NOTICE: This PDF file was adapted from an on-line training module of the EPA’s

Watershed Academy Web, found at http://www.epa.gov/watertrain. To the extent

possible, it contains the same material as the on-line version. Some interactive parts of the
module had to be reformatted for this non-interactive text presentation.

This document does not constitute EPA policy. Mention of trade names or commercial
products does not constitute endorsement or recommendation for use.

Links to non-EPA web sites do not imply any official EPA endorsement of or responsibility
for the opinions, ideas, data, or products presented at those locations or guarantee the
validity of the information provided. Links to non-EPA servers are provided solely as a
pointer to information that might be useful to EPA staff and the public.

WATERSHED ACADEMY WEB 1 Watershed Modeling

http://www.epa.gov/watertrain
Welcome to the Watershed Modeling Tools module. Modeling is one among many assessment
tools used in watershed planning and management.

This module has three main purposes. It should help you to:

1. Understand when and how modeling can contribute to watershed assessment.

2. Learn approaches and tools that are useful for watershed modeling. Note that the
requirements of watershed assessment can necessitate different tools and approaches
from a traditional point source modeling problem.
3. Understand the considerations in choosing models for watershed assessments.

Throughout the module, underlined terms in bold are in the glossary on page 35.

What is a model?
This module introduces the topic of water quality modeling to support watershed studies. Let’s
start with the simplest question: what are models? In general, models are representations of
systems or processes. Some of the oldest forms of models were actual miniature physical
representations of natural systems. Mathematical models are also representations of systems, but
use a series of mathematical equations. The number, form, and interconnections of these
equations in a model can range from very simple to highly sophisticated.

An example of a collection of processes that might be included in a watershed nonpoint source

model is shown in Figure 1 on the next page. As you can see, watershed nonpoint source
modeling can be rather complex and require knowledge of several ongoing processes. The
processes are divided into those applicable into pervious and impervious areas. The linkages
describe how the model handles the simulation of several state variables, which are the basic
components for which mass is conserved. In this model, the state variables are water, sediment,
and a generic pollutant (such as phosphorus). The model first represents the movement of water,
starting with precipitation and snow melt. In pervious areas (the left hand side) water may
evaporate, run off on the surface, or infiltrate and move through the soil or ground water. In
impervious areas water does not infiltrate, so only evaporation and surface runoff are considered.
The model next simulates the movement of particles, which often carry pollutants. This is done
through erosion in the pervious area, and washoff of particulate matter in the impervious area.
Finally, the model represents movement of pollutants, in association with either water or
particulate matter.

WATERSHED ACADEMY WEB 2 Watershed Modeling

http://www.epa.gov/watertrain
 Figure 1

There are two points to remember as we discuss models:

• Models are a type of tool, and are used in combination with many other assessment
techniques.
• Models are a reflection of our understanding of watershed systems. As with any tool, the
answers they give are dependent on how we apply them, and the quality of these answers
is no better than the quality of our understanding of the system.

WATERSHED ACADEMY WEB 3 Watershed Modeling

http://www.epa.gov/watertrain
Ways in which issues are identified for modeling
Let’s step back and consider the position of a water quality manager, who, confronted by various
kinds of information, has to make a decision about whether modeling would be a useful tool for
assessment. Water quality issues may have been brought to the water quality manager’s attention
in various ways, including:

Informal Assessment: For example, issues might be identified after receiving a public complaint
or observing a nuisance condition. Issues might also be identified as a result of a decline in
resource condition or quality perceived by general public or by experts.

Formal Assessment: Section 106(e)(1) of the Clean Water Act requires each state to establish
appropriate monitoring methods and procedures necessary to compile and analyze data on the
quality of waters of the United States. Under Section 305(b), each State, Territory, and Interstate
Commission provides a biennial report of the current status of water quality to EPA. These
305(b) lists contain a formal assessment of water quality issues identified in monitoring. (See the
web site: www.epa.gov/305b/ for information on section 305(b) programs). For instance,
monitoring might reveal a chronic occurrence of a standard violation, a declining trend in water
quality, or a documented impairment of aquatic or wildlife habitat.

Proactive Need. Issues also arise from a desire to protect existing resources and prevent future
degradation. For example, a local community may want to ensure protection of its drinking water
supply for the next 50-100 years, or a wildlife agency may wish to protect critical habitat.

Since issues are identified

in such various ways
(Figure 2), our background
knowledge of the issues Informal
assessment
may vary considerably. In
(observation/
carrying out an assessment perception)
of a given problem, the
manager needs to ask a
number of questions to
identify what is already
known about magnitude, Informal assessment
sources and causes of (monitoring and
problems or threats to water evaluation)
quality:

Do we know the Proactive

extent of the need
problem or threat (protection/
(spatial scale)? prevention)

Do we know the
existing and Figure 2
projected persistence
of the problem (time scale)?

WATERSHED ACADEMY WEB 4 Watershed Modeling

http://www.epa.gov/watertrain
 • Do we know the severity of the problem or threat (level of risk)?

o Is there a threat to human health?

o Is it causing irreversible ecological damage?
o Is it repairable or restorable?

• Do we know the cause(s) of the problem or threat?

o Is it due to point and/or nonpoint sources?

o Are there multiple sources of the problem?
o Is the problem exacerbated by interaction with other stressors, including chemical
stressors, physical stressors, or the alteration or loss of habitat?

Having asked and answered these questions to the extent possible with existing information, the
manager must decide if further assessment is required, and if so, what type of assessment. Often
the manager will be confronted by choice: monitor or model? Modeling is useful for many
purposes, but it may not always be the best tool for a given situation. If one has the choice, real
monitoring data are always preferable to model predictions (and, of course, the quality of any
modeling effort depends a great deal on the quality and quantity of data available to it).

The first step in choosing a model for watershed assessment is to step back and decide if a model
is needed at all, or whether another assessment tool might be better for the given situation. The
next several screens present questions which will often clarify whether a model would be useful.

Is modeling needed to scope or quantify a problem?

When a water quality issue is first identified, the level of understanding of the severity and
sources of the problem is often limited. Modeling is frequently used to help build understanding
of a water quality problem. Typically, simple “scoping” models are useful to help you quickly
estimate the extent and severity of a problem. Figure 3 shows a checklist of problem descriptors
that a scoping model can address.

For example, consider a lake which is

believed to receive excess loads of
nutrients. Does this loading result in a
degree of eutrophication which impairs
uses?

Both monitoring and modeling can be

used to help answer this question.
Monitoring actual lake responses would
be more reliable, but there is not always
the luxury of time and funding necessary
to collect the data. Further, a few scattered
measurements may not be very
informative. Lakes often exhibit a high
Figure 3

WATERSHED ACADEMY WEB 5 Watershed Modeling

http://www.epa.gov/watertrain
degree of variability in algal response both within a season, and year to year, so a few
chlorophyll a measurements of algal concentrations may not provide a clear answer.

In some cases, a simple scoping model can quickly be applied to give a general answer or
estimate of the risk posed to water quality. For example, you could use an empirical model,
which predicts trophic state based on a statistical relationship between phosphorus load and lake
retention time, and algal concentration.

Simple models can also help to estimate runoff flows or contaminant loads for the purpose of
assessing relative magnitudes, and thus targeting areas of greatest risk. This is another area
where modeling can be more cost-effective than monitoring if a high degree of accuracy is not
required for initial estimates. For example, simple models, such as loading coefficients, can aid
in identifying areas where runoff is greatest, and areas which are likely to generate the largest
loads of a given pollutant. Such modeling is particularly useful for obtaining initial estimates of
nonpoint loads; it is difficult to gather monitoring data on nonpoint runoff flow and pollutant
loads. This makes modeling of runoff-generated loading an attractive option.

Recommended further reading: Mills, W.B. et al. 1985. Water Quality Assessment: A Screening
Procedure for Toxic and Conventional Pollutants in Surface and Ground Water. EPA/600/6-
85/002.

Is modeling needed to predict how conditions are expected

to change over time?
Models are also useful for extrapolating from current conditions to potential future conditions.
Indeed, it is not possible to monitor the future, so modeling is the default choice.

An example of the type of question a manager may wish to answer is:

Given the expected rate of population growth and development within a

watershed, and accompanying conversion of land use and land cover, what do we
expect to happen to environmental resources under existing management
measures and regulatory programs? Are these protections adequate, or are
additional management strategies needed?

To answer this type of question, we could either consult a clairvoyant or use a model that helps
convert projections concerning land use and population change into a prediction of watershed
conditions and waterbody response (see Figure 4 on the next page).

WATERSHED ACADEMY WEB 6 Watershed Modeling

http://www.epa.gov/watertrain
 Look into the Crystal
Ball Or

Figure 4

Is modeling needed to evaluate alternative management

options?
The ability of models to predict future conditions (Figure 5) is very useful for projecting the
outcome(s) of various possible management measures and strategies (see:
www.epa.gov/OWOW/NPS/MMGI/ for guidance specifying management measures for
sources of nonpoint pollution in coastal waters). Modeling is thus a tool to aid in selecting
desired management options. For example, you may wish to compare the impacts of various
possible ways of allocating waste among a number of known sources of some pollutant, as is
required in the Total Maximum Daily Load (TMDL) provisions of the Clean Water Act. Or,
you may be considering a range of management strategies, and you would like to know the
answers to questions like:

• Should we consider modifying local land

use ordinances as a means of managing
runoff loading of pollutants to a
waterbody?
• What best management practices (BMPs)
are likely to be the most effective, or most
cost-effective in controlling pollutant
loads?
• Would resource management measures
like protection of riparian buffers or flow
controls be effective?
• Would increased treatment of waste water
(point source control) or increased
treatment of drinking water (an Figure 5
engineering approach to controlling risks
to human health) suffice to solve the water
quality problem?

WATERSHED ACADEMY WEB 7 Watershed Modeling

http://www.epa.gov/watertrain
 • Is there a useful role for ecological restoration techniques, such as restoring channel form
or riparian shading, in managing the problem?
• What combinations of management options are likely to be most effective?

What’s different about watershed modeling?

How will it change my job? What additional difficulties will it present?

Given that modeling can and often is needed to answer questions related to watershed
assessment, let’s consider the aspects of watershed modeling that differ from more traditional
modeling of point source impacts. You may be quite accustomed to using mathematical models
to evaluate water quality problems associated with point sources. You may now be wondering
how the new paradigm of watershed management is going to change your job. In other words:
What other types of models may be needed to address watershed systems? What additional
difficulties does it present?

Let’s start by considering an example of a typical point source problem. We’ll then expand this
problem to a watershed scale to see what sort of additional factors will need to be considered.

Consider that a publicly owned treatment works (POTW) has requested an increase in the
permitted amount of oxygen-demanding waste it is allowed to discharge to a river. As shown in
Figure 6 on the next page, we can apply a steady-state water quality model to estimate the
dissolved oxygen (DO) depletion associated with this source at a design low flow condition in
the receiving water (e.g., the 7Q10 flow).

In this case, the steady-state EPA model QUAL2E was applied (See
www.epa.gov/QUAL2E_WINDOWS/ for more on QUAL2E). River miles increase in the
upstream direction, by convention. The outfall of interest is at river mile 17. As the oxygen-
demanding waste is carried downstream it begins to deplete oxygen within the river. This effect
increases as river velocity slows, decreasing aeration, below river mile 14. The analysis was
made at design conditions, consisting of proposed new permitted effluent flow, 7Q10 low flow in
the receiving water, and late summer monthly average water temperatures. (Higher temperatures
result in more biological oxygen demand and lower potential for the water to hold oxygen).
Under these conditions, DO concentrations are predicted to be below the state minimum DO
standard of 5 mg/l between river mile 13 and river mile 6. Therefore, the proposed increased load
is not acceptable.

The careful observer will note that there is also some depression of DO levels above the effluent
discharge. This is due to nonpoint sources, which can also be incorporated into a watershed scale
analysis, as described in Figure 6.

WATERSHED ACADEMY WEB 8 Watershed Modeling

http://www.epa.gov/watertrain
 A Typical Point Source Modeling Problem:
• Steady State
• Low Flow Conditions

Figure 6. Predicted DO concentration (mg/l) along river

miles

Moving to the watershed scale

Continuing the previous example, now suppose that there are other nonpoint sources of oxygen
demanding waste in this watershed. These include runoff from a number of small and large
animal operations, urban runoff containing leaf litter and other organic wastes (both nonpoint
sources), and combined sewer overflows during storm events (an episodic point source).

What should the modeler consider in evaluating the watershed-scale DO problem? A traditional
approach has been to treat these various nonpoint and episodic point sources as background
against which the wasteload allocation for the point source is evaluated. If the estimate of
“background” is based on detailed and comprehensive monitoring, such an approach can yield an
accurate estimate of the acceptable load from the point sources consistent with meeting water
quality standards. But, if the background sources are based on incomplete monitoring or guesses,
an inaccurate wasteload allocation may result. For instance, precipitation-driven nonpoint
sources like combined sewer overflows only occur during rain events in which case the receiving
water is unlikely to be at 7Q10 low flow. Considering only the background present during
drought conditions may not be the right answer either. Indeed, if the nonpoint sources provide a
sufficiently large load of oxygen demanding waste, the most critical condition, in which the least
assimilative capacity is available to the point source effluent, may occur at flows well above a
7Q10 low flow, in which case a steady-state wasteload allocation may miss the point entirely.

Perhaps more importantly, for effective management, which aims both to protect water quality
and to allow economic development, it is often desirable to look at the watershed-scale big
picture, including management strategies for both point and nonpoint sources. For instance, it

WATERSHED ACADEMY WEB 9 Watershed Modeling

http://www.epa.gov/watertrain
would be unwise to impose expensive new treatment costs on the POTW, resulting in significant
increases in local utility rates, if the same level of water quality protection could be obtained
with less expense through nonpoint source controls.

In sum, it is important to understand how the

nonpoint sources interact with the point
source, and how these jointly affect water
quality. A watershed-scale analysis may thus
need to address both the POTW and the
nonpoint sources simultaneously (Figure 7).

But, undertaking the watershed-scale

analysis of oxygen-demanding waste and
DO may introduce a number of
complications. Some of the issues that may
arise are:

Oxygen-demanding wastes from different

sources may have different characteristics:
Some sources may load highly reactive,
quickly degrading substances, while others
Figure 7. A Typical Watershed Scale Problem may load more slowly reacting refractory
substances. The highly reactive substances
are most likely to cause an impact near the source. Slowly degrading wastes are likely to be
problematic further from the source, or in a place where they reside longer and accumulate (e.g.,
a slow-moving river segment, lake, or estuary). The DO profile in a watershed may be the net
result of the interaction of a variety of types of wastes.

Quantifying nonpoint sources: Most nonpoint sources are precipitation driven, and vary widely
in time. Further, because they are not discharged through a managed discrete conveyance, they
are inherently difficult to monitor, and even rough estimates of loading may be subject to
considerable uncertainty.

“Design” conditions for analysis: When both steady and episodic sources are present, there may
no longer be a clear design condition for analysis. In other words, the maximum watershed-scale
impact may occur at some flow higher than 7Q10 flow. Further, if some of the sources are
episodic it may be very difficult to estimate the actual frequency or risk of excursions of the
water quality standard. One recourse is to go to a fully time-dependent (dynamic) modeling
analysis, but this is likely to require a high level of effort.

So, what will the purpose of modeling be? This will include refining the wasteload allocation for
the point source, but also evaluating trade-offs and control strategies at the watershed scale
which consider both the point and the nonpoint sources. These trade-offs involve deciding who
will bear the burden of the necessary pollution reductions. Do we place the whole burden on
point sources, which can be controlled by enforcement, or do we encourage implementation of
BMPs on farms? Do we require the same BMPs of small and large farms, or place a greater
burden on large operations responsible for more of the loading? What is the most equitable and
cost effective approach to management?

WATERSHED ACADEMY WEB 10 Watershed Modeling

http://www.epa.gov/watertrain
Many potential complicating factors have been mentioned above, which may make the prospects
of watershed scale analysis appear daunting. There are, however, also many ways in which the
analysis may be simplified. It is therefore essential to start by asking how detailed an assessment
is required that is, to develop a modeling strategy. In general, the level of detail that is required
will be closely related to the types of decisions that need to be made, the level of uncertainty that
is acceptable in results, and the financial and ecological implications of decisions.

Watershed Modeling Strategy

Key points of a modeling strategy are best illustrated by example. Let’s now go through some
examples of watershed assessment questions you may be called upon to address, and discuss the
types of things each situation requires us to consider when choosing an appropriate model (or
models). For each example presented in the following pages, a number of common issues will
need to be addressed, including:

• The need for a model to estimate the runoff of water and loading of contaminants from
land areas in the watershed (nonpoint sources).
• How to represent spatial components in the model.
• How to represent time in the model.
• Whether it is necessary to link together more than one model to answer assessment
questions.
• The appropriate level of complexity of the components of the modeling system.

Notice that model complexity is not

independent of the other considerations listed
here. It depends on each of these other
factors: whether a loading model is used, how
time and space are represented, whether more
than one model is needed, and what decisions
need to be made. These factors are illustrated
in Figure 8.

Figure 8. Considerations while choosing a model for

Watershed Assessment

WATERSHED ACADEMY WEB 11 Watershed Modeling

http://www.epa.gov/watertrain
Appropriate model complexity is an important issue, so let’s look at that subject a little further
before going through the examples. (Additional information on this topic will be provided during
and after the examples). The complexity of the model used for watershed assessment will depend
on many factors. First, there are a number of technical needs regarding the complexity of the
assessment questions being addressed, and the inherent complexity of the problem being studied.
Determining whether a water quality standard will be met at the edge of a mixing zone is
inherently simpler than evaluating actual ecological risks. In addition, watershed-wide problems
are usually more complex than localized problems, multiple sources are more complex than
single sources, and nonpoint sources are usually more complex to analyze than point sources.
The characteristics of the contaminant are also important: contaminants which undergo
degradation or phase changes, or which persist and bioaccumulate in the environment often
require a more complex analysis. In addition, appropriate model complexity is also usually
influenced by a number of practical and sociological issues. Often, stakeholder demands about
detail and scientific certainty of model results will play an important part in model selection.
When model results are likely to be used to support costly management actions, the model must
have the detail and certainty sufficient to satisfy the affected stakeholders. Finally, practical
issues such as available funding, staff time, availability of historic data, and user expertise play
an important role in the appropriate level of model complexity.

We now present some specific examples of the use of watershed modeling tools.

Example 1

For the first example,

consider a lake at risk of
eutrophication. For
planning purposes, we
wish to project the effects
of land use changes in the
surrounding watershed on
nutrient loading to the
lake (Figure 9).

Figure 9

WATERSHED ACADEMY WEB 12 Watershed Modeling

http://www.epa.gov/watertrain
Will land use changes result in an increase in nutrient loading? Assume for this case that we are
conducting a scoping study, designed to provi de a rough estimate of possible impacts. The study
does not require a level of rigor sufficient to make immediate regulatory decisions or implement
management strategies based on the model results.

The first step in modeling is to develop a model strategy. As described previously, several
important issues help to determine an appropriate model strategy (Figure 10).

Stressor and systems considered

The map shows the watershed which drains to the lake. Land use types in the watershed are
depicted by different colors. At present, the watershed is predominantly agricultural (shown in
state color), but the town in the watershed is growing rapidly, and agricultural land is being
converted to suburban, commercial, and light industrial uses. Based on information obtained
from the local planning department, estimates of rate of conversion of agricultural and forest
land can be used to predict the future distribution of land uses in the watershed.

Box

Figure 10

WATERSHED ACADEMY WEB 13 Watershed Modeling

http://www.epa.gov/watertrain
Level of study / complexity
Because the purpose of modeling is to provide an initial scoping-level estimate, rather than to
make final management decisions based on model results, it is appropriate to undertake the
initial modeling with a simple, but appropriate model. We know that model results will be
subject to considerable uncertainty, but this is not a great concern for initial scoping. Once we
see the initial results we can then decide how much additional rigor should be introduced.

A detailed, quantitative uncertainty analysis is also not needed at the scoping level. We will,
however, wish to have some information on the quality of predictions in order to interpret results
and make an informed decision about the need for additional modeling. For instance, if the
scoping model indicates that no water quality impacts are expected will we be confident that no
problems will arise, or is there a good chance that the model prediction of no adverse impact is
wrong.

There are two ways to address this issue at the scoping level. On the one hand, we could
structure the model with conservative (worst case) estimates of nutrient loading rates, so that we
can be pretty sure that a prediction of no adverse impact means no adverse impact. On the other
hand, a simple sensitivity analysis could be used to estimate the likely range of model
predictions.

Spatial considerations
The purpose of the scoping application is to predict nutrient loads to the lake. Therefore, it is not
necessary to try to accurately predict nutrient concentrations in each tributary stream, as long as
the general loading pattern is adequately represented. In addition, since this is a scoping level
study, it is not critical that we separate the impact of each individual source of nutrient pollution,
as it might be if the modeling was to be used to impose pollutant limitations on various
stakeholders.

Time considerations
The receiving water of interest is a lake, with a reasonably long residence time, and the intention
of scoping is to predict an approximate average nutrient concentration in the lake in response to
land use changes. Thus, we need not focus on transient changes in nutrient concentrations, but
will be more interested in growing-season or yearly average concentrations. A continuous
simulation model is not needed; a steady-state model should be sufficient.

Questions to be answered
In sum, what is needed for the scoping analysis is a model that can predict loading of nutrients to
the lake from all points in the watershed, on an annual or seasonal basis. This can be compared
with existing rates of nutrient loading.

Based on the analysis of model selection considerations, an annual-average steady-state model of

nutrient loading from the watershed will be sufficient. Watershed loading of nutrients can thus be
described using an export coefficient model. In an export coefficient model, the average
nonpoint loading from the watershed is calculated simply from a coefficient which represents an
amount of load per unit area per unit time - e.g., pounds per acre per year. This unit load
typically varies with land use, and its value is based on past studies, and is thus an empirical
estimate. Total load to the lake under future land use conditions can then be estimated by

WATERSHED ACADEMY WEB 14 Watershed Modeling

http://www.epa.gov/watertrain
summing the export coefficient estimates of nonpoint loading with the total point source loads
(usually established from effluent monitoring data), as shown in Figure 11.

EXAMPLE LOADING ANALYSIS

Source: Vanoni, V.A., ed. 1975.
Sedimentation Engineering
American Society of Civil Engineers,
New York.

Figure 11

Export coefficients are most reliable when determined from studies on similar land uses in the
same general geographic region. Various sources are also available to provide default values of
export coefficients for different land uses.

The export coefficient analysis is very simplistic, but may be accurate enough for the scoping-
level analysis. More subtlety could be added, with little additional effort, by making the unit load
estimates relatively higher in areas closer to the receiving waterbody to account for different
rates of trapping of nutrients in the watershed between source areas and the lake. A common
approach is to divide the watershed into sub-watersheds, and, for each of these, to calculate an
adjustment factor based on sub-watershed size. Large sub-watersheds will tend, on average, to
have more land area farther from the lake than smaller sub-watersheds, and thus the average load
from the larger sub-watersheds should be smaller, all other factors being the same. Because
phosphorus tends to sorb to particulate material, nonpoint loading of phosphorus is usually most
strongly affected by sediment erosion and transport. It is therefore often useful to adjust
estimates of phosphorus loading from a sub-watershed by a sediment delivery ratio, which
relates the amount of sediment delivered at the watershed mouth to the amount of sediment
eroded as a function of watershed area. Commonly-used sediment delivery ratios are shown in
the graph, shown in Figure 12 on the next page.

WATERSHED ACADEMY WEB 15 Watershed Modeling

http://www.epa.gov/watertrain
 Source: Vanoni, V.A., ed.
1975.
Sedimentation Engineering.
American Society of Civil
Engineers, New York.

Figure 12

Example 2

The second example continues the analysis of the impacts of projected land use changes on the
lake begun in Example 1 (Figure 13).

Figure 13

WATERSHED ACADEMY WEB 16 Watershed Modeling

http://www.epa.gov/watertrain
The analysis is still at the scoping level, but one additional layer of complexity is added by
attempting to assess what changes in nutrient loading may imply for in-lake conditions. We have
thus refined the “questions to be answered”, and need to refine the modeling strategy
accordingly.

The motivation for this additional complexity is that nutrients are not in themselves the subject
of our interest. Instead, nutrients are proxies for what we are really interesting in assessing: the
ability of this lake to meet uses such as swimming, fishing, support of aquatic habitat, etc. The
actual objectives of the assessment are often referred to as “assessment endpoints”, while the
proxy measurements are referred to as “indicators”. Indicators are chosen as parameters which
are closely related to assessment endpoints (i.e., nutrients are a cause of eutrophication in the
lake, which can lead to impairment of uses), but are more readily measured and predicted than
the assessment endpoints.

The strength of the linkage between assessment endpoints and indicators, and thus the quality of
our assessment depends on which indicators are chosen. In general, indicators which are more
strongly tied to the assessment endpoints provide a stronger basis for assessment. We can get
closer to the actual assessment endpoints by selecting predicted trophic status as an indicator of
the ability of the lake to meet its existing and designated uses.

Instead of simply predicting nutrient loads, we now want to take the loading information and use
it to predict an in-lake indicator: an index of trophic status. To do this, we will need to link the
loading model described previously with a receiving water model that will predict the effects of
nutrient loading on the trophic status of the lake. As shown in Figure 14, predicted land use
changes lead to new estimates of nutrient loads, provided by the nutrient loading model. These
new nutrient loads are then used to predict trophic status using the lake water quality model.

For example, for a scoping level analysis we might use the Generalized Watershed Loading
Functions (GWLF) model (Haith et al., 1992) to predict annual average nutrient loads from each
subwatershed, then use these predictions as input to the Corps of Engineers BATHTUB model
(Walker, 1987) to estimate (1) nutrient balance within the lake segments, and (2) average
growing season algal response measured as concentration of chlorophyll a. Other, similar
models such as EUTROMOD (Reckhow et al. 1992) are available which combine the loading
model and receiving water response model in a single software package.

Figure 14

WATERSHED ACADEMY WEB 17 Watershed Modeling

http://www.epa.gov/watertrain
References used in this example:

Haith, D.A., R. Mandel, and R.S. Wu. 1992. GWLF, Generalized Watershed Loading Functions,
Version 2.0, User’s Manual. Department of Agricultural & Biological Engineering, Cornell
University, Ithaca, NY.

Reckhow, K.H., S.C. Coffey, M.H. Henning, K. Smith, and R. Banting, 1992.
EUTROMOD:Technical Guidance and Spreadsheet Models for Nutrient Loading and Lake
Eutropication. North American Lake Management Society.

Walker, W.W., Jr. 1987. Empirical Methods for Predicting Eutrophication in Impoundments.
Report 4-Phase III: Applications Manual. U.S. Army Corps of Engineers Technical Report E-81-
9. U.S. Army Waterways Experiment Station, Environmental Laboratory, Vicksburg, MS.

Example 3

In Example 3 we continue the study of the lake described in Examples 1 and 2, but again
increase the level of complexity and sophistication. Suppose the results of the scoping study in
Example 2 told us that the lake was likely to become highly eutrophic under future land use
conditions, with summer chlorophyll a concentrations expected to be frequently in excess of “40
micrograms per liter”. This suggests the need for proactive management. To plan effective
management we need to begin pinpointing the causes of the problem, and specifically the
relative magnitude of concentrations from different sources, so that we can start to figure out
where to target management efforts and source control.

We are now asking more detailed questions of the modeling; this requires a revision of the
modeling strategy. As before, we can structure the discussion in terms of the space, time, and
complexity considerations of the modeling (Figure 15).

Spatial considerations
The primary refinement to the
modeling objectives is the need
to assess specific sources. We
want to be able to assess the
relative contribution of
individual point sources of
nutrients, as well as nonpoint
sources within different areas of
the watershed. For instance,
some areas of the watershed
may have much greater
susceptibility for generating
nutrient loads to the lake (due,
for instance, to proximity or
higher erodibility of soils) and
Figure 15. Model Selection Considerations may thus be targets for best
management practice (BMP)

WATERSHED ACADEMY WEB 18 Watershed Modeling

http://www.epa.gov/watertrain
implementation or zoning restrictions on development. Accordingly, we now need a loading
model with more spatial detail.

Time considerations
The revised modeling objectives can still be addressed in terms of seasonal or annual average
conditions, since eutrophication is a long-term process, which is relatively unaffected by short-
term variability in nutrient loads. We may, however, wish to increase the sophistication of the
modeling to look at average conditions in response to a variety of flow regimes. What is the
range of expected responses under average, low-precipitation, and high-precipitation years?

Level of study/complexity
As we move to tentative source identification in a planning-level study, certainty in model results
becomes more important, because the results may be used to make decisions with economic and
ecological impacts and to target future management efforts. In general, high levels of uncertainty
in modeling results become less acceptable the closer we get to making real management
decisions that affect stakeholders, including wastewater treatment plants which may need to
install more advanced treatment, towns which may need to improve stormwater controls, and
farmers who may be asked to install more BMPs.

This need to reduce uncertainty in model predictions will justify additional model complexity.
Keep in mind that the rule of thumb concerning appropriate model complexity is: Do not add any
more detail than is necessary. The potential need for management decisions makes more detail
necessary but more data, more funds, and more expertise are required for every increment in
model complexity.

One important way to reduce

prediction uncertainty is to make the
model more site-specific. In this
example, we could continue to use the
unit loads if we know for each sub-
watershed the break-down of area
within each land use. However, this
approach is likely to maintain a
significant level of uncertainty or
error, since the unit loads are probably
derived from averages of historical
data from many areas. We can make
the model more site-specific by
shifting from simple export
coefficients to somewhat more detailed
models of runoff processes in urban
and rural areas (Figure 16).
Figure 16
Urban runoff
The dominant characteristic of urban land areas is a high percentage of impervious land cover.
Precipitation runs off directly from impervious surfaces, rather than percolating into the soil, and
can often transport significant pollutant loads (See www.epa.gov/OWOW/NPS/facts/point7.htm).
A more sophisticated, yet still simple, approach to estimating pollutant loads generated by urban

WATERSHED ACADEMY WEB 19 Watershed Modeling

http://www.epa.gov/watertrain
areas is the use of buildup-washoff models, for which there is a well-developed literature (Novotny
and Olem, Water Quality: Prevention, Identification, and Management of Diffuse Pollution). This
approach is based on the observation that almost all runoff from urban areas comes from the paved
or impervious area, and that most of the polluting material carried in runoff accumulates within 1
meter of curbs. Thus, between precipitation events, the model estimates the buildup or amount of
material that accumulates per curb length. In addition to an increase with time, buildup may be
correlated with other variables such as time of year, curb height, street width, traffic speed,
atmospheric deposition rate, traffic emission rate, and frequency of street cleaning. The model then
estimates the amount of washoff of material that occurs in response to a precipitation event.
Washoff is correlated with rainfall intensity, and the amount of available accumulated solids.

Buildup-washoff calculations can be implemented with simple equations, but require additional
data, including estimates of buildup rates and the intensity and inter-event timing of
precipitation. The buildup-washoff formulation is also used within more complex simulation
models, such as EPA’s SWMM (http://ccee.oregonstate.edu/swmm/).

Note that a buildup-washoff formulation is, at least in part, a representation of the actual process
of pollutant load generation, rather than simply an empirical average. Such a process-based
modeling technique is inherently time-variable, because it depends on the time available for
buildup, and the timing and characteristics of storms which drive washoff.

Non-urban runoff
Non-urban sources of runoff and pollution include agriculture
(www.epa.gov/OWOW/NPS/facts/point6.htm), forestry
(www.epa.gov/OWOW/NPS/facts/point8.htm), and other rural land uses. Non-urban areas are
generally characterized by pervious surfaces, into which water may infiltrate. Pollutant loading is
often separated into a dissolved component, which moves with the flow of water, and a
sediment-attached component, which moves with the erosion of sediment. A simple process-
based approach is to calculate the amount of runoff volume and the amount of erosion, then
apply a concentration factor to estimate pollutant loads.

Overland runoff is often estimated with the Natural Resource Conservation Service (NRCS)
Curve Number method (Ogrosky and Mockus, 1964). This method relates runoff volume to
precipitation volume, antecedent soil moisture conditions, and a so-called “curve number” (CN).
CN is taken from tables compiled by NRCS and depends on land use, land cover, and hydrologic
soil group.

Erosion from non-urban pervious areas is often estimated using the Universal Soil Loss
Equation (USLE) or one of its modifications (Wischmeier and Smith, 1978). The USLE includes
factors for rainfall erosivity, soil erodibility, slope length and steepness, and land cover and
management. Depending on the units used for erosivity, the USLE may be expressed on an
annual or event basis. The USLE is designed to predict long-term average rates of soil losses
from fields and other land uses. The rate of soil loss is not, however, the same as the yield of
eroded sediment, as a substantial amount of the eroded soil may be trapped or redeposited before
reaching a water body. Therefore, in watershed models the USLE is usually coupled with an
estimate of fraction of sediment delivery (“delivery ratio”).

WATERSHED ACADEMY WEB 20 Watershed Modeling

http://www.epa.gov/watertrain
In many watersheds, delivery of dissolved nutrients via ground water flux is also significant,
particularly for nitrogen. At the simple process-model level, simple mass balance models of
precipitation infiltration and ground water delivery to streams are often used to account for
ground water loading.

Use of process-based models

In an attempt to reduce uncertainty in model predictions, we have moved to process-based
models, albeit simple ones. These use mathematical relationships to convert data on land use,
land cover, and time series of precipitation into estimates of time series of pollutant loads to a
waterbody. With a process-based model of loading in hand, we can begin to examine specific
management questions, such as the following:

• What effects can be expected from increasing development and impervious area?
• How much can increased street sweeping reduce loading from urban areas?
• What is the effect of various agricultural and erosion control practices on sediment and
nutrient loads?
• What areas of the watershed generate the highest loads of pollutants?

It should be cautioned that even the most sophisticated process-based models of nutrient loading
contain many empirical parameters, which cannot be measured directly. To increase confidence
in model predictions, it will be necessary to go through a process of model calibration, in which
model parameters are adjusted to provide a better fit to observations.

Example 4

The previous examples considered the case of nutrient enrichment of a lake, for which it was
appropriate to think in terms of long-term average conditions. There are many other cases in
which it is not sufficient to address only long-term averages. A typical problem in which we are
concerned with short-term
impacts is the case of
violations of a dissolved
oxygen (DO) standard in a
river segment (Figure 17).
The state has specified a
daily average DO standard of
5 mg/l, and an instantaneous
value of 4 mg/l for the
protection of aquatic life.
This water quality standard is
being violated due to a
combination of sources, and
the state wishes to develop a
management strategy to bring
the river segment back into
compliance.
Figure 17

WATERSHED ACADEMY WEB 21 Watershed Modeling

http://www.epa.gov/watertrain
As in the previous examples, model selection and model strategy are determined by a number of
considerations, including stressors, and required treatment of time, space, and the level of
study/complexity (Figure 18).

Stressors
Depletion of DO in the river segment results from the interaction of a number of different
sources of oxygen demand, including both point sources and diffuse watershed sources (Figure
18). These sources include:

1. A wastewater treatment plant, which is a permitted point source discharging a relatively

steady load of oxygen-demanding waste;
2. Urban runoff routed through a separate storm sewer system, which is also a permitted
point source, but discharges a highly episodic load of oxygen demanding waste;
3. Runoff from agricultural crop farming and animal operations, which constitute a diffuse
nonpoint source of oxygen demanding waste; and
4. Sediment oxygen demand, exerted by organic material stored in the river sediment.

The stressors have a variety of different spatial and temporal characteristics, which will affect the
choice of a model strategy.

Figure 18

WATERSHED ACADEMY WEB 22 Watershed Modeling

http://www.epa.gov/watertrain
Time considerations
The objective of management is to prevent excursions of the DO water quality standard. This
should be protected at all times, except under extraordinary conditions. Equivalently, the
objective could be stated as maintaining a frequency of excursions of the standard that is below a
certain acceptable low frequency, such as once in three years.

Not all the sources are constant in time, and we are concerned that the standard be met at almost
all times. For these reasons, long-term average predictions of DO concentrations in the river
segment do not provide us what we need to know. Instead, we need a model that will capture
transient depression of DO concentrations in the river segment.

There are two ways in which we could try to capture the time-varying nature of impacts. The
direct approach would be to implement a continuous model with adequate temporal resolution to
predict the actual time-series of DO concentrations. This would typically require a substantial
effort. The alternative is to apply a “worst-case” type of model that predicts only what happens at
critical conditions, which are those conditions at which the greatest impact is expected.

In modeling the impacts of point sources a worst-case approach is typically used. This consists of
assuming a minimal instream dilution capacity or design flow and applying a steady-state water
quality model such as QUAL2E - which is much simpler, and less expensive to implement than a
continuous model (See www.epa.gov/ORD/WebPubs/QUAL2E/). By choosing a conservative
design flow and other conservative design conditions (such as the high end of the expected water
temperature range, which increases oxygen demand), a wasteload allocation can be assigned to
the point source, which is protective of the waterbody under most conditions. Typically, the
design flow is assumed to be the 7Q10 flow which is the 7-day average low flow which recurs,
on average, once every 10 years. Note that this still potentially allows occasional excursions of
the DO standard, during those time periods when the instream flow is less than the 7Q10 flow.

Thus, a relatively simple method is available for the analysis of point source impacts. The
addition of episodic and nonpoint sources of load (typical in watershed assessment) complicates
this analysis. For these sources, lowest dilution flows and highest source loads often do not
coincide, particularly when there are significant precipitation-driven sources. In these cases, the
“worst case” may be at some flow higher than the 7Q10 low flow.

The simplest and most conservative approach to the watershed-scale analysis in this example
would be to apply a steady-state model at the design low flow, as was done for the point source,
and assume a worst-case loading from the storm sewers, agricultural areas, and other
precipitation-driven sources. This would certainly be protective of water quality, but is likely to
be unrealistic, since maximum loading from the runoff would normally be associated with higher
than 7Q10 flows in the river. Indeed, such an ultra-conservative approach is likely to suggest that
there is no assimilative capacity available for the point source, even when observations indicate
that this is clearly not the case.

To make the modeling assessment more realistic, while continuing to use a simple, conservative
approach we could run two steady-state model applications, intended to represent the range of
expected impacts in the receiving water segment. The first application would be intended to
reflect the impact of the point source at drought condition flows, with no contribution from
episodic precipitation-driven sources. The second model application would be designed to reflect

WATERSHED ACADEMY WEB 23 Watershed Modeling

http://www.epa.gov/watertrain
the impact of nonpoint loading during a large runoff event, plus the point source. To make this
application conservative (but not completely unrealistic) we would combine the large runoff
loads with an unusually low dilution flow for such an event in the receiving water. One way to
do this is to analyze historical records of flows during large rain events, and choose the lowest
instream flow (or once-in-10-years recurrence flow) observed in association with large rain
events.

Alternatively, the results can be made more realistic by running a continuous simulation model
that will provide a full representation of the correlation of runoff events and instream flows.
Continuous modeling allows us to assess the interaction of point sources and nonpoint sources
over all flow conditions, including the period following a runoff loading event when flow drops
off, but concentrations may be high. The main drawback of using a continuous modeling analysis
is the additional data, time, and user expertise required to implement it.

The choice of which method to use will depend on the level of detail required to answer the
question at hand, and the resources available for the assessment.

Model linkage considerations

Developing a watershed modeling tool for this example will require linking several different
types of models together. Or, we may be able to use an existing software package that has all
these parts already linked together (i.e., it includes sub-models for each of the components) (see
Figure 19 on the next page). The three main types of models which must be linked are:

1. A model for flow and pollutant loading in urban runoff;.

2. A model for flow and pollutant loading in rural runoff; and
3. A model of the receiving water body, which calculates the effects of oxygen demanding,
loads on DO in the river segment.

Spatial detail
As with the previous example of the lake, the amount of spatial detail we need to include
depends largely on how well we wish to be able to separate out the effects of different sources.
This is less important at the scoping level, when one perhaps only wishes to assess the
approximate severity of a problem, but more important when results of the modeling study will
be used to evaluate management strategies. For instance, we may wish to consider the potential
effectiveness and trade-offs between reducing overall input of BOD via requiring a higher level
of treatment at the wastewater treatment plant, implementing street cleaning to reduce urban
stormwater loading, requiring the town to build infiltration basins to reduce the total flow and
loading from urban areas, or encouraging more use of best management practices in agricultural
areas. A model is a useful tool for comparing the effectiveness of various possible control
options. However, to make the comparison the model must include sufficient complexity to
assess the relative impacts, and interactions of individual sources.

WATERSHED ACADEMY WEB 24 Watershed Modeling

http://www.epa.gov/watertrain
 Figure 19 Model Components

Example 5

We will now move on to a more complex example in which the joint impacts of many sources
must be considered. This example addresses a large (7,000 square mile) watershed which
contains many point sources, including wastewater treatment plants and industrial discharges.
There are also urban and rural sources of nonpoint load, including combined sewer overflows
(CSOs), separate storm sewer systems, septic systems, and runoff and groundwater flux from
agricultural fields and animal feed lots. In the estuary at the downstream end of this watershed
massive fish kills are occurring as a result of a complex interaction of natural tidal events with
eutrophication caused by nutrient inputs. The nutrient loads promote excessive algal growth,
including some toxic species. In addition, algal die off, particularly at the salt-freshwater
interface, exerts a large oxygen demand and results in intermittent depressed dissolved oxygen
levels. Degraded conditions in the estuary are attributed primarily to excess nitrogen loads,
which derive from the entire watershed.

WATERSHED ACADEMY WEB 25 Watershed Modeling

http://www.epa.gov/watertrain
 Objective

Develop a Photo: USDA

strategy to
significantly reduce
nutrient inputs to
a river in a large
watershed with
many sources and
diverse stakeholders.

Figure 20. Controlling Eutrophication in a Large Watershed

The state wishes to reduce nutrient inputs to the estuary, but must decide by how much, and how
to do it (Figure 20). A key question is who bears the burden of additional pollution controls?
There are many stakeholders in this watershed who would be affected by any management
decisions. The issue is contentious: agricultural interests in the lower watershed claim that
problems are due primarily to increased wastewater flows from rapid growth in upstream urban
areas, and they should not have to bear the costs of management. The urban areas blame
agriculture and contend that loads from their plants are significantly attenuated before reaching
the estuary, and it would be unfair to restrict their ability to grow. Landowners in rural areas are
afraid that nutrient restrictions will preclude them from developing their property. Meanwhile,
fishermen and the tourist industry in the estuarine area are losing money and demanding
immediate action.

In such a contentious atmosphere, with major financial interests at stake, any modeling
prediction which appears to favor one stakeholder group’s point of view is likely to be
challenged by other groups in the watershed, who may claim that the results of the model are too
uncertain to justify the restrictive measures the state wishes to impose on them. Because of these

WATERSHED ACADEMY WEB 26 Watershed Modeling

http://www.epa.gov/watertrain
demands, and the need for defensibility in results, it is judged that a high level of complexity is
warranted in the modeling application.

This case demonstrates that the stakeholders or affected parties will often drive the nature and
complexity of a watershed modeling study. Even though it may be possible to reach a correct
management decision using a simpler scoping model this will not provide a high enough degree
of certainty to justify politically difficult management decisions. A high degree of model
complexity may be required to assure all stakeholders that their interests are being addressed.

Model selection considerations for this case may be summarized as follows:

Stressors
The stressor of interest is nitrogen loading. Nitrogen is linked to support for uses through
eutrophication in the estuary. A means of predicting algal response to nitrogen loads is also
required; however, it may be advisable to handle this component separately by first determining
an acceptable target nitrogen load at the estuary head, then using the watershed model to assess
attainment of the target load.

Time considerations
There is a fairly long lag time in estuarine response, and estimates of loading to the estuary at the
time scale of weeks would be sufficient. However, some important watershed sources are
precipitation driven, and contention over the importance of these sources may demand use of a
continuous (daily time step) simulation model driven by precipitation data. The model will,
however, need to be run over a long time series of precipitation data (decades) to derive a
reasonable estimate of the efficacy of control strategies under the expected range of meteorologic
and hydrologic conditions.

Spatial considerations
Both types of land use and the location of these land uses within the watershed are at issue in
evaluating management options. Therefore, good spatial resolution of land uses is required, using
a GIS-based approach. Evaluating loss or attenuation of nitrogen loads during transport in stream
will require use of an appropriate river model. Finally, ground water can be an important
pathway for transport of nitrogen, so the model must also be able to address surface-ground
water interactions.

The model strategy for this case requires linking of a GIS-based continuous model of watershed
nitrogen loading with a continuous model of river transport. This is accomplished using the EPA
model BASINS (See: www.epa.gov/OST/BASINS). BASINS combines spatially distributed
land use data, EPA’s Reach File 1, and the HSPF simulation model for watersheds and rivers
into a convenient package with an ArcView GIS interface (see Figure 21 on the next page). Only
by using the power of a GIS interface is it convenient to undertake a continuous simulation
model of such a large watershed. (Note that the ability to address routing of stream flows and
transformations of pollutant loads during stream transport is included in BASINS Version 2,
currently undergoing testing and scheduled for full release in early 1998).

WATERSHED ACADEMY WEB 27 Watershed Modeling

http://www.epa.gov/watertrain
 Figure 21

Model Calibration, Validation, and Verification

Most environmental models include parameters which must be tuned or adjusted to obtain
reasonable match between model predictions and observed conditions (see Figures 22 – 24 on
the following pages). All models require checking and testing to evaluate how well they perform.
The first activity is referred to as model calibration, and the latter as model validation. Without
calibration and validation, a modeling application is only an educated guess. This may be
sufficient for some scoping applications, but generally not for management decisions.

WATERSHED ACADEMY WEB 28 Watershed Modeling

http://www.epa.gov/watertrain
Model calibration and validation
can be defined as follows:

Model calibration involves

minimization of the deviation
between measured field
conditions and model output by
adjusting parameters of the
model. Data required for this
step are a set of known input
values along with corresponding
field observation results.
Calibration typically includes a
sensitivity analysis, which
provides information as to
which parameters have the Comparison of Model Hydrographic (Using Interception–
greatest effect on output. Careful Infiltration Sub-Models) with the Observed Hydrograph
consideration should be given to
adjustment of model parameters Figure 22
to ensure that values are within
the range of reasonable possibility. EPA provides guidance on reasonable values of
parameters for water quality models (www.epa.gov/ORD/
WebPubs/surfaceH2O/surface.html).

Model validation involves the use of a second, independent set of information to check the model
calibration. The data used for validation should consist of field measurements of the same type as
the data output from the model. Specific features such as mean values, variability, extreme
values, or all predicted values
may be of interest to the modeler
and require testing.

Model validation is sometimes

referred to as verification. Under
current usage, this terminology is
discouraged. Instead, model
verification is used to refer to
another process, the examination
of the numerical technique and
computer code to ascertain that it
truly represents the conceptual
model and that there are no
inherent numerical problems with
obtaining a solution.

Figure 23

WATERSHED ACADEMY WEB 29 Watershed Modeling

http://www.epa.gov/watertrain
A variety of statistical tests are available for assessing model goodness of fit during calibration
and validation. A useful introductory summary is provided in: Reckhow, K.H., J.T. Clements,
and R.C. Dodd. 1990. Statistical evaluation of mechanistic water quality models. J. Environ.
Eng., 116(2): 250-268.

How Do We Deal With the Added Complexity?

Watershed assessment may
require additional complexity
beyond traditional modeling
exercises for the evaluation of
point source impacts. Looking
back over our examples, we can
see that modeling at the
watershed level can add
complexity for several reasons:

• Involvement of diverse
stakeholders who demand
a certain level of
modeling detail and
certainty
• The need to consider
nonpoint source loading
of pollutants from land
areas
• The need to consider Figure 24
multiple sources of
stressors
• Sometimes (though not necessarily), watershed modeling requires representation of
greater spatial and temporal detail than a point source problem. Often there will be
interaction between point sources and episodic nonpoint sources which may require
evaluation of temporal patterns. In addition, evaluating management options may require
separating out the effects of many different sources.

How can limited resources be stretched to handle the additional demands of watershed
modeling? While watershed management can add levels of complexity to the jobs of assessment
and modeling, it also provides a number of new opportunities for partnerships with other
agencies and groups, allowing the state to deal with water quality problems more effectively (see
Figure 25 on the next page). You can be more cost-effective by drawing on existing expertise in
other agencies and organizations (e.g., using agricultural agency expertise on runoff processes).
You can also be more effective by leveraging resources through monitoring consortia or
watershed associations. A watershed model may be more complex than a point source model, but
may be useful for the evaluation of multiple sources within a watershed. These arrangements
allow you to reduce duplication, be more strategic with monitoring plans, and leverage available
lab resources. In many cases, you may also be able to pool public and private funds to conduct

WATERSHED ACADEMY WEB 30 Watershed Modeling

http://www.epa.gov/watertrain
joint analyses. This type of jointly funded study has the added benefit that when you get buy-in
at the beginning, the stakeholders are more likely to accept the results.

Overarching Principles
This module concludes by reviewing
some of the important points discussed
above in the form of some overarching
principles to follow when choosing
appropriate models for watershed
assessment.

1. Have a clear statement of project

objectives and verify the need for
modeling. (Figure 26)

Ask yourself: How can a model help

address the questions and problems Figure 25
relevant to decisions? How can a model be used to link stressors or management actions to
quantitative measures (endpoints) of waterbody condition? Is modeling appropriate for
examination of the stressors of concern in this situation?

Some of the types of problems for which

modeling could be useful include:

• Quantifying loads from nonpoint sources

• Estimating load reductions needed to
meet a water quality standard
• Establishing a total maximum daily load
(TMDL)
• Distinguishing between levels of
effectiveness of different management
strategies
• Determining if management criteria can
be met by a proposed strategy.

Remember that, where practically and

economically feasible, real data are always
Figure 26 preferable to model predictions as a basis for
decisions.

WATERSHED ACADEMY WEB 31 Watershed Modeling

http://www.epa.gov/watertrain
2. Use the simplest model(s) that will satisfy the project objectives. (Figure 27)

Adding complexity to the analysis means that

more time, money, expertise, and data will be
required. A common mistake is to assume that
more complex models automatically yield
greater accuracy. This is often not true. More
complex model formulations will often contain
a large number of unobservable parameters, and
may make it easier to obtain a spurious match
between model predictions and observations.
Complexity often gives a false sense of security
that is not justified by careful examination of
sources of model uncertainty. Therefore, it is a
good rule not to include any more temporal and
spatial detail than is necessary to address the
problem at hand.

Figure 27

3. Define carefully the parameter(s) of interest. (Figure 28)

Before beginning a modeling study, it is

important to define carefully which
parameter(s) are to be predicted. It is also
advisable to specify what an acceptable level
of accuracy will be in model predictions
before starting the calibration process.

Modeled parameters should be selected to

coincide with an appropriate endpoint for the
analysis which can serve as an indicator of
water quality prediction. This ties in closely
with model complexity. In the examples
presented above for management of a lake, we
saw that we might choose to predict nutrient
loads, nutrient concentrations, lake trophic
Figure 28 status, or biological effects resulting from
lake trophic status. Choosing to model effects
on the waterbody or biota will add to the complexity of the modeling effort, but may be
worthwhile in order to produce results that are more useful for making informed management
choices.

WATERSHED ACADEMY WEB 32 Watershed Modeling

http://www.epa.gov/watertrain
It is also important to be sure that predictions will be available in appropriate units (e.g.,
dissolved versus particle-attached concentrations, daily versus annual average concentrations) for
use in decision making. For example, for a eutrophication analysis it may be sufficient to
predict total phosphorus concentrations on a seasonal average basis, while a wasteload allocation
to prevent metal toxicity may require predicting the dissolved-phase concentration of the metal
on a continuous basis.

4. To the extent possible, use a prediction method consistent with available data. (Figure 29)

Data availability should be evaluated

before beginning the model selection
process. Models selected for analysis
should have input requirements which
match up well with data which is already
available, or which can be collected in the
course of the study.

Figure 29

5. It will often be necessary to use multiple models or link models to address watershed
assessment problems. (Figure 30)

As we have seen in the examples, it is often

necessary to use one model to predict loading
to a waterbody from nonpoint sources, and a
second model to predict fate and transport of
pollutants in the waterbody, and possibly a
third model to predict effects of the pollutant
on biota or communities. It is important to
choose models with compatible data
input/output. For instance, if the fate and
transport model requires daily load values,
then the loading model should provide daily
output of the appropriate parameters.

Figure 30

WATERSHED ACADEMY WEB 33 Watershed Modeling

http://www.epa.gov/watertrain
Components of the Strategy
Developing an effective modeling tool requires some up-front development of an appropriate
modeling strategy. The modeling strategy represents each of the over-arching principles (Figure
31), and should address the following items:

1. Types of stressors to be addressed

2. Spatial scale and resolution of the
application.
3. Time scale and resolution of the
application.
4. Level of complexity and detail
required.
5. Linkage between model output and
management/decision-making
needs.

Figure 31

WATERSHED ACADEMY WEB 34 Watershed Modeling

http://www.epa.gov/watertrain
Glossary of Underlined Terms
Pervious & Impervious Surfaces:
Pervious surfaces allow water to penetrate or infiltrate into the underlying soil or rock.
Impervious surfaces do not allow water to pass through. For instance, tiled soil is highly
pervious, while asphalt is impervious.

Eutrophication:
The process of physical, chemical, and biological changes associated with nutrient and organic
matter enrichment of a water body. Most frequently used to refer to nuisance growth of algae or
other aquatic plants associated with nutrient enrichment.

Chlorophyll a:
A green pigment used by algae and green plants during the process of photosynthesis to convert
light, carbon dioxide, and water to sugar. Chlorophyll a concentration is often used as an
approximate index of algal biomass.

Trophic Status:
A measure of the degree of eutrophication of a water body.

BMPs:
Best Management Practices

7Q10 Flow:
The 7-day average low flow which recurs, on average, once every 10 years.

POTW:
Publically Owned Treatment Works

Erosivity:
The susceptibility of a surface to be eroded or worn away.

WATERSHED ACADEMY WEB 35 Watershed Modeling

http://www.epa.gov/watertrain