METHODOLOGY FOR CHOOSING AMONG ALTERNATIVES




     TO REDUCE POLLUTANT CONTRIBUTIONS FROM WATERSHEDS
                            by




                    William C. Sonzogni




                     Thomas M. Heidtke




                    Timothy J. Monteith









            Great Lakes Basin Commission Staff
Prepared for U.S. Environmental Protection Agency, Region V




       (Under Interagency Agreement EPA-29-D-F0857)
                      November, 1979

-------
                                TABLE  OF CONTENTS
OVERVIEW MODELING PROCESS  - WHAT  IS  IT?









SYNOPSIS OF MAJOR STEPS  IN THE  PROCESS









INFORMATION REQUIREMENTS




         Point Sources




         Rural Runoff




         Urban Runoff









ACCOUNTING FOR POLLUTANT INPUTS









CALIBRATION




         Transmission




         Flow Variation









REMEDIAL PROGRAMS









COST-EFFECTIVENESS




         Biological  Availability









LONG-TERM VALUE OF MODELING PROCESS









REFERENCES CITED

-------
OVERVIEW MODELING PROCESS — WHAT IS IT ? -
En the Pollution from Land Use Activities Reference Group (PLUARG)
study, a U.S.—Canada cooperative investigation of nonpoint source pollution of
the Great Lakes, a process called “overview modeling” was used to identify the
most cost—effective mix of point and nonpoint controls in Great Lakes basin
watersheds. This process provided detailed estimates of pollutant inputs from
each source in each of the drainage areas or streams discharging to the lakes.
Because of the large number of computations sometimes involved in arriving at
these estimates, a computer program was used to simplify the procedure.
The overview modeling process takes into account factors such as
varying land use (farmland, forest, wetlands, etc.) and land form (soil
texture, slope, etc.), as well as different types of sewer systems (combined,
separate, or unsewered). More importantly, the modeling process is dynamic,
enabling one to take into account changing conditions such as population
growth, urbanization of rural areas, or the natural removal of pollutants from
the water as it moves downstream to the lake.
The real value of the process becomes clear when pollutant control
information is introduced into the computations. In this stage, alternatives
for reducing phosphorus inputs from each source are tested, and those measures
which produce the greatest pollutant reduction at least cost can be easily
identified. The overview modeling process is also very adaptable, and can be
used in watersheds outside the Great Lakes basin.
The following discussion is based on a number of reports dealing with
3

-------
Overview modeling (Johnson et al. , 1978; Johnson et al. , 1979; Heidtke, 1978;
Sonzogni et al. , 1980; Heidtke et al. , 1979a, 1979b). These reports should be
consulted for details on the process or for examples of specific applications.
SYNOPSIS OF MAJOR STEPS IN OVERVIEW MODELING PROCESS
A river basin or watershed is divided into sub—watershed units as
shown in the geographic schema in Figure 1. Point and nonpoint sources are
then identified and their respective pollutant inputs are estimated. An
accounting system (the accounting system could range from a simple to a
complex mathematical algorithm) is then used to route the inputs downstream to
the receiving water as shown in the model schema in Figure 1. Transmission
losses, which may occur, for example, due to a reservoir as shown in Figure 1,
are accounted for through the application of “transmission coefficients” in
various stretches of the tributary. Once the information base is established,
the effect of remedial measures at different points in the system can be
compared in terms of the cost of the remedial measure per unit reduction in
the pollutant input at the receiving water. This basic “accounting system” is
readily adaptable to large or small watersheds and can be as general or
detailed as the user desires.
INFORMATION REQUIREMENTS
The information base is probably the most critical requirement of the
process. The temporal and spatial detail of the input data will largely
determine the degree of sophistication of the work. The information
requirements center on the location and extent of pollution sources, namely
4

-------
municipal and industrial point sources, rural nonpoint sources and urban
nonpoint sources. Other factors, such as demographic influences and land use
changes, provide the basis for long—term assessment. Fortunately, much of the
information needed in the process can be obtained from local areawide water
quality planning studies (208 studies) now being conducted across the U.S.
Point Sources
Pollutant loads to a receiving water from a municipal treatment plant
are estimated from three items of information: (1) sewered population, (2)
the per capita input of a pollutant, and (3) the treatment efficiency of the
plant. For some pollutants, such as phosphorus, the amount of pollutant
discharged may be readily available from direct measurements.
In applying the overview modeling process to phosphorus inputs to the
Great Lakes, the following expression was used to estimate pollutant loads
from treatment plants:
w i’ x pci x (l—T)
w
(1.—T) = P x pci
w
T = P x pci
5

-------
where W = estimate of pollutant load from the treatment plant (kglyr)
(often obtained from 208 data),
P = estimate of population served by the plant,
pci = annual per capita input of pollutant to the plant
(kg/person/yr) (can be estimated from the literature or
derived from available data),
T = percent of incoming load removed by treatment.
As explained in Johnson et al . (1978) and Heidtke et al . (1979b), 1.5
kg of total phosphorus per year was assumed as a representative per capita
input. Sewered populations can usually be drawn from the 208 information base
or similar data sources, and treatment efficiencies can be calculated to yield
a load from each facility consistent with any assumed treatment scheme. For
example, when simulating loads to the lake under the assumption that treatment
plants are achieving a fixed total phøsphorus effluent concentration (C 0 ),
the appropriate treatment level (T) can be computed using the following
equations:
w. = p pci
in
W C
out out
w. -w
in out
w.
in
6

-------
where P = sewered population
pci = per capita input of pollutant (kg/year)
W r = pollutant load entering treatment plant
C 0 = pollutant concentration in treatment plant effluent
Q = average wastewater flow
W 0 = pollutant load Leaving treatment plant
T = fraction of incoming load removed by treatment
Alternately, if the wastewater flow from a municipal treatment plant
is known (which is almost always the case), the pollutant load can be
calculated when the mean concentration is known or estimated. Without
measurements, the pollutant concentration in the effluent can often be
estimated from information on the type of treatment, industries served and
general chemical characteristics of the water supply.
Rural Runoff
Contributions of pollutants from a given sub—basin (as depicted in
Figure 1) can be determined by a number of means. For some pollutants the
universal soil loss equation (USLE) can be used to estimate a load under
certain precipitation conditions. The U.S. EPA Nonpoint Source Model (Cahill
et aL. , 1979), which is a variation of the USLE approach, can also be used.
In some sub—basins monitoring data (for example, USGS water quality data) may
be available so that the contribution is defined. If only flow data is
available for a sub—basin, it may be possible to estimate loads by
7

-------
extrapolating from other monitored watersheds (while this type of estimate was
not possible a few years ago, the extensive studies of nonpoint pollution have
increased the state of knowledge to the stage that reasonable estimates can
often be made). Finally, a procedure called the modified unit area load
approach can be used.
In the overview modeling studies to date, pollutant loads
attributable to land drainage were generally estimated using a modified unit
area load (UAL) concept. The annual diffuse pollutant load generated from a
given hydrologic area is estimated based on two characteristics of the area——
land use (urban, agriculture, forest, etc.) and land form (soil texture and
topography).
Explaining the modified unit area load concept further, Table 1 shows
that, while unit area loads of phosphorus for a particular land use may vary
by an order of magnitude or more, knowledge of certain characteristics of the
watershed permit a more refined estimate of a representative value. For
example, certain combinations of factors (such as row crops grown on a soil of
high clay content) produce a high unit area load of phosphorus. Statistical
studies done on some of the rural watersheds in the Great Lakes basin show
that close to 90 percent of the variability in measured unit area loads of
total phosphorus between watersheds are accounted for by differences in soil
texture and the percentage of the area in row crops. Aside from phosphorus,
the modified unit area load approach has been used to estimate suspended
solids and heavy metals loads.
Despite the use of the modified unit area load approach in past
8

-------
applications of the overview modeling approach, other possibilities exist for
estimating rural runoff inputs. The overview modeling approach is flexible;
the usefulness is not affected by the technique used to generate a load from a
sub—basin.
Urban Runoff
Runoff contributions from urban areas can also be estimated by
various means. In some sub—basins actual measurements will exist. Recent
urban runoff studies by U.S. EPA (Nationwide Urban Runoff Program) could
provide some techniques for estimating sub—basin contributions.
As discussed under “Rural Runoff”, pollutant loadings from urban
runoff were estimated in previous overview modeling work using a unit area
load approach. Urban phosphorus unit area loads are given in Table 2. This
table was developed from an extensive review of the literature (including
several PLUARG studies). As shown, loads are a function of the degree of
urbanization. Some unit area loads for urban areas are significantly higher
than those for rural areas. The unit area load for urbanizing land
(construction sites) is particularly high. If construction occurred on sandy
soil, the phosphorus unit area load would likely be less than that given in
Table 2. Most large urban areas in the Great Lakes basin are located on clay
plains, however.
The effect of urban expansion can also be considered. Past overview
modeling work was arranged so that rural sub—watersheds could gradually
decrease in area to accommodate urban expansion. The effect on future
9

-------
phosphorus inputs due to projected urban expansion has in fact been estimated
for the entire U.S. Great Lakes basin (Heidtke et al. , 1979a). These
estimates were derived largely from 208 planning data.
ACCOUNTING SYSTEM
In order to simplify accounting for all the inputs from a large
watershed, a computer program can be used. However, for some applications
(such as to a small watershed) it is not necessary to use a computer, as
computations can readily be made by hand. The larger and more complex a
watershed, the more useful a computer program becomes.
In the previous overview modeling work, a specialized computer
language——APL (A Programming Language)——was used (see appendix in Johnson et
al., 1978, for further description of the APL program). The algorithm
developed uses a cascading system approach to represent a drainage basin and
involves a three step process:
(1) a mathematical description of a river drainage basin is
developed using a sub—watershed infrastructure,
(2) unit area load tables and other point source information sources
are used to estimate pollutant loads,
and (3) effects of remedial measures are estimated through manipulations
and iterations in various program functions.
10

-------
These steps have been described earlier.
Despite the utility of APL for this type of work, it is not a
commonly used computer language. However, a program could be written in a
more familiar language, such as Fortran, which would satisfy user needs.
Further, a program or technique that hand held or programmable calculators
would utilize could likely be developed that would be satisfactory for
uncomplex applications. Consequently, sophisticated computer use need not
limit the process.
CALIBRATION
Whenever possible, the pollutant load from a watershed to a receiving
point should be compared to actual measurements or monitoring data. In the
Great Lakes area, extensive scream or river monitoring has been conducted by
the U.S. Geological Survey. Monitoring stations exist at or near river mouths
for most major tributaries. This data can be used to adjust or calibrate the
model.
Transmission
One way in which the model can be “calibrated” is by adjusting the
transmission factor associated with each pollutant input (see Figure 1). The
transmission factor indicates the fraction of the pollutant load which may be
Lost in transit to the lake. Upstream sources, or sources above an
impoundment or lake—like widening of the river, would be expected to exhibit
greater transmission losses than downstream sources.
11

-------
The results for the Great Lakes indicate that for total phosphorus
transmission losses are small. That is, the estimated Load calculated with
the model assuming 100 percent transmission were generally in agreement with
measured loads. Only those tributaries with large impoundments required
adjustments for transmission losses. As explained in Sonzogni et al. (1979),
very little empirical data actually exists on transmission losses. When
available, it of course could be used in the process of model calibration.
One area of the Creat Lakes basin where some transmission loss occurs
is the eastern basin of Lake Michigan. Previous analysis of Lake Michigan
phosphorus loads revealed that transmission losses over the entire lake may
amount to 15—20 percent of the total phosphorus load to tributary receiving
waters. Much of this loss occurs because the eastern portion of the drainage
basin contains a number of large inland lakes. Also, several large municipal
point sources (Lansing, Jackson, Kalamazoo) are located in the upstream
portion of the watersheds, resulting in greater transport distances and,
therefore, greater opportunity for entrapment. Additionally, many of eastern
Lake Michigan tributaries have lake—like widenings at their mouths which
probably reduce the delivery of phosphorus.
Flow Variation
In order to account for different flow levels and thus adjust
nonpoint source inputs to average conditions, model estimates should be
calibrated with “average” loads (over a long historical record, where
possible), rather than the load for any particular year.
12

-------
Rather dramatic changes in annual river flow can occur from one year
to the next. These flow fluctuations, resulting from differences in the
amount, intensity, and time of occurrence of precipitation, can greatly affect
loads of land—derived pollutants. Using any one year’s results for
calibration purposes could be misleading if that year was a high or low flow
year.
Sonzogni et al . (1979) have described a technique to adjust loads for
any one year to “average” conditions. The technique is based on a historical
average flow computed from gaging station records. Pollutant loads are then
adjusted to average conditions in proportion to flow. The technique thus
assumes that load is proportional to flow. Such a technique is most
applicable on a gross scale and may not be appropriate for an individual small
watershed. However, simply considering the range in annual flows can provide
a qualitative appreciation of the variability expected from diffuse inputs due
to meteorological (and runoff) conditions. Such qualitative appreciation may
often be sufficient to enable decisions to be made for watershed management.
Information on flow variations is also used in many methods to
estimate nonpoint source inputs. For example, the unit area loads described
previously are designed to represent average conditions. The universal soil
loss equation, of course, considers meteorological and climatic variables
which affect runoff, and can be adjusted to reflect average conditions (if the
data are available). The U.S. EPA Nonpoinc Source Model (Cahill, 1979) can
also be used to estimate nonpoint source inputs from a sub—basin for an
average year, but again the availability of necessary data is often a
constraint. Overall, however, a number of techniques exist to normalize
13

-------
resuLts to an “average year” using the overview modeling process.
REMEDIAL PROGRAMS
The real value of the overview model process becomes clear when
pollutant control information is introduced into the computations. En this
stage, alternatives for reducing pollutant inputs from each source are tested,
and those measures which produce the greatest reduction at least cost can be
easily identified. The remedial measures can be applied to point or nonpoint
sources and can consist of any number or type of remedial measures. Again,
the modeling process is flexible.
In previous overview modeling work a number of alternative plans were
proposed for reducing phosphorus loads to the Great Lakes. Two alternatives
considered were the limitation of total phosphorus in the effluents from
sewage treatment plants to either 0.5 mg/L or 0.3 mgIL (down from the current
1.0 mg/L requirement). Programs for reducing loads from urban runoff ranged
from simple streetsweeping to capturing and treating stormwater runoff and
combined sewer overflows.
For agricultural runoff, three increasingly expensive levels of
phosphorus control were considered. These programs vary from voluntary sound
soil conservation or “good stewardship” on all agricultural land to the
application of more intensive “best management practices” in areas producing
characteristically high phosphorus inputs to the lakes, for example, row crop
fields on fine—textured soils.
Examples of sound soil conservation or “good stewardship” include
properly incorporating fertilizers and manure into the soil, avoiding the
14

-------
addition of excessive amounts of inorganic fertilizers, and avoiding farming
on slopes near streams. Best management practices, on the other hand, include
various farming techniques such as minimum tillage, winter cover crops, and
stripcropping.
Recent information from the U.S. Army Corps of Engineers’ Lake Erie
Wastewater Management Study indicates new no—till measures in parts of the
Erie basin can actually be beneficial to the farmer (and, thus, voluntarily
implemented) and still result in substantial nonpoint phosphorus reductions.
In a given rural watershed or sub—basin, a particular set of actions
may have been recon uended from previous study (e.g., by a 208 agency in
conjunction with the Soil Conservation Service). The effect of these measures
on overall watershed pollutant dynamics can be determined if the pollutant
reduction from the measure is known. In some cases a variety of measures may
be appropriate for a single sub—basin, i.e., different measures can be
considered for individual farms. Realistically, however, it is often
difficult to know exactly how much - the pollutant load will be reduced.
However, it can often be approximated with sufficient accuracy for planning
and management purposes. If desired, a likely range of reductions (e.g., a
pollutant would be reduced from 10 to 50 percent) could be factored into the
modeling process to give the upper and lower bounds for the possible
reduction.
COST—EFFECTIVENESS
In order to decide whether a control program or mix of programs is
15

-------
useful, cost—effectiveness is an important consideration. In the overview
modeling process, cost—effectiveness is defined as the ratio of the annual
cost of a control program to the amount of pollutant a control program will
prevent from entering the lake or receiving water. Those programs with the
lowest cost—effectiveness ratio are able to remove a given amount of pollutant
at lower cost than other remedial programs. The ability to generate cost—
effectiveness data for various remedial measures is one of the main features
of the overview modeling process.
Importantly, cost—effectiveness in the overview modeling process
is based on the amount of pollutant removed at the receiving water. Some
control programs may be effective in preventing pollutant inputs to the
system, but overall the program has little effect on the pollutant load
delivered to the receiving water. For example, the cost—effectiveness of a
program to control an upstream point source may be unfavorable if, due to high
transmission losses, little of the pollutant reaches the receiving water.
Figure 2 shows a simple comparison of the cost—effectiveness of
various phosphorus reduction schemes for Lake Erie. It illustrates how cost—
effectiveness can be used as a criteria (although not always the only
criteria) in making decisions. Although not shown in this figure, the cost—
effectiveness of achieving a 1 mg/L total phosphorus effluent concentration at
municipal treatment plants is very cost—effective. Thus, this strategy
(already agreed upon for Lake Erie) appears to have been a wise one from a
cost—effectiveness point of view.
16

-------
Estimating the cost—effectiveness of remedial options is extremely
tritical today, since many of the obvious pollution control measures have
already been undertaken. Given the economic realities of the ‘80s, we can no
longer afford to choose among alternatives without assessing their cost—
effectiveness. The overview modeling process is designed to help in this
process.
Biological Availability
The availability of a pollutant (e.g., phosphorus, heavy metals) is
of current concern, since some of the pollutant delivered to a receiving water
may not be in a chemical form which causes pollution. For example, of the
total phosphorus delivered to the Great Lakes from U.S. tributaries, 40
percent or more is likely to be in a biologically unavailable form. Our water
quality control efforts could be ineffective if directed at unavailable forms
of various pollutants.
To date, no attempt has been made to evaluate the cost—effectiveness
of pollutant loads according to reductions in available pollutants. However,
this could be done, at least for some pollutants. For example, it appears
that the percent availability of phosphorus contributed by point sources is
considerably greater than nonpoint sources. Consequently, the relative cost—
effectiveness between point source and nonpoint source control in terms of
removing available phosphorus would greatly favor point source control. That
is, the ratio of the cost of control to the amount of available phosphorus
removed (i.e., cost—effectiveness) will be less for point source control than
for nonpoint source control.
17

-------
Adjusting cost—effectiveness estimates to account for the biological
availability of a pollutant should greatly assist decision makers in choosing
among alternatives. If the percent availability is known or can be closely
approximated (as is the case for phosphorus), an estimate of the cost—
effectiveness of various control options can be made. Alternately, a range of
upper and lower extremes can be determined which is often all that is needed
to make responsible decisions. Such an approach avoids the need for more
detailed information about the biological availability of each source (which
in most cases is not available anyway).
Of course, if sufficient information is known about the amount of
available pollutant generated by each source in all parts of a watershed, the
appropriate loads can be generated directly. As discussed previously, the
modeling process is flexible in this respect.
LONG-TERM VALUE OF MODELING PROCESS
A major advantage of the ove;view modeling process is that it can
accommodate a large and dynamic data base. As new and more detailed
information becomes available on such critical variables as population growth
and remedial program costs, the model may be used to reevaluate which
pollution control programs offer the best results for our tax dollar.
As research on technologies to reduce pollutant inputs to the surface
waters continues, new and more reliable information will continually be
surfacing on treatment costs and efficiencies. Once the basin information is
compiled (identification of point sources, division of the watershed into sub—
18

-------
basins, etc.), the capability exists to assess the cost—effectiveness of
suggested remedial programs, both now and in the future. Thus, a planning
tool is created which should have many applications and will form the basis
for long—term watershed management.
19

-------
REFERENCE CITED
Cahill, T.H., Pierson, R.W., Jr., and B.R. Cohen (1979). “Nonpoint Source
Model Calibration in Honey Creek Watershed.” EPA—60013—79—054, 134 p.
Heidtke, T.M. (1978). “Comparing Costs of Pollution Control.” Great Lakes
Basin Communicator , Great Lakes Basin Commission, Ann Arbor, Michigan, 9,
1.
Heidtke, T.M., Monteith, T.J., Sullivan, R.A., Scheflow, D.J., Skimin, W.E.,
and W.C. Sonzogni (1979a). “Future U.S. Phosphorus Loadings to the Great
Lakes: An Integration of Water Quality Management Planning Information.”
Great Lakes Environmental Planning Study Contribution, Great Lakes Basin
Commission, Ann Arbor, Michigan.
Heidtke, T.M., Sonzogni, W.C., and T.J. Monteith (1979b). “Management
Information Base and Overview Modeling: Update of Projected Loadings to
the Great Lakes.” Great Lakes Basin Commission, Ann Arbor, Michigan, 38
p.
Johnson, M.G., Comeau, J.C., Heidtke, T.M., Sonzogni, W.C., and B.W. Stahlbauin
(1978). “Management Information Base and Overview Modelling.” Prepared
for the Pollution from Land Use Activities Reference Group, International
Joint Commission, Windsor, Ontario, 90 p.
20

-------
Johnson, M.G., Comeau, J.C., Sonzogni, W.C., Heidtke, T.M., and B.W. Stahibaum
(1980). “Modelling Effects of Remedial Programs to Aid Great Lakes
Environmental Management,” J. Great Lakes Res. , in publication.
Sonzogni, W.C., Jeffs, D.N., Konrad, J.C., Robinson, J.B., Chesters, C.,
Coote, D.R., and R.C. Ostry (1980). “Pollution from Land Runoff,” Envir.
Sci. and Tech. , in publication.
Sonzogni, W.C., Monteith, T.J., Skimin, W.E., and S.C. Chapra (1978).
“Critical Assessment of U.S. Land Derived Pollutant Loadings to the Great
Lakes.” Task D Report, Pollution from Land Use Activities Reference
Group, International Joint Cotmiission, Windsor, Ontario.
21

-------
GEOGRAPHK SCHEMA
MODEL SCHEMA
SIMPLE LJ T.FALITY A
(6
‘COMPLEX
____ lMuIacPALITV B
/
I
©
\
/
/
/
\
(5)
SEIWIALE
SEWERS
SEWERS
fl
LAKE LAKE
FIGURE 1:
WATERSHED MODEL LLUSTRATION

-------
Cu , , .,
UJ4)
OL.
=4,
a-a.
30--
FIGURE 2
SAMPLE STRATEGY FOR PHOSPHORUS REDUCTION - LAKE ERIE
20-
10 -
— 25.5
®
21.0
——-——--17.5
Streetsweeping
.
.
Best
Management
.
Practices
.•.
0.5ppm
Municipal
Effluent
Limit
————4.5
© Best
Stewdardship
Practices - ..
.
. .
.
.
Minimal Cost
a)
çø
C”

50 60
10 20
1 •
30 40
COST per year (S Million)
70

-------
Table 1
TYPICAL TOTAL PHOSPHORUS UNIT AREA LOADS
for RURAL LAND, FORESTED LAND and WETLANDS
IN THE GREAT LAKES BASIN
(kg/ha — year)
Soil Texture
Coarse Medium Fine
L.and Use and Intensity Sand Low Loam Loam Clay Organic
Rural
Cultivated Fields — Row Crop
(low animal density) 0.25 .65 0.85 1.05
Cultivated Fields — Mixed Farming
(medium animal density) 0.10 0.20 0.30 0.55 0.85
Pasture/Range — Dairy
(medium animal density) 0.05 0.05 0.10 0.40 O.6d
Grassland 0.05 0.05 0.10 0.15 0.25
Forest
General 0.05
Wetlands
Natural Area
Muck Farm
Unit area loads may be higher when soil has an unusually high clay content.
it area loads may be higher in certain unique forested areas with clay soils.
For example, the Nemadji River basin which flows into Lake Superior contributes
about 1.0 kg/ha — year.

-------
Table 2
TYPICAL TOTAL PHOSPHORUS
URBAN UNIT AREA LOADS
for GREAT LAKES BASIN
(kg/ha — year)
Industrialization Level
Urban Low Mediun High
Combined Sewer 9 10 ii
Separate Sewer 2.5 3.0
Unsewered 1.2
Small Urban 2.5
(Sewer System not
Differentiated)
Urbanizing Land 75

-------