United States
Environmental Protection
Agency
Environmental Research
Laboratory
Athens GA 30613
vn
Research and Development
EPA-600/S3-82-057 Sept. 1982
Project Summary
River Basin Validation of the
Water Quality Assessment
Methodology for Screening
Nondesignated 208 Areas:
Volumes I and II
Michael J. Davis, Michael K. Snyder, and John W. Nebgen
Techniques for estimating diffuse
nutrient loads and their water quality
effects throughout large watersheds
were tested under various field situa-
tions for a wide range of data availability
circumstances, water quality param-
eters, and hydrologic-hydraulic condi-
tions. Loadings were calculated using
methods from Loading Functions for
Assessment of Water Pollution from
Nonpoint Sources (EPA-600/2-76-
151) for five river basins, including the
Sandusky River in Ohio, the Chester
and Patuxent Rivers in Maryland, and
the Ware and Occoquan Rivers in
Virginia. Water quality response to
these loads was calculated using
methods from Water Quality Assess-
ment A Screening Method for
Nondesignated 208 Areas (EPA-
600/9-77-023) for the Sandusky,
Chester, Patuxent, and Ware rivers
and estuaries.
Obtaining sufficient data to operate
the model was the biggest problem in
applying the loading methodology to
specific river basins. Given the inher-
ent inaccuracy in the basic Universal
Soil Loss Equation (USLE) approach,
use of the national data base provided
in the original document was justified,
especially if supplemented with speci-
fic county cover (R) factors. Despite
the inaccuracies resulting from the
use of a national data base, two to
three person weeks of effort per basin
should produce useful inputs to water
quality screening assessments.
The loading methods and the water
quality methods tested in this study
were highly compatible and gave rea-
sonably accurate predictions of in-
stream, lake, and estuary water
quality constituent concentrations.
The river modeling methods are the
most accurate, followed by the tech-
niques for estuaries and then impound-
ments. Low flow, steady state condi-
tions are better predicted than high
flow, unsteady loading situations.
Applying these techniques, the water-
shed management planner should be
able to recommend appropriate actions
to investigate pollutant problem areas
more closely in specific watersheds.
This Project Summary was devel-
oped by EPA's Environmental Research
Laboratory, Athens, GA, to announce
key findings of the research project
that is fully documented in two
separate reports (see Project Report
ordering information at back).
Introduction
In August 1977 the U.S. Environmen-
tal Protection Agency (EPA) published
Water Quality Assessment A Screen-
ing Method for Nondesignated 2O8
-------
Areas (EPA-600/9-77-023) This docu-
ment is a compendium of techniques
designed to aid in the assessment of
water quality problems in large areas
that encompass a wide spectrum of
human activities and water quality
conditions These include agriculture and
silviculture, as well as industrial and
municipal activities. In this Water
Quality Assessment Methodology (WQAM),
Tetra Tech, Inc., under EPA contract,
brought together a number of methods
designed to accommodate both urban
and non-urban nonpomt sources, as
well as municipal and industrial point
sources of pollutants. In addition to the
assessment of effluent water quality,
WQAM provided for systematic routing
of these pollutants through rivers and
streams, impoundments, and estuary
systems All algorithms were designed
to be used as hand calculation tools.
In a separate study, Midwest Research
Institute, under EPA contract, developed
methods for estimating diffuse loads
entering receiving waters. The study
was described in Loading Functions for
Assessment of Water Pollution from
Nonpoint Sources (EPA-600/2-76-
151).
The primary goal of the study described
in this Project Summary was to demon-
strate Midwest Research Institute's
loading functions and Tetra Tech's
water quality screening procedures
under authentic field situations. The
demonstration was designed to subject
the procedures to a wide range of data
availability, water quality parameters,
and hydrologic/hydrauhc situations. In
addition to the primary goal, secondary
goals were to:
1. Provide a report demonstrating
the WQAM, to be used as a guide
by planners.
2. Show the degree of compatibility
between the nonpomt-loading
methods and the water quality
screening methods in the WQAM.
3. Develop firmer insight into the
strengths and weaknesses of the
nonpoint loading methodology.
4. Evaluate the sensitivity of non-
point load estimates to varying
degrees of data availability.
5. Determine how critical or neces-
sary the quality and quantity of
nonpoint source details are with
regard to reliably modeling in-
stream processes as they are
affected by nonpoint loading.
6. Demonstrate strengths and weak-
nesses of the WQAM screening
methodology.
Five river basins were examined.
These are the Sandusky River in Ohio,
the Chester and the Patuxent Rivers in
Maryland, and the Ware and Occoquan
Rivers in Virginia. Loading analyses
were also performed on the Potomac
River Basin and the Susquehanna River
Basin.
This summary describes work reported
in two volumes River Basin Validation
of the Water Quality Assessment
Methodology for Screening Nondesig-
nated 208 Areas, Volume I. Nonpoint
Load Estimation; Volume II. Chesapeake-
Sandusky Nondesignated 208 Screening
Methodology Demonstration. Volume I
is a discussion of the application of the
nonpomt load assessment methodology
to a number of nx^er basins Volume II
considers the application of the water
quality screening methodologies to these
same basins. The nonpoint source load
estimates given in Volume I were used
as inputs to the calculations involving
wet weather conditions that are pre-
sented in Volume II The two volumes
are organized similarly; the river basins
were considered in the same order in
both. There is cross-referencing in
Volume I to portions of Volume II so that
the interested reader can see how
results obtained in Volume I are used in
the second volume.
Although the two volumes are related,
each can stand alone as a separate
demonstration of the different meth-
odologies.
Volume I
Data Availability Application of the
nonpomt source loading methodology
requires a large volume of data in spite
of the relative simplicity of the overall
approach Therefore, a major problem
in this, and probably in any other,
application was data availability. A
screening analysis should by its nature
not require the generation of significant
quantities of new data. Those data that
are used in the analysis should already
be available and should require a
minimum of manipulation prior to use.
For example, the Sandusky Basin has
been well studied, but there was a
definite lack of applicable data readily
available for this study. Generally, the
available data were aggregated to the
county level, e.g., land use information.
Also, as might be anticipated, in all the
basins there was a problem in estimating
sediment delivery ratios. Long-term
sediment yield data were generally not
available for the basins, therefore,
average delivery ratios could not be
properly estimated. Furthermore, there
was a lack of the sediment yield data
needed to define variability within the
basins. Finally, there was generally, no
way to estimate how average delivery
ratios might vary with the season. Use
of a single average value for the delivery
ratio could result in a considerable
over- or underestimation of loads for
particular subbasins and seasons. This
difficulty was nearly universal It was, in
fact, a problem of less concern in the
Sandusky Basin than for most basins
because some sediment measurements
were available and because the effi-
ciency of delivery is thought to be
relatively uniform throughout the basin
Information on pollutant loading rates
and pollutant characteristics in urban
areas was also not readily available. It
was purely a matter of chance that
actual measurements were available
for use in one of the urban areas
(Bucyrus) in one of the basins (Sandusky)
Value of Parameter Refinement
The most important parameter refine-
ments involved land use data and the R,
K, and C factors in the universal soil loss
equation The national data base used
provided R values by the Land Resource
Area (LRA). These could easily be
replaced by values that more nearly
represent each county For example, in
the Sandusky, the change was from an
annual value of 150 to 125. (Of course,
individual event values were calculated
for use in the demonstration ) Cover (C)
factor values were also changed by the
refinement process. The C values used
for the individual counties reflect
changes in the stage of crop growth,
which is an improvement over the
average annual C values in the data
base The level of resolution of soil
credibility values was improved from
the LRA level (in the data base) to the
county level. In some cases, resolution
was at the subbasin level. As much as a
20 to 40% decrease in erodibility value
was noted for some subbasins in the
Sandusky due to refinements in the K
values Land use changes in that basin
mostly increased soil loss in the interval
between 1967 (data base) and the base
year used in the individual basin
calculations.
On balance then, as compared to the
data base values, the refinements for
the Sandusky led to decreases in R, K,
and C and, therefore, to a decrease in
annual soil loss over that which would
be obtained using the data base. Land
use changes partially offset the decrease.
Again, using the Sandusky Basin as an
-------
example, in six out of the eight counties,
average soil losses decreased by 20%or
more due to the refinements. Given the
level of effort required to produce the
refinements and the inherent inaccuracy
in the approach, the use of the data base
is a cost-effective approach. For annual
soil loss calculations in the basins
studied, the original data base would
provide useful results (as'compared to
the refined values) if one merely
modified the R factors for each county
and accounted for the major change in
cropland. That is, a significant improve-
ment was possible in this case with only
a limited amount of effort.
Certain problems occurred in the
attempt to improve the estimates for
some of the parameters, as already
noted. A particular problem was esti-
mating sediment delivery ratios, a
problem that was exacerbated in the
case of individual subbasins. Reasonable
estimates of delivery ratios were
essential for accurate estimates of
sediment delivered to a stream. Lacking
a general approach to the problem, the
delivery ratio issue will continue to
frustrate many applications of the
methodology Although less difficult,
problems also occurred with other
parameters as well The LS and the P
factors in the USLE were not modified
and were used directly from the existing
national data base Improvement of the
estimates used requires substantial
information on topography and soil
conservation practices in each basin.
As already noted, data on the loading
rates for pollutants on city streets were
generally lacking and recourse must be
made to tabulated, crude averages
Sensitivity Analysis In the case of
urban nonpomt loads, the important
matter of sensitivity to assumptions
was considered. The major problem
centered on determination of street
loading rates and use of an annual
average approach. For rural nonpomt
loads, the various factors used in
determining sediment of nutrient loads
(except rainfall inputs) were multiplied
together to obtain the final result.
Therefore, uncertainties in the factors
were multiplied For example, for
sediment loads, a 20% error in each
factor involved in determining the load
gave approximately a 300% error in
the load, assuming no compensation
among the errors Similar errors in the
case of nutrient calculations yielded a
total error of about 400%. Because most
of the factors could not be determined
with an error of less than 20%, the
possible error in the results could be
quite large unless there is compensation
among the errors. This fact indicated
that the results obtained are always
rather uncertain.
Uncertainties in land use information
in the present application related
primarily to the resolution of the
information. Agricultural land use data
were generally available; however, they
were at the county level of resolution
Therefore, specifying land use conditions
in a subbasin was difficult The primary
need in land use data was for accurate
specification of the cropland its area
and type of crop. Land use affects
pollutant load calculations through the
C factor in the USLE In an application in
which a pollutant Ipad was needed for a
subbasin that covers a fraction of a
county and in which land use and other
data were available only at the county
level of resolution or lower, the loads
may be grossly overestimated. This
overestimation could occur in a subbasin
for which a higher than average fraction
(for the county) is cropped, for which
slopes are steeper than average, or for
which there are no conservation prac-
tices applied (P = 1). Poor resolution of
needed data can result in substantial
errors for particular locations within a
basin
Results were also somewhat sensitive
to errors in describing agricultural
practices in a basin. The significance of
errors in practices related primarily to
the problem of timing of agricultural
operations and, therefore, the degree of
cover on the ground at particular times.
In summary, errors in results were
directly proportional to errors in the
various parameters used in the analysis
because they are multiplicative. Assess-
ment of sensitivity to errors in the
description of practices or land use, or
the degree or resolution in the available
data is an involved exercise that will
yield results that vary considerably from
basin to basin. Such variability was
anticipated because of different rainfall
patterns and the degree of non-homoge-
neity of land use among basins.
Level of Effort Required m an Applica-
tion Application of the nonpoint load
estimation methodology to basins
such as those examined in this study
should require on the order of two to
three person weeks of effort per basin.
This estimate assumes an analyst
familiar with the procedure and with the
general subject of rural nonpoint source
loads. It also assumes familiarity with
use of the nonpoint calculator program
The availability and use of more
extensive data than considered in this
demonstration would increase the time
required Report preparation is not
included in the time estimate
Verification of the Load Estimation
Procedures Considering the lack of
measured nonpoint loads (both rural
and urban) available for comparison
and the long-term average nature of the
estimates that have been made, veri-
fication of the procedure by direct
comparison with measured loads was
quite difficult Comparison with measured
instream concentrations was a more
promising 'approach The results pre-
sented in Volume II indicate the level of
verification that can be expected for the
approach used, particularly in the case
of the Sandusky Basin
Future Applications In the present
study, considerable effort was expended
in selecting a series of events for each
basin so that consistent flow data were
available for use in the instream
assessment This effort was necessary
to assure compatibility and to allow an
attempt at verification of the results In
actual applications, such a selection of
actual events may be unnecessary. A
possible approach would be to define
typical average events for various
stages in cover occurring throughout
the year These "typical" events could
be equivalent to events that produce
some fraction of the total soil loss that
occurs during some fraction of the year
The information needed to define such
an event is available in terms of the
annual distribution of the R factor
Therefore, in an application it is possible
to consider design storms with charac-
teristics that can be defined independently
of an actual watershed
Load Estimation in Specialized Appli-
cation Volume I provides an example
of the application of much of the basic
rural nonpoint source methodology to
the problem of estimating long-term
nutrient fluxes in streams. This applica-
tion showed that the procedures can be
applied in ways that overcome some of
their fundamental weaknesses (e.g.,
the need for a delivery ratio), while
providing useful results. It is likely that
other specialized applications can be
developed also.
Attainment of Study Goals The
primary goal of the study was to
demonstrate the nonpoint loading
methodology under actual field condi-
tions This goal was accomplished. The
nonpoint loading procedures and the
water quality screening methodology
-------
were also shown to be compatible,
which was one of the subgoals of the
program.
The application under field conditions
pointed out the primary strengths of the
methodology its relative simplicity
and the ease with which basic calcula-
tions can be done and its weaknesses
dependence on a delivery ratio, a
higher level of special aggregation in
the case of practical applications in
large basins, and the need for large
amounts of data. These characteristics
were well demonstrated in the studies
of the various basins, which illustrated
the degrees of data availability likely in
practice. These applications indicated
that major parameter refinements tend
to be time consuming and, in many
cases, of limited value. They also
indicated the difficulty of determining or
assigning sediment delivery ratios in
most cases.
Impact of Methodological Shortcom-
ings Several important features in
the rural nonpomt methodology limited
the accuracy that can be expected from
the results of an assessment. These
features include: (a) a high level of
spatial aggregation in the analysis an
important fact because the USLE is
intended for rather small, homogeneous
areas, (b) the use of a delivery ratio to
account for sediment transport; (c) the
assumption that pollutants such as
phosphorus are associated with sedi-
ment, and (d) the long-term average,
nonhydrologic nature of the USLE
An attempt was made to overcome
the lack of suitability of the USLE for
analyzing actual events by averaging
over many events. Dealing with an
average event in this manner was
acceptable, however, proper averaging
required many events occurring over a
long period of time. Data were not
always available to carry out such
averaging.
Additional shortcomings occurred in
the urban methodology used, which
dealt with annual loads and which
depended upon street loading rates that
were not well established.
A screening methodology such as
was applied here is intended for
relatively easy application using existing
data. Overcoming some of the limita-
tions listed above would require greatly
increased amounts of data to reduce
spatial resolution problems, to provide
increased information on sediment
transport, and to provide data on runoff
needed to allow soluble forms of
constituents to be included and to allow
Table 1.
Water Quality Simulation Results Summary for Rivers
System
Sandusky
Patuxent
LOW FLOW
Temperature
BOD
Dissolved Oxygen
Coliforms
HIGH FLOW
Sediment
BOD
Total N
Total P
o
o
Key:
Results good to excellent
Q Results fair to good
* Simulation performed, no comparative data available
(blank) No simulation performed
a more hydrologically-based approach.
Because this demonstration illustrated
the fact that needed data may not be
available even for the screening
approach used, it seems reasonable to
conclude that more rigorous approaches
can result in even more obstacles due to
data limitations, Especially when large
areas must be considered.
The users of the nonpoint methodol-
ogy should be well aware of its limita-
tions. These limitations, however,
should not prevent the use of the
approach. As the present study showed,
applications can be made that result in
useful inputs to water quality assess-
ments in spite of certain methodological
shortcomings of the procedures used
The user should always recall that the
methodology was intended for screening
purposes.
Volume II
Applicability of Techniques The
nonpoint source calculator and the non-
designated 208 screening methodology
were highly compatible. Outputs from
the nonpoint source calculator were
easily adapted and in some cases were
used directly in the mass balance
equations of the screening methods.
Event-based urban nonpoint loads were
not readily predictable by the nonpoint
calculator, but it is questionable whether
the non-designated 208 screening
methods are applicable under these
high flow - unsteady loading scenarios
except to provide approximate upper
and lower limits of mstream pollutant
levels.
Loading predicted by the nonpoint
source calculator in conjunction with
mass balance techniques employed by
the non-designated 208 screening
methods provided reasonably accurate
predictions of instream, lake, and
estuary water quality constituent con-
centrations No effects due to basin size
or location were noted that detracted
from either the applicability or accuracy
of the methods. Generally, loss of
accuracy due to a loss in resolution was
mitigated by the averaging effects
intrinsic to larger systems.
A qualitative assessment of the
rivers, estuaries and impoundments
methods is shown in Tables 1, 2 and 3.
In general, the tables imply that the
river methods were the most accurate
followed by estuaries and then im-
poundments. Within each method it
should be mentioned that low flow -
steady state conditions were more
readily reproducible than high flow -
unsteady loading situations. The im-
poundment methods probably required
the least time and background skills to
apply The riverine methods usually
required more time to apply than the
estuary methods. The riverine results,
however, should be easier to interpret
for the uninitiated user than the results
of the estuary methods.
Loadings predicted by the nonpoint
source calculator in which all parameters
were assumed to be correlated with
-------
sediment loss were more accurate for
sediment and phosphorus than for
nitrogen and BODs. This was an ex-
pected result. In general, predicted
nonpoint source nitrogen and BODs
loads were too low based on compari-
son of observed and predicted instream
concentrations.
For conservative parameters, linear
increases or decreases in load esti-
mates (either point or nonpoint) resulted
m approximately linear changes in the
concentrations of those constituents in
the water bodies. Therefore, an approxi-
mate error analysis could be performed
directly using load estimates. For
Table 2. Water Quality Simulation
Results Summary for Im-
poundments
Occoquan
IMPOUNDMENTS
Temperature
BOD
Dissolved Oxygen
Sediment
Total N
Total P
©
o
Key:
Results good to excellent
Q Results fair to good
nonconservative parameters, changes
in stream, lake, or estuary concentra-
tions caused by increases or decreases
in loadings could only be determined by
routing the pollutants through the
receiving water system. An error
analysis using loading changes and
assuming the constituents behaved
conservatively gave an upper limit for
the concentration changes likely to be
encountered.
Although the methods appeared to be
a powerful tool for quickly identifying
water quality problem areas, the use of
the predictive techniques in conjunction
with observed data further added to
their effectiveness. By doing this, the
planner could identify specific problem
areas in which quality cannot adequately
be described by the simple techniques.
In most cases, the planner will be able
to recommend action, based on an
understanding of the methods he has
already applied, to investigate the
problem area more closely. These
further investigations may include
sampling programs or the use of a more
sophisticated analytical tool
Rivers and Streams Hydraulic
characterization of rivers and streams
was one of the most error-prone steps
in the methods A major reason for this
was that flow, in many cases, was a
function of phenomena that could not
be estimated directly from the surface
topography Unless the user had ground-
Table 3. Water Quality Simulation Results Summary for Estuaries
System
Chester
Patuxent
Ware
LOW FLOW
BOD
Coliforms
Total N
Total P
HIGH FLOW
Sediment
BOD
Total N
Total P
0
*
*
O
O
*
O
o
o
0
0
Key:
Results good to excellent
0 Results fair to good
O Results poor to fair
Simulation performed, no comparative data available
(blank) No simulation performed
water measurements or detailed poten-
tiometric maps, these effects could not
be properly characterized.
Dissolved oxygen prediction was far
more sensitive to errors in estimating
reaeration rates than m estimating
deoxygenation rates Predictive tech-
niques for stream reaeration deserve
additional attention Comparison of
predicted instream fecal or total cohform
concentrations with observed data was
impractical
Impoundments Thermal plots from
the impoundment thermal model accu-
rately described water temperature,
thermal gradients, and time of the onset
of stratification The greatest difficulty
in using the thermal profiles lay in the
selection of the correct plot to apply in a
borderline case In such cases, selection
of the maximum depth parameter was
aided by also considering the mean
impoundment depth. The hydraulic
residence time strongly affected the
thermal profile of an impoundment.
Accuracy of sedimentation calcula-
tions for impoundments depended
primarily on accurate load estimates.
Although on-site data should be used,
the demonstration watershed results
indicated that the Universal Soil Loss
Equation may be used with some
confidence. Adequate knowledge of
sediment diameters is required because
trapping efficiencies were sensitive to
particle size
The ability of the methods toquantita-
tively predict parameter values associ-
ated with eutrophication was limited
Plant growth could only be approximated
and seasonal effects could not be
represented adequately
The hypohmnion dissolved oxygen
calculations for impoundments were
sensitive to the BOD loading rate and
decay rates Qualitatively useful results
were predicted by the simplified hypo-
limnion dissolved oxygen model even
when BOD decay rates were not
accurately known.
Estuaries The stratification-circu-
lation method was preferred for estuarme
classification, but the required data was
not always available To obtain a
complete picture of the hydrodynamic
variation that an estuary might undergo,
the surface velocity, the net freshwater
velocity, and the surface and bottom
salinity should be available for high and
low freshwater inflows Use of the
flow ratio method underestimated the.
degree of vertical stratification.
The tidal prism and modified tidal
prism flushing times were related, and
-------
their ratio seemed to depend on the
estuary volume. The fraction of fresh-
water method was fairly insensitive to
the number of segments used to
estimate flushing times. For flushing
times derived by the modified tidal
prism method that were similar at high
and low flows, mechanisms other than
advective flow were more important in
flushing an estuary The fraction of
freshwater and the modified tidal prism
methods predicted more similar flushing
times for small estuaries
Low flow predictions of pollutant
distributions in estuaries were good for
conservative constituents. For non-
conservative constituents, the modified
tidal prism method must be used.
Pollutant distributions predicted for
unsteady flow or unsteady loading
represented upper and lower limit
concentrations. Contamination in the
estuary caused by replacement waters
could be estimated by comparing
observed profiles to those predicted by
the estuarme methods.
Michael J. Davis, Michael K. Snyder. and John W. Nebgen are with Midwest
Research Institute. Kansas City. MO 64110.
Robert B. Ambrose is the EPA Project Officer (see below).
The complete report consists of two volumes, entitled "River Basin Validation of
the Water Quality Assessment Methodology for Screening Nondesignated
208 Areas:"
"Volume I. Nonpoint Source Load Estimation," (Order No. PB 82-260 837;
Cost: $15.00, subject to change}
"Volume II. Chesapeake-Sandusky Nondesignated208 Screening Method-
ology Demonstration," (Order No. PB 82-260 845; Cost: $19.50, subject to
change)
The above reports will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Environmental Research Laboratory
U.S. Environmental Protection Agency
Athens, GA 30613
-------
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Postage and
Fees Paid
Environmental
Protection
Agency
EPA 335
Official Business
Penalty for Private Use $300
0000329
------- |