest root-mean-square deviation (RMSE) values.
After determining which variable selection method provided the best models for each
stream grouping, 10-fold cross-validation was used to train, test, and select the best ANFIS
model. In the10-fold cross validation technique, the models are trained on 90% of the data (9
folds) and tested on the remaining 10% (1 fold). This was repeated 10 times for each model a

71

different fold being used for testing each time (Hamaamin et al., 2013). Similar results for R2 and
RMSE were used to determine the final best model.
5.3.3.2 Scenario Phase
5.3.3.2.1

Develop the Reference Condition

Stream reference condition is a benchmark condition in which the environmental impacts
of anthropogenic activities can be measured (Stoddard et al., 2006). In order to develop a
reference condition for the Honeyoey Creek-Pine Creek Watershed a new SWAT project was
created that used pre-settlement landuse. Vegetation circa1800 maps were obtained from the
Michigan Natural Features Inventory (MNFI). These maps were based on detailed fields notes
taken from a General Land Office survey of the state from 1816 to 1856 (MNFI, 2014).
To identify which stream order experienced the most change between the pre-settlement
and current conditions a Wilcoxon Signed-Ranks statistical test was performed. The Wilcoxon
Signed-Ranks test is a non-parametric technique that calculates the difference between two sets
of observations by ranking the differences between paired values (Pratt, 1959). Due to the
presence of ties and zeros in the differences between paired values, the method presented by
Pratt (1959) was used to calculate the p-values for the variables within each set of stream orders.
Furthermore, percent changes were also calculated between all data sets.
5.3.3.2.2

Best Management Practices

To reduce the degradation within the stream systems a variety of best management
practices were applied to agricultural land throughout the region based on Natural Resources
Conservation Service recommendations. These included cover crop, forest, native grass, no
tillage, and residue management.

72

Cover crops are primarily used to reduce sediment erosion during the times when the
field would normally be barren (post-harvest till planting) (Arabi et al., 2007). However they can
help control the flow of water during this time as well by slowing run off and reducing peak
flow. Winter wheat was chosen as the cover crop due to its popularity in the study area. Use of
no tillage or conservation tillage reduces the amount of soil disturbances that occur on the field
from agricultural practices (Tuppad and Srinivasan, 2008). This increases the amount of residue
on the field that can also reduce runoff from the field. Application of residue management
reduces the amount of tillage applied to a field post-harvest. This leaves behind crop residue that
helps protect the soil and controls the flow of water on the site during the time between the
harvest and planting. A residue management of 1000 kg/ha was chosen for use in this study.
Returning the agricultural land to forest has been successfully used to improve stream flow
conditions within the region of application (Qui, et al., 2011). In addition, the pre-settlement
conditions for the majority of the agricultural land in the Honeyoey Creek-Pine Creek Watershed
was forest. Therefore this BMP attempts to return the region to its original pristine condition, by
improving both water quality and quantity of the region. Converting the agricultural land to
native grass, is an attempt to restore the land to a more natural condition by cultivating native
grasses on the land (Einheuser et al., 2013b). This reduces the amount of runoff from the site and
can significantly lower the peak flow discharge.
In addition to environmental flow benefits of BMP implementation, associate cost is an
important factor to consider in order to develop practical solution. This cost includes the
implementation and maintenance costs for each BMP per unit area (Table 5.2). BMP costs were
obtained from the NRCS Typical Statewide Average Practice Costs for 2014 (NRCS, 2014b).

73

Table 5.2. BMP costs per unit area (NRCS, 2014b).
BMP

Cost/ha

Cover Crop

$167.29

No Tillage

$42.01

Residue Management

$29.65

Forest

$355.19

Native Grass

$160.62

SWAT BMP implementation procedures were adopted from other studies (Arabi et al.,
2007; Tuppad et al., 2010; Woznicki and Nejadhashemi, 2012; Giri et al., 2014). However, for
the forest BMP, land use condition from pre-settlement map was used to replace current
agricultural land.
5.3.3.2.3

Optimizing BMP Placement

In order to evaluate the overall stream health condition and identify near optimal
locations in which to apply Best Management Practices while minimizing implementation cost a
GA technique was proposed. This is because the intractable size of the solution space. Given six
scenarios (five control practices and one no-BMP) and 185 target agricultural land parcels, there
are 6185 ≈ 10144 possible implementation options. To accelerate the process, first we maximized
the overall stream health score and next calculated the implementation cost in postprocessing. The overall stream health scores were calculated for every stream in the subbasin (m
= 553 stream segments) and then used to calculate a weighted average stream health score for the
entire region. Equation 1 was used to calculate the weighted average stream health score:
𝑆𝑡𝑟𝑒𝑎𝑚 𝐻𝑒𝑎𝑙𝑡ℎ 𝑆𝑐𝑜𝑟𝑒 =

74

∑𝑚
𝑖=1 𝐼𝐵𝐼𝑖 ×𝐿𝑖
∑𝑚
𝑖=1 𝐿𝑖

(1)

Where IBIi is the individual stream health scores, Li is the individual stream lengths, and
m is the number of subbasins in the region.
The fitness is the quality of a solution under a defined metric, in our case this is stream
health. So given an assignment of practices to all target reaches, 𝑃𝑖 , from the set of all possible
assignments, 𝒫, the fitness is optimized at:
𝑚𝑎𝑥
𝑃∈𝒫ℎ𝑒𝑎𝑙𝑡ℎ(𝑃)

(2)

We do not require the cost to be constrained during the application of the algorithm, but
we do save the cost value for distinguishing the feasibility of results with similar health scores.
The actual algorithm can be described as an evolutionary DNA encoding process. The
data are similar to a DNA strand where the array of control practices, ranging from zero to five
(corresponding to the possible control practices), represent a string of nucleotides. Each
candidate solution then gives an assignment of actions which can be quantified in terms of
fitness. With such an arrangement we need information on the river reaches, the control
practices, and the cost per practice to derive the watershed's overall health and the total cost of
any applied actions.
The candidates are initialized by providing a pseudo-random assignment of control
practices to each targeted reach. There are 𝑛 such candidates produced and the top 𝑚 solutions
are chosen for the initial set. No fitness control is applied during generation, but the highest 𝑚
members will survive to the evolutionary stages. Both 𝑚 and 𝑛 are user-selectable values which
we choose to be 𝑚 = 20 and 𝑛 = 100. Through a series of generations new candidates are
created from the most fitted candidates of the previous generation. In each step a crossover with

75

mutation is performed to create offspring instances capturing the desirable traits of the parent
candidates.
Each generation is pruned to contain only the 20 members with the highest stream health
score, these candidates will become the parents of the new generation. This combination of the
high performing parents into a child candidate is called crossover. This process exploits the good
genetic traits of the parents while approaching a locally maximum solution. Crossover occurs for
the control practices by first selecting two candidates from the group of parents. Next a random
point is chosen to break the genes in half. The left of the break point is copied from the first
parent's chromosome's genes (list of control practices) and the right similarly with the second
parent's chromosome's genes. Similarly create another child by swapping in parent one's genes
on the right and parent two's on the left. Note that if the split point is chosen on one of the
endpoints no crossover will occur.
With only crossover applied a local solution will be reached, but this may be a poor
solution. To combat this, a random selection of control practices of a selected candidate are
changed randomly to inject diversity into the population of candidate solutions. This effect is
called mutation. Mutation allows for new solutions to be explored by increasing diversity and
lowers the chance remaining fixed on a local maximum solution. The probability of mutation is
done randomly on a gene by gene basis at a rate of 5%. Rarely, due to its probabilistic nature,
mutation may not occur on a chromosome.
After the new set of children are created from crossover and mutation, the old generation
is discarded and the set of children become the new potential parents to again be pruned into the
20 fittest candidates. There are several possible termination conditions. The simplest is to
terminate after a fixed number generations have been evaluated and selecting the most fit
76

member of this generation. Because we were interested in finding many solutions of high stream
health we terminated after sufficiently many candidates had IBI scored of over 65, which
represent an excellent stream health condition according to Lyons’ (1992) warmwater stream IBI
classification.
5.4.

Results and Discussion

5.4.1 SWAT Model Calibration and Validation
The SWAT model evaluation criteria (NSE, RSR, and PBIAS) for the calibration and
validation periods are reported in Table 5.3. As can be seen in Table 5.3 all NSE values are
above 0.5 with a range from 0.534 to 0.776 for calibration and 0.518 to 0.788 for validation, all
RSR values are under 0.7 with a range of 0.474 to 0.682 for calibration and 0.46 to 0.694 for
validation period, and all PBIAS values are within a range of +/- 25 with a maximum value of
23.167 for calibration and -24.267 for validation. This indicates that the developed SWAT
models met the satisfactory evaluation criteria and can be used to estimate daily streamflow data.

77

Table 5.3. Statistical criteria for the calibrated SWAT model at different USGS gauging stations
(Figure S2).
USGS
Station
Number

NSE
RSR
PBIAS
NSE
RSR
PBIAS
Calibration Calibration Calibration Validation Validation Validation

04144500

0.625

0.612

21.300

0.601

0.632

9.421

04148500

0.699

0.548

23.167

0.751

0.499

8.276

04148140

0.534

0.682

10.389

0.539

0.679

-24.267

04147500

0.631

0.608

8.508

0.647

0.594

-10.969

04151500

0.586

0.643

15.144

0.679

0.566

12.624

04157000

0.776

0.474

16.215

0.788

0.460

11.746

04155500

0.632

0.607

19.483

0.580

0.648

0.953

04156000

0.734

0.516

6.479

0.733

0.517

9.850

04154000

0.577

0.650

6.324

0.518

0.694

13.792

5.4.2 Variable Selection for the Best Stream Health Model
ANFIS was used to determine which dimensionality reduction technique (Spearman’s
Rank Coefficient, PCA, and Bayesian Variable Selection) and variables should be used for
development of stream health predictor model. This was done by evaluating all combinations of
variables and number of MFs (two, three, and four). Table 5.4 reports the R2 and RMSE for all
selected sets of variables. For stream grouping, all, 1 – 3, 4 – 7, the Bayesian Variable Selection
had the highest overall R2 values; 0.571, 0.514, and 0.699, respectively. This indicates that the
variables selected by this method represented the region most accurately when compared to all
sets of selected variables. This led to the selection of variables from Bayesian technique for the
1–3 and 4–7 stream grouping for the final stream health model. The second best technique is
Spearman’s Rank Coefficient based on average R2 value. In addition, this technique produced the
78

lowest average RMSE, which is a robust performance. For the PCA, the performance was poor
for stream order 1-3 (the lowest R2 value); however, perform better than Spearman’s Rank
Coefficient for stream order 4-7 based on higher R2 and the lowest RMSE between all techniques
and stream grouping.
Table 5.4. Coefficients of determination for each variable selection technique.
Number of
Variables

Stream
Orders

RMSE

R2

Spearman’s Rank Coefficient

3

All

15.869

0.447

Spearman’s Rank Coefficient

2

1–3

14.371

0.481

Spearman’s Rank Coefficient

2

4–7

16.507

0.547

Principal Component Analysis

3

All

18.695

0.569

Principal Component Analysis

2

1–3

18.329

0.228

Principal Component Analysis

3

4–7

13.082

0.634

Bayesian Variable Selection

4

All

14.703

0.571

Bayesian Variable Selection

3

1–3

19.895

0.514

Bayesian Variable Selection

2

4–7

13.313

0.699

Variable Selection Technique

The final stream health ANFIS model for stream orders 1 – 3 is a linear model with three
variables as follow: variability of flow values in January (MA24), variability across annual
maximum flows (MH18), and variability in reversals (RA9). All variables had two Gaussian
MFs. The RMSE and R2 for this model are 19.895 and 0.514, respectively. For stream orders 4 –
7, a linear model with two variables was selected, these variables were: the number of days
where flow increased from the previous day (RA5) and the skewness of the annual maximum
flows (MH19). RA5 had four Gaussian MFs, while MH19 had 3 Gaussian MFs. The RMSE and
R2 for the model are 13.313 and 0.699, respectively.
79

5.4.3 The Reference versus Current Conditions
A comparison was done between the pre-settlement (reference) and current conditions in
order to identify the impacts of human activates on local and regional stream health. The streams
were classified according to Lyons’ (1992) stream IBI classification method (Table 5.1 and
Figure 5.3),

Figure 5.3. Pre-settlement and current stream health classes.
80

Overall the stream health score for the Honeyoey Creek-Pine Creek Watershed was
reduced from 73.1 for the pre-settlement to 59.3 for the current condition. To determine the
differences between the pre-settlement and current stream health scores the percentage of the
total stream length for the study area was calculated (Table 5.5). Excellent and Good health
scores were combined together as Satisfactory, and Poor and Very Poor health scores were
combined together as Unsatisfactory. Based on these percentages, the pre-settlement conditions
have more stream length within the Satisfactory stream health group, with 80% of the total
stream length compared to the current conditions value of 66%. However the greatest difference
between the two conditions is the percentage of stream length that falls within the Fair stream
health group. This means that 14% (80%-60%) of the stream with the Satisfactory conditions
were downgraded to Fair condition while only 4% (14%-10%) of the streams were upgraded to
Fair condition from Unsatisfactory. As the land use changed from the pre-settlement to the
current conditions, the stream health score for 58 % of the streams decreased. While 26 % of the
streams showed improvement and 16 % showed no change (S3). Therefore, as it was expected,
the overall stream health pattern shown degradation in the majority of the watershed.

81

Table 5.5. Comparison of pre-settlement and current conditions based on the length of stream for
each health class.
Stream
Health Class

Presettlement

Current

Excellent

69.36%

44.70%

Good

11.13%

21.06%

Fair

5.69%

23.92%

Poor

1.49%

1.66%

Very Poor

12.33%

8.66%

Overall Condition

Presettlement

Current

Satisfactory

80.49%

65.76%

Fair

5.69%

23.92%

Unsatisfactory

13.82%

10.32%

The Wilcoxon Signed-Ranks statistical test was performed to evaluate the significant
differences between reference and current conditions for different stream orders. p-values and
percent changes are presented in Tables 6 and 7. The maximum stream order in the Honeyoey
Creek-Pine Creek Watershed was five.
Table 5.6. Wilcoxon Signed-Ranks test and percent changes between the pre-settlement and
current conditions for stream orders 1 through 3.
Stream Order 1

Stream Order 2

Stream Order 3

Variables
% change

p-value

% change

p-value

% change

p-value

Stream Health Score

-22

<0.01

-22

<0.01

-23

<0.01

MA24

120

<0.01

122

<0.01

135

<0.01

MH18

-70

<0.01

46

<0.01

-38

<0.01

RA9

1

<0.01

0

0.075

-1

0.053

RA5

3

<0.01

2

<0.01

4

<0.01

-30

<0.01

-28

<0.01

-35

<0.01

MH19

*Bold values indicate significance probability at the 0.01 level

82

Table 5.7. Wilcoxon Signed-Ranks test and percent changes between the pre-settlement and
current conditions for stream orders 4 through 5 and for all streams.
Stream Order 4

Stream Order 5

All Stream Orders

Variables
% change

p-value

% change

p-value

% change

p-value

Stream Health Score

-20

<0.01

18

0.373

-21

<0.01

MA24

133

<0.01

83

<0.01

123

<0.01

MH18

-114

<0.01

-48

0.011

-83

<0.01

RA9

-1

0.960

1

0.343

0

<0.01

RA5

3

<0.01

5

0.048

3

<0.01

-34

<0.01

-37

<0.01

-31

<0.01

MH19

*Bold values indicate significance probability at the 0.01 level

Of all streams within the region, 5th order streams were the least affected by the land use
changes; with three variables (MH18, RA9 and RA5) as well as the stream health scores showing
no significant change (p > 0.01) between the two scenarios (Table 5.7). In contrast, 1st order
streams were the most impacted with all variables and the stream health scores showing
significant change (p < 0.01) between the two scenarios (Table 5.6). This significance was
mirrored by the evaluation of all streams. But this is expected since the region it predominantly
1st order streams.
Of all the variables evaluated, the largest positive percent change was seen in the MA24
variable (variability of flow values in January), with an average increase of 123 % between all
stream orders (Table 5.7) and a maximum change of 135 % for 3rd order streams (Table 5.6).
This shows that there is more variability of flow values within the month of January in the
current condition compared to the pre-settlement condition. The largest negative percent change
was seen in the MH18 variable (variability across annual maximum flows), with and average
83

decrease of 83 % between all stream orders (Table 5.7) and a maximum decrease of 114 % for
4th order streams (Table 5.7). This shows that the variability of annual maximum flows has
decreased in the current condition when compared to the pre-settlement condition. The variable
with the least percent change was the RA9 variable (variability in reversals), with an average
change of 0 % between all streams orders (Table 5.7). This variable also was the variable that
showed the least significant change between the two conditions. For the stream health score
overall, there was an average decrease of 21 % for all stream orders (Table 5.7) with a maximum
decrease of 23 % for 3rd order streams (Table 5.6). This indicates that on average the health
streams within the region has decreased by 21%.
5.4.4 Optimizing BMP Placement
After the development of the stream health model, the Genetic Algorithm was used to
create different BMP scenarios. Without the guide, the selection and placement of the BMPs can
be random; however, the genetic algorithm was used to improve the overall stream health score
by optimizing the type and placement of various BMPs. After one hundred and eighty two
iterations, the maximum stream health score reached a plateau at a value of 71.036. Because IBI
scores of 65 and above present the excellent stream health condition, it was decided that only
scenarios with the scores ≥ 65 would be further analyzed to identify the best scenario. From
these one hundred and eighty two scenarios, eight unique scenarios satisfied the aforementioned
criteria even though they had different BMP compositions. Table 5.8 shows the percentage of the
total agricultural area allocated to BMP implementation.

84

Table 5.8. Percentage of allocated area to BMP implementation within the agricultural lands in
scenarios that overall IBI score is greater than 65.
BMP Practices
BMP
Scenario

No
BMP

Cover
Crop

Forest

Native
Grass

No
Tillage

Residue
Management

1

14.98%

16.11%

11.99%

14.13%

20.47%

22.32%

2

24.63%

21.31%

7.76%

15.95%

14.62%

15.73%

3

19.70%

13.90%

13.58%

14.01%

24.93%

13.88%

4

13.89%

19.92%

12.53%

13.44%

20.31%

19.91%

5

18.48%

11.30%

14.70%

12.31%

20.03%

23.18%

6

10.94%

16.97%

21.57%

13.97%

22.92%

13.63%

7

11.35%

13.71%

24.84%

13.04%

22.78%

14.29%

8

11.38%

17.18%

20.08%

18.84%

17.39%

15.13%

* Bold values indicate dominant BMP(s) per scenario

Based on Table 5.8, no tillage was the dominant BMP for three of the eight scenarios
(Scenarios 3, 4, and 6) and had an average implementation of 20.43%. This indicates that no
tillage was the most often implemented BMP. After no tillage, both residue management and
forest were the dominant BMP for two of the eight scenarios (Scenarios 1 and 5 and Scenarios 7
and 8 respectively). Even though both residue management and forest were dominant for two
scenarios, residue management was applied on more area than forest (17.26% and 15.88%
respectively). Another interesting observation is the dominant BMP for Scenario 2. No BMP was
the most often implemented BMP for this scenario, covering 25% of all agricultural land (Table
5.8). This is interesting, because it shows that we still can achieve excellent stream health
condition (IBI ≥ 65) in the study area while not implementing BMPs on all agricultural lands.

85

For the eight BMP implementation scenarios with the overall IBI scores above 65, the
individual stream health scores are different. Figure 5.4 shows the distribution of the individual
stream health scores for all scenarios. To better represent this data, scores were categorized by
stream order.

Figure 5.4. Distributions of individual stream health scores for all BMP scenarios.
As can be seen in Figure 5.4, all stream orders had maximum stream health scores of 100
and a median stream health score between 60 and 80. However, 1st and 3rd order streams were
the only stream orders that spanned the entire stream health range (0 to 100). While 2nd and 4th
order stream had outliers within the lower ranges of the stream health scores. In contrast, 5th
order streams never dropped below a stream health score of 35. This shows that 5th order streams

86

had the smallest range of scores. Individual scenario stream health score distributions can be
found in Figures S4 through S11 in the Appendix.
To better examine the upper ranges of the individual stream health scores. The percentage
of stream health scores greater than 65 for each scenario were calculated and presented in Table
5.9. While this is not indicative of the overall steam health score, it does show which scenario
was able to reach the excellent stream health class for more of the total stream length. As can be
seen in Table 5.9, Scenarios 4, 6, and 7 had the largest amount stream length within the excellent
stream health class, with a percentage of 61%. While Scenarios 3 and 8 had the least amount of
stream length within the excellent stream health class, with a percentage of 41%. Therefore if
decision makers wanted to improve individual stream health scores, Scenarios 4, 6, or 7 are the
logical choice. However, if improving the overall stream health score is the focus any of the
scenarios can be used
Table 5.9. Percentage of individual streams with a stream health score greater than 65 for all
BMP scenarios.

BMP Scenario

Percentage of Individual Steams
within the Excellent Stream
Health Class

1

46%

2

46%

3

41%

4

61%

5

46%

6

61%

7

61%

8

41%

87

5.4.5 Associated cost of BMP Implementation Scenarios
Finally, the costs of the eight acceptable scenarios were calculated. This was the final
criteria for selecting a near optimum solution that account for both stream health and associated
implementation cost. Furthermore since the overall stream health scores for all scenarios was
identical, it became the key factor for identifying the best scenario. Table 5.10 displays each
scenario’s overall steam health scores and total cost of implementation.
Table 5.10. All unique BMP scenarios with an overall stream health score above 65.
BMP Scenario

Stream Health Score

Cost

1

71.036

$1,964,889

2

71.036

$1,822,138

3

71.036

$1,985,994

4

71.036

$2,081,966

5

71.036

$1,941,840

6

71.036

$2,580,815

7

71.036

$2,668,190

8

71.036

$2,599,286

* Bold values indicate scenario with the lowest cost.

As can be seen in Table 5.10, Scenario 2 had the lowest cost among all scenarios with a
stream health score greater than 65. This made it the best of all the scenarios because it was able
to achieve the maximum stream health score (71.036) while costing the least to implement
($1,822,138). This result matches the BMP selection percentages presented in Table 5.8 in which
no BMP was selected for the majority of agricultural areas, and a solution with fewer
implemented BMPs will cost less.

88

Figure 5.5 displays the locations of the BMP implementations within the study area as
well as the individual stream health class for Scenario 2. A final comparison between Scenario 2
and the current condition was preformed to evaluate the impacts made by implementing Scenario
2. Any increase of stream health scores was considered as improvement while any decrease of
stream health score was considered as decline. These changes are presented in Figure 5.6.

89

Figure 5.5. BMP placement and individual stream health scores for Scenario 2.

90

Figure 5.6. Improvements and declines in stream health scores between the pre-settlement and
current conditions.
As can be seen in Figure 5.6 the implementation of Scenario 2 resulted in the increase in
stream health score for 52 % of the streams. While 11 % of the streams showed no change and
36 % showed decline. This shows that the majority of streams improved with the implementation
of Scenario 2.
5.5

Conclusion
In order to optimize best management practice implementation plan in the context of

stream health, a series of models including SWAT, HIT, Stream Health Predictor (ANFIS), and
Genetic Algorithm were coupled. Daily stream flow data were incorporated in the HIT model to
estimate 171 biologically relevant stream flow indices. Three dimensionality reduction
techniques including Spearman’s Rank Correlation, PCA, and Bayesian variable selection were
used to reduce number of variables that were incorporated in development of stream health
predictor models using ANFIS. The stream health predictor models were used to estimate
individual stream IBI scores and overall stream health scores for the watershed. The genetic
91

algorithm was used to guide BMP selection and placement in the study area to achieve the
highest overall stream score at the lowest price.
The Bayesian variable selection was selected as the best approach to reduce the number
of variables and eliminate redundancy in the stream health predictor model. Two ANFIS models
were selected to evaluate stream health conditions within the region; one for stream orders 1 – 3
and one for stream orders 4 – 7. For stream orders 1 – 3, the final stream health model included
three variables (MA24, MH18, and RA9). RA represents rate of change for an average event
while MA and MH prefix represent magnitude of average and high flow events, respectively. For
the stream orders 4 – 7 two variables (RA5, and MH19) were selected. The main deference
between these two models is that the model for the stream orders 4 – 7 is not as sensitive as the
model for stream orders 1 – 3 to average flow magnitude. These models were used to evaluate
the stream health scores for all stream segments within the region.
Analysis was done to identify the changes in stream health due to the alteration of land
use from pre-settlement to current conditions. It was found that in the current conditions 58% of
streams had a decrease in stream health, 26% showed improvement, and 16% showed no change.
In term of overall stream health score 21% decline was observed. Furthermore, analysis
indicated that 5th order streams were the least affected by this change, while 1st order streams
were the most impacted.
The implementation of different BMP scenarios was preformed to find a near optimum
solution that maximized the stream health score and minimized the cost. Eight unique BMP
scenarios with different BMP compositions were identified that had a maximum stream health
score of 71.036. The most commonly implemented BMP was no tillage. Scenarios 4, 6, and 7
had the highest percentage of individual streams with health scores greater than 65. However,
92

cost analysis of the scenarios, identified Scenario 2 as the lowest costing option, thus matching
the criteria set for this study. Comparison of the stream health scores from Scenario 2 to the
Current condition yielded an increase of stream health scores for 52 % of the streams, no change
for 11% of streams, and a decrease of stream health for 36 % of streams.
In order to improve the predictability and reliability of the stream health predictor
models, future studies should include wide range of water quality and quantity variables along
with both fish and macroinvertebrates biological indicators. Studies like the one perform here
can provide valuable information to decision makers, allowing them to develop a watershed level
BMP implementation scenarios that will improve the stream health conditions at the lowest cost.
5.6

Acknowledgments
This work is supported by the USDA National Institute of Food and Agriculture, Hatch

project MICL02212. We also would like to thank the Michigan Institute for Fisheries Research
for providing stream and catchment data.

93

6.

CONCLUSIONS

This research evaluated the application of a genetic algorithm in combination with a stream
health model to find a near-optimum solution that maximizes stream health while minimizing
BMP implementation cost. In the first study, a review of currently existing macroinvertebrate
and fish stream health indices was used to identify the most appropriate biological index that can
represent stream health conditions in a large and diverse watershed. The Index of Biological
Integrity (IBI) was selected because the study area is comprised of warm water streams (Lyons,
1992) and degradation of fisheries in the region has be identified as a major issue (EPA, 2013).
Then, the relationship between IBI score, stream health condition, and hydroecological variables
was explored to develop a new stream health predictor models. Finally, a genetic algorithm was
used to identify a near-optimal solution of BMP implementation. The following can be
concluded from the results of both studies:


The most commonly used and modified macroinvertebrate and fish indices were the
Benthic Index of Biotic Integrity (B-IBI), Ephemeroptera/ Plechoptera/ Trichoptera
(Index) index, Hilsenhoff Biotic Index (HBI), and Index of Biological Integrity (IBI).



The IBI was the most commonly used and modified fish index.



Macroinvertebrate indices can be used to describe local degradation, while fish indices
can be used to describe regional degradation.



Stream health indices often need to be modified to take into account local stream
conditions and biological communities.



Watershed models built using the Soil and Water Assessment Tool were able to
satisfactorily represented observed streamflow conditions.

94



Bayesian variable selection and ANFIS techniques were able to develop stream health
models using hydroecological variables and IBI scores. These models were applicable for
all stream segments within the study area.



Comparison of the pre-settlement land use (reference conditions) and current conditions
indicated that la nude change in the study area has had significant negative impacts on
stream health.



The Genetic Algorithm identified no tillage, residue management, and forest as the most
appropriate BMPs to be implemented in the study area.



The coupled models and genetic algorithm were identified a near-optimum solution that
maximized stream health condition while minimizing BMP implementation cost.

95

7.

FUTURE RESEARCH RECOMMENDATIONS

This research provides valuable insight into using genetic algorithms with stream health
predictor models to explore BMP implementation. However, additional research should be
perform on applicability of the developed technique in different regions and to improve our
understanding of the relationships between hydroecological variables and biological indices. The
following are few suggestions for future research:


Apply this technique to a larger watershed allows for better understanding of the link
between BMP selection and placement, streamflow, and stream health.



Examine the uncertainty of the collected data and model components. Quantifying
uncertainty will aid water resource managers and stakeholders in the decision making
process.



Explore the impacts of climate change on stream health and BMP selection. To ensure
freshwater resource sustainability, future impacts need to be taken into account.



Evaluation of additional BMPs. For this study, five different BMPs were evaluated.
However, there are many more BMPs that may have a greater impact on stream health.



Explore the use of other stream health indices as the basis for more comprehensive
stream health predictor models. Only the IBI was addressed in this study, where
macroinvertebrate indices or a combination of macroinvertebrate and fish indices should
be used to determine if a more accurate stream health model can be developed.



Explore the addition of social aspects, such as stakeholder and decision maker
preferences, to the decision making and genetic algorithm framework developed in this
study.

96

APPENDIX

97

Table A1. Comparisons between B-IBI and other indices that either used or modified it for
application elsewhere.
Index
B-IBI

Index
Macroinvertebrate Index of
Biotic Integrity

Invertebrate Community
Index

Chesapeake Bay IBI

Metrics included
Total taxa richness
Intolerant snail and mussel species richness
Mayfly richness
Caddisfly richness
Stonefly richness
Relative abundance of Corbicula
% Oligochaetes
% Omnivores
% Filterers
% Grazers
% Predators
Proportion of individuals in two most abundant taxa
Total abundance.
Metrics Added to B-IBI
Metrics Removed from B-IBI
Relitive Abundance Crustacea Total taxa richness
and Mollusca
Intolerant snail and mussel
Chironomidae genera richness species richness
Relative Abundance five most Relative abundance of Corbicula
dominant genera
% Oligochaetes
Macroinvertebrate density
% Omnivores
Relitive Abundance EPT
% Filterers
Orthocladiinae/Chironomidae % Grazers
ratio
% Predators
Tanytarsini/Chironomidae
Proportion of individuals in two
ratio
most abundant taxa
Hilsenhoff’s biotic index
Total abundance.
Number of Dipteran taxa
Intolerant snail and mussel
% Mayfly composition
species richness
% Caddisfly composition
Stonefly richness
% Tribe Tanytarsini Midge
Relative abundance of Corbicula
composition
% Oligochaetes
% other Dipteran and non% Omnivores
insect composition
% Filterers
% tolerant organisms
% Grazers
Number of qualitative EPT
% Predators
taxa
Proportion of individuals in two
most abundant taxa
Total abundance.
Shannon-Weiner diversity
Total taxa richness
Biomass
Intolerant snail and mussel
species richness
98

Modified B-IBI

Modified ICI

% of abundance as pollutionindicative taxa
% of biomass as pollutionindicative taxa
% of abundance as pollutionsensitive taxa
% of biomass as pollutionsensitive taxa
% of abundance as carnivores
and omnivores
% of abundance as deep
deposit feeders
% of biomass deeper than 5
cm
% of taxa deeper than 5 cm
Proportion of Scrapers

Number Dipteran taxa
% Mayfly composition§
% Caddisfly composition‡
% predatory Chironomidae
composition§
% other dipteran and noninsects§
% tolerant organisms§
Number EPT taxa

99

Mayfly richness
Caddisfly richness
Stonefly richness
Relative abundance of Corbicula
% Oligochaetes
% Filterers
% Grazers
% Predators
Proportion of individuals in two
most abundant taxa

% Grazers
Intolerant snail and mussel
species richness
Intolerant snail and mussel
species richness
Stonefly richness
Relative abundance of Corbicula
% Oligochaetes
% Omnivores
% Filterers
% Grazers
% Predators
Proportion of individuals in two
most abundant taxa
Total abundance.

Table A2. Comparisons between HBI and other indices that either used or modified it for
application elsewhere.
Index
Hilsenhoff’s
Biotic Index
Index
Nutrient Biotic
Index
Tolerance
Indicator Values

Metrics included
Organism tolerance to organic pollutants
Metrics Added to HBI
Organism tolerance to nitrogen
Organism tolerance to
phosphorous
Organism tolerance to dissolved
oxygen
Organism tolerance to nitrite plus
nitrate
(nitrate)
Organism tolerance to total
phosphorus
Organism tolerance to water
temperature
Organism tolerance to nutrient
loading
No changes made

Invertebrate
Species Index
Family-level
Biotic Index
Macroinvertebrate Relative Abundance Crustacea
Index of Biotic
and Mollusca
Integrity
Chironomidae genera richness
Relative Abundance (five most
dominant genera)
Macroinvertebrate density
Orthocladiinae/Chironomidae
ratio
Tanytarsini/Chironomidae ratio
EPT richness
EPT % abundance
Benthic
# Ephemeroptera
Community Index # Diptera
Richness
Shannon–Wiener Diversity
% Trichoptera
% Crustacea and Mollusca
# Filterers
# Scrapers
EPT richness
EPT % abundance

100

Metrics Subtracted from HBI
Organism tolerance to organic
pollutants
Organism tolerance to organic
pollutants

Organism tolerance to organic
pollutants

Table A3. Comparisons between EPT and other indices that either used or modified it for
application elsewhere.
Index
EPT
Index
Benthic Community
Index

Non-wadeable
Macroinvertebrate
Assemblage Condition
Index

Macroinvertebrate Index
of Biotic Integrity

Macroinvertebrate
Multimetric Index

EPTC
Guapiacu-Macau
Multimetric Index

Metrics included
EPT richness
EPT % abundance
Metrics Added to EPT
# Ephemeroptera
# Diptera
Richness
Shannon–Wiener Diversity
% Trichoptera
% Crustacea and Mollusca
# Filterers
# Scrapers
Hilsenhoff’s biotic index
Diptera taxa richness
% Coleoptera taxa
% Oligochaete and leech taxa
% Collector-filterer individuals
Predator taxa richness
% Burrower taxa
Tolerant taxa richness
% Facultative individuals
Relative Abundance Crustacea
and Mollusca
Chironomidae genera richness
Relative Abundance (five most
dominant genera)
Macroinvertebrate density
Orthocladiinae/Chironomidae
ratio
Tanytarsini/Chironomidae ratio
Hilsenhoff’s biotic index
#Family
Sensitive taxa
EPT/Chironomidae
%Gathering-collector
%Shredder
Coleoptera richness
% Plecoptera
% Shredders
Family richness
Trichoptera richness (families)
101

Metrics Subtracted from EPT

EPT % abundance

EPT % abundance
EPT richness

Shannon diversity (families)
% Mollusca and Diptera
Hydropsychidae/Trichoptera
Chironomidae/Diptera
Alabama Department of
Environmental
Management Index of
Stream Health

Total # taxa
# Chironomidae taxa
NCBI
Dominant taxon* (%)
Chironomidae (%)
Filterers* (%)
Florida Department of
Total # taxa
Environmental Protection # Chironomidae taxa
Index of Stream Health
Florida Index
Dominant taxon* (%)
Chironomidae (%)
Filterers* (%)
Mississippi Department
Total # taxa
of Environmental Quality NCBI
Index of Stream Health
Dominant taxon* (%)
# Scrapers/# gatherers
# Shredders/total # collected
# EPT/# Chironomidae
Community loss Index
North Carolina Division
Total # taxa
of Water Quality Index of NCBI
Stream Health

EPT % abundance

South Carolina
NCBI
Department of Health and
Environmental Control
Index of Stream Health

EPT % abundance

EIPT

Ephemeroptera richness
Ephemeroptera % abundance

Elmidae richness
Elmidae % abundance

102

EPT % abundance

EPT % abundance

EPT % abundance

Table A4. Comparisons between IBI and other indices that either used or modified it for
application elsewhere.
Index
IBI

Index
B-IBI

Mebane IBI

Metrics included
Number of Species
Presence of Intolerant Species
Species Richness and Composition of Darters
Species Richness and Composition of Suckers
Species Richness and Composition of Sunfish (except Green
Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid Individuals
Number of Individuals in Sample
Proportion of Omnivores (Individuals)
Proportion of Insectivorous Cyprinids
Proportion of Top Carnivores
Proportion with Disease, Tumors, Fin Damage, and Other
Anomalies
Metrics Added to IBI
Metrics Subtracted from IBI
Intolerant snail and mussel
Presence of Intolerant Species
species richness
Species Richness and
Mayfly richness
Composition of Darters
Caddisfly richness
Species Richness and
Stonefly richness
Composition of Suckers
Relative abundance of
Species Richness and
Corbicula
Composition of Sunfish (except
% Oligochaetes
Green Sunfish)
% Filterers
Proportion of Green Sunfish
% Grazers
Proportion of Hybrid
Proportion of individuals in
Individuals
two most abundant taxa
Number of Individuals in
Total abundance.
Sample
Proportion of Insectivorous
Cyprinids
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Native coldwater species,
Number of Species
number
Presence of Intolerant Species
Coldwater individuals,
Species Richness and
percent
Composition of Darters
Alien species, number
Species Richness and
Sensitive native individuals,
Composition of Suckers
percent
Tolerant individuals, percent
103

Common carp Cyprinus
carpio individuals, percent
Sculpin age-classes, number
Sculpin individuals, percent
Salmonid age-classes, minus
mountain whitefish
Prosopium
williamsoni, number

Northern Glaciated Plains
Index of Biotic Integrity

Centrarchidae species
richness plus Micropterus
salmoides
Tolerant species richness
% Lithophilic spawners
% Alien fish
% Native coolwater species

Yangtze River Index of
Biotic Integrity

% Number of species in the
family Cyprinidae
% Number of species in
Bagridae catfishes
% Number of species in the
family Cobitidae
Percent of tolerance
individuals
Number of families in fishery
catches
Individual condition Percent
of non-native fish species
% Characiform individuals

Multi-metric Index for
Atlantic Rain Forest Streams

104

Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Proportion of Omnivores
(Individuals)
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Number of Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Presence of Intolerant Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Cyprinids
Number of Species

% Water column native
individuals
% Tolerant species
% Detritivorous individuals

Cool–cold transition IBI

Number of madtom and
sculpin species
Number of coolwater species
Percentage tolerant species
Percentage generalist feeders

Cool–warm transition IBI

Number of native minnow
species
Percentage tolerants
Percentage omnivores

105

Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Number of Species
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Number of Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)

Fish Based Index for Lakes

Fish Based Index for
Reservoirs

Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Number of planktivore
Number of Species
species
Species Richness and
Total Biomass of strict
Composition of Darters
lithophilic individuals
Species Richness and
% Total biomass of tolerant
Composition of Suckers
individuals
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Number of strict lithophilic
Number of Species
species
Presence of Intolerant Species
% strict lithophilic species
Species Richness and
% Species Piscivores
Composition of Darters
Number of Herbivores
Species Richness and
Total biomass of tolerant
Composition of Suckers
species
Species Richness and
Total biomass of Planktovores Composition of Sunfish (except
% Total biomass of
Green Sunfish)
lithophilic species
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
106

European Fish Index

Czech Multi-metric Index

Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Density of omnivorous
Number of Species
species
Species Richness and
Density of phytophilic species Composition of Darters
Relative abundance of
Species Richness and
lithophilic species
Composition of Suckers
Number of benthic species
Species Richness and
Number of rheophilic species Composition of Sunfish (except
Relative number of tolerant
Green Sunfish)
species
Proportion of Green Sunfish
Number of species migrating Proportion of Hybrid
over long distances
Individuals
Number of potamodromous
Number of Individuals in
species
Sample
Proportion of Omnivores
(Individuals)
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Ecological-quality ratio of the Number of Species
typical species presence
Presence of Intolerant Species
Ecological-quality ratio
Species Richness and
of the overall abundance
Composition of Darters
Ecological-quality ratio of the Species Richness and
relative abundance of
Composition of Suckers
rheophilic species
Species Richness and
Ecological-quality ratio of the Composition of Sunfish (except
relative abundance of
Green Sunfish)
eurytopic species
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies

107

Minnesota fish index of biotic The number of taxa
integrity
designated as darter, sclupin,
and madtoms taxa
The number of insectivore
taxa minus the number of
tolerant taxa
The number of headwater
taxa minus the number of
tolerant taxa
The number of minnow taxa
minus the number of tolerant
taxa
The number of piscivore taxa
The number of wetland taxa
minus the number of tolerant
taxa
The abundance of fish per
100m minus that of tolerant
taxa
The percentage of total
abundance of the two most
dominant taxa
The percentage of total
abundance of the piscivore
taxa
The percentage of total
abundance of the lithophilic
taxa
The percentage of total
abundance of the tolerant taxa
Multi-metric Index for the
% native lotic taxa
Coastal Plain Ecoregion

108

Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Cyprinids
Proportion of Top Carnivores

Number of Species
Presence of Intolerant Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Top Carnivores

Multi-metric Index for the
Northern Appalachians
Ecoregion

% native large river taxa
% native egg hider taxa
% invertivore/piscivore
individuals

Multi-metric Index for the
Southern Appalachians
Ecoregion

% native large river taxa
% native egg hider
individuals
Herbivore richness
% threatened and endangered
individuals

109

Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Number of Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Number of Species
Presence of Intolerant Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Proportion of Insectivorous
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies

Multi-metric Index for the
Southern Plains Ecoregion

% native lotic individuals
Native large river species
richness

Multi-metric Index for the
Temperate Plains Ecoregion

% native large river taxa
% native rheophilic
individuals
% lithophilic spawner taxa

Multi-metric Index for the
Upper Midwest Ecoregion

Threatened and endangered
species richness
% native large river taxa
% lithophilic spawner
individuals

110

Number of Species
Presence of Intolerant Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Number of Species
Presence of Intolerant Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Number of Species
Presence of Intolerant Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers

Multi-metric Index for the
Western Mountain Ecoregion

% native egg hider
individuals
% piscivore individuals
% native large river
individuals

Multi-metric Index for the
Xeric West Ecoregion

% lithophilic spawner taxa
% native water column
individuals
Threatened and endangered
species richness
Herbivore richness

111

Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Number of Species
of Intolerant Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Number of Species
Presence of Intolerant Species
Species Richness and
Composition of Darters
Species Richness and
Composition of Suckers
Species Richness and
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish

Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Coldwater Multi-metric Index # Brook trout individuals per Number of Species
100m2
Presence of Intolerant Species
% Fluvial dependent
Species Richness and
individuals
Composition of Darters
# Warmwater species (stream- Species Richness and
size-corrected)
Composition of Suckers
% Warmwater individuals
Species Richness and
% Brook trout individuals
Composition of Sunfish (except
Green Sunfish)
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Proportion of Insectivorous
Cyprinids
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies
Mixed-water Multi-metric
% White sucker individuals
Number of Species
Index
% Fluvial-specialist
Presence of Intolerant Species
individuals, except blacknose Species Richness and
dace
Composition of Darters
% Non-tolerant general feeder Species Richness and
individuals
Composition of Suckers
% Native warmwater
Species Richness and
individuals
Composition of Sunfish (except
% Intolerant individuals
Green Sunfish)
# Fluvial specialist species
Proportion of Green Sunfish
Proportion of Hybrid
Individuals
112

Number of Individuals in
Sample
Proportion of Omnivores
(Individuals)
Proportion of Insectivorous
Proportion of Top Carnivores
Proportion with Disease,
Tumors, Fin Damage, and
Other Anomalies

113

Figure S1. IBI monitoring stations in the Saginaw Bay Watershed.

114

Figure S2. USGS streamflow gauging stations.

115

Figure S3. Improvements and declines in stream health scores between the pre-settlement and
current conditions.

116

Figure S4. Distributions of individual stream health scores against stream order (scenarios 1).
117

Figure S5. Distributions of individual stream health scores against stream order (scenarios 2).
118

Figure S6. Distributions of individual stream health scores against stream order (scenarios 3).

119

Figure S7. Distributions of individual stream health scores against stream order (scenarios 4).

120

Figure S8. Distributions of individual stream health scores against stream order (scenarios 5).
121

Figure S9. Distributions of individual stream health scores against stream order (scenarios 6).
122

Figure S10. Distributions of individual stream health scores against stream order (scenarios 7).
123

Figure S11. Distributions of individual stream health scores against stream order (scenarios 8).
124

REFERENCES

Abbaspour, K. C. (2007). SWAT-CUP 2012: SWAT Calibration and Uncertainty Programs - A
User Manual. . Eawag, Swiss Federal Institute of Aquatic Science and Technology,
Duebendorf, Switzerland.
Adriaenssens, V., Baets, B. D., Goethals, P. L., and Pauw, N. D. (2004). Fuzzy rule-based
models for decision support in ecosystem management. Science of the Total
Environment, 319(1), 1-12.
Alcázar, J., and Palau, A. (2010). Establishing environmental flow regimes in a Mediterranean
watershed based on a regional classification. Journal of Hydrology, 388(1-2), 41-51.
Alcalá-Fdez, J., Sánchez, L., García, S., del Jesus, M. J., Ventura, S., Garrell, J. M., Otero, J.,
Romero, C., Bacardit, J., Rivas, V. M., Fernández, J. C., and Herrera, F. (2009). KEEL: a
software tool to assess evolutionary algorithms for data mining problems. Soft
Computing, 13(3), 307-318.
Alfredsen, K., Harby, A., Linnansaari, T., and Ugedal, O. (2012). Development of an inflowcontrolled environmental flow regime for a norwegian river. River Research and
Applications, 28(6), 731-739.
Angradi, T. R., and Jicha, T. M. (2010). Mesohabitat-specific macroinvertebrate assemblage
responses to water quality variation in mid-continent (North America) great rivers.
Ecological Indicators, 10(5), 943-954.
Angradi, T. R., Pearson, M. S., Jicha, T. M., Taylor, D. L., Bolgrien, D. W., Moffett, M. F.,
Blocksom, K. A., Hill, B. H. (2009). Using stressor gradients to determine reference
expectations for great river fish assemblages. Ecological Indicators, 9(4), 748-764.
Arabi, M., Govindaraju, R. S., and Hantush, M. M. (2006). Cost‐effective allocation of
watershed management practices using a genetic algorithm. Water Resources
Research, 42(10).
Arabi, M., Frankenberger, J.R., Engel, B.A., and Arnold, J.G. (2007). Representation of
agricultural conservation practices with SWAT. Hydrologic Processes, 22(16), 30423055.
Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling. 1999. Rapid Bioassessment
Protocols for Use in Streams and Wadeable Rivers: Periphyton, Benthic

125

Macroinvertebrates and Fish, Second Edition. EPA 841-B-99-002. U.S. Environmental
Protection Agency; Office of Water; Washington, D.C.
Beck, M. W., Wilson, B. N., Vondracek, B., and Hatch, L. K. (2014). Application of neural
networks to quantify the utility of indices of biotic integrity for biological monitoring.
Ecological Indicators, 45, 195-208.
Belmar, O., Velasco, J., and Martinez-Capel, F. (2001). Hydrological classification of natural
flow regimes to support environmental flow assessments in intensively regulated
Mediterranean rivers, Segura River Basin (Spain). Environmental Management, 47(5),
992-1004.
Besley, C. H., and Chessman, B. C. (2008). Rapid biological assessment charts the recovery of
stream macroinvertebrate assemblages after sewage discharges cease. Ecological
Indicators, 8(5), 625-638.
Blocksom, K., and Johnson, B. (2009). Development of a regional macroinvertebrate index for
large river bioassessment. Ecological Indicators, 9.2, 313-328.
Boyle, T. P., and Fraleigh, H. D. (2003). Natural and anthropogenic factors affecting the
structure of the benthic macroinvertebrate community in an effluent-dominated reach of
the Santa Cruz River, AZ. Ecological Indicators, 3(2), 93-117.
Brazner, J., Danz, N., Niemi, G., Regal, R., Trebitz, A., Howe, R., Hanowski, J., Johnson, L.,
Ciborowski, J., Johnston, C., Reavie, E., Brady, V., and Sgro, G. (2007). Evaluation of
geographic, geomorphic and human influences on Great Lakes wetland indicators: A
multi-assemblage approach. Ecological Indicators, 7(3), 610-635.
Bunn, S. E., Abal, E. G., Smith, M. J., Choy, S. C., Fellows, C. S., Harch, B. D., Kennard, M. J.,
Sheldon, F. (2010, January). Integration of science and monitoring of river ecosystem
health to guide investments in catchment protection and rehabilitation. Freshwater
Biology, 55(s1), 223-240.
Butcher, J. T., Stewart, P. M., and Simon, T. P. (2003a). A Benthic Community Index for
streams in the Northern Lakes and Forests Ecoregion. Ecological Indicators, 3(3), 181193.
Butcher, J. T., Stewart, P. M., and Simon, T. P. (2003b). Effects of two classification strategies
on a Benthic Community Index for streams in the Northern Lakes and Forests Ecoregion.
Ecological Indicators, 3(3), 195-202.
Carlin, B. P., and Chib, S. (1995). Bayesian model choice via Markov chain Monte Carlo
methods. Journal of the Royal Statistical Society. Series B (Methodological), 473-484.

126

Carlson, T. N., Haase, A. T., Cudaback, C. N., and Anderson, M. C. (2012, October). Assessing
Stream Health in the Chesapeake Basin Using the SHARP Model. Open Journal of
Modern Hydrology, 2(4), 71-78.
Chen, Q., and Mynett, A. E. (2003). Integration of data mining techniques and heuristic
knowledge in fuzzy logic modelling of eutrophication in Taihu Lake.Ecological
Modelling, 162(1), 55-67.
Chen, H., and Zhao, Y. W. (2011). Evaluating the environmental flows of China's Wolonghu
wetland and land use changes using a hydrological model, a water balance model, and
remote sensing. Ecological Modelling, 222(2), 253-260.
Chessman, B. C. (1995). Rapid assessment of rivers using macroinvertebrates: a procedure based
on habitat‐specific sampling, family level identification and a biotic index. Australian
Journal of Ecology, 20(1), 122-129.
Cibin, R., Sudheer, K. P., and Chaubey, I. (2010). Sensitivity and identifiability of stream flow
generation parameters of the SWAT model. Hydrological Processes, 24(9), 1133-1148.
Coe, M. T., Latrubesse, E. M., Ferreira, M. E., and Amsler, M. L. (2011). The effects of
deforestation and climate variability on the streamflow of the Araguaia River, Brazil.
Biogeochemisty, 105, 119-131.
Compin, A., and Céréghino, R. (2003). Sensitivity of aquatic insect species richness to
disturbance in the Adour–Garonne stream system (France). Ecological Indicators, 3(2),
135-142.
Couceiro, S., Hamada, N., Forsberg, B., Pimentel, T., and Luz, S. (2012). A macroinvertebrate
multimetric index to evaluate the biological condition of streams in the Central Amazon
region of Brazil. Ecological Indicators, 18, 118-125.
Cummins, K. W., and Klug, M. J. (1979). Feeding ecology of stream invertebrates.Annual
review of ecology and systematics, 10, 147-172.
Dos Santos, D. A., Molineri, C., Reynaga, M. C., and Basualdo, C. (2011). Which index is the
best to assess stream health? Ecological Indicators, 11(2), 582-589.
Einheuser, M. (2011). Exploring relationships between in-stream conditions and ecological
health while assessing landuse and climate scenarios (A Thesis). East Lansing: Michigan
State University.
Einheuser, M. D., Nejadhashemi, A. P., Sowa, S. P., Wang, L., Hamaamin, Y. A., and Woznicki,
S. A. (2012). Modeling the effects of conservation practices on stream health. Science of
the Total Environment, 435, 380-391.
127

Einheuser, M. D., Nejadhashemi, A. P., and Woznicki, S. A. (2013a). Simulating stream health
sensitivity to landscape changes due to bioenergy crops expansion. Biomass and
Bioenergy, 58, 198-209.
Einheuser, M. D., Nejadhashemi, A. P., Wang, L., Sowa, S. P., and Woznicki, S. A. (2013b).
Linking biological integrity and watershed models to assess the impacts of historical land
use and climate changes on stream health. Environmental management, 51(6), 11471163.
EPA. (2006). Wadeable Streams Assessment: A Collaborative Survey of the Nation's Streams.
Washington, D.C.: Office of Research and Development: Office of Water.EPA. (2011,
March 6). Aquatic Resource Monitoring. Retrieved 1 25, 2014, from Aquatic Indicators:
www.epa.gov/nhrlsup1/arm/indicators/indicators.htm
EPA. (2011, April 6). Aquatic Resource Monitoring. Retrieved 1 25, 2014, from Aquatic
Indicators: www.epa.gov/nhrlsup1/arm/indicators/indicators.htm
EPA. (2012a, March 6). Chapter 7 (part A): Benthic Macroinvertebrate Protocols. Retrieved
from Water: Bioassessment:
http://water.epa.gov/scitech/monitoring/rsl/bioassessment/ch07main.cfm
EPA. (2012b, March 6). Water: Monitoring and Assessment. Retrieved from Assessing
Biological Integrity of Surface Water:
http://water.epa.gov/type/wetlands/assessment/fact1.cfm
EPA. (2014, November 4). Water: Healthy Watersheds. Retrieved from Hydrology Assessments:
http://water.epa.gov/polwaste/nps/watershed/hydrologic.cfm
Esselman, P. C., Infante, D. M., Wang, L., Cooper, A. R., Wieferich, D., Tsang, Y.-P.,
Thornbrugh, D. J., Taylor, W. W. (2013). Regional fish community indicators of
landscape disturbance to catchments of the conterminous United States. Ecological
Indicators, 26, 163-173.
Fahlbusch, H. (2009, February). Eary Dams. Engineering History and Heritage, 162(1), 13-18.
Flinders, C., Horwitz, R., and Belton, T. (2008). Relationship of fish and macroinvertebrate
communities in the mid-Atlantic uplands: Implications for integrated assessments.
Ecological Indicators, 8(5), 588-598.
Gassman, P.W., Reyes, M.R., Green, C.H., and Arnold, J.G. (2007). The soil and water
assessment tool: historical development, applications, and future research directions.
Transactions of ASABE; 50: 1211–1250.

128

Giri, S., Nejadhashemi, A. P., and Woznicki, S. A. (2012). Evaluation of targeting methods for
implementation of best management practices in the Saginaw River Watershed. Journal
of Environmental Management, 103, 24-40.
Giri, S., Nejadhashemi, A. P., Woznicki, S., and Zhang, Z. (2014). Analysis of best management
practice effectiveness and spatiotemporal variability based on different targeting
strategies. Hydrological Processes, 28(3), 431-445.
Goetz, S., and Fiske, G. J. (2013). On the Relationship Between Stream Biotic Diversity and
Exurbanization in the Northeastern USA. In Geospatial Tools for Urban Water Resources
(pp. 61-78). Springer Science+Business Media.
Griffith, M. B., Hill, B. H., McCormick, F. H., Kaufmann, P. R., Herlihy, A. T., and Selle, A. R.
(2005). Comparative application of indices of biotic integrity based on periphyton,
macroinvertebrates, and fish to southern Rocky Mountain streams. Ecological Indicators,
5(2), 117-136.
Gupta, A. (1995). Magnitude, frequency, and special factors affecting channel form and
processes in the seasonal tropics. Natural and anthropogenic influences in fluvial
geomorphology, 125-136.
Haase, R., and Nolte, U. (2008). The invertebrate species index (ISI) for streams in southeast
Queensland, Australia. Ecological Indicators, 8(5), 599-613.
Hamaamin, Y. A. (2014). Applications of soft computing and statistical methods in water
resources management (Doctoral dissertation, MICHIGAN STATE UNIVERSITY).
Hamaamin, Y. A., Nejadhashemi, A. P., and Einheuser, M. D. (2013). Application of fuzzy logic
techniques in estimating the regional index flow for michigan. Transactions of the
ASABE, 56(1), 103-115.
Harrison, T. D., and Kelly, F. L. (2013). Development of an estuarine multi-metric fish index
and its application to Irish transitional waters. Ecological Indicators, 34, 494-506.
Henriksen, J. A., Heasley, J., Kennen, J. G., and Nieswand, S. (2006). Users’ Manual for the
Hydroecological Integrity Assessment Process Software (including the New Jersey
Assessment Tools). Reston, VA: USGS
Hilsenhoff, W. L. (1987). An Improved Biotic Index of Organic Stream Pollution. The Great
Lakes Entomologist, 20(1), 31-39.
Houston, L., Barbour, M., Lenat, D., and Penrose, D. (2002). A multi-agency comparison of
aquatic macroinvertebrate-based stream bioassessment methodologies. Ecological
Indicators, 1(4), 279-292.
129

Hu, T.-J., Wang, H.-W., and Lee, H.-Y. (2007). Assessment of environmental conditions of NanShih stream in Taiwan. Ecological Indicators, 7(2), 430-441.
Huang, M., and Asner, G. P. (2010, September 30). Long‐term carbon loss and recovery
following selective logging in Amazon forests. Global Biogeochemical Cycles, 24(3).
Iliopoulou-Georgudaki, J., Kantzaris, V., Katharios, P., Kaspiris, P., Georgiadis, T., and
Montesantou, B. (2003). An application of different bioindicators for assessing water
quality: a case study in the rivers Alfeios and Pineios (Peloponnisos, Greece). Ecological
Indicators, 2(4), 345-360.
International Rivers. (2014, June 6). Environmental Impacts of Dams. Retrieved from
http://www.internationalrivers.org/environmental-impacts-of-dams
Jang J.S.R. (1993). ANFIS: adaptive-network-based fuzzy inference system. IEEE Trans Syst
Man Cybern, 23(3), 665–685.
Jeong, J., Kannan, N., and Arnold, J. G. (2012). Effects of Urbanization and Climate Change on
Stream Health in North-Central Texas. Journal of Environmental Quality, 43(1), 100-109.
Johnson, R. C., Carreiro, M. M., Jin, H.-S., and Jack, J. D. (2013). Within-year temporal
variation and life-cycle seasonality affect stream macroinvertebrate community structure
and biotic metrics. Ecological Indicators, 13(1), 206-214.
Jordan, S. J., Lewis, M. A., Harwell, L. M., and Goodman, L. R. (2010). Summer fish
communities in northern Gulf of Mexico estuaries: Indices of ecological condition.
Ecological Indicators, 10(2), 504-515.
Justus, B., Petersen, J. C., Femmer, S. R., Davis, J. V., and Wallace, J. (2010). A comparison of
algal, macroinvertebrate, and fish assemblage indices for assessing low-level nutrient
enrichment in wadeable Ozark streams. Ecological Indicators, 10(3), 627-638.
Kanno, Y., Vokoun, J., and Beauchene, M. (2010). Development of dual fish multi-metric
indices of biological condition for streams with characteristic thermal gradients and low
species richness. Ecological Indicators, 10(3), 565-571.
Karr, J. R. (1981). Assessment of Biotic Integrity Using Fish Communities. Fisheries, 6(6), 2127.
Karr, J. R., and Dudley, D. R. (1981). Ecological perspective on water quality goals.
Environmental Management, 5(1), 55-68.
Kerans, B. L., and Karr, J. R. (1994). A Benthic Index of Biological Integrity (B-IBI) for Rivers
of the Tennessee Valley. Ecological Applications, 4(4), 768-785.

130

King, A. J., Tonkin, Z., and Mahoney, J. (2009). Environmental flow enhances native fish
spawning and recruitment in the Murray River, Australia. River Research and
Applications, 25(10), 1205-1218.
Krause, J. R., Bertrand, K. N., Kafle, A., and Troelstrup Jr., N. H. (2013). A fish index of biotic
integrity for South Dakota’s Northern Glaciated Plains Ecoregion. Ecological Indicators,
34, 313-322.
Lam, Q. D., Schmalz, B., and Fohrer, N. (2010). Modeling point and diffuse source pollution of
nitrate in a rural lowland catchemnt using the SWAT model. Agricultural Water
Management, 97(2), 317-325.
Launois, L., Veslot, J., Irz, P., and Argillier, C. (2011). Development of a fish-based index (FBI)
of biotic integrity for French lakes using the hindcasting approach. Ecological Indicators,
11(6), 1572-1583.
Love, B. J., and Nejadhashemi, A. P. (2011). Water quality impact assessment of large-scale
biofuel crops expansion in agricultural regions of Michigan.Biomass and
Bioenergy, 35(5), 2200-2216.
Luo, P., Takara, K., He, B., Cao, W., Yamashiki, Y., and Nover, D. (2011). Calibration and
uncertainty analysis of SWAT model in a Japanese river catchment. Journal of Japan
Society of Civil Engineers, Ser. B1 (Hydraulic Engineering), 67.
Lyons, J. (1992).Using the index of biotic integrity (IBI) to measure environmental quality in
warmwater streams of Wisconsin. U.S. Department of Agriculture, Forest Service, North
Central Experiment Station, St. Paul, Minnesota, General Technical Report NC-149.
Lyons, J. (2012). Development and validation of two fish-based indices of biotic integrity for
assessing perennial coolwater streams in Wisconsin, USA. Ecological Indicators, 23,
402-412.
Mack, J. J. (2007). Developing a wetland IBI with statewide application after multiple testing
iterations. Ecological Indicators, 7(4), 864-881.
Magbanua, F. S. (2012). Agricultural intensification and stream health: combined impacts of
pesticide and sediment (Thesis, Doctor of Philosophy). University of Otago.
Marzin, A., Archaimbault, V., Belliard, J., Chauvin, C., Delmas, F., and Pont, D. (2012).
Ecological assessment of running waters: Do macrophytes, macroinvertebrates, diatoms
and fish show similar responses to human pressures? Ecological Indicators, 23, 56-65.
MathWorks. (2014). Fuzzy Logic Toolbox. Retrieved from Design and simulate fuzzy logic
systems: http://www.mathworks.com/products/fuzzy-logic/
131

McCartney, M. P., Shiferaw, A., and Seleshi, Y. (2009). Estimating environmental flow
requirements downstream of the Chara Chara weir on the Blue Nile River. Hydrological
Processes, 23(26), 3751-3758.
Meador, M. R., Carlisle, D. M., and Coles, J. F. (2008). Use of tolerance values to diagnose
water-quality stressors to aquatic biota in New England streams. Ecological Indicators,
8(5), 718-728.
Mebane, C. A., Maret, T. R., and Hughes, R. M. (2003). An Index of Biological Integrity (IBI)
for Pacific Northwest Rivers. Transactions of the American Fisheries Society, 132(2),
239-261.
Merritt, D. M., Scott, M. L., Poff, N. L., Auble, G. T., and Lytle, D. A. (2010, January). Theory,
methods and tools for determining environmental flows for riparian vegetation: riparian
vegetation-flow response guilds. Freshwater Biology, 55(1), 206-225.
Metternicht, G. (2001). Assessing temporal and spatial changes of salinity using fuzzy logic,
remote sensing and GIS. Foundations of an expert system.Ecological Modelling, 144(2),
163-179.
Meyer, J. L., Sale, M. J., Mulholland, P. J., and Poff, N. L. (1999). Impacts of climate change on
aquatic ecosystem functioning and health. Journal of the American Water Resources
Association, 1373-1386.
MNFI. (2014). Vegitation circa1800 Maps. Retrieved from
http://mnfi.anr.msu.edu/data/veg1800.cfm
Monaghan, K. A., and Soares, A. M. (2010). The bioassessment of fish and macroinvertebrates
in a Mediterranean–Atlantic climate: Habitat assessment and concordance between
contrasting ecological samples. Ecological Indicators, 10(2), 184-191.
Musil, J., Horky, P., Slavík, O., Zboril, A., and Horká, P. (2012). The response of the young of
the year fish to river obstacles: Functional and numerical linkages between dams, weirs,
fish habitat guilds and biotic integrity across large spatial scale. Ecological Indicators, 23,
634-640.
Muxika, I., Borja, A., and Bonne, W. (2005). The suitability of the marine biotic index (AMBI)
to new impact sources along European coasts. Ecological Indicators, 5(1), 19-31.
NASS. (2012). CropScape-Cropland Data Layer. Retrieved from United States Department of
Agriculture: Nattional Agricultural Statistics Service:
http://nassgeodata.gmu.edu/CropScape/

132

Navarro-Llácer, C., Baeza, D., and Heras, J. (2010). Assessment of regulated rivers with indices
based on macroinvertebrates, fish and riparian forest in the southeast of Spain. Ecological
Indicators, 10(5), 935-942.
NED. (2014, April 15). National Elevation Dataset. Retrieved from
http://ned.usgs.gov/index.html
Neitsch, S.L., Arnold, J.G., Kiniry, J.R., and Williams, J.R. (2011) Soil and water assessment
tool theoretical documentation, version 2009. Temple, Tex.: USDA-ARS Grassland, Soil
and Water Research Laboratory; http://swat.tamu.edu/media/99192/swat2009-theory.pdf.
Nel, J. L., Turak, E., Linke, S., and Brown, C. (2011). Integration of environmental flow
assessment and freshwater conservation planning: a new era in catchment management.
Marine and Freshwater Research, 62(3), 290-299.
Neumann, M., Baumeister, J., Liess, M., and Schulz, R. (2003a). An expert system to estimate
the pesticide contamination of small streams using benthic macroinvertebrates as
bioindicators II. The knowledge base of LIMPACT. Ecological Indicators, 2(4), 391-401.
Neumann, M., Liess, M., and Schulz, R. (2003b). An expert system to estimate the pesticide
contamination of small streams using benthic macroinvertebrates as bioindicators Part 1.
The database of LIMPACT. Ecological Indicators, 2(4), 379-389.
NLCD. (2014). Multi-Resolution Land Characteristics Consortium (MRLC). Retrieved from
National Land Cover Database 2006 (NLCD2006): http://www.mrlc.gov/nlcd06_leg.php
NRCS. (2014a). Wed Soil Survey. Retrieved from
http://websoilsurvey.sc.egov.usda.gov/App/WebSoilSurvey.aspx
NRCS. (2014b). NRCS. Retrieved from Typical Statewide Average Practice Costs 2014:
http://efotg.sc.egov.usda.gov/references/public/MI/Typical_Statewide_Average_Practice
Cost2014.pdf
Olden, J. D., and Naiman, R. J. (2010, January). Incorporating thermal regimes into
environmental flows assessments: modifying dam operations to restore freshwater
ecosystem integrity. Freshwater Biology, 55(1), 86-107.
Olden, J. D., and Poff, N. L. (2003). Redundancy and the choice of hydrologic indices for
characterizing streamflow regimes. River Research and Applications, 19(2), 101-121.
Oliveira, R. B., Baptista, D. F., Mugnai, R., Castro, C. M., and Hughes, R. M. (2011). Towards
rapid bioassessment of wadeable streams in Brazil: Development of the Guapiacu-Macau
Multimetric Index (GMMI) based on benthic macroinvertebrates. Ecological Indicators,
11(6), 1584-1593.
133

Pander, J., and Geist, J. (2013). Ecological indicators for stream restoration success. Ecological
Indicators, 30, 106-118.
Pelletier, M. C., Gold, A. J., Gonzalez, L., and Oviatt, C. (2012). Application of multiple index
development approaches to benthic invertebrate data from the Virginian Biogeographic
Province, USA. Ecological Indicators, 23, 176-188.
Pearson, K. (1901). Principal components analysis. The London, Edinburgh, and Dublin
Philosophical Magazine and Journal of Science, 6(2), 559.
Piniewski, M., Acreman, M. C., Stratford, C. J., Okruszko, T., Gielczewski, M., Teodorowicz,
M., Rycharski, M., and Oświecimska-Piasko, Z. (2011). Estimation of environmental
flows in semi-natural lowland rivers - the Narew basin case study. Pol. J. Environ. Stud.,
20(5), 1281-1293.
Poff, N.L., Allan, J.D., Bain, M.B., Karr, J.R., Prestegaard, K.L., Richter, B.D., Sparks, R.E., and
Stromberg, J.C. (1997). The natural flow regime. Bioscience 47: 769–784
Poff, N. L., Richter, B. D., Arthington, A. H., Bunn, S. E., Naiman, R. J., Kendy, E., Acreman,
M., Apse, C., Bledsoe, B. P., Freeman, M. C., Henriksen, J., Jacobson, R. B., Kennen, J.
G., Merritt, D. M., O’Keeffe, J. H., Olden, J. D., Rogers, K., Tharme, R. E. and Warner,
A. (2010). The ecological limits of hydrologic alteration (ELOHA): a new framework for
developing regional environmental flow standards. Freshwater Biology, 55(1), 147-170.
Poff, N. L., and Zimmerman, J. K. (2010). Ecological responses to altered flow regimes: a
literature review to inform the science and management of environmental flows.
Freshwater Biology, 55(1), 194-205.
Power, M. E., Parker, G., Dietrich, W. E., and Sun, A. (1995). How does floodplain width affect
floodplain river ecology? A preliminary exploration using
simulations. Geomorphology, 13(1), 301-317.
Pratt, J. W. (1959). Remarks on zeros and ties in the Wilcoxon signed rank procedures. Journal
of the American Statistical Association, 54(287), 655-667.
Puente, A., Juanes, J., Garcia, A., Alvarez, C., Revilla, J., and Carranza, I. (2008). Ecological
assessment of soft bottom benthic communities in northern Spanish estuaries. Ecological
Indicators, 8(4), 373-388.
Putz, F. E., Zuidema, P. A., Synnott, T., Pena-Claros, M., Pinard, M. A., Sheil, D., . . . Zagt, R.
(2012, May 16). Sustaining conservation values in selectively logged tropical forests: the
attained and the attainable. Convervation Letters, 5(4), 296-303.
Richter, B.D., Baumgartner, J.V., Powell, J., and Braun, D.P. (1996). A method for assessing
hydrologic alteration within ecosystems. Conservation Biology 10: 1163–1174.
134

Ridoutt, B. G., and Pfister, S. (2010). A revised approach to water footprinting to make
transparent the impacts of consumption and production on global freshwater scaricity.
Global Environmental Change, 20(1), 113-120.
Rossaro, B., Marziali, L., Cardoso, A. C., Solimini, A., Free, G., and Giacchini, R. (2007). A
biotic index using benthic macroinvertebrates for Italian lakes. Ecological Indicators,
7(2), 412-429.
Roy, A. H., Rosemond, A. D., Paul, M. J., Leigh, D. S., and Wallace, J. B. (2003). Stream
macroinvertebrate response to catchment urbanisation (Georgia, U.S.A.). Freshwater
Biology, 48(2), 329-346.
Scanlon, B. R., Faunt, C. C., Longuevergne, L., Reedy, R. C., Alley, W. M., McGuire, V. L., and
McMahon, P. B. (2012). Groundwater depletion and sustainability of irrigation in the US
High Plains and Centray Valley. PNAS, 109(24), 9320-9325.
Schölkopf, B., Smola, A., and Müller, K. R. (1998). Nonlinear component analysis as a kernel
eigenvalue problem. Neural computation, 10(5), 1299-1319.
Seelbach, P.W., Wiley M.J., Kotanchik J.C. and Baker M.E., (1997). State of Michigan
Department of Natural Resources. A Landscape-based Ecological Classification System
for River Valley Segments in Lower Michigan (MI VSEC Version 1.0): Fisheries
Research Report 2036.
SEMCOG. (2008). SEMCOG...Shaping the future of Southeast Michigan. Retrieved August 2,
2013, from Low Impact Development:
http://library.semcog.org/InmagicGenie/DocumentFolder/LIDManualWeb.pdf
Setegn, S. G., Srinivasan, R., Melesse, A. M., and Dargahi, B. (2010). SWAT model application
and prediction uncertainty analysis in the Lake Tana Basin, Ethiopia. Hydrological
Processes, 24(3), 357-367
Sharma, R. C., and Rawat, J. S. (2009). Monitoring of aquatic macroinvertebrates as bioindicator
for assessing the health of wetlands: A case study in the Central Himalayas, India.
Ecological Indicators, 9(1), 118-128.
Singh, K. P., Basant, A., Malik, A., and Jain, G. (2009). Artificial neural network modeling of
the river quality - A case study. Ecological Modeling, 220(6), 888-895.
Smith, A. J., Bode, R. W., and Kleppel, G. S. (2007). A nutrient biotic index (NBI) for use with
benthic macroinvertebrate communities. Ecological Indicators, 7(2), 371-386.
Smith, K. A., and Sklarew, D. (2012). A stream suitability index for brook trout (Savelinus
fontinalis) in the Mid-Atlantic United States of America. Ecological Indicators, 23, 242249.
135

Sophocleous, M. (2012). Conserving and extending the useful life of the largest aquifer in North
America: the future of the high plains/Ogallala aquifer. Groundwater, 50(6), 831-839
Stoddard, J.L., Larsen, D.P., Hawkins, C.P., Johnson, R.K., and Norris, R.H., (2006). Setting
expectations for the ecological condition of streams: the concept of reference condition.
Ecological applications, 16(4), 1267-1276.
Terra, B. d., Hughes, R. M., Francelino, M. R., and Araujo, F. G. (2013). Assessment of biotic
condition of Atlantic Rain Forest streams: A fish-based multimetric approach. Ecological
Indicators, 34, 136-148.
Terzi, Ö. (2011). Monthly river flow forecasting by data mining process.Knowledge-Oriented
Applications in Data Mining. (pp. 127-134)
Texas A&M University System. (2013). SWAT: Soil and Water Assessment Tool. Retrieved
August 15, 2013, from Fact Sheet: swat.tamu.edu
Tuppad, P. and Srinivasan, R., (2008). Bosque River Environmental Infrastructure Improvement
Plan: Phase II BMP Modeling Report. Publ. No. TR-313. Texas A&M Univ., Texas
AgriLife Research, College Station, Tex
Tuppad, P., Mankin, K.R., and Mcvay, K.A. (2010). Strategic targeting of cropland management
using watershed modeling. Agricultural Engineering International: CIGR Journal 12, 3-4
USGS. (2011). A Health Assessment of the Nation’s Streams. Retrieved December 7, 2014,
http://www.usgs.gov/blogs/features/usgs_top_story/a-health-assessment-of-the-nationsstreams/
USGS. (2012, May 1). Nevada Water Science Center. Retrieved August 14, 2013, from
Colorado River: nevada.usgs.gov/water/lakes_rivers/colorado_river.htm
USGS. (2013c, August 14). The USGS Water Science School. Retrieved August 15, 2013, from
The Water Cycle: Freshwater Storage:
ga.water.usgs.gov/edu/watercyclefreshstorage.html
USGS. (2013a, August 14). The USGS Water Science School. Retrieved August 15, 2013, from
Irrigation water use: ga.water.usgs.gov/edu/wuir.html
USGS. (2013b, January 10). The USGS Water Science School. Retrieved August 14, 2013, from
How urbanization affects the hydrologic system: ga.water.usgs.gov/edu/urbaneffects.htm
Vannote, R. L., Minshall, G. W., Cummins, K. W., Sedell, J. R., and Cushing, C. E. (1980). The
river continuum concept. Canadian journal of fisheries and aquatic sciences, 37(1), 130137.

136

Waite, I., Brown, L., Kennen, J., May, J., Cuffney, T., Orlando, J., and Jones, K. (2010)
Comparison of watershed disturbance predictive models for stream benthic
macroinvertebrates for three distinct ecoregions in western US. Ecological Indicators;
10(6): 1125-1136.
Walters, D., Roy, A., and Leigh, D. (2009). Environmental indicators of macroinvertebrate and
fish assemblage integrity in urbanizing watersheds. Ecological Indicators, 9(6), 12221233.
Wan, H., Chizinski, C. J., Dolph, C. L., Vondracek, B., and Wilson, B. N. (2010). The impact of
rare taxa on a fish index of biotic integrity. Ecological Indicators, 10(4), 781-788.
Weisberg, S. B., Ranasinghe, J. A., Dauer, D. M., Schaffner, L. C., Diaz, R. J., and Frithsen, J.
B. (1997). An estuarine benthic index of biotic integrity (B-IBI) for Chesapeake Bay.
Estuaries, 20(1), 149-158.
Woznicki, S.A. and Nejadhashemi, A.P (2012). Sensitivity Analysis of Best Management
Practices Under Climate Change Scenarios. Journal of the American Water Resources
Association 48(1), 90-112.
Worldometers, (2014). Water consumed this year (billions of liters). Retrieved December 8,
2014 from Worldometers: http://www.worldometers.info/water/
Wu, Y., and Liu, S. (2012, May). Automating calibration, sensitivity and uncertainty analysis of
complex models using the R package Flexible Modeling Environment (FME): SWAT as
an example. Environmental Modelling and Software, 31, 99-109
Young, R. G., and Collier, K. J. (2009). Contrasting responses to catchment modification among
a range of functional and structural indicators of river ecosystem health. Freshwater
Biology, 54(10), 2155-2170.
Zhu, D., and Chang, J. (2008). Annual variations of biotic integrity in the upper Yangtze River
using an adapted index of biotic integrity (IBI). Ecological Indicators, 8(5), 564-572.

137