&EPA
United States Office of Water EPA-822-R-08-023
Environmental Protection Office of Science and Technology December 2008
Agency Washington, DC 20460 www.epa.gov
METHODS FOR EVALUATING WETLAND CONDITION
#19 Nutrient Loading
-------
vvEPA
United States Office of Water EPA-822-R-08-023
Environmental Protection Office of Science and Technology December 2008
Agency Washington, DC 20460 www.epa.gov
METHODS FOR EVALUATING WETLAND CONDITION
#19 Nutrient Loading
Major Contributors
Iowa State University
U. Sunday Tim and William Crumpton
Prepared jointly by:
The U.S. Environmental Protection Agency
Health and Ecological Criteria Division (Office of Science and Technology)
and
Wetlands Division (Office of Wetlands, Oceans, and Watersheds)
-------
NOTICE
The material in this document has been subjected to U.S. Environmental Protection
Agency (EPA) technical review and has been approved for publication as an EPA document.
The information contained herein is offered to the reader as a review of the "state of the
science" concerning wetland bioassessment and nutrient enrichment and is not intended to
be prescriptive guidance or firm advice. Mention of trade names, products or services does
not convey, and should not be interpreted as conveying official EPA approval, endorsement,
or recommendation.
APPROPRIATE CITATION
U. S. EPA. 2008. Methods for Evaluating Wetland Condition: Nutrient Loading.
Office of Water, U.S. Environmental Protection Agency, Washington, DC.
EPA-822-R-08-023.
ACKNOWLEDGEMENTS
EPA acknowledges the contributions of the following people in the writing of this module:
U. Sunday Tim and William Crumpton both of Iowa Sate University.
This entire document can be downloaded from the following U.S. EPA websites:
http://www.epa.gov/waterscience/criteria/wetlands/
http://www.epa.gov/owow/wetlands/bawwg/publicat.html
111
-------
CONTENTS
FOREWORD v
LIST OF "METHODS FOR EVALUATING
WETLAND CONDITION" MODULES vi
NUTRIENT LOADING 1
SUMMARY 1
PURPOSE 1
INTRODUCTION 1
LOADING FUNCTION MODELS 2
PROCESS-ORIENTED MODELS 7
LIMITATIONS AND MODEL VALIDATION 25
CHOOSING A SUITABLE MODEL 31
REFERENCES 34
LIST OF TABLES
TABLE 1: INPUT PARAMETERS REQUIRED BY GWLF MODEL 4
TABLE 2: INPUT PARAMETERS REQUIRED BY AGNPS MODEL 8
TABLE 3: INPUT PARAMETERS REQUIRED BY HSPF 15
TABLE 4: SWAT/SWAT 2OO INPUT PARAMETERS 19
LIST OF FIGURES
FIGURE 1: MODEL QUALITY ASSURANCE COMPONENTS 26
IV
-------
FOREWORD
In 1999, the U.S. Environmental Protection Agency (EPA) began work on this series of reports
entitled Methods for Evaluating Wetland Condition. The purpose of these reports is to help
States and Tribes develop methods to evaluate (1) the overall ecological condition of wetlands
using biological assessments and (2) nutrient enrichment of wetlands, which is one of the pri-
mary stressors damaging wetlands in many parts of the country. This information is intended
to serve as a starting point for States and Tribes to eventually establish biological and nutrient
water quality criteria specifically refined for wetland waterbodies.
This purpose was to be accomplished by providing a series of "state of the science" modules
concerning wetland bioassessment as well as the nutrient enrichment of wetlands. The individual
module format was used instead of one large publication to facilitate the addition of other
reports as wetland science progresses and wetlands are further incorporated into water quality
programs. Also, this modular approach allows EPA to revise reports without having to reprint
them all. A list of the inaugural set of 20 modules can be found at the end of this section.
This last set of reports is the product of a collaborative effort between EPAs Health and
Ecological Criteria Division of the Office of Science and Technology (OST) and the Wetlands
Division of the Office of Wetlands, Oceans and Watersheds (OWOW). The reports were
initiated with the support and oversight of Thomas J. Danielson then of OWOW, Amanda K.
Parker and Susan K. Jackson (OST), and seen to completion by Ifeyinwa F. Davis (OST). EPA
relied on the input and expertise of the contributing authors to publish the remaining modules.
More information about biological and nutrient criteria is available at the following
EPA website:
http://www. epa.gov/ost/standards
More information about wetland biological assessments is available at the following
EPA website:
http://www.epa.gov/owow/wetlands/bawwg
-------
LIST OF "METHODS FOR EVALUATING WETLAND
CONDITION" MODULES
MODULE # MODULE TITLE
1 INTRODUCTION TO WETLAND BIOLOGICAL ASSESSMENT
2 INTRODUCTION TO WETLAND NUTRIENT ASSESSMENT
3 THE STATE OF WETLAND SCIENCE
4 STUDY DESIGN FOR MONITORING WETLANDS
5 ADMINISTRATIVE FRAMEWORK FOR THE IMPLEMENTATION OF
A WETLAND BIOASSESSMENT PROGRAM
6 DEVELOPING METRICS AND INDEXES OF BIOLOGICAL INTEGRITY
7 WETLANDS CLASSIFICATION
8 VOLUNTEERS AND WETLAND BIOMONITORING
9 DEVELOPING AN INVERTEBRATE INDEX OF BIOLOGICAL
INTEGRITY FOR WETLANDS
1O USING VEGETATION TO ASSESS ENVIRONMENTAL CONDITIONS
IN WETLANDS
11 USING ALGAE TO ASSESS ENVIRONMENTAL CONDITIONS IN WETLANDS
12 USING AMPHIBIANS IN BlOASSESSMENTS OF WETLANDS
13 BIOLOGICAL ASSESSMENT METHODS FOR BIRDS
14 WETLAND BIOASSESSMENT CASE STUDIES
15 BIOASSESSMENT METHODS FOR FISH
16 VEGETATION-BASED INDICATORS OF WETLAND NUTRIENT ENRICHMENT
17 LAND-USE CHARACTERIZATION FOR NUTRIENT AND SEDIMENT RISK
ASSESSMENT
18 BlOGEOCHEMICAL INDICATORS
19 NUTRIENT LOADING
2O WETLAND HYDROLOGY
VI
-------
NUTRIENT LOADING
PURPOSE
Jhe purpose of this module is to describe
and discuss the general hydrologic prop-
erties that make wetlands unique, and to pro-
vide an overview of the processes that control
wetland hydrologic behavior. The intent is to
provide a general discussion of wetland hy-
drologic processes and methods in the hope
of fostering an understanding of the impor-
tant attributes of wetland hydrology relevant
to the monitoring and assessment of these
systems. As such, it is not intended to address
the narrower definition of wetland hydrology
for jurisdictional or classification purposes.
Also, this module should not replace more
advanced wetland texts. If the need arises to
obtain more specific information, the reader
is advised to refer to wetland books or ar-
ticles, including those referenced within this
document.
SUMMARY
A7\itrient loading to wetlands is deter-
L \ mined primarily by surface and subsur-
face transport from the contributing land-
scape, and varies significantly as a function
of weather and landscape characteristics such
as soils, topography, and land use. In the ab-
sence of sufficient measurements, nutrient
loading can only be estimated using an appro-
priate loading model. This module provides
an overview of hydrologic and contaminant
transport models that can be used to estimate
nutrient loads to wetlands.
Jhe purpose of this module is to provide
an overview of hydrologic and contami-
nant transport models that can be used to es-
timate nutrient loads to wetlands.
INTRODUCTION
r the past three decades, considerable
effort has been expended in developing
models to simulate watershed hydrology and
nutrient transport, particularly the estima-
tion of cumulative field/watershed contribu-
tions of flow, sediment, nutrients, and other
contaminants of interest. Appropriately used,
existing models may apply when in evaluat-
ing wetland reference conditions or establish-
ing nutrient criteria for wetlands or guiding
management decisions once nutrient criteria
are established.
Several reviews have summarized the char-
acteristics, features, strengths, and limitations
of models that are used for estimating wa-
tershed hydrology and water quality (Doni-
gian, et al., et al. 1991b, 1995b; DeVries and
Hromadka 1993; Novotny and Olem 1994;
Tim 1996a, 1996b). These models vary wide-
ly in structure and in spatial and temporal
scale, and can be classified as (i) empirical or
semi-empirical loading function models and
ii) process-oriented simulation models.
-------
LOADING
FUNCTION MODELS
T oading function models are based on em-
M^ pirical or semi-empirical relationships that
provide estimates of pollutant loads based on
long-term measurements of flow and contam-
inant concentration. They provide for rapid
estimation of critical pollutant loads with
minimal effort and data requirements. Load-
ing function models are widely used to esti-
mate pollutant loads in areas where limited
data sets are available for process-based mod-
eling. A major advantage of loading function
models is their simplicity. Generally, loading
function models contain procedures for esti-
mating pollutant load based on either heuris-
tics or on the empirical relationships between
landscape physiographic characteristics and
phenomena that control pollutant export.
McElroy et al. (1976) and Mills (1985) de-
scribed components of several screening models
developed by EPA's Environmental Research
Laboratory at Athens, Georgia to facilitate
estimation of nutrient loads from point and
nonpoint sources and to enhance preliminary
assessment of water quality. The model con-
tains simple empirical expressions that relate
the magnitude of nonpoint pollutant load to
readily available or measurable input param-
eters such as soils, land use and land cover,
land management practices, and topography.
This model is attractive because it can be
applied to very large watersheds often with
minimal effort and little or no calibration is
required.
Regression modeling, an approach based on
statistical descriptions of historic flow and
pollutant concentration data, is an alternative
to the screening model Regression models are
used to obtain preliminary estimates of pollut-
ant load under limiting and incomplete data.
These models require primary input param-
eters such as drainage area, percent impervi-
ousness, mean annual precipitation, land use
pattern, and ambient temperature. Regression
models can determine storm-event mean pol-
lutant load with confidence intervals for the
estimated loads.
In addition to regression modeling, sever-
al less complex, process-based models have
been used to estimate flow and contaminant
transport in terrestrial environments. Ex-
amples of process-based models include the
Generalized Watershed Loading Function
(GWLF), the Spatially Referenced Regres-
sions on Watersheds (SPARROW), and the
Pollutant Load model (PLOAD).
GENERALIZED WATERSHED
LOADING FUNCTION MODEL
The Generalized Watershed Loading Func-
tion Model (GWLF), developed at Cornell
University, estimates stream flow, nutrient
load and sediment load from watersheds
management areas. The model allows simu-
lation of point and nonpoint loadings of nu-
trients and pesticides from urban and agricul-
tural watersheds, including septic systems.
The model also provides data to evaluate the
effectiveness of certain land use management
practices. The GWLF is a temporally-contin-
uous simulation model with daily time steps,
but it is not spatially distributed. It simulates
overland flow and channel flow using a water
balance approach based on measurements of
daily precipitation and average temperature.
Precipitation is partitioned into direct surface
runoff and infiltration using the SCS Curve
2
-------
Number technique. Here, the Curve Number
determines the amount of precipitation that
runs off directly, adjusted for antecedent soil
moisture based on total precipitation dur-
ing the previous five days. A separate Curve
Number is specified for each combination of
land use types and soil hydrologic groups.
The amount of water available to the shallow
groundwater zone is influenced by evapo-
transpiration. This is estimated in GWLF us-
ing the available moisture in the unsaturated
zone, the evapo-transpiration potential, and a
cover coefficient. Potential evapo-transpira-
tion is estimated from a relationship to mean
daily temperature and the number of daylight
hours. GWLF calculates the groundwater dis-
charge by performing a lumped parameter
water balance on the saturated and shallow
saturated zones.
Soil erosion is modeled by the Revised Uni-
versal Soil-loss Equation (RULSE). Nutrient
fluxes in GWLF are estimated empirically
using daily nutrient fluxes from surface run-
off from pervious and impervious surfaces,
sediment erosion, groundwater base-flow,
and septic runoff. The monthly nutrient load
is calculated by totaling the daily nutrient
fluxes. In GWLF, the nitrogen and phospho-
rus loads from surface runoff are estimated by
multiplying excess runoff by their flow-weight-
ed average concentrations, respectively.
The model assumes that each specific land-
cover type has unique event-mean-concen-
tration processes that affect transport and
storage, and are unique to the land use. The
nutrient-loading model for urban land use is
based on an accumulation/wash off model.
Nutrient fluxes from impervious surfaces
and urban lands are estimated using chemical
build-up and wash-off parameterization. Both
nitrogen and phosphorus from eroded sedi-
ments are estimated using the sediment load,
enrichment ratio, and the concentration of
nitrogen and phosphorus in the top layer of
the soil. As with many mid-range terrestrial
models, GWLF calculates concentrations of
dissolved and sediment-bound nitrogen and
phosphorus in stream flow as the sum total
of base flow, stream flow (overland flow) and
point sources. Groundwater only contributes
dissolved nitrogen and phosphorus values re-
flecting the effects of local land use. Nutri-
ent losses in urban runoff are assumed to be
entirely in the solid-phase, while point source
losses are assumed to be dissolved.
The GWLF requires three categories of
input parameters: meteorological; hydrology
and landscape; and chemical and biophysical
(see Table 1). The model requires daily pre-
cipitation and temperature. The GWLF also
requires information related to land use, land
cover, soil, and parameters that govern run-
off, erosion, and nutrient load generation. The
strength of GWLF model is that data required
by this model are readily available from most
resource management agency databases.
In general, GWLF is an empirically de-
rived, statistically based process that uses
daily inputs of precipitation and temperatures
to compute nutrient fluxes. A major strength
of GWLF is its simplicity in estimating pol-
lutant load. Because of this, the model has
been used for screening landscapes accord-
ing to their pollutant delivery potentials or for
identifying critical areas of nonpoint pollu-
tion. However, it does not account for rainfall
intensity or storage along channels. Because
it uses a simplified technique for estimating
base flow, the model cannot reproduce the
precise history of overland flow and fluxes as
3
-------
TABLE 1: INPUT PARAMETERS REQUIRED BY GWLF MODEL
Basin/watershed size
Land use and land cover distribution
Curve Number by source area
USLE factors by source area
ET cover coefficient
Erosivity coefficients
Daylight hours by month
Growing season months
Initial saturated storage
Initial unsaturated storage
Recession coefficient
Seepage coefficient
Initial snow amount
Sediment delivery ratio
Soil water ava ilable cap acity
3.Chemical and Biophysical:
Dissolved N and P in runoff by land cover type
N and P concentrations in manure runoff
N and P buildup in urban areas
N and P from point sources
Background N and P in groundwater
Background N and P in top soil layer
Duration of manure spreading
Population on septic systems
Per capita septic system loads for N and P
do event-based models. It can, however, repro-
duce the frequency and magnitude of monthly
nutrient fluxes from undisturbed watersheds.
The GWLF model does not have a sufficient-
ly long history of application and may not be
applicable to land areas with a high degree of
altered hydrology.
SPATIALLY REFERENCED
REGRESSIONS ON WATERSHEDS
As described in Preston and Brakebill (1999)
the Spatially Referenced Regressions on Wa-
tersheds (SPARROW) model was developed
to relate the water quality conditions within
a watershed to sources of nutrients as well as
those factors that influence transport of the nu-
trients. Developed specifically for conditions
within the Chesapeake Bay watershed, the
SPARROW methodology utilizes statistical
techniques and spatially distributed landscape
data to estimate nutrient loads. Specifically,
the SPARROW methodology was designed
to provide statistically based relationships
between water quality and anthropogenic
factors (e.g., sources of contamination with-
in the watershed), land surface characteris-
tics that influence delivery of pollutants to
the stream, and in-stream transformation of
pollutants through chemical and biological
-------
pathways. The general form of the statistical
regression model for SPARROW is (Preston
and Brakebill 1999):
L, =
in which L^ = nutrient load in stream reach i; n,
N = pollutant source index; N = total number
of sources; J(i) = number of upstream stream
reaches; Rn = estimated source parameter; Sn,j =
contaminant mass from source n in drainage to
reach j; a = estimated vector of land-to-water
delivery parameters; z} = land surface charac-
teristics associated with drainage reach j (e.g.,
temperature, slope, stream density, irrigated
land, precipitation, and wetland); 8 = estimated
vector of in-stream loss parameter; and T13J =
channel transport characteristics. The source
parameter (3 consists of point sources, nutrient
applications in the form of animal manure,
commercial fertilizer, and atmospheric depo-
sition of pollutants. The parameter, a, deter-
mines the relative influence of different types
of land-surface characteristics on the deliv-
ery of nutrients from land surfaces to stream
channels.
The literature reports a number of applica-
tions of SPARROW model, primarily applied
to the Chesapeake Bay ecosystem. These ar-
ticles document water quality conditions and
assess the effectiveness of best management
practices in controlling nonpoint pollution.
The results provided the basis for not only
delineating watershed areas that are most
critical to the export of nutrients, but also for
targeting and prioritizing remedial control
strategies and conservation programs (Smith,
et al. 1997).
MIDRANGE MODELS
In addition to the GWLF and SPARROW
models described above, other modeling ap-
proaches utilize a compromise between em-
piricism and more complex mechanistic ap-
proaches. Typical examples of such models
include the Storm water Intercept and Treat-
ment Evaluation Model for Analysis and
Planning (SITEMAP) (Omnicron Associates
1990) and Pollutant Load model or PLOAD.
These models use daily time steps. Both can
be used to examine seasonal variability and
the load response to landscape characteristics
of specific watersheds. Due to their complex-
ity, they may have greater data requirements
and may require more site-specific data.
SITEMAP is a dynamic simulation model
developed to assist with simulating stream
segment waste-load allocations from point
and non-point sources. This model calculates
daily runoff and pollutant loading and can be
used for storm-event or continuous simula-
tions (including probability distributions) of
runoff, pollutant loads, infiltration, soil mois-
ture, and evapo-transpiration. SITEMAP can
be used in either single or mixed land uses,
and for event-based or continuous simulation
of surface runoff and pollutant load. Users of
the model are able to assess the effectiveness
of alternative management strategies and to
estimate load and waste-load from point and
nonpoint sources, respectively. The primary
outputs from the model include probabilis-
tic estimates of runoff volume and nutrient
loadings. A typical example application of
SITEMAP involved the assessment of pol-
lutant load and surface runoff in the Tualatin
River Basin and Fairview Creek watershed in
Oregon.
-------
PLOAD is a simplified GIS-based water-
shed-loading model. It can model combined
point and non-point source loads in either
small urban areas or in rural watersheds of
any size. As a loading model, PLOAD pro-
vides annualized estimates of pollutant ex-
port to waterbodies. Pollutants most com-
monly analyzed include sediments (TSS and
IDS), oxygen demand (BOD and COD), nu-
trients (nitrogen, nitrate plus nitrite, TKN,
ammonia, phosphorous), metals (lead, zinc)
and bacteria (fecal coliform), or any other
user-specified pollutant. The model addresses
pollutant loading by land use categories and
sub-watersheds, but does not as certain indi-
vidual non-point sources or at actual pollut-
ant fate and transport processes. Additional
features of the model include: (i) the ability to
estimate average annual pollutant load, (ii) a
user-friendly interface that enhances manipu-
lation of input parameters and the assessment
of alternative pollution control strategies,
(iii) tools to facilitate evaluation of land use
change impacts, and (iv) the ability to gener-
ate outputs at user-defined formats.
To use the PLOAD model, users are re-
quired to provide reasonably accurate val-
ues of input parameters describing wa-
tershed land use and land cover, pollutant
loading functions—based on land cover
types, location of point source inputs, land
areas with specified BMPs, and other gen-
eral watershed characteristics. When sup-
plied with these input variables, PLOAD
generates outputs that include average an-
nual loads, aggregated by sub-watershed,
and reported in tables and maps of loads by
watershed. In addition, users of PLOAD can
view and compare multiple loading scenarios
simultaneously.
The PLOAD is a part of the comprehensive
modeling tools in the EPAs Better Assess-
ment Science Integrating Point and Nonpoint
Sources (BASINS). The literature also re-
ports the applications of the PLOAD model
for assessing the effects of land use change
and BMPs for watersheds in North Carolina
and Maryland.
SUMMARY OF LOADING
FUNCTION MODELS
In summary, the many and diverse loading
function models developed to allow estima-
tion of point and non-point source pollution
loads are based on simplistic, functional and
empirical expressions that integrate flow and
pollutant concentration. Attractive features
of these models are that they: (i) require very
limited data and computer modeling experi-
ence; (ii) contain relatively simple procedures
for estimating pollutant load; and, (iii) pro-
vide tools for rapid assessment of point and
non-point contributions to the watershed pol-
lutant load. However, these advantages come
at some expense regarding accuracy, nature of
environmental process and conceptualization
of the physical system. In particular, most
loading function models fail to incorporate
the complex, nonlinear biogeochemical and
physical processes that influence the physical
system. Furthermore, loading function mod-
els are limited in how spatial and temporal
processes are handled and how landscape
variability is characterized. Despite these
limitations, there are situations in which these
models are logical and legitimate.
6
-------
PROCESS-ORIENTED
MODELS
/n contrast to the empirical and simplified
loading function models described above,
process-oriented simulation models integrate
knowledge of physical, chemical, and biologi-
cal processes with empirical data, and allow
users to evaluate interactions among human,
economic and societal factors. This section
provides an overview of some of the process-
oriented simulation models that have been
used to predict watershed hydrology and
water quality, and that could provide mod-
eling tools for predicting nutrient loading to
wetlands. These models include AGNPS and
AnnAGNPS, HEC-HMS, HEC-5Q, HSPF,
STORM, SWAT, SWMM, and SWRRB
models.
AGRICULTURAL NONPOINT
SOURCE MODEL
The Agricultural Non-Point Source Pollu-
tion Model (AGNPS) is event-based, as well
as a continuous or annualized AGNPS (An-
nAGNPS) simulation model. These models
predict surface runoff, sediment yield, and
nutrient transport primarily from agricultur-
al watersheds. The two main nutrients simu-
lated are nitrogen and phosphorus, which are
essential plant nutrients and are major con-
tributors to eutrophication and surface water
pollution. The basic model components include
hydrology, erosion, sediment, and chemical
transport (primarily nutrients and pesticides).
The model also considers point sources of
water, sediment, nutrients, and chemical oxy-
gen demand (COD from various sources in-
cluding feedlots). Water impoundments are
also considered as deposit!onal areas for sed-
iment-associated nutrients. The model also
has the ability to output water quality charac-
teristics at intermediate or user-defined points
throughout the watershed stream network.
The AGNPS model uses a grid-cell-based
subdivision of the watershed, in which each
cell is considered homogeneous. The cells
are linked together through the aspect or flow
direction, and all watershed characteristics
and primary biophysical inputs are expressed
at the grid-cell level. The components of the
model use equations and methodologies that
have been well established in the water quality
modeling literature and are extensively used
by resource management agencies. For exam-
ple, the runoff volume is estimated using the
SCS curve number technique. The peak run-
off rate for each grid-cell is estimated using an
empirical relationship in the CREAMS mod-
el (Knisel 1980). Soil erosion and sediment
yield are computed by using the USLE and a
bedload equation, a relationship—developed
by Foster et al. (1981) based on the continuity
equation. In the model, feedlots are treated
as point sources and pollutant contributions
from these sources are estimated by using the
feedlot pollution model developed by Young
(1982). Other point sources are accounted for
by incorporating incoming flow rates and
concentrations of nutrients to the cells where
they occur.
In the AGNPS model, the resolution for
the individual grid cells can range from 2.5
acres to greater than 40 acres (or 1 ha to
more than 10 ha) depending on the problem
being addressed, the size and complexity of
the watershed, and the technical expertise of
the modeler. Smaller grid-cell sizes such as
10 acres (4 ha) are recommended for water-
shed less than 2000 acres (800 ha). However,
7
-------
TABLE 2: INPUT PARAMETERS REQUIRED BY AGNPS MODEL
Watershed-level Input Parameters:
Watershed identification
Cell area (Acres)
Total number of grid cells
Precipitation (inches)
Energy-In tensity
Storm type
Cell-level Input Parameters:
Cell number
Aspect
SCS Curve Number
Average land slope (°o)
Slope shape factor (uniform, convex, COHC£IVC)
Average field slope length
Manning roughness coefficient
Soil erodibility factor (K-U SLE)
Cropping factor (C-USLE)
Practice (P-USLE)
Surface condition constant
Soil texture (sand, silt, clay, peat)
Fertilization level
Fertilization availability factor
Point source indication
Gully source level
Chemical oxygen demand factor
for watershed and catchments that are larger
than 2000 acres (800 ha), grid-cell sizes of 40
acres (16 ha) are normally used. The calcula-
tion of flow and transport processes in AG-
NPS occurs in three stages based on a set of
twenty or more parameters for each grid cell,
with the initial calculations for all cells in the
watershed made in the first stage. The sec-
ond stage calculates the runoff volume and
sediment yield for each of the cells contain-
ing impoundments and the sediment yields
for primary cells. A primary cell is one into
which no other cell drains.
The non-point source pollution component
of the model estimates transport and trans-
formation of nitrogen, phosphorus, chemi-
cal oxygen demand, and pesticides. Pollutant
transport is subdivided into soluble or dis-
solved phase and the sediment-attached or
sediment-bound phase. Soluble nitrogen and
phosphorus compounds are calculated using
a relationship adapted from the CREAMS
model (Knisel 1980); along with sediment
yield equations taken from the CREAMS and
the WEPP models. The input parameters for
the AGNPS model include: cell number, re-
ceiving cell number, SCS curve number, land
slope, field slope length, channel slope, chan-
nel side-slope, soil erodibility factor, cover
and management factor, support practice fac-
tor, surface condition constant, aspect, and
many other parameters related to land cover,
land topography, management practices, and
climate. The watershed-level parameters re-
quired include: area, area of each grid-cell,
characteristics of storm precipitation, and
8
-------
storm energy-intensity. Table 2 summarizes
the major input parameters required by the
AGNPS model.
The AGNPS model is, by far one of the
more widely used water quality models for
estimating the relative effects of agricultural
management practices in small to large wa-
tersheds. However, the model has many limi-
tations, including: lack of process-level de-
scription of nutrient transformation processes
or the biochemical cycling of major plant ele-
ments to document the biochemical cycling
during transport; inability to characterize
the transport and transformation of nutrients
and pesticides in stream channels or similar
waterbodies; inability to handle sub-surface
flow and transport processes, as well as sub-
surface interactions; the lack of a process to
route flow or pollutants from individual grid-
cells to the watershed outlet; and the model is
event-based.
ANNUAL AGNPS
To eliminate some of these limiting factors,
the AGNPS model has undergone numerous
refinements. The term "AGNPS" now refers
to the system of modeling components instead
of the single-event AGNPS described above.
These enhancements made to the event-based
AGNPS of the 1980s and early 1990s are in-
tended to improve the capability of the pro-
gram and to automate many of the input data
preparation steps needed for use with large
watershed systems. The current version of the
model is called AnnAGNPS, which is virtu-
ally the same computer program as AGNPS
5.x except that it allows for continuous simu-
lations of surface runoff, peak flow rate, and
pollutant transport for longer time periods
and on a daily basis. AnnAGNPS is designed
to handle watershed areas of up to 300,000
ha, and it divides the watershed area into sub-
divisions of homogenous cells with respect to
soil type, land use, and land management.
In contrast to the event-based model,
AnnAGNPS operates on a daily time step.
It simulates water, sediment, nutrients, and
pesticide transport at the cell and watershed
levels. Special components are included to
handle concentrated sources of nutrients
from feedlots and point sources, concentrat-
ed sediment sources with attached chemicals
from gullies, and irrigation (water with dis-
solved chemicals and sediment with attached
chemicals). Each day the applied water and
resulting runoff are routed through the wa-
tershed system before the next day is consid-
ered. The model partitions soluble nutrients
and pesticides between surface runoff and
infiltration. Sediment-transported nutrients
and pesticides are estimated and equilibrated
within the stream system, with the sediment
assumed to consist of five particle size class-
es (clay, silt, sand, small aggregate, and large
aggregate).
The soil profile is divided into two layers.
For estimating surface runoff, infiltration
and soil water storage. The top 200 mm are
used as a tillage layer whose properties can
change; the second layer's properties remain
static. A daily soil moisture water budget con-
siders applied water (rainfall, irrigation, and
snowmelt), runoff, evapo-transpiration, and
percolation. Surface runoff is estimated by
using the SCS Runoff Curve Number equa-
tion where the Curve Number can be modi-
fied daily, based on tillage operations, soil
moisture, and crop stage. Evapotranspiration
is estimated as a function of potential evapo-
transpiration by using the Penman equation
9
-------
(Penman 1948) and soil moisture content.
Erosion and sediment transport is predicted
within a watershed landscape according to
RUSLE (Renard, et al. 1997).
For each day and each grid cell, the model
calculates mass balances of nutrients (pri-
marily nitrogen, phosphorous), and organic
carbon. The model considers plant uptake of
nitrogen and phosphorus, fertilization, resi-
due decomposition, and nutrient transport.
Soluble and sediment-adsorbed nutrients are
estimated, and they are further partitioned
into organic and mineral phases. Each nu-
trient component is decayed based upon the
reach travel time, water temperature, and
appropriate decay-constant. The soluble nu-
trients are decreased further by infiltration.
Attached nutrients are adjusted for deposition
of clay particles Based on a first-order rela-
tionship, equilibrium concentrations are cal-
culated at both the upstream and downstream
points of reach. Plant uptake of nutrients is
modeled through a simple crop growth stage
index. A daily mass balance adapted from
GLEAMS (Leonard, et al. 1987) is estimated
for each pesticide. The pesticides have unique
chemodynamic properties, including half-life
and organic matter partitioning coefficient.
Major components of the pesticide model in-
clude foliage wash-off, vertical transport in
the soil profile, and degradation. Soluble and
sediment adsorbed fractions are calculated
for each grid cell on a daily basis.
AnnAGNPS also contains simplified meth-
ods to route sediment, nutrients, and pesti-
cides through the watershed. Peak flow for
each reach is calculated using an extension of
the TR-55 graphical peak-discharge method.
Sediment routing is calculated based upon
transport capacity relationships using the
Bagnold stream power equation. Sediments
are routed by particle size class, where each
particular size class can be deposited, more
entrained, or transported unchanged; de-
pending upon the amount entering the reach,
the availability of that size class in the chan-
nel and banks, and the transport capacity of
each size class. If the sum of all incoming
sediment is greater than the sediment trans-
port capacity, then the sediment is deposited.
If that sum is less than the sediment trans-
port capacity, the sediment discharge at the
downstream end of the reach will include bed
and bank material (if it is an erodible reach).
Nutrients and pesticides are subdivided into
soluble and sediment attached components
for routing. Attached phosphorus is further
subdivided into organic and inorganic. Each
nutrient component is decayed based upon
the reach travel time, water temperature, and
appropriate decay constant. Soluble nutrients
are further reduced by infiltration. Attached
nutrients are adjusted for deposition of clay
particles. Based on a first-order relationship,
equilibrium concentrations are calculated at
both the upstream and downstream points of
the reach.
AnnAGNPS includes 34 different input
data categories, which can be grouped into
climate, landscape characterization, agricul-
tural management, chemical characteristics,
and feedlot operations. The climatic data con-
sist of precipitation, maximum and minimum
air temperature, relative humidity, sky cover,
and wind speed. Land characterization data
include soil characterization, curve number,
RUSLE parameters, and watershed drainage
characterization. Agricultural management
relates to data on tillage, planting, harvest,
rotation, chemical operations, and irrigation
schedules. Feedlot operations include daily
10
-------
manure production rates, times of manure re-
moval, and residual amount from previous op-
erations. Indeed, there are over 400 separate
input parameters necessary for model execu-
tion. Some of these parameters are repeated
for each cell, soil type, land use, feedlot, and/
or channel reach. Separate parameters are
necessary for the model verification section.
Default values are available for some of the
input parameters. The daily climate data in-
put set includes twenty-two parameters, eight
of which are repeated for each day simulat-
ed. A climate generator, GEM, can be used
to generate the precipitation and minimum/
maximum air temperatures for AnnAGNPS.
The development of other input data can be
simplified because of duplication over a given
watershed. Some of the geographical inputs
including cell boundaries, land slope, slope
direction, and land use, can be generated by
GIS and digital elevation models. Model in-
put is facilitated by an input editor, which is
currently available with the model. The input
editor interface provides a page format for
data input, with each of the 34 major data cat-
egories on a separate input page. Input and
output can be in either all English or all met-
ric units. Separate input files for watershed
and climate data allow for quickly changing
climatic input.
Extensive data checks (with appropriate er-
ror messages) are performed as data are en-
tered and, to a lesser extent, after all data are
read. Output is expressed on an event basis
for selected stream channel reaches and as
source accounting from land or reach com-
ponents over the simulation period. Primary
outputs parameters generated by the model
relate to soluble and attached sediment-nu-
trients and pesticides, surface runoff volume
and peak flow, and sediment yield based on
particle size classes. Each output parameters
can be selected by the user for the desired wa-
tershed source locations (specific cells, reach-
es, feedlots, point sources, and gullies) and
for any simulation period. Source accounting
indicates the fraction of a pollutant load pass-
ing through any reach in the stream network
that came from the user-identified watershed
source location. In addition, event quantities
for user-selected parameters can be extracted
at desired stream reach locations.
A major limitation of the AnnAGNPS is that
it does not estimate transport of pesticide me-
tabolites or daughter products. Other limita-
tions of AnnAGNPS models include: (1) they
lack a nutrient transformation component for
both nutrient, and pesticides; (2) they lack a
subsurface or near-surface water flow com-
ponents; (3) they lack flow and contaminant
routing component; (4) all runoff and associ-
ated pollutant (sediment, nutrient, and pesti-
cide) loads for a single day are routed to the
watershed outlet before the next day simula-
tion begins (regardless of how many days this
may actually take); (5) there are no mass bal-
ance calculations tracking inflow and outflow
of water; (6) there is no tracking of sediment-
bound pollutants in the stream reaches; (7)
point sources are limited to constant loading
rates (water and nutrients); and, (8) there is no
provision for using spatially variable rainfall
inputs. Detailed information on AGNPS and
AnnAGNPS can be found at http://www.sed-
lab.olemiss.edu/PLM/AnnAGNPS.html.
HYDROLOGIC ENGINEERING
COMPUTATION- HYDROLOGIC
MODELING SYSTEM
The Hydrologic Engineering Computation -
Hydrologic Modeling System or HEC-HMS,
11
-------
developed by the U.S. Army Corps of Engi-
neers, is a physically-based model designed to
simulate precipitation runoff processes of den-
dritic watersheds. The model was developed
to allow the simulation of large river basins
and flood hydrology, as well as small urban
watersheds. HEC-HMS is the latest version
of the HEC-1 model and exhibits a number of
similar options for simulating precipitation-
runoff processes. In addition to unit hydro-
graphic and hydrologic routing functions, ca-
pabilities available with HEC-HMS include a
linear-distributed runoff transformation that
can be applied with gridded rainfall data, a
simple "moisture-depletion" option that can
be used for simulations over extended time
periods, and a versatile parameter optimiza-
tion option.
HEC-HMS also provides the capability for
continuous soil moisture accounting and res-
ervoir routing operations. Several options are
included in HEC-HMS to compute overland
flow and infiltration. These include the SCS
Curve Number equation, gridded SCS Curve
Number equation, and the Green-Ampt equa-
tion. In addition to unit hydrographic and
hydrologic routing options, other capabilities
of the model include: linear quasi-distributed
runoff transformation for use with gridded
precipitation and terrain data such as DEM;
continuous simulation with either one layer or
a more complex five layer soil moisture meth-
od; and, a versatile parameter estimation op-
tion. The modified Clark method, ModClark,
is a linear quasi-distributed unit hydrograph
method that can be applied with gridded pre-
cipitation. A variety of flow routing schemes
are included in the model. Hydrographs pro-
duced by the model can be used directly or in
conjunction with other model for studies of
water quality, urban drainage, flow forecasting,
reservoir spillway design, flood mitigation,
and flood management.
The HEC-HMS modeling environment has
been enhanced by geospatial technologies.
For example, the GEOspatial Hydrologic
Modeling Extension or HEC-GEOHMS is a
software package that integrates HEC-HMS
with ArcView GIS. GEOHMS also incor-
porates ArcView Spatial Analyst Extension
to allow users to generate model inputs for
HEC-HMS. Using the digital terrain data
from GIS databases, HEC-GEOHMS trans-
forms the drainage paths and watershed
boundaries into a hydrologic data structure
that represents watershed response to precipi-
tation. It provides an integrated, spatially-ex-
plicit simulation environment with data man-
agement and customized toolkit capabilities.
Other interactive capabilities allow users to
construct a hydrologic schematic of the wa-
tershed at stream gages, hydraulic structures,
and control points within the waterbody.
HEC-HMS also features a Windows-based
graphical user interface (GUI), integrated hy-
drological analysis components, data storage
and management capabilities, and graphics
and reporting tools. The data storage and ma-
nipulation component is used for the storage
and retrieval of time series, paired functions,
and gridded data, in a manner that is largely
transparent to the user. The HEC-HMS GUI
provides a means for specifying watershed
components, inputting data for each compo-
nent, and examining the results interactively.
It also contains global editors for entering or
examining data for all applicable landscape
elements.
Both HEC-HMS and HEC-GEOHMS
have long history of application as a quasi-
12
-------
dynamic hydrologic model. They are both in
the public domain and their technical refer-
ence manuals contain useful information on
how to model hydrological processes in gen-
eral, and the implementation of HEC-HMS
or HEC-GEOHMS in particular. In addition,
technical and users supports are adequate.
However, several factors limit the use of the
model in many situations, particularly when
assessing wetland hydrology. First, the model
was developed to predict the hydrologic re-
sponses of rural landscapes due to precipita-
tion and no water quality component is in-
cluded. Second, the model is unsuitable for
landscapes with significantly altered surface
hydrology due to, for example, tiling or other
landscape modification strategies. Finally, the
model does not have an explicit subsurface
modeling capability.
HYDROLOGIC ENGINEERING
COMPUTATION-5 QUALITY
The Hydrologic Engineering Computa-
tion-5 Quality or HEC-5Q is a water quality
model for use with U. S. Army Corps of En-
gineers' hydraulic model, HEC-5. The water
flow simulation module, HEC-5, was devel-
oped to assist in planning studies for evalu-
ating proposed reservoirs in a system and to
assist in sizing the flood control and conser-
vation storage requirements for each project
recommended for the system. It can also be
useful for selecting proper reservoir opera-
tional releases for hydropower, water supply,
and flood control.
The water quality simulation module, HEC-
5Q, is used to simulate concentrations of
various combinations of the following water
quality constituents: temperature, dissolved
oxygen, nitrate (NO3) - nitrogen, phosphate
(PO4) - phosphorus, ammonia (NH3) - nitro-
gen, phytoplankton, C-biochemical oxygen
demand, benthic oxygen demand, benthic
source for nitrogen, benthic source for phos-
phorus, chloride, alkalinity, pH, coliform
bacteria, three user-specified conservative
constituents, three user-specified non-con-
servative constituents, water column and
sediment dissolved organic chemicals, water
column and sediment heavy metals, water
column and sediment dioxins and furans, or-
ganic and inorganic particulate matter, sulfur,
iron and manganese.
Using estimates of system flows generated
by HEC-5, the HEC-5Q model computes the
distribution of temperature and other water
quality constituents in the reservoir and in
the associated downstream reaches. For those
constituents modeled, the water quality mod-
ule can be used in conjunction with the flow
simulation module to determine concentra-
tions resulting from operation of the reservoir
system for flow and storage considerations,
or alternately, for determination of flow rates
necessary to meet water quality objectives.
HEC-5Q can be used to evaluate options for
coordinating reservoir releases among proj-
ects to examine the effects on flow and water
quality at a specified location in the system.
Examples of applications of the flow simula-
tion model include examination of reservoir
capacities for flood control, hydropower, and
reservoir release requirements to meet water
supply and irrigation diversions. The model
may be used in applications including the
evaluation of in-stream temperatures and
constituent concentrations at critical loca-
tions in the system, examination of the poten-
tial effects of changing reservoir operations
on temperature, or water quality constituent
concentrations.
13
-------
Reservoirs equipped with selective with-
drawal structures may be simulated to de-
termine operations necessary to meet down-
stream water quality objectives. With these
capabilities, planners could evaluate the ef-
fects on water quality of proposed reservoir-
stream system modifications and determine
how a reservoir intake structure could be op-
erated to achieve desired water quality objec-
tives within the system.
The 1997 version of HEC-5Q, modified
by Resource Management Associates, Inc.,
under contract to the HEC, provides flex-
ibility when applying it to systems consist-
ing of multiple branches of streams flowing
into or out of reservoirs, which may be placed
in tandem or in parallel configurations. The
user can specify the number of streams and
reservoirs that can be modeled, and program
dimensions can be increased to meet project
needs.
HYDROLOGIC SIMULATION
PROGRAM-FORTRAN
The Hydrologic Simulation Program-For-
tran or HSPF (Johansen, et al. 1984; Bicknell, et
al. 1993; Donigian, et al. 1995a) is a physically
based, semi-distributed and deterministic mod-
el developed during the mid-1970's to predict
watershed hydrology and water quality for both
conventional and toxic organic pollutants. It
provides an analytical tool for: (i) planning,
design and operation of water resource sys-
tems; (ii) watershed, water-quality manage-
ment and planning; (iii) point and non-point
source pollution analyses; (iv) fate, transport
exposure assessment and control of conven-
tional and toxic pollutants; and, (v) evaluation
of urban and rural agricultural management
practices. HSPF combines three process-ori-
ented models: the Agricultural Runoff Man-
agement Model or ARM (Donigian and Davis
1978); the Non-point Source Runoff Model or
NPM (Donigian and Crawford 1979); and,
the Hydrologic Simulation Program or HSP
and its water quality component (Hydrocomp
1977). All of these components were seam-
lessly combined into a basin-scale framework
for simulating water quantity and water qual-
ity conditions of terrestrial and aquatic sys-
tems (Bicknell, et al. 1993) and for integrated
analysis of in-stream hydraulic process.
HSPF provides continuous simulations of
hydrological water balance, chemical trans-
port and fate in the terrestrial environment.
It also includes an in-stream water quality
component for evaluating nutrient fate and
transport, biochemical oxygen demand, dis-
solved oxygen, phytoplankton, zooplankton,
and benthic algae. In general, the model con-
sists of three primary application modules: (1)
PERLND, which simulates water budget and
runoff processes, snowmelt and accumula-
tion, sedimentation, nutrients (e.g., nitrogen,
phosphorous) and pesticide fate and transport
in runoff, and movement of a chemical tracer
(e.g., bromide); (2) IMPLND, which simu-
lates impervious land area runoff and water
quality; and (3) RCHRES, which predicts
movement of runoff water and water quality
constituents in stream channels and mixed
reservoirs.
The PERLND module includes process-
based functions for predicting: (1) Ambient
temperature as a function of elevation dif-
ferences between land segment and weather
station (ATEMP); (2) Water budget resulting
from precipitation on each previous land seg-
ment (PWATER); (3) Sediment deposition and
detachment from the land areas (SEDMNT);
14
-------
TABLE 3: INPUT PARAMETERS REQUIRED BY HSPF
Soils
Geology
Land-surface elevation (DEM)
Land use and land cover
Hydrography/natural drainage network
Artificial drainage network
Drainage basin delineation
Innut Time
Str.
•earn How
Precipitation (daily/breakpoint)
Air Te mperatures (Maximum/Minimum)
Water use
Auxiliarv Data for Hvdrolop-ic Modeling
Channel geometry, roughness and gradient
Discrete-sample data foe water quality modeling
Nutrient concentrations
Sediment concentrations (total suspended sediment)
Sediment size distribution
Field parameters (e.g., dissolved oxygen, pH, etc.)
GIS and Auxiliarv data for Water Oualitv Modeling
Cropland
Pasture
CAFOs
Fertilizer application rates
Manure application rates
Atmospheric deposition
We tlands
Point Sources
(4) Soil temperature for surface and subsur-
face layers and its impact on flow and contam-
inant transport, (PSTEMP); (5) Surface run-
off water temperature and dissolved oxygen
and carbon dioxide concentrations in over-
land flow (PWTGAS); (6) Water quality con-
stituents in the surface and subsurface flows
from each previous land segment (PQUAL);
(7) Storage and moisture fluxes and solute
transport in each soil layer or compartment
(INSTLAY); and, (8) (Movement and behav-
ior of pesticides (PEST), nitrogen (NITR),
phosphorus (PHOS) and tracers (TRACER)
through the top surface soil profile.
The EVIPLND module of HSPF predicts rele-
vant flow and transport processes in the imper-
vious land segments. It contains compartment
equations for simulating air temperature at
different locations within the watershed or
basin (ATEMP) as in the PERLND module,
snow accumulation and snowmelt (SNOW),
hydrologic water budget that includes infiltra-
tion and other interactions (IWATER), solids
accumulation and removal (SOLIDS), surface
runoff water temperature and gas concentra-
tions (IWTGAS), and generalized water quality
constituents. These modeling compartments
are similar to the PERLND module except
that little or no infiltration and other surface-
subsurface interactions occur.
In the RCHRES module, constitutive equa-
tions are used to route runoff and water
quaity constituents predicted by the PER-
LND and IMPLND modules through stream
-------
channel networks and reservoirs. The
RCHRES module also simulates those pro-
cesses that occur in open channels, such as
sediment detachment and deposition; chemi-
cal phase partitioning and transformation
(e.g., oxygen and biochemical oxygen de-
mand); plankton population; nitrogen and
phosphorus mass balances; and total carbon
and carbon dioxide concentrations. Embed-
ded within RCHRES module are compart-
ment equations for describing channel flow
hydrodynamics (HYDR), sediment transport
(SEDTRN) advection of water quality con-
stituents (ADCALC), transport of conserva-
tive chemicals and water quality constituents
(CONS and EQUAL) and including synthetic
organic chemicals and pesticides.
The HSPF modeling environment also con-
tains five utility models that enhance access,
manipulation and analysis of time-series of
model parameters, including hourly precipita-
tion, daily evaporation and daily stream flow
(Table 3). These utility modules include the
following: (i) COPY, which copies data resid-
ing in the time series store or watershed man-
agement titles to another file; (ii) PLTGEN,
which creates an ASCII file for display on
a plotter or for input to other programs; (iii)
DISPLAY, which generates summary data in
tabular form, (iv) DU RANL, a utility pro-
gram for frequency, duration and statistical
analyses; and, (v) GENER, which transforms
one or more time series to produce a new
or different time series. In addition to these
utility programs, ancillary programs such as
ANNIE (Lumb, et al. 1990) and HSPEXP
(Lumb and Kittle 1993) are used with HSPF
to interactively manipulate, store, retrieve,
list, plot, and update spatial, parametric and
time-series data. ANNIE and other similar
interactive pre and post-processing software
programs greatly reduce the massive data
size and intensive data demands of HSPF.
HSPEXP is a stand-alone land-surface hydro-
logic computation module that incorporates
an expert system component for model cali-
bration and for other modeling support.
Since its debut in the early 1980s, HSPF has
undergone a number of enhancements. Some
of these improvements were in direct response
to changes in computer operating systems
(e.g., shift from DOS to Windows), comput-
ing environment (e.g., from mainframe to
minicomputer), human-computer interaction
(e.g., paradigm shift from command line inter-
faces to GUIs), and user requirements (e.g.,
the need to predict hydrology and water qual-
ity of mixed land-use watersheds.) Today,
HSPF can be implemented on most computer
platforms, from laptops to the largest super-
computers using DOS, Windows, UNIX, or
other platforms. Depending on the size of the
watershed or basin, an HSPF simulation can
be efficiently executed on a 486-based mi-
crocomputer or a Pentium III (or greater) mi-
crocomputer with/without extended memory.
Overall, the HSPF modeling code accommo-
dates a wide range of operating environments
and user competencies. However, for water-
sheds and basins with complex land-use and
significant spatial heterogeneity, powerful
computing resources and high levels of mod-
eling competency are required.
The capabilities, strengths, and weaknesses
of HSPF have been demonstrated by its many
applications to urban and rural watersheds
(e.g., Donigian, et al. 1990; Moore, et al.
1992; and Ball, etal. 1993). Some applications
have featured more comprehensive and innova-
tive uses of the model, particularly its ability
to handle complex landscapes and environmental
16
-------
conditions. For example, Donigian et al.
(1990, 1991a) and Donigian and Patwardhan
(1992) describe the application of HSPF with-
in the framework of the Chesapeake Bay pro-
gram to determine total contributions of flow,
sediment, and other water quality constituents
(e.g., dissolved oxygen and nutrients) to the tidal
region of the Chesapeake Bay estuary. They
use HSPF to estimate total loads of nitrogen
and phosphorus entering the Chesapeake Bay
from contributing sub-basins under a range
of land management scenarios and to evalu-
ate the feasibility of the 40% reduction in
non-point polluted loads to the Bay.
In another application of the model, the
Maryland Department of the Environ-
ment use HSPF to quantify nonpoint source
contributions to the water quality impairment
in the Patuxent River and to evaluate alterna-
tive strategies for improving downstream wa-
ter quality in the Patuxent River Estuary. In
this application, the HSPF provides estimates
of non-point pollution loads from complex
mixed land-use areas of the drainage basin,
and the in-stream water quality throughout
the river system.
As part of the EPA's Better Assessment Sci-
ence Integrating point and Nonpoint Sources
(BASINS) tool, HSPF is being applied to wa-
tersheds and basins for watersheds and wa-
ter-quality based assessment for developing
the Clean Water Act Total Maximum Daily
Loads. Linked to Windows-based user inter-
face, HSPF constitutes the major component
of BASINS' nonpoint source model (NPSM)
that estimates land-use-specific nonpoint
source loadings for selected pollutants within
the watershed.
STORAGE TREATMENT
OVERFLOW RUNOFF MODEL
The Storage Treatment Overflow Runoff
Model or STORM is a model designed by the
Hydrologic Engineering Center of the U.S.
Army Corps of Engineers to simulate run-
off from urbanized landscapes. This model
consist of components that facilitate rainfall-
runoff assessment, water quality simulation,
and statistical and sensitivity analysis of the
modeling results. In general, STORM's ad-
vantage over other continuous simulation
models because of its relatively simple struc-
ture and moderate data requirements. It par-
ticularly addresses combined sewer outflows,
although it may be used to simulate storm-
water runoff quality and quantity. The hydro-
logic modeling procedures in STORM adopt
a modified rational formula with a simplified
runoff coefficient and depressive storage. Wa-
ter quality constituents are estimated based
on buildup or wash-off functions, and include
total suspended and settled solids, BOD, total
coliform, ortho-phosphate, and total nitrogen.
The model does have capability of continuous
and diffuse source release and uses the USLE
to estimate soil erosion by water. Limitations
of the STORM include minimal flexibility in
parameters with which to calibrate model to
observed hydrographs, lack of a desktop ver-
sion that operates in desktop environment,
and the large amount of input data required
for its application.
SOIL AND WATER
ASSESSMENT TOOL
The Soil and Water Assessment Tool or
SWAT (Arnold, et al. 1995) was developed
by the USDA,Agricultural Research Services
17
-------
by combining the modeling components of
SWRRB-WQ, EPIC, and ROTO, with a
weather generator. SWAT provides continu-
ous, long-term simulation of the impact of
land management practices on water, sedi-
ment, and agricultural chemical yields in
large complex watersheds. The SWAT model
assists resource planners in assessing non-
point source pollution impacts on watersheds
and large river basins. According to Arnold et
al. (1998), the model: (i) is based on physical
processes—associated with water flow, sedi-
ment detachment and transport, crop growth,
nutrient cycling, and pesticide fate and trans-
port; (ii) uses readily available input param-
eters and standard environmental databases;
(iii) is computationally efficient and supports
simulation of large basins or a variety of man-
agement scenarios and practices; and, (iv)
enables users to examine long-term implica-
tions of current and alternatives agricultural
management practices that can be juxtaposed
on the rural landscape.
In the development of the SWAT model, em-
phasis was placed on: (i) reasonably accurate
depiction and characterization of the agricul-
tural land management and spatial variability;
(ii) accurate prediction of pollutant load; (iii)
flexibility in discretization of the watershed
into homogeneous, manageable sub-basins;
and, (iv) continuous, long-term simulations
as opposed to discrete storm-event simula-
tions of most quasi-distributed models.
The SWAT modeling code consists of eight
major components: hydrology, weather, sedi-
mentation, soil temperature, crop growth,
nutrients, pesticides, and agricultural man-
agement. Hydrologic processes simulated by
the model include surface runoff, estimated
using the curve number methodology with an
option to simulate infiltration on the basis of
the Green-Ampt equation; percolation mod-
eled with a layered storage routing technique
combined with a crack flow model; lateral
subsurface flow; groundwater flow to streams
from shallow aquifers; potential evapora-
tion by the Hargraves, Priestley-Taylor, and
Penman-Montheith techniques; snow melt;
transmission losses from stream; and, water
storage losses from pond and reservoirs. Me-
teorological variables that drive the hydrologic
modeling component of SWAT include: daily
precipitation, daily minimum and maximum
temperatures, solar radiation, relative humid-
ity, and wind speed. For watersheds without
historical or current measurements of these
climatic data variables, a weather generator
can be used to synthetically simulate all or
some variables based on monthly histori-
cal statistics. Different climatic data can be
associated with specific sections of the wa-
tershed.
Sediment yield from individual sub-basins
and hydrologic response units is computed by
using the modified Universal Soil Loss Equa-
tion. Crop growth is predicted by using algo-
rithms from the EPIC model that character-
izes plant phenological developments based
on daily accumulation of heat units, harvest
index for partitioning grain yield, Montheit's
approach for potential biomass, and adjust-
ments for temperature and water stress.
Nitrate-N losses in runoff, deep percolation,
and lateral subsurface flow are simulated us-
ing methodologies in CREAMS and SWRRB-
WQ models. The transformation processes of
nitrogen (N) considered in SWAT include
mineralization (residue and humus), nitrifica-
tion, denitrification, volatilization, and plant
uptake. For phosphorus (P), the transforma-
tion processes include mineralization, soluble
18
-------
TABLE 4: SWAT/SWAT 2OO INPUT PARAMETERS
Sub-basins
Reach and main channels
Hydrologic respons e units
Groundwater aquifer data
Channel characteristics
General water quality information
Stream and lake water quality
Point sources
Ponds/wetlands/ reservoir days
Tributary channels
Precipitation (daily)
Solar radiation
Min/max temperatures
Solar radiation and wind speed
Relative humidity
Potential evapo-transpi ration
Soils and soil properties
Management practices
Fertilizer application
Manure application
Pesticide application
Urban data
P in runoff, sediment-bound P, P fixation by
soil particles, and crop uptake. Pesticide trans-
port and transformation follow algorithms in
the GLEAMS model and include equations
for describing interception by crop canopy,
volatilization, soil degradation, losses in run-
off and sediment, and leaching. Agricultural
management practices in the SWAT model
include tillage effects on soil and residue mix-
ing, bulk density and residue decomposition,
irrigation, and chemical management.
water balance). Algorithms are included to
characterize in-stream parameters such as
chlorophyll, dissolved oxygen, organic N,
ammonia-N, and biological oxygen demand.
Within stream and reservoirs, the model
facilitates the simulation of major processes
including outflow, nutrient and pesticide load-
ing, nutrient and pesticide transformations,
volatilization, diffusive transport of chemical
constituents, and chemical/sediment resus-
pension.
In the SWAT model, the stream channel
processes include channel routing (flood, sed-
iment, nutrients, and pesticides) and reservoir
routing (sediment, nutrients, pesticides, and
Because of its semi-distributed parameter
nature, coupled with its extensive climatic,
soil, and management databases, the SWAT
model is probably one of the most widely
19
-------
used hydrologic and water quality model for
large watersheds and basins. To enhance the
use of the model, several interfaces that link
the modeling code with geographic informa-
tion systems (GIS) have been developed. For
example, Srinivasan and Arnold (1994) de-
scribe an interface that links the SWAT mod-
el to the GRASS (Geographical Resources
Analysis Support System), a raster-based GIS
software package. This interface supports
watershed delineation into hydrologically
homogeneous units and enhances the extrac-
tion of appropriate soil, topographic, climate,
agricultural management, and land use data
for modeling and the display of the results in
the results in the form of maps and graphs.
Building on the popularity and the look and
feel of the ArcView GIS (Environmental Sys-
tems Research Institute, Redlands, CA), an-
other interface was developed for the SWAT
model.
The SWAT-ArcView user interface con-
tains appropriately structured components
and functions for generating sub-basin topo-
graphic attributes and model parameters, ed-
iting of input coverages and data, running the
SWAT model, and displaying model outputs
in a user-defined format. With more than 500,
000 copies in use worldwide ArcView GIS is
probably the most versatile desktop software
for the manipulation, analysis, modeling, and
visualization of geographically referenced
data. The interface uses the many capabili-
ties of ArcView GIS to offer users desirable
housekeeping functions such as creating a
new SWAT project (wherein a project refers
to a set of model parameters and model ap-
plication), editing of the modeling database,
and opening, copying, and deleting of a SWAT
project. In general, the interface consists of
customized menus and dialog boxes that fa-
cilitate interactive manipulation of watershed
and modeling database and for interrogating
the modeling code.
As a quasi-distributed model, one of the
many limitations of SWAT is that it is input
data intensiveness andit requires the specifi-
cation of an appropriate data format that en-
sures error-free simulation (see Table 4 for a
partial list). The primary input parameters in-
cludes those that describe the watershed (e.g.,
area), the watershed landscape (e.g., number
of hydrologic response units, number of sub-
basins, average sub-basin slope, etc.), agri-
cultural management (e.g., date of planting,
chemical application, tillage, and harvesting),
and the climatic conditions within the wa-
tershed. These input data categories are ar-
ranged in different hierarchically structured
data files with definable extensions. For ex-
ample, parameters that describe the different
hydrologic response units within a sub-basin
are constituted under the *.sub input file.
They include tributary channels, amount of
topographic relief and its influence on climatic
conditions within a sub-basin, parameters af-
fecting surface and subsurface water flow and
contaminant transport. Likewise, the param-
eters describing soil physical and chemical
properties within each hydrologic response
unit are arranged as input files with *.sol and
*.chm extension, respectively, while the 14
different types of agricultural management
operations simulated by SWAT are defined in
the *.mgt input file extension.
To further assist users in creating and or-
ganizing input data for modeling, a digital
database and customized menus are provid-
ed with the modeling code. Users of SWAT
can select and use the following data sets:
(i) USDA-NRCS STATSGO soil-association
20
-------
database- consisting of soil map unit polygons
and attribute data; (ii) digital elevation model
(DEM) for the contiguous United States as
derived from 1:250,000 scale USGS topo-
graphic data; (iii) Anderson Level III classi-
fied land use/land cover data created by using
the l:250,000-scale USGS LUDA; and, (iv)
historical climatic database for 1130 weather
stations located across the U.S.
The SWAT model and the ArcView GIS
modeling interface are available to users
worldwide through the model's Web site
(http://www.brc.tamu.edu/swat/swatdoc.
html) or by sending an e-mail request to the
principals at the Blackland Research Center,
Temple, TX. The SWAT models runs on a
number of operating environments including
Windows (95, 98, NT, and 2000) as well as
Unix workstations. Version 99.2 of the SWAT
model, for example, requires about 16 MB of
RAM, a 486 or Pentium processor, and 10 to
15 MB of disk storage.
The SWAT model has found widespread
application in many modeling studies that
involve systemic evaluation of the impact of
agricultural management on water quality.
Several case studies are available in the lit-
erature that demonstrate the reliability of the
model. For example, as part of the national
Coastal Pollutant Discharge Inventory, the
National Oceanic and Atmospheric Adminis-
tration utilized the SWAT model to estimate
nonpoint source loading into all U.S. coastal
areas. Srinivasan et al. (1998) describe the
application of SWAT to selected watersheds
in the Upper Trinity River Basin in Texas.
Manguerra and Engel (1998) report the use of
SWAT model to evaluate runoff from two ag-
ricultural watersheds in west central Indiana.
More recent applications of the SWAT model
include watershed assessments and nonpoint
source pollution control in Texas (Rosenthal,
et al. 1995), Mississippi (Bingner 1996), and
Indiana (Engel and Arnold 1991). The U.S.
Environmental Protection Agency is consid-
ering adopting SWAT as a nonpoint source
modeling component of its BASINS (Better
Assessment Science Integrating Point and
Nonpoint Sources) modeling environment.
The current version of BASINS uses the
HSPF model to assist in delineating impaired
and critical watersheds and for analyzing
baseline nonpoint source loadings and for ex-
amining total maximum daily load allocation
scenarios and TMDL compliance assessment
within watersheds.
STORM WATER
MANAGEMENT MODEL
The Storm Water Management Model or
SWMM is a comprehensive computer model
used for the analysis of water quantity and
quality of runoff. The model has been widely
used to perform either single event or contin-
uous simulation (i.e. long-term) of hydrologic
and hydraulic problems of both combined and
separate sewer systems, as well as for assess-
ing urban nonpoint pollution problems. The
model predicts flows, stages, and pollution
concentrations. SWMM also simulates all
components of the hydrologic cycle includ-
ing, rainfall, snowmelt, surface runoff and
subsurface flow, flow/flood routing through
drainage networks, storage, and treatment.
SWMM can be used both for planning and
designing sewers and for evaluating the hy-
drology of urban watersheds including those
with wetlands. In planning mode, the model
can be used as an overall assessment of the
urban runoff problems and potential pollutant
21
-------
abatement options. This mode is realized by
continuous simulation of hydrology and hydro-
logic conditions using long-term precipitation
data. Users can perform frequency analysis of
predicted hydrographs and pollutographs, and
examine hydrological events of specific inter-
est. In design mode, event simulation may
also be performed using a detailed watershed
schematization and shorter time steps for the
precipitation input. SWMM is structured
around six different, but related, modules or
blocks including: (i) RAIN, which processes
precipitation data for input into the RUNOFF
block; (ii) RUNOFF, which generates runoff
volume and quality from precipitation on the
watershed; (iii) TEMP, which processes are
temperature data for snowmelt computations;
(iv) TRANSPORT, which is based on kine-
matic wave routing of flow and quality, base
flow generation, and infiltration; (v) STOR-
AGE and TREATMENT, which handles de-
tention; and, (vi) EXTRAN, which handles
dynamic flow routing equations (Saint Venant's
equations) for accurate simulation of back-
water, looped connections, surcharging, and
pressure flow. Within the EXTRAN block,
users can perform sophisticated hydrau-
lic analysis of urban drainage networks us-
ing either the Saint Venant's hydrodynamic
equations or the kinematic wave equations.
The RAIN block facilitates the processing of
hourly and 15-minute (breakpoint) precipita-
tion time series for input to continuous simu-
lation. It also includes the statistical analysis
procedures of the EPA SYNOP model used
to characterize storm events. By using these
blocks, users can simulate all aspects of the
urban hydrologic and quality cycles, includ-
ing rainfall, snowmelt, surface and subsur-
face runoff, flow routing through the drain-
age network, storage and treatment.
In SWMM, the watershed/basin is divided
into basic spatial units called sub-watersheds
or subbasins. Each sub-watershed requires
specification of a number of parameters char-
acterizing its landscape. Data requirements
for hydraulic and hydrologic simulation in-
clude area, imperviousness, slope, surface
roughness, and depression storage and infil-
tration parameters. Land use information is
used to determine type of ground cover for
each model sub-area. Depression storage can
be estimated from rainfall and stream flow
data or from published literature values. Soil
infiltration parameters are calculated from ei-
ther the Horton equation or the Green-Ampt
equation. Manning roughness values for pre-
vious and impervious areas are estimated
form published values for each land cover
type. Water quality processes in SWMM are
simulated by a variety of options, including
constant concentrations, regression relation-
ships (load vs. flow), buildup and wash-off
Other water quality processes in SWMM are
those associated with precipitation, land sur-
face, erosion, sedimentation, soil, deposition
and treatment. SWMM can predict up to ten
different pollutants during a single simula-
tion session. Pollutants that can be simulated
include total nitrogen, total phosphorus, or-
tho-phosphate, copper and zinc. Ten different
land uses can be simulated and land uses are
grouped as appropriate. The event-mean con-
centration can be calculated for each pollut-
ant and each land use.
Depending on the objective of the model
application, the input data requirements by
SWMM can be minimal or extensive. The
data collection and data preparation activities
for simulation modeling can be intensive,
particularly for large watersheds and drain-
age networks. For example, the simulation
22
-------
of sewer hydraulics requires expensive and
time-consuming field verification of sewer
invert elevations. Extensive data is also need-
ed for the model calibration and validation.
The outputs generated from SWMM consist
of hydrographs and pollutographs (concentra-
tion vs. time) at any desired point within the
drainage system. Users can output depths and
velocities of flow as well as summary statistics
defining surcharging, volumes, continuity, and
other water quality parameters. The statistics
block can be used to separate the hydrographs
and pollutographs into storm events and to
compute summary statistics on parameters
such as volume, duration, inter-event time,
load, average concentration, and peak con-
centration. Model outputs can be in tabular
or geographical format. There are options
for dynamic plots of the hydraulic grade line
produced by the EXTRAN module. Linkages
have also been developed between SWMM
and GIS.
SWMM is perhaps one of the most widely
used models developed by the EPA for urban
runoff simulations. Originally developed be-
tween 1969 and 1971, SWMM has withstood
many verification tests. It continues to be
used in countries throughout the world in-
cluding the United States as well as in Aus-
tralia, Canada and Europe. A large body of
literature exists describing the applicability of
the model. Within the United States, applica-
tions of SWMM are many and varied Span-
ning states as varied as California, Florida,
and Virginia. The U.S. Geological Survey
has used the model to predict hydrology of a
watershed in Rolla, Missouri. The model was
applied to the Winter Haven chain of lakes
and its watersheds to predict pollutant load-
ing to the lake and to examine the effects of
human activities on lake water quality.
One of the major strengths of SWMM is
its ability to predict hydraulic systems such
as drains, detention basins, wetlands, sew-
ers, and related flow controls. The SWMM,
however, does have a number of limitations
including: (i) the lack of component equa-
tions and functions to route subsurface flow
and water quality; (ii) limited interactions
between the relevant biophysical and chemi-
cal processes; (iii) the reliance on first-order
rate kinetics to describe pollutant transfor-
mation in the TRANSPORT block; and, (- iv)
the lack of explicit functional components to
predict biogeochemical cycling in receiving
waterbodies and control structures.
One drawback, when using of earlier ver-
sions of SWMM, is the lack of an appropriate
user interface. Over the past decade develop-
ers have worked to enhance the "look-and-
feel" of the model's interface using interfaces
such as MIKE-SWMM, PC_SWMM, and
XP-SWMM. In response to EPAs clients'
need for improved computational tools for
managing urban runoff and wet weather wa-
ter quality problems, the agency has support-
ed development of a new version of SWMM
that incorporates recent advancements in
software engineering methods and updated
computational techniques. In this new ver-
sion, the architecture of SWMM's compu-
tational scheme has been revised by using
object-oriented programming techniques.
This revision of SWMM resulted from a col-
laborative effort between EPA-NRMRL's
Water Supply and Water Resources Division
and Camp Dresser McKee, Inc. New fea-
tures include: improved prediction of infil-
tration, soil moisture accounting, functions
23
-------
for estimating groundwater flow and energy
balance, and techniques for routing surface
water flow. They also incorporated features
such as—Lagrangian water quality transport
model, bed/suspended load sediment trans-
port model, and interactive real-time control
of sewer flow routing.
SIMULATION OF WATER
RESOURCES IN RURAL
BASINS-WATER QUALITY
The Simulation of Water Resources in Ru-
ral Basins-Water Quality (SWRRB-WQ)
(Arnold, et al. 1990) adapts the CREAMS
(Knisel 1980) model to provide predictions
of hydrologic, sedimentation, nutrient and
pesticide transport in large, complex rural
watersheds and basins. The primary objec-
tive of the model is to predict the effects of
alternative management decisions on water
flow, sediment yields, and chemical trans-
port with an acceptable level of accuracy for
un-gauged rural basins and watersheds. The
major modifications to the CREAMS model
which resulted in the SWRRB-WQ are: (i)
the modeling code now allows simultaneous
computation of several sub basins to predict
water and sediment yields and chemical load-
ing, and each sub-basin was considered a
homogeneous entity; (ii) a return flow com-
ponent appropriately simulates the soil water
balances; (iii) reservoir storage routing com-
ponent provides estimates of effects of ponds
and reservoirs on water flow and sediment
yield; (iv) a weather simulation model pro-
vides statistical, daily estimates of weather
inputs such as precipitation, solar radiation,
and minimum and maximum temperatures;
(v) plant growth model provides predictions
of management and natural and anthropogenic
inputs on variation in crop growth; and, (vi)
components are incorporated to enable simu-
lation of sediment movement in ponds, reser-
voirs and streams. In general, the SWRRB-
WQ handles the major biophysical processes
including surface runoff, percolation, return
flow, evapo-transpiration, transmission losses,
pond and reservoir storage, sedimentation,
nutrient cycling, pesticides fate and transport,
and plant growth.
In the SWRRB-WQ model, the water bal-
ance in the soil-plant-water atmosphere sys-
tem is represented by the hydrologic model-
ing component. Thus, the hydrological cycle,
particularly the soil water balance, is de-
scribed by the equation:
SW.-SW =
i-Qi- ETi- Pi-
in which SW = soil-water content less 15-bar
water content; t = time in days; R, Q, ET, P
and QR = daily amount of precipitation, run-
off, evapo-transpiration, percolation, and re-
turn flow, respectively surface runoff, Q is es-
timated by using modified form of the runoff
curve number technique and sediment yield is
predicted by using modified USLE (Williams
and Berndt 1977).
Nutrient yield and nutrient cycling in
SWRRB-WQ adopts the expressions developed
in the EPIC model (Williams, et al. 1989)
and the quantities calculated for each sub-
watershed is routed to watershed outlet. The
nutrient load is distributed between the sol-
uble and sediment-bound phases. Pesticides
fate and transport modeling in SWRRB-WQ
adopts the methodology and equations in
GLEAMS model (Leonard, et al. 1987). As
with nutrients, the pesticides are distributed
between the soluble and adsorbed phases
according to the organic matter content of the
soil.
24
-------
Inputs parameters required for SWRRB-
WQ model simulations are related to process-
es such as hydrology, sediment yield, chemical
fate and transport, and channel routing. The
basic inputs include time history of precipi-
tation, meteorological data, characteristics of
land surface including management practices,
vegetation cover, and terrain, conversations
and structural management practices within
sub-basins, chemical characteristics of pol-
lutants, stream channel characteristics, and
point source impacts such as reservoirs and
ponds. The SWRRB-WQ also requires input
parameters that describe the entire drainage
basin (e.g., total drainage area, basin slope,
and field capacity), pesticide parameters (e.g.,
soil partition coefficient, wash-off fraction,
soil biological half-life, and water solubility),
and sub-basin characteristics (e.g., slope, area,
curve number, and type of vegetation cover).
The hardware and software requirements
for implementing the SWRRB-WQ model are
fairly standard. Depending on the area of the
watershed and the degree of variability in hy-
drologic (e.g., ponds, gullies, and reservoirs)
and landscape features, the model can be ex-
pected to run efficiently on standard desktop
computers operating under the Windows en-
vironment.
Several applications of SWRRB-WQ eval-
uate the hydrology and water quality of com-
plex, large rural watersheds and basins and
are reported in the literature. For example, the
National Oceanic and Atmospheric Adminis-
tration used SWRRB-WQ to estimate loading
of nonpoint pollutants from rural basins in all
coastal counties in the United States (Sing-
er, et al. 1988). In this application, disparate
data from the National Weather Service sta-
tions, Natural Resource Conservation Service
(NRCS) Soils 5 database, the U.S. Geological
Survey's digital land use land cover data, and
other watershed parameters were used with the
model to provide simulations of water qual-
ity variables for cropland, forest, and range-
land in about 770 watersheds that comprise
the Gulf Coast, eastern, and western coastal
zones of the United States. In another appli-
cation, Arnold et al. (1987) predicted the ef-
fects of urbanization on watershed water yield
and reservoir sedimentation. As a component
of the HUMUS (Hydrologic Unit Model of
the United States) project, the SWRRB-WQ
model was integrated with EPIC and ROTO
(Arnold, et al. 1995) to provide a tool for the
1997 Resource Conservation Assessment of
the NRCS. Lastly, a Windows interface to
enhance the use of the model was developed
by the Office of Science and Technology of
the U.S. EPA, to assist regional planning ju-
risdictions in developing the total maximum
daily loads for agricultural watersheds. This
can be found at http://www.epa.gov/docs/
SWRRB WINDOWS/metadata.txt.html.
LIMITATIONS AND
MODEL VALIDATION
Mathematical models of ecological sys-
tems provide a simplified, approximate
representation of real-world processes and
phenomena. Indeed, researchers describe
models as "metaphors for reality" or "delib-
erately simplified construct of nature erected
for purposes of understanding a system or
phenomena" (Batchelor 1994). Bear (1979)
defines a model as: "a simplified version of
a real investigated system that approximately
simulates the latter's excitation-response relations
that are relevant to the considered problems."
Application of models to ecological problems
25
-------
^
^
mod eling
uncertainties
Water quality
standards
.
Determine
purpose exi
tence
FIGURE 1: MODEL QUALITY ASSURANCE COMPONENTS
requires well-designed protocols for model
reliability assessment and quality assurance,
including model validation.
Mathematical models are routinely used in
most disciplines and fields related to earth and
environmental sciences. Their use in problem
solving and decision-making is increasing.
Examples abound in many application areas
on the potential benefits of modeling. How-
ever, there is an area of concern to developers
and users of these models as well as the deci-
sion makers using information derived from
the output of the models. Indeed, the ability
of models to replicate real-world processes
and system responses is greatly influenced
by (i) errors in the underlying theory upon
which the model is based, (ii) uncertainty
in the input parameters, and (iii) unpredict-
ability of the system's phenomena. These
factors not only affect the integrity of model
outputs, but also the decisions that these out-
puts support. Because mathematical models
are increasingly relied upon in environmental
decision-making, it has become imperative to
document their reliability. In addition, mod-
els used to describe earth system processes
are becoming increasingly complex, often
involving multiple media, multiple pathways
and widely varying endpoints. This complex-
ity could lead to errors and uncertainty in the
predicted endpoints and outcomes, making
it increasingly necessary to develop meth-
odologies to convey critical uncertainties in
environmental models. Techniques and ap-
proaches to convey errors and uncertainties
in mathematical models fall under the do-
main of quality assurance and quality control
(QA/QC). In modeling, components of a QA/
QC protocol often include pre- and post-au-
dit analyses that involve model verification,
sensitivity analysis, model calibration and
validation, and the assessment of model un-
certainty (Figure 1).
26
-------
In hydrologic and water quality modeling, a
wide range of techniques are used to establish
the veracity and reliability of environmental
models. These include model verification, mod-
el sensitivity analysis, model calibration, model
validation, and model uncertainty analysis.
MODEL VERIFICATION
Model verification constitutes the process
of assessing the reliability of the modeling
computer code in generating both accurate
and "numerically stable" outputs that rep-
resent the conceptualized physical system.
Often compared to or confused with model
validation, verification of a model generally
involves comparing the results of the numeri-
cal solution to those obtained using analytical
or "closed-form" techniques. Through model
verification, illogical statements in the com-
puter code or incorrect assumptions that re-
quire significant model modifications can be
identified and corrected. In the use of hydro-
dynamic models, for example, it is desirable
that the computational scheme (e.g., numeri-
cal finite difference or finite element) be free
of numerical dispersion due to the choice of
input parameters for the advection component
of flow. A model verification process assures
that the numerical results are reasonably cor-
rect and matches prior specifications and as-
sumptions.
SENSITIVITY ANALYSIS
For mathematical models, sensitivity anal-
ysis is required to help identify key input
parameters and predictions errors. The aim of
sensitivity analysis, in general is to estimate
the rate of change in the predicted model
output with respect to changes in the model
inputs. Such information is important for: (i)
assessing the range and limits of applicability
of the model, (ii) determining parameters for
which it is important to have highly accurate
values, and, (iii) understanding the behavior
of the physical system being modeled. The
choice of method of sensitivity analysis de-
pends largely on the sensitivity measure em-
ployed, the desired accuracy in the estimates of
the sensitivity measure, and the computational
demands and costs involved.
Methods of sensitivity analysis can be
broadly divided into three main categories:
(i) variations of parameters or model formula-
tions in which the models is run for different
combinations of input parameters of concern,
or a straightforward change is made to the
model structure; (ii) domain-wide sensitivity
analysis involving the evaluation of the sys-
tem behavior response over the entire range
of parameter variations; and (iii) local sen-
sitivity analysis which focuses on estimates
of model sensitivity to input and parameter
variation in the vicinity of a point. One wide-
ly used method of sensitivity analysis is the
normalized gradient technique. For a math-
ematical model of the form:
F(u,k) = 0
where k is a set of m parameters, and u is a
vector of n output variables. Thus the normal-
ized gradient sensitivity analysis takes the
form:
Other techniques include the normalized
response and the local gradient approxima-
tion represented mathematically as:
D =du /u(k)
i i iv '
du = [S;j] 5k; S;j = du/dk
in which S and D are sensitivity coefficients.
27
-------
MODEL CALIBRATION
MODEL VALIDATION
This process of model QA involves adjust-
ing model input parameters until the system
output and the model output (predicted values)
show an acceptable level of agreement. Typi-
cally, this level of agreement is measured
using an objective function (or some aggre-
gation function of the model residuals), usu-
ally supported by visual inspection of the
computed or predicted time series. Thus, the
modeling structure and parameter combina-
tion producing the best performance is com-
monly assumed to represent the conceptual-
ized physical system.
Fundamentally, model calibration is an in-
teractive process involving: (i) simulations
using parameter sets from the search space
to document model performance; (ii) deter-
mination of parameter sets that are likely to
perform better than those used in the previ-
ous simulations, and model simulation us-
ing the new or revised parameter sets; and,
(iii) repetition of step (ii) until a satisfactory
measure of performance is obtained or until
further improvements are negligible. During
calibration, model performance is quantified
by an objective function and coefficients. Some
commonly used coefficients include the coef-
ficient of determination, modeling error or
bias, and the root mean square error. Graphi-
cal plots such as hydrographs (in hydrody-
namic models) and scatter plots can be used.
The three steps for model calibration could
be undertaken manually or automatically us-
ing some form of optimization.
An inherent issue in many modeling appli-
cations is what constitutes an acceptable bias
or difference between model predictions and
corresponding observations in the real-world.
Model reliability and quality assurance can
also be assessed through a validation process.
Model validation is probably one technique
of model performance assessment that has re-
ceived the most attention in the modeling lit-
erature. Differing opinions exist as to the def-
inition of model validation or what constitutes
a model validation process. For example, the
U.S. Department of Energy defines validation
as the determination "that the model indeed
reflects the behavior of the real world." The
International Atomic Energy Agency (IAEA)
defines a validated model as one that provides
"a good representation of the actual process
occurring in a real (physical) system". Fur-
thermore, the IAEA, in its Radioactive Waste
Management Glossary provides yet another
definition of model validation as "a process
carried out by comparison of model predic-
tions with independent field observations
and experimental measurements". Wigman
(1972) defines validation as "the process of
discriminating between sets of postulates by
reference to fresh data not used in setting up,
fitting, and a calibration process". From these
definitions, the purposes of model validation
are to: (i) objectively assess the performance
and trustworthiness of the model, (ii) charac-
terize the effects of parameter variability and
parameter uncertainty on model outputs, and,
(iii) evaluate the results of model simulations
without human bias and interpretation.
28
-------
In general, model validation is the process
which determines the accuracy of a model by
comparing model outputs to data measured
from the natural world that the model is simu-
lating. The initial conditions for the model are
matched to those at the time of collection of
the field (observed) data. From a collection of
those comparisons, the overall model perfor-
mance is analyzed, evaluated, and documented.
Furthermore, model validation involves iden-
tifying those factors that contribute to differ-
ences between model predictions and field
observations.
In model validation, numerous attempts
have been made to develop practical and
quantitative performance measures to es-
tablish whether to accept, modify, or refute
a model. For example, Whitmore (1991) sug-
gests a combination of graphical and statisti-
cal techniques for assessing model reliability.
The discrepancy between model predictions
and field observations, whether random or
systematic, can be classified as space-time-
independent residuals. The sum of the squares
of the residual error is partitioned into two
other sums of squares: one derived from ran-
dom variations and the other due to systemat-
ic variation or mismatch between predictions
and confirming real-world observations. The
performance criteria for assessing model reli-
ability based on replicated field experiments, as
summarized by Whitman (1991) are as follows:
LOFIT = lad/ = Ifl. (y. - x/
in which RSS is the residual sum of squares;
SSE is the sum of squares of the error; LOFIT
(or lack of fit) is sum of squares attributed to
the lack of fit, an indication of model bias;
d = deviation or residual error (y - x); d =
1J VJjj j75 j
the mean deviation (y -x); y = mean of the
\Jj j75 Jj
measurements in they'th experiment; and, x
= mean of the predictions of the y'th experi-
ment. League and Green (1991) and Green
and Stephenson (1986) propose a combina-
tion of approaches for assessing model valid-
ity. They suggested the use of goodness-of-fit
tests that include: maximum error (ME), not
mean square error (RMSE), modeling effi-
ciency (EF), coefficient of determination (CD)
and coefficient of residual mass (CRM). The
expressions for three performance measures
are as follows:
= max_xi-yiJoralll
RMSE = 100/y[ £(* - y^/N]0-5
EF = [ Ify. - y)2 - Z(Xi - y/] / QXy. - y)2}
= E(yi-y)2]/(I(xi-y)2}
where N is the number of pair of model-pre-
dicted (x ) and field observed (y ) values, and y
is the mean value of the observations. For the
models to be considered fully validated and
representative of real-world physical system,
values of ME, RMSE, EF, CD, and CRM must
be equal to 0, 0, 1.0, 1.0, and 0, respectively.
MODEL UNCERTAINTY
ANALYSIS
Analysis of model errors and uncertainty
is rapidly becoming an acceptable practice
in environmental modeling. It is essential
for making reliable predictions of complex
phenomena. Well informed and technically
defensible environmental policy decisions
based on model simulations demand that we
identify and document: the significance of the
29
-------
inherent variability of the physical system, the
impact of the approximations and simplifica-
tions made in formulating the model problem,
the consequences of simulation errors, the
sensitivity of the predictions to limited un-
derstanding of governing processes and sys-
tem dynamics, and, the probabilistic implica-
tions of inherent stochastic effects that exist
in most physical systems. A systematic anal-
ysis of model uncertainty provides valuable
insights into the level of confidence in model
predictions and assists in assessing how the
model predictions should be weighed in any
decision making process. Furthermore, mod-
el uncertainty analysis can suggest to model
users reasons for strengthening or weakening
their belief in the model results.
Increasingly, the reliability of mathematical
models requires that we gain a better under-
standing of the simplifying assumptions in
the model, the influence of potential model-
ing error and uncertainties on the response of
the model, and the sources of the modeling
uncertainty. A number of sources of model
uncertainty have been reported in the litera-
ture, including uncertainties due to model
structure, model comprehensibility, choice of
boundary conditions, and model spatial and
temporal resolution.
Uncertainty from modeling structure arises
when there are alternative sets of scientific or
technical assumptions for developing the mod-
el. Thus, when a competing model is used and
the results are compared; similar conclusion
could provide some level of confidence with
the model. If, however, an alternate model
formulation provides different conclusions,
then further evaluation of model structure
may be necessary.
In the development of mathematical mod-
els, processes that describe the dynamics of
the physical system are simplified for pur-
poses of tractability. Examples of model un-
certainty due to comprehensibility include
assumptions of nonlinearity, compressibil-
ity, unidirectional flow, or the conversion of
nonlinear process to linear processes to allow
simplified analytical solutions to be obtained.
Uncertainty of predictions from simplified
models can be characterized by comparing
predictions to those obtained from more in-
clusive and detailed models.
Mathematical models that are validated for
a section of the input space could be com-
pletely inappropriate when used for decisions
in other regions of the parameter space. For
example, in predicting components of the hy-
drologic cycle, models that are calibrated for
certain precipitation events may not be appro-
priately verified if similar events are applied
during the validation process.
Model uncertainty can arise from the se-
lection of the spatial and temporal resolution.
There is a trade-off between model prediction
accuracy and the computation time. Trade-off
also exists between the choice of the spatial
resolution (e.g., lumped or distributed) and
the validity of the governing equations. Quite
often, coarse spatial resolution introduces ap-
proximations and uncertainties in the model
results due to aggregation. However, a finer
resolution, in some situations, does not necessarily
result in predictions that are more accurate.
A number of techniques have been utilized
to attempt to represent and/or reduce un-
certainty in mathematical modeling. Some
widely used techniques involve: (i) classical
set theory, in which uncertainty is expressed
30
-------
by sets of mutually exclusive alternatives in
situation where one alternative is desired, (ii)
probability theory, where model uncertainty,
notwithstanding its origin, is expressed in
forms of a measure or subsets of a universal
set of alternatives; and, (iii) fuzzy set theory,
which unlike the classical set theory, is capa-
ble of incorporating vagueness that emerges
from imprecision of definitions rather than
from non-specificity. Modeling uncertain-
ty using fuzzy set theory is expressed as a
degree rather than an affirmation.
CHOOSING A
SUITABLE MODEL
Models are increasingly used in many as-
pects of environmental management and
planning, ranging from evaluating changes
in watershed management to extending data-
sets to areas with little or no measurement,
and to assessing impact of external influ-
ences such as climate change. While there
are many mathematical models of hydrology
and water quality in use, the skill in selecting
the right model for an application and balanc-
ing the data requirements against the cost of
model implementation is an art as well as sci-
ence. For a critical and rigorous assessment of
model suitability, users need to ask the follow-
ing questions:
THE MODELING PROCESS
(a) Hydrology:
• Does the model have a built-in stochastic
climate generator for constituting syn-
thetic climate data if measurements are
not available?
• Does the model compute overland flow
(runoff) using a processes-oriented ap-
proach or physically-based approach
(e.g. SCS Curve Number technique)?
• Does the model compute flow in a stream
channel and route this downstream?
• What method of flow routing is used?
• Does the model account for flow into and
out of artificial impoundments (e.g. lakes
and wetlands)?
• Does the model explicitly incorporate
flow into and out of marshes and ponds?
• Does the model contain specialized
functions to deal with outflow or outfalls
into estuaries, tidal flows, and saltwater
intrusions?
• How does the model deal with irrigation
water?
(b) Sedimentation:
• What technique is used in the model
to estimate soil erosion by water? Is
it USLE, MUSLE, or RUSLE?
• How is ephemeral gully erosion
simulated?
• How is streambed and bank erosion
predicted?
31
-------
What physically based or process-oriented
approach is adopted by the model to pre-
dict sediment detachment, transport, and
deposition?
In estimating sediment yield, is an
expression for the delivery ratio stated
explicitly? If so, how does the model
route sediment to the domain outlet
point?
Are there any provisions to handle other
nutrient sources (e.g. organic wastes
from municipal sludge and food process-
ing residues or atmospheric inputs)?
Are nitrogen and phosphorus predicted
as total amounts or concentrations?
MODEL PARAMETERS
(c) Nutrient export
• How does the model handle the fate and
transport of nutrients in the landscape?
• What forms of nutrients does the model
handle? Nitrogen or phosphorus?
• Does the model contain components that
predict the fate and movement of nitro-
gen in surface runoff?
• Does the model handle inorganic forms
of nutrients?
• What forms of nitrogen does the model
predict?
• Does the model include manure manage-
ment and nitrogen transformation?
• How does the model handle the fate and
transport of phosphorus in runoff?
• Are there component equations to differ-
entiate between dissolved and particulate
matter?
• What forms of phosphorus does the
model predict?
• Does the model handle subsurface leaching
losses of both nitrogen and phosphorus?
(a) Meteorological:
• Does the model require breakpoint,
hourly, daily, or monthly values of
precipitation?
• Does the model include a climate generator
for constituting climate data where mea-
surements are unavailable or inadequate?
• Does the model require air temperature
for each time-step of the modeled period?
• Does the model require wind speed,
relative humidity, and solar radiation
data for each time-step?
• Is precipitation data considered spatially
distributed or lumped?
• Does the model require information on
percent cloud cover, sunshine hours, or
other related surface air data?
(b) Landscape:
• What topographic information is re-
quired by the model (e.g. elevation, slope,
aspect, drainage network)?
• What soil properties and characteristics
does the model require?
32
-------
Does the model require users to specify
the type of land cover, land-use and land
management?
Are the landscape-related parameters re-
quired by the model spatially distributed?
What management factors (agricultural,
urban, and forest) are considered in the
model?
In terms of tillage practices, are there
any component of the model that incor-
porates the different effects of these
practices in hydrology and water quality?
Does the model handle crop rotation or
changes in crop growth parameters with
respect to time and location?
Are irrigation practices (application rate,
type of irrigation system) and chemigation
handled by the model?
Does the model accept data on artificial
drainage of the subsurface soil and the
associated effects on hydrology and wa-
ter quality?
What conservation practices can be ad-
equately incorporated into the model?
Does the model incorporate information
on nutrient and pesticide management?
(c) Model Output Parameters
• At what time-step does the model
produce flow and water quality results?
• Does the model incorporate information
on nutrient and pesticide management?
Does the model lump output results with
respect to watershed area?
In what time-space format does the
model generate the outputs?
For the spatially distributed models,
can users examine outputs at specified
locations within the landscape?
For water quantity and quality variables,
can users track the source of the
contaminant?
Can users evaluate or assess the effects
of model and parameter uncertainty on
the predicted outputs?
Is the output generated in a tabular
or graphical format?
What output data format is provided
in the model?
(d) Space and Time Scale
• Does the model simulate discrete events
or can it utilize long-term continuous
data?
• Is the model designed for the plot,
field, whole-farm, or watershed-scale?
• Does the model allow the use of
GIS in extending its spatial scale?
• Was the model developed for a specific
geographical area?
• Is the modeling technology applicable
on a national basis?
33
-------
REFERENCES
Arnold, J. G., M. D. Birchet, J. R. Williams,
W. F. Smith, and H. N. McGill. 1987. Mod-
eling the effects of urbanization on basin
water yield and reservoir sedimentation.
Water Resources Bulletin. 23:1101-07.
Arnold, J. G., J. R. Williams, A. D. Nicks,
and N. B. Sammons, 1990, SWRRBWQ,
A Basin Scale Simulation Model for Soil
and Water Resources Management.
College Station, TX Texas A&M Press.
Arnold, J. G., J. R. Williams, R. Srinivasan,
K. W King, and R. H. Griggs. 1995.
SWAT—Soil and Water Assessment Tool:
Draft
Users Manual. Temple, TX: USDA-ARS.
Arnold, J.G., R. Srinivasin, R.S. Muttiah, and
J. R. Williams. 1998. Large Area Hydrologic
Modeling and Assessment: Part I. Model
Development. Journal of American Water
Resources Association. 34:73-89.
Ball, J. E., M. J. White, G. de R. Innes, and
L. Chen. 1993. Application of HSPF on the
Upper Nepean Catchment. Proceedings of
Hydrology and Water Resources Symposium.
Newcastle, New South Wales, Australia.
pp. 343-48.
Batchelor, P. 1994. Models as metaphors: the
role of modeling in pollution prevention.
Waste Management. 14: 223-251.
Bear, I, 1979. Use of models in decision-
making. Transport and Reactive Processes
in Aquifers. T. Draco and F. Stanffer, (eds.)
Balkema, Rotherdam. pp. 3-9.
Bicknell, B. R., J. C. Imhoff, J. L. Kittle, A.
S. Donigian, andR. C. Johanson, 1993.
Hydrological Simulation Program - FOR-
TRAN (HSPF): User's manual for release
10.0. Environmental Research Laboratory,
U S. Environmental Protection Agency,
Athens, GA. PA 600/3-84-066.
Bingner, R. L. 1996. Runoff simulated from
Goodwin Creek Watershed using SWAT.
Transactions of the ASAE. 39: 85-90.
DeVries, J. J. and T. V. Hromadka. 1993.
Streamflow. Handbook of Hydrology, D. R.
Maidment, ed. New York: McGraw-Hill.
pp. 21.1-21.39.
Donigian, A. S., Jr. and H. H. Davis, Jr. 1978.
User's Manual for Agricultural Runoff
Management (ARM) Model. EPA-600/
3-78-080. U.S. Environmental Protection
Agency, Athens, GA
Donigian, A. S. Jr. and N. H. Crawford, 1979.
User's Manual for the Nonpoint Source
(NFS)Model. U.S. Environmental
Protection Agency, Athens, GA
Donigian, A. S., Jr., B. R. Bicknell, L. C.
Linker, J. Hannawald, C. Chang, and R.
Reynolds. 1990. Chesapeake Bay Program
Watershed Model Application to Calculate
Bay Nutrient Loadings: Preliminary Phase
I Findings and Recommendations. Pre-
pared by Aqua Terra Consultants for U.S.
EPA Chesapeake Bay Program, Annapolis,
MD.
34
-------
Donigian, A. S. Jr., B. R. Bicknell, A. S.
Patwardhan, L. C. Linker, D. Y. Alegre, C.
H. Chang and R. Reynolds. 1991a. Chesa-
peake Bay Program Watershed Model Ap-
plication to Calculate Bay Nutrient Load-
ings. Prepared by Aqua Terra Consultants
for U.S. EPA Chesapeake Bay Program,
Annapolis, MD.
Donigian, A. S., Jr. and W. C. Huber. 1991b.
Modeling ofNonpoint Source Water
Quality in Urban and Non-Urban Areas.
EPA/600/3-91/039. U.S. Environmental
Protection Agency, Environmental
Research Laboratory, Athens, GA.
Donigian, A. S. and A. S. Patwardhan. 1992.
Modeling Nutrient Loadings from Crop-
lands in the Chesapeake Bay Watershed.
Proceedings of Water Resources sessions
at Water Forum '92. Baltimore, MD. pp.
817-22.
Donigian, A. S. Jr., B. R. Bicknell, and J. C.
Imhoff. 1995a. Hydrological Simulation
Program - FORTRAN (HSPF). Computer
Models of Water shed Hydrology. V. P.
Singh, ed. Littleton, CO: Water Resources
Publications, pp. 395-442.
Donigian, A. S., J. C. Imhoff, and R. B. Am-
brose. 1995b. Modeling Watershed Water
Quality. Environmental Hydrology. V. P.
Singh, ed. The Netherlands: Kluwer
Academic Publishers, pp. 377-426.
Engel, B. A. and J. G. Arnold. 1991. Agri-
cultural nonpoint source pollution control
using spatial decision support system.
Internal report. Department of Agricultural
Engineering. West Lafayette, IN: Purdue
University.
Foster, G.R., L.J. Lane, J.D. Nowlin, J. M.
Laflen, and R.A. Young. 1981. Estimating
erosion and sediment yield on field size areas.
Transactions. oftheASAE. 24:1253-62.
Foster, G. R., K. G. Renard, D. C. Yoder,
D. K. McCool, and G. A. Weesies. 1996.
RUSLE User's Guide. Ankey, IA: Soil and
Water Conservation Society. 173 pp
Green, I.R A. and D. Stephenson. 1986. Cri-
teria for comparison of single event mod-
els. Hydrological Sciences. 31: 395-411.
Haith, D. A. and L. L. Shoemaker. 1987.
Generalized watershed loading functions
for stream flow nutrients. Water Resources
Bulletin. 107:121-37.
Haith, D. A., R. Mandel, and R. S. Wu. 1992.
GWLF - Generalized Watershed Loading
Functions, Version 2.0—User's manual.
Department of Agricultural Engineering,
Ithaca, NY: Cornell University.
Hasssanizadeh, S., and J. Carrera. 1992. Edi-
torial on model validation and verification.
Advances in Water Resources. 15:1-3.
Hydrocomp, Inc. 1977 Hydrocomp Water
Quality Operations Manual. Palo Alto,
CA: Hydrocomp, Inc.
International Atomic Energy Agency. 1982.
Radioactive waste management glossary.
Report IAEA-TECDOC-264. The Inter-
national Atomic Energy Agency. Vienna,
Austria.
35
-------
Johansen, R. C, J. C. Imhoff, J. L. Kittle, Jr.
and A. S. Donigian. 1984. Hydrological
Simulation Program—FORTRAN (HSPF):
Users Manual for Release 8.0. EPA-600/3-
84-066, Environmental Research Laboratory,
U. S. Environmental Protection Agency,
Athens, GA. 30613.
Knisel, W. G.1980. CREAMS: A Field-Scale
Model for Chemicals, Runoff, and Erosion
from Agricultural Management Systems.
U.S. Department of Agriculture, Conserva-
tion Research Report No. 26. U.S. Depart-
ment of Agriculture, Washington, DC.
Leonard, R. A., W. G. Knisel, and D. A. Still.
1987. GLEAMS: Groundwater Loading
Effects of Agricultural Management Sys-
tems. Transactions of the ASAE. 30:1403-
18.
League, K. and Green R.E. 1991. Statistical
and graphical methods for evaluating solute
transport models. Journal of Contaminate
Hydrology. 7:51-73.
Lumb, A. M., J. L. Kittle, and K. M. Flynn.
1990. User's manual for ANNIE, A com-
puter program for interactive hydrologic
analyses and data management. Water
Resources Investigations Report 89-4080.
Reston, VA: U.S. Geological Survey.
Lumb, A. M., and J. L. Kittle. 1993. Expert
System for calibration and application of
watershed models. Proceedings of the Fed-
eral Interagency Workshop on Hydrologic
Modeling Demands for the 90s. Water
Resources Investigation Report 93-4018.
Fort Collins, CO: US. Geological Survey.
Penman, H.L. 1948. Natural evaporation
from open water, bare soil and grass.
Proceedings of the Royal Society, London.
A193: 120-146.
Manguerra, H. B. and B. A. Engel. 1998.
Hydrologic parameterization of watershed
for runoff prediction using SWAT. Journal
of the American Water Resources Associa-
tion. 34-5:1149-62.
McElroy, A. D., S. W. Chiu, J. W. Nabgen,
A. Aleti, and R. W. Bennett. 1976. Loading
functions for assessment of water pollution
for non-point sources. EPA 600/2-76/151.
U. S. Environmental Protection Agency,
Washington, DC.
Mills, W. B. 1985. Water quality assessment:
A screening procedure for toxic and con-
ventional pollutants in surface and ground
water. EPA/600/6-85/002a. Environmental
Research Laboratory, U.S. Environmental
Protection Agency, Athens, GA.
Moore, L. W, C. Y. Chew, R. H. Smith, and
S. Sahoo. 1992. Modeling of best manage-
ment practices on North Reelfoot Creek,
Tennessee. Water Environment Research.
64:241-47.
Novotny, V and H. Olem. 1994. Water Quality:
Prevention, Identification, and Management
of Diffuse Pollution. New York, NY: van
Nostrand Reinhold.
Omnicron Associates. 1990. Nonpoint
Pollution Source Model for Analysis
and Planning—User's Manual. Portland,
OR: Omnicron Associates.
Preston, S. D. and J. W. Brakebill. 1999.
Application of Spatially Referenced Re-
gression Modeling for the Evaluation of
Total Nitrogen Loading in the Chesapeake
Bay Watershed. Water Resources Investiga-
tions Report 99-4054. Baltimore, MD: US.
Geological Survey.
36
-------
Renard, K.G., G.R. Foster. G.A. Weesies, D.
K. McCool. And D.C. Yoder. 1997. Pre-
dicting Soil Erosion by Water: A Guide to
Conservation Planning with the Revised
Soil Loss Equation (RUSLE). Agricultural
Handbook No. 703. Washington, DC: U.S.
Government Printing Office.
Rosenthal, W D., R. Srinivasan, and J. G. Ar-
nold. 1995. Alternative river management
using a linked GIS-hydrology model.
Transactions of the ASAE. 38:783-90.
Singer, M. P., F. D. Arnold, R. H. Cole, J. G.
Arnold, and J. R. Williams. 1988. Use of
SWRRB computer model for the National
Pollutant Discharge Inventory. W. L. Lyle
and T. J. Hoban, eds. Proceeding of the
Symposium on Coastal Water Resources.
pp. 119-31.
Smith, R. A., G. E. Schwarz, and R. B. Alexan-
der. December, 1997 Regional interpretation
of water-quality monitoring data. Water
Resources Research. 33(12):2781-98.
Srinivasan, R. and J. G. Arnold. 1994.
Integrating a basin-scale water quality
model with GIS. Water Resources
Bulletin. 30:441-52.
Srinivasan, R., J. G. Arnold, R. S. Muttiah,
and J. R. Williams. 1998. Large area hy-
drologic modeling and assessment, Part I:
model development. Journal of the American
Water Resources Association. 34:73-89
Tim, U. S. 1996a. Emerging technologies
for hydrologic and water quality modeling
research. Transactions of the ASAE.
39:465-76.
Tim, U. S. 1996b. Coupling vadose zone
models with GIS: emerging trends and
potential bottlenecks. Journal of Environ-
mental Quality. 25:535-44.
Whitmore, A.P. 1991. A method for assess-
ing the goodness of computer simulation
of soil processes. Journal of Soil Science
42:289-99.
Wigman, M. R. 1972. The fitting, calibration,
and validation of simulation models.
Simulation. 18:188-92
Williams, J. R. and H. D. Berndt, 1977
Sediment yield prediction based on
watershed hydrology. Transactions
oftheASAE.20:lWO-04.
Williams, J. R., Jones, C. A., Kiniry, J. R. and
Spanel, D. A. 1989. The EPIC crop growth
model. Transactions of the ASAE. 32:497-511.
Young, R. A., M. A. Otterby , and A. Roos.
1982. A technique for evaluation feedlot
pollution potential. Journal of Soil and
Water Conservation. 37: 21-23.
Young, R., C. A. Onstad, D. D. Bosch, and
W. P. Anderson. 1987. AGNPS: Agricultur-
al Non-Point Source Pollution Model: A
watershed analysis tool. USDA, Agri-
cultural Research Service. Conservation
Research Report 35. 77 pp.
37
------- |