Transactions of the ASABE
Vol. 49(4): 967−986 E 2006 American Society of Agricultural and Biological Engineers ISSN 0001−2351 967
SEDIMENT AND NUTRIENT MODELING FOR TMDL
D
EVELOPMENT AND IMPLEMENTATION
T
M
D
L
D. K. Borah, G. Yagow, A. Saleh, P. L. Barnes, W. Rosenthal, E. C. Krug, L. M. Hauck
ABSTRACT. At present, there are over 34,000 impaired waters and over 58,000 associated impairments officially listed in the
U.S. Nutrients and sediment are two of the most common pollutants included in the list. States are required to identify and
list those waters within their boundaries that are not meeting standards, to prioritize them, and to develop Total Maximum
Daily Loads (TMDLs) for the pollutants of concern. Models are used to support development of TMDLs, typically to estimate
source loading inputs, evaluate receiving water quality, and determine source load allocations so that receiving water quality
standards are met. Numerous models are available today, and selection of the most suitable model for a specific TMDL project
can be daunting. This article presents a critical review of models simulating sediment and nutrients in watersheds and
receiving waters that have potential for use with TMDL development and implementation. The water quality models
discussed, especially those with sediment and/or nutrient components, include loading models (GWLF and PLOAD),
receiving water models (AQUATOX, BATHTUB, CE-QUAL-W2, QUAL2E, and QUAL2K), and watershed models having
both loading and receiving components (AGNPS, AnnAGNPS, CASC2D/GSSHA, DWSM, HSPF, KINEROS2, LSPC, MIKE
SHE, and SWAT). Additional models mentioned include another receiving water quality model (WASP), watershed models
(ANSWERS storm event, ANSWERS continuous, PRMS storm event, SWMM, and WEPP), and BMP models (APEX, REMM,
and VFSMOD). Model sources, structures, and procedures for simulating hydrology, sediment, and nutrients are briefly
described for the reviewed models along with an assessment of their strengths, limitations, robustness, and potentials for using
in sediment and/or nutrient TMDLs. Applications of AGNPS, APEX, BATHTUB, CE-QUAL-W2, GWLF, and SWAT in TMDL
developments are presented. Applications of some of the other models (DWSM, GSSHA, and KINEROS2) relevant to TMDL
studies are also presented. The models proved to be useful; however, they require a learning process. Simple models are easy
to use but have limitations; comprehensive models are labor and data intensive but offer extensive analysis tools. Finally,
recommendations are offered for advancing the sediment and nutrient modeling technologies as applied to TMDL
development and implementation. Advances could be made towards: making the best use of existing models, enhancing the
existing models, combining strengths of existing models, developing new models or supplemental components with physically
based robust routines, numerous field applications, sensitivity analyses, full documentation, and rigorous education and training.
Keywords. BMP, Hydrology, Loading, Modeling, Nutrients, Receiving water, Sediment, TMDL, Water quality, Watershed.
ection 303(d) of the Clean Water Act and the U.S.
Environmental Protection Agency’s (USEPA) Wa-
ter Quality Planning and Management Regulations
(40 Code of Federal Regulations Part 130) require
states to identify and list those waters within their boundaries
Submitted for review in February 2006 as manuscript number SW
6355; approved for publication by the Soil & Water Division of ASABE in
April 2006.
The authors are Deva K. Borah, ASABE Member Engineer, Principal
Modeler, Borah Hydro−Environmental Modeling, Champaign, Illinois;
Gene Yagow, ASABE Member Engineer, Research Scientist, Virginia
Tech, Blacksburg, Virginia; Ali Saleh, ASABE Member Engineer,
Research Scientist, Tarleton State University, Stephenville, Texas; Philip L.
Barnes, ASABE Member Engineer, Research Engineer, Kansas State
University, Manhattan, Kansas; Wesley Rosenthal, ASABE Member
Engineer, Associate Professor, Texas A&M University, Texas Agricultural
Experiment Station, Temple, Texas; Edward C. Krug, Biogeochemist,
Borah Hydro−Environmental Modeling, Champaign, Illinois; and Larry
M. Hauck, Deputy Director, Texas Institute for Applied Environmental
Research (TIAER), Tarleton State University, Stephenville, Texas.
Corresponding author: Deva K. Borah, Illinois State Water Survey, 2204
Griffith Dr., Champaign, IL 61820; phone: 217-244-8856; fax:
217-333-2304; e-mail: borah@uiuc.edu.
that are water quality limited, to prioritize them, and to devel-
op Total Maximum Daily Loads (TMDLs) for the pollutants
of concern. Based on most recent state 303(d) lists, there are
over 34,000 impaired waters and over 58,000 associated im-
pairments in the U.S. (USEPA, 2006). Metals, pathogens, nu-
trients, and sediment are the most common pollutants
included in the list. Nutrients are the third leading listed cause
of water quality impairments, representing 9% of all im-
paired segments. Sediment ranks fourth, representing 8% of
all impaired water bodies in the U.S. Sediment is listed for its
physical impairment to aquatic habitats, whereas nutrients
are listed for impairment by organic enrichment, i.e., through
excessive algal growth, which can lead to depletion of dis-
solved oxygen, among other problems. Additionally, several
nitrogen (N) species have detrimental impacts on water qual-
ity: ammonia is toxic to fish, and high levels of nitrates cause
methemoglobenemia in infants.
A TMDL is the allowable load (amount) of any pollutant
that a stream can receive and still meet applicable water
quality standards and support its designated use(s). A TMDL
is comprised of loads from permitted point, diffused
(nonpoint), and natural background sources (Shoemaker et
al., 2005). TMDLs must be calculated with seasonal
S
Copyright by the American Society of Agricultural and Biological Engineers. Borah, D. K.; Yagow, G.; Saleh, A.;
Barnes, P. L.; Rosenthal, W.; Krug, E. C.; Hauck, L. M., "Sediment and nutrient modeling for TMDL development
and implementation," Transactions of the ASABE. Vol. 49(4): 967-986. (doi: 10.13031/2013.21742) @2006
968 TRANSACTIONS OF THE ASABE
variations, account for critical conditions related to stream
flow, loading, and water quality parameters, and incorporate
a margin of safety, which represents uncertainty in the
modeling process (see companion article Shirmohammadi et
al., 2006).
Models are used to support development of TMDLs,
typically to estimate source loading inputs, evaluate receiv-
ing water quality, and determine load allocation to sources so
that receiving water quality standards are met. The models
help users to understand the dynamics of physical watershed
systems that include sources of water and pollutant, and the
receiving waters such as lakes, rivers, estuaries, and coastal
areas. Models can also answer questions, such as: how do
inputs from human sources and land management activities
affect loadings to the receiving waters and their conditions,
and how should these inputs be changed to improve the
conditions? Development of models that can reliably repre-
sent the physical systems is challenging. Numerous models
are available today, and many of them have been used in
environmental and water quality management since the
1970s. However, not all models are appropriate for TMDLs.
A model must be capable of quantifying the potential
response of the selected endpoints to changes in source
pollution loadings. Physically based models are the best
suited models.
Shoemaker et al. (2005) conducted an extensive review of
existing models and described their various characteristics
and capabilities. Some of those models are currently being
used for TMDLs, and others have potentials for use in TMDL
development. Detailed descriptions of many of the models
may be found in Singh (1995) and Singh and Frevert (2002a,
2002b, 2006). Kalin and Hantush (2003) also conducted
reviews of many of the models, applied two of the watershed
models to an experimental watershed, and evaluated their
performances. Borah and Bera (2003) critically reviewed the
mathematical bases of eleven leading watershed-scale mod-
els and categorized the models by their capabilities, specifi-
cally identifying their computational efficiencies and
long-term continuous and storm event simulation capabili-
ties. In another study, Borah and Bera (2004) reviewed
twelve to eighteen applications of each of three watershed
models and discussed their performance abilities. Numerous
other applications of the models, especially a few leading
ones, may be found in the literature, some comparing relative
performances (e.g., Saleh and Du, 2004). Selection of the
most suitable model for a particular TMDL project is still a
challenging task.
The primary objectives of this article are to conduct a
critical review of models simulating sediment and nutrient
generations and transport (distribution) that have already
been used or have potential for use in TMDL development
and implementation; categorize the models into loading
models, receiving water models, and watershed models;
describe briefly their modeling procedures for simulating
hydrology (basic to sediment and nutrient simulations),
sediment, and nutrients along with an assessment of their
strengths, limitations, robustness, and potentials for use in
sediment and/or nutrient TMDLs; review applications of
some of the models in TMDL and related studies along with
an assessment of their performances, suitability, and short-
comings; and finally, discuss advancing hydrology, sedi-
ment, and nutrient modeling technologies with the mention
of a few ongoing efforts. Unlike the previous reviews
mentioned above, this review focuses on models for sediment
and nutrient TMDLs and their distinct categories, strengths,
limitations, robustness (computational efficiencies), and
other insightful information not commonly available to help
in the selection of the most suitable model for a sediment
and/or nutrient TMDL. The discussion on advancing model-
ing technology is another unique feature of this article.
This article is part of a special ASABE TMDL modeling
collection effort. The purpose and overview of this collection
is outlined in a companion article (Muñoz-Carpena et al.,
2006). Five other related topics are discussed in five
companion articles, including pathogens (Benham et al.,
2006), dissolved oxygen (DO) (Vellidis et al., 2006),
biological indicators (Yagow et al., 2006), uncertainty in
models (Shirmohammadi et al., 2006), and economics
(Bosch et al., 2006). The materials presented here are based
on information available in the literature and may not always
be from the original sources.
SEDIMENT AND NUTRIENT MODELS
The common model characteristics currently needed for
TMDL development include: (1) watershed-based, (2) allow
for continuous simulation, (3) consider both point and
nonpoint pollutants, (4) consider contributions from both
surface and groundwater, (5) consider pollutants in both
dissolved and particulate phases, where appropriate, and
(6) have the ability to represent the major important wa-
tershed and land use characteristics influencing the pollutant
of concern. In general, suitable models will still differ in their
degree of representation of various watershed processes,
their need for calibration, their ability to represent spatial and
temporal variability throughout the watershed, and in the
level of user support available.
USEPA has recognized the importance of modeling for
development of TMDLs and has supported the development
of a Geographic Information System (GIS) based user
interface called BASINS (Better Assessment Science Inte-
grating Point and Nonpoint Sources; USEPA, 2001). This
interface serves as a gateway for estimation of parameter
values and model input formatting for the following models:
AQUATOX, HSPF, PLOAD, QUAL2E, and SWAT. Howev-
er, because of considerable variability from state to state in
standards, data availability, and user preference, these
models may not always be the most appropriate for use in a
given impaired water body, and therefore are not the only
ones used and accepted by USEPA for TMDL development.
Some of the other models currently in use include AGNPS,
AnnAGNPS, BATHTUB, CE-QUAL-W2, GWLF, and
LSPC (an enhanced version of HSPF).
These models can be categorized into three general
classes for TMDL development: (1) loading models, (2) re-
ceiving water models, and (3) watershed models. GWLF and
PLOAD are loading models that estimate loadings of water,
sediment, and/or chemicals from a watershed outlet into a
water body. AQUATOX, BATHTUB, CE-QUAL-W2, and
QUAL2E are receiving water models that analyze water
quantity and/or quality in a receiving water body (stream,
impoundment, lake, estuary, etc.) in response to loadings
from its contributing watershed(s). Watershed models, such
as AGNPS, AnnAGNPS, HSPF, LSPC, and SWAT, have
capabilities of both the loading models and some capabilities
969Vol. 49(4): 967−986
of the receiving water models because they estimate water,
sediment, and chemical loadings from different parts of a
watershed into its streams and/or lakes and analyze water
quantities and/or qualities of the receiving waters as well.
Watershed models can be further subdivided into two
distinct categories: (1) long-term continuous models (simply
referred to as continuous watershed models), and (2) storm
event watershed models. AnnAGNPS, HSPF, LSPC, and
SWAT are continuous models, used for analyzing long-term
impacts of climate and hydrological changes and manage-
ment practices on water quantity and quality in a watershed.
AGNPS is a storm event model. Storm event models are used
to predict water quantities and qualities, and analyze impacts
of management practices, during and after individual storm
events, especially extreme storm events that may cause
flooding and move disproportionately large amounts of
sediment and nutrients. Storm event model results are used
in engineering designs of control or conservation structures
and would be most useful for evaluating appropriate best
management practices (BMPs) for TMDL implementation.
As discussed below, some models, such as CASC2D/GSSHA
and MIKE SHE, have the ability to perform both long-term
continuous and storm event simulations.
The above models and a few other promising ones for
TMDL are described and discussed here. Most of these
models are included in the extensive review by Shoemaker
et al. (2005), which includes general discussions on model-
ing, categorization of models based on their structures and
capabilities, and outlines of research needs. In addition,
Shoemaker et al. (2005) provide fact sheets for each of the
models reviewed containing: contact information, download
information, model overview (abstract), model features,
model areas supported, model capabilities (conceptual basis,
scientific detail, and model framework), scale (spatial and
temporal), assumptions, model strengths, model limitations,
application history, model evaluation, model inputs, user’s
guide, technical hardware and software requirements (com-
puter hardware, operating system, programming language,
and runtime estimates), linkage supported, related systems,
sensitivity-uncertainty-calibration, model interface capabili-
ties, and a complete list of source references.
L
OADING MODELS
GWLF
The Generalized Watershed Loading Functions (GWLF)
model (Haith and Shoemaker, 1987; Haith et al., 1992) is
used to predict monthly loadings of water, sediment, and
nutrients, including N and phosphorous (P), from non-
gauged watersheds with mixed land uses. The model
considers the watershed as a single unit and aggregates loads
from all land use areas into a watershed total. Surface runoff
is based on the USDA Soil Conservation Service (SCS, 1972)
runoff curve number method. Various storages and other
water components are adjusted to satisfy daily water balance.
Erosion from pervious areas is simulated using the Universal
Soil Loss Equation (USLE) (Wischmeier and Smith, 1978)
with calculations modified by a daily rainfall factor. An
area-based sediment delivery factor is applied to erosion. The
monthly sediment loads are computed from annual sediment
load using relative monthly transport abilities computed as a
power function of monthly runoff. Sediment can also be
contributed from impervious areas through a daily buildup/
washoff routine. The model is based on the assumption that
the relationship between erosion and sediment transport
varies on a monthly basis, but that for each simulated year
(April-March) there is no net deposition and no carryover of
detached sediment from year to year.
For nutrient simulations, GWLF uses a loading function
approach, where dissolved or particulate concentrations are
associated with flow volumes or sediment loads, respective-
ly, from various land uses or pollutant source inputs, such as
groundwater, manure application, and septic system effluent.
Best Management Practices (BMPs) can be simulated
through land use changes, changes in loading factors, and
application of efficiency factors during post-processing
(Yagow, 2004).
Therefore, GWLF can be used for both sediment and
nutrient (N and P) TMDLs for estimating loadings. Its
strengths include: simplicity, the relatively small amount of
input data required, the helpful user guidance for parameter
evaluation, and the fact that it does not require calibration,
although hydrologic calibration has been shown to be helpful
where monitored data are available. The model’s runtime
estimates are seconds to minutes. GWLF’s limitations
include: the lack of ongoing user support, and the require-
ment for auxiliary post-processing procedures to simulate
subwatersheds and account for many BMPs.
PLOAD
PLOAD (USEPA, 2001) is a simplified Geographic
Information System (GIS) based model developed by CH2M
HILL, a consulting firm, for calculating pollutant loads from
watersheds. PLOAD estimates nonpoint loads of pollution on
an annual average basis, for any user-specified pollutant,
including total suspended solids (suspended sediment), total
dissolved solids, biological oxygen demand, chemical oxy-
gen demand, P, N, nitrate plus nitrite, total Kjeldahl N,
ammonia, fecal coliform, lead, and zinc. The user may use
either the export coefficient or the EPA’s Simple Method
approach to calculate nonpoint-source loads. The model
requires GIS data and/or tabular inputs on land use,
watershed, BMP site and area (optional), pollutant loading
rates, impervious terrain factors, and point-source facility
locations and loads (optional). PLOAD can be used for both
sediment and nutrient TMDLs for estimating loadings.
RECEIVING WATER MODELS
AQUATOX
AQUATOX (USEPA, 2004; www.epa.gov/waterscience/
models/aquatox/about.html; accessed 9 Jan. 2006) can
represent a variety of aquatic ecosystems, including vertical-
ly stratified lakes, reservoirs and ponds, and rivers and
streams, and can simulate multiple environmental stressors
(including nutrients, organic loadings and chemicals, and
temperature) and their effects on a user-specified variety of
algal, macrophyte, invertebrate, and fish communities.
Therefore, AQUATOX can help to identify and quantify the
cause and effect relationships between chemical water
quality, the physical environment, and aquatic life in those
aquatic ecosystems (Shoemaker et al., 2005).
The model uses differential equations to represent chang-
ing values of state variables and solves the equations using a
numerical solution method. Time steps of 15 minutes to one
day (smaller steps during rapid changes) are used in solving
the equations. However, the reporting time step could vary
from 0.1 day to 99 days (one day is normally used). Runtime
970 TRANSACTIONS OF THE ASABE
estimates are reported to range from one second to several
minutes. The model is suitable for nutrient TMDLs for
analyzing receiving water qualities.
BATHTUB
BATHTUB (Walker, 1986) predicts lake and reservoir
responses to nutrient loadings on a yearly basis based on
in-lake/reservoir water quality state. It can compute average
loads over one year or over the growing season. It can
distinguish between the characteristics of a reservoir and
those of a natural lake, including the effects of non-algal
turbidity on transparency and algae responses to phospho-
rous. It summarizes information on in-lake water quality data
and can also estimate nutrient loadings based on correlations
of concentrations and flows (rating curves), although it is
commonly used in combination with a loading model, such
as GWLF (e.g., IEPA, 2004).
BATHTUB contains a number of regression equations
based on a wide range of lake and reservoir data sets. It can
treat the lake or reservoir as a continuously stirred, mixed
reactor, or it can predict longitudinal gradients in trophic state
variables in a reservoir or narrow lake. These trophic state
variables include in-lake total P (TP), ortho-P, organic N,
hypolimnetic DO, metalimnetic DO, chlorophyll concentra-
tions, and Secchi depth (transparency). Uncertainty esti-
mates are provided with predicted trophic state variables.
The model can be used for nutrient TMDLs for analyzing
receiving water quality.
CE-QUAL-W2
CE-QUAL-W2 (Cole and Wells, 2003) is a laterally
averaged, two-dimensional (longitudinal and vertical) hy-
drodynamic and water quality model for rivers, lakes,
reservoirs, and estuaries (best suitable for relatively long and
narrow water bodies). The model is based on finite difference
solutions of the governing partial differential equations using
variable grid spacing (segment length and layer thickness)
and internally adjustable time steps to ensure numerical
stability. The hydrodynamic component of the model
predicts water surface elevations, velocities, and tempera-
tures, while the water quality component simulates 21 con-
stituents, including nutrients, phytoplankton, and DO
interactions during anoxic conditions. Runtime estimates are
reported to be minutes to hours. The model can be used for
nutrient TMDLs for analyzing receiving water quality.
QUAL2E
QUAL2E (USEPA, 1995; www.epa.gov/docs/QUAL2E_
WINDOWS/; accessed 9 Jan. 2006) is a one-dimensional,
steady-state, and nonuniform flow and water quality model
that simulates nutrient dynamics, algal production, and DO
with the impact of benthic and carbonaceous demand in
streams and rivers. It predicts temperature, DO, biochemical
oxygen demand, ammonia, nitrate, organic N, inorganic P,
organic P, algae, and conservative and non-conservative
substances. The streams or rivers are assumed to be
trapezoidal and divided into homogeneous reaches, with
each reach further subdivided into uniform segments or
control volumes. Although the model is steady state, it can
accommodate diurnal variations of temperature and dis-
solved oxygen. The governing equations are advection-dis-
persion-reaction equations with external sources and sinks. A
flow balance is assumed, and steady nonuniform flow is used
to solve the advection equation. The governing equations are
numerically solved using an implicit finite difference
scheme. The dispersion coefficient is an empirical function.
The model accommodates uncertainty analyses, and runtime
estimates are in minutes. More descriptions and discussion
on QUAL2E are given in a companion article while
discussing DO modeling (Vellidis et al., 2006).
QUAL2K (Chapra and Pelletier, 2003; www.epa.gov/
ATHENS/wwqtsc/html/qual2k.html; accessed 3 Feb. 2006),
an enhanced version of QUAL2E, is also available. The
enhancements include two forms of carbonaceous biological
oxygen demand (slowly oxidizing and rapidly oxidizing),
internal sediment processes, and simulations of pH and
alkalinity. Runtime estimates are reported to be minutes to
hours. Both QUAL2E and QUAL2K can be used for nutrient
TMDLs for analyzing receiving water qualities.
Other Receiving Water Models
There are many other receiving water quality models that
have potential for TMDL development and implementation,
most of which are reviewed and categorized by Shoemaker
et al. (2005) and Kalin and Hantush (2003). It is noteworthy
to mention the Water Quality Analysis Simulation Program
(WASP) (www.epa.gov/athens/wwqtsc/html/wasp.html; ac-
cessed 9 Jan. 2006). WASP is a detailed and versatile
state-of-the-art receiving water quality model with dynamic
one-, two-, or three-dimensional spatial simulation capabili-
ties simulating both eutrophication, nutrient, and dissolved
oxygen (EUTRO), as well as metals, toxics, and sediment
(TOXI); it is also a computationally intensive model with
runtime estimates of minutes to hours. More description and
discussion on WASP can be found in the companion article
discussing DO modeling (Vellidis et al., 2006). The model is
suitable for both sediment and nutrient TMDLs for analyzing
receiving water qualities.
W
ATERSHED MODELS
HSPF
The Hydrological Simulation Program − Fortran (HSPF)
(Bicknell et al., 2001) is a continuous model for simulating
watershed hydrology and water quality for a wide range of
conventional and toxic organic pollutants. It performs
typically at an hourly time step and produces a time history
of water quantity and quality at any point in a watershed. The
watershed is divided into subwatersheds, each conceptual-
ized as a group of pervious and impervious land uses all
routed to a representative stream segment or a mixed
reservoir. Routing is performed by assuming that the
subwatersheds, streams, and the reservoirs (impoundments)
are a series of one-dimensional reservoirs.
HSPF uses a comprehensive, physically based water
budgeting procedure with interaction among the various
storages and processes. It accounts for interception, infiltra-
tion, evapotranspiration, snowmelt, surface runoff, interflow,
groundwater loss and recharge, and base flow; these are
mostly represented by empirical equations. HSPF allows
routing of in-stream flows and can simulate reservoir
behavior as well.
Pervious land surface erosion and transport are modeled
using exponential relationships for soil detachment, de-
tached sediment washoff, and gully erosion. Sediment from
impervious areas is also modeled with buildup/washoff
routines. In-stream sediment transport, deposition, and scour
of sediment are simulated for each of three particle-size
971Vol. 49(4): 967−986
classes (sand, silt, and clay) based on physical properties and
using published equations.
HSPF includes very detailed subroutines of nutrient
dynamics and calculates individual nutrient balances at a
user-specified time step, representing a series of storages and
phases with transport either by runoff in the dissolved phase
or attached to sediment in the particulate phase. HSPF allows
for detailed inputs of field operations and fertilization rates
(management activities) through its special actions module.
It simulates in-stream fate and transport of a wide variety of
pollutants, such as nutrients, sediment, tracers, DO,
biochemical oxygen demand, temperature, bacteria, and
user-defined constituents, including pesticides.
BMPs can be simulated either through land use changes,
a variety of special action functions that include direct
reductions of input source loads and distributions, or through
the Best Management Practice (BMPRAC) module. The
BMPRAC module simulates simple removal fractions for a
wide variety of constituents, including sediment and many
forms of nutrients. These removal fractions can vary monthly
or be constant.
Primary strengths of HSPF include: flexibility, ability to
simulate a wide range of user-configurable inputs, modular
structure that allows use of only those components needed for
a specific application, and USEPA and USGS support.
HSPF’s limitations include large input data requirements, the
need for monitored data in order to perform calibration, and
a steep learning curve. Its runtime estimates are seconds,
minutes, or even hours depending on the application. More
descriptions and discussion on HSPF are given in the
companion article discussing DO modeling (Vellidis et al.,
2006). It is suitable for both sediment and nutrient TMDLs.
LSPC
The Loading Simulation Program in C++ (LSPC) (Tetra
Tech, Inc., and USEPA, 2002) is a continuous watershed
modeling system that uses HSPF algorithms for simulating
hydrology, sediment, and general water quality on land, and
includes a simplified component for simulating in-stream
transport. The model can be configured for use in either
simple or complex watersheds, depending on the application
requirement and data availability. It has no inherent limita-
tions in terms of the numbers of modeled subdivisions of land
use and subwatersheds or of model operations, and is
applicable to large, complex watersheds. The model pro-
duces output on the subwatershed or reach-segment level.
Data management tools are included for evaluating output
from multiple watersheds and multiple scenarios simulta-
neously. The Microsoft Visual C++ programming architec-
ture allows for integration of output with widely available
database and spreadsheet programs.
LSPC simulates the hydraulics of complex natural and
man-made drainage networks and variable groundwater
surfaces. It is capable of simulating both peak flow and low
flows at a variety of time steps, from sub-hourly to daily. For
simulation within each subwatershed, processes are lumped
by land use category and, therefore, the relative location of
one land parcel to another is not represented. The model
approaches a distributed model when smaller subwatersheds
are used, although as in all models, the smaller the modeling
sub-unit, the greater the resources needed to estimate
parameter values and the longer the simulation times.
In-stream simulation is limited to well-mixed rivers and
reservoirs with one-directional flow.
LSPC uses empirical relationships to represent physical
processes and requires extensive calibration. Similar to
HSPF, its runtime estimates range from seconds to hours,
depending on spatial and temporal resolution and length of
simulation period, and it is suitable for both sediment and
nutrient TMDLs.
SWAT
The Soil and Water Assessment Tool (SWAT) (Neitsch et
al., 2002) is a continuous simulation model that runs typically
on a daily time step, simulates hydrology, weather, sedimen-
tation, soil temperature, crop growth, nutrients, pesticides,
and agricultural management. The watershed is divided into
subwatersheds, each connected through a stream channel and
further subdivided into hydrologic response units (HRUs)
with unique combinations of soils and land uses. Simulations
are performed at the HRU level and summarized in each
subwatershed. The simulated variables (water, sediment,
nutrients, and other pollutants) are routed through the stream
network to the watershed outlet.
Hydrologic simulations are based on a daily water budget
where change in soil water content is equal to precipitation
minus surface runoff, evapotranspiration, percolation, and
groundwater return (base) flow. Surface runoff is computed
using the SCS (1972) runoff curve number method; the
amount reaching the stream channel is computed using an
exponential function with a lag coefficient. Evapotranspira-
tion can be computed using three alternative well-publicized
methods. Percolation and groundwater return flow are
computed using exponential functions with empirical param-
eters. Lateral subsurface flow is computed simultaneously
with percolation using a well-publicized kinematic storage
model. All the surface, lateral subsurface, and base flow
waters reaching the stream channels are routed through the
channel network using a variable storage coefficient method
(Williams, 1969) or the Muskingum routing method (Linsley
et al., 1958). Transmission losses are determined while
routing water through channels using the SCS (1983)
method. In addition, canopy storage (interception) is com-
puted for simulating crop growth. Infiltration rates are
computed using the Green and Ampt (1911) infiltration
equation for sub-daily (smaller time increment) simulations.
Erosion and sediment yield are estimated for each HRU
with the Modified Universal Soil Loss Equation (MUSLE)
(Williams, 1975), an enhancement of the USLE. Sediment is
routed through the stream channel considering deposition
and degradation processes and using a simplified equation
(Williams, 1980) based on Bagnold’s definition of stream
power.
SWAT assumes that nitrate and organic N may be removed
from the soil via mass flow of water. Amounts of nitrate-N
contained in the runoff, lateral flow, and percolation are
estimated as products of the volume of water and the average
concentration of nitrate in a soil layer. Organic N transport
with sediment is calculated with a loading function (McElroy
et al., 1976; Williams and Hann, 1978) for individual runoff
events. Plant use of N is estimated with a supply and demand
approach where the daily plant N demands are calculated as
the difference between the actual concentration of the
element in the plant and the optimal concentration. The