United States
Environmental Protection
Agency
Office of
Research and Development
Washington, DC 20460
EPA/600/3-91/039
June 1991
EPA
Modeling of Nonpoint
Source Water Quality in
Urban and Non-urban
Areas
-------
-------
EPA/600/3-91/039
June 1991
Modeling of Nonpoint Source Water Quality
in Urban and Non-urban Areas
by
Anthony S. Donigian, Jr.
AQUA TERRA Consultants
Mountain View, California 94093-1004
and
Wayne C. Huber
University of Florida
Gainesville, Florida 32611-2013
Project Technical Monitor
Thomas O. Barnwell, Jr.
Assessment Branch
Environmental Research Laboratory
Athens, Georgia 30613
ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ATHENS, GA 30613
Printed on Recycled Paper
-------
Disclaimer
The work presented in this document has been funded by the
U, S. Environmental Protection Agency under Contract No.
68-03-3513 to AQUA TERRA Consultants. It has been
subjected to the Agency's peer and administrative review
and has been approved as an EPA document. Mention of
trade names or commercial products does not constitute
endorsement or recommendation for use.
11
-------
Foreword
As environmental controls become more costly to implement
and the penalties of judgment errors become more severe,
environmental quality management requires more efficient
analytical tools based on greater knowledge of the
environmental phenomena to be managed. As part of this
Laboratory's research on the occurrence, movement,
transformation, impact, and control of environmental
contaminants, the Assessment Branch develops management
and engineering tools to help pollution control officials
address environmental problems.
Pollutants in runoff and seepage from urban, agricultural,
and forested areas contribute significantly to water pollution
problems in many areas of the United States. The
development and application of computer-operated
mathematical models to simulate the movement of pollutants
and thus to anticipate environmental problems has been the
subject of extensive research by government agencies,
universities, and private companies for many years. This
review and model-selection guidance document was
developed under the direction of EPA's Office of Water and
Office of Research and Development to assist water quality
planners in applying modeling techniques to the development
of cost-effective nonpoint source controls.
Rosemarie C. Russo, Ph.D.
Director
Environmental Research Laboratory
Athens, Georgia
-------
Abstract
Nonpoint source assessment procedures and modeling
techniques are reviewed and discussed for both urban and
non-urban land areas. Detailed reviews of specific meth-
odologies and models are presented, along with overview
discussions focussing on urban methods and models, and
on non-urban (primarily agricultural) methods and models.
Simple procedures, such as constant concentration, regres-
sion, statistical, and loading function approaches are de-
scribed, along with complex models such as SWMM,
HSPF, STORM, CREAMS, SWRRB, and others. Briefcase
studies of ongoing and recently completed modeling efforts
are described. Recommendations for nonpoint runoff quality
modeling are presented to elucidate expected directions of
future modeling efforts. This work was performed as Work
Assignment No. 29 under EPA Contract No. 68-03-3513
with AQUA TERRA Consultants. Work was complete as of
March 1990.
IV
-------
Contents
Page
Disclaimer .....,„ . . . ii
Foreword iii
Abstract iv
Acknowledgments vii
Section 1.0 - Nonpoint Source Modeling Objectives and Considerations .1
Section 2.0 - Overview of Nonpoint Source Quality Modeling . . . 3
- 2.1 Modeling Fundamentals 3
2.2 Operational Models ; "3
2.3 Surveys and Reviews of Nonpoint Source Models 3
2.4 Summary of Data Needs . 4
Section 3.0 - Overview of Available Urban Modeling Options . 5
3.1 Introduction . 5
3.2 Constant Concentration or Unit Loads 5
3.3 Spreadsheets 5
3.4 Statistical Method . . 6
3.5 Regression—Rating Curve Approaches 7
3.6 Buildup and Washoff fe. . 7
3.7 Related Mechanisms : 8
Section 4.0 - Overview of Available Non-urban Modeling Options 10
4.1 Nonpoint Source Loading Functions . 10
4.2 Simulation Models . 12
Section 5.0 - Nonpoint Source Runoff Quality Simulation Models and Methods 18
5.1 Urban Runoff Quality Models . . . 18
5.2 Non-urban Runoff Quality Models and Methods 19
5.3 Discussion 22
Section 6.0 - Brief Case Studies • 26
6.1 Urban Model Applications 26
6.2 Non-urban Model Applications . 27
Section 7.0 - Summary and Recommendations for Nonpoint Source Runoff Quality Modeling 29
Section 8.0 - References • • • • 31
Section 9.0 - Appendix: Detailed Model Descriptions 37
Storage, Treatment, Overflow, Runoff Model (STORM) 43
Distributed Routing Rainfall Runoff Model—Quality DR3M-QUAL , . 46
EPA Screening Procedures 49
Agricultural NonPoint Source Pollution Model—(AGNPS) 52
Areal Nonpoint Source Watershed Environment Response Simulation—(ANSWERS) 55
Chemicals, Runoff, and Erosion from Agricultural Management Systems—(CREAMS) 57
Groundwater Loading Effects of Agricultural Management Systems—(GLEAMS) 57
Hydrological Simulation Program—Fortran (HSPF) . . . .' . . 60
Stream Transport and Agricultural Runoff of Pesticides for Exposure Assessment (STREAM) . 60
Pesticide Root Zone Model—PRZM .'. 64
Simulator for Water Resources in Rural Basins—(SWRRB) . 67
Pesticide Runoff Simulator—(PRS) 67
Unified Transport Model for Toxic Materials—UTM-TOX 70
-------
Acknowledgments
This nonpoint model review was sponsored by the EPA
Office of Water Regulations and Standards, through the
Environmental Research Laboratory in Athens, Georgia.
Mr. Thomas O. Bamwell was the EPA Project Technical
Monitor, with the Assessment Branch in Athens, and Dr.
Hiranmay Biswas was the liaison with EPA Headquarters.
Their support and guidance for this effort is gratefully
acknowledged.
Dr. Wayne Huber of the University of Florida performed
the urban model review, and Mr. Anthony Donigian of
AQUA TERRA Consultants performed the review of the
non-urban models. Both authors reviewed the entire re-
port. Mr. Avinash Patwardhan assisted Mr. Donigian with
the information gathering and review of the non-urban
models. Also, the draft final report was peer reviewed
by Dr. Vladimir Novotny of Milwaukee, Wisconsin; and
Dr. Matt C. Smith of the University of Georgia, Athens,
Georgia. Their insightful comments and suggestions were
sincerely appreciated.
VI
-------
Section 1.0
Nonpoint Source Modeling Objectives and Considerations
Studies and projects involving stonnwater runoff quality
from all categories of land use—urban, agricultural crop-
land, pasture, forest—can relate to many environmental
problems. In the broadest sense, water quality studies may
be performed to protect the environment under various state
and federal legislation. For example, Section 304(1) of the
Clean Water Act requires States to identify waterbodies
impaired by both point and nonpoint source pollution and
develop appropriate control strategies; while Section 405
will eventually require analysis of stonnwater outfalls in all
urban areas in the U.S. In a narrower sense, a study may
address a particular water quality issue in a particular
receiving water, such as bacterial contamination of a beach,
release of oxygen demanding material into a stream or river,
unacceptable aesthetics of an open channel receiving urban
and non-urban runoff, eutropbication of a lake,
contamination of basements from surcharged sewers due to
wet-weather flooding, etc.
By no means should it be assumed that every water quality
problem requires a water quality modeling effort. Some
problems may be mostly hydraulic in nature, e.g., the base-
ment flooding problem. That is, the solution may often
reside primarily in a hydrologic or hydraulic analysis in
which die concentration or load of pollutants is irrelevant.
In some instances, local or state regulations may prescribe
a nominal "solution" without recourse to any water quality
analysis as such. For example, stonnwater runoff in Florida
is considered "controlled" through retention or detention
with filtration of the runoff from the first inch of rainfall for
areas of 100 acres or less. Other problems may be resolved
through the use of measured data without the need to model.
In other words, many problems do not require water quality
modeling at all.
If a problem does require modeling, specific modeling
objectives will need to be defined to guide the modeling
exercise and approach. Models may be used for objectives
such as the following:
1. Characterize runoff quantity and quality as to temporal
and spatial detail, concentration/load ranges, etc.
2. Provide input to a receiving water quality analysis, e.g.,
drive a receiving water quality model.
3. Determine effects, magnitudes, locations, combinations,
etc. of control options.
4. Perform frequency analysis on quality parameters, e.g.,
to determine return periods of concentrations/loads.
5. Provide input to cost-benefit analyses.
Objectives 1 and 2 characterize the magnitude of the prob-
lem, and objectives 2 through 5 are related to the analysis
and solution of the problem. Computer models allow some
types of analysis, such as frequency analysis, to be per-
formed that could rarely be performed otherwise since
periods of water quality measurements are seldom very
long. It should always be borne in mind, however, that use
of measured data is usually preferable to use of simulated
data, particularly for objectives 1 and 2 in which accurate
concentration values are needed. In general, models are not
good substitutes for good field sampling programs. On the
other hand, models can sometimes be used to extend and
extrapolate measured data.
Careful consideration should be given to objective number
2. The first urban runoff quality model (SWMM) inadver-
tently overemphasized the concept of simulation of detailed
intra-storm quality variations, e.g., production of a "polluto-
graph" (concentration vs. time) at 5 or 10 minute intervals
during a storm for input to a receiving water quality model.
The early agricultural runoff quality models (e. g., ARM and
ACTMO models) followed a similar detailed approach
primarily to evaluate and demonstrate the models' abilities
to represent observed data from small (less than 10 hectare)
monitored fields.
But the fact is that the quality response of most receiving
waters is insensitive to such short-term variations, as illus-
trated in Table 1. In most instances, the total storm load
will suffice to determine the receiving water response,
eliminating the necessity of becoming embroiled in calibra-
tion against detailed pollutographs especially for conven-
tional pollutants. Instead, only the total storm loads need be
matched, a much easier task. Exceptions to this general
observation may be appropriate when considering toxic pol-
lutants, e.g., pesticides, when short-term concentrations may
be lethal to aquatic organisms. Also, simulation of short
time increment changes in concentrations and loads is gener-
ally necessary for analysis of control options, such as
-------
storage or high-rate treatment, whose efficiency may depend
on the transient behavior of the quality constituents.
Any consideration of water qualify modeling means that
some additional data will be required for model input. As
described later, such requirements may be as simple as a
constant concentration, or much more complex such as soil
nitrification or mineralization rates. Data may be obtained
from existing studies or their acquisition may require
extensive field monitoring. For some conceptualizations of
the urban quality cycle, e.g., buildup and washoff, it may
not be routinely possible to physically measure fundamental
input parameters, and such parameters will only be obtained
through model calibration. Involvement in acquisition of
quality data, be it through literature reviews or field sur-
veys, profoundly escalates the level of effort required for
the study. Details on data requirements for modeling will be
deferred until modeling techniques are described.
Tabt* 1. Required temporal detail for receiving water analysis
(After Driscoll, 1979, and Hydroscience, 1979)
Type of
Receiving Water
Key
Constituents
Response
Time
Lakes, Bays
Estuaries
Large Rivers
Streams
Ponds
Beaches
Nutrients
Nutrients, OD"
Bacteria
OD, Nitrogen
OD, Nitrogen
Bacteria
OD, Nutrients
Bacteria
Weeks—Years
Days—Weeks
Days
Hours—Days
Hours
Hours-Weeks
Hours
•OD - oxygon demand, e.g., BOD, that affects dissolved oxygen.
-------
Section 2.0
Overview of Nonpoint Source Quality Modeling
2.1 Modeling Fundamentals
Modeling caveats and an introduction to modeling are pre-
sented by several authors including James and Burges
(1982), Kibler (1982), Huber (1985,1986) and summarized
in a recent manual of practice (WPCF, 1989). Space does
not permit a full presentation here; a few items are high-
lighted below.
1. Have a clear statement of project objectives. Verify
. the need for quality modeling. (Perhaps the objectives
can be satisfied without quality modeling).
2. Use the simplest model that will satisfy the project
objectives. Often a screening model, e.g., regression
or statistical, can determine whether more complex
simulation models are needed.
3. To the extent possible, utilize a quality prediction
method consistent with available data. This would
ordinarily rule against buildup-washoff formulations,
although these might still be useful for detailed
simulation, especially if calibration data exist.
4. Only predict the quality parameters of interest and
only over a suitable time scale. That is, storm event
loads and EMCs will usually be the most detailed pre-
diction necessary, and seasonal or annual loads will
sometimes be all that is required. Do not attempt to
simulate intra-storm variations in quality unless it is
necessary.
5. Perform a sensitivity analysis on the selected model
and familiarize yourself with the model characteristics.
6. If possible, calibrate and verify the model results. Use
one set of data for calibration and another independent
set for verification. If no such data exist for die appli-
cation site, perhaps they exist for a similar catchment
nearby.
2.2 Operational Models
Implementation of an off-the-shelf model or method will be
easiest if the model can be characterized as "operational" in
the sense of:
1. Documentation. This should include a user's manual,
explanation of theory and numerical procedures, data
needs, data input format, etc. Documentation most
often separates the many computerized procedures
found in the literature from a model that can be
accessed and easily used by others.
2. Support. This is sometimes provided by the model
developer but often by a federal agency such as the
HEC or EPA.
3. Experience. Every model must be used a "first time"
but it is best to rely on a model with a proven track
record.
The models described in Sections 3.0 and 4.0 are all opera-
tional in this sense. New methods and models are constantly
under development and should not be neglected simply be-
cause they lack one of these characteristics, but the user
should be aware of potential difficulties if any characteristic
is lacking.
2.3 Surveys and Reviews of Nonpoint
Source Models
Several publications, often somewhat put of date, provide
reviews of available models. Some models (e.g., SWMM,
STORM, HSPF, CREAMS) have persisted for many years
and are included in both older and newer reviews, while
other models (e.g., USGS, Statistical, spreadsheet, AGNPS,
SWRRB) are more recent. Reviews that consider surface
runoff quality models include Huber and Heaney (1982),
Kibler (1982), Whipple et al. (1983), Barnwell (1984,
1987), Huber (1985, 1986), Donigian and Beyerlein (1985),
Bedient and Huber (1988), and Viessman et al. (1989).
HEC models are described in detail by Feldman (1981).
Descriptions of EPA nonpoint source water quality models
are provided by Ambrose et al. (1988) and Ambrose and
Barnwell (1989). Selected pesticide runoff models have been
reviewed by Mulkey et al. (1986) and Lorber and Mulkey
(1982). Agricultural nonpoint source models have been
showcased in a number of conferences and symposia over
the past few years, including a 1983 Symposium on Natural
Resources Modeling (DeCoursey, 1985); a 1984 Conference
on Agricultural Nonpoint source Pollution: Model Selection
and Application, in Venice, Italy (Giorgini and Zingales,
-------
1986); and a June 1988 International Symposium on Water
Quality Modeling of Agricultural Nonpoint Sources (De-
Coursey, In Press). Beasley and Thomas (1989) describe
recent model enhancements and applications for five selected
models to agricultural and forested watersheds in the south-
eastern U.S. These reports and proceedings provide a wealth
of information on current efforts and recent developments in
modeling nonpoint contributions and water quality impacts
of agricultural activities.
2.4- Summary of Data Needs
In application of most models, there are two fundamental
types of data requirements. First, there are the data needed
simply to make the model function, that is, input parameters
and timeseries data for the model. These typically include
precipitation (rainfall) and other meteorologic informa-
tion, drainage area, imperviousness, runoff coefficient and
other quantity prediction parameters, plus quality predic-
tion parameters such as constant concentration, constituent
median and CV, regression relationships, buildup and wash-
off parameters, soil/chemical characteristics, partition co-
efficients, reaction rates, etc. In other words, each mode
will have a fundamental list of required input data. Although
it is difficult to generalize for the entire universe of
both simple and complex nonpoint source models, Novotny
and Chesters (1981) have prepared a summary table of re-
quired input data from which Table 2 was adapted.
The second type of information is required for calibration
and verification of more complex models, namely, sets of
measured runoff and quality samples (coincident with the
input precipitation and meteorologic data) with which to test
the model. Such data exist (e.g., Huber et al., 1982; Driver
et al. (1985), Noel et al., 1987) but seldom for the site of
interest. If the project objectives absolutely require such data
(e.g., if a model must be calibrated in order to drive a re-
ceiving water quality model), then expensive local monitor-
ing may be necessary.
This summary relates primarily to quality prediction and
may not represent a comprehensive statement of data needs
for quantity prediction. However, since rainfall and runoff
are required for virtually every study, certain quantity-
related parameters are also necessary for various methods.
Table 2. Input data needs for nonpoint source models
(after Novotny and Chesters, 1981)
1. System Parameters
a. Watershed Size
b. Subdivision of the Watershed into Homogenous Subareas
c. Imperviousness of Each Subarea
d. Slopes
e. Fraction of Impervious Areas Directly Connected to a Channel
f. Maximum Surface Storage (depression plus interception storage)
g. Soil Characteristics Including Texture, Permeability, Erodibility, and Composition
h. Crop and Vegetation Cover
i. Curb Density or Street Gutter Length
k. Sewer System or Natural Drainage Characteristics
2. State Variables
a. Ambient Temperature
b. Reaction Rate Coefficients
c. Adsorption/Desorption Coefficients
d. Growth Stage of Crops
e. Daily Accumulation Rates of Litter
f. Traffic Density and Speed
0. Potency Factors for Pollutants (pollutant strength on sediment)
h. Solar Radiation (for some models)
3. Input Variables
a. Precipitation
b. Atmospheric Fallout
c. Evaporation Rates
-------
Section 3.0
Overview of Available Urban Modeling Options
3.1 Introduction
Several quality modeling options exist for simulation of
quality in urban storm and combined sewer systems. These
have been reviewed by Huber (1985; 1986) and range from
simple to involved, although some "simple" methods, e.g.,
the EPA statistical methods, can incorporate quite sophis-
ticated concepts. The principal methods available to the
contemporary engineer are outlined below, in a rough order
of complexity. Their data requirements are summarized in
Table 3. The methods are:
1. Constant concentration
2. Spreadsheet
3. Statistical
4. Rating curve or regression
5. Buildup/washoff
3.2 Constant Concentration or Unit Loads
As its name implies, all runoff is assumed to have the same,
constant concentration for a given pollutant. At its very
simplest, an annual runoff volume can be multiplied by a
concentration to produce an annual runoff load. However,
this option may be coupled with a hydrologic model, where-
in loads (product of concentration and flow) will vary if
the model produces variable flows. This option may be quite
useful because it may be used with any hydrologic or hy-
draulic model to produce loads, merely by multiplying by
the constant concentration. For instance, the highly sophis-
ticated SWMM Extran Block may be used for hydraulic
analysis of sewer system, prediction of overflows and diver-
sions to receiving waters, etc., yet it performs no quality
simulation as such. In many instances, it may be most im-
portant to get the volume and timing of such overflows and
diversions correctly, and simply estimate loads by multi-
plying by a concentration.
An obvious question is what (constant) concentration to use?
The EPA NURP studies (EPA, 1983) have produced a large
and invaluable data base from which to select numbers,
but the 30 city coverage of NURP will most often not
include a site representative of the area under study.
Nonetheless, a large data base does exist from which to
review concentrations. Another option is to use measured
values from the study area. This might be done from a
limited sampling program. However, the NURP study con-
clusively demonstrated the variation that exists in event
mean concentrations (EMCs, total storm event load divid-
ed by total storm event runoff volume) at a site, within a
city, and within a region or the country as a whole. Thus,
while use of a constant concentration may produce load
variations, EMC variations will not be replicated. These
variations may be important in the study of control options
and receiving water responses.
Unit loads are perhaps an even simpler concept. These con-
sist of values of mass per area per time, typically Ib/ac-yr
or kg/ha-yr, for various pollutants, although other normali-
zations such as Ib/curb-mile are sometimes encountered.
Annual (or other time unit) loads are thus produced upon
multiplication by the contributing area. Such loadings are
obviously highly site-specific and depend upon both demo-
graphic and hydrologic factors. They must be based on an
average or "typical" runoff volume and cannot vary from
year to year, but they can conveniently be subject to reduc-
tion by best management practices (BMPs), if the BMP ef-
fect is known. Although early EPA references provide some
information for various land used (EPA, 1973; EPA, 1976a;
McElroy et al., 1976), unit loading rates are exceedingly
variable and difficult to transpose from one area to another.
Constant concentrations can sometimes be used for this pur-
pose, since mg/1 x 0.2265 = Ib/ac per inch of runoff. Thus,
if a concentration estimate is available, the annual loading rate,
for example, may be calculated by multiplying by the inches
per year of runoff. Finally, the Universal Soil Loss Equation
(Wischmeier and Smith, 1978; Heaney et al., 1975) was devel-
oped to estimate tons per acre per year of sediment loss
from land surfaces. If a pollutant may be considered as a
fraction ("potency factor") of suspended solids concentration
or load, this offers another option for prediction of annual
loads. Lager et al. (1977), Manning et al. (1977) and Zison
(1980) provide summaries of such values.
3.3 Spreadsheets
Microcomputer spreadsheet software, e.g., Lotus, Quattro,
Excel, is now ubiquitous in engineering practice. Very
extensive and highly sophisticated engineering analysis is
routinely implemented on spreadsheets, and water quality
-------
simulation is no exception. In essence, the spreadsheet may
be used to automate and extend the concept of the constant
concentration idea. In the usual manifestation of this spread-
sheet application, runoff volumes are calculated very sim-
plistically, usually using a runoff coefficient times a rainfall
depth. The coefficient may vary according to land use, or an
SCS procedure may be used, but the hydrology is inherently
simplistic in the spreadsheet predictions. The runoff volume
is then multiplied by a constant concentration to predict
runoff loads. The advantage of the spreadsheet is that a
mixture of land uses (with varying concentrations) may
easily be simulated, and an overall load and flow-weighted
concentration obtained from the study area (Walker et al.,
1989). The study area itself may range from a single
catchment to an entire urban area. The relative contributions
of different land uses may be easily identified, and handy
spreadsheet graphics tools used for display of the results.
As an enhancement, control options may be simulated by
application of a constant removal fraction for an assumed
BMP. Although spreadsheet computations can be amazingly
complex, BMP simulation is rarely more complicated than
a simple removal fraction because anything further would
require simulation of the dynamics of the removal device
(e.g., a wet detention pond), which is usually beyond the
scope of the hydrologic component of the spreadsheet
model. Nonetheless, if simple BMP removal fractions can
be believed, the spreadsheet can easily be used to estimate
the effectiveness of control options. Loads with and without
controls can be estimated and problem areas, by contributing
basin and land use, can be determined. Since most engineers
are familiar with spreadsheets, such models can be devel-
oped in-house in a logical manner.
The spreadsheet approach is best suited to estimation of
long-term loads, such as annual or seasonal, because very
simple prediction methods generally perform better over a
long averaging time and poorly at the level of a single storm
event. Hence, although the spreadsheet could be used at the
microscale (at or within a storm event) it is most often
applied for much longer time periods. It is harder to obtain
the variation of predicted loads and concentrations using the
spreadsheet method because this can ordinarily only be done
by varying the input concentrations or rainfall values. A
Monte Carlo simulation may be attempted (i.e., systematic
variation of all input parameters according to an assumed
frequency distribution) if the number of such parameters is
not too large. These results may then be used to estimate the
range and/or frequency distribution of predicted loads and
concentrations.
In a generic sense, the spreadsheet idea may be used in
methods programmed in other languages, e.g., Fortran. For
example, comprehensive assessments of coastal zone pollu-
tion from urban areas are made by NOAA (1987) by as-
sembling land use data with different runoff coefficients,
predicting daily and seasonal runoff volumes from daily
rainfall, and predicting seasonal pollutant loads using con-
stant concentrations. Although the demographic data base
and use of magnetic tapes may dictate use of mainframes,
the computational concept is still that of a spreadsheet.
Again, the question arises of what concentrations to use, this
time potentially for multiple land uses and subareas. And
again, the NURP data base will usually be the first one to turn
to, with the possibility of local monitoring to augment it.
3.4 Statistical Method
The so-called "EPA Statistical Method" is somewhat generic
and until recently was not implemented in any off-the-shelf
model or even very well in any single report (Hydroscience,
1979; EPA, 1983). A new FHWA study (Driscoll et al.,
1989) partially remedies this situation. The concept is
straightforward, namely that of a derived frequency
distribution for EMCs. This idea has been used extensively
for urban runoff quantity (e.g., Howard, 1976; Loganathan
and Delleur, 1984; Zukovs et al., 1986) but not as much for
quality predictions.
The EPA Statistical Method utilizes the fact that EMCs are
not constant but tend to exhibit a lognormal frequency
distribution. When coupled with an assumed distribution of
runoff volumes (also lognormal), the distribution of runoff
loads may be derived. When coupled again to the distribu-
tion of streamflow, an approximate (lognormal) probability
distribution of in-stream concentrations may be derived
(Di Toro, 1984)—a very useful result, although assumptions
and limitations of the method have been pointed out by
Novotny (1985) and Roesner and Dendrou (1985). Further
analytical methods have been developed to account for
storage and treatment (Di Toro and Small, 1979; Small and
Di Toro, 1979). The method was used as the primary
screening tool in the EPA NURP studies (EPA, 1983) and
has also been adapted to combined sewer overflows (Dris-
coll, 1981) and highway-related runoff (Driscoll etal.,
1989). This latter publication is one of the best for a
concise explanation of the procedure and assumptions and
includes spreadsheet software for easy implementation of
the method.
A primary assumption is that EMCs are distributed log-
normally at a site and across a selection of sites. The con-
centrations may thus be characterized by their median value
and by their coefficient of variation (CV = standard devia-
tion divided by the mean). There is little doubt that the
lognormality assumption is good (Driscoll, 1986), but simi-
lar to the spreadsheet approach, the method is then usually
combined with weak hydrologic assumptions, e.g., predic-
tion of runoff using a runoff coefficient. (The accuracy of a
runoff coefficient increases as urbanization and impervious-
ness increase.) However, since many streams of concern in
an urban area consist primarily of storinwater runoff during
wet weather, the ability to predict the distribution of EMCs
is very useful for assessment of levels of exceedance of
water quality standards. The effect of BMPs can again be
estimated crudely through constant removal fractions that
effect on the coefficient of variation. Overall, the method
has been very successfully applied as a screening tool.
-------
Input to the method as implemented for the FHWA (Driscoll
et al., 1989) includes statistical properties of rainfall (mean
and coefficient of variation of storm event depth, duration,
intensity and interevent time), area, and runoff coefficient
for the hydrologic component, plus EMC median and coeffi-
cient of variation for the pollutant. Generalized rainfall
statistics have already been calculated for many locations in
the U.S. Otherwise, the EPA SYNOP model (EPA, 1976b;
Hydroscience, 1979; EPA, 1983; Woodward-Clyde, 1989)
must be run on long-term hourly rainfall records. If receiv-
ing water impact is to be evaluated the mean and CV of the
streamflow are required plus the upstream concentration. A
Vollenweider-type lake impact analysis is also provided
based on phosphorus loadings.
As with the first two methods discussed, the choice of
median concentration may be difficult, and the Statistical
Method requires a coefficient of variation as well. For-
tunately, from NURP and highway studies, CV values for
most urban runoff pollutants are fairly consistent, and a
value of 0.75 is typical. If local and/or NURP data are
not available or inappropriate, local monitoring may be
required, as in virtually every quality prediction method.
The estimation of the whole EMC frequency distribution for
a pollutant is a definite advantage of the Statistical Method
over some applications of constant concentration and simple
spreadsheet approaches. Frequency analyses of water quan-
tity and quality parameters may also be performed on the
output of continuous simulation models such as HSPF,
SWMM and STORM. The derived distribution approach of
the Statistical Method avoids the considerable effort required
for continuous simulation at the expense of simplifying
assumptions that may or may not reflect the prototype
situation adequately.
3.5 Regression—Rating Curve Approaches
With the completion of the NURP studies in 1983, there are
measurements of rainfall, runoff and water quality at well
over 100 sites in over 30 cities. Some regression analysis
has been performed to try to relate loads and EMCs to
catchment, demographic and hydrologic characteristics (e. g.,
McElroy etal., 1976; Miller etal., 1978; Brown, 1984),
the best of which are recent results of the USGS (Tasker
and Driver, 1988; Driver and Tasker, 1988), to be de-
scribed briefly below. Regression approaches have also been
used to estimate dry-weather pollutant deposition in com-
bined sewers (Pisano and Queiroz, 1977), a task at which no
model is very successful. What are termed "rating curves"
herein are just a special form of regression analysis, in
which concentration and/or loads are related to flow rates
and/or volumes. This is an obvious exercise attempted at
most monitoring sites and has a historical basis in sediment
discharge rating curves developed as a function of flow rate
in natural river channels.
A rating curve approach is most often performed using total
storm event load and runoff volume although intra-storm
variations can sometimes be simulated in this manner as
well (e.g., Huber and Dickinson, 1988). It is usually
observed (Huber, 1980; EPA, 1983; Driscoll et al., 1989)
that concentration (EMC) is poorly or not correlated with
runoff flow or volume, implying that a constant concentra-
tion assumption is adequate. Since the load is the product of
concentration and flow, load is usually well correlated with
flow regardless of whether or not concentration correlates
well. Manifestation of spurious correlation (Benson, 1962)
is often ignored in urban runoff studies. If load is propor-
tional to flow to the first power (i.e., linear), then the
constant concentration assumption holds; if not, some
relationship of concentration with flow is implied. Rating
curve results can be used by themselves for load and EMC
estimates and can be incorporated into some models (e.g.,
SWMM, HSPF).
Rainfall, runoff and quality data were assembled for 98
urban stations in 30 cities (NURP and other) hi the U.S. for
multiple regression analysis by the USGS (Driver and
Tasker, 1988; Tasker and Driver 1988). Thirty-four multi-
ple regression models (mostly log-linear) of storm runoff
constituent loads and storm runoff volumes were developed,
and 31 models of storm runoff EMCs were developed. Re-
gional and seasonal effects were also considered. The two
most significant explanatory variables were total storm
rainfall and total contributing drainage area. Impervious
area, land use, and mean annual climatic characteristics also
were significant explanatory variables in some of the
models. Models for estimating loads of dissolved solids,
total nitrogen, and total ammonia plus organic nitrogen
(TKN) generally were the most accurate, whereas models
for suspended solids were the least accurate. The most
accurate models were those for the more arid Western U.S.,
and the least accurate models were those for areas that had
large mean annual rainfall.
These USGS equations represent the best generalized
regression equations currently available for urban runoff
quality prediction. Note that such equations do not require
preliminary estimates of EMCs or local quality monitoring
data except for the very useful exercise of verification of the
regression predictions. Regression equations only predict the
mean and do not provide the frequency distribution of pre-
dicted variable, a disadvantage compared to the statistical
approach. (The USGS documentation describes procedures
for calculation of statistical error bounds, however). Finally,
regression approaches, including rating curves, are notori-
ously difficult to apply beyond the original data set from
which the relationships were derived. That is, they are sub-
ject to very large potential errors when used to extrapolate
to different conditions. Thus, the usual caveats about use of
regression relationships continue to hold when applied to
prediction of urban runoff quality.
3.6 Buildup and Washoff
In the late 1960s, a Chicago study by the American Public
Works Association (1969) demonstrated the (assumed linear)
buildup of "dust and dirt" and associated pollutants on urban
street surfaces. During a similar time frame, Sartor and
Boyd (1972) also demonstrated buildup mechanisms on the
-------
surface as well' as an exponential washoff of pollutants
during rainfall events. These concepts were incorporated
into the original SWMM model (Metcalf and Eddy et al.,
1971) as well as into the STORM, USGS and HSPF models
to a greater or lesser degree (Huber, 1985). "Buildup" is a
term that represents all of the complex spectrum of dry-
weather processes that occur between storms, including
deposition, wind erosion, street cleaning, etc. The idea is
simply that all such processes lead to an accumulation of
solids and perhaps other pollutants that are then "washed
off during storm events.
Although ostensibly physically based, models that include
buildup and washoff mechanisms really employ conceptual
algorithms because the true physics is related to principles
of sediment transport and erosion that are poorly understood
in this framework. Furthermore, the inherent heterogeneity
of urban surfaces leads to use of average buildup and wash-
off parameters that may vary significantly from what may
occur in an isolated street gutter, for example. Thus, except
in rare instances of measurements of accumulations of sur-
face solids, the use of buildup and washoff formulations
inevitably results in a calibration exercise against measured
end-of-pipe quality data. It then holds that in the absence of
such data, inaccurate predictions can be expected.
Different models offer different options for conceptual
buildup and washoff mechanisms, with SWMM having the
greatest flexibility. In fact, with calibration, good agreement
can be produced between predicted and measured concentra-
tions and loads with such models, including intra-storm vari-
ations that cannot be duplicated with most of the methods
discussed earlier. (When a rating curve is used in SWMM
instead of buildup and washoff, it is also possible to
simulate intra-storm variations in concentration and load.)
A survey of linear buildup rates for many pollutants by
Manning et al. (1977) is probably the best source of gener-
alized buildup data, and some information is available in
the literature to aid in selection of washoff coefficients
(Huber 1985; Huber and Dickinson, 1988). However, such
first estimates may not even get the user in the ball park
(i.e., quality—not quantity—predictions may be off by more
than an order of magnitude); the only way to be sure is to
use local monitoring data for calibration and verification.
Thus, as for most of the other quality prediction options
discussed herein, the buildup-washoff model may provide
adequate comparisons of control measures, ranking of loads,
etc. but cannot be used for prediction of absolute values of
concentrations and loads, e.g., to drive a receiving water
quality model, without adequate calibration and verification
data. Since buildup and washoff are somewhat appealing
conceptually, it is somewhat easier to simulate potential
control measures such as street cleaning and surface infil-
tration using these mechanisms than with, say, a constant
concentration or rating curve method. In the relatively
unusual instance in which intra-storm variations in con-
centration and load must be simulated, as opposed to total
storm event EMC or load, buildup and washoff also offer
the most flexibility. This is sometimes important for the
design of storage facilities in which first-flush mechanisms
may be influential.
As mentioned above, generalized data for buildup and wash-
off are sparse (Manning et al., 1977) and such measure-
ments almost never conducted as part of a routine monitor-
ing program. For buildup, normalized loadings, e.g., mass/
day-area or mass/day per curb-length, or just mass/day, are
required, along with an assumed functional form for buildup
vs. time, e.g., linear, exponential, Michaelis-Menton, etc.
For washoff, the relationship of washoff (mass/time) vs.
runoff rate must be assumed, usually in the form of a power
equation. When end-of-pipe concentration and load data are
all that are available, all buildup and washoff coefficients
end up being calibration parameters.
3.7 Related Mechanisms
In the discussion above, washoff is assumed proportional to
the runoff rate, as for sediment transport. Erosion from
pervious areas may instead be proportional to the rainfall
rate. HSPF does the best job of including this mechanism in
its algorithms for erosion of sediment from pervious areas.
SWMM includes a weaker algorithm based on the Universal
Soil Loss Equation.
Many pollutants, particularly metals and organics, are ad-
sorbed onto solid particles and are transported in particulate
form. The ability of a model to include "potency factors"
(HSPF) or "pollutant fractions" (SWMM) enhance the abil-
ity to estimate the concentration or load of one constituent
as a fraction of that of another constituent, e.g., solids
(Zison, 1980).
The groundwater contribution to flow in urban areas can be
important in areas with unlined and open channel drainage.
Of the urban models discussed, HSPF far and away has the
most complex mechanisms for simulation of subsurface
water quality processes in both the saturated and unsaturated
zones. Although SWMM includes subsurface flow routing,
the quality of subsurface water can only be approximated
using a constant concentration.
The precipitation load may be input in some models
(SWMM, HSPF), usually as a constant concentration. Point
source and dry-weather flow (baseflow) loads and concen-
trations can also be input to SWMM, STORM and HSPF to
simulate background conditions. Other quality sources of
potential importance include catchbasins (SWMM) and
snowmelt (SWMM, STORM, HSPF).
Scour and deposition within the sewer system can be very
important in combined sewer systems and some separate
storm sewer systems. The state of the art in simulation of
such processes is poor (Huber, 1985). SWMM offers a
crude but calibratable attempt at simulation of such
processes.
-------
Table 3. Data need* for various quality prediction method*
Method
Data
Potential Source
Unit Load
Constant Concentration
Spreadsheet
Statistical
Regression
Rating Curve
Buildup
Washoff
Mass per time per unit tributary area.
Runoff prediction mechanism (simple to
complex).
Constant concentration for each
constituent.
Simple runoff prediction mechanism.
Constant concentration or concentration
range.
Removal fractions for controls.
Rainfall statistics.
Area, imperviousness. Pollutant median
and CV.
Receiving water characteristics and
statistics.
Storm rainfall, area, imperviousness, land
use.
Measured flow rates/volumes and quality
EMCs/loads.
Loading rates and rate constants.
Street cleaning removals.
Power relationship with runoff.
Derive from constant concentration
and runoff. Literature values.
Existing model; runoff coefficient or
simple method.
NURP; local monitoring.
e.g., runoff coefficient, perhaps as
function of land use.
NURP; local monitoring.
NURP; Schueler (1987); local and state
publications.
NURP; Driscoll et a). (1989);
Woodward-Clyde (1989); EPA SYNOP
model.
NURP; Driscoll (1986); Driscoll et al.
(1989); local monitoring.
Local or generalized data.
Local data.
NURP; local data.
Literature values'.
Literature values.
Literature values.',
'Usually must be calibrated using end-of-pipe monitored quality data.
-------
Section 4.0
Overview of Available Non-urban Modeling Options
As with the options for modeling nonpoint source pollutants
from urban areas, a wide range of techniques are available
for modeling these contributions from non-urban land uses,
from simple annual 'loading functions' to detailed process
simulation models. The key issue in estimating nonpoint pol-
lution loads from a watershed, or parcel of land, is the type
and extent of human activities occurring (or not occurring)
on the land. The same hydrologic, physical, and chemical/
biological processes that determine nonpoint pollutant loads
occur on all land surfaces (and in the soil profile) whether
it is urban, forest, agricultural cropland, pasture, mining,
etc. The relative importance and magnitude of these pro-
cesses, in determining nonpoint loads, will vary among land
use categories and associated human activities. Even within
an urban region, the parameters required for the various
modeling options described in Section 4 will differ for
commercial, industrial, transportation, and various densi-
ties of residential land. Many of these same urban model-
ing options have been used for non-urban land areas with
parameters (e.g., constant concentrations) estimated for the
specific non-urban land use.
The focus of the majority of non-urban nonpoint source
estimation procedures and models has been on agricultural
cropland, although the procedures and models have often
been adapted and applied to many other land use categories.
The agricultural research community, comprised of the U.S. •
Department of Agriculture (including the Soil Conservation
Service and Agricultural Research Service) regional labora-
tories and state universities, have developed a significant
body of knowledge of soil processes and procedures for
estimating runoff and soil erosion that have formed the
'building blocks' for loading functions and nonpoint source
models. This section discusses some of the loading function
procedures available for agricultural and other non-urban
areas and general concepts underlying the more detailed
process simulation models. The individual detailed models
will be discussed briefly in Section 5.2 with additional
details provided in the Appendix.
4.1 Nonpoint Source Loading Functions
The term 'loading function* has been used in the nonpoint
pollution literature to describe simple calculational proce-
dures usually for estimating the average annual load, and
sometimes the storm event load, of a pollutant from an
individual land use category. A number of different loading
functions have been developed and proposed over the past
two decades, the most widely used of which are the EPA
Screening Procedures, also referred to as the EPA Water
Quality Assessment Methodology. These procedures are
described below, followed by a brief discussion of a few
other loading functions in the literature.
4.1.1 EPA Screening Procedures
The EPA Screening Procedures (Mills et al., 1985; Mills
et al., 1982) are a revision and expansion of water quality
assessment procedures initially developed for nondesig-
nated 208 areas (Zison et al., 1977). The Procedures have
been expanded and revised to include consideration of the
accumulation, transport, and .fate of toxic chemicals, in
addition to conventional pollutants included in the ear-
lier versions. The manual includes a separate chapter
describing calculation procedures for estimating nonpoint
loads for urban and non-urban land areas in addition to
chapters on procedures for rivers and streams, impound-
ments, and estuaries. The most recent update includes
consideration of toxics loadings and fate/transport in
groundwater systems.
The procedures for nonpoint load assessments described in
the manual are essentially a compilation and integration of
techniques developed earlier by Midwest Research Institute
(MRI) (McElroy et al., 1976), Amy et al. (1974), Heaney
et al. (1976) and Haith (1980). However, the presentation
of the procedures is well-integrated, supplemented with
additional parameter estimation guidance, and includes
sample calculations. The procedures for non-urban areas
are derived from the MRI loading functions for average
annual estimates based on the Universal Soil Loss Equation
(USLE) (Wischmeier and Smith, 1978), while the storm
event procedures use the Modified USLE (Williams, 1975)
and the SCS Runoff Curve Number procedure (Mockus,
1972) for storm runoff volume. Pollutant concentrations in
runoff and soil, and enrichment ratios are required for
estimating pollutant loads; precipitation contributions of
nutrients can be included. Specific information is included
for estimation of salinity loads in irrigation return flows.
Separate equations are provided for estimating loads for
sorbed pollutants, dissolved pollutants, and partitioned
pollutants (i.e., both sorbed and dissolved phases); this latter
category is primarily for pesticides for which the procedures
were developed by Haith (1980) for storm event loads.
Guidance is provided for estimating all required parameters.
10
-------
The primary strengths and advantages of the EPA Screening
Procedures are as follows:
a. Excellent user documentation and guidance, including
occasional workshops sponsored by the EPA Center
for Exposure Assessment Modeling, in Athens, GA.
b. No computer requirements since the procedures can be
performed on hand calculators; associated programs
have been developed for river quality analyses (Mills
etal.,1979).
c. Loading calculations and procedures can be linked to
water quality procedures in other chapters to assess
water quality impacts of nonpoint source loads.
d. Relatively simple procedures with minimal data re-
quirements that can be satisfied from the user manual
when site-specific data are lacking.
These screening procedures are well suited for general
screening-level assessments; however, they suffer from the
same disadvantages as all such gross estimation techniques.
As with the urban options discussed above, the accuracy of
the loads depends on the accuracy of the user-assumed pol-
lutant concentration; the impacts of management options is
usually represented by a simple, constant 'removal fraction';
snowmelt and associated loadings are not represented; and
calculations can be tedious and time consuming for complex
multi-land use basins. la spite of these disadvantages and
limitations, the EPA Screening Procedures are appropriate
for many types of nonpoint load assessments. They have en-
joyed wide popularity, partly due to the availability of
training workshops sponsored by EPA, and have been ap-
plied in a number of regions, including the Sandusky River
in northern Ohio and the Patuxent, Ware, Chester, and
Occoquan basins in the Chesapeake Bay region (Davis
etal., 1981; Dean etal., 1981a; Dean etal., 1981b).
Although the procedures are quite amenable to a computer-
ized or spreadsheet implementation, to our knowledge no
effort has been made to implement such a format.
4.1.2 Other Loading Functions
As noted earlier, other loading functions have been
developed and proposed by various groups and authors,
though none have the support nor have they demonstrated
the longevity of the EPA Screening Procedures. The
WRENS handbook (Water Resources Evaluation of Non-
point Silvicultural Sources, U.S. Forest Service (1980)) is
similar to the EPA procedures but its focus is directed to the
effects of forestry activities on water quality. The handbook
provides quantitative techniques for estimating potential
changes in streamflow, surface erosion, soil mass movement,
total potential sediment discharge, and water temperature for
comparative analyses of alternative silvicultural management
practices. Runoff and erosion estimation techniques are
similar to those used in the EPA Screening Procedures with
parameters modified for forestry conditions.
Haith and Tubbs (1981) developed watershed loading
functions as a screening tool to evaluate agricultural non-
point source pollution in large watersheds. These functions
also use the SCS Curve Number procedure for runofff esti-
mation and the USLE for erosion; then, based on user-
defined pollutant concentrations in runoff and attached to
sediment, the procedures allow calculation of loadings to
receiving waters. These functions have been added to the
most recent update of the EPA Screening Procedures manual
(i.e., 1985). A validation of the loading functions for a 850
km2 watershed is described by Haith and Shoemaker (1987).
More recently, Li et al. (1989) have proposed loading
functions for estimating the average annual pesticide loads
in surface runoff. They developed regression equations
derived from 100-year simulations of daily pesticide
runoff using the Haith (1980) pesticide model. The regres-
sion equation coefficients are based on pesticide half life
and soil partition coefficients, and are tabulated in the
article for a wide range of values. Two different regressions
are described: one based simply on mean annual soil ero-
sion, and the other based on mean annual soil erosion and
surface runoff volume during the month of pesticide
application.
4.1.3 Discussion
The loading functions discussed above, and other similar
techniques in the literature, differ from the simulation
models primarily in time scale definition and their simpli-
fied, mostly empirical techniques for estimating nonpoint
loads. They are used primarily to estimate average annual
or event loads, and potential changes in these loads with
land use and management practice. These procedures and
associated calculations can usually be performed with a hand
calculator, and with proliferation of personal computers and
advances in computer technology we can expect to see more
and more of these techniques available on PCs. However,
users should not interpret the aura of implementation on a
PC as an improvement in the capabilities, accuracy, or
validity of these techniques. There are significant limitations
in these procedures, because of their simplified nature,
especially for evaluation of the impacts of management prac-
tices. As described above, they often require user-specified
concentrations of pollutants in runoff and/or attached to
sediment; some assume the total pollutant load can be esti-
mated as a function of sediment alone.
Unfortunately, a comprehensive data base, comparable to
the NURP data base for urban areas, does not exist for
estimating the needed input concentrations for the wide
range of non-urban land use categories. Also, there appears
to be much greater variability in runoff concentrations from
non-urban land than from urban land areas; consequently,
extrapolation of concentrations from other sites may be less
appropriate for non-urban land categories. Agricultural crop-
land is especially difficult to represent by single-valued
'representative' concentrations due to differences in crops,
fertilizer applications, tillage practices, agronomic practices,
soils characteristics, etc.
11
-------
In spite of these limitations, loading functions can be
useful for general screening assessments to identify relative
nonpoint contributions under different conditions as long as
their assumptions and limitations are recognized. They are
more popular than the detailed simulation models and have
thus been applied more frequently, primarily because of
their ease of use. Also, the loading functions are a very use-
ful precursor to more detailed modeling studies for general
problem assessment and identification and to determine if
such detailed studies are warranted.
4.2 Simulation Models
The primary differences between nonpoint runoff simulation
models and the loading functions described above relate to
the temporal and spatial detail of the analysis, along with
(usually) a more refined representation of the processes that
determine nonpoint pollutant loadings. Whereas the loading
functions can be used with only a hand calculator (or
spreadsheet), the added detail of most simulation models
requires a computer code, computer facilities, and signifi-
cantly more input data, such as daily rainfall and possibly
other meteorologic timeseries. These models are most often
computerized procedures that perform hydrologic (runoff),
sediment erosion, and pollutant (chemical/ biological) calcu-
lations on short time intervals, usually ranging from one
hour to one day, for many years. The resulting values for
each time interval, e.g., runoff, sediment, pollutant load or
concentration, can be analyzed statistically and/or aggre-
gated to daily, monthly, or annual values for estimates of
nonpoint loadings under the conditions simulated.
As with the urban models, a wide range of nonpoint models
appropriate for non-urban areas are available and have been
used for many different types of land categories. The avail-
able models also cover a large range of complexity depend-
ing on the extent to which hydrologic, sediment erosion, and
chemical/ biological processes are modeled in a mechanistic
manner or based on empirical procedures. Similar to urban
modeling, many of the same simple procedures and assump-
tions used in the loading functions are also incorporated into
a number of simulation models, e.g., USLE, SCS Curve
Number, constant pollutant concentration. Section 5.2
provides brief summaries of a number of the more widely
used and 'operational' non-urban models, along with a brief
discussion of relative strengths and weaknesses; additional
details for each of the models is provided in the Appendix.
In the remainder of this section, we discuss the two major
modeling efforts that have dominated the non-urban (pri-
marily agricultural) nonpoint modeling arena over the past
two decades as a basis for describing the types of modeling
techniques used for non-urban land uses. In addition, we
will briefly discuss the key differences between urban and
non-urban models and identify a number of ongoing model
development efforts.
4.2.1 HSPF and CREAMS Model Development
The 1970's and early 1980's was a period of increasing
pollution, and corresponding development of mathematical
models to both characterize the pollutant loadings and water
quality impacts, and evaluate alternative means of control.
During this period the EPA, through the Athens-ERL, spon-
sored a number of model development and testing efforts
(i.e., PTR, ARM, NPS, WEST) culminating in the HSPF
model—Hydrologic Simulation Program-Fortran (Johanson
et al, 1980). Barnwell and Johanson (1981) discuss the
various model development and testing efforts leading to the
initial release of HSPF in 1980; HSPF is currently in
Release No. 9 (Johanson et al., 1984) as a result of numer-
ous enhancements and code corrections. The focus of the
model development was the ability to represent contributions
of sediment, pesticides, and nutrients from agricultural
areas, and evaluate resulting water quality conditions at the
watershed scale considering both nonpoint contributions and
instream water quality processes. Only the nonpoint capabil-
ities of HSPF (i.e., PERLND and IMPLND modules) are
discussed in this report.
Coincident with the HSPF (and predecessor models) devel-
opment, the U.S. Department of Agriculture through the
Agricultural Research Service (ARS) assembled a group of
ARS scientists to refine, improve, and integrate existing
models into a package for representing runoff, sediment,
nutrient, and pesticide runoff from agricultural fields. The
effort was initiated in 1977 and the resulting CREAMS
model (Chemicals, Runoff, and Erosion from Agricultural
Management Systems) (Knisel, 1980) was first published in
1980. Since its initial release, CREAMS has undergone
testing and application and a companion version, called
GLEAMS (Leonard et al., 1986) has been developed with
special emphasis on vadose zone processes to represent
movement of chemicals to groundwater.
CREAMS and HSPF PERLND are the most detailed, opera-
tional models of agricultural runoff available at the
current time. In many ways, they are more alike than they
are different. Both models simulate runoff and erosion from
field size areas, using different methods, and both simulate
land surface and soil profile chemical/biological processes
(using similar methods) that determine the fate and transport
of pesticides and nutrients. Figure 1 shows the structure of
the various subroutines that comprise the HSPF PERLND
module; note that the Agrichemical Modules of PERLND
perform the soil chemical/biological process simulation.
Figure 2 conceptually shows the structure and processes
simulated for pesticides and nutrients in the ARM model
which was the basis for the HSPF Agrichemical Modules.
Figure 3 includes analogous diagrams for CREAMS, show-
ing the structure of the model and the processes involved in
estimating nutrient losses in runoff and through leaching.
Although the hydrology and sediment algorithms are dif-
ferent for the two models, the soil processes that determine
the availability of chemicals for runoff and leaching are
quite similar; both consider sorption/desorption, plant up-
take, soil transformations (e.g., mineralization, nitrifica-
tion), attenuation/decay, etc. that control the fate and
migration of chemicals in the soil.
12
-------
On the other hand, CREAMS is a product of the agricultural
research community with specific emphasis on representing
soil profile and field-scale processes at the level of detail
appropriate for design of field-based agricultural manage-
ment systems. Thus, CREAMS allows more detailed repre-
sentation of field terraces, drainage systems, field topog-
raphy, etc. and associated sediment erosion processes. A
detailed hydrology option is available, requiring breakpoint
rainfall (i.e., short time interval rainfall, hourly or less), or
the popular SCS Curve Number procedure can be used with
daily rainfall. Because of its field-scale focus, CREAMS is
limited to representing only surface runoff contributions;
subsurface and leaching losses of chemicals are simply
removed from the system. The original CREAMS documen-
tation published in -1980 indicated that an effort to expand
the model to a basin-scale was underway. The SWAM
model (DeCoursey, 1982; Alonso and DeCoursey, 1985)
uses CREAMS as a source area component and adds the
capabilities to consider watershed and basin scale analyses;
however, the model development effort is still underway at
this time and SWAM is not currently considered 'opera-
tional' in terms of documentation and use by non-developers
(DeCoursey, personal communication, 1990). CREAMS has
been used as the source area model by a number of investi-
gators for specific studies (e.g. SWRRB (Williams et al.,
1985), ADAPT (Ward et al., 1988), ARDBSN (Devaurs et
al., 1988)) but no fully integrated, operational package with
CREAMS for use at the basin/watershed scale is available
comparable to HSPF.
4.2.2 Other Non-Urban Nonpoint Models
Many other non-urban nonpoint models exist in the model-
ing community and have been used to varying CREAMS,
but each has been used by the developers, a few, by outside
users. A comprehensive review of all available and relevant
nonpoint models was well beyond the scope of this review
effort. Below we discuss a few additional models, selected
by the authors, that have been applied more often than typi-
cal 'research' models, but they may not fully satisfy the
definition of an 'operational' model described in Section
2.2. The ANSWERS, AGNPS, PRZM, SWRRB/PRS, and
UTM-TOX models are discussed briefly below; a paragraph
description is included in Section 5.2 and additional details
are provided in the Appendix.
The ANSWERS model (Areal Nonpoint Source Watershed
Environment Response Simulation) developed by Beasley
and Huggins (1981) at Purdue University differs from most
other nonpoint models in that it is an 'event'," distributed-
parameter model, as opposed to a continuous, lumped pa-
rameter modeling approach. ANSWERS is designed primar-
ily to simulate single storm events, and requires that the
watershed be subdivided into grid elements with parameter
information provided for each element; most continuous
nonpoint models only require specification of average or
mean parameter values for a watershed or subwatershed
area. The ANSWERS approach imposes greater computa-
tional burden and spatial data requirements, thus limiting
most analyses to single 'design' storms. However, the addi-
tional spatial detail allows greater evaluation of source areas
within a specific watershed area if required by the problem
assessment. ANSWERS is primarily a runoff and sediment
model; the nutrient simulation is based on simple correla-
tions between concentration and sediment yield/runoff
volume; soil nutrient processes are not simulated.
The AGNPS model (Agricultural Nonpoint Source Pollution
Model) developed by the USDA Agricultural Research
Service (Young etal., 1986) is one of the most recent
nonpoint models and thus has limited demonstrated expe-
rience. It is designed to simulate runoff, sediment, and
nutrients from watershed-scale areas for either single event
or continuous periods. It uses a distributed approach, similar
to ANSWERS, whereby the watershed area is divided into
cells, model computations are done at the cell level, and
runoff, sediment, and nutrients are routed from cell to cell
from the watershed boundaries to the outlet. AGNPS uses
the SCS curve number approach combined with a unit hy-
drograph routing procedure, the Modified USLE, and sim-
ple correlations of extraction coefficients of nutrients in
runoff and sediment. AGNPS can accommodate point source
inputs from feedlots, wastewater treatment plants, and user-
defined stream bank and gully erosion. Because of its dis-
tributed approach, its spatial data requirements are similar
to ANSWERS.
The PRZM model (Pesticide Root Zone Model) (Carsel
et al., 1984) was developed by the EPA Athens laboratory
for modeling the fate of pesticides within the crop root
zone, and subsequent leaching to groundwater. However, it
includes a runoff and erosion component based on the SCS
curve number and Modified USLE, respectively. PRZM
represents dissolved, adsorbed, and vapor phase chemical
concentrations in the soil by modeling the processes of
13
-------
surface runoff, erosion, evapotranspiration, plant uptake,
soil temperature, pesticide decay, volatilization, foliar
washoff, advection, dispersion, and decay. The most recent
version of PRZM is included in an integrated root/vadose/
groundwater model called RUSTIC recently released by the
EPA Athens Laboratory (Dean et al., 1989). PRZM is cur-
rently limited to simulation of organic chemicals like
pesticides, but its runoff and erosion components are similar
to many other nonpoint models.
The SWRRB model (Simulator for Water Resources in
Rural Basins) was developed by USDA (Williams et al.,
1985; Arnold et al., 1989) for basin scale water quality
modeling. Its runoff (SCS curve number) and erosion
(Modified USLE) components are similar to the other non-
point models, but SWRRB also includes channel processes
and subsurface flow components to allow representation of
large basin areas. It performs calculations on a daily
Umestep, and simulates hydrology, crop growth, sediment
erosion, sediment transport, and nitrogen/phosphorus/
pesticide movement in runoff. Its nutrient and pesticide
capabilities are derived from CREAMS; these are the most
recent additions to the model and they are still undergoing
testing and validation by the developers.
The UTM-TOX model (Unified Transport Model for Toxic
Materials) (Patterson et al., 1983), developed by the Oak
Ridge National Laboratory for the U.S. EPA Office of
Pesticides and Toxic Substances, is a multimedia model that
combines hydrologic, atmospheric, and sediment transport
in one computer code. It is similar to HSPF in many ways,
in terms of its comprehensive scope; its representation of
soil, land surface, and channel processes; and its use of the
Stanford Watershed Model as its hydrologic module. UTM-
TOX provides a more detailed simulation of soil-plant
processes, includes atmospheric transport and deposition,
and is designed primarily for organic chemicals; no specific
capabilities are included for nutrients or agricultural condi-
tions. UTM-TOX, to our knowledge; has had limited appli-
cation, possibly because of its relatively complex nature and
the lack of user support.
14
-------
Figure 1. Subroutine structure for HSPF PERLND.
15
-------
N2 I I PLHT-N I
T m m. f
1
i
KH4
\
,K
-------
HYDROLOGY I EROSION
MODEL 1 MODEL
, 1 ' -
\ 7
\ NUTR 1 ENT
\ DATA
/
\ * f
1 NUTR 1 ENT 1
I MODEL 1
^ \
N ** P NUTRIENT
' N LEACH 1 NG
RUNOFF
N b P
WITH
SEDIMENT
F ow diagram of Input and output -for the nutrient model
N IN
RAINFALL
*
- 1 IN*ll_TnATION
N LEACHED
DEEPER 1 NTO
SOIL
Diagram for estimating nutrient
SOLUBLE N
«««.£_ N 8. P
RUNOFF
osses i n runoff
FERT 1 LZER
RESIDUES
WASTES
Inn ltratlon>v^ _X^ Incorporation
sol 1 water and ^^ jf^^
SOIL temperature ^
ORGANIC f NITRATE nooT
N.TROGEN mln«-.ll»*lon _ ZONE
denltr Iflcatlon 1
so l water
t PLANT . :'
. p,=nx growth _ NITROGEN
1 . IMT perco at I on
1 .• _ T
1 NITROGEN ^ | NITRATE
GASES LEACHED
Diagram for estimating nitrate leaching
Figure 3. Nutrient simulation in CREAMS.
17
-------
Section 5.0
Nonpoint Source Runoff Quality Simulation Models and Methods
5.1 Urban Runoff Quality Models
5.1.1 Introduction
Four models (USGS, HSPF, STORM, SWMM) will be
described briefly at this point; extensive details about the
four models may be found in the Appendix. These four
models essentially make up the best choice of full-scale
simulation models for urban areas. Other models have been
adapted from SWMM (e.g., FHWA, RUNQUAL) and
STORM (e.g., SEMSTORM) and given modified names,
but the principles are fairly similar. Still other models, such
as the Illinois State Water Survey ILLUDAS model
(Terstriep and Stall, 1974) have sometimes been adapted for
water quality simulation for a specific project (Noel and
Terstriep, 1982), but such modifications and quality pro-
cedures remain undocumented, and the quality model can be
considered not operational. At least two European models
are available that simulate water quality. These are de-
scribed briefly at the end of this section. Finally, there are
many models well known in the hydrologic literature, such
as those developed by the HEC and SCS, that might be use-
ful in the hydrologic aspect of water quality studies but that
do not simulate water quality directly. This review is limited
to models that directly simulate water quality. A general com-
parison of model attributes is given in Table 4. This table
includes the EPA Statistical Method since with the publica-
tion of the recent FHWA study, it can be considered a for-
malized procedure (Driscoll et al., 1989). The constant con-
centration, spreadsheet, and regression approaches described
earlier are more generic in nature and not included in Table
4, but their attributes were provided in the earlier text.
5.1.2 DR3M-QUAL
A version of the USGS Distributed Routing Rainfall Runoff
Model that includes quality simulation (DR3M-QUAL) js
available from that agency for general use (Alley and Smith,
1982a, 1982b). Runoff generation and subsequent routing
use the kinematic wave method, and parameter estimation
assistance is included in the model. Quality is simulated
using buildup and washoff functions, with settling of solids
in storage units dependent on a particle size distribution.
The model has been used in some of the NURP studies that
were conducted by the USGS (see the Appendix and Alley,
1986). No microcomputer version is available.
5.1,3 HSPF
The Hydrological Simulation Program—Fortran (HSPF) is
the culmination of hydrologic routines that originated with
flie Stanford Watershed Model in 1966 and eventually incorpo-
rated many nonpoint source modeling efforts of the EPA
Athens laboratory (Johanson et al., 1984). This model has been
widely used for non-urban nonpoint source modeling and is de-
scribed additionally in that section of this report as well as in
detail in the Appendix. The user's manual includes informa-
tion on all hydrologic and water quality routines, including
the IMPLND (impervious land) segment for use in urban
area. Additional guidelines for application are provided by
Donigian et al. (1984). The model has special provisions for
management of time series that result from continuous simu-
lation. A microcomputer version is available.
5.1.4 STORM
The first significant use of continuous simulation in urban
hydrology came with the Storage, Treatment, Overflow,
Runoff Model (STORM), developed by the Corps of Engi-
neers Hydrologic Engineering Center (HEC, 1977; Roesner
et al., 1974) for application to the San Francisco master
plan for CSO pollution abatement. The HEC also provides
application guidelines (Abbott, 1977). The current version
includes dry-weather flow input for combined sewer simula-
tion. The support of the HEC led to the wide use of
STORM for planning purposes, especially for evaluation of
the trade-off between treatment and storage as control
options for CSOs (e.g., Heaney et al., 1977). Statistics of
long-term runoff and quality time series permit optimization
of control measures.
STORM utilizes simple runoff coefficient, SCS and unit
hydrograph methods for generation of hourly runoff depths
from hourly rainfall inputs. No flow routing is performed,
but runoff may be routed through a constant-rate treatment
device, with excess flow diverted to a storage device. Flows
exceeding the treatment rate cause CSOs when storage is
filled. The build-up and wash-off formulations are used for
simulation of six pre-specified pollutants. However, the
model can be manipulated to provide loads for arbitrary
conservative pollutants (e.g., Najarian et al., 1986). The
model is hampered somewhat by lack of an operational mi-
crocomputer version. However, various individual consul-
tants have adapted the nonproprietary code to their own
project needs.
18
-------
5.1.5 SWMM
The original version of the Storm Water Management Model
(SWMM) was developed for EPA as single-event model
specifically for the analysis of CSOs (Metcalf and Eddy et
al., 1971), but its scope has vastly broadened since the
original release. Version 4 (Huber and Dickinson, 1988;
Roesner et al., 1988) of the model performs both continuous
and single-event simulation throughout the whole model, can
simulate backwater, surcharging, pressure flow and looped
connections (by solving the complete dynamic wave equa-
tions) in its Extran Block, and has a variety of options for
quality simulation, including traditional build-up and wash-
off formulations as well as rating curves and regression
techniques. Subsurface flow routing (constant quality) may
be performed in the Runoff Block in addition to surface
quantity and quality routing, and treatment devices may be
simulated in the Storage/Treatment Block using removal
functions and sedimentation theory. A hydraulic design
routine is included for sizing of pipes, and a variety of
regulator devices may be simulated, including orifices (fixed
and variable), weirs, pumps, and storage. A bibliography of
SWMM usage is available (Huber et al., 1986) that contains
many references to case studies.
SWMM is segmented into the Runoff, Transport, Extran,
Storage/Treatment and Statistics blocks for rainfall-runoff,
routing, and statistical computations. Water quality may be
simulated in all blocks except Extran, and metric units are
optional. Since the model is non-proprietary, portions have
been adapted for various specific purposes and locales by
individual consultants and other federal agencies, e.g.,
FHWA. A microcomputer version is available.
5.1.6 Two European Models
The four U.S. models discussed above do not take
advantage of graphics and other "user-friendly" capabilities
of microcomputers. Two well-known and commercially-
available European models, MOUSE and Wallingford, are
excellent examples of application of the full power of the
microcomputer when used in conjunction with recent pro-
gramming languages and graphics hardware and software.
Both models feature menu-driven pre-processors for data
input, graphical display and interactive editing of catchment
boundaries and sewer networks, and post-processing of pre-
dicted hydrographs and pollutographs, including graphical
displays and statistical analysis. Although the quality algo-
rithms are relatively simple, the hydrologic and hydraulic
'components of both models are relatively sophisticated. The
cost of each model is approximately $15,000, including
training and documentation but not the source code.
The Danish Hydraulic Institute, in cooperation with various
other laboratories and private software firms has produced
the MOUSE (Modeling of Urban Sewers) model. Included
in the package are modules for generation of runoff from
rainfall, sewer routing (the SI IS model, comparable to the
SWMM Extran Block), and a simple quality routine that
uses the constant concentration approach (Jacobsen et al.,
1984; Johansen et al., 1984). Further information on
MOUSE is available from Danish Hydraulic Institute, Agern
Alle 5, DK-2970 Hersholm, Denmark.
The Wallingford model is maintained by Hydraulics
Research Ltd. in the United Kingdom. It also consists of a
cluster of modules, including runoff generation from rainfall
(WASSP), simple and fully-dynamic sewer routing
(WALLRUS and SPIDA, respectively), and a quality
routine (MOSQITO)featuring processes similar to those in
SWMM (Henderson and Moys, 1987). Further information
on the group of Wallingford models is available from
Hydraulics Research Ltd., Wallingford, Oxfordshire OX10
8BA, United Kingdom.
5.2 Non-urban Runoff Quality Models and
Methods
In this section we provide brief summaries of the primary
non-urban runoff quality models reviewed; as noted earlier
additional details on each model are provided in the
Appendix. Below summaries are presented for HSPF,
CREAMS/GLEAMS, ANSWERS, AGNPS, PRZM,
SWRRB, and UTM-TOX, and Table 5 shows a comparison
of selected model attributes and capabilities.
5.2.1 HSPF
The Hydrological Simulation Program—FORTRAN (HSPF)
(Johanson et al., 1981; 1984) is a comprehensive package
for simulation of watershed hydrology and water quality for
both conventional and toxic organic pollutants. HSPF incor-
porates the watershed scale ARM and NFS models into a
basin-scale analysis framework that includes fate and
transport in one-dimensional stream channels. It is the only
comprehensive model of watershed hydrology and water
quality that allows the integrated simulation of land and soil
contaminant runoff processes with instream hydraulic, water
temperature, sediment transport, nutrient, and sediment-
chemical interactions. The runoff quality capabilities include
both simple relationships (i.e. empirical buildup/washoff,
constant concentrations) and detailed soil process options
(i.e., leaching, sorption, soil attenuation and soil nutrient
transformations).
The result of this simulation is a time history of the runoff
flow rate, sediment load, nutrient, pesticide, and/or user-
specified pollutant concentrations, along with a time history
of water quantity and quality at any point in a watershed.
HSPF simulates three sediment types (sand, silt, and clay)
in addition to a single organic chemical and transformation
products of that chemical. The instream nutrient processes
include DO, BOD, nitrogen and phosphorus reactions, pH,
phytoplankton, zooplankton, and benthic algae.
The' organic chemical transfer and reaction processes
included are hydrolysis, oxidation, photolysis, biode-
gradation, a volatilization, and sorption. Sorption is modeled
as a first-order kinetic process in which the user must
specify a desorption rate and an equilibrium partition
coefficient for each of the three solid types. Resuspension
and settling of silts and clays (cohesive solids) are defined
19
-------
in terms of shear stress at the sediment-water interface. For
sands, the capacity of the system to transport sand at a par-
ticular flow is calculated and resuspension or settling is
defined by the difference between the sand in suspension
and the capacity. Calibration of the model requires data for
each of the three solids types. Benthic exchange is modeled
as sorption/desorption and desorption/scour with surficial
benthic sediments. Underlying sediment and pore water are
not modeled.
5.2.2 CREAMS/GLEAMS
Chemicals, Runoff, and Erosion from Agricultural Manage-
ment Systems (CREAMS) was developed by the U.S.
Department of Agriculture—Agricultural Research Service
(Knisel, 1980; Leonard and Ferreira, 1984) for the analysis
of agricultural best management practices for pollution
control. CREAMS is a field scale model that uses separate
hydrology, erosion, and chemistry submodels connected
together by pass files.
Runoff volume, peak flow, infiltration, evapotranspiration,
soil water content, and percolation are computed on a daily
basis. If detailed precipitation data are available then
infiltration is calculated at histogram breakpoints. Daily
erosion and sediment yield, including particle size distri-
bution, are estimated at the edge of the field. Plant nutrients
and pesticides are simulated and storm load and average
concentrations of sediment-associated and dissolved chemi-
cals are determined in the runoff, sediment, and percolation
through the root zone (Leonard and Knisel, 1984).
User defined management activities can be simulated by
CREAMS. These activities include aerial spraying (foliar or
soil directed) or soil incorporation of pesticides, animal
waste management, and agricultural best management prac-
tices (minimum tillage, terracing, etc.).
Calibration is not specifically required for CREAMS
simulation, but is usually desirable. The model provides
accurate representation of the various soil processes. Most
of the CREAMS parameter values are physically measur-
able. The model has the capability of simulating 20 pesti-
cides at one time.
Groundwater Loading Effects of Agricultural Management
Systems (GLEAMS) was developed by the United States
Department of Agriculture—Agriculture Research Service
(Leonard et al., 1987) to utilize the management oriented
physically based CREAMS model (Knisel, 1980) and incor-
porate a component for vertical flux of pesticides. GLEAMS
is the vadose zone component of the CREAMS model.
GLEAMS consists of three major components namely hy-
drology, erosion/sediment yield, and pesticides. Precipi-
tation is partitioned between surface runoff and infiltration
and water balance computations are done on a daily basis.
Surface runoff is estimated using the Soil Conservation
Service Curve Number Method as modified by Williams and
Nicks (1982). The soil is divided into various layers, with
a minimum of 3 and a maximum of 12 layers of variable
thickness are used for water and pesticide routing (Knisel
et al., 1989)
5.2.3 ANSWERS
Areal Nonpoint Source Watershed Environment Response
Simulation (ANSWERS) was developed at the Agricultural
Engineering Department of Purdue University (Beasley and
Huggins, 1981). It is an event based, distributed parame-
ter model capable of predicting the hydrologic and ero-
sion response of agricultural watersheds. Application of
ANSWERS requires that the watershed to be subdivided into
a grid of square elements. Each element must be small
enough so that all important parameter values within its
boundaries are uniform. For a practical application element
sizes range from one to four hectares. Within each element
the model simulates the processes of interception, infil-
tration, surface storage, surface flow, subsurface drainage,
and sediment drainage, and sediment detachment, transport,
and deposition. The output from one element then becomes
a source of input to an adjacent element.
As the model is based on a modular program structure it
allows easier modification of existing program code and/or
addition of user supplied algorithms. Model parameter
values are allowed to vary between elements, thus, any
degree of spatial variability within the watershed is easily
represented.
**
Nutrients (nitrogen and phosphorus) are simulated using
correlation relationships between chemical concentrations,
sediment yield and runoff volume. A research version
(Amin-Sichani, 1982) of the model uses "clay enrichment"
information and a very descriptive phosphorus fate model to
predict total, particulate, and soluble phosphorus yields.
5.2.4 AGNPS
Agricultural Nonpoint Source Pollution Model (AGNPS)
was developed by the U.S. Department of Agricul-
ture—Agriculture Research Service (Young et al., 1986) to
obtain uniform and accurate estimates of runoff quality with
primary emphasis on nutrients and sediments and to com-
pare the effects of various pollution control practices that
could be incorporated into the management of watersheds.
The AGNPS model simulates sediments and nutrients from
agricultural watersheds for a single storm event or for
continuous simulation. Watersheds examined by AGNPS -
must be divided into square working areas called cells.
Grouping of cells results in the formation of subwatersheds,
which can be individually examined. The output from the
model can be used to compare the watershed examined against
other watersheds to point sources of water quality problems,
and to investigate possible solutions to these problems.
AGNPS is also capable of handling point source inputs from
feedlots, waste water treatment plant discharges, and stream
bank and gully erosion (user specified). In the model,
pollutants are routed from the top of the watershed to the
outlet in a series of steps so that flow and water quality at
20
-------
any point in the watershed may be examined. The Modified
Universal Soil Loss Equation is used for predicting soil
erosion, and a unit hydrograph approach used for the flow
in the watershed. Erosion is predicted in five different
particle sizes namely sand, silt, clay, small aggregates, and
large aggregates.
The pollutant transport portion is subdivided into one part
handling soluble pollutants and another part handling sedi-
ment attached pollutants. The methods used to predict nitrogen
and phosphorus yields from the watershed and individual cells
were developed by Frere et al. (1980) and are also used in
CREAMS (Knisel, 1980). The nitrogen and phosphorus
calculations are performed using relationships between
chemical concentration, sediment yield and runoff volume.
Data needed for the model can be classified into two
categories: watershed data and cell data. Watershed data
includes information applying to the entire watershed which
would include watershed size, number of cells in the water-
shed, and if running for a single storm event then the storm
intensity. The cell data includes information on the parame-
ters based on the land practices in the cell.
Additional model components that are under development
are unsaturated/saturated zone routines, economic analysis,
and linkage to Geographic Information System.
5.2.5 PRZM
Pesticide Root Zone Model (PRZM) was developed at the
U.S. EPA Environmental Research Laboratory in Athens,
Georgia by Carsel et al. (1984). It is a one-dimensional,
dynamic, compartmental model that can be used to simulate
chemical movement in unsaturated zone within and immedi-
ately below the plant root zone. The model is divided into
two major components namely, the hydrology (and hydrau-
lics) and chemical transport. The hydrology component
which calculates runoff and erosion is based upon the Soil
Conservation Service curve number procedure and the
Universal Soil Loss Equation respectively. Evapotranspira-
tion is estimated directly from pan evaporation or by an
empirical formula if pan evaporation data is not available.
Soil-water capacity terms including field capacity, wilting
point, and saturation water content are used for simulating
water movement within the unsaturated zone. Irrigation
application is also within model capabilities.
Pesticide application on soil or on the plant foliage are
considered in the chemical transport simulation. Dissolved,
adsorbed, and vapor-phase concentrations in the soil are
estimated by simultaneously considering the processes of
pesticide uptake by plants, surface runoff, erosion, decay,
volatilization, foliar washoff, advection, dispersion, and
retardation. The user has two options to solve the transport
equations using the original backward difference implicit
scheme or the method of characteristics (Dean et al., 1989).
As the model is dynamic it allows considerations of pulse
loads.
PRZM is an integral part of a unsaturated/saturated zone
model RUSTIC (Dean et al., 1989). RUSTIC (Risk of
Unsaturated/Saturated Transport and Transformation of
Chemical Concentrations) links three subordinate models in
order to predict pesticide fate and transport through the crop
root zone, and saturated zone to drinking water wells
through PRZM, VADOFT, SAFTMOD.
VADOFT is a one-dimensional finite element model which
solves Richard's equation for water flow in the unsaturated
zone. VADOFT can also simulate the fate and transport of
two parent and two daughter products. SAFTMOD is a two-
dimensional finite element model which simulates flow and
transport in the saturated zone in either an X-Y or X-Z
configuration. The three codes PRZM, VADOFT, and
SAFTMOD are linked together through an execution super-
visor which allows users to build models for site specific
situation. In order to perform exposure assessments, the
code is equipped with a Monte Carlo pre and post processor
(Deanetal., 1989).
5.2.6 SWRRB
Simulator for Water Resources in Rural Basins (SWRRB)
was developed by Williams et al. (1985), and Arnold et al.
(1989) for evaluating basin scale water quality. SWRRB
operates on a daily time step and simulates weather, hydrol-
ogy, crop growth, sedimentation, and nitrogen, phospho-
rous, and pesticide movement. The model was developed by
modifying the CREAMS (Knisel, 1980) daily rainfall
hydrology model for application to large, complex, rural
basins.
Surface runoff is calculated using the Soil Conservation
Service Curve Number technique. Sediment yield is com-
puted for each basin by using the Modified Universal Soil
Loss Equation (Williams and Berndt, 1977). The channel
and floodplain sediment routing model is composed of two
components operating simultaneously (deposition and degra-
dation). Degradation is based on Bagnold's stream power
concept and deposition is based on the fall velocity of the
sediment particles (Arnold et al., 1989).
Return flow is calculated as a function of soil water content
and return flow travel time. The percolation component uses
a storage routing model combined with a crack flow model
to predict the flow through the root zone. The crop growth
model (Arnold et al., 1989) computes total biomass each
day during the growing season as a function of solar radia-
tion and leaf area index.
The pollutant transport portion is subdivided into one part
handling soluble pollutants and another part handling sedi-
ment attached pollutants. The methods used to predict nitro-
gen and phosphorus yields from the rural basins are adopted
from CREAMS (Knisel, 1980), The nitrogen and phospho-
rus calculations are performed using relationships between
chemical concentration, sediment yield and runoff volume.
The nutrient capabilities are still undergoing testing and
validation at this time.
21
-------
The pesticide component is directly taken from Hoist and
Kutney (1989) and is a modification of the CREAMS (Smith
and Williams, 1980) pesticide model. The amount of pesti-
cide reaching the ground or plants is based on a pesticide
application efficiency factor. Empirical equations are used
for calculating pesticide washoff which are based on thresh-
old rainfall amount. Pesticide decay from the plants and the
soil are predicted using exponential functions based on the
decay constant for pesticide in the soil, and half life of
pesticide on foliar residue.
The Pesticide Runoff Simulator (PRS) was developed for the
U.S. EPA Office of Pesticide and Toxic Substances by
Computer Sciences Corporation (1980) to simulate pesticide
runoff and adsorption into the soil on small agricultural
watersheds. PRS is based on SWRRB. Thus, the PRS
hydrology and sediment simulation is based on the USDA
CREAMS model, and the SCS curve number technique is
used to predict surface runoff. Sediment yield is simulated
using a modified version of the Universal Soil Loss Equa-
tion and a sediment routing model.
The pesticide component of PRS is a modified version of
the CREAMS pesticide model. Pesticide application (foliar
and soil applied) can be removed by atmospheric loss, wash
off by rainfall, and leaching into the soil. Pesticide yield is
divided into a soluble fraction and an adsorbed phase based
on an enrichment ratio.
The model includes a built in weather generator based on tem-
perature, solar radiation, and precipitation statistics. Calibration
is not specifically required, but is usually desirable.
5.2.7 UTM-TOX
Unified Transport Model for Toxic Materials (UTM-TOX)
was developed by Oak Ridge National Laboratory for the
U.S. EPA Office of Pesticides and Toxic Substances,
Washington, D.C. (Patterson et al., 1983). UTM-TOX is a
multimedia model that combines hydrologic, atmospheric,
and sediment transport in one computer code. The model
calculates rates of flux of a chemical from release to the
atmosphere, through deposition on a watershed, infiltration
and runoff from the soil, to flow in a stream channel and
associated sediment transport. From these calculations mass
balances can be established, chemical budgets made, and
concentrations in the environment estimated. The atmos-
pheric transport model (ATM) portion of UTM-TOX is a
Gaussian plume model that calculates dispersion of pol-
lutants emitted from point (stack), area, or line sources.
ATM operates on a monthly time step, which is longer than
the hydrologic portion of the model and results in the use of
an average chemical deposition falling on the watershed.
The Terrestrial Ecology and Hydrology Model (TEHM)
describes soil-plant water fluxes, interception, infiltration,
and storm and groundwater flow. The hydrologic portion of
the model is from the Wisconsin Hydrologic Transport
Model (WHTM), which is a modified version of the Stan-
ford Watershed Model (SWM). WHTM includes all of the
hydrologic processes of the SWM and also simulates soluble
chemical movement, litter and vegetation interception of the
chemical, erosion of sorbed chemical, chemical degradation
in soil and litter, and sorption in top layers of the soil.
Stream transport includes transfer between three sediment
components (suspended, bed, and resident bed).
5.3 Discussion
The models discussed briefly here (and more extensively in
the Appendix) do not represent all of the modeling options
available for runoff quality simulation, but they are certainly
the most notable, widely used and most operational. Selec-
tion from among these models is often made on the basis of
personal preference and familiarity, in addition to needed
model capabilities. For example, for urban modeling various
in-house versions of STORM are still used by consultants
even though the "official" HEC version has not been
updated since 1977, because these versions have been
adapted to the needs of the firm and because STORM has
proven to provide useful continuous simulation results.
The USGS DR3M-QUAL model has perhaps been used the
least by persons outside that agency, but has worked
satisfactorily in several applications documented in the
Appendix. Support for both STORM and DR3M-QUAL
would be minimal. CREAMS has been used most extensive-
ly for field-scale agricultural runoff modeling because of its
agricultural origins and ties to the agricultural research
community.
HSPF and SWMM are probably the most versatile and most
widely applicable of the models, with the nod to SWMM if
the urban hydrology and hydraulics must be simulated in
detail. On the other hand, the water quality routines in
HSPF for sediment erosion, pollutant interaction and
groundwater quality are superior in HSPF, and the capabil-
ity to efficiently handle all types of land uses and pollutant
sources, (including urban and agriculture, point and non-
point), is a definite advantage when needed for large com-
plex basins. Both models appear somewhat overwhelming in
terms of size to the novice user, but only the components of
interest of either model need be used in a given study, and
the catchment schematization can often be coarse for pur-
poses of simulation of water quality at the outlet. Thus,
although the installation of these models on a microcomputer
may occupy several megabytes of a hard disk, they may be
applied in simple ways (i.e., applied to a simplified
schematization of the catchment) with a significant reduction
in data requirements. Furthermore, the several quality
modeling options within SWMM permit simple conceptual
water quality simulation using constant concentration and
rating curves as well as the more formidable buildup-
washoff methods. Similarly for HSPF, the ability to use the
simple SWMM-type formulations for urban and non-
agricultural areas, and detailed soil/runoff process
simulation for agricultural areas provides the user with great
flexibility in representing the watershed system.
Continuing model development and testing within the agri-
cultural research community will likely lead to further en-
hancements and development of many of the agricultural
22
-------
models, like CREAMS, SWRKB, and AGNPS. In fact, USDA
has supported, and continues to support, a wide range of model
development work in individual research facilities, many of
which are (or at least appear to be) very similar in terms of
using similar algorithms or model formulations (e.g. EPIC
(Williams et al., 1984), Opus (V. Ferreira, 1989, personal
communication), SWAM (DeCoursey, 1982). The SWRRB
development effort appears to be focussing in on a middle
ground (in terms of complexity) between HSPF and the de-
tailed field-scale models which are limited to small areas; its
use of daily rainfall, as opposed to smaller time interval
measurements (usually hourly is needed for HSPF) is seen
as a definite advantage by many users. However, most of
these efforts still focus primarily on agricultural areas, with
limited abilities to be used in large, complex multi-land use
basins.
Regression, spreadsheet, statistical methods, and loading
functions are most useful as screening tools. Indeed if the
Statistical Method or EPA's Screening Procedures, indicate
that there should be no water quality problem (as defined by
exceedance of a specified concentration level with a speci-
fied frequency), then more detailed water quality simulation
may not be required at all. If sensitivity analyses and
'worst-case' evaluations further support the conclusions,
detailed water quality modeling will probably not be needed.
23
-------
Table 4. Comparison of urban model attribute*
Attribute
Sponsoring agancy
Simulation type*
No. pollutants
Rainfall/runoff
analysis
Sewer system flow
routing
Full, dynamic flow
routing equations
Surcharge
Regulators, overflow
Model:
DR3M-QUAL
uses
C.SE
4
Y
Y
N
Y'
N
HSPF
EPA
C,SE
10
Y
Y
N
N
N
Statistical*
EPA
N/A
Any
N°
N/A
N/A
N/A
N/A
STORM
HEC
C
6
Y
N
N
N
Y
SWMM
EPA
C.SE
10
Y
Y
Yd
Y"
Y
structures, e.g.,
weirs, orificies, etc,
Special solids routines
Storage analysis
Treatment analysis
Suitable for screening
(S), design (D)
Available on micro-
computer
Data and personnel
requirements'1
Overall model
complexity1
Y
Y
Y
S,D
N
Medium
Medium
Y
Y
Y
S,D
Y
High
High
N
Y'
Y'
S
Y»
Medium
Medium
N
Y
Y
S
N
Low
Medium
Y
Y
Y
S,D
Y
High
High
'EPA procedure.
kC - continuous simulation, SE = single event simulation.
•Runoff coefficient used to obtain runoff volumes.
*Futl dynamic equations and surcharge calculations only in Extran Block of SWMM.
•Surcharge simulated by storing excess inflow at upstream end of pipe. Pressure flow not simulated.
'Storage and treatment analyzed analytically.
•FHWA study, Driscoll et al. (1989)
''General requirements for model installation, familiarization, data require ments, etc. To be interaretted onlv verv
generally. '
'Reflection of general size and overall model capabilities. Note that complex models may still be used to simulate
very simple systems with attendant minimal data requirements.
24
-------
Tabl* 5. Comparison of non-urban model attributes
Attribute
Sponsoring Agency
Simulation type
Rainfall/Runoff
analysis
Erosion Modeling
Pesticides
'Nutrients
User-Defined
Constituents
Soil Processes
Pesticides
Nutrients
Multiple Land Type
Capability
Instream Water
Quality Simulation
Available- on
Micro-computer
Data and Personnel
Requirements
Overall Model
Complexity -
x AGNPS
USDA
C.SE
Y
Y
Y
Y
N
N
N
Y
N
Y
M
M
ANSWERS
Purdue
SE
Y
Y
N
'' ' ••'• Y -
N
N
N
Y
N
Y
M/H
M
CREAMS
USDA
C,SE
Y
Y
Y
Y
N
-
Y
Y
N
N
Y
H
H
HSPF
EPA
C,SE
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
H
H
PRZM
EPA
C
Y
Y
Y
N
N
Y
N
N
N
Y
M
M
SWRRB
USDA
C
Y
• Y
Y
Y
N
Y
Y
Y
N
Y
M
. M/H
;
UTM-TOX
ORNL&
EPA
C,SE
Y
.
y
N
N
Y
N
N
Y
Y
... •-'- -:.
N
H
H
Y - yes, N = no, M = Moderate, H = High
C = Continuous, SE = Storm Event
25
-------
Section 6.0
Brief Case Studies
How are quality processes being simulated in studies of
urban and rural runoff quality problems? Below, the authors
draw upon their personal knowledge of a few ongoing and
recently completed studies (listed alphabetically).
6.1 Urban Model Applications
6.1.1 Boston
CH2M-HH1 (Gainesville, FL) used continuous SWMM
modeling for the development of TSS and BOD loads from
CSOs to Boston Harbor (Morrissey and Harleman, 1989).
After first estimates from Sartor and Boyd (1972) and Pitt
(1979), buildup and washoff functions were calibrated to
estimates of annual totals based on monitoring. A "typical"
five years of hourly precipitation data selected from 40
years of available record were input to SWMM to develop
CSO loads, and the effectiveness of street cleaning and
catchbasin cleaning BMPs was studied using the model (V.
Adderly, CH2M-Hill, Inc., Gainesville, FL, Personal Com-
munication, 1989).
6.1.2 Delevan Lake. Wisconsin
A joint project of the USGS (Madison) and the University
of Wisconsin investigated suspended solids and phosphorus
loads to 1800-ac Delevan Lake in southeastern Wisconsin
(Walker et al.t 1989). A spreadsheet approach was imple-
mented using Multiplan, with unit load estimates for the
surrounding basin (agricultural, urban, industrial). The
Universal Soil Loss Equation was used for sediment loads
from agricultural areas. Some calibration was possible using
measurements on four tributaries. The cost-effectiveness of
agricultural control options was evaluated based on cost
estimates for various agricultural BMPs.
6.1.3 Hackensack River Basin
Pollution problems in the lower and estuarine portion of
the Hackensack River in New Jersey are being studied by
Najarian and Associates (Eatontown, NJ) using SWMM
coupled with monitoring data from four CSO and five storm
sewered areas (Huang and DiLorenzo, 1990; Najarian et al.,
1990). The pollutants of primary interest are BOD and
ammonia for input to a dynamic receiving water quality
model of the river and estuary, with emphasis upon the rela-
tive contributions of CSOs, separate storm sewered areas
and point sources. Although rating curve results were very
good predictors for the monitored catchments from which
they were derived, it was found that they could not be
extrapolated (transferred) to the ungaged catchments. Hence,
Michaelis-Menton buildup and exponential washoff parame-
ters were calibrated for the basins and transferable gen-
eralized coefficients developed as a function of land use.
Intra-storm variations were simulated in order to use
SWMM to drive a short time increment dynamic model of
the river and estuary.
6.1.4 Jacksonville
Camp, Dresser and McKee (Jacksonville) will use SWMM
for quantity predictions and both a spreadsheet and SWMM
or STORM with constant concentrations for load estimates
to the St. Johns River (Camp, Dresser and McKee, Inc.,
1989). The constant concentrations are based on NURP data
and limited Florida data. If SWMM or STORM is used to
drive a receiving water quality model for the river, local
data will be used for better calibration. At the moment,
CDM feels that both quantity and quality control options can
be compared on the basis of present data, with a minimum
of expensive local sampling.
6.1.5 Orlando
To help alleviate nonpoint source pollution to lakes down-
stream from the Boggy Creek Watershed south of Orlando,
Camp, Dresser and McKee (Orlando) developed a spread-
sheet model to assess nutrient loadings resulting from
existing and future land uses (Camp, Dresser and McKee,
1987). Runoff coefficients were calibrated to match mea-
sured creek runoff volumes, and EMCs as a function of land
use were estimated from sampling in Orlando and Tampa.
An overall calibration factor was used to obtain agreement
between the total estimated TN and TP loads produced by
the product of flows and EMCs for the various land uses
and measured annual nutrient loads in Boggy Creek. Thus,
relative contributions from various land uses remained the.
same while the overall loads were adjusted. BMP removal
efficiencies were applied in conjunction with changing land
uses to obtain control strategies for future watershed
development.
6.1.6 Providence
\
SWMM is being used by Greeley and Hanson (Philadelphia)
to simulate CSO loads from Providence using three moni-
tored storms for calibration and verification (R. Janga,
Greeley and Hansen, Inc., Philadelphia, PA, Personal
26
-------
Communication, 1989). Quality is being simulated using
constant concentration in the Runoff Block and the quality
routing routines in the Transport Block. SWMM may be
used to drive a receiving water model before the project is
completed. Extran is also being used to simulate some of the
overflow hydraulics.
/ --'-"-'
6.1.7 San Francisco Bay
Woodward-Clyde (Oakland) is using SWMM to simulate loads
from the Santa Clara Valley into South San Francisco Bay
(P. Mangarella, Woodward-Clyde Consultants, Oakland,
CA, Personal Communication, 1989). Measured runoff and
flow data are being used to calibrate the Runoff Block
quantity routines, and constant concentrations are being used
(no buildup or washoff) based on one year of monitoring of
a selection of land use types. The model may not be used to
drive a receiving water model but it will be used to compare
alternatives to reduce loads of toxics to the Bay.
6.1.8 Tallahassee
The Northwest Florida Water Management District
(Havana, FL) is using SWMM to develop the stormwater
master plan for Tallahassee and Leon County (R. Ortega,
Northwest Florida Water Management District, Havana, FL,
Personal Communication, 1989). Extensive use of the model
has already been made for quantity predictions. The present
plan is to develop rating curve relationships on the basis of
considerable quality monitoring data gathered during the
study for input into SWMM. BMPs will also be studied with
the model, especially storage. Final control decisions will be
made on the basis of 28-year SWMM simulations using 15-
min rainfall data.
6.2 Non-urban Model Applications
6.2.1 Chesapeake Bay Program
The EPA Chesapeake Bay Program has been using the
HSPF model as the framework for modeling total watershed
contributions of flow, sediment, and nutrients (and
associated constituents such as water temperature,' DO,
BOD, etc.) to the tidal region of the Chesapeake Bay
(Donigian et al., 1986; 1990). The watershed modeling
represents pollutant contributions from an area of more than
68,000 sq. mi., and provides input to drive a fully dynamic
three-dimensional, hydrodynamic/water quality model of the
Bay. The watershed drainage area is divided into land seg-
ments and stream channel segments; the land areas modeled
include forest, agricultural cropland (conventional and
conservation tillage systems), pasture, urban (pervious and
impervious areas), and uncontrolled animal waste contribu-
tions. The stream channel simulation includes flow routing
and oxygen and nutrient biochemical modeling (through
phytoplankton) in order to account for instream processes
affecting nutrient delivery to the Bay.
Currently, buildup/washoff type algorithms are being used
for urban impervious areas, potency factors for all pervious
areas, and constant (or seasonally variable) concentrations
for all subsurface contributions and animal waste compo-
nents. Enhancements are underway to utilize the detailed
process (i.e. Agri-chemical modules) simulation for crop-
land areas to better represent the impacts of agricultural
BMPs. The watershed modeling is being used to evaluate
nutrient management alternatives for attaining a 40 % reduc-
tion in nutrient loads delivered to the Bay, as defined in a
joint agreement among the governors of the member states.
6.2.2 Alachlor Special Review
TheEPA Office of Pesticide Programs performed a 'Special
Review' of the herbicide alachlor, which is widely used on
corn and soybeans, to determine estimated concentrations in
.surface waters resulting from agricultural applications.
HSPF was applied to selected watersheds in three separate
agricultural regions—Iowa River Basin, IA; Honey Creek,
OH; and Little River, GA—under different usage assump-
tions to evaluate. a likely range of both mean annual and
maximum daily alachlor concentrations (Mulkey and Doni-
gian, 1984). The modeling results provided input to the
human health risk assessment in which EPA decided to
allow continued use of alachlor in the U.S.
6.2.3 CREAMS Application in Pennsylvania
The CREAMS model was applied by the University of
Maryland Department of Agriculture and Engineering to
selected subbasins of the Susquehanna and Potomac river
basins in Pennsylvania to evaluate the effects of agricultural
BMPs on nutrient loadings to surface water and to ground-
water (Shirmohammadi and Shoemaker, 1988). The study
was sponsored by the Interstate Commission on the Potomac
River to evaluate the relative nutrient loading impacts of a
wide range of potential BMPs, including no till, contouring,
terracing, strip cropping, diversions, grass waterways, nutri-
ent management, and various joint combination scenarios.
Although the model was applied without calibration or ob-
served, runoff/leaching data for comparison, the results
showed the relative effectiveness of the alternatives
analyzed. The study was performed in support of efforts, to-
evaluate alternative means of achieving the 40 % reduction
goal of the Chesapeake Bay Agreement.
6.2.4 Use of SWRRB in NOAA's National
Coastal Pollutant Discharge Inventory
The National Oceanic and Atmospheric Administration
(NOAA) is using the SWRRB model to evaluate pollutant
loadings to coastal estuaries and embayments as part of its
National Coastal Pollutant Discharge Inventory (NOAA,
1987a; 1987b). SWRRB is being used for loadings from all
non-urban areas, while a separate procedure is proposed for
all urban areas. SWRRB has been run for all major estuaries
on the East Coast, Gulf Coast, and West Coast for a wide
range of pollutants.
6.2.5 Use of AGNPS in Virginia
The Virginia Department of Soil and Water Conservation is
applying the AGNPS model to evaluate sediment erosion
and nutrient loadings from land uses within the Owl Creek
and Nomini Creek watersheds. Both watersheds have been
27
-------
instrumented to 'monitor runoff quality from forest, agri-
cultural cropland, and animal feedlot areas. AGNPS is being
applied to the watersheds to analyze potential reductions
ia nutrient loadings for alternative management scenarios
(M. Flagg, Virginia Dept. of Soil and Water Conservation,
Richmond, VA, Personal Communication, 1990). The re-
sults of these applications, and planned use of the calibrated
HSPF model resulting from the Chesapeake Bay Program
application noted above, will be used to evaluate the means
of achieving Virginia's 40% nutrient reductions required by
the Chesapeake Bay Agreement.
6.2.6 Use of HSPF in Metropolitan Washington
The Metropolitan Washington Council of Governments has
used HSPF in a number of modeling studies to evaluate
water quality impacts of nonpoint sources, potential changes
resulting from proposed urban stormwater management
practices, and water quality changes resulting from alter-
native wastewater treatment levels (Sullivan and Schueler,
1982; Schueler, 1983; Metropolitan Washington Council of
Governments, 1985). Studies on Piscataway and Seneca
Creeks demonstrated reasonable agreement with observed
instreara water quality variables, and then the calibrated
model was used to analyze water quality impacts of alter-
native scenarios including increased street sweeping,
stormwater detention, and stormwater treatment. Since the
watershed areas are primarily urban, the buildup/washoff
algorithms were used to calculate loadings from all land
areas. In a separate study, the HSPF instream module was
used with pre-defined nonpoint and point source loadings to
evaluate the impacts of proposed alternative wastewater
treatment levels on the segment of the Potomac River near
Washington, D.C.
6.2.7 Patuxent River Nonpoint Source
Management Study
The Maryland Department of the Environment is conducting
a study of the Patuxent River to quantify nonpoint source
contributions and evaluate alternative means of improving
downstream water quality in the Patuxent River Estuary
(Summers, 1986). The study includes a 7-year monitoring
program that involves observations of runoff quantity and
quality at both field size (single land use) locations and
instream, multi-land use sites. HSPF is being applied to cal-
culate nonpoint loadings from the forest, agricultural, and
urban land areas of the watershed and the instream water
quality throughout the river system. The Patuxent is a
microcosm of the larger Chesapeake Bay, a complex water-
shed with multiple land uses, point and nonpoint sources,
and reservoirs draining to a tidal estuary. Like the larger
Chesapeake Bay study, both simple and complex nonpoint
runoff algorithms will be used to represent all land uses and
effects of potential management practices.
6.2.8 European Case Studies in Application
of the CREAMS Model
Svetlosanov and Knisel (1982) provide a compendium of
case studies describing applications of CREAMS in Europe.
The work was sponsored by the International Institute of
Applied Systems Analysis in Laxenburg, Austria, with the
dual objectives of demonstrating the use of CREAMS for
quantitative evaluation of the impacts of agricultural
management in different countries, and performing model
testing and validation studies. The report describes
applications in Finland, West Germany, Poland, Sweden,
the United Kingdom, and the Soviet Union; comparisons of
model results with observations were made in a few of the
studies, while in others CREAMS was used without compar-
ison to observed data to investigate alternative practices.
Analysis of the case studies identified some of the potential
benefits from the model applications, and elucidated some
of the generic model weaknesses (e.g., smowmelt) and spe-
cific refinements needed for European conditions.
28
-------
Section 7.0
Summary and Recommendations for Ndnpoint Source Runoff Quality Modeling
Simulation of runoff quality will increase in importance as
regulation and control of nonpoint sources increases in the
next several years. The implementation of Section 405 of
the Clean Water Act is especially important if stormwater
outfalls will be required to have NPDES permits. The EPA
is currently establishing guidelines for data collection,
quality monitoring and forms of analysis such that urban
areas can meet their obligations under these regulations.
Waste load allocations and appropriate control strategies
required under Section 303 (d) and 319 will demand more
detailed analyses of'nonpoint contributions for comprehen-
sive water quality management.
Some form of modeling will almost assuredly become part
of routine analyses performed at some portion of the
thousands upon thousands of CSO and stormwater discharge
locations around the country. Several modeling options
exist, but none of them are truly "deterministic" in the sense
of fully characterizing the physical, chemical and biological
mechanisms that underlie conceptual buildup, erosion, trans-
port and degradation processes that occur in an urban drain-
age system. Even if fully deterministic models were avail-
able, it is doubtful that they could be routinely applied
without calibration data. But this is essentially true of almost
all methods. Because a method is simple, e.g., constant con-
centration, does not make it more correct. Rather, the as-
sumption is made that there will be some error in prediction
regardless, of the method, and there may be no point in com-
piling many hypothetical input parameters for a more com-
plex model lacking a guarantee of a better prediction. For
example, a study in Denver showed that regression equa-
tions could predict about as well as DR3M-QUAL given the
available quality information (Ellis and Lindner-Lunsford,
1986). But physically-based (conceptual) models do have
certain advantages, discussed later.
Physically-based urban models depend upon conceptual
buildup and washoff processes incorporated into the quality
algorithms. Such models have withstood the test of time and
have been applied in major urban runoff quality studies.
However, the relative lack of fundamental data on buildup
and washoff parameters has lead to simpler methods more
often being applied, starting with the assumption of a con-
stant concentration and becoming more complex. For exam-
ple, the derived distribution approach of the EPA Statistical
Method provides very useful screening information with
minimal data—but more than are required by just assuming
a constant concentration. With the mass of NURP and other
data, regression approaches are now more viable but still
subject to the usual restrictions of regression analysis.
Spreadsheets are ubiquitous on microcomputers and serve as
a convenient mechanism to implement several of the simple
approaches, especially those that rely upon sets of
coefficients and EMCs as a function of land use Or other
demographic information. For example, the EPA Screening
Procedures could easily be implemented in a spreadsheet
format, and would be an appropriate tool for nonpoint source
wasteload allocation assessments, at least for screening
purposes.
Minimal data requirements and ease of application are the
principal advantages of simpler simulation methods (constant
concentration, statistical, regression, loading functions). How-
ever, in spite of their more complex data requirements, con-
ceptual models (DR3M-QUAL, HSPF, STORM, SWMM,
CREAMS, SWRRB) have advantages in terms of simulation
of routing effects and control options as well as superior
statistical properties of continuous time series. For example,
the EPA Statistical Method assumes that stream flow is not
correlated with the urban runoff flow. This may or may not
be true in a given situation, but it is not necessary to require
such an assumption when running a model such as HSPF or
SWMM. The urban and non-urban conceptual models dis-
cussed in detail all have a means of simulating storage and
treatment effects, and/or impacts of a significant number of
management options. Other than a constant removal, this is
difficult to do with the simpler methods. The conceptual
models generally have very much superior hydrologic and
hydraulic simulation capabilities (not true for STORM ex-
cept that it can also use real rainfall hyetographs as input).
This alone usually leads to better prediction of loads
(product of flow times concentration). It should also be
borne in mind that even complex models such as SWMM
can be run with minimal quality (and quantity) data require-
ments, such as using only a constant concentration. Finally,
some of the case studies imply that transferability of coeffi-
cients and parameters is easier with buildup and washoff
than with rating curve and constant concentration methods.
If a more complex conceptual model is to be applied, which
one should it be from among the ones described herein?
SWMM is certainly the most widely used and probably the
most versatile for urban areas, but all have their advocates.
HSPF may be more appropriate in large multiple land use
29
-------
watersheds, in areas with more open space where ground-
water contributions increase in importance, where rainfall-
induced erosion occurs, or where quality interactions are
important along the runoff pathway. The simplicity of
STORM remains attractive, and various consultants have
utilized their own version as a planning tool. The USGS
DR3M-QUAL model has been successfully applied in
several USGS studies but has not seen much use outside the
agency. It contains useful techniques for quality calibration.
SWMM and HSPF retain limited support from the EPA
Center for Exposure Assessment Modeling (CEAM) at
Athens, Georgia, and a similar level of support is available
for CREAMS from the USDA Southeast Watershed Re-
search Laboratory in Tifton, Georgia. Unfortunately, this
support is limited mainly to distribution and implementation
on a computer system. STORM and DR3M-QUAL will re-
main useful, but it is unlikely that either of these two
models will enjoy enhancements or support from their spon-
soring agencies in the near future. Extramural support for
all major operational models is highly desirable for main-
tenance and improvements, especially in light of the general
models in nonpoint source studies in the U.S. No model can
exist for long without continuing sustenance in the form
of user support, maintenance, and refinements in re-
sponse to changing technology. All agencies who have
sponsored, or are currently sponsoring, model development
efforts need to recognize the critical importance of these
activities if their efforts are to produce 'operational' models
with associated wide-spread usage.
Agricultural model development will continue largely under
the continuing sponsorship of the U.S.D.A. Currently,
CREAMS is the most used model for strictly agricultural
land,, but a number of model development efforts are on-
going at various agricultural research stations across the
country. The continuing development and testing of the
SWRRB model will likely lead to its increased use in a
number of non-urban studies; its use of a daily time step is
attractive to users because of the less intensive data needs
than for HSPF. However, for large complex watersheds, in-
volving both urban and non-urban areas, HSPF will remain
the model of choice for many users. Ongoing agricultural
research will likely lead to improved understanding, of
processes, with improved algorithms that should be incor-
porated into current models. For example, the U.S.D.A.
effort to develop a more process-oriented replacement for
the USLE (i.e. the WEPP—Watershed Erosion Prediction Proj-
ect) will likely lead to improved soil erosion algorithms that
may be appropriate for incorporation into current models.
What is a reasonable approach to simulation of runoff
quality? The main idea, for both urban and non-urban areas,
is to use the simplest approach that will address the project
objectives at the time. This usually means to start simple
with a screening tool such as constant concentration (usually
implemented in a spreadsheet), regression, statistical, or
loading function approach. If these methods indicate that
more detailed study is necessary' or if they are unable to
address all the aspects of the problem, e.g.; the effective-
ness of control options or management alternatives, then one
of the more complex models must be run. No method cur-
rently available (or likely to be available) can predict
absolute (accurate) values of concentrations and loads with-
out local calibration data, including complex buildup and
washoff models for urban areas, and soil process models for
agricultural croplands. Thus, if a study objective is to
provide input loads to a receiving water quality model, local
site-specific data will probably be required. On the other
hand, several methods and models might be able to compare
the relative contributions from different source areas, or to
determine the relative effectiveness of control and/or
management options (if the controls can be characterized by
simple removal fractions). When used for purposes such as
these, the methods, including buildup and washoff models,
can usually be applied on the basis of NURP data (for urban
models) and/or the best currently available source of quality
data, such as data from agricultural research stations for the
non-urban models.
When properly applied and their assumptions respected,
models can be tremendously useful tools in analysis of urban
and non-urban runoff quality problems. Methods and models
are evolving that utilize the large and currently expanding
data base of quality information. As increasing attention is
paid to runoff problems in the future, the methods and
models can only be expected to improve.
30
-------
Section 8.0
References
Abbott, J. 1977. Guidelines for Calibration and Application
of STORM. Training Document No. 8. Hydrologic Engi-
neering Center, Corps of Engineers, Davis, CA.
Alley, W.M. 1986. Summary of Experience with the Dis-
tributed ^Routing Rainfall-Runoff Model (DR3M). In:
Urban Drainage Modeling, C. Maksimovic and M.
Radojkovic, (eds). Proc. of International Symposium on
Comparison of Urban Drainage Models with Real Catch-
ment Data, Dubrovnik, Yugoslavia. Pergamon Press,
New York, NY. pp 403-415.
Alley, W.M. and P.E. Smith. 1982a. Distributed Routing
Rainfall-Runoff Model—Version II. USGS Open File
Report 82-344, Gulf Coast Hydroscience Center, NSTL
Station, MS.
Alley, W.M. and P.E. Smith. 1982b. Multi-Event Urban
Runoff Quality Model. USGS Open File Report 82-764,
Reston, VA.
Alonzo, C.V. and D.G. DeCoursey. 1985. Small Watershed
Model. In: Proceedings of the Natural Resources Model-
ing Symposium. DeCoursey, D.G. (ed). Pingree Park,
CO October 16-21, 1983. ARS-30. U.S. Department of
Agricultural, Agricultural Research Service, pp. 40-46.
Ambrose, R.B., Jr. and T.O. Barnwell, Jr. 1989. En-
vironmental Software at the U.S. Environmental Pro-
tection Agency's Center for Exposure Assessment Model-
ing. CEAM, EPA, Athens, GA, Environmental Software
4(2):76-93.
Ambrose, R.B., Jr., J.P. Connolly, E. Southerland, T.O.
Barnwell, Jr., and J.L. Schnoor. 1988. Waste Allocation
Simulation Models. J. Water Pollution Control Federa-
tion, 60(9): 1646-1655.
American Public Works Association. 1969. Water Pollution
Aspects of Urban Runoff. Report 11030DNS01/69 (NTIS
PB-215532), Federal Water Pollution Control Adminis-
tration, Washington, DC.
American Society of Agricultural Engineers. 1988. Model-
ing Agricultural, Forest, and Rangeland Hydrology.
Proceedings of the 1988 International Symposium. ASAE
Publication No. 07-88, pp. 510. St Joseph, MI.
Amy, G., R. Pitt, R. Singh, W.L. Bradford and M.B.
LaGraff. 1974. Water Quality Management Planning for
Urban Runoff. EPA 440/9-75-004 (NTIS PB 241 689/
AS), U.S. Environmental Protection Agency, Washing-
ton, DC.
Barnwell, T.O., Jr. 1984. EPA's Center for Water Quality
Modeling. Proc. of the Third International Conference on
Urban Storm Drainage. Chalmers University, Goteborg,
Sweden, Vol. 2, pp. 463-466.
Bamwell, T.O., Jr. 1987. EPA Computer Models are
Available to All. Water Quality International, IAWPRC,
No. 2, pp. 19-21.
Beasley, D.B. and D.L. Thomas. 1989, Application of
Water Quality Models for Agricultural and Forested
Watersheds. Southern Cooperative Series Bulletin No.
338. University of Georgia—Coastal Plain Experiment
Station, Tifton, GA pp. 116. .
Bedient, P.B. and W.C. Huber. 1989. Hydrology and
Floodplain Analysis. Addison-Wesley Publishers, Read-
ing, MA.
Bensen, M.A. 1962. Spurious Correlation in Hydraulics and
Hydrology. J. Hydraulics Div., Proc. ASGE, 91(HY4):
57-71.
Brown, R.G. 1984. Relationship between Quantity and
Quality of Storm Runoff and Various Watershed Charac-
teristics in Minnesota, USA. Proc.. Third International
Conference on Urban Storm Drainage, Chalmers Univer-
sity, Goteborg, Sweden, Vol.3, pp. 791-800.
Camp, Dresser and McKee. 1987. Boggy Creek Basin Non-
point Source Water Quality Study. Report to the South
Florida Water Management District from Camp, Dresser
and McKee, Maitland, FL.
Camp, Dresser and McKee. 1989. Phase I Final Report,
Master Stormwater Management Plan, City of Jackson-
ville, FL. COM, Jacksonville, FL.
Davis, M.J., M.K. Snyder and J.W. Nebgen. 1981. River
Basin Validation of the Water Quality Assessment Methr
odology for Screening Nondesignated 208 Areas—Volume
31
-------
I: Nonpoint Source Load Estimation. U.S. Environmental
Protection Agency, Athens, GA.
Dean, J.D., W.B. Mills and D.B. Porcella. 1981a. A Screen-
ing Methodology for Basin Wide Water Quality Manage-
ment. Symposium on Unified River Basin Management.
R.M. North, L.B. Dworsky and D.J. Allee (eds.) May 4-7,
1980, Gatlinburg, TN.
Dean, J.D., B. Hudson and W.B. Mills. 1981b. River Basin
Validation of the MRI Nonpoint Calculator and Tetra Tech's
Nondesignated 208 Screening Methodologies, Volume n.
Chesapeake-Sandusky Nondesignated 208 Screening Meth-
odology Demonstration. U.S. Environmental Protection
Agency, Athens, GA.
DeCoursey, D.G. 1982. ARS Small Watershed Model. Paper
No. 82-2094. ASAE Summer Meeting, University of Wis-
consin, Madison, WI. June 27-30. pp. 33.
DeCoursey, D.G. 1985. Proceedings of the Natural Resources
Modeling Symposium. Pingree Park, CO October 16-21,
1983. ARS-30. U.S. Department of Agricultural, Agri-
cultural Research Service, pp. 532.
Devaurs, M.A., E.P. Springer, L.J. Lane and G.J. Langhorst.
1988. Prediction Methodology for Contaminant Transport
from Rangeland Watersheds. In: Modeling Agricultural,
Forest, and Rangeland Hydrology. Proceedings of the 1988
International Symposium. American Society of Agricultural
Engineers. ASAE Publication No. 07-88, pp. 79-94. St.
Joseph, MI.
Di Toro, D.M. 1984. Probability Model of Stream Quality
Due to Runoff. J. Environmental Engineering Div., ASCE,
110(3):6X)7-628.
DiToro, D.M. andMJ. Small. 1979. Stormwater Interception
and Storage. J. Environmental Engineering Div., ASCE,
105(EEl):43-54.
Donigian, A.S., Jr., J.C. Lnhoff, B.R. Bicknell and J.L.
Kittle, Jr. 1984. Application Guide for Hydrological
Simulation Program Fortran (HSPF). EPA-600/3-84-065,
U.S. Environmental Protection Agency, Athens, GA.
Donigian, A.S. Jr. and D.C. Beyerlein. 1985. Review and
Analysis of Available NFS and Integrated Watershed
Models. Prepared by Anderson-Nichols & Co., Palo Alto,
CA, for Woodward-Clyde Consultants, Walnut Creek, CA.
Donigian, A.S. Jr., B.R. Bicknell and J.L. Kittle, Jr. 1986.
Conversion of the Chesapeake Bay Basin Model to HSPF
Operation. Prepared by AQUA TERRA Consultants for
Computer Sciences Corporation, Annapolis, MD and U.S.
EPA Chesapeake Bay Program, Annapolis, MD.
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 Recommenda-
tions. Prepared by AQUA TERRA Consultants for U.S.
EPA Chesapeake Bay Program, Annapolis, MD.
Driscoll, E.D. 1979. In: Benefit Analysis for Combined Sewer
Overflow Control. Seminar Publication, EPA-4525/4-79-013,
U.S. Environmental Protection Agency, Cincinnati, OH.
Driscoll, E.D. 1981. Combined Sewer Overflow Analysis
Handbook for Use in 201 Facility Planning. Two Volumes,
Final Report for Contract 68-01-6148, U.S. Environmental
Protection Agency, Facility Requirements Division, Policy
and Guidance Branch, Washington, DC.
Driscoll, E.D.. 1986. Lognormality of Point and Nonpoint
Source Pollutant Concentrations. Proc. Stormwater and
. Water Quality Model Users Group Meeting, Orlando, FL,
EPA/600/9-86/023, pp. 157-176, (NITS PB87-117438/AS),
U.S. Environmental Protection Agency, Athens, GA.
Driscoll, E.D., P.E. Shelley, and E.W. Strecker. 1989.
Pollutant Loadings and Impacts from Highway Stormwater
Runoff. Vol. I Design Procedure (FHWA-RD-88-006) and
Vol. ffl Analytical Investigation and Research Report
(FHWA-RD-88-008). Office of Engineering and Highway
Operations R & D, Federal Highway Administration,
McLean, VA.
Driver, N.E., M.H. Mustard, R.B. Rhinesmith and R.F.
Middelburg. 1985. US. Geological Survey Urban Storm-
water Data Base for 22 Metropolitan Areas Throughout the
United States. U.S.G.S. Open File Report 85-337, Lake-
wood, CO.
Driver, N.E. and G.D. Tasker. 1988. Techniques for Esti-
mation of Storm-Runoff Loads, Volumes, and Selected
Constituent Concentrations in Urban Watersheds in the
United States. U.S.G.S. Open-File Report 88-191, Denver,
CO.
Ellis, S.R. and J.B. Lindner-Lunsford. 1986. Comparison of
Conceptually Based and Regression Rainfall-Runoff Models
in the Denver Metropolitan Area, Colorado, USA. In:
Urban Drainage Modelling, Proceedings of International
Symposium on Comparison of Urban Drainage Models with
Real Catchment Data, C. Maksimovic and M. Radojkovic,
eds., Pergamon Press, New York, NY. pp. 263-273.
Environmental Protection Agency. 1973. Methods for Identi-
fying and Evaluating the Nature and Extent of Non-point
Sources of Pollutants. EPA-430/9-73-014, U.S. Environ-
mental Protection Agency, Washington, DC.
Environmental Protection Agency. 1976a. Land Use—Water
Quality Relationship. WPD-3-76-02, Water Planning
Division, U.S. Environmental Protection Agency, Washing-
ton, DC.
Environmental Protection Agency. 1976b et seq. Areawide
Assessment Procedures Manual. Three Volumes, EPA-600/
32
-------
9-76-014 (NTIS PB-271863/SET), U.S. Environmental Pro-
tection Agency, Cincinnati, OH.
Environmental Protection Agency. 1983b. Results of the
Nationwide Urban Runoff Program, Volume I, Final
Report. NTIS PB84-185552, U.S. Environmental Protection
Agency, Washington, DC.
Feldman, A.D. 1981. HEC Models for Water Resources Sys-
tem Simulation: Theory and Experience. Advances in Hy-
droscience, Vol. 12, pp. 297-423, Academic Press, New
York, NY. '" :. , .
s'' _ _ -
Giorgini, A. and F. Zingales. 1986. Agricultural Nonpoint
Source Pollution: Model Selection and Application. Develop-
ments in Environmental Modeling Series (No. 10) pp. 409.
Elsevier Science Publishers, Amsterdam, The Netherlands.
Haith, D. A. 1980. A Mathematical Model for Estimating Pesti-
cide Losses in Runoff. J. of Env. Qua! 9(3):428-433.
Haith, D. A. and L.J. Tubbs. 1981. Watershed Loading
Functions for Nonpoint Sources. ASCE. J. Env. Engr.
107(EE1): 121-137.
Haith, D. A. andL.L. Shoemaker. 1987. Generalized Loading
Functions for Stream Flow Nutrients. AWRA. Wat Res.
Bull. 23(3):471-478.
Heaney, J.P., W.C. Huber, M.A. Medina, Jr., M.P. Murphy,
S.J. Nix, and S.M. Hasan. 1977. Nationwide Evaluation of
Combined Sewer Overflows and Urban Stormwater Dis-
charges, Vol. II: Cost Assessment and Impacts. EPA-6007
2-064b (NTIS PB-266005), U.S. Environmental Protection
Agency, Cincinnati, OH.
Heaney, J.P., W.C. Huber and S.J. Nix. 1976, Storm Water
. Management Model, Level I, Preliminary Screening Pro-
cedures. EPA-€00/2-76-275 (NTIS PB-259916). U.S. En-
vironmental Protection Agency, Cincinnati, OH.
Heaney, J.P., W.C. Huber, H. Sheikh, M.A. Medina, J.R.
Doyle, W.A. Peltz and I.E. Darling. 1975. Urban Storm-
water Management Modeling and Decision-Making. EPA-
670/2-75-022 (NTIS PB-242290), U.S. Environmental
Protection Agency, Cincinnati, OH.
Henderson, R.J. and G.D. Moys. 1987. "Development of a
Sewer Flow Quality Model for the United Kingdom,"
Topics in Urban Storm Water Quality, Planning and
Management, Proc. Fourth Int. Conference on Urban Storm
Drainage, W. Gujer and V. Krejci, eds., Ecole Polytech-
nique Federate, Lausanne, Switzerland, pp. 201-207.
Howard, C.D.D. 1976. Theory of Storage and Treatment Plant
Overflows. J. Environmental Engineering Division, Proc.
ASCE, 102(EE4):709-722.
Huang, P. and J.L. DiLorenzo. 1990. "Lower Hackensack
River Watershed Planning Using SWMM-4," Proc. Storm-
water and Water Quality Modeler Users Group Meeting,
Eatontown, NJ, EPA Report in press, Environmental
Protection Agency, Athens, GA.
Huber, W.C. 1980. Urban Wasteload Generation by Multiple
Regression Analysis of Nationwide Urban Runoff Data.
Proc. Workshop on Verification of Water Quality Models,
R.V. Thomann and T.O. Barnwell, eds., EPA-600/9-80-0-
16, pp. 167-175 (NTIS PB80-186539), U.S. Environmental
Protection Agency, Athens, GA. ,
Huber, W.C. and J.R Heaney. 1982. Analyzing Residuals
Generation and Discharge from Urban and Non-urban Land
Surfaces. Li: Analyzing Natural Systems, Analysis for Re-
gional Residuals—Environmental Quality Management, D. J.
Basta and B.T. Bower, eds., Resources for the Future,
Johns Hopkins University Press, Baltimore, MD (also avail-
able from NTIS as PB83-223321), Chapter 3, pp. 121-243.
Huber, W.C., J.P. Heaney, D.A. Aggidis, R.E. Dickinson,
K.J. SmolenyakandR.W. Wallace. 1982. Urban Rainfall-
Runoff-Quality Data Base, EPA-600/2-81-238 (NTIS PB82-
221094), U.S. Environmental Protection Agency, Cincin-
nati, OH.
Huber, W.C. 1985. Deterministic Modeling of Urban Runoff
Quality. In: Urban Runoff Pollution, H.C. Torno, J. Mar-
salek and M. Desbordes, eds., NATO ASI Series, Series G:
Ecological Sciences, 10:167-242, Springer-Verlag, New
York, NY. ...
Huber, W.C. 1986. Modeling Urban Runoff Quality: State of
the Art. Proceedings of Conference on Urban Runoff Qual-
ity, Impact and Quality Enhancement Technology, B. Ur-
bonas and L.A. Roesner, eds., Engineering Foundation,
ASCE, pp. 3448, New York, NY.
Huber, W.C., J.P. Heaney and B.A. Cunningham. 1986.
Storm Water Management Model (SWMM) Bibliography.
EPA/600/3-85/077 (NTIS PB86-136041/AS), U.S. Environ-
mental Protection Agency, Athens, GA.
Huber, W.C. and R.E. Dickinson. 1988. Storm Water Man-
agement Model User's Manual, Version 4. EPA/600/3-88/
OOla (NTIS PB88-236641/AS), U.S. Environmental Pro-
tection Agency, Athens, GA.
Hydrolpgic Engineering Center. 1977. Storage, Treatment,
Overflow, Runoff Model, STORM, User's Manual. Gener-
alized Computer Program 723-S8-L7520, Corps of Engi-
neers, Davis, CA.
Hydroscience, Inc. 1979. A Statistical Method for Assessment
'of Urban Storm Water Loads^-Impacts— Controls. EPA-
440/3-79-023 (NTIS PB-299185/9), U.S. Environmental
Protection Agency, Washington, DC.
33
-------
Jacobsen, P., P. Harremoes and C Jakobsen. 1984. "The
Danish Stormwater Modelling Package: The SVK-System,"
Proc. Third Int. Conference on Urban Storm Drainage, P.
Balmer, P-A. Malmqyist and A. Sjoberg, eds., Chalmers
University, Goteborg, Sweden, 2:453-562.
James, L.D. and S. J. Burges. 1982. Selection, Calibration, and
Testing of Hydrologic Models. In: Hydrologic Modeling of
Small Watersheds, C.T. Haan, H.P. Johnson and D.L Brak-
ensiek, eds., Monograph No. 5, Chapter 11, pp. 435-472,
American Society of Agricultural Engineers, St. Joseph, ML
Johansea, N.B., J.J. linde-Jensen and P. Harremoes. 1984.
"Computing Combined System Overflow Based on Histori-
cal Rain Series," Proc. Third Int. Conference on Urban
Storm Drainage, P. Balmer,, P-A. Malmqvist and A.
Sjoberg, eds., Chalmers University, Goteborg, Sweden,
3:909-918.
Johanson, R.C., J.C. Lnhoff, J.L. Kittle and A.S. Donigian.
1984. Hydrological Simulation Program—Fortran (HSPF):
User's Manual for Release 8. EPA-600/3-84-066. U.S.
Environmental Protection Agency, Athens, GA.
Kibler, D.F., ed. 1982. Urban Stormwater Hydrology. Ameri-
can Geophysical Union, Water Resources Monograph 7,
Washington, DC.
Knisel, W. (ed). 1980. CREAMS: A Field-Scale Model for
Chemicals, Runoff, and Erosign from Agricultural Manage-
ment Systems. U.S. Department of Agriculture. Conserva-
tion Research Report No. 26, 640 pp.
Lager, J.A., W.G. Smith, W.G. Lynard, R.M. Finn and E.J.
Finnernore. 1977. Urban Stormwater Management and
Technology: Update and Users' Guide. EPA-600/8-77-014
(NTIS PB-275654), U.S. Environmental Protection Agency,
Cincinnati, OH.
Li, W., D.E. Merrill and D.A. Haith. 1989. Loading Func-
tions for Pesticide Runoff. J. WPCF 62(1): 16-26.
Loganathan, V.G. and J.W. Delleur. 1984. Effects of Urban-
ization on Frequencies of Overflows and Pollutant Loadings
from Storm Sewer Overflows: A Derived Distribution
Approach. Water Resources Research, 20(7):857-865.
Lorber, M.N. and L.A. Mulkey. 1982. An Evaluation of
Three Pesticide Runoff Loading^Models. J. Environ. Qua!
ll(3):519-529.
Manning, M.J., R.H. Sullivan and T.M. Kipp. 1977. Nation-
wide Evaluation of Combined Sewer Overflows and Urban
Stormwater Discharges—Vol. HI: Characteristics of Dis-
charges. EPA-600/2-77-064c (NTIS PB-272107), U.S. En-
vironmental Protection Agency, Cincinnati, OH.
Metropolitan Washington Council of Governments. 1985.
Documentation of the Calibration and Verification of the
Upper Potomac River Model. Prepared for Maryland Office
of Environmental Programs, Baltimore, MD.
McElroy, A.D., S.Y. Chiu, J.W. Nebgen, A. Aleti and R.W.
Bennett. 1976. Loading Functions for Assessment of Water
Pollution from Non-Point Sources. EPA-600/2-76-151
(NTIS PB-253325), U.S. Environmental Protection Agency,
Washington, DC.
Metcalf and Eddy, Inc. 1971. University of Florida, and Water
Resources Engineers, Inc., Storm Water Management
Model, Volume I—Final Report. EPA Report 11024DOCO-
7/71 (NTIS PB-203289), U.S. Environmental Protection
Agency, Washington, DC.
Miller, R.A., H.C. Mattraw, Jr. and M.E. Jennings. 1978.
Statistical Modeling of Urban Storm Water Processes,
Broward County, Florida. Proc. International Symposium
on Urban Storm Water Management, University of Ken-
tucky, Lexington, KY. pp. 269-273.
Mills, W.B., V.H. Colber and J.D. Dean. 1979. Hand-Held
Calculator Programs for Analysis of River Water Quality
Interactions. Tetra Tech, Inc., Lafayette, CA.
Mills, W.B., J.D. Dean, D.B. Porcella, S.A. Gherini, R.J.M.
Hudson, W.E. Frick, G.L. Rupp and G.L. Bowie. 1982.
Water Quality Assessment: A Screening Procedure for Tox-
ic and Conventional Pollutants. EPA-600/6-82-O04a and b.
Volumes I and n. U.S. Environmental Protection Agency.
Mockus, J. 1972. Estimation of Direct Runoff from Storm
Rainfall. In: National Engineering Handbook, Sec. 4, Hy-
drology. U.S. Soil Conservation Service, Washington, DC.
Morrissey, S.P. and D.R.F. Harleman. 1989. "Technology
and Policy Issues Involved in the Boston Harbor Cleanup,"
Parsons Laboratory Report No. 327, M.I.T., Cambridge,
MA.
Mulkey, L.A. and A.S. Donigian, Jr. 1984. Modeling Alach-
lor Behavior in Three Agricultural River Basins. U.S.
Environmental Protection Agency, Environmental Research
Laboratory, Athens, GA.
Mulkey, L.A., R.F. Carsel and C.N. Smith. 1986. Devel-
opment, Testing, and Applications of Nonpoint Source
Models for Evaluation of Pesticides Risk to the Environ-
• meht. In: Agricultural Nonpoint Source Pollution: Model
Selection and Application. Giorgini, A. and F. Zingales
(eds). Developments in Environmental Modeling Series
(No. 10). Elsevier Science Publishers, Amsterdam, The
Netherlands.
Najarian, T.O., T.T. Griffin and V.K. Gunawardana. 1986.
Development Impacts on Water Quality: A Case Study. J.
Water Resources Planning and Management, ASCE, 112(1)-
20-35.
34
-------
Najarian, T.O., V.K. Gunawardana, R.V. Ram and D. Lai.
1990. "Modeling the Impact of Point/Nonpoint Discharges
on Lower Hackensack River Water Quality," Proc. Storm-
water and Water Quality Modeler Users Group Meeting,
Eatontown, NJ, EPA Report in press, Environmental
Protection Agency, Athens, GA.
National Oceanic and Atmospheric Administration. 1987a. The
National Coastal Pollutant Discharge Inventory: Urban
Runoff Methods Document; Office of Oceanography and
Marine Assessment, NOAA, Rockville, MD.
National Oceanic and Atmospheric Administration. 1987b. The
National Coastal Pollutant Discharge Inventory: Nonurban
Runoff Methods Document. Office of Oceanography and
Marine Assessment, NOAA, Rockville, MD.
Noel, D.D. and M.L. Terstriep. 1982. Q-ILLUDAS—A Con-
tinuous Urban RunoffAVashoff Model. Proc. Int. Symp. on
Urban Hydrology, Hydraulics, and Sediment Control, UKY
BU128, University of Kentucky, Lexington, KY pp. 1-10.
Noel, D.D., M.L. Terstriep and C.A. Chenoweth. 1987.
Nationwide Urban Runoff Program Data Reports. Illinois
State Water Survey, Dept. of Energy and Natural Re-
sources, Champaign, IL.
Novotnyj V. and G. Chesters. 1981. Handbook of Nonpoint
Pollution: Sources and Management, Van Nostrand Rein-
hold Company, New York, NY. 555 pp.
Novotoy, V. 1985. Discussion of "Probability Model of
Stream Quality Due to Runoff," by D.M. Di Toro. Journal
of Environmental Engineering, ASCE, lll(5):736-737.
Pisano, W.C. and C.S. Queiroz. 1977. Procedures for Estimat-
ing Dry Weather Pollutant Deposition in Sewerage Systems.
EPA-600/2-77-120 (NTIS PB-270695), U.S. Environmental
Protection Agency, Cincinnati, OH.
Pitt, R. 1979. Demonstration of Non-Point Pollution Abatement
Through Improved Street Cleaning Practices. EPA-600/2-
79-161 (NTIS PB80-108988), U.S. Environmental Protec-
tion Agency, Cincinnati, OH. .
/
Roesner, L.A., J.A. Aldrich and R.E. Dickinson. 1988. Storm
Water Management Model User's Manual Version 4: Ad-
dendum I, EXTRAN. EPA/600/3-88/001b (NTIS PB88-
236658/AS), U.S. Environmental Protection Agency,
Athens, GA.
Roesner, L.A. and S.A. Dendrou. 1985. Discussion of
"Probability Model of Stream Quality Due to Runoff," by
D.M. Di Toro. Journal of Environmental Engineering,
ASCE, 111(5):738-740.
Roesner, L.A., H.M. Nichandros, R.P. Shubinski, A.D.
Feldman, J.W. Abbott and A.O Friedland. 1974. A Model
for Evaluating Runoff-Quality in Metropolitan Master
Planning. ASCE Urban Water Resources Research Pror
gram, Technical Memo. No. 23 (NTIS PB-234312), ASCE,
New York, NY.
Sartor, J.D. and G.B. Boyd. 1982. Water Pollution Aspects of
Street Surface Contaminants. EPA-R2-72-081 (NTIS PB-
214408), U.S. Environmental Protection Agency, Washing-
s ton, DC.
Schueler, T.R. 1983. Seneca Creek Watershed Management
Study, Final Report, Volumes I and n. Metropolitan
Washington Council of Governments, Washington, DC.
Schueler, T.R. 1987. Controlling Urban Runoff: A Practical
Manual for Planning and Designing Urban BMPs. Metro-
politan Information Center, Metropolitan Washington,
Council of Governments, Washington, DC.
Shirmohammadi, A. and L.L. Shoemaker. 1988. Impact of
Best Management Practices on Water Quality in Pennsyl-
vania. ICPRB Report No. 88-7. Prepared for Interstate
Commission on the Potomac River Basin, Rockville, MD.
Small, M.J. and D.M. Di Toro. 1979. Stormwater Treatment
Systems. Journal of the Environmental Engineering Divi-
sion. Proc. ASCE, 105(EE3):557-569.
Sullivan, M.P. and T.R. Schueler. 1982. "The Piscataway
Creek Watershed Model: A Stormwater and Nonpoint
Source Management Tool." In: Proceedings Stormwater and
Water Quality Management Modeling and SWMM Users
Group Meeting. October 18-19, 1982. Paul E. Wisner ed.,
Univ. of Ottawa, Dept. Civil Engr., Ottawa, Ont, Canada.
Sveflosanov, W. and W.G. Knisel (eds.). 1982. European and
United States Case Studies in Application of the CREAMS
Model. CP-82-S11. International Institute of Applied
Systems Analysis, Laxenburg, Austria, pp. 147.
Summers, R.M. 1986. HSPF Modeling of the Patuxent River
Basin. Maryland Office of Environmental Programs,
Baltimore, MD. Presented at Chesapeake Bay Research
Conference, March 1986. ^
Tasker, G.D. and N.E. Driver. 1988. Nationwide Regression
Models for Predicting Urban Runoff Water Quality at
Unmonitored Sites. Water Resources Bulletin, 24(5):
1091-1101.
Terstriep, M.L. and J.B. Stall. 1974. The Illinois Urban
Drainage Area Simulator, ILLUDAS. Bulletin 58, Illinois
State Water Survey, Urbana, IL.
U.S. Forest Service. 1980. An Approach to Water Resources
Evaluation of Non-Point Silvicultural Sources (A Procedural
Handbook). EPA-600/8-80-012. U.S. Environmental Protec-
tion Agency, Athens, GA. 861 pp.
Viessman, W., Jr., G.L. Lewis and J.W. Knapp. 1989. Intro-
duction to Hydrology: Third Edition, Harper and Row, New
York, NY.
35
-------
Walker, J.F., S.A. Pickard and W.C. Sonzogni. 1989. Spread-
^wet Watershed Modeling for Nonpoint-Source Pollution
Management in a Wisconsin Area. Water Resources Bulle-
tin, 25(1): 139-147.
Ward, A.D., C.A. Alexander, N.R. Fausey and J.D. Dorsey.
1988. In: Modeling Agricultural, Forest, and Rangeland
Hfydrology. Proceedings of the 1988 International Sympo-
sium. American Society of Agricultural Engineers. ASAE
Publication No. 07-88, pp. 129-141, St. Joseph, MI.
Water Pollution Control Federation. 1989. Manual of Practice
on CSO Pollution Abatement, No. FD-17, WPCF, Alexan-
dria, VA.
Whipple, DJ., N.S. Grigg, T. Grizzard, C.W. Randall, R.P.
Shubinski and L.S. Tucker. 1983. StormwaterManagement
in Urbanizing Areas. Prentice-Hall, Englewood Cliffs, NJ.
Williams, J.R. 1975. Sediment-Yield Prediction with Universal
Soil Loss Equation Using Runoff Energy Factor. In: Present
and Prospective Technology for Predicting Sediment Yields
and Sources. U.S. Dept. of Agricultures. ARS-S-40.
Williams, J.R., C.A. Jones andP.T. Dyke. 1984. A Modeling
Approach to Determining the Relationship Between Erosion
and Soil Productivity. Trans. ASAE 27(l):129-144l
Williams, J.R., A.D. Nicks and J.G. Arnold. 1985. Simulator
for Water Resources in Rural Basins. ASCE J. Hydraulic
Engineering. lll(6):970-986.
Wischmeier, W.H. and D.D. Smith. 1978. Predicting Rainfall-
Erosion Losses: A Guide to Conservation Planning. Agricul-
tural Handbook No. 537. U.S. Dept. of Agriculture, Agri-
cultural Research Service.
Woodward-Clyde Consultants. 1989. Synoptic Analysis of Se-
lected Rainfall Gages Throughout the United States. Report
to EPA, Woodward-Clyde Consultants, Oakland, CA.
Zison, S.W. 1980. Sediment-Pollutant Relationships in Runoff
from Selected Agricultural, Suburban and Urban Water-
sheds. EPA-600/3-80-022, U.S. Environmental Protection
Agency, Athens, GA.
Zison, S.W., K. Haven and W.B. Mills. 1977. Water Quality
Assessment: A Screening Methodology for Nondesignated
208 Areas. EPA-€00/6-77-023. U.S. Environmental Protec-
tion Agency, Athens, GA.
Zukovs, G., J. Kollar and M. Shanahan. 1986. Development
of the HAZPRED Model. Proc. Stormwater and Water
Quality Model Users Group Meeting, Orlando, FL. EPA/
600/9-86V023, pp. 128-146 (NTTS PB87-117438/AS), U.S.
Environmental Protection Agency, Athens, GA.
36
-------
Section 9.0
Appendix
Detailed Model Descriptions
Urban
SWMM
STORM
DR3M-QUAL
Non-Urban
EPA Screening Procedures
AGNPS
ANSWERS
CREAMS/GLEAMS
HSPF
PRZM
SWRRB
UTM-TOX
37
-------
Nonpoint Source Model Review
1. Name of Method
Storm Water Management Model (SWMM)
2. Type of Method
. Surface Water Model:
xx Surface Water Model:
Simple Approach
Refined Approach
XX Soil (Groundwater) Model: Simple Approach
Soil (Groundwater) Model: Refined Approach
3. Purpose/Scope
Purpose: Predict rainfall/runoff/quality processes in urban and other areas. Predict hydrographs and polluto-
graphs (concentration vs. time) in
xx runoff waters
XX surface waters
xx ground waters
Source/Release Types:
xx n Continuous
xx. Single
Level of Application:
XX, Screening
Types of Chemicals:
xx Conventional
Unique Features:
xx Intermittent
xx Multiple
xx Intermediate
xx Organic
xx Diffuse
xx Detailed
xx Metals
XX Addresses degradation products
xx Integral database/database manager
TXX Integral uncertainty analysis capabilities
Interactive input/execution manager
4. Level of Effort
System setup:
Assessments:
xx mandays
mandays
xx manweeks
xx manweeks
manmonths
xx manmonths
(Estimates reflect order-of-magnitude values and depend heavily on the experience and ability of the assessor.)
38
-------
5. Description of the Method/Techniques
The Storm Water Management Model (SWMM) consists of
several modules or "blocks" designed to simulate most
quantity and quality processes in the urban hydrologic cycle.
Storm sewers, combined sewers, and natural drainage sys-
tems can be simulated. For generation of hydrographs, the
Runoff Block simulates the^rainfall-runoff process using a
nonlinear reservoir approach, with an option for snowmelt
simulation. Groundwater and unsaturated zone flow and out-
flow are simulated using a simple lumped storage scheme.
A water balance is maintained between storms, and the
entire model may be used for both continuous and single
event simulation. Flow routing is accomplished in order of
increasing complexity in the Runoff Block (nonlinear reser-
voir), Transport Block (kinematic wave) and/or Extran
Block (complete dynamic equations). The Extran Block is
the most comprehensive simulation program available in the
public domain for a drainage system domain and is capable
of simulating backwater, surcharging, looped sewer connec-
tions and a variety of hydraulic structures and appurte-
nances. The Storage/Treatment Block may be used for
storage-indication flow routing. Output from all blocks
includes both flow and stage hydrographs.
Quality processes in the Runoff Block include generation of
surface runoff constituent loads through a variety of options:
1) build-up of constituents during dry weather and wash-off
during wet weather, 2) "rating curve" approach in which
load is proportional to flow rate to a power, 3) constant
concentration (including precipitation loads), and/or 4)
Universal Soil Loss Equation. Removal of "built-up" loads
can occur by street cleaning during dry weather, and special
options are available for snow. One constituent can be taken
as a fraction ("potency factor") of another to simulate ad-
sorption onto solids, for example. The concentration of
groundwater outflow may only be treated as a constant.
Routing and first-order decay may be simulated in the
Runoff and Transport Blocks. (Extran includes no quality
simulation.) The Transport Block may also be used to
simulate scour and deposition within the sewer system based
Shield's criterion for initiation of motion, and generation of
dry-weather flow and quality. The Storage/Treatment Block
simulates removal in storage/treatment devices by 1) first-
order decay coupled with complete mixing or plug flow,
2) removal functions (e.g., solids deposition as a function
of detention time), or 3) sedimentation dynamics. Residuals
(e.g., sludge) are accounted for. Any constituent may be
simulated, but sediment processes, e.g., scour and deposi-
tion, are generally weak with the exception of the Storage/
Treatment Block.
The blocks can be run independently or in any sequence.
Additional blocks are available for statistical analysis of the
output time series (Statistics Block), input and manipulation
of precipitation, evaporation, and temperature time series
(Rain and Temp Blocks), line printer graphics (Graph
Block), and output time series manipulation (Combine
Block).
6. Data Needs/Availability
SWMM can be run in a very simple configuration, e.g., a
single subcatchment and no drainage network, or in a very
detailed configuration, e.g., many subcatchments and chan-
nels and pipes. The volume of data varies accordingly, but
at a minimum will require information on area, impervious-
ness, slope, roughness, depression storage and infiltration
characteristics at the desired level of detail. Channel/pipe
data include shapes, dimensions, slopes or invert elevations,
roughnesses, etc. Quantity data are usually available from
the urban municipality in the form of contour maps and
drainage plans.
If quality is generated in the Runoff Block using a build-up/
wash-off formulation, then build-up coefficients are needed
for alternative build-up formulations (i.e., linear, power,
exponential or Michaelis-Menton) as well as for wash-off
equations. Due to lack of data availability for these con-
ceptual mechanisms, users often resort to rating curves that
may be more readily derived from measurements, or to con-
stant concentrations.
Precipitation input can be in the form of hyetographs for
individual storm events, or long-term hourly or 15-minute
precipitation (or arbitrary time interval) records from the
National Climatic Data Center. Measured hydrographs and
pollutographs (or storm event loads or event mean concen-
trations) are required for calibration and verification.
Quantity results are often reasonably good with little or no
calibration, whereas quality results need local, site-specific
measurements to ensure accurate predictions. Since quality
measurements are expensive and sparse, reliance is often
placed on generalized data from other sites, e.g., EPA
Nationwide Urban Runoff Program (NURP) data, from
which reasonable comparative assessments may usually be
made. - However, if SWMM (or any urban ndnpoinfsource
model) is used to drive a receiving water quality model,
local, site-specific quality measurements are normally
required.
7. Output of the Assessment
SWMM produces a time history of flow, stage and constit-
uent concentration at any point hi the watershed. Seasonal
and annual summaries are also produced, along with conti-
nuity checks and other summary output. The Statistics Block
may be used for storm event separation and frequency anal-
ysis of any of the time series. Somewhat crude (by today's
standards) line printer graphics are available for hydrograph
and pollutograph plots, including comparison with measured
data.
8. Limitations
The simulation methods used in various blocks are exten-
sively described in the SWMM documentation and need to
conceptually represent the prototype watershed for a satis-
factory simulation. Experience has shown SWMM quantity
simulations to compare favorably with measurements and
with other conventional hydrologic methods, e.g., unit
39
-------
hydrographs. Qiiality simulation is especially weak in repre-
sentation of the true physical, chemical and biological
processes that occur in nature and is often a calibration or
curve-fitting exercise. This accounts for the tendency for
users to bypass build-up/wasn-off formulations in favor of
constant concentrations or rating curves that may be more
easily calibrated. Perhaps the weakest component of the
model is simulation of solids transport, although it should be
borne in mind that this is difficult to do in any model. The
microcomputer version of the program is not especially
"user friendly" and lacks good graphics routines. However,
on-screen messages are provided during program execution
to inform the user of the current program status.
9. Hardware/Software Requirements
SWMM is written in ANSI standard Fortran 77. Executable
code prepared using the Ryan-McFarland Fortran compiler
is distributed for IBM XT/AT-compatibles. A hard disk is
required and SL math co-processor is usually required (see
Contact section, below). No special peripherals other man
a. printer are required. Over 200 sample input files are also
distributed, along with the Fortran source code. The pro-
gram is also maintained for DEC/VAX systems; however,
the user must perform his/her own compilation from the
Fortran source code. Input files must be prepared using an
editor capable of producing an ASCII file as output. No
automatic or menu-driven input routine is available. The
program can be obtained in either floppy disk format for
MS-DOS applications or on a 9-track magnetic tape (from
EPA only) with installation instructions for the DEC/VAX
VMS environment. SWMM has been installed on many
computers world-wide with no or minor modifications.
10. Experience
The original version of SWMM was developed in 1969-71
by a consortium of Metcalf and Eddy, Inc., Water Re-
sources Engineers, Inc. (now a part of Camp, Dresser and
McKee, Inc.), and the University of Florida (Metcalf and
Eddy et al., 1971). The program was maintained intermit-
tently for several years at the University of Florida under
the sponsorship first of the EPA Storm and Combined
Sewer Branch (Cincinnati) and later by the EPA Center for
Exposure Assessment Modeling (Athens, GA). Limited (un-
sponsored) maintenance and development work continues at
the University of Florida. Continuing updates to the Extran
Block have been placed in the public domain by Camp,
Dresser and McKee, Inc. and incorporated into the model.
The most current version of the model is Version 4 (Huber
and Dickinson, 1988; Roesner et al., 1988) following earlier
Version 2 (Heaney et al., 1975; Huber et al., 1975) and
Version 3 (Huber et al., 1981; Roesner et al., 1981). A
SWMM bibliography (Huber et al., 1985) contains over 200
SWMM-related references, including many references to
case studies. The model is often referenced in texts (e.g.,
Viessman et al., 1989) and has been used in applications
ranging from routine drainage design to highly complex
hydraulic routing and analysis to large nonpoint and point
source pollution abatement studies, e.g., Boston, Detroit,
Washington DC.
Maintaining SWMM as a non-proprietary model and in the
public domain has produced massive user feedback to
the model developers and led to many improvements
(Huber, 1989). This has been greatly facilitated by
approximately semi-annual meetings of the Storm and Water
Quality Model Users Group and publication of proceedings
by EPA. The model has achieved the status of a "standard
of comparison" for modelers wishing to develop new and
improved techniques.
SWMM has been applied to urban hydrologic quantity and
quality problems in many locations world-wide, including
probably over 100 locations in the U.S. and Canada. Appli-
cations in major U.S. cities include Albuquerque, Atlanta,
Boston, Chicago, Cincinnati, Denver, Detroit, Hartford,
Jacksonville, New Haven, New Orleans, Newark, New
York, Philadelphia, Providence, Rochester, San Jose, San
Francisco, Seattle, Syracuse, Tallahassee, Tampa, Wash-
ington DC.
Because SWMM is in the public domain, portions of the
model have been adapted for related modeling purposes. A
model similar to SWMM in many ways was developed for
the Federal Highway Administration (Dever et al., 1981,
1983). It contains components of the Runoff and Extran
Blocks along with specialized hydraulic simulation capabil-
ities for highway drainage. Because SWMM contains much
more versatile water quality routines, the FHWA model will
not be discussed further here; however, some of its charac-
teristics are compared to those of other models by Huber
(1985).
11. Validation/Review
The program has been calibrated and verified on many inde-
pendent data sets, and the model algorithms have been vali-
dated through extensive outside analysis and review during
the 20 years of model history. Several comparisons of
SWMM and other models have been conducted (e.g., Heeps
and Mein, 1974; Brandstetter, 1977; Huber and Heaney,
1982; Huber, 1985; Water Pollution Control Federation,
1989). In 1982 the U.S. Office of Technology Assessment
said that its "reliability and widespread availability have
made SWMM the most widely used model of its type in the
United States and Canada, and have been important in in-
creasing the use of models by engineers and planners. The
SWMM User's Group [now the Storm and Water Quality
Model Users Group, sponsored by the EPA CEAM] has
been instrumental in achieving the widespread dissemination
and acceptance enjoyed by this important modeling tool."
(Office of Technology Assessment, 1982).
A final caveat about water quality predictions is still in
order: neither SWMM or any other'urban nonpoint model
can predict accurate concentrations and loads in urban
runoff without local, site-specific data. Such data would
likely be required in order to drive a receiving water quality
model, for example. However, relative values and compari-
son of alternatives can usually still be studied from approxi-
mate quality predictions.
40
-------
12. Contact
SWMM is available from the EPA Center for Exposure
Assessment Modeling (CEAM) at no charge. It may be
down-loaded from the CEAM computer bulletin board or
obtained on floppy disks (on two high-density disks in
archived format). The CEAM also maintains the Fortran
source code for a DEC/VAX VMS version. The same pro-
gram is also available for $40 from the Department of En-
vironmental Engineering, University of Florida, Gainesville.
The Florida version for PCs requires a math co-processor
whereas the CEAM version contains an emulation routine
for PCs without co-processors. At the CEAM, contact
Mr. David Disney
Environmental Research Laboratory
U.S. Environmental Protection Agency
College Station Road
Athens, Georgia 30613
(404) 546-3123
Documentation (Huber et al., 1988; Roesner et al., 1988)
can be obtained from the NTIS or from the University of
Florida ($56):
Dr. Wayne C. Huber
Dept. of Environmental Engineering Sciences
University of Florida
Gainesville, Florida 32611-2013
(904) 392-0846
13. References
Brandstetter, A. 1976. Assessment of Mathematical Models
for Storm and Combined Sewer Management, EPA-600/
2-76-175a (NTIS PB-259597), Environmental Protection
Agency, Cincinnati, OH.
Cunningham, B.A. and W.C. Huber. 1987. Economic and
Predictive Reliability Implications of Stormwater Design
Methodologies, Publication No. 98, Florida Water Re-
sources Research Center, University of Florida, Gaines-
ville, 146 pp.
Dever, R.J., Jr., L.A. Roesner and I.A. Aldrich. 1983.
Urban Highway Storm Drainage Model Vol. 4, Surface
Runoff Program User's Manual and Documentation,
FHWA/RD-83/044, Federal Highway Administration,
Washington, DC.
Dever, R.J., Jr., L.A. Roesnef, and D-C., Woo. 1981.
"Development and Application of a Dynamic Urban
Highway Drainage Model," Urban Stormwater Hydraul-
ics and Hydrology, Proc. Second International
Conference on Urban Storm Drainage, Urbana, IL,
Water Resources Publications, Littleton, CO, pp.
229-235.
Heaney, J.P., W.C. Huber, H. Sheikh, M.A. Medina, J.R.
Doyle, W.A. Peltz and J.E. Darling. 1975. Urban
Stormwater ManagementModelingandDecision-Maldng,
EPA-670/2-75-022 (NTIS PB-242290), Environmental
Protection Agency, Cincinnati, OH, 186 pp.
Heeps, D.P. and R.G. Mein. 1974. "Independent Compari-
son of Three Urban Runoff Models," J. Hydraulics Divi-
sion, Proc. ASCE, 100(HY7):995-1009.
Huber, W.C. 1986. "Deterministic Modeling of Urban
Runoff Quality." In: Urban Runoff Pollution, Pro-
ceedings of the NATO Advanced Research Workshop on
Urban Runoff Pollution^ Montpellier, France, August
1985, H.C. Torno, J. Marsalek, and M. Desbordes,
eds., Springer-Verlag, New York, Series G: Ecological
Sciences 10:167-242.
Huber, W.C. 1989. "User Feedback on Public Domain
Software: SWMM Case Study" In: Proc. Sixth Confer-
ence on Computing in Civil Engineering, Atlanta,
Georgia, ASCE, New York.
Huber, W.C. and R.E. Dickinson. 1988. Storm Water
Management Model Version 4, User's Manual, EPA/
600/3-88/OOla (NTIS PB88-236641/AS), Environmental
Protection Agency, Athens, GA, 569 pp.
Huber, W.C. and J.P. Heaney. 1982. "Analyzing Residuals
Discharge and Generation from Urban and Non-Urban
Land Surfaces," Chapter 3, pp. 121-243, In: Analyzing
Natural Systems, Analysis for Regional Residuals-^
Environmental Quality Management, Basta, D.J., and
B.T. Bower, eds., Resources for the Future, Washing-
ton, D.C., The Johns Hopkins University Press, Balti-
more, MD, (also available as EPA-600/3-83-046, NTIS
PB83-223321).
Huber, W.C., J.P. Heaney and B.A. Cunningham. 1985.
Storm Water Management Model (SWMM) Bibliog-
raphy, ' EPA/600/3-85/077 (NTIS PB86-136041/AS),
. Environmental Protection Agency, Athens, GA, 58 pp.
Huber, W.C., J.P. Heaney, M.A Medina, W.A. Peltz, H.
Sheikh and G.F. Smith. 1975. Storm Water Management
Model User's Manual, Version U, EPA-670/2-75-017
(NTIS PB-257809), Environmental Protection Agency,
Cincinnati, OH, 351 pp. ,
Huber, W.C., J.P. Heaney, S.J. Nix, R.E. Dickinson and
D.J. Polmann. 1981. Storm Water Management Model
User's Manual, Version HI, EPA-600/2-84-109a (NTIS
PB84-198423), Environmental Protection Agency, Cin-
cinnati, OH, 531 pp.
Metcalf and Eddy, University of Florida, and Water
Resources Engineers. 1971. Storm Water Management
Model, Vol. I, Final Report, 11024DOC07/71 (NTIS
PB-203289)EPA Water Pollution Control Research Series
Washington, D.C., 352 pp.
41
-------
Office of Technology Assessment. 1982. Use of Models for
Water Resources Management, Planning, and Policy,
U.S. Govt. Printing Office, Washington, DC.
Roesner, L.A., R.P. Shubinski and J.A. Aldrich. 1981.
Storm Water Management Model User's Manual, Ver-
sion HI: Addendum I, EXTRAN, EPA-600/2-84-109b
(NTIS PB84-198431), Environmental Protection Agency,
Athens, GA.
Roesner, L.A., J.A. Aldrich and R.E. Dickinson. 1988.
Storm Water Management Model User's Manual, Ver-
sion4, EXTRAN Addendum, EPA/600/3-88/001b (NTIS
PB88-236658/AS), Environmental Protection Agency,
Athens, GA.
42
-------
Nonpoint Source Model Review
1. Name of Method
Storage, Treatment, Overflow, Runoff Model (STORM)
2. Type of Method
xx Surface Water Model:
Surface Water Model:
, Soil (Groundwater) Model:
Simple Approach
Refined Approach
Simple Approach
Soil (Groundwater) Model: Refined Approach
3- Purpose/Scope
Purpose: Predict rainfall/runoff/quality processes in urban areas. Predict hydrographs and pollutographs (concentration vs.
time) in . . • •
xx runoff waters
surface waters
ground waters
Source/Release Types:
xx Continuous
. Single
Level of Application:
xx Screening
Types of Chemicals:
xx Conventional
Unique Features:
Intermittent
. Multiple
Intermediate
Organic
xx Diffuse
Detailed
Metals
Addresses degradation products
Integral database/database manager
xx Integral uncertainty analysis capabilities .
Interactive input/execution manager
4. Level of Effort
System setup: xx mandays
Assessments: . mandays
xx manweeks
xx manweeks
manmonths
xx manmonths
(Estimates reflect brder-of-magnitude values and depend heavily on the experience and ability of the assessor.)
43
-------
5. Description of Method/Techniques
The Storage, Treatment, Overflow, Runoff Model
(STORM) contains simplified hydrologic and water quality
routines for continuous simulation in urban areas. Hourly
runoff depths are computed by means of an area-weighted
runoff coefficient for the pervious and impervious portions,
with recovery of depression storage between events. Alter-
natively, the SCS method can be used to generate hourly
runoff volumes. There is no flow routing as such. Runoff
passes through a treatment device up to its capacity and is
otherwise diverted to a storage unit. If treatment is at capac-
ity and the storage is full, an "overflow" occurs to the re-
ceiving water. This scheme results from the origins of the
model for simulation of combined sewer overflows. At the
end of the storm, remaining storage is routed through treat-
ment. The program is driven by hourly precipitation records
obtained from the National Climatic Data Center, and snow-
melt can be simulated using the degree-day method. Dry-
weather flows can also be simulated; these provide a base-
flow and are mixed with stormwater runoff during a storm
event.
Linear build-up and first-order exponential wash-off is used
to simulate concentrations of up to six pre-specified pol-
lutants (suspended solids, settleable solids, BOD, total
coliforms, ortho-phosphate, and nitrogen). Because build-up
and wash-off parameters are adjustable, other conservative
pollutants could be simulated under the guise of the above
names. Erosion may be simulated using the Universal Soil
Loss Equation. Runoff is mixed with dry-weather flow, if
any, and transported without quality routing or decay to the
treatment device. Flows passing through treatment are not
included as loads to receiving waters (i.e., treatment re-
leases are handled as if there is 100% removal). However,
a summary is maintained of concentrations and loads that
bypass the storage/treatment option and overflow to the
receiving water. No treatment occurs in the storage device.
6. Data Needs/AvaBabatty
Data needs are somewhat less for STORM than for compa-
rable continuous simulation models because of its simple
hydrologic and water quality routines. Runoff coefficients
may be estimated from standard handbooks and textbooks,
and SCS parameters are widely available if the soil types are
known. Build-up and wash-off parameters are less flexible
than in models such as SWMM, HSPF or DR3M-QUAL
and no more easy to estimate. Calibration is advisable but
somewhat awkward because the model is primarily set up to
run in only a continuous mode. However, since STORM is
usually run only in a screening or planning mode, compara-
tive evaluations can usually be made without calibration.
7. Output of the Assessment
Output includes storm event summaries of runoff volume,
concentrations and loads plus summaries of storage and
treatment utilization and total overflow loads and concen-
trations. Hourly hydrographs and pollutographs (concentra-
tion vs. time) may be computed but rarely are because there
is no way to do this for only brief periods during a long,
continuous simulation. Useful statistical summaries on an
annual and total simulation period basis that enable an esti-
mation of "percent control," i.e., percentage of runoff pass-
ing through storage and also the number of overflows, both
as a function of the treatment rate and storage capacity. The
storage-treatment combination can then be optimized. The
utilization of storage is also summarized, which helps in
defining the duration of critical events, e.g., time required
between events for complete drainage of the storage.
8. Limitations
STORM's hydrologic routines (runoff coefficient, SCS, no
routing) are the simplest of any simulation model con-
sidered. Although they lead to minimal data requirements,
they also lead to less flexibility in matching observed hydro-
graphs. Similarly, STORM uses the quality routines embod-
ied in the original SWMM program (Metcalf and Eddy et
al., 1971) with very few modifications. These have been
shown to be relatively inflexible in matching observed pol-
lutographs and have been updated in SWMM and other
models. Only hourly precipitation inputs are possible, mak-
ing it difficult to work with more recent continuous records
of 15-minute precipitation data or arbitrary input hyeto-
graphs, e.g., measured rainfall and runoff data. Lack of an
agency-supported microcomputer version currently hampers
the model's use.
9. Hardware/Software Requirements
STORM is written in Fortran IV and must be compiled by
the user on a mainframe. It is available only on a 9-track
magnetic tape and has been installed on IBM and CDC sys-
tems, among others. Although individual users have pre-
pared versions for IBM PC/AT-compatibles, the model
developer does not sell or support such a version.
10. Experience
STORM represents the first significant use of continuous
simulation, especially quality applications, in the urban
setting. (The Stanford Watershed Model, incorporated into
HSPF, was the first continuous model.) The model was
developed by the Army Corps of Engineers, Hydrologic
Engineering Center (HEC) originally for application to the
San Francisco master drainage plan (Roesner et al., 1974;
HEC, 1977) for abatement of combined sewer overflows
(CSOs) into San Francisco Bay (McPherson, 1974). Sixty-
two years of hourly rainfall data were analyzed with
STORM to provide optimal combinations of storage and
treatment for the master plan. The HEC also provides addi-
tional guidelines for model application (Abbott, 1977).
The support of the HEC, renowned for the suite of hydro-
logic models, led to extensive use of STORM in the late 70s
and early 80s, but the lack of HEC updates in recent years
has caused a decline in model use. Nonetheless, many appli-
cations may be found, including Heaney et al. (1977) for a
nationwide assessment, Shubinski et al. (1977) for the
Detroit area, and Najarian et al. (1986) in New Jersey.
44
-------
A simplified receiving water quality model was developed
for streams for use with the STORM model (Medina, 1979),
although Medina's model could also be driven by other con-
tinuous simulation models.
A very similar model to STORM was developed by Lager
et al. (1976) for application in San Francisco and Rochester,
known as "Simplified SWMM." This model is not currently
available, and STORM is essentially a substitute.
11. Validation/Review
STORM has been used in many applications although sel-
dom compared against measured rainfall-runoff data; one
such comparison is by Abbott (1978) in which model predic-
tions compared favorably with hourly runoff values. The
model's many applications by diverse users serve as a form
of validation.
12. Contact
In 1989, the distribution of the most popular of the HEC
programs, e.g., HEC-1 and HEC-2 and some others, was
transferred to selected private vendors. .Direct HEC support
for these programs is limited to Corps of Engineers or other
federal agencies. However, STORM is still available direct-
ly only from the HEC for $200 on a 9-track magnetic tape.
No microcomputer version is available. Only the Fortran TV
source code is distributed; the user must perform the com-
pilation. Contact
The Hydrologic Engineering Center
Corps of Engineers
609 Second Street
Davis, California 95616
13. References
Abbott, J. 1977. Guidelines for Calibration and Application of
STORM, Training Document No. 8, Hydrologic Engineer-
ing Center, Corps of Engineers, Davis, CA.
Abbott, J. 1978. Testing of Several Runoff Models on an
Urban Watershed, ASCE Urban Water Resources Research
Program, Technical Memorandum No. 34, ASCE, New
York.
Heaney, J.P., W.C. Huber, M.A. Medina, M.P. Murphy, SJ.
Nix and S.M. Hasan. 1977. Nationwide Evaluation of
Combined Sewer Overflows and Urban Stonnwater Dis-
charges, Vol. II: Cost Assessment and Impacts, EPA-600/
2-77-064b (NTIS PB-242290), Environfcnental Protection
Agency, Cincinnati, OH.
Hydrologic Engineering Center. 1977. Storage, Treatment,
Overflow, Runoff Model, STORM, User's Manual, Gener-
alized Computer Program 723-S8-L7520, Corps of Engi-
neers, Davis, CA.
McPherson, M.B. 1974. "Innovation: A Case Study," ASCE
Urban Water Resources Research Program, Technical
Memorandum No. 21 (NTIS PB-232166), ASCE, New
York.
Medina, M.A. 1979. Level ffl: Receiving Water Quality
Modeling for Urban Stonnwater Management, EPA-600/
2-79-100 (NTIS PB80-134406), Environmental Protection
Agency, Cincinnati, OH.
Metcalf and Eddy, Inc., University of Florida, Water
Resources Engineers, Inc. 1971. Storm Water Management
Model, Volume I: Final Report, Report 11024DOC07/71
(NTIS PB-203289), Environmental Protection Agency,
Washington, DC.
Najarian, T.O, T.T. Griffin and V.K. Gunawardana. 1986.
"Development Impacts on Water Quality: A Case Study"
Journal Water Resources Planning and Management, ASCE,
112(l):20-35.
Roesner, L.A., H.M. Nichandros, R.P. Shubinski, A.D.
Feldman, J.W. Abbott and A.O. Friedland. 1974. A Model
for Evaluating Runoff-Quality in Metropolitan Master
Planning, ASCE Urban Water Resources Research Pro-
gram/Technical Memorandum No. 23 (NTIS PB-234312);
ASCE, New York.
Shubinski, R.P., A.J. Knepp and C.R. Bristol. 1977.
Computer Program Documentation for 'the Continuous
Storm Runoff ModelSEM-STORM, Report to the "Southeast
Michigan Council of Governments, Detroit, MI.
45
-------
Nonpoint Source Model Review
1. Name of Method
Distributed Routing Rainfall Runoff Model—Quality DR3M-QUAL
2. Type of Method
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Surface Water Model:
XX Surface Water Model:
Soil (Groundwater) Model:
Soil (Groundwater) Model:
3. Purpose/Scope
tip • ' .
Purpose: Predict rainfall/runoff/quality processes in urban and other areas. Predict hydrographs and pollutographs (concentra-
tion vs. time) in
xx runoff waters
_xx, surface waters
ground waters
Source/Release Types:
xx Continuous
Single
Level of Application:
Screening
Types of Chemicals:
xx, Conventional
Unique Features:
xx Intermittent
Multiple
xx Intermediate
xx Organic
xx Diffuse
xx Detailed
xx Metals
xx Addresses degradation products
Integral database/database manager
Integral uncertainty analysis capabilities
__ Interactive input/execution manager
4. Level of Effort
System setup:
Assessments:
xx mandays
___ mandays
xx manweeks
xx manweeks
manmonths
xx manmonths
(Estimates reflect order-of-magnitude values and depend heavily on the experience and ability of the assessor.)
46
-------
5. Description of the Method/Techniques
The Distributed Routing Rainfall Runoff Model—Quality
(DR3M-QUAL) incorporates water quality routines into an
updated version of an earlier U.S. Geological Survey
(USGS) urban hydrologic model (Dawdy et al., 1972). Run-
off is generated from rainfall using the kinematic wave
method over multiple subcatchments and routed through
drainage pathways by the same technique. Storage-indication
routing is available for storage basins. The model can be
run over any time period and is sometimes used to simulate
a group of storms while bypassing simulation of the inter-
vening dry periods (although a moisture balance is main-
tained). A built-in optimization routine aids in estimation of
quantity parameters.
Quality is simulated for arbitrary parameters using exponen-
tial build-up functions plus wash-off functions determined
from experience with model calibration. Considerable guid-
ance is provided for parameter estimation. Removal of built-
up solids can occur during dry weather by street cleaning.
Erosion is simulated using empirical equations relating
sediment yield to runoff volume and peak. Some guidance
is provided for the erosion parameters using relationships
based on the Universal Soil Loss Equation. Concentrations
of other constituents can be taken as a fraction ("potency
factors") of sediment concentration. Precipitation can con-
tribute a constant concentration.
Quality routing through the drainage network is done by a
Lagrangian scheme to simulate plug flow and no decay.
Plug flow routing is also performed in storage basins, with
settling based on sedimentation theory and dependent on a
particle size distribution.
6. Data Needs/Availability
Data needs depend on the degree of schematization. Quan-
tity parameters for a subcatchment include area, impervious-
ness, length, slope, roughness and infiltration parameters.
Channels are characterized by trapezoidal or circular dimen-
sions and kinematic wave parameters. Storage basins require
stage-area-discharge relationships. Quality parameters in-
clude build-up and wash-off coefficients. Rainfall input can
be for single or multiple storms.
Quantity data are similar in form to those required by other
urban hydrologic models and may be derived from contour
maps and drainage plans available from municipalities.
Build-up and wash-off parameters are very difficult to esti-
mate a priori and require local, site-specific quality mea-
surements for calibration if accurate quality predictions are
needed, e.g., to drive a receiving water quality model.
7. Output of the Assessment
DR3M-QUAL produces a time history of runoff hydro-
graphs and quality pollutographs (concentration or load vs.
time) at any location in the drainage system. Summaries for
storm events are provided as well as line printer graphics
for hydrographs and pollutographs.
8. Limitations
The kinematic wave is perhaps as good a conceptual model
as exists for overland flow but approximations are always
inherent in its application, as for any conceptual model. The
quantity model can be expected to perform reasonably well
with minimat calibration. No interaction among quality pa-
rameters exists (other than the ability to treat one pollutant
as a fraction of sediment concentration). Except for the
sedimentation algorithm within storage units, simulation of
sediment transport processes is weak, as with virtually all
other models of this type. Generally, quality predictions
must be calibrated if accurate concentrations and loads are
required, e.g., to drive a receiving water, quality model. On
the other hand, relative comparisons can be made using gen-
eralized U.S. data for calibration purposes.
3. Hardware/Software Requirements
The program is written in Fortran 77 for IBM or Prime
mainframes. Source code is provided on a 9-track magnetic
tape for compilation and installation,on the user's computer.
10. Experience
The basis for the hydrologic components of DR3M-QUAL
is the earlier modeling work of Dawdy et al. (1972) that
was updated first for the quantity portion of the model
(Dawdy et al., 1978; Alley and Smith, 1982a) and then for
additional quality routines (Alley and Smith, 1982b). Much
of the emphasis on the form and calibration of build-up and
wash-off parameters is based on research done by Alley et
al. (1980), Alley (1981), and Alley and Smith (1981). The
program has received extensive internal review within the
USGS and has been applied to their urban modeling studies
in South Florida (Doyle and Miller, 1980), Rochester (Kap-
pel et al., 1985; Zarriello, 1988), Anchorage (Brabets,
1986), Denver (Lindner-Lunsford and Ellis, 1987), and
Fresno (Guay and Smith, 1988). A summary of experience
with the model is given by Alley (1986).
11. Validation/Review
The program has undergone the extensive internal peer
review process of the USGS and has been tested and com-
pared with field data on,at least 37 catchments by various
individuals (Alley, 1986). It has primarily been used by
persons within the USGS although it is freely available to
anyone.
12. Contact
The Fortran 77 source code and documentation are available
on a 9-track magnetic tape (to be supplied by the user) from
the USGS National Center:
Ms. Kate Flynn
U.S. Geological Survey
410 National Center
. Reston, Virginia 22092
(703) 648-5313
47
-------
Minimal support is available from the program authors
(Alley and Smith).
13. References
Alley, W.M. 1981. "Estimation of Impervious-Area Washoff
Parameters," Water Resources Research, 17(4): 1161-1 166.
Alley, W.M. 1986. "Summary of Experience with the Distrib-
uted Routing Rainfall-Runoff Model (DR3M)." In: Urban
Drainage Modeling, C. Maksimovic and M. Radojkovic,
eds., Proc. of International Symp. on Comparison of Urban
Drainage Models with Real Catchment Data, Dubrovnik,
Yugoslavia, Pergamon Press, New York, pp. 403-415.
Alley, W.M., F.W. Ellis and R.C. Sutherland. 1980. "Toward
A More Deterministic Urban Runoff Quality Model," Proc.
of the International Symposium on Urban Runoff, Univer-
sity of Kentucky, Lexington, pp. 171-182.
Alley, W.M. and P.E. Smith. 1981. "Estimation of Accumu-
lation Parameters for Urban Runoff Quality Modeling,"
Water Resources Research, 17(6): 1657-1664.
Alley, W.M. and P.E. Smith. 1982a. Distributed Routing
Rainfall-Runoff Model— Version H, USGS Open File Report
82-344, Reston, VA.
Alley, W.M. and P.E. Smith. 1982b. Multi-Event Urban Run-
off Quality Model, USGS Open File Report 82-764, Reston,
VA*
Brabets, T.P. 1987. Quantity and Quality of Urban Runoff
from the Chester Creek Basin Anchorage, Alaska, USGS
Water-Resources Investigations Report 86-4312, Anchorage,
Dawdy, D.R., R.W. Lichty and J.M. Bergmann. 1972. A
Rainfall-Runoff Simulation Model for Estimation of Flood
Peaks for Small Drainage Basins, USGS Professional Paper
506-B, Reston, VA.
Dawdy, D.R., J.C. Schaake, Jr. and W.M. Alley. 1972.
User's Guide for Distributed Routing Rainfall-Runoff
Model, USGS Water Resources Investigations 78-90, Gulf
Coast Hydroscience Center, NSTL Station, MS.
Doyle, W.H., Jr. and J.E. Miller. 1980. Calibration of a
Distributed Routing Rainfall-Runoff Model at Four Urban
Sites Near Miami, Florida, USGS Water Resources Investi-
gations 80-1, Gulf Coast Hydroscience Center, NSTL
Station, MS.
Guiy, J.R. and P.E. Smith. 1988. Simulation of Quantity and
Quality of Storm Runoff for Urban Catchments in Fresno,
California, USGS Water-Resources Investigations Report
88-4125, Sacramento, CA.
Kappel, W.M., R.M. Yager and P.J. Zarriello. 1986. Quantity
and Quality of Urban Storm Runoff in the Irondequoit
Creek Basin near Rochester, New York, Part 2. Quality of
Storm Runoff and Atmospheric Deposition, Rainfall-Runoff-
Quality Modeling, and Potential of Wetlands for Sediment
and Nutrient Retention, USGS Water-Resources Investiga-
tions Report 85-4113, Ithaca, New York.
Lindner-Lunsford, J.B. and S.R. Ellis. 1987. Comparison of
Conceptually Based and Regression Rainfall-Runoff Models,
Denver Metropolitan Area, Colorado, and Potential Appli-
cations in Urban Areas, USGS Water-Resources Investiga-
tions Report 87-4104, Denver, CO.
Zarriello, P.J. 1988. "Simulated Water-Quality Changes in
Detention Basins." In Design of Urban Runoff Quality
Controls, L.A. Roesner, B. Urbonas and M.B. Sonnen,
eds., Proc. of Engineering Foundation Conference, Potosi,
MO, ASCE, New York, pp. 268-277.
48
-------
Nonpoint Source Model Review
1. Name of the Method
EPA Screening Procedures
2. Type of Method
xxx Surface Water Model:
Surface Water Model:
Air Model:
Air Model:
Soil (Groundwater) Model:
Soil (Groundwater) Model:
Multi-media Model:
Multi-media Model:
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
3. Purpose/Scope
Purpose: Predict concentrations of contaminants in
xxx runoff waters
xxx surface waters
ground waters
Source/Release Types:
Continuous
' Single
Level of Application:
xxx Screening
Type of Chemicals:
xxx Conventional
Unique Features:
Intermittent
xxx Multiple
Intermediate
xxx Organic
xxx Diffuse
Detailed
xxx Metals
Addresses degradation products
Integral Database/Database manager
Integral Uncertainty Analysis Capabilities
Interactive Input/Execution Manager
4. Level of Effort
System setup: xx mandays manweeks
Assessments: xx mandays xx manweeks
. manmonths
manmohths
manyear
manyear
(Estimates reflect order-of-magnitude values and depend heavily on the experience and ability of the assessor.)
49
-------
5. Description of the Method/Techniques
EPA Screening Procedures (Mills et al., 1982, 1985) are a
revision and expansion of water quality assessment proce-
dures initially developed for nondesignated 208 areas (Zison
et al., 1977). The procedures have been expanded and re-
vised to include consideration of the accumulation, trans-
port, and fate of toxic chemicals, in addition to conventional
pollutants included in the earliest version. The manual in-
cludes a separate chapter describing calculation procedures
for estimating NFS loads for urban and nonurban land areas
in addition to chapters on procedures for rivers and streams,
impoundments, and estuaries. A useful overview chapter on
the aquatic fate processes of toxic organisms is also
included.
The procedures for NFS load assessments described in the
manual are essentially a compilation and integration of esti-
mation techniques developed earlier by Midwest Research
Institute (McElroy et al., 1976), Amy et al. (1974), Heany
et al. (1976) (i.e., SWMM-Level I), and Haith (1980). How-
ever, the presentation of the procedures is well integrated,
supplemented with additional parameter estimation information,
and includes sample calculations. The procedures for nonurban
areas are derived from MRI loading functions for average
annual estimates based on Universal Soil Loss Equation
(Wischmeier and Smith, 1978), while the storm event proce-
dures use the modified USLE (Williams, 1975) and the SCS
Runoff Curve Number procedure (Mockus, 1972) for storm
runoff. Pollutant concentrations in soils and enrichment
ratios are required for estimating pollutant loads; precipi-
tation contributions of nutrients can be included. Specific
information is included for estimation of salinity loads in
irrigation return flows, and storm event pesticide losses
are estimated separately based on procedures developed by
Haith (1980).
Annual and storm event loads from urban areas follow the
procedures included in the SWMM-Level I (Heaney et al.,
1976) and Amy et al. (1974) respectively. The basic proce-
dures are relatively standard for urban areas, calculating
pollutant loads as a function of solids accumulation and
washoff.
The EPA Screening Procedures have excellent user docu-
mentation and guidance, including occasional workshops
sponsored by the EPA Water Quality Modeling Center,
Athens, Georgia.
6. Data Needs/Availability
The data requirement is minimal and is also available from
tho manual when site specific data is lacking. Soil and land
use data is required for estimating runoff and sediment
yield. This data is readily available from USDA-SCS soil
survey reports. Tables for estimating water quality input
parameters for nitrogen, phosphorous, heavy metals, and
pesticides are included in the manual.
7. Output of the Assessment
Output available from the model include predicted stream
concentrations of BOD, DO, total N, total P, temperature,
and conservative and organic pollutants by reach; total lake
concentrations, organic pollutants eutrophic status, and
hypolimnion DO deficit; and estuary concentrations of
BOD, DO, total N, total P, and conservative pollutants by
reach. As calculations are done on a hand calculator they
can be arranged according to user's convenience.
8. Limitations
It is a very simplified procedure requiring the user to
estimate pollutant concentrations. It use is very limited while
evaluating impacts of various management practices. As cal-
culations are done on a hand calculator, they can be tedious
and time consuming for complex multi-media use basins.
Snowmelt runoff and loadings are ignored.
9. Hardware/Software Requirements
No computer requirement, calculations can be done by hand
calculators.
10. Experience
These procedures have enjoyed wide popularity, partly due
to the availability of training workshops sponsored by EPA,
and have been applied in a number of regions, including the
Sandusky River in northern Ohio and the Patuxent, Ware,
Chester, and Occuquan basins in the Chesapeake Bay region
(Davis et al., 1981; Dean et al., 1981a; and Dean et al.,
1981b).
11. Validation/Review
12. Contact
For copies of the manual contact
Center for Exposure Assessment Modeling
U.S. Environmental Protection Agency
Environmental Research Laboratory
Athens, GA. 30613
(404) 546-3123
13. References
Amy, G., R. Pitt, R. Singh, W.L. Bradford and M.B.
LaGraff. 1974. Water Quality Management Planning for
Urban Runoff. U.S. EPA, Washington, D.C., EPA 440/9-
75-004. (NTIS PB 241 689/AS).
Davis, M.J., M.K. Snyder and J.W. Nebgen. 1981. River
Basin Validation of the Water Quality Assessment
Methodology for Screening Nondesignated 208 Areas—
Volume I: Nonpoint Source Load Estimation. U.S. EPA,
Athens, GA.
50
-------
Dean, 7.D., W.B. Mills and D.B. Porcella. 1981a. A
Screening Methodology for Basin Wide Water Quality
Management. Symposium on Unified River Basin Manage-
ment. R.M. North, LIB. Dwoesky, and D.J. Alice (edi-
tors), May 4-7, 1980, Gaflinburg, TN.
Dean, J.D., B. Hudson and W.B. Mills. 1981b. River Basin
Validation of the MRI Nonpoint Calculator and Tetra Tech's
Nondesignated 208 Screening Methodologies, Volume n.
Chesapeake-Sandusky Nondesignated 208 Screening Meth-
odology Demonstration. U.S. EPA, Athens, GA.
Haith, D.A. 1980. "A Mathematical Model for Estimating
Pesticide Losses in Runoff". J. ofEnv. Qual. 9(3):428-433.
Heaney, J.P., W.C. Huber and SJ. Nix. 1976. Storm Water
Management Model, Level I, Preliminary Screening Proce-
dures. EPA-600/2-76-275, U.S. Environmental Protection
Agency. Cincinnati, OH.
McElroy, A.D., S.Y. Chiu, J.W. Nebgen, A. Aleti and F.W.
Bennett. 1976. Loading Functions for Assessment of Water
Pollution from Nonpoint Sources. EPA-600/2-76-151. U.S.
Environmental Protection Agency. Washington, D.C.
Mills, W.B., V.H. Colber and J.D. Dean. 1979. Hand-Held
Calculator Programs for Analysis of River Water Quality
Interactions. Tetra Tech Inc., Lafayette, CA.
Mills, W.B., J.D. Dean, D.B. Porcella, S.A. Gherini, R.J.M.
Hudson, W.E. Frick, G.L. Rupp and G.L. Bowie. 1982.
Water Quality Assessment: A Screening Procedure for Tox-
ic and Conventional Pollutants. EPA-600/6-82-004a and b.
Volumes I and n. U.S. Environmental Protection Agency.
Mills, W.B., D.B. Porcella, M.J. Ungs, S.A. Gherini, K.V.
Summers, L. Mok, G.L. Rupp, G.L. Bowie and D.A.
Haith. 1985. Water Quality Assessment: A Screening Proce-
dure for Toxic and Conventional Pollutants in Surface and
Ground Water. EPA/600/6-85/002a. U.S. Environmental
Protection Agency.
Mockus, J. 1972. "Estimation of Direct Runoff from Storm
Rainfall." In: National Engineering Handbook, Sec. 4,
Hydrology. U.S. Soil Conservation Service, Washington,
D.C.
Williams, J.R. 1975. "Sediment-Yield Prediction with
Universal Soil Loss Equation Using Runoff Energy Factor."
In: Present and Prospective Technology for Predicting
Sediment Yields and Sources. U.S. Dept. of Agriculture.
ARS-S-40.
Wischmeier, W.H. andD.D. Smim. 1978. Predicting Rainfall-
Erosion Losses: A Guide to Conservation Planning. U.S.
Department of Agriculture, Agriculture Research Service.
Zison, S.W., K. Haven and W.B. Mills. 1977. Water Quality
Assessment: A Screening Methodology for Nondesignated
208 Areas. EPA-600/6-77-023, U.S. EPA, Athens, GA.
51
-------
Nonpoint Source Model Review
1. Name of the Method
Agricultural NonPoint Source Pollution Model—(AGNPS)
2. Type of Method
Surface Water Model:
SXX Surface Water Model:
__ Air Model:
Air Model:
Soil (Groundwatcr) Model:
Soil (Groundwater) Model:
Multi-media Model:
___ Multi-media Model:
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
3. Purpose/Scope
Purpose: Predict concentrations of contaminants in
xxx runoff waters
surface waters
ground waters
Source/Release Types:
Continuous
Single
Level of Application:
xx% Screening
Type of Chemicals:
xxx Conventional
Unique Features:
Intermittent
xxx Multiple ,
xxx Intermediate
, Organic
Addresses degradation products
Integral Database/Database manager
Integral Uncertainty Analysis Capabilities
XXX Interactive Input/Execution Manager
4. Level of Effort
xxx Diffuse
xxx Detailed
Metals
System setup:
Assessments:
xx mandays
mandays
xx manweeks
xx manweeks
manmonths
xx manmonths
manyear
. manyear
(Estimates reflect order-of-magnitude values and depend heavily on the experience and ability of the assessor.)
52
-------
5. Description of the Method/Techniques
Agricultural Nonpoint Source Pollution Model (AGNPS) was
developed by the U.S. Department of Agriculture—Agricul-
ture Research Service (Young et al., 1986) to obtain uni-
form and accurate estimates of runoff quality with primary
emphasis on nutrients and sediments and to compare the
effects of various pollution control practices that could be
incorporated into the management of watersheds.
The AGNPS model simulates sediments and nutrients from
agricultural watersheds for a single storm event or for con-
tinuous simulation. Watersheds examined by AGNPS must
be divided into square working areas called cells. Grouping
of cells results in the formation of subwatersheds, which can
be individually examined. The output from the model can be
used to compare the watershed examined against other
watersheds to point sources of water quality problems, and
to investigate possible solutions to these problems.
AGNPS is also capable of handling point source inputs from
feedlots, waste water treatment plant discharges, and stream
bank and gully erosion (user specified). In the model, pol-
lutants are routed from the top of the watershed to the outlet
in a series of steps so that flow and water quality at any
point in the watershed may be examined. The Modified Uni-
versal Soil Loss Equation is used for predicting soil erosion,
and a unit hydrograph approach used for the flow in the
watershed. Erosion is predicted in five different particle
sizes namely sand, silt, clay, small aggregates, and large
aggregates. »
The pollutant transport portion is subdivided into one part
handling soluble pollutants and another part handling sedi-
ment attached pollutants. The methods used to predict nitro-
gen and phosphorus yields from the watershed and individ-
ual cells were developed by Frere et al. (1980) and are also
used in CREAMS (Knisel, 1980). The nitrogen and phos-
phorus calculations are performed using relationships be-
tween chemical concentration, sediment yield and runoff
volume.
Data needed for the model can be classified into two
categories: watershed data and cell data. Watershed data
includes information applying to the entire watershed which
would include watershed size, number of cells in the
watershed, and if running for a single storm event then the
storm intensity. The cell data includes information on the
parameters based on the land practices in the cell.
Additional model components that are under development
are unsaturated/saturated zone routines, economic analysis,
and linkage to Geographic Information System.
6. Data Needs/AvaQabDity
The input data needed are extensive however, can be ob-
tained through visual field observations, maps (topographic
and soils) and from various publications, table and graphs
(Young et al. 1986). Meteorologic data consisting of daily
rainfall is needed for hydrology simulation. The model also
has an option to evaluate watershed response to a single
storm event. Data on soil and land use can be obtained from
local USDA-SCS field offices.
7. Output of the Assessment
The model provides estimates of: (1) hydrology, with esti-
mates of both runoff volume and peak runoff rate; (2) sedi-
ment, with estimates of upland erosion, channel erosion, and
sediment yield; and nutrients (both sediment attached and
dissolved), with estimates of pollution loadings to receiving
cells. A graphics option in the program allows the user to
plot different variables within the watershed.
8. Limitations
The model does not handle pesticides. The pollutant
transport component needs further field testing. Nutrient
transformation and instream processes are not within model
capabilities.
9. Hardware/Software Requirements
The program is written in standard FORTRAN 77 and has
been installed on IBM PC/AT and compatibles. A hard disk
is required for operation of the program and a math co-
processor is highly recommended. Executable code prepared
with Ryan-McFarland compiler and is available only for
MS/DOS environment. Source code is only available for
MS/DOS environment.
10. Experience
The model is being used extensively within United States to
evaluate nonpoint source pollution by various government
agencies and consultants. The model has been used by Shi
(1987) to perform economic assessment of soil erosion and
water quality in Idaho. Setia and Magleby (1987) and Setia
et al. (1988) used AGNPS for evaluating the economic ef-
fect of nonpoint pollution control alternatives. The model
has also been used by Koelliker and Humbert (1989) for
water quality planning. APNPS was used by Frevert and
Crowder (1987) to analyze agricultural nonpoint pollution
control options in the St. Albans Bay watershed.
11. Validation/Review
The model has been validated using field data from agricul-
tural watersheds in Minnesota, Iowa, and Nebraska (Young
et al., 1986). Lee (1987) validated the model in an Illinois
watershed using the single storm option of the model. The
author found that the simulated and observed data for runoff
volume and sediment yield were well represented when
compared with observed data.
53
-------
12. Contact
The copies of the AGNPS program and the manual are
available from
Dr. Robert Young
USDA-ARS
North Central Research Laboratory
Morris, MN 56267
Phone: (612) 589-3411
13. References
Frere, M.H., J.D. Ross and L.J. Lane. 1980. "The Nutrient
Submodel." In: Khisel, W.G. (ed.). 1980. CREAMS: A
Field Scale Model for Chemicals, Runoff, and Erosion from
Agricultural Management Systems. U.S. Department of Ag-
riculture. Conservation Research Report No. 26. pp. 65-86.
Frevert, K. and B.M. Crowder. 1987. "Analysis of agricul-
tural nonpoint pollution control options in the St. Albans
Bay watershed." Economic Division, ERS, USDA. Staff
Report No. AGES870423.
Knisel, W.G. (ed.). 1980. CREAMS: A Field Scale Model for
Chemicals, Runoff, and Erosion from Agricultural Manage-
ment Systems. U.S. Department of Agriculture. Conserva-
tion Research Report No. 26. 640 pp.
Lee, M.T. 1987. "Verification and Applications of a Nonpoint
Source Pollution Model." In: Proceedings of the National
Engineering Hydrology Symposium. ASCE, New York,
NY.
Setia, P.P. and R.S. Magleby. 1987. "An Economic Analysis
of Agricultural Nonpoint Pollution Control Alternatives."
Jour, of Soil and Water Conservation. 42:427-431.
Seb'a, P.P., R.S. Magleby and D.G Carvey. 1988. "Illinois
Rural Clean Water Project—An Economic Analysis." Re-
sources and Technology Division, ERS, USDA. Staff
Report No. AGES830617.
Shi, H.Q. 1987. "Integrated Economic Assessment of Soil
Erosion and Water Quality in Idaho's Tom Beall Water-
shed." M.Sc. Thesis. University of Idaho.
Young, R.A., C.A. Onstad, D.D. Bosch and W.P. Anderson.
1986. Agricultural Nonpoint Source Pollution Model: A
Watershed Analysis Tool. Agriculture Research Service,
U.S. Department of Agriculture, Morris, MN.
54
-------
Nonpoint Source Model Review
1. Name of the Method
Areal Nonpoint Source Watershed Environment Response Simulation—(ANSWERS)
2. Type of Method
Surface Water Model:
xxx Surface Water Model:
Air Model:
Air Model:
Soil (Groundwater) Model:
xxx Soil (Groundwater) Model:
Multi-media Model:
Multi-media Model:
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
3. Purpose/Scope
Purpose: Predict concentrations of contaminants in
xxx runoff waters
surface waters
ground waters
Source/Release Types:
Continuous
Single
Level of Application:
xxx Screening
Type of Chemicals:
xxx Conventional
Unique Features:
Intermittent
Multiple
xxx Intermediate
Organic
Addresses degradation products
xxx Integral Database/Database manager
Integral Uncertainty Analysis Capabilities
Interactive Input/Execution Manager
4. Level of Effort
xxx Diffuse
Detailed
Metals
System setup:
Assessments:
xx mandays
mandays
.manweeks
xx manweeks
manmonths
xx manmonths
manyear
manyear
(Estimates reflect order-of-magnitude values and depend heavily on the experience and ability of the assessor.)
55
-------
5. Description of the Method/Techniques
Areal Nonpoint Source Watershed Environment Response
Simulation (ANSWERS) was developed at the Agricultural
Engineering Department of Purdue University (Beasley and
Huggins, 1981). It is an event based, distributed parameter
model capable of predicting the hydrologic and erosion
response of agricultural watersheds.
Application of ANSWERS requires that the watershed to be
subdivided into a grid of square elements. Each element
must be small enough so that all important parameter values
within its boundaries are uniform. For a practical applica-
tion element sizes range from one to four hectares. Within
each element the model simulates the processes of intercep-
tion, infiltration, surface storage, surface flow, subsurface
drainage, and sediment drainage, and sediment detachment,
transport, and deposition. The output from one element then
becomes a source of input to an adjacent element.
As the model is based on a modular program structure it
allows easier modification of existing program code and/or
addition of user supplied algorithms. Model parameter
values are allowed to vary between elements, thus, any
degree of spatial variability within the watershed is easily
represented.
Nutrients (nitrogen and phosphorus) are simulated using
correlation relationships between chemical concentrations,
sediment yield and runoff volume. A research version
(Amin-Sichani, 1982) of the model uses "clay enrichment"
information and a very descriptive phosphorus fate model to
predict total, particulate, and soluble phosphorus yields.
6. Data Needs/AvaHabilfty
Data need comprise of detailed description of the watershed
topography, drainage network, soils, and land use. Most of
the data can be obtained from USDA-SCS soil surveys, land
use and cropping surveys.
7. Output of the Assessment
The model can evaluate alternative erosion control manage-
ment practices for both agricultural land and construction
sites (Dillaha et al., 1982). Output can be obtained on an
element basis or for the entire watershed in terms of flow
and sediment. The ANSWERS program comes with a plot-
ting program.
8. Limitations
For a simulation run of ANSWERS on a large watershed a
mainframe computer is required. However, for smaller
watersheds a IBM-PC compatible version of ANSWERS is
available. The model is a storm event model and the input
data file is quite complex to prepare. Snowmelt processes or
pesticides cannot be simulated by the model. The water
quality constituents modeled are limited to nitrogen and
phosphorous. These constituents are represented by rela-
tionships between chemical concentrations with sediment
yield and runoff volume. No transformation of nitrogen and
phosphorus is accounted for in the model.
3. Hardware/Software Requirements
The program is written in standard FORTRAN 77 and has
been installed on IBM PC/AT and compatibles. A hard disk
is required for operation of the program and a math co-
processor is highly recommended. Executable code prepared
with Ryan-McFarland compiler.
10. Experience
The model has been successfully applied in Indiana on an
agricultural watershed and a construction site to evaluate
best management practices by Beasley (1986).
11. Validation/Review
Individual components of the model have been validated by
the developers.
12. Contact
To obtain copies of the model please write or call Dr. David
Beasley at the following address:
Dr. David Beasley, Professor and Head
Dept. of Agricultural Engineering
University of Georgia
Coastal Plain Experiment Station
P.O. Box 748
Tifton, GA 31793
(912)386-3377
13. References
Amin-Sichani, S. 1982. "Modeling of Phosphorus Transport in
Surface Runoff from Agricultural Watersheds." Ph.D.
Thesis, Purdue University, W. Lafayette, Indiana, 157 pp.
Beasley, D.B. and L.F. Huggins. 1981. ANSWERS Users
Manual. EPA-905/9-82-001, U.S. EPA, Region V.
Chicago, IL.
Beasley, D.B. 1986. "Distributed Parameter Hydrologic and
Water Quality Modeling." In: Agricultural Nonpoint Source
Pollution: Model Selection and Application. A. Giorginiand
F. Zingales (editors), pp. 345-362.
Beasley, D.B. and D.L. Thomas. 1989. Application of Water
Quality Models for Agricultural and Forested Watersheds.
Southern Cooperative Series Bulletin No. 338. Agricultural
Experiment Station, University of Georgia, Athens, GA.
Dillaha, T.A. ffl, D.B. Beasley and L.F. Huggins. 1982.
"Using the ANSWERS Model to Estimate Sediment Yields
on Construction Sites." J. of Soil and Water Cons.,
37(2): 117-120.
56
-------
Nonpoint Source Model Review
1. Name of the Method
Chemicals, Runoff, and Erosion from Agricultural Management Systems—(CREAMS)
Groundwater Loading Effects of Agricultural Management Systems—(GLEAMS)
2. Type of Method
Surface Water Model:
xxx Surface Water Model:
Air Model:
Air Model:
Soil (Groundwater) Model:
xxx Soil (Groundwater) Model:
Multi-media Model:
Multi-media Model:
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
3. Purpose/Scope
Purpose: Predict concentrations of contaminants in
xxx runoff waters
surface waters
xxx ground waters (vadose and root zone)
Source/Release Types:
Continuous
Single
Level of Application:
xxx Screening
Type of Chemicals:
xxx Conventional
Unique Features:
Intermittent
xxx Multiple
xxx Intermediate
xxx Organic
xxx Diffuse
xxx Detailed
Metals
xxx Addresses degradation products
Integral Database/Database manager
Integral Uncertainty Analysis Capabilities
Interactive Input/Execution Manager
4. Level of Effort
System setup: xx mandays xx manweeks
Assessments: mandays xx manweeks
mamnonths
xx manmonths
manyear
manyear
(Estimates reflect order-of-magnitude values and depend heavily on the experience and ability of the assessor.)
57
-------
5. Description of the Method/Techniques
Chemicals, Runoff, and Erosion from Agricultural Manage-
ment Systems (CREAMS) was developed by the U.S.
Department of Agriculture—Agricultural Research Service
(Knisel, 1980; Leonard and Ferreira, 1984) for the analysis
of agricultural best management practices for pollution
control. CREAMS is a field scale model that uses separate
hydrology, erosion, and chemistry submodels connected
together by pass files.
Runoff volume, peak flow, infiltration, evapotranspiration,
soil water content, and percolation are computed on a daily
basis. If detailed precipitation data are available then
infiltration is calculated at histogram breakpoints. Daily
erosion and sediment yield, including particle size distri-
bution, are estimated at the edge of the field. Plant nutrients
and pesticides are simulated and storm load and average
concentrations of sediment-associated and dissolved chemi-
cals are determined in the runoff, sediment, and percolation
through the root zone (Leonard and Knisel, 1984).
User defined management activities can be simulated by
CREAMS. These activities include aerial spraying (foliar or
soil directed) or soil incorporation of pesticides, animal
waste management, and agricultural best management prac-
tices (minimum tillage, terracing, etc.).
Calibration is not specifically required for CREAMS simula-
tion, but is usually desirable. The model provides accurate
representation of the various soil processes. Most of the
CREAMS parameter values are physically measurable. The
model has the capability of simulating 20 pesticides at one
time.
Groundwater Loading Effects of Agricultural Management
Systems (GLEAMS) was developed by the United States
Department of Agriculture—Agriculture Research Service
(Leonard et al., 1987) to utilize the management oriented
physically based CREAMS model (Knisel, 1980) and incor-
porate a component for vertical flux of pesticides in the root
zone. A nitrogen component for the root zone is in
development.
GLEAMS consists of three major components namely
hydrology, erosion/sediment yield, and pesticides. Precipi-
tation is partitioned between surface runoff and infiltration
and water balance computations are done on a daily basis.
Surface runoff is estimated using the Soil Conservation
Service Curve Number Method as modified by Williams and
Nicks (1982). The soil is divided into various layers, with
a minimum of 3 and a maximum of 12 layers of variable thick-
ness are used for water and pesticide routing (Knisel et al.,
1989).
6. Data Needs/Availability
The data needed for CREAMS are quite detailed. As
CREAMS is a continuous simulation model the data needs
are extensive. Meteorologic data consisting of daily or
breakpoint precipitation is required for hydrology simula-
tion. Monthly solar radiation and air temperature data is also
needed for estimating components of the hydrological cycle.
Data regarding soil type and properties along with informa-
tion on crops to be grown is needed. A broad range of
values for various model parameters can be obtained from
the user's manual.
7. Output of the Assessment
Various output options are available for hydrology and
nutrient simulations, including storm, monthly, or annual
summary. Output for each segment of the overland flow and
channel elements is available from areas in the watershed
where intense erosion or deposition can be identified.
8. Limitations
The maximum size of the simulated area is limited to a field
plots. A watershed scale version (Opus) of CREAMS/
GLEAMS is currently under development (Ferreira, person-
al communication). The model is limited in data manage-
ment and handling. The model cannot simulate instream
processes. Although CREAMS has been applied in a wide
range of climatic regimes, there is concern regarding its
simulation capability for snow accumulation, melt, and
resulting runoff, and hydrologic impacts of frozen ground
conditions (see Jamieson and Clausen, 1988; Kauppi, 1982;
Knisel etal., 1983).
9. Hardware/Software Requirements
The program is written in standard FORTRAN 77 and has
been installed on IBM PC/AT-compatibles. A hard disk is
required for operation of the program and a math co-
processor is highly recommended. The program can be
obtained on floppy disk for MS/DOS operating systems.
10. Experience
CREAMS has been extensively applied in a wide variety of
hydrologic settings with good success. CREAMS has been
used for in a wide variety of hydrologic and water quality
studies (Smith and Williams, 1980; Morgan and Morgan,
1982; Lane and Ferreira, 1980; Kauppi, 1982; Knisel et al.,
1983; and Jamieson and Clausen, 1988). Crowder et al.
(1985) have used CREAMS in conjunction with an eco-
nomic model to evaluate effects of conservation practices.
11. Validation/Review
The model has been validated by the developers along with
independent experts. Erosion/sedimentation component of
the CREAMS model has been verified by Foster and Fer-
reira (1981). Smith and Williams (1980) have validated the
hydrology submodel at 46 sites in the southern and mid-
western portions of United States. GLEAMS model has
been validated with field data for Fenamiphos and its
metabolites by Leonard et al., 1990. The authors report
satisfactory comparison between observed and simulated
results.
58
-------
12. Contact
To obtain copies of the model please write or call Dr. Walt
Knisel or Frank Davis at the following address:
USDA-Agricultural Research Service
Southeast Watershed Research Lab
P.O. Box 946
Tifton, Georgia 31793
(912) 386-3462
13. References
Beasley, D.B. and D.L. Thomas. 1989. Application of Water
Quality Models for Agricultural and Forested Watersheds.
Southern Cooperative Series Bulletin No. 338. Agricultural
Experiment Station, University of Georgia, Athens, GA.
Crowder, B.M., H.B. Pionke, D.J. Epp and C.E. Young.
1985. "Using CREAMS and Economic Modeling to Evalu-
ate Conservation Practices: An Application." Journal of
Environmental Quality, 14(3):428-434.
Foster, G.R. and V.A. Ferreira. 1981. "Deposition in Uniform
Grade Terrace Channels." In: Crop Production with Con-
servation in the 8O's. American Society of Agricultural
Engineers, St. Joseph, Michigan, pp. 185-197.
Jamieson, C.A. and J.C. Clausen. 1988. Tests of the
CREAMS Model on Agricultural Fields in Vermont. Water
Resources Bulletin, 24(6): 1219-1226.
Kauppi, L. 1982. "Testing the Application of CREAMS to
Finnish Conditions." In: European and United States Case
Studies in Application of the CREAMS Model. V. Svetlosa-
nov and W.G. Knisel (editors). International Institute for
Applied Systems Analysis. Laxenberg, Austria, pp. 43-47.
Knisel, W. (ed). 1980. CREAMS: A Field Scale Model for
Chemicals, Runoff, and Erosion from Agricultural Manage-
ment Systems. U.S. Department of Agriculture. Conserva-
tion Research Report No. 26. 640 pp.
Knisel, W.G., G.R. Foster and R.A. Leonard. 1983.
"CREAMS: A System for Evaluating Management Prac-
tices." In: Agricultural Management and Water Quality.
F.W. Schaller and G.W. Bailey (editors). Iowa State
Uniyersity Press, Ames, Iowa, pp. 178-199.
Knisel, W.G., R.A. Leonard and P.M. Davis. 1989.
"Agricultural Management Alternatives: GLEAMS Model
Simulations." Proceedings of the Computer Simulation
Conference. Austin, Texas, July 24-27, pp. 701-706.
Lane, L.J. and V.A. Ferreira. 1980. "Sensitivity Analysis."
In: CREAMS: A Field-Scale Model for Chemicals, Runoff,
and Erosion from Agricultural Management Systems. W.G.
Knisel (editor). U.S. Department of Agriculture, Conser-
vation Report No. 26, U.S. Government Printing Office,
Washington, D.C., pp. 113-158.
Leonard, R.A., W.G. Knisel and D.A. Still. 1987.
"GLEAMS: Groundwater Loading Effects of Agricultural
Management Systems." Trans, of the ASAE, 30(5): 1403-
1418.
Leonard, R.A. and V.A. Ferreira. 1984. CREAMS2—The
Nutrient and Pesticide Models. Proc. Natural Resources
Modeling Symposium. Agricultural Research Service. U.S.
Department of Agriculture.
Leonard, R.A. and W.G. Knisel. 1984. Model Selection for
Nonpoint Source Pollution and Resource Conservation. In:
Proc. of the International Conference on Agriculture and
Environment 1984. Venice, Italy, pp. E1-E18.
Leonard, R.A., W.G. Knisel, P.M. Davis and A.W. Johnson.
1990. "Validating GLEAMS with Field Data for Fenami-
phos and its Metabolites." Journal of Irrigation and
Drainage Engineering, 116(l):24-35.
Morgan, R.P.C. and D.D.V. Morgan. 1982. "Predicting
Hillslope Runoff and Erosion in the United Kingdom:
Preliminary Trials with the CREAMS Model." In: Euro-
pean and United States Case Studies in Application of the
CREAMS Model. V. Sveflosanov and W.G. Knisel (edi-
tors). International Institute for Applied Systems Analysis.
Laxenburg, Austria, pp. 83-97.
Smith, R.E. and J.R. Williams. 1980. "Simulation of the
Surface Water Hydrology." In: CREAMS: A Field-Scale
Model for Chemicals, Runoff, and Erosion from Agricul-
tural Management Systems. W.G. Knisel (editor). U.S.
Department of Agriculture, Conservation Report No. 26,
U.S. Government Printing Office, Washington, D.C., pp.
13-35.
59
-------
Nonpoint Source Model Review
1. Name of the Method
Hydrological Simulation Program—Fortran (HSPF)
Stream Transport and Agricultural Runoff of Pesticides for Exposure Assessment (STREAM)
2. Type of Method
Surface Water Model:
xxx Surface Water Model:
Air Model:
Air Model:
Soil (Groundwater) Model:
. Soil (Groundwater) Model:
Multi-media Model:
Multi-media Model:
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
3. Purpose/Scope
Purpose: Predict concentrations of contaminants in
225 runoff waters
xxx surface waters
xxx ground waters
Source/Release Types:
xxx Continuous
xxx Single
Level of Application:
XXX Screening
Type of Chemicals:
xxx Conventional
Unique Features:
xxx Intermittent
xxx Multiple
xxx Intermediate
xxx Organic
XXX Addresses degradation products
xxx Integral Database/Database manager
___ Integral Uncertainty Analysis Capabilities
xxx Interactive Input/Execution Manager
4. Level of Effort
xxx Diffuse
* xxx Detailed
Metals
System setup:
Assessments:
xx mandays
mandays
xx manweeks
xx manweeks
manmonths
xx manmonths
. manyear
. manyear
(Estimates reflect order-of-magnitude values and depend heavily on the experience and ability of the assessor.)
60
-------
5. Description of the Method/Techniques
The result of this simulation is a time history of the runoff
flow rate, sediment load, and nutrient and pesticide concen-
trations, along with a time history of water quantity and
quality at any point in a watershed. HSPF simulates three
sediment types (sand, silt, and clay) in addition to a single
organic chemical and transformation products of that chemi-
cal. The transfer and reaction processes included are hydrol-
ysis, oxidation, photolysis, biodegradation, a volatilization,
and sorption. Sorption is modeled as a first-order kinetic
process in which the user must specify a desorption rate and
an equilibrium partition coefficient for each of the three
solid types. Resuspensioa and settling of silts and clays
(cohesive solids) are defined in terms of shear stress at the
sediment-water interface. For sands, the capacity of the
system to transport sand at a particular flow is calculated
and resuspension or settling is defined by the difference
between the sand in suspension and the capacity. Calibration
of the model requires data for each of the three solids types.
Benthic exchange is modeled as sorption/desorption and
desorption/scour with surficial benthic sediments. Underly-
ing sediment and pore water are not modeled.
6. Data Needs/AvaDabDity
Data needs for HSPF are extensive. HSPF is a continuous
simulation program and requires continuous data to drive the
simulations. As a minimum, continuous rainfall records are
required to drive the runoff model and additional records of
evapotranspiration, temperature, and solar intensity are
desirable. A large number of model parameters can also be
specified although default values are provided where reason-
able values are available. HSPF is a general-purpose pro-
gram and special attention has been paid to cases where
input parameters are omitted. Option flags allow bypassing
of whole sections of the program where data are not
available.
7. Output of the Assessment
HSPF produces a time history of the runoff flow rate,
sediment load, and nutrient and pesticide concentrations,
along with a time history of water quantity and quality at
any point in a watershed. Simulation results can be pro-
cessed through a frequency and duration analysis routine
that produces output compatible with conventional lexicolog-
ical measures (e.g., 96-hour LC50).
8. Limitations
HSPF assumes that the "Stanford Watershed Model" hydro-
logic model is appropriate for the area being modeled.
Further, the instream model assumes the receiving water
body model is well-mixed with width and depth and is thus
limited to well-mixed rivers and reservoirs. Application of
this methodology generally requires a team effort because of
its comprehensive nature.
9. Hardware/Software Requirements
The program is written in standard FORTRAN 77 and has
been installed on systems as small as IBM PC/AT-
compatibles. A hard disk is required for operation of the
program and a math co-processor is highly recommended.
No special peripherals other than a printer are required. The
program is maintained for both the IBM PC-compatible and
the DEC/VAX with VMS operating system. Executable
code prepared with the Ryan-McFarland FORTRAN com-
piler and PLINK86 linkage editor is available for the
MS/DOS environment. Source code only is available for the
VAX environment.
The program can be obtained in either floppy disk format
for MS/DOS operation systems or on a 9-TRK magnetic tape
with installation instructions for the DEC VAX VMS
environment. This program has been installed on a wide range
of computers world-wide with no or minor modifications.
10. Experience
HSPF and the earlier models from which it was developed
have been extensively applied in a wide variety of hydrolog-
ic and water quality studies (Barnwell and Johanson, 1981;
Barnwell and Kittle, 1984) including pesticide runoff testing
(Lorber and Mulkey, 1981), aquatic fate and transport mod-
el testing (Mulkey et al., 1986; Schnoor et al., 1987) analy-
ses of agricultural best management practices (Donigian et
al., 1983a; 1983b; Imhoff et al., 1983) and as part of pesti-
cide exposure assessments in surface waters (Mulkey and
Donigian, 1984).
An application of HSPF to five agricultural watersheds in a
screening methodology for pesticide review is given in
Donigian (1986). The Stream Transport and Agricultural
Runoff for Exposure Assessment (STREAM) Methodology
applies the HSPF program to various test watersheds for
five major crops in four agricultural regions in the U.S.,
defines a "representative watershed" based on regional con-
ditions and an extrapolation of the calibration for the test
watershed, and performs a sensitivity analysis on key pesti-
cide parameters to generate cumulative frequency distribu-
tions of pesticide loads and concentrations in each region.
The resulting methodology requires the user to evaluate only
the crops and regions of interest, the pesticide application
rate, and three pesticide parameters—the partition coeffi-
cient, the soil/sediment decay rate, and the solution decay
rate.
61
-------
11. Validation/Review
The program has been validated with both field data and
model experiments and has been reviewed by independent
experts. Numerous citations for model applications are
included in the References below. Recently, model refine-
ments for instream algorithms related to pH and sediment-
nutrient interactions have been sponsored by the USGS and
the EPA Chesapeake Bay Program, respectively.
12. Contact
The model is available from the Center for Exposure
Assessment Modeling at no charge. Mainframe versions of
the programs compatible with the DEC VAX systems are
available on standard on-half inch, 9-track magnetic tape.
When ordering tapes, please specify the type of computer
system that the model will be installed on (VAX, PRIME,
HP, Cyber, IBM, etc.), whether the tape should be non-
labeled, if non-labeled specify the storage formate (EBCDIC
or ASCII), or if the tape should be formatted as a VAX
files-11, labeled (ASCII) tape for DEC systems. Model
distributions tapes contain documentation covering instal-
lation instructions on DEC systems, FORTRAN source code
files, and test input data sets and output files that may be
used to test and confirm the installation of the model on
your system. Users are responsible for installing programs.
Requests for PC versions of the models should be accompa-
nied by 8 formatted double-sided, double-density (DS/DD),
error free diskettes. Please do not send high density (DD/
HD) diskettes. Model distribution diskettes contain docu-
mentation covering installation instructions on PC systems,
DOS batch files for compiling, linking, and executing the
model, executable task image(s) ready for execution of the
model(s), all associated runtime files, and test input data
sets and corresponding output files that may be used to test
and confirm the installation of the model on your PC or
compatible system.
To obtain copies of the models, please send 9-track specifi-
cations or the appropriate number of formatted diskettes to
the attention of David Disney at the following address:
Center for Exposure Assessment Modeling
U.S. Environmental Protection Agency
Environmental Research Laboratory
Athens, Georgia 30613
(404) 546-3123
USA
Program and/or user documentation, or instructions on how
to order documentation, will accompany each response.
13. References
Bamwell, T.O. 1980. An Overview of the Hydrologic Simula-
tion Program—FORTRAN, a Simulation Model for Chemi-
cal Transport and Aquatic Risk Assessment. In: Aquatic
Toxicology and Hazard Assessment: Proceedings of the Fifth
Annual Symposium on Aquatic Toxicology, ASTM Special
Tech. Pub. 766, ASTM, 1916 Race Street, Philadelphia,
PA 19103.
Bamwell, T.O. and R. Johanson. 1981. HSPF: A Compre-
hensive Package for Simulation of Watershed Hydrology
and Water Quality. In: Nonpoint Pollution Control: Tools
and Techniques for the Future. Interstate Commission on the
Potomac River Basin, 1055 First Street, Rockville, MD
20850.
Barnwell, T.O. and J.L. Kittle. 1984. "Hydrologic Simulation
Program—FORTRAN: Development, Maintenance and
Applications." In: Proceedings Third International Confer-
ence on Urban Storm Drainage. Chalmers Institute of Tech-
nology, Goteborg, Sweden.
Bicknell, B.R., A.S. Donigian Jr. and T.O. Barnwell. 1984.
Modeling Water Quality and the Effects of Best Manage-
ment Practices in the Iowa River Basin. J. Wat. Sci. Tech.,
17:1141-1153.
Chew, Y.C., L.W. Moore, and R.H. Smith. 1991. "Hydro-
logic Simulation of Tennessee's North Reelfoot Creek
watershed" J. Water Pollution Control Federation 63(1):
10-16.
Donigian, A.S., Jr., J.C. Imhoff and B.R. Bicknell. 1983.
Modeling Water Quality and the Effects of Best Manage-
ment Practices in Four Mile Creek, Iowa. EPA Contract
No. 68-03-2895, Environmental Research Laboratory, U.S.
EPA, Athens, GA. 30613.
Donigian, A.S., Jr., J.C. Imhoff, B.R. Bicknell and J.L.
Kittle, Jr. 1984. Application Guide for the Hydrological
Simulation Program—FORTRAN EPA 600/3-84-066, Envi-
ronmental Research Laboratory, U.S. EPA, Athens, GA.
30613.
Donigian, A.S., Jr., D.W. Meier and P.P. Jowise. 1986.
Stream Transport and Agricultural Runoff for Exposure
Assessment: A Methodology. EPA/600/3-86-011, Environ-
mental Research Laboratory, U.S. EPA, Athens, GA.
30613.
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. Prepared by AQUA TERRA Consultants
for U.S. EPA Chesapeake Bay Program, Annapolis, MD.
Hicks, C.N., W.C. Huber and J.P. Heaney. 1985. Simulation
of Possible Effects of Deep Pumping on Surface Hydrology
Using HSPF. In: Proceedings of Stormwater and Water
Quality Model User Group Meeting. January 31-February
1, 1985. T.O. Barnwell, Jr., ed. EPA-600/9-85/016.
Environmental Research Laboratory, Athens, GA.
Johanson, R.C., J.C. Imhoff, J.L. Kittle, Jr. and A.S.
Donigian. 1984. Hydrological Simulation Program—
62
-------
FORTRAN (HSPF): Users Manual for Release 8.0, EPA-
600/3-84-066, Environmental Research Laboratory, U.S.
EPA, Athens, GA. 30613.
Johansson, R.C. 1989. Application of the HSPF Model to
Water Management in Africa. In: Proceedings of Storm-
water and Water Quality Model Users Group Meeting.
October 3-4, 1988. Guo, et al., eds. EPA-600/9-89/001.
Environmental Research Laboratory, Athens, GA.
Lorber, M.N. and L.A. Mulkey. 1982. An Evaluation of
Three Pesticide Runoff Loading Models. J. Environ. Qual.
11:519-529.
Moore, L.W., H. Matheny, T. Tyree, D. Sabatini and S.J.
Klaine. 1988. "Agricultural Runoff Modeling in a Small
West Tennessee Watershed," J. Water Pollution Control
Federation 60(2):242-249.
Motta, D.J. and M.S. Cheng. 1987. "The Henson Creek
Watershed Study" In: Proceedings of Stormwater and Water
Quality Users Group Meeting. October 15-16, 1987. H.C.
Tomo, ed. Charles Howard and Assoc., Victoria, BC,
Canada.
Mulkey, L.A., R.B. Ambrose, and T.O. Barnwell. 1986.
"Aquatic Fate and Transport Modeling Techniques for
Predicting Environmental Exposure to Organic Pesticides
and Other Toxicants—A Comparative Study." In: Urban
Runoff Pollution, Springer-Verlag, New York, NY.
Nichols, J.C. and M.P. Timpe. 1985. Use of HSPF to simu-
late Dynamics of Phosphorus in Floodplain Wetlands over
a Wide Range of Hydrologic Regimes. In: Proceedings of
Stormwater and Water Quality Model Users Group Meet-
ing, January 31-February 1, 1985. T.O. Bamwell, Jr., ed.
EPA-600/9-85/016, Environmental Research Laboratory,
Athens, GA.
Schnoor, J.L., C. Sato, D. McKetchnie, and D. Sahoo. 1987.
Processes, Coefficients, and Models for Simulating Toxic
Organics and Heavy Metals in Surface Waters. EPA/600/3-
87/015. U.S. Environmental Protection Agency, Athens,
GA. 30613.
Schueler, T.R. 1983. Seneca Creek Watershed Management
Study, Final Report, Volumes I and JJ. Metropolitan
Washington Council of Governments, Washington, DC.
Song, J.A., G.F. Rawl, W.R. Howard. 1983. Lake Manatee
Watershed Water Resources Evaluation using Hydrologic
Simulation Program—FORTRAN (HSPF) m: Colloque sur
la Modelisation des Eaux Pluviales. Septembre 8-9, 1983.
P. Beron, et al., T. Barnwell, editeurs. GREMU—83/03
Ecole Polytechnique de Montreal, Quebec, Canada.
Sullivan, M.P. and T.R. Schueler. 1982. The Piscataway
Creek Watershed Model: A Stormwater and Nonpoint
Source Management Tool. In: Proceedings Stormwabr and
Water Quality Management Modeling and SWMM Users
Group Meeting, October 18-19, 1982. Paul E. Wisner, ed.
Univ. of Ottawa, Dept Civil Engr., Ottawa, Ont., Canada.
Weatherbe, D.G. and Z. Novak. 1985. Development of Water
Management Strategy for the Humber River. In: Proceed-
ings Conference on Stormwater and Water Quality Manage-
ment Modeling, September 6-7,1984. E.M. and W. James,
ed. Computational Hydraulics Group, McMaster University,
Hamilton, Ont., Canada.
Woodruff, D.A., D.R. Gaboury, R.J. Hughto and O.K.
Young. 1981. Calibration of Pesticide Behavior on a
Georgia Agricultural Watershed Using HSP-F. In; Proceed-
ings Stormwater and Water Quality Model Users Group
Meeting, September 28-29, 1981. W. James, ed. Computa-
tional Hydraulics Group, McMaster University, Hamilton,
Ont., Canada.
Udhiri, S., M-S Cheng and R.L. Powell. 1985. The Impact of
Snow Addition on Watershed Analysis Using HSPF. In:
Proceedings of Stormwater and Water Quality Model Users
Group Meeting, January 31-February 1, 1985. T.O. Barn-
well, Jr., ed. EPA-600/9-85/016, Environmental Research
Laboratory, Athens, GA.
63
-------
Nonpoint Source Model Review
1. Name of the Method
Pesticide Root Zone Model—PRZM
2. Type of Method
xxx Surface Water Model:
Surface Water Model:
Air Model:
___ Air Model:
Soil (Groundwater) Model:
xxx Soil (Groundwater) Model:
Multi-media Model:
Multi-media Model:
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
3. Purpose/Scope
Purpose: Predict concentrations of contaminants in
xj£ runoff waters
surface waters
xxx ground waters (vadose and root zone)
Source/Release Types:
Continuous
Single
Level of Application:
xxx Screening
Type of Chemicals:
Conventional
Unique Features:
xxx Intermittent
xxx Multiple
xxx Intermediate
xxx Organic
Addresses degradation products
_ Integral Database/Database manager
_ Integral Uncertainty Analysis Capabilities
_ Interactive Input/Execution Manager
4. Level of Effort
System setup: xx mandays
Assessments: _ mandays
xxx Diffuse
xxx Detailed
Metals (pesticides)
xx manweeks
xx manweeks
manmonths
xx manmonths
manyear
manyear
(Estimates reflect order-of-magnitude values and depend heavily on the experience and ability of the assessor.)
64
-------
5. Description of the Method/Techniques
Pesticide Root Zone Model (PRZM) was developed at the
U.S. EPA Environmental Research Laboratory in Athens,
Georgia by Carsel et al. (1984). It is a one-dimensional,
dynamic, compartmental model that can be used to simulate
chemical movement in unsaturated zone within and imme-
diately below the plant root zone. The model is divided into
two major components namely, the hydrology (and hydrau-
lics) and chemical transport. The hydrology component
which calculates runoff and erosion is based upon the Soil
Conservation Service curve number procedure and the Uni-
versal Soil Loss Equation respectively. Evapotranspiration
is estimated directly from pan evaporation or by an empiri-
cal formula if pan evaporation data is not available. Soil-
water capacity terms including field capacity, wilting point,
and saturation water content are used for simulating water
movement within the unsaturated zone. Irrigation application
is also within model capabilities.
Pesticide application on soil or on the plant foliage are
considered in the chemical transport simulation. Dissolved,
adsorbed, and vapor-phase concentrations in the soil are
estimated by simultaneously considering the processes of
pesticide uptake by plants, surface runoff, erosion, decay,
volatilization, foliar washoff, advection, dispersion, and
retardation. The user has two options to solve the transport
equations using the original backward difference implicit
scheme or the method of characteristics (Dean et al., 1989).
As the model is dynamic it allows considerations of pulse
loads.
PRZM is an integral part of a unsaturated/saturated zone
model RUSTIC (Dean et al., 1989). RUSTIC (Risk of
Unsaturated/Saturated Transport and Transformation of
Chemical Concentrations) links three subordinate models in
order to predict pesticide fate and transport through the
crop root zone, and saturated zone to drinking water wells
through PRZM, VADOFT, SAFTMOD.
VADOFT is a one-dimensional finite element model which
solves Richard's equation for water flow in the unsaturated
zone. VADOFT can also simulate the fate and transport of
two parent and two daughter products. SAFTMOD is a two-
dimensional finite element model which simulates flow and
transport in the saturated zone in either an X-Y or X-Z
configuration. The three codes PRZM, VADOFT, and
SAFTMOD are linked together through an execution super-
visor which allows users to build models for site specific
situation. In order to perform exposure assessments, the
code is equipped with a Monte Carlo pre and post processor
(Dean etal., 1989).
6. Data Needs/AvaHabBity
The meteorological data needed by the model consist of
daily rainfall, potential evaporation, and air temperature. If
pesticide volatilization is to be simulated then additional
meteorological data consisting daily wind speed and solar
radiation are needed. Soils and land use data are required
which can be obtained from local USDA-SCS field offices.
The data regarding pesticide input parameters can be
obtained from the user's manual or from published research.
7. Output of the Assessment
Predictions are made on a daily basis. Output can also be
summarized for a daily, monthly, or annual period. Daily
time series values of various fluxes and soil storages can be
written to sequential files for further evaluation. In addition,
a Special Action' option allows the user to output soil pro-.,
file pesticide concentrations at user specified times (Dean et
al., 1989).
8. Limitations
One of the model limitation is that it one-dimensional in the
vertical direction and hence does not handle lateral flow.
PRZM only simulates downward movement of water and
does not account for diffusive movement due to soil water
gradients. This process has been identified to be important
when simulating the effects of volatilization by Jury et al.
(1984). The model only simulates organic chemicals, for
example pesticides.
9. Hardware/Software Requirements
The program is written in standard FORTRAN 77 and has
been installed on IBM PC/AT-compatibles. A hard disk is
required for operation of the program and a math co-
processor if highly recommended. The program can be
obtained on floppy disk for MS/DOS operating systems.
10. Experience
PRZM has been used to study Aldicarb application to citrus
in Florida (Jones et al., 1983), and potatoes in New York
(Carsel et al., 1985) and Wisconsin (Jones, 1983). It has
also been used for Metalaxyl application to tobacco in
Florida and Maryland (Carsel et al., 1986) and to Atrazine
and chloride application to corn in Georgia (Carsel et al.,
1985).
11. Validation/Review
The PRZM model has undergone testing with field data in
New York and Wisconsin (potatoes), Florida (citrus), and
Georgia (corn) (Carsel, et al., 1985, Jones 1983, Jones et
al., 1983). The results of these tests demonstrate that PRZM
is a useful tool for evaluating groundwater threats for pesti-
cide use.
12. Contact
To obtain copies of the user's manual and the computer pro-
gram contact
Center for Exposure Assessment Modeling
U.S. Environmental Protection Agency
Environmental Research Laboratory
Athens, GA. 30613.
(404) 546-3123
65
-------
13. References
Carsel, R.F., C.N. Smith, L.A. Mulkey, J.D. Dean and P.
Jowise. 1984. User's Manual for the Pesticide Root Zone
Model (PRZM): Release 1. EPA-600/3-84-109. U.S. Envi-
ronmental Protection Agency. Environmental Research
Laboratory, Athens, GA.
Carsel, R.F., L.A. Mulkey, M.N. Lorber and L.B. Baskin.
1985. "The Pesticide Root Zone Model (PRZM): A Proce-
dure for Evaluating Pesticide Leaching Threats to Ground
Water" Ecological Modeling 30:49-69.
Carsel, R.F., W.B. Nixon and L.G. Balentine. 1986.
"Comparison of Pesticide Root Zone Model Predictions
with Observed Concentrations for the Tobacco Pesticide
Metalaxyl in Unsaturated Zone Soils" Environ. Toxicol.
Cfaem. 5:345-353.
Dean, J.D., P.S. Huyakom, A.S. Donigian, Jr., K.A. Voos,
R.W. Schanz, Y.J. Meeks and R.F. Carsel. 1989. Risk of
Unsaturated/Sattirated Transport and Transformation of
Chemical Concentrations (RUSTIC). EPA/600/3-89/048a.
U.S. Environmental Protection Agency. Environmental
Research Laboratory, Athens, GA.
Jones, R.L. 1983. "Movement and Degradation of Aldicarb
Residues in Soil and Ground Water." Presented at the
CETAC Conference on Multidisciplinary Approaches to
Environmental Problems, November 6-9, Crystal City, VA.
Jones, R.L., P.S.C. Rao and A.G. Homsby. 1983. "Fate of
Aldicarb in Florida Citrus Soil 2, Model Evaluation."
Presented at the Conference on Characterization and
Monitoring of Vadose (Unsaturated) Zone, December 8-10,
Las Vegas, NV.
66
-------
Nonpoint Source Model Review
1. Name of the Method
Simulator for Water Resources in Rural Basins—(SWRKB)
Pesticide Runoff Simulator—(PRS)
2. Type of Method
Surface Water Model:
xxx Surface Water Model:
Air Model:
Air Model:
Soil (Groundwater) Model:
xxx Soil (Groundwater) Model:
Multi-media Model:
Multi-media Model:
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
3. Purpose/Scope
Purpose: Predict concentrations of contaminants in
xxx runoff waters
surface waters
xxx ground waters
Source/Release Types:
Continuous
Single
Level of Application:
xxx Screening
Type of Chemicals:
xxx Conventional
Unique Features:
. Intermittent
. Multiple
xxx Intermediate
xxx Organic
Addresses degradation products
Integral Database/Database manager
xxx Integral Uncertainty Analysis Capabilities
Interactive Input/Execution Manager
4. Level of Effort
xxx Diffuse
xxx Detailed
Metals
System setup:
Assessments:
xx mandays
mandays
xx manweeks
xx manweeks
manmonths
xx manmonths
manyear
. manyear
(Estimates reflect order-of-magnitude values and depend heavily on the experience and ability of the assessor.)
67
-------
5. Description of the Method/Techniques
Pesticide Runoff Simulator (PRS) was developed for the
U.S. EPA Office of Pesticide and Toxic Substances by
Computer Sciences Corporation (1980) to simulate pesticide
runoff and adsorption into the soil on small agricultural
watersheds. PRS is based on SWRRB (Simulator for Water
Resources in Rural Basins) originally developed by Williams
et al. (1985) at the USDA.
The PRS hydrology and sediment simulation is based on the
USDA CREAMS model. The SCS curve number technique
is used to predict surface runoff. Sediment yield is simulated
using a modified version of the Universal Soil Loss Equa-
tion and a sediment routing model.
The pesticide component of PRS is a modified version of
the CREAMS pesticide model. Pesticide application (foliar
and soil applied) can be removed by atmospheric loss, wash
off by rainfall, and leaching into the soil. Pesticide yield is
divided into a soluble fraction and an adsorbed phase based
on an enrichment ratio.
The model includes a built in weather generator based on
temperature, solar radiation, and precipitation statistics.
Calibration is not specifically required, but is usually
desirable.
Simulator for Water Resources in Rural Basins (SWRRB)
was developed by Williams et al. (1985), and Xrnold et al.,
(1989) for evaluating basin scale water quality. SWRRB
operates on a daily time step and simulates weather, hydrol-
ogy, crop growth, sedimentation, and nitrogen, phospho-
rous, and pesticide movement. The model was developed by
modifying the CREAMS (Knisel, 1980) daily rainfall hy-
drology model for application to large, complex, rural
basins.
Surface runoff is calculated using the Soil Conservation
Service Curve Number technique. Sediment yield is com-
puted for each basin by using the Modified Universal Soil
Loss Equation (Williams and Berndt, 1977). The channel
and floodplain sediment routing model is composed of two
components operating simultaneously (deposition and degra-
dation). Degradation is based on Bagnold's stream power
concept and deposition is based on the fall velocity of the
sediment particles (Arnold et al., 1989).
Return flow is calculated as a function of soil water content
and return flow travel time. The percolation component uses
a storage routing model combined with a crack flow model
to predict the flow through the root zone. The crop growth
model (Arnold et al., 1989) computes total biomass each
day during the growing season as a function of solar radia-
tion and leaf area index.
The pollutant transport portion is subdivided into one part
handling soluble pollutants and another part handling sedi-
ment attached pollutants. The methods used to predict nitro-
gen and phosphorus yields from the rural basins are adopted
from CREAMS (Knisel, 1980). The nitrogen and phospho-
rus calculations are performed using relationships between
chemical concentration, sediment yield and runoff volume.
The pesticide component is directly taken from Hoist and
Kutney (1989) and is a modification of the CREAMS (Smith
and Williams, 1980) pesticide model. The amount of pesti-
cide reaching the ground or plants is based on a pesticide
application efficiency factor. Empirical equations are used
for calculating pesticide washoff which are based on thresh-
old rainfall amount. Pesticide decay from the plants and the
soil are predicted using exponential functions based on the
decay constant for pesticide in the soil, and half life of
pesticide on foliar residue.
6. Data Needs/AvaHabflrty
Meteorologic data comprising of daily precipitation and
solar radiation are required for hydrology simulations.
Another set of input data consists of soils, land use,
fertilizer, and pesticide application. The soils and land use
data can be obtained from USDA-SCS soil survey maps.
Some guidance is available in the manual for estimating
parameters required for nutrient and pesticide simulation.
7. Output of the Assessment
The model predicts daily runoff volume and peak rate, sedi-
ment yield, evapotranspiration, percolation, return flow, and
pesticide concentration in runoff and sediment.
8. Limitations
There is very minimal model documentation. In the hydrol-
ogy component the snow accumulation processes are ig-
nored, and for the'case of pesticides no comprehensive
instream simulation is available. Nutrient transformations
along with pesticide daughter products are not accounted for
in the model.
' !' ,» ' ...I ',''•' „ ' :: ,<
9. Hardware/Software Requirements
The PRS model is operational on the EPA National Com-
puter Center on an IBM 370/168 computer under MVS. The
model may be accessed via WYLBUR for modification of
the source code, creation or modification of input datasets,
and submission of batch executions of the model.
The SWRRB program is written in standard FORTRAN 77
and has been installed on IBM PC/AT and compatibles. A
hard disk is required for operation of the program and a
math co-processor is highly recommended.
10. Experience
The SWRRB model has been used by the Exposure Assess-
ment Branch, Hazard Evaluation Division, and the Office of
Pesticide Programs of the USEPA (Arnold et al., 1989).
11. Validation/Review
SWRRB was tested on 11 large watersheds by Arnold and
Williams (1987). These watersheds were located at eight
68
-------
Agricultural Research Service locations throughout the
United States. The results showed that SWRRB can realis-
tically simulate water and sediment yield under a wide range
of soils, climate, land-use, topography, and management
systems. •
12. Contact
For copies of the SWRRB program and the user manual
contact
Nancy Sammons
808 East Blackland Road
Temple, Texas 76502
(817)770-6512
13. References
Arnold, J.G., J.R. Williams, A.D. Nicks and N.B. Sammons.
1989. "SWRRB, A Basin Scale Simulation Model for Soil
and Water Resources Management." Texas A&M Press.
255 pp. (La Press).
Arnold, J.G. and J.R. Williams. 1987. "Validation of
SWRRB—Simulator for Water Resources in Rural Basins."
J. of Water Resources Planning and Management.
113(2):243-256.
Computer Science Corporation. 1980. Pesticide Runoff
Simulator User's Manual. U.S. EPA, Office of Pesticides
and Toxic Substances, Washington, D.C.
Hoist, R.W. andL.L. Kutney. 1989. "U.S. EPA Simulator for
Water Resources in Rural Basins." Exposure Assessment
Branch, Hazard Evaluation Division, Office of Pesticide
Programs, U.S. EPA (draft).
Knisel, W.G. (ed.). 1980. "CREAMS, A Field Scale Model
for Chemicals, Runoff, and Erosion from Agricultural
Management Systems." USDA Conservation Research
Report No. 26, 643 pp.
Smith, R.E. and J.R. Williams. 1980. "Simulation of the
Surface Water Hydrology." In: CREAMS, A Field Scale
Model for Chemicals, Runoff, and Erosion from Agri-
cultural Management Systems. W.G. Knisel editor. USDA
Conservation Research Report No. 26, pp. 13-35.
Williams, J.R. and H.D. Berndt. 1977. Sediment yield pre-
diction based on watershed hydrology. Transactions of the
ASAE. 20(6): 1100-1104.
Williams, J.R., A.D. Nicks and J.G. Arnold. 1985. "Simulator
for Water Resources in Rural Basins." ASCE J. Hydraulic
Engineering. lll(6):970-986.
69
-------
Nonpoint Source Model Review
1. Name of the Method
Unified Transport Model for Toxic Materials—UTM-TOX
2. Type of Method
Surface Water Model:
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
Simple Approach
Refined Approach
3. Purpose/Scope
Purpose: Predict concentrations of contaminants in
X2S runoff waters
xjx. surface waters
ground waters (vadose and root zone)
xjcx. Surface Water Model:
Air Model:
xxx, Air Model:
Soil (Groundwater) Model:
Soil (Groundwater) Model:
Multi-media Model:
Multi-media Model:
Source/Release Types:
xxx Continuous
Single
Level of Application:
Screening
Type of Chemicals:
xxx Conventional
Unique Features:
Intermittent
xxx Multiple
Intermediate
xxx Organic
Addresses degradation products
Integral Database/Database manager
, Integral Uncertainty Analysis Capabilities
Interactive Input/Execution Manager
4. Level of Effort
xxx Diffuse
xxx Detailed
xxx Metals
System setup:
Assessments:
xx mandays
mandays
xx manweeks
xx manweeks
manmonths
xx manmonths
. manyear
. manyear
(Estimates reflect order-of-magnitude values and depend heavily on the experience and ability of the assessor.)
70
-------
5. Description of the Method/Techniques
Unified Transport Model for Toxic Materials (UTM-TOX)
was developed by Oak Ridge National Laboratory for the
U.S. EPA Office of Pesticides and Toxic Substances,
Washington, D.C. (Patterson et al., 1983). UTM-TOX is a
multimedia model that combines hydrologic, atmospheric,
and sediment transport in one computer code.
The model calculates rates of flux of a chemical from
release to the atmosphere, through deposition on a water-
shed, infiltration and runoff from the soil, to flow in a
stream channel and associated sediment transport. From
these calculations mass balances can be established, chemi-
cal budgets made, and concentrations in the environment
estimated.
The atmospheric transport model (ATM) portion of UTM-
TOX is a Gaussian plume model mat calculates dispersion
of pollutants emitted from point (stack), area, or line
sources. ATM operates on a monthly time step; which is
longer than the hydrologic portion of the model and results
in the use of an average chemical deposition falling on the
watershed.
The Terrestrial Ecology and Hydrology Model (TEHM) de-
scribes soil-plant water fluxes, interception, infiltration, and
storm and groundwater flow.
The hydrologic portion of the model is from the Wisconsin
Hydrologic Transport Model (WHTM), which is a modified
version of the Stanford Watershed Model (SWM). WHTM
includes all of the hydrologic processes of the SWM and
also simulates soluble chemical movement, litter and
vegetation interception of the chemical, erosion of sorbed
chemical, chemical degradation in soil and litter, and
sorption in top layers of the soil. Stream transport includes
transfer between three sediment components (suspended,
bed, and resident bed).
6. Data Needs/Availability
The input data includes monthly wind, hourly precipitation,
solar radiation, daily maximum and minimum temperatures,
soil characteristics, topographic information, surface water
characteristics, sediment characteristics, and the physio-
chemical properties and transformation rates associated with
the chemical.
7. Output of the Assessment
The output can be obtained in terms of plots and tables
summarizing the average monthly and annual chemical con-
centrations in the 8 wind sectors, in saturated and un-
saturated soil layers, in runoff, out of each reach, and in the
stems, leaves, roots and fruits of vegetation.
8. Limitations
The model ignores the interaction between chemicals and
sediment in streams. There is a large time and spatial reso-
lution of ATM portion of the model relative to the hydro-
logic processes. The model is quite complex and requires
significant user expertise.
3. Hardware/Software Requirements
UTM-TOX is large computer program and is written is
FORTRAN IV. The program was developed for IBM 3707
3033 or VAX 11/780 systems.
10. Experience
An earlier version of the model has been applied by Munro
et al. (1976) to evaluate the movement of lead, cadmium,
zinc, copper and sulphur through Crooked Creek Water-
shed. This earlier version was also used by Huff et al.
(1977) to Walker Branch Watershed. One of the current ap-
plication of the model is reported by Patterson (1986) for
estimating lead transport budget in the Crooked Creek
Watershed. No current references since 1986 on the applica-
tion of the model are available.
11. Validation/Review
The model components have been field validated by several
researchers (Culkowski and Patterson, 1976; Munro et al.,
1976; and Raridon, Fields, and Henderson, 1976).
12. Contact
To obtain copies of the model and the manual contact
Dr. M.R. Patterson
Oak Ridge National Laboratory
Mail Stop 6243
Oak Ridge, Tennessee 37831
(615) 574-5442
13. References
Culkowski, W.M. and M.R. Patterson. 1976. A Comprehen-
sive Atmospheric Transport and Diffusion Model. ORNL/
NSF/EATC-17.
Huff, D.D., G.S. Henderson, C.L. Begovich, R.L. Luxmoore
and J.R. Jones. 1977. The Application of Analytic and
Mechanistic Hydrologic Models to the Study of Walker
Branch Watershed. In: Watershed Research in Eastern
North America. A Workshop to Compare Results. D.L.
Corell (editor). Vol. H, pp. 741-766.
Munro, J.K., R.J. Luxmoore, C.L. Begovich, K.R. Dixon,
A.P. Watson, M.R. Patterson and D.R. Jackson. 1976.
Application of the Unified Transport Model to the
Movement of Pb, Cd, Zn, Cu, and S through the Crooked
. Creek Watershed. ORNL/NSF/EATC-28.
Patterson, M.R., T.J. Sworski, A.L. Sjoreen, M.G. Browman,
C.C. Coutant, D.M. Hetrick, B.D. Murphy and R.J.
Raridon. 1983. A User's Manual for UTM-TOX, A Unified
Transport Model. Draft. Prepared by Oak Ridge National
Laboratory, Oak Ridge, TN, for U.S. EPA Office of Toxic
Substances, Washington, DC.
71
-------
Patterson, M.R.* 1986. Lead Transport Budget in Crooked
Creek Watershed. In: Pollutants in Multimedia Environment.
Plenum Press. 1986. pp. 93-118.
Raridon, R.J., D.E. Fields and G.S. Henderson. 1976.
Hydrologic and Chemical Budgets on Walker Branch
Watershed—Observations and Modeling Applications.
OKNL/NSF/EATC-24.
72
A-U.S. GOVERNMENT PRINTING OFFICE: 1991 - 548-187/40511
------- |