EPA
United States
Environmental Protection
Agency
Office Of Water
(WH-585)
EPA-823-R-92-005
August 1992
Technical Guidance Manual
For Performing Waste Load
Allocations
Book III: Estuaries
Part 4
Critical Review Of
Coastal Embayment And
Estuarine Waste Load
Allocation Modeling
* 1 9 9 2
THE YEAR OF
CLEAN WATER
Recycled/Recyclable
Printed on paper that contains
at least 50% recycled fiber
-------
-------
TECHNICAL GUIDANCE MANUAL
FOR PERFORMING WASTE LOAD ALLOCATIONS
BOOK III: ESTUARIES
PART 4: Critical Review of Coastal Embayment and Estuarine
Waste Load Allocation Modeling
Project Officer
Hiranmay Biswas, Ph.D.
Edited By
Robert B. Ambrose, Jr. P.E.1
Prepared by
Paul L. Freedman, P.E.2
David W. Diiks, Ph.D.2
Bruce A. Monson2
1. Center for Exposure Assessment Modeling,
Environmental Research Laboratory, U.S. EPA, Athens, GA
2. LTI, Limno-Tech, Inc., Ann Arbor, Ml
Prepared for
U.S. Environmental Protection Agency
401M. Street, S.W.
Washington, D.C. 20460
-------
-------
Table of Contents
Preface v
Acknowledgements vjj
10. Great Lakes Embayment Seasonal Phytoplankton Model of Saginaw Bay .... 10-1
10.1. Background 10-1
10.2. Problem Setting ..'..".' 10-1
10.3. Model Application 10-2
10.4. Post-Audit '.'.'.'. 10-8
10.5. References 10-8
11. Potomac Estuary Water Quality Modeling 11-1
11.1. Background H_1
11.2. Problem Setting H_1
11.3. Dynamic Estuary Model (DEM) of Dissolved Oxygen 11-2
11.4. Potomac Eutrophication Model (PEM) 11-6
11.5. Finite Element Model 11-13
11.6. References 11-18
12. Manasquan Estuary Real Time Modeling . 12-1
12.1. Background 12-1
12.2. Problem Setting " 12-1
12.3. Model Calibration 12-2
12.4. References 12-11
13. Calcasieu River Estuary Modeling 13-1
13.1. Background 13_1
13.2. Problem Setting 13_1
13.3. Model Application 13_2
13.4. Total Maximum Daily Loads 13-7
13.5. References 13_8
14. Expert Critique of Case Studies 14-1
14.1. Robert V. Thomann, Ph.D. „ . . 14-1
14.2. Donald R.F. Harleman, Ph.D. . . 14-10
14.3 Gerald T. Orlob, Ph.D., P.E. :...... 14-14
HI
-------
-------
Preface
The document is the third of a series of manuals provid-
ing information and guidance for the preparation of
waste load allocations. The first documents provided
general guidance for performing waste load allocations
(Book I), as well as guidance specifically directed
toward streams and rivers (Book II). This document
provides technical information and guidance for the
preparation of waste load allocations in estuaries. The
document is divided into four parts:
Part 1 of this document provides technical information
and policy guidance for the preparation of estuarine
waste load allocations. It summarizes the important
water quality problems, estuarine characterisitics and
processes affecting those problems, and the simula-
tion models available for addressing these problems.
Part 2 provides a guide to monitoring and model
calibration and testing, and a case study tutorial on
simulation of waste load allocation problems in
simplified estuarine systems. Part 3 summarizes initial
dilution and mixing zone processes, available models,
and their application in waste load allocation.
This part, "Part 4: Critical Review of Coastal Embay-
ment and Estuarine Waste Load Allocation Modeling,"
summarizes several historical case studies, with critical
review by noted experts. The reader should refer to the
preceding parts for information on model processes,
available models, and guidance to monitoring and
calibration.
The technical guidance is comprehensive and state-of-
the-art. Case studies of applications serve as the best
teacher of the proper and improper use of this technical
guidance. Similar models are often used in large fresh-
water coastal embayments and estuaries because
there are some similarities in their hydrodynamic
transport processes.Therefore, included in this part are
one freshwater embayment study and three estuarine
studies where models were used for waste load alloca-
tion. These studies have been selected to provide a
range of representative geographic areas, freshwater
bays and embayments, estuaries, and models. The
studies were not selected because they were ex-
emplary but rather because they represented applica-
tions of diverse approaches.
Each of the studies has particular merits and deficien-
cies; the balance is different in each study. Perfect
examples are not always the best teachers. By exam-
ing the strengths and weaknesses of each application
the reader can appreciate how to best use the technical
guidance and how to avoid misuse and common
problems.
The examples are summarized with only limited com-
mentary. The information for each is presented with
sufficient detail to allow the reader to understand what
was done and to highlight certain noteworthy aspects.
Following the case examples, three experts critique the
relative merits and deficiencies in each case study and
provide their opinions on the proper approach to es-
tuarine modeling.
A draft version of this document received scientific peer
review from the following modeling experts:
Steven C. Chapra,
University of Colorado-Boulder
Donald R.F. Harleman,
Massachusetts Institute of Technology
Gerald T. Orlob,
University of California-Davis
Robert V. Thomann,
Manhattan College
Their comments have been incorporated into the final
version.
Organization: "Technical Guidance Manual for Performing Waste Load Allocations.
Book III: Estuaries"
Part
1
2
3
4
Title
Estuaries and Waste Load Allocation Models
Application of Estuarine Waste Load Allocation Models
Use of Mixing Zone Models in Estuarine Waste Load Allocation Modeling
Critical Review of Coastal Embayment and Estuarine Waste Load
Allocation Modeling
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
JUL 2 0 1992
OFFICE OF
WATER
MEMORANDUM
SUBJECT: Final Technical Guidance Manual For Performing Wasteload
FROM:
TO:
Allocations: Book III, Estiiar£es/..P,a£,tS('"3) and 4
"
Tudor T. Davies, j Director
Office of Science and Technology (WH-551)
Regional Water Management Division Directors
Regional Environmental Services Division Directors
Regional TMDL/WLA Coordinators
Attached, for national use, is the final version of the
Technical Guidance Manuals; for Performing Wasteload Allocations:
Book III, Estuaries, Parts 3 and 4. Parts 1 and 2 were finalized
during FY 91 and have been Jin distribution ever since for national
use. We are sending extra copies of Parts 3 and 4 of the guidance
document to the TMDL/WLA coordinators for distribution to the
States to use in conducting wasteload allocations.
i
An earlier draft of Parts 3 and 4 were reviewed by your staff
and their comments were considered in finalizing this guidance.
Major modifications to the I earlier draft include:
o The discussion on mixing zone criteria in Part 3 (see page
7-3) is now consistent with the March 1991 version of the
Technical Support Document for Water Quality-based Toxics
Control. '
o The title of Part 4 hab been modified from Critical Review of
Estuarine Wasteload Allocation Modeling to Critical Review of
Coastal Embayment and Estuarine Wasteload Allocation Modeling.
This change was necessary because the Saginaw Bay example in
Part 4 of this guidance does not meet the strict regulatory
definition of an estuary.
If you have any questions or comments or desire additional
information, please contact Russell S. Kinerson, Exposure
Assessment Branch, Standards and Applied Science Division (WH-585),
Telephone (202) 260-1330. !
Attachments
Printed on Recycled Paper
-------
Acknowledgements
This document represents the efforts of several people
and the integration of several documents. Site selec-
tion and national expert review were managed by
Hiranmay Biswas, U.S. EPA Office of Science and
Technology (formerly Monitoring and Data Support
Division). Chapter 12 on the Manasquan along with
portions of Chapter 11 were excerpted from an earlier
draft Technical Guidance Manual prepared by Richard
Wagner, Jane Metcalfe, and Elizabeth Southerland of
JRB Associates. Chapter 10 on Saginaw Bay was
prepared by Bruce Monson, Chapter 11 by David Dilks,
and Chapter 13 by Paul Freed man, all of Limno-Tech,
Inc. Scott Hinz and Susan Johnson of Limno-Tech are
acknowledged for their work in editing this draft docu-
ment.
National expert peer review of this manual was
provided by Robert V. Thomann, Manhattan College,
Donald R.F. Harleman, Massachusetts Institute of
Technology, and Steven C. Chapra, University of
Colorado at Boulder.
In addition, helpful internal review comments were
received from Thomas Barnwell, U.S. EPA Athens En-
vironmental Reasearch Laboratory; James Martin,
ASci Corporation; Rick Brandes, U.S. EPA Office of
Wastewater Enforcement and Compliance; Steve
Glomb, U.S. EPA Office of Wetlands, Oceans, and
Watersheds; Dale Bryson, U.S. EPA Region V; Mimi
Dannel, U.S. EPA Region VI; Aaron Setran, U.S. EPA
Region IX; Clyde Bohmfalk, Texas Water Commission,
Michael Waldon, The University of Southern Louisiana;
and June Harrigan, State of Hawaii. A significant part
of this internal review process was managed by Karen
Gourdine, U.S. EPA Office of Science and Technology.
Bruce Monson of Limno-Tech was responsible for the
draft formatting of this document, and the final layout
was done by Tad Slawecki, also of Limno-Tech.
VII
-------
-------
10. Great Lakes Embayment Seasonal Phytoplankton Model
of Saginaw Bay
10.1. Background
The Saginaw Bay is not an estuary. However, its
hydrodynamic processes are similarto those observed
in some shallow estuaries with wind-driven circulation.
The Saginaw Bay phytoplankton model of Bierman and
Dolan (1986a,b) is presented here to illustrate the ap-
plication of a dynamic and kinetically complex box
model to a Great Lakes embayment. This model was
calibrated with two comprehensive data sets. Follow-
ing significant reductions in loadings and changes in
the Bay's water quality, the model projections were
tested and validated (post-audit) with another com-
prehensive data set. The model was developed as part
of a long-term study of eutrophication in Saginaw Bay.
It was designed as a management and research tool
to estimate phytoplankton response to various phos-
phorus control strategies. The model was used exten-
SAGINAW
FLINT
SCALE
6 IO ao 30mil«»
6 10 20 30 40 50km
Source: Bierman and Dolon, 198
Figure 10-1. Saginaw Bay site map [Bierman and Dolan
(1986a). Reprinted from ASCE Journal of
Environmental Engineering, Vol. 112, No. 2. p.
401. With permission].
sively by the USEPA and International Joint Commis-
sion to evaluate nutrient loading reductions for
Saginaw Bay.
The authors describe the model as "a deterministic,
spatially segmented, multi-class phytoplankton
model." The phytoplankton comprise five functional
groups: diatoms, greens, non-nitrogen-fixing blue-
greens, nitrogen-fixing blue-greens, and "others."
Nutrient uptake is considered for phosphorus,
nitrogen, and silica. Herbivory, settling, and decom-
position are mechanisms of phytoplankton depletion.
10.2. Problem Setting
Located on the western shore of Lake Huron (Figure
10-1), the Saginaw Bay watershed is approximately
21,000 km (8108 mi2). It is dominated by agriculture,
forest, and four urban-industrial centers: Bay City, Flint,
Midland, and Saginaw. The 1980 population for the
area was slightly over 1,200,000. The area is drained
by the Bay's major tributary, the Saginaw River. The
River accounts for 90 percent of the tributary inflow to
the Bay.
Saginaw Bay extends 90 km from the River's mouth to
the Bay's opening to Lake Huron. It is broad (42 km),
shallow (10 m average depth), and vertically well-
mixed. The average hydraulic residence time is ap-
proximately four months.
The Bay has been characterized as behaving like a
simple estuary (Ayers et al, 1956). Like estuaries, it is a
nutrient-rich arm of a larger nutrient-poor water body,
Lake Huron (Richardson, 1974). Furthermore, water
levels and flow directions of the Bay change. Unlike an
estuary, the water level is influenced by wind patterns
rather than tides. Northern gales can create a seiche in
the Bay that raises the water level at the mouth of the
Saginaw River by more than a meter (Fish and Wildlife
Service, 1956; cited by Richardson, 1974).
The International Joint Commission identified Saginaw
Bay as one of forty-two Great Lakes Areas of Concern
needing remedial action. Eutrophication of the Bay had
caused taste, odor, and filter-clogging problems for
municipal water supplies. Waste discharges and runoff
have been major contributors to water quality degrada-
tion. In the late 1970's, phosphorus reduction
programs were implemented at wastewater treatment
plants and resulted in large reductions of phosphorus
loading to the Bay. From 1975 to 1980 the phosphorus
loads were reduced over 65 percent. The model was
10-1
-------
calibrated and verified when the phosphorus loadings
were high (1974 and 1975) and tested in a post-kudit
following the large reductions of phosphorus (1980).
10.3. Mode! Application
Although this model's development began in a pore
simple form, it is presented here in its most advanced
form as a spatially segmented, temporally dynamic
model. A more spatially simplified precursor model
(Bierman and Dolan, 1980) provided some valuable
conclusions aboutthe biological and chemical proces-
ses In the waterbody. These findings were used to
develop the kinetic structure and calibrate the more
spatially detailed model. For example, the factors in-
fluencing phytoplankton dynamics in the model are
temperature, light, nutrients, and zooplankton grazing.
Temperature and light were generally more grpwth-
rate limiting than nutrients. However, nutrient limitation
became Importantforpeak phytoplankton crops.jlnthe
spring and fall, the primary source of phosphorus was
external loading, which fed the dominant diatom crops.
In mid-summer, the primary source of phosphorus was
recycling within the water column and from sediments,
which fed the summer blue-green crops. ;
The multi-class phytoplankton model was developed
to predict the response of the Saginaw Bay phyto-
plankton to various phosphorus control strategies. Of
primary concern were the nuisance, bloom-forming
blue-greens that cause taste and odor problems. The
emphasis In the model was on nutrient cycling since it
Is a limiting and controllable factor in phytoplankton
growth. Several hypothetical scenarios and a| post-
audit are presented below following examination of the
calibration and validation of the model.
10.3.1. Model Description
The model developed for Saginaw Bay falls jinto a
general class of models called "box models'." The
approach involves dividing the water body into several
cells (or boxes), each of which is considered complete-
ly mixed (Figure 10-2). Transport of chemicals,
biomass, and water between cells occurs through ad-
vectlve transport and dispersion. !
The mass of pollutants, algae or other constituents in
each cell changes in response to loadings, transport,
mixing, settling, and reaction kinetics. A mass balance
equation is written for each cell and the resulting dif-
ferential equation solved simultaneously through time
for all cells by a numerical method. '
LEGEND: • BOAT STATION
A WATER INTAKE
10 KM
10 Ml
Figure 10-2. Model segmentation of Saginaw Bay [Bierman
and Dolan (1986a). Reprinted from ASCE
Journal of Environmental Engineering, Vol. 112,
No. 2. p. 402. With permission].
The model incorporates three nutrients - nitrogen,
phosphorus, and silica - each with biologically avail-
able and unavailable components, and a biomass
component. It includes five classes of algae and two
classes of zooplankton. The interaction of the com-
ponents are shown in Figure 10-3.
The model is structured in a format to simulate a
specified number of phytoplankton and zooplankton
classes. The model developers chose to use multiple
classes of phytoplankton and zooplankton to predict
the desired decline in blue-green algae. Phytoplankton
groups respond differently to zooplankton grazing and
have different nutrient requirements. Unlike the many
eutrophication models that use chlorophyll a as a sur-
rogate for phytoplankton, this model uses
phytoplankton cell biomass.
A number of mechanisms are considered in this model,
including:
• Internal nutrient pool kinetics for phosphorus,
nitrogen, and silicon.
• A reaction-diffusion mechanism for carrier-
mediated uptake of phosphorus and nitrogen that
includes luxury uptake of nutrients.
Advective transport is defined as a flow based on system hydrodynamics (modeled or measured). Dispersion transports mass
from areas of high concentration to areas of low cohcentration with no net flow of water.
10-2
-------
HIGHER
PREDATORS
1
-
CARNIVOROUS
ZOOPLANKTON
i
HERBIVOROUS
ZOOPLANKTON
)f
-x
DIATOMS
* \.._.
1 \ V-.V.-.V.--V.V? 1
GREENS
[
OTHERS ' BLUE-GREENS BLUE-GREENS
IHOHHflltlHSI INflXtHSI
f ,_L_ t I
-f
-*•
AVAILABLE
SILICON
AVAILABLE
PHOSPHORUS
j
UNAVAILABLE
SILICON
4
i
~
AVAILABLE ATMOSPHERIC
NITROGEN NITROGEN
A
UNAVAILABLE
PHOSPHORUS
1
SEDIMENT
TOTAL
SILICON
UNAVAILABLE
NITROGEN
•
SEDIMENT
TOTAL
PHOSPHORUS
SEDIMENT
TOTAL
NITROGEN
Figure 10-3. Schematic diagram of principal model compartments and interaction pathways [Bierman and Dolan (1986a)
Reprinted from ASCE Journal of Environmental Engineering, Vol. 112, No. 2. p. 403. With permission].
• Biological-chemical kinetics, included in sediment
compartments for total concentrations of phos-
phorus, nitrogen, and silicon.
• Zooplankton grazing.
• Saturation kinetics for water column nutrient
mineralization.
• Saturation kinetics for phytoplankton decomposi-
tion.
• An advective-dispersive model for transport of
chloride used to determine water exchange
among the segments.
• Wind-dependent resuspension for sediment
nutrients.
The internal nutrient pool kinetics are a noteworthy
aspect of the model because they treat cell growth as
a two-step process: 1) uptake of nutrients and 2)
biomass growth. The common approach is a one step
use of the Monod (Michaeiis-Menten) equation, where
cell growth is a direct function of external nutrient
concentrations. The internal pool kinetics allow for
accumulation of surplus internal nutrients when exter-
nal nutrient concentration is high and use of internal
stores when external nutrient concentration is low. The
recycling of nutrients is a function of the phytoplankton
losses. This more realistic approach requires greater
model complexity and additional model coefficients.
Furthermore, it exacts a severe computational burden
because all cell history must be tracked to follow ex-
posure patterns.
While phytoplankton growth is a function of nutrient
kinetics, phytoplankton loss mechanisms include
respiration, decomposition, sinking, and zooplankton
grazing. Respiration loss is a temperature-dependent,
first-order decay term. Microbial decomposition is a
temperature-dependent, second-order decay term
proportional to total phytoplankton concentration and
specific growth rate. Sinking loss is set at a constant
velocity for each phytoplankton class. Zooplankton
10-3
-------
grazing loss is a temperature-dependent, two-cbm-
ponent loss mechanism. It was included for diatbms,
greens, and "other" phytoplankton, but not for blue-
greens. The zooplankton response function included
losses to higher-order predators. A constant "refuge
concentration" is specified for both phytoplankton|and
zooplankton below which there is no grazing or preda-
tion. l
10.3.2. Model Inputs
The complexity of the model required many
parameters and boundary conditions. Model coeffi-
cients are defined in Table 10-1 and values summarized
In Table 10-2 (a-c) to provide the reader a sense o|f the
model complexity.
Each phytoplankton group was characterized by a
maximum growth rate, a temperature growth adjust-
ment factor, and a saturation light intensity. Qther
phytoplankton coefficients included respiration |rate,
decomposition rate, sinking rate, and conversion ifates
of nutrient forms (from unavailable to available).
Zooplankton kinetic coefficients, taken from literature
or data collected for this study, included assimilation
efficiency, maximum ingestion rate, and phytoplarjkton
preference factor. Growth and death rates were es-
timated or calibrated to field data. Coefficients were
assigned for each of the two functional groups of
zooplankton: fast ingesters and slow ingesters.
Nutrient uptake and cell growth were treated ip the
model as separate processes, but have parallel sets of
equations and coefficients. Nutrient kinetics for phos-
phorus and nitrogen depended on the variables of
percent dry weight and minimum cell quotas for these
nutrients. Minimum concentrations were also assigned
for external nutrients, which corresponded to the| min-
imum levels to which the phytoplankton could deplete
the environment. External and internal half-saturation
levels were specified for both processes. The latter
were set equal to the minimum cell quotas. Maximum
phosphorus and nitrogen uptake rates were the same
for all phytoplankton groups. Silica coefficients Iwere
specified only for diatoms. !
Another Important assumption of the model was the
partitioning of phosphorus into available and un|avail-
able components. Dissolved ortho-phosphorus was
considered to be available for immediate uptake by
phytoplankton. Unavailable phosphorus! was
equivalent to total phosphorus minus dissolved ortho-
phosphorus. For scenarios discussed below, available
and unavailable phosphorus ratios were estimated for
point and nonpoint sources. The effective ratio of avail-
able to total phosphorus for point sources at the
Saglnaw River mouth was assumed to be 34 percent.
It was also assumed that the ratio did not change with
different treatment levels.
Environmental forcing functions varied for each year
and included water temperature, incident solar radia-
tion, pollutant and tributary loadings, boundary condi-
tions, and water transport rates. They were determined
independently of the model and supplied as input.
Table 10-3 is a summary of selected examples of these
inputs designed to provide the reader a sense of the
range of values. These environmental factors were
supplied to the model as time series input. Water
transport rates were obtained from a separate time-
variable model of a conservative tracer, chloride
[Richardson (1974)].
10.3.3. Calibration/Verification
The approach to calibration was to match general
trends of the seasonal changes in the data and obtain
model output within one standard deviation of the
mean value of the observed data for each cruise. When
this was not achieved, model coefficients were ad-
justed to best approximate the peak concentrations. A
Student's t-test was used to compare mean values
from field data and the model.
The first test of the model was to visually compare the
model calculations with the observed data for each of
the model segments. Figure 10-4 presents the phos-
phorus calculations. As seen here, the simulation of
trends is reasonable, but variability in the data and
model discrepancies do exist. This may be caused by
short term variation not considered in the model or
other factors. As an additional test, statistical analyses
were performed.
The results of the statistical analysis are presented in
Table 10-4 as percent of sampling cruises in which
predicted and observed means were not significantly
different at a 95 percent confidence level. Segments 1
and 3 had the lowest scores, but represent only 3.5 and
5.0 percent, respectively, of the total volume of the Bay.
Also these are shoreline segments most influenced by
changes in wind and tributary loading. Because seg-
ments 1 and 3 represent a small percentage of the total
area, they were not emphasized in the calibration.
The model did a good job at matching total phos-
phorus despite large differences in total phosphorus
concentrations among segments. The model was less
effective in simulating the dissolved available phos-
phorus. Overall, the calibration resulted in ap-
proximately 86 percent of the model output being not
significantly different than the field data for the thirteen
principal variables.
10-4
-------
Table 10-1.
Description of Model Coefficients [Bierman and Dolan (1981)].
FACT phytoplankton cell size in mg dry wt/cell
f(L) , phytoplankton light reduction factor
(dimensionless)
f(T) phytoplankton temperature reduction factor
(dimensionless)
Ke light extinction coefficient in meter'1
KNCELL intracellular half-saturation constant for nitrogen-
dependent growth in moles N/cell
KPCELL intracellular half-saturation constant for phosphorus-
dependent growth in moles P/cell
KSCM half-saturation constant for silicon-dependent
growth of diatoms in moles Si/L
KZSATk half-saturation concentration of phytoplankton
for grazing by zooplankton k in mg/L
P, N actual moles of phosphorus (nitrogen) per phyto-
plankton cell
PCA, NCA intracellular available phosphorus (nitrogen)
concentrations in moles/liter cell volume
PCAMIN, minimum intracellular concentrations, correspond-
NCAMIN ing to PSAMIN and NSAMIN, respectively, for avail-
able phosphorus (nitrogen) in moles/liter cell
volume
PCM, NCM concentrations of available nutrients (phosphorus,
SCM nitrogen, silicon) in water column in moles/L
PDETH maximum predatory death rate for zooplankton in
liter/mg-day
PHOTO photoperiod (dimensionless)
PKI, NKI affinity coefficient for phosphorus (nitrogen) uptake
mechanism in liters/mole
PO, NO . minimum cell quota of phosphorus (nitrogen) per
phytoplankton cell in moles/cell
PSA, NSA actual total phosphorus (nitrogen) in phytoplankton
cells in moles/mg dry wt
PSAMIN, minimum quota of phosphorus (nitrogen) in
NSAMIN phytoplankton cells in moles/mg dry wt
PSATh saturation concentration of zooplankton k above
which predatory death rate remains constant, in
mg/L
Q water circulation rate in volume/day
RIPM, RINM maximum phosphorus (nitrogen) uptake rate in
day "1
RADINC incident solar radiation in langleys/day
RADSAT saturation light intensity for phytoplankton growth in
langleys/day
RAGRZDi rate at which phytoplankton I is ingested (grazed) by
zooplankton in mg A/liter day
RAMAX phytoplankton maximum growth rate at 20 C in day'1
RLYS
RRESP
RTOP,
RTON,
RTOS
RZ
RZMAX
RZPEX,
RZNEX,
RZSEX
SPGR
SSA
T
CROP
TOP, TON,
TOS
TOPSNK,
TONSNK,
TOSSNK
V
WPCM,
WNCM,
WSCM
WTOP,
WTON,
WTOS
Z
ZASSIM
ZEFFkl
ZDETH
ZKDUM
ZSAFE
phytoplankton decomposition rate in liter/mg day
phytoplankton respiration rate in day"1
rates of transformation from unavailable nutrient
forms (phosphorus, nitrogen, silicon) to available
forms in day "1
zooplankton specific growth rate in day"1
zooplankton maximum ingestion rate in day'1
nutrient (phosphorus, nitrogen, silicon) excretion
by zooplankton to unavailable nutrient pool in
moles/mg zooplankton-day
phytoplankton specific growth rate in day "1
silicon composition of diatoms in moles/mg dry wt
temperature in C
total phytoplankton concentration in mg dry wt/L
concentration of unavailable nutrient forms (phos-
phorus, nitrogen, silicon) in moles/L
sinking rates of unavailable nutrient forms (phos-
phorus, nitrogen, silicon) in meters/day
inner bay volume in liters
external loading rates of available nutrients (phos-
phorus, nitrogen, silicon) in moles/day
external loading rates of unavailable nutrients
(phosphorus, nitrogen, silicon) in moles/day
zooplankton concentration in mg dry wt/L
zooplankton assimilation efficiency
(dimensionless)
ingestion efficiency of zooplankton k for phyto-
plankton I (dimensionless)
specific zooplankton death rate in day"1
effective half-saturation concentration of total
phytoplankton for grazing by zooplankton
refuge concentration of zooplankton below which
predatory grazing does not occur, in mg/L
NOTE:
The addition of the suffix "BD" to a variable refers
to the boundary value of the variable.
10-5
-------
Table 10-2. Summary of Selected Mode! Coefficients [Bierman and Dolan (1 981)].
a. Summary of Phytoplankton Coefficients ;
Coaf.
R1PM
PK1
PO
CONCP
KPCELL
RINM
NKI
NO
CONCN
KNCELL
SSA
KSCM
RAMAX
ASINK
RLYS
FACT
RAOSAT
RRESP
Units
day1
liters/mole
mole P/cell
mole P/cell
da/1
liters/mole
mole N/Cell
mole N/cell
mole Si/mg
mole Si/liter
da/1
meter/day
liters/mg/day
mg/cell
langleys/day
da/1
b. Summary of Zooplankton
Coof.
RZMAX
ZASSIM
KZSATk
AZMIN
BDETH
PDETH
2SAFE
PSATk
Unit
da/1
mg/liter
mg/liter
da/1
da/1
mg/liter
mg/liter
Diatoms
0.500
0.518x106
0.724X10'13
0.250x106
0.724x1 0'13
0.125
0.100x107
0801 X10'11
0.208X107
0.801X10'11
0.334X10'5
0.357X1 0'5
1.6
0.05
0.004
0.450x10-s
100
0.05
Coefficients
Faster
Ingester
0.70
0.60
1.0
0.20
0.05
0.50
0.01
1.0
Greens
o.5oo :
0.167xl07
0.312X-J014
0.250S106
0.312XJO'14
0.125
O.IOOxlO7
0.345x1 0'12
0.208x1 07
0.345x1 0'12
1.4
0.05
0.004 !
0.194x;10"6
100 i
0.05
Others
0.500
0.1 58x 10s
0.148x10"13
0.250x1 06
0.148X10'13
0.125
0.1 00x1 07
0.163X10"11
0.208x1 07
0.163x10'11
1.2
0.05
0.004
0.918x10"6
100
0.05
Blue-Greens
(non-Ng)
0.500
0.200X 107
0.566X1014
0.356x106
0.566x1 0'14
0.125
0.100X107
0.438X10'12
0.208X107
0.438x1 0'12
1.0
0.05
0.012
0.246. 10"6
50
0.05
c. Summary of Coefficients for
Slow
Ingester
0.10
0.60
1.0 ,
0.20
0.01
6.10
0.01
1.0
Coefficient
RTOP, RTON,
RTOS, TOPSNK,
TONSNK
Units
da/1
meters/day
Blue-Greens
(N2-Fixing)
0.500
0.518x106
0.488x1014
0.356x1 08
0.488X10'14
0.125
0.100x107
0.377X10'12
0.208X107
0.377X10'12
0.70
0.05
0.012
0.212x10'6
50
0.05
Unavailable Nutrients
Value
0.005
0.05
10.3.4. Projections '
In a report to the International Joint Commission (Bier-
man and Dolan, 1980) the model was applied to seven
scenarios of phosphorus loadings. The scenarios con-
sisted of various combinations of advanced 'was-
tewater treatment and non-point source reduction. The
results were presented as annual average total phos-
phorus concentration, total phytoplankton biorjnass,
total blue-green phytoplankton biomass, and taste and
odor In the municipal water supply. Although the Bay
was partitioned into five segments, only two contrast-
Ing segments (segments 2 and 4) were analyzed. Seg-
ment 2 contained 73 percent of the total water volume
of the inner Bay and was the most degraded portjon of
the Bay. Segment 4 had the highest water quality in the
Bay. These segments represented the two extremes in
the Bay.
In the model simulations, peak total biomass con-
centrations did not change significantly with reduc-
tions in phosphorus loads; however, the blue-green
phytoplankton responded in direct proportion to phos-
phorus reduction in segment 2 and in a lower propor-
tion in segment 4. This was the first objective of nutrient
control in the Bay.* This simulation of algal species
change is a unique aspect of this multi-class
phytoplankton model. The model has the ability to
distinguish nutrient limitation among different types of
phytoplankton and hence allows changes in composi-
tion.
In general, the model showed phytoplankton growth to
be nitrogen-limited, but for a two month period (May
and June) diatoms were silica-limited. This agreed with
actual observations of nutrient depletion. In mid-
August, the nitrogen-fixing blue-greens capitalized on
the depletion of nitrogen and proliferated. Their growth
was then restricted by phosphorus limitation.
Later application of the model to 1980 data in a post-^udit shows the blue-greens actually responded in a much greater
proportion to phosphorus reduction. i
10-6
-------
100
75
50
25
_j
«^
^ 100
o 75
l-
< 50
u
8 100
75
50
25
0
III
SEGMENT 1
SEGMENT 2
I I I I
SEGMENT 3
J F M A M J J A S^T
SAGINAW BAY 1974
TOTAL PHOSPHORUS
SEGMENT 4
SEGMENT 5
J'F M'A M j1 •TA'S'O'N'D
01
100
75
50
25
100
O 75
'<
o:
IL)
o
z
o
o
50
25
75
50
25
0
I—i—«—I—i—I—r~i—i—r
SEGMENT 2
SEGMENT 3
I"..I .I..I
I.I.I
I ..I
J'F'M'A'M'J'J'A'S'O'N'D
SAGINAW BAY 1975
TOTAL PHOSPHORUS
SEGMENT A
i i i i—r—i—r—r
T
SEGMENT 5
J'F'M'A'M'J'J'A'S'O'N'D
Figure 10-4. Model output and field data comparison for total phosphorus (solid line is model output; data are sampling
cruise means and three standard deviations) [Bierman and Dolan (1986a). Reprinted from ASCE Journal of
Environmental Engineering, Vol. 112, No. 2. p. 409. With permission].
10-7
-------
Table 10-3. Summary of Selected Model Inputs [Biermah and Dolan (1986b). Reprinted from ASCE Journal of Environmental
Engineering, Vol. 112, No. 2. p. 422. With permission].
Parameter
Sample Year
1974 1975
1980
Saglnaw River Loadings (Metric Tons): '
Total
Total
Total
Environmental
Phosphorus ;
Nitrogen ;
Silicon
Forcing Factors:
Number of Days Where Wind Speed Exceeded :
Threshold for Resuspension (Annual)
1266 1470
14,100 15,290
23,000 31 ,000
29 40
493
11,030
12,250
35
Annual Average Water Temperature (°C)
Segment 2
Segment 4
12.0 13.9
9.8 11.1
14.9
12.8
10.4. Post-Audit
In 1980, a survey was conducted and used in a post-
audit of the model. A post-audit provides a test of the
model for use in projections by comparing forecasts to
actual observations. Environmental conditions
changed substantially in five years. From 1975 to 1980,
total annual load of total phosphorus decreased 66
percent and available phosphorus decreased 78| per-
cent. It was estimated that 44 percent of the drop in
phosphorus load was because of decreases in
tributary flow. The other 56 percent was attributed to
point source controls and a detergent phosphorus ban
for the State of Michigan initiated in 1977. Total
phytoplankton biomass also decreased substantially,
with the nitrogen-fixing blue-greens being njearly
eliminated.
The predictive capability of the model was tested using
the 1980 data. The model was rerun using the 1974 and
1975 model coefficients but loading and environmental
conditions for 1980. The results are presented!as a
comparison of predicted and observed percent reduc-
tions between the 1974-75 calibration years and the
1980 resurvey year (Figure 10-5). In general, the rpodel
overestimated the percent reduction in total phos-
phorus, and underestimated reductions in diatomls and
blue-green algae.
Table 10-4. Statistical Comparison between Model Results
and Field Data [Bierman and Dolan (1986b)].
Year
1974
1975
1980
Cell* !
1 2 3
72 85 65
64 80 72
57 64 52
4 i5
88 87
86 87
85 86
* Percent of sampling cruises for which computed mean vjalues
not significantly different from observed mean values at 95%
confidence level; average of 13 variables.
Underestimation of phosphorus concentrations was a
characteristic of model results during the calibration
years and in the post-audit survey. This discrepancy
was attributed to the underestimation of wind-driven
resuspension of sediments. Nevertheless, the model's
prediction of elimination of threshold odor violations at
the water treatment plant agreed with the data. This
was the primary management need for the model.
Blue-green phytoplankton biomass in segment 4 was
correlated with threshold odor in the drinking water
intake. The model predictions for threshold odor viola-
tions in the drinking water intake agreed with observa-
tions because both were below the blue-green
biomass threshold.
Overall, the model predictions did not match observed
concentrations closely, but were consistent with ob-
served trends. The model correctly predicted that if
phosphorus loadings were reduced to 400-500 metric
ton/year, blue-green algae would decrease more than
other species and threshold odor would be eliminated.
The response of the blue-greens exceeded the predic-
tion of the model in absolute values.
10.5. References
Ayers., J.C., Anderson, D.V., Chandler, D.C., and G.H.
Lauff, 1956. Currents and water masses of Lake Huron,
(1954) Synoptic Surveys. Ontario Dept. Lands and
Forests, Division of Research and University of
Michigan, Great Lakes Res. Inst. (Referenced in
Richardson, 1974).
Bierman, V.J. and D.M. Dolan. 1980. Responses of
Saginaw Bay, Lake Huron, to reductions in phosphorus
loadings from Saginaw River. Report prepared for the
International Joint Commission.
Bierman, V.J. and D.M. Dolan. 1981. Modeling of
phytoplankton- nutrient dynamics in Saginaw Bay,
Lake Huron. J. Great Lakes Res. 7(4): 409-439.
10-8
-------
100
80
<
z in
u i:
o _j
CC UI
^ 60
40
20
SAGINAW BAY
AVERAGE OF RESULTS
FOR SEGMENTS 2 AND 4
73 z
64
•
<
0
t-
UJ
m
_j
g
<
O
s
ce
0.
3lJ
%
•-"»
'"'•'
\~
^i *
/*_'
T
|
-* x
IE
f
5"
^
o
1 1
T"|
'•Jr.
V
10 O IOO
95 T
8ZT
DT_
j \
'1 J
-« N
T
'
••2
^
v\
^ J
?
"«
"•?,
-
14 ^I-
v*"^
/.
4
ft
''!'-
IF
t*
A
^
•^
-.••^
, -1
••-
"t
^f
^
^•-
f
• *?
\'
~
;-;\
^
' s'
^
•ff
"• -.
%^
-;
x'
r>
*'
'
;'
PHOSPHORUS SPRING FALL
LOADS TOTAL PHOSPHORUS
CONCENTRATION
SPRING FALL THRESHOLD
DIATOMS BLUE GREENS OOOR
VIOLATIONS
Figure 10-5. Changes in water quality constituents between 1974 and 1980 in segments 2 and 4 [Bierman and Dolan (1986b)
Reprinted from ASCE Journal of Environmental Engineering, Vol. 112, No. 2. p. 409. With permission].
Bierman, V.J. and D.M. Dolan. 1986. Modeling of
phytoplankton in Saginaw Bay: I. Calibration Phase. J.
Env. Eng., ASCE 112(2): 400-414.
Bierman, V.J. and D.M. Dolan. 1986. Modeling of
phytoplankton in Saginaw Bay: II. Post-Audit Phase. J;
Env. Eng., ASCE 112(2): 415-429.
Bierman, V.J., Dolan, D.M., Stoermer, E.F., Gannon,
J.E., and V.E. Smith. 1980. The development and
calibration of a spatially simplified multi-class
phytoplankton model for Saginaw Bay, Lake Huron.
Great Lakes Environmental Planning Study (GLEPS) :
Contribution No. 33.
Fish and Wildlife Service, U.S. Department of Interior.
1956. Surface current studies of Saginaw Bay and Lake
Huron. (Referenced in Richardson, 1974).
Richardson, W.L 1974. Modeling Chloride Distribution
in Saginaw Bay. Proc. 17th Conf. Great Lakes Res.
1974: 462-470. Internal. Assoc. Great Lakes Res.
10-9
-------
-------
11. Potomac Estuary Water Quality Modeling
11.1. Background
The studies discussed here include application of three
different waste load allocation-related models for the
Potomac Estuary near Washington, D.C. The three
models, although covering basically the same location,
have markedly different structures to address three
different water quality issues. The water quality con-
cerns consisted of:
• Dissolved Oxygen Depression
• Nutrient Enrichment and Algal Proliferation
• Total Residual Chlorine
These three water quality concerns each had unique
spatial and temporal considerations, such that all con-
cerns could not be properly addressed by a single
model. In this regard, three separate (but inter-related)
models were developed to specifically address each
issue. The Dynamic Estuary Model (DEM), a one-
dimensional, spatially detailed real-time dissolved
oxygen model was applied to determine effluent limita-
tions for oxygen-demanding materials. The Potomac
Eutrophication Model (PEM), a tidally-averaged
eutrophication model, was applied to determine the
impact of nutrient control strategies on regional algal
concentrations. Neleus, a real-time, two-dimensional
finite element model, was applied to determine very
localized total residual chlorine impacts and the poten-
tial for forming a barrier to fish passage.
11.2. Problem Setting
The Potomac Estuary drains an 11,560 square-mile
area, comprising portions of Maryland, Virginia, West
Virginia, and Pennsylvania. It is used for a wide variety
of activities, ranging from industrial water supply
(primarily cooling water supplies), to navigation, boat-
ing and commercial and sport fishing.
The Potomac Estuary extends 114 miles from the fall
line at Chain Bridge in Washington, D.C. to its junction
with the Chesapeake Bay (see Figure 11-1). The es-
Quantlc
Figure 11-1. Location map of Potomac Estuary [USGS (1985)].
11-1
-------
tuary can be divided into three zones: the freshwater
or tidal river zone, the transition zone, and the saline
zone. The upper reach, although tidal, contains bnly
freshwater, and extends from Chain Bridge to |ust
above Quantico. The middle zone is characterized by
a transition from fresh to brackish water and extends
from Quantico to the Highway 301 bridge. The lower
reach is highly saline, vertically stratified, and often
anoxic near the bottom. The modeling and waste load
allocation discussed herein focuses on the freshwater
zone. [
The major source of pollutants in the upper Potomac
Estuary is the District of Columbia and its suburbs.
Population in the Washington, D.C. area increased
from 2.1 million in 1960 to 3.2 million in 1980. At ikast
14 wastewater treatment plants with a combined flow
well over 500 MGD discharged into the Potomac; Es-
tuary in 1980. This discharge is a significant increase
over the 325 MGD wastewater flow in 1966. While
effluentflow has increased, the load of phosphorus and
BODs from these point sources has decreased; ap-
proximately seventy-five and fifty per cent respectively
during this period because of substantial improve-
ments in wastewater treatment.
The most significant point source discharge to j the
estuary is the Blue Plains Wastewater Treatment Plant
in Washington, D.C., which has an average annual flow
of 227 MGD. Other sources of nutrients and oxygen
demanding material to the Potomac Estuary include
nonpoint source discharges from upper basin
drainage and downstream tributaries, combined sewer
overflows, and atmospheric pollutants.
The upper portion of the Potomac Estuary has tieen
plagued with occurrences of low levels of dissolved
oxygen, floating algal mats, and high concentrations of
chlorophyll a, indicating a relatively advanced state of
eutrophication. In recent years, these problems r^ave
dramatically declined because of increased waste-
water treatment.
11.3. Dynamic Estuary Model (DEM) of
Dissolved Oxygen |
The Potomac Estuary was regularly depleted of jdis-
solved oxygen during the 1960s and early 1970s in
response to point sources of pollution and combined
sewer overflows in the Washington, D.C. area. U.S.
EPA Region III, in their "Potomac Strategy", highlighted
the need to develop and validate water quality models
for the Potomac that could be used for waste load
allocation purposes (Clark, 1982). The Potomac
Strategy State/EPA Technical Committee sub-
sequently recommended DEM as the appropriate
model to use to assess dissolved oxygen impacts in
the Upper Potomac. Their decision was, in large part,
based on the capability of the model to provide good
spatial resolution and diurnal calculations.
DEM represents the Potomac using a series of inter-
connected channels and junctions. These channels
and junctions can be arranged to simulate simple
two-dimensional features of the estuary, but are
primarily one-dimensional (i.e. no lateral variation) with
branching. DEM as configured for the Potomac ex-
tends from Chain Bridge (River Mile 0.0) as an
upstream boundary to Piney Point (River Mile 96.2) as
a downstream boundary. This configuration consists
of 133 junctions and 139 channels, but the focus of the
water quality modeling was in the upper 20 miles. DEM
simulates in "real time", meaning that the model
predicts conditions as they vary through diurnal and
tidal variations.
DEM consists of two separate but closely related
models. The first, a hydrodynamic model, simulates
both the tidal and net advective movement of water.
This model provides predictions of water depth,
velocity, and direction of flow based upon input infor-
mation on geometry, roughness, tributary inflows and
tidal variations in depth at Piney Point. The results of
the hydrodynamic model are input to the second
model, which simulates water quality.
The water quality model predicts the transport and
transformation of pollutants in the Potomac Estuary.
The model, as applied for Potomac dissolved oxygen,
simulates three variables: dissolved oxygen, car-
bonaceous biochemical oxygen demand (CBOD), and
ammonia.
Dissolved oxygen concentrations are increased by
atmospheric reaeration and algal photosynthesis, and
are decreased by oxidation of CBOD, nitrification of
ammonia, sediment oxygen demand, and algal
respiration. The model does not predict algal
photosynthesis or respiration. Instead, these values
must be input by the modeler based upon observed
data or calculations performed external to the model.
CBOD concentrations are increased by point and non-
point loadings, and are decreased by settling and
deoxygenation of CBOD. Ammonia concentrations are
increased by point and nonpoint loadings, and are
decreased by a first-order loss term defined in DEM as
nitrification.
Water quality data for model calibration and verification
consisted of both wet and dry weather surveys con-
ducted in 1965, 1966, 1967, 1968, 1970, 1977, 1978,
1979, and 1980. The Blue Plains Wastewater Treatment
Plant, the primary point source of pollutants to the river,
implemented secondary treatment in 1977.
11-2
-------
0)
E
01
q
— Modal Predictions
X Tide Table Data
2O
100
Miles Below Chain Bridge
Figure 11 -2. Potomac Estuary hydraulic calibration high water phasing — Mean Tide [Adapted from Clark (1982)].
11.3.1. Model Calibration/Verification
Calibration of DEM to the Potomac required separate
calibration of both the hydrodynamic and water quality
submodels. Hydrodynamic calibration focused on the
channel roughness coefficient to best describe the
magnitude and phasing of predicted tides. The model
was calibrated using mean upstream freshwater flow
(11,000 cfs) and elevation data published in the Nation-
al Oceanographic and Atmospheric Administration
(NOAA) Tide Tables. Sample model calibration results
are shown in Figure 11 -2. The hydrodynamic submodel
was then verified to observed data from the periods
January 11-13, 1971 and July 22-28, 1981. The
calibrated roughness coefficients accurately
reproduced tidal range and phasing for all data sets.
The water quality submodel calibration was divided
into two separate tasks: 1) calibration of dispersion
(using conservative tracers) and 2) calibration of reac-
tion rate coefficients (using water quality concentra-
tions). The dispersive transport coefficient was
calibrated to chloride data collected during the period
August 1 to September 8,1977, and verified to chloride
data from the period September 15 to November 12,
1969. The model predicted the majority of data quite
well, but was unable to simulate the steepest portion of
the chloride gradient due to numerical dispersion (Fig-
ure 11-3). The dispersion rates determined through
calibration and verification of the chloride data were
also tested against a 1978 dye survey. The model was
able to simulate observed far-field data quite well, with
discrepancies in near-field embayments.
Water quality data for model calibration of reaction
kinetics consisted of surveys conducted in 1965,1966,
1967,1968,1970,1977,1978, and 1979. The objective
of the calibration procedure was to simulate as many
data sets as possible and to provide a test of the
model's ability to duplicate a wide range of conditions.
Model calibration (coefficient adjustment) was con-
ducted on data sets through 1977, with the later data
sets used for verification (without coefficient adjust-
ment). The data sets from 1965 to 1970 were collected
during periods of relatively constant environmental
conditions and used for steady-state model com-
parisons. The 1977 data set was collected over a two
month period characterized by a massive algal bloom
(100-300 ug/l chlorophyll a) and die-off, and used a
real-time model to characterize the significant transient
processes. Example model calibration to data is shown
in Figures 11-4 to 11-6 for the parameters ammonia,
nitrate + nitrite, and dissolved oxygen. Comparisons of
BOD were not provided because algae complicated its
measurement and comparisons. The model generally
reproduced trends in observed data quite well and was
also very successful in matching 1978 and 1979 data
during model validation.
11-3
-------
35
|g
10
Miles Below Chain Bridge
Figure 11-3. Potomac Estuary chloride verification, time period: September-October, 1969 [Adapted from Clark (1982)].
n
z
3
2.8 -
2.6 -
2.4 -
2.2 -
2 -
1.8 -
1.6 -
1.4 -
1.2 -
1 -
0.8 -
0.6 -
0.4 -
0.2 -
0
— Predicted
O Max. Observed
X Win. Observed
T
T
—7—
16
! 12
Miles Below Chain Bridge
T~
20
24
28
32
Figure 11-4. DEM calibration results for ammonia, time period: August 31-September 16,1965 [Adapted from Clark (1982)].
11-4
-------
I
+
CM
O
1.7 - -
1.6 -
1.5 -
1.4 -
1.3 -
1.2 -
1.1 -
1 -
0.9 -
0.8 -
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 ->
0.1 -=.1
0 --
FLOW
— Predicted
* Max. Observed
X Min. Observed
12
—i—
16
—r—
20
24
28
32
Miles Below Chain Bridge
Figure 11-5. DEM calibration results for nitrate-h nitrite, time period: August 31-September 16,1965 [Adapted from Clark (1982)].
o
Q
12
11
10
9
B
7
6
5 •
3 -
2 -
1 -
0
Predicted
O Max. Observed
X Mm. Observed
X
X. X_
-ii 1 1 1 r
12 16 20
Miles Below Chain Bridge
24
1 1 T
28
32
Figure 11 -6. DEM calibration results for dissolved oxygen, time period: August 31 -September 16,1965
[Adapted from Clark (1982)].
11-5
-------
11.3.2. Model Application
Application of DEM was conducted over the course of
several years and modeling efforts. Initial waste load
allocation projections were made by U.S. EPA (Clark,
1982). A revised and updated examination was per-
formed in 1984, but recommendations from this effort
were deferred when data from the mid-1980's ap-
peared inconsistent with model predictions (MWCOG,
1987). The model was then revalidated in 1987 to mpre
recent water quality data, and new waste load alloca-
tion projections performed. ,
DEM was applied by Greeley and Hansen (1984)|as
part of the Washington D.C. Blue Plains Feasibility
Study, to detemnine regional capacity treatment needs
and establish allowable effluent loads for dischargers
to the Upper Estuary. Numerous alternatives were [ex-
amined for water quality compliance and other factors.
Seven final regional wastewater treatment scenarios
were evaluated for their ability to lead to compliance
with water quality standards for dissolved oxygen.
Model projections were made at critical environmental
conditions consisting of drought (7Q10) freshwater
flow and a water temperature of 28°C, the upper 90th
percentile temperature at summer low flow. Model
coefficients were based on the average of post-1977
simulations. Algal productivity and respiration inputs
were derived from drought flow simulations using |the
Potomac Eutrophication Model (see later discussion).
Sediment oxygen demand (SOD) was proportionately
reduced with loadings toward background values.
Boundary concentrations were representative of Ithe
period 1977-1979.
DEM model results for both daily averaged and daily
minimum dissolved oxygen indicated that all final alter-
natives evaluated would lead to compliance with dis-
solved oxygen standards for the critical conditions
scenario. Water quality differences between scenarios
were viewed as small in comparison to the substantial
differences in cost. The recommended treatment
scenario was subsequently based upon cost, ien-
gineering and other considerations. !
11.3.3. DEM Post Audit
State and Federal regulators originally rejected the
OEM-based waste load allocation recommendations,
due primarily to a review of 1982-1985 dissolved
oxygen data from the Upper Potomac. These data
indicated that dissolved oxygen standards violatipns
were still occurring, even though treatment plants were
performing at recommended levels. Given that DEM
predicted that additional nitrification treatment at two
area POTWs would improve minimum dissolved
oxygen concentrations by 0.8 mg/l, they recom-
mended nitrification treatment at these plants.
Local governments expressed considerable reserva-
tion regarding the need for improved treatment, and
conducted a study to revisitthe DEM modeling analysis
and examine regulatory agency concerns (MWCOG,
1987). Extensive water quality surveys were conducted
in the Upper Potomac in 1986 to validate (or refute) the
predictive capability of DEM. In addition, special
studies were conducted investigating current pollutant
decay rates, sediment oxygen demand, and occur-
rence and cause of water quality standard violations.
Limno-Tech (1987) applied DEM to simulate 1985 and
1986 conditions. This analysis determined that DEM
calculations of dissolved oxygen were very sensitive
(±3 mg/l) to algal-productivity related parameters
which were not directly measured. Given judicious
selection of inputs, DEM could simulate recent dis-
solved oxygen data. Since neither observed (nor
eutrophication model predicted) algal productivity in-
formation was available, DEM predictions could not be
explicitly confirmed or refuted. An important outcome
of this analysis was that transient changes in algal
productivity could be responsible for dissolved oxygen
standards violations, irrespective of point source im-
pacts. Furthermore, detailed examination of DEM indi-
cated that it over-calculated the benefits from
additional nitrification treatment because it simplistical-
ly assumed all ammonia loss was due to nitrification.
The ammonia mass balance is a net combination of
nitrification, algal uptake of ammonia, sediment am-
monia release, and hydrolysis of organic nitrogen.
Re-evaluation indicated a reduced nitrification rate and
a benefit due to additional nitrification treatment of 0.2
to 0.5 mg/l.
As a result of these findings, the dominance of net algal
productivity and the small benefits from additional
nitrification treatment, further nitrification treatment re-
quirements were deferred.
11.4. Potomac Eutrophication Model (PEM)
The Potomac Estuary began exhibiting signs of
eutrophication (algal blooms, floating mats of vegeta-
tion) in the late 1940s and continued through the 1960s.
In an effort to control these problems, point source
discharges of total phosphorus to the estuary were
reduced by seventy-five percent over the period 1968
to 1979. However, algal bloom conditions persisted
into the late 1970's, causing concern as to whether the
decrease in point source phosphorus was controlling
eutrophication. The Potomac Eutrophication Model
(PEM) was developed to determine the impact of his-
torical pollution controls on Potomac Estuary
eutrophication, and to guide regulators in setting future
effluent limitations.
The PEM model was developed because the existing
DEM model focused more on spatial resolution than on
11-6
-------
the kinetic complexities of eutrophication which were
necessary to forecast the benefits of nutrient controls.
In addition, the tidally averaged and large segment
approach of PEM is more consistent with the regional
and seasonal focus of eutrophication. PEM is a version
of the EPA supported Water Quality Analysis Simula-
tion Program (WASP), but developed specifically for
the Potomac (Hydroqual, 1982). Compartment or box
modeling techniques are used to represent the estuary
as a series of water column and sediment segments.
There is no hydrodynamic submodel included in PEM.
Average flows, velocities, and dispersion coefficients
are not computed by the model; they are specified as
model inputs. The hydrodynamic inputs are tidally
averaged and reflect seasonal changes, not daily or
hourly changes. The kinetic equations employed in
PEM link phytoplankton growth and death to non-linear
nutrient interactions and recycle mechanisms, directly
couple phytoplankton to dissolved oxygen concentra-
tions, and internally compute sediment nutrient release
and oxygen demand. The following state variables are
included in PEM:
• Chlorides
• Phytoplankton carbon
• Total organic nitrogen
• Ammonia nitrogen
• Nitrite-nitrate nitrogen
» Dissolved and paniculate organic phosphorus
• Dissolved and paniculate inorganic phosphorus
• Carbonaceous biochemical oxygen demand
• Dissolved oxygen
PEM computes water column concentrations on a
daily basis. The focus in calibrating the model was on
matching monthly and annual trends over a regional
scale of 75-100 miles. Such spatial and temporal scales
represent the global response of the estuary to
seasonally transient nonpoint source inputs from the
upper Potomac Basin and tributaries, and point sour-
ces from wastewater treatment plants.
The PEM network consists of 23 main channel seg-
ments and 15 tidal embayment segments, each with a
sediment layer segment below. These segments range
in length from one to two miles in the upper tidal
freshwater portion of the estuary, to 10-15 miles in the
lower, saline portion of the estuary. The focus of the
modeling was on the freshwater segments.
11.4.1. Model Calibration/Verification
Historical data from several sources were used for both
the calibration and verification of PEM. Data sets were
selected that provided spatial coverage of at least the
upper 50 miles of the estuary on a biweekly or monthly
basis for the crucial summer period, and that included
simultaneous measurements of chlorophyll a,
nutrients, and dissolved oxygen. Data from different
sources were often combined to produce a more
robust characterization of the estuary. The data sets
generally had biweekly sampling during the warm
weather season at stations 1 to 2 miles apart in the
freshwater portions of the upper estuary. Data col-
lected during 1966, and 1968 through 1970 were used
in the calibration, and are representative of water
quality conditions prior to the implementation of phos-
phorus removal at the major sewage treatment plants
along the estuary.
USGS data from the years 1977 through 1979 were
used to verify PEM. These years were selected be-
cause they offered the chance to study the changes in
the estuary after institution of phosphorus removal at
;Blue Plains. Thus, the verification period provided an
opportunity to further test the model's ability to simu-
late the eutrophication process in the Potomac Es-
tuary.
The verification data set involved short, intensive week-
long surveys in 1977 and 1978. The entire length of the
estuary was usually sampled twice during the 1977 and
1978 surveys, with vertical samples collected at a num-
ber of stations. In 1979, the spatial and temporal
coverage was reduced, and sampling was limited to
twice a week at five major stations.
11.4.2. Environmental Inputs
The PEM application for 1966 to 1979 required exten-
sive inputs for environmental conditions including
flows, loads, and boundary conditions which are sum-
marized below.
PEM does not include a hydrodynamic submodel, so
flows must be input for each segment of the model. To
simplify model input during the calibration period, only
the two major and dominant sources of freshwater flow
were included, the Potomac River at Little Falls (the
upstream boundary) and the Blue Plains Wastewater
Treatment plant effluent. Downstream tributary flows
and other treatment plant discharges were deemed
minimal. Both upstream freshwater flows and Blue
Plains effluent flows were input to the model using
piece-wise linear approximations of seasonal flow pat-
terns, not actual day-to-day fluctuation. For model
verification, the model also included flows for the
Anacostia River and Occoquan Reservoir. These flows
were insignificant during the extreme drought of the
calibration period, but were of sufficient magnitude
during verification that they had to be considered.
11-7
-------
i
Pollutant loads to the Potomac were divided into trpee
categories: 1) Point Sources, 2) Combined Setoer
Overflows, and 3) Nonpoint Sources. Point source
Inputs of pollutants were defined by monitoring data
and daily operating reports from the area's municipal
wastewater treatment plants. The Blue Plains treatment
facility accounted forthe large majority of these inputs.
In addition to permitted outfalls, an unregulated "gap"
In a major sewer line contributed approximately 6 MGD
of raw sewage until closed in July 1972. Estimates of
monthly averaged combined sewer overflow pollutant
loadings for Washington D.C. were generated wijh a
SWMM model simulation of the D.C. sewer network.
Combined sewer overflows for Alexandria were- es-
timated based on calculated stormwater runoff anq the
average CSO concentrations measured in the D.C.
sewer system.
Nonpoint source loads to the estuary were estimated
for all tributaries to the main stem of the Upper
Potomac Estuary. The nonpoint source flow for 6ach
tributary was based on data from USGS gagingista-
tions. Estimates of flow for ungaged tributaries were
based on the gaged discharge in neighboring
tributaries. Seasonal flow trends were defined for each
year by smoothing out many of the small peak flows
using linear approximations. Water quality concentra-
tions associated with nonpoint runoff were based on
predictions of the Nonpoint Source (NPS) model.
Simulated daily flows and pollutant loadings from 11977
to 1979 were analyzed, and a mean concentratioh for
each of three flow ranges were determined and used
In the model inputs. Slight reductions in concentrations
were used for the 1960's simulations to reflect the| less
developed land use.
714.3. Boundary Conditions
Model inputs for upstream boundary conditions yvere
based on data but required considerable extrapolation
and Interpolation to simulate the several years of icon-
ditions. Available data were statistically analyzed and
correlated to flow. Where applicable, relationships
were used between pollutant concentration ;(e.g.
nitrate) and flow; otherwise, average concentrations
were matched to observed USGS flow. All inputs jwere
smoothed to characterize seasonal trends, not day to
day transients.
11.4A. Calibration
The model calibration included reaction ratejs for
phytoplankton growth, nitrogen and phosphorite cy-
cling, and the distribution of CBOD and dissolved
oxygen. Calibration was accomplished by varying rate
coefficients until a satisfactory fit was obtained be-
tween the predicted and observed water qualityidata.
Model coefficients were identical for all calibration sur-
veys. External inputs such as flow, temperature, solar
radiation, and light extinction coefficients were as
measured during the surveys.
Figures 11-7 to 11-10 show predicted and observed
water quality data. These figures present calibration
results for chlorides, chlorophyll a, DO, BODs, total
organic phosphorus, total inorganic phosphorus, am-
monia nitrogen, and nitrite-nitrate nitrogen during May
and September of 1966, the year with the lowest
recorded flow. The model predicted the overall varia-
tion in the data well. Of particular note is the chloride
calibration, which validated water transport. Other
calibration runs were similar.
11.4.5. Verification
Initial verification used 1977-1979 environmental con-
ditions and the model coefficients derived during
calibration. Some of the calibrated coefficients had to
be modified for the verification period to reflect im-
proved treatment and the altered settling charac-
teristics of inorganic phosphorus. These changes
included the relocation of the Blue Plains outfall and
the use of ferric chloride to precipitate phosphorus. To
account forthe altered settling characteristics, a spatial
settling function was developed that was unique to the
verification. The instream nitrification rate and the
oxidation rate for carbonaceous BOD were also
changed to reflect improved treatment levels.
Predicted and observed water quality are compared in
Figures 11 -11 and 11 -12, which illustrate the July 1977
PEM verification for chlorides, chlorophyll a, dissolved
inorganic phosphorus, total phosphorus, ammonia
nitrogen, nitrite-nitrate nitrogen, BODs and DO. Similar
results were attained for other surveys.
11.4.6. Statistical Assessment of Validation
In addition to the graphical comparisons, statistical
measurements of goodness-of-fit tested the adequacy
of PEM for future predictions. The three statistical pro-
cedures used in the PEM study are:
• Regression analyses
• Relative error
• Comparison of means.
In regression analyses, the calculated values from the
model are compared to the observed values, and a
number of standard statistics computed, including the
correlation coefficient and the standard error of the
estimate. Table 11 -1 shows that 73 to 88 percent of the
variability in the observed chlorophyll a data and 60 to
93 percent of the variability in the observed dissolved
oxygen are explained by the model.
11-8
-------
• Nay 26. June 1
— May 30
40 60
RIVER MILE
2SO
200
ISO
^ 100-
5. '
" 50-
d
_i
X
• Hoy 26, June 1
— Hay JO
*
• * j ' 1 I, |
• i i i . . , ,-,- iTpTTr, , , , ,
• Sept 7, Sept 8
— Sept 7
20
40 60
RIVER MILE
Figure 11 -7. PEM calibration for chlorides and chlorophyll a, May and September, 1966 [Hydroqual (1982)].
• Hey 26, June 1
— "ay 30
20
40 6O
RIVER MILE
• Hay 26. June 1
— Hey 30
4O
RIVER MILE
I I I I
80 100
Figure 11-8. PEM calibration for BODS and dissolved oxygen, May and September, 1966 [Hydroqual (1982)].
11-9
-------
1.00-
-> O.80-
£
j= 0.60-
f/1
| 0.40-
X
a.
Z
y o.oo
o
g 1.00-
| oJ
0.6O-
a40
02O
o.oo
0.50-
. M.y 26. June Z
~~N*yI° ___ 0.2O-
f 0.90-
': V)
g 0.60
o
• x
th 030-
.y>^-^__ i s
-^x • T->- . : ' '• — : i— °- „„„
i i i i i i i i i i i ' i" a °-°0
11 -z
o i 5Q-
i CC ••**'
- Sept 7, Sept S O
-Sept7 3 1.20-
t-
*- O.90
1 O.60
rl\» ' t : °-30
^"n^Lj^xj-i : : J
0 ' ' ' 20 " ' ' 40 ' ' ' 60 80 '100
RIVER MILE
• HIV 26. June 2
— H«y 30
^
1 1 1 | 1 ' I | ' | 1 I
. . • sept 7, sept 8
A '. — Sept 7
•\ t .
r
A ^~^-.. !rJ = ~~ ' ~7
0 ' ' ' 20 ' ' ' 40 ' 60 80 100
RIVER MILE
Flauro 11-9. PEM calibration for total organic and total inorganic phosphorus (mg P/L), May and September, 1966
[Hydroqual (1982)].
4.00
3.20
f, 2.40
a 1.60
Z 0.80
UJ
§ 0.0&
May 26, June 2
• Niy 30
4.00
3.20-
^ 2.40
z
S, 1.60
O.BO
0.00
Hoy 26, June 2
• H.y 30
AMMONIA
4.OO-
3.20-
2.40
1.60-
080-
0.00-
1 • Sept 7, Sept 8
M Sept 7
\ :
1 '
. 1:. \
^•1--V-^. 1 :: : . . i
20
40
RIVER MILE
IOO
" 4.00-1
<
c
5 3aoH
i
ui
t 2.40-
o.eo-
o-oo^-f, ^0'-^40
60
RIVER MILE
Sept 7, Sept B
Sept 7
80 100
Figure 11-10. PEM calibration for ammonia and nitrite+;nitrate (mg N/L), May and September, 1966 [Hydroqual (1982)].
11-10
-------
10,000
i °-32
s
gO.24
a.
« 0.16
tr.
O
0.00
. July 18, July 20
— July 19
25O
Jl 200
-J ISO
>-
I 'OCH
3
S 50
. July IB, July 20
— July 19
'—I—I—i—I—i—i
40 60
RIVER MILE
100
§0.80-
1 '
0)°'64'
tr
? 0.48-
a.
en
I 0.32-
1 0.'6-
1-
O.OO-
• July It, July 20
July 19
X"\C_._^ : * 3
d •• :•: - • ^
20
40
RIVER MILE
Figure 11-11. PEM verification for chlorides (mg/L), chlorophyll-a (ug/l), dissolved inorganic phosphorus (mg P/L) and total
phosphorus (mg P/L), July 1977 [Hydroqual (1982)].
— 3.0-
z
J 2.4-
C ~
| 1.8-
z
D
"c
0.0-
3.0-
|M:
g 1.8-
S
i 1.2-
|
£
z I
. July IB, July 20
— July 19
••
JI-.
1 1 ' • | • • i j i • . |
• July 18, July 20
— July 19
A-
/ \ '
-J- V'1" :
> 20 40' ' ' 60' ' 80' ' ' IOC
RIVER MILE
a" 6
o
m
20
• July IB, July 20
— July 19
=
D»
C
«
Ok
>*
S
•a
"o
w
M
0
i
15-
12-
9-
6
3-
0-
• July IB. July 20
— July 19
-.-• ^LL^
' • • * .
'•- : ' :*•••:
t
40 60
RIVER MILE
80
lob
Figure 11-12. PEM verification for ammonia nitrogen (mg N/L), nitrite-nitrate nitrogen (mg N/L), bottle BOD5 (mg/L) and
dissolved oxygen (mg/L), July 1977 [Hydroqual (1982)].
11-11
-------
Table 11-1. Unear Regression Statistics
Chlorophyll a
Year
1968
1969
1977
1978
1979
I*
0.81
0.75
0.73
0.88
0.78
Standard Error
(uall) I
10.8 i
10.8 '
17.6
4.3
2.3
Slope
0.79
0.76
0.84
0.70
0.35
Intercept fag/l)
7.7
2.5
12.99
12.26
20.00
Hypothesis
R
R
A
R
R
Dissolved Oxygen
Year
1968
1969
1970
1977
1978
1979
/
0.60
0.93
0.74
0.73
0.75
0.68
Standard Error
(mall) '•
0.74 !
0.38
1.41
0.93
0.59
0.56
Slope
0.83
1.16
0.58
1.21
0.68
1.08
Intercept (mg/l)
1.55
-0.76
2.45
-1.12
2.77
-0.29
Hypothesis
A
A
R
A
R
A
The relative errors of the summer average means of the
principal state variables were also calculated in jthe
PEM study. These values indicate a large degree of
variation among variables for any one year, as wejl as
across years for any one variable. The median relative
errors, ranged from 10 to 30 percent for chlorophyll a,
5 to 10 percent for DO, and 15 to 25 percent acrosjs all
variables.
In comparing the means, a Student's "t" test was used
to determine the difference between the obseryed
mean and the computed mean. If there was no [sig-
nificant statistical difference between the means,! the
model was assumed to be verified. This statistic indi-
cates that there was no statistical difference between
observed and computed summer means for 77 perfcent
of the variable-segment pairs for which a comparison
could be made.
11.4.8. Post-Audit
Despite the continued reduction in point source phos-
phorus loading and gradual improvement in \ftjater
quality, a massive and unexpected bloom of blue-
green algae occurred in the Upper Potomac during the
summer of 1983. By August, the bloom had exceeded
200 /*g/l of chlorophyll a. The bloom continued into the
months of September and October. The occurrence of
the 1983 algal bloom offered a unique opportunity to
evaluate the predictive capability of PEM. A post-audit
PEM simulation was performed to test the ability of the
model to predict the observed bloom conditions
(Hydroqual, 1989).
The PEM post-audit was conducted in conjunction with
an Expert Panel convened to investigate the cause of
the bloom. Their conclusions (Thomann et al, 1985)
can be summarized as follows:
• PEM was able to successfully predict chlorophyll
concentrations in the portions of the estuary
upstream of the bloom, and was able to predict the
onset of the bloom to nuisance levels through the
end of July.
• PEM was not able to predict the intensification of
the bloom, neither in magnitude nor spatial or
temporal extent.
• Model comparison to data indicated that there was
a significant source of phosphorus to the bloom
area that was not being considered by PEM.
The Expert Panel subsequently recommended that
investigations be undertaken to define the source of
increased nutrients. These investigations were to in-
clude evaluation of pH effects on sediment nutrient
release, and evaluation of the factors controlling
alkalinity and pH in the Potomac. The Expert Panel also
11-12
-------
recommended that PEM be updated to include newly
identified factors.
The first revision of PEM incorporated the results of
bloom-related experiments that indicated increases in
water column pH could significantly increase the mag-
nitude of sediment nutrient flux. This resulted in the
addition of two components to PEM: 1) simulation of
pH, and 2) inclusion of a pH-mediated sediment flux.
The simulation of pH required the addition of a separate
submodel to simulate the equilibria between the multi-
ple forms of inorganic carbon. This pH-driven equi-
librium is also affected by algal photosynthesis, which
increases water column pH. The second submodel
added to PEM related to pH-mediated sediment
release. The original version of PEM simulated sedi-
ment quality and the flux of nutrients across the sedi-
ment water interface. The updated PEM removed these
sediment computations and replaced the predicted
nutrient flux as a pH driven boundary condition.
This "first revision" of PEM provided improved predic-
tion of 1983 conditions over the original version, but
was still unsatisfactory for the relationship between
phytoplankton, dissolved oxygen, nutrients, and the
carbonate system. PEM was then further updated to
include a second algal species representative of the
blue-green alga Mlcrocystls, which was the primary
component of the observed bloom. Re-calibration of
the model provided an improved description of the
observed data.
11.5. Finite Element Model
Chlorine has been used extensively as a wastewater
disinfectant and as an agent to prevent biofouling in
cooling waters. Concerns have been raised that the
discharge of chlorine in wastewater to the Upper
Potomac Estuary might pose ecological health risks.
In particular, discharges from opposing shorelines
might result in a cross channel barrier that could
prevent fish movement and migration. This study was
conducted to determine the occurrence and fate of
residual chlorine in the Potomac and to evaluate the
likelihood of the formation of a toxic cross channel
barrier.
A comprehensive study was conducted involving field
surveys of discharge and Potomac Estuary total
residual chlorine (TRC) concentrations. The objectives
of the study were to document the current spatial
extent of TRC; to develop and calibrate a two dimen-
sional TRC model for testing various environmental
scenarios; and to conduct model analysis of the
various scenarios to establish the risk of a chlorine
barrier.
The study area of the Potomac Estuary is freshwater
but hydraulically influenced by ocean tides. The con-
fluence with the Anacostia River, numerous embay-
ments, and highly variable channel physiography
make this section of the Potomac Estuary
hydrodynamically complex. The data available to sup-
port a TRC model were limited to grab samples in only
the longitudinal and lateral dimensions. Modeling was
therefore constrained to two dimensions. This was,
however, consistent with the purpose of the modeling
— to define the lateral and longitudinal extent of effluent
iresidual chlorine plumes as a potential barrier to fish
migration.
The complex physiography of the upper Potomac Es-
tuary did not allow use of simple analytical models.
One-dimensional water quality models were of little use
for evaluating the chloride discharges because the
lateral extent of contamination could not be simulated.
Branching one-dimensional estuary models, such as
the Dynamic Estuary Model (DEM) may be configured
to run as pseudo-two-dimensional models but have
unrealistically high dispersion for localized calculations
and poor characterization of two-dimensional
transport. For these reasons, a true two-dimensional
hydrodynamic and water quality model was required.
The Neleus chlorine model selected for this study
consists of a hydrodynamic model linked to a water
quality model. The hydrodynamic model solves the
complete non-linear, two-dimensional, partial differen-
tial equations of fluid motion (Katopodes, 1987; LTI,
1987). The equations are integrated over time using a
modified Petrov-Galerkin finite element model numeri-
cal technique yielding surface elevation and velocity at
each of the model grid nodes. The results are input as
mass transport terms to the water quality model.
The water quality model uses the same grid.framework
as the hydrodynamic model and is represented by a
two-dimensional, vertically averaged, partial differen-
tial equation of mass transport. The equation includes
terms for advective and diffusive mass transfer, mass
sources and/or sinks, and first-order decay. The
numerical solution is obtained in the same manner as
with theflow equations except that an iterative solution
is not required since the mass transport equation be-
comes linear with the assumption of zero diffusive flux
at the model boundaries.
11.5.1. Model Inputs
The Neleus model required a finite element grid com-
prising 1171 quadrilateral elements with 1408 nodes
(element intersections) as shown in Figure 11-13. This
fine detail was required because of complex
bathymetry. In addition, grid resolution had to be high
near pollutant sources to maintain numerical stability
11-13
-------
during computation and to provide accurate model
predictions within fairly short distances of discharge
locations.
After setting the model grid, model inputs for boundary
conditions and loadings were determined. These in-
cluded tidal elevation and flows. The NOAA Tide Tables
provided minimum and maximum tidal elevations afid
a sinusoidal Interpolation scheme was used to provide
tidal elevations for each hydrodynamic model time
•5
step. Some actual recorded tidal elevation data were
available for use in modeling the residual chlorine
surveys. Minimum and maximum elevations and time
(NOAA, 1984) were abstracted from the continuous
record. Advective freshwater discharges were
specified as nodal velocities at the upstream ends of
the model for each simulation. These were determined
using information from USGS flow records for both the
Potomac (at Chain Bridge) and Anacostia channels.
Blue Plains
W.VTP I
i 4-
Figure 11-13. Chlorine model finite element grid network [UTI (1987)].
11-14
-------
Daily variations in discharge were incorporated in
simulations when appropriate.
In terms of pollutant inputs, four chlorine discharge
locations were identified in the study area as:
• Blue Plains WWTP • Alexandria WWTP
Arlington WWTP
PEPCO Power Plant
The Blue Plains Wastewater Treatment Plant was the
only source for which information was known about
outfall configuration and precise location. As a result,
the other chlorine sources were treated as mass pol-
lutant loadings with no momentum effects. The impact
of this simplification on main channel model results
was minimal since Arlington and Alexandria discharge
to embayments and PEPCO discharges chlorine inter-
mittently at very low levels.
11.5.2. Available Data
Four surveys conducted prior to the modeling effort
were available for model calibration. First, the USGS
conducted a dye survey over a six day period in
August, 1980 (Hearn, 1984). Dye was injected for one
tidal day (24.8 hours) from the Blue Plains outfall and
subsequently measured throughout the study area.
Three surveys conducted by the District of Columbia
Department of Consumer and Regulatory Affairs
provided effluent and ambient TRC concentrations
throughout the tidal cycle.
11.5.3. Model Calibration/Verification
The Neleus model involved validation for both
hydrodynamic and water quality models. The
hydrodynamic model has one calibration parameter —
Manning's n, which reflects the hydrodynamic effects
of bottom roughness. The lack of hydrodynamic field
data limited the calibration of the hydrodynamic model.
However, previous work by Katopodes (1987) resulted
in a limited calibration of the model hydrodynamics
through comparison with DEM hydrodynamic predic-
tions. A constant Manning coefficient of 0.026 was
used by Katopodes (1987) and was chosen for use in
the chlorine study. The water quality model has three
parameters that require calibration: longitudinal and
lateral dispersion coefficient, and the first-order
chlorine decay rate. The dispersion terms were ad-
justed through simulation of the August 1980 USGS
dye study, while the chlorine decay rate was selected
through simulation of two of the 1984 chlorine field
studies.
The 1980 USGS dye study was used to calibrate the
lateral and longitudinal dispersion coefficients. The
model simulation began on the 10th of August with the
dye release simulation starting on the 11th. Discharge
from the Blue Plains outfall was constant with a flow of
517 cfs (334 MGD) and dye concentration was 0.03446
mg/l over the release period of 24.8 hours.
Longitudinal and lateral dispersion coefficients were
first estimated from literature information (Fischer et
al., 1979 and McDowell and O'Connor, 1977), but
refined to values of 120 ft2/sec for longitudinal disper-
sion and 10 ft /sec for lateral dispersion. Figure 11-14
presents the model dye predictions compared to
measured dye concentrations for two survey stations.
These simulations assumed no decay of dye.
The model predictions follow the trends in the dye data
for all stations. Evidence of dye loss is seen for stations
B and C beginning on approximately August 13th. The
inclusion of dye decay would improve the fit of the
model to dye data, but would not affect the calibration
of the dispersion terms. Since dye decay was not
important to the modeling of TRC, no further model
refinement for simulation of dye was performed.
The July, 1984 survey was selected for initial chlorine
modeling because the sampling covered a longer time
period than the other surveys. Data were collected
during both day and night. The effects of daytime
photolysis on chlorine decay could then be analyzed
by comparing day versus night results.
Loading during the survey included a total residual
chlorine concentration in the Blue Plains effluent of
0.333 mg/l at 330 MGD, in the Arlington WWTP effluent
of 1.9 mg/l at 26 MGD, and the Alexandria wastewater
treatment facility produced a total residual chlorine
level of 1.9 mg/l at a discharge rate of 43 MGD. The
PEPCO discharge was 401 MGD with intermittent ef-
fluent chlorine levels. The exact times during which
chlorination occurred were not known, but the levels
of chlorine applied to the cooling water were low. A
constant residual chlorine concentration of 0.02 mg/l
was used to represent the likely level of discharge from
PEPCO.
For initial simulations, a chlorine decay rate of 12.8 per
day was determined experimentally. A more conserva-
tive decay rate of 6.4 per day was also tested. The
comparison of model versus data is shown in Figures
11-15 and 11-16 for averaged field data and model
predictions. Averaging was used to simplify the presen-
tation of results and because the field data were not
sufficient to justify detailed comparisons. Contour lines
of constant concentration are used to depict model
output whereas field data are shown as singular
numeric values. In general, measured chlorine levels at
most field stations were too near detection limits to be
considered accurate except as order of magnitude
11-15
-------
estimates. Therefore, the averaging represents the
plume character well.
The comparisons of model to TRC data were cofi-
sidered reasonable for both loss rates. The differences
in the simulations were not dramatic and and indicated
that physical transport was dominant. The model char-
acterized the dissipation of TRC especially when co;n-
sidering the data can only be best relied on as an order
of magnitude indicator. The value of 96 jigl\ just to the
north of Blue Plains represents only one observation,
and appears to be an anomaly. Further sensitivity
analyses suggest that the lower decay rate of 6.4/day
might be more representative of nighttime conditions,
while the higher rate of 12.8/day may be appropriate
for daytime.
t>
'§
o
•o
o
o:
JD
o.
O
£
1O.O
9.O -
8.0 -
7.0 -
&.O -
5.0 -
4.0 -
3.0-
2.0 -
1.O -
O.O
11
10.0
13
15
17
11 — 16 August 198O
Figure 11-14. August 1980 dye survey calibration at stations B and C [LTI (1987)].
11-16
-------
- i,., Time Averaged
38.00 to 53.50 hrs
,a I
iCunaui
-t-
S.
4
Model Contours
ug/l TRC
Field Data
ug/l TRC.
Discharge Site
Figure 11-15. July 11,1984 TRC survey calibration at 12.8/day loss rate [LTI (1987)].
Time Averaged
38.00 to 53.50 hrs
-•-
S,
4
^
#g&
\fc$*
iffcC^TK*
Model Contours
ug/l TRC
Field Data
ug/l TRC
Discharge Site
Figure 11-16. July 11,1984 TRC survey calibration at 6.4/day loss rate [LTI (1987)].
11-17
-------
The October, 1984 survey was chosen for model
validation because it had the greatest spatial quantity
of chlorine data. The dispersion and chlorine dechy
rates adopted for the July 1984 calibration runs we're
used for modeling of this survey, but inputs for actual
observed loadings and ambient environmental condi-
tions were used.
The chlorine model predictions for this survey did not
compare well at certain stations but the predicted
plume front and general decrease in chlorine levels
moving away from the Blue Plains outfall compared
well with data. As a result, the model was considerpd
sufficiently validated to evaluate the potential for: a
cross channel barrier. Refined model calibration of the
chlorine loss rate was not possible because of data
limitations and variability. The model was still deemed
well suited to assess the presence or absence of a
cross channel barrier. The more conservative loss rate
was used forthis purpose. The modeling effort was riot
considered to be well suited for highly precise predic-
tions or for waste load allocations.
11.5 A Model Application
The potential existence of a chlorine concentration
barrier was examined by model simulation overa range
of conditions. These included variations in effluent
loads, river flow, and tidal conditions. All other aspects
of the model were Identical to those used in the calibra-
tion procedure. Effluent loads to the Upper Potornac
Estuary included the Blue Plains wastewater facility
(370 MGD) on the east shoreline, and the Alexandria
(54 MGD) and Arlington (30 MGD) wastewaterfacilities
plus the PEPCO cooling water discharge (350 MGD)
on the Western shoreline. The discharge values for the
wastewater treatment facilities represent estimated
capacity needs for the period 2005-2010 (MWWRPB,
1986). The Arlington and Alexandria discharges were
examined at a 1.0 mg/l total residual chlorine (TRC)
concentration. The PEPCO chlorine discharge level
used was 0.02 mg/i. Blue Plains, the largest w|as-
tewater plant was examined under two TRC scenarios,
0.02 mg/l and 0.40 mg/l TRC. This represented condi-
tions with and without dechlorination.
Three Potomac river flow conditions were examined.
The critical seven day, ten year drought lowflow (7Q10)
of 470 cfs, and two April flow conditions to characterize
a period of likely fish migration. The long term average
April flow (19,900 cfs) was simulated as well as [the
lowest recorded mean monthly April flow (7,573 cfs).
For the three conditions the actual corresponding
Anacostia River flows were 8,165, and 345 cfs, respec-
tively.
Model results for the various simulations are sum-
marized in Figures 11 -17 to 11 -19. For each simulation,
model results were examined for all phases of the tide:
ebb, flood, and slack. For these purposes, the model
output has been displayed for the most critical condi-
tion where the chlorine residual extends the furthest
distance across the Potomac. Model results for other
periods in the tidal cycle were less critical and are not
shown. A 10/tg/l criterion for TRC was used to charac-
terize the plume boundary because this is the District
of Columbia water quality standard. The figures display
the boundaries of the 10 and 2fig/\ TRC concentration
contours. Higher concentrations were only apparent
in the immediate vicinity of the discharge pipes and
dissipated quickly. These very near zone discrimina-
tions were not a model objective, and cannot be ex-
amined accurately by this model. A jet plume model
that incorporates the hydraulic characteristics of the
the discharge itself would be required to evaluate water
quality impacts in the immediate vicinity of the dis-
charge.
Among all examined scenarios, no conditions were
simulated where the 10 ^g/l concentration boundary
extended across the entire Potomac and presented a
potential TRC barrier. At the 0.4 mg/l level of TRC in
Blue Plains effluent, the boundary extends ap-
proximately one third of the river width. Discharges
from Alexandria and Arlington were largely dissipated
in their respective embayments. The PEPCO discharge
only minimally impacted the main channel. For 0.02
mg/l TRC (dechlorination) at Blue Plains, the plume is
barely observable in the main channel. All predicted
main channel concentrations were less than 10/ig/l.
In the model calibration section of this case study, the
deficiencies in the calibration data sets were noted, as
was their significance to model uncertainty. Nonethe-
less, the uncertainty in model rates would not be suffi-
cient to alter the basic findings. Reasonable changes
in dispersion rates had small effects on the plume
width. In addition, for forecast purposes a conserva-
tively low chlorine loss rate was used. This maximized
the predicted plume persistence.
11.6. References
Clark, L J. 1982. A Modeling Study of the Upper
Potomac Estuary: Application of the Dynamic Estuary
Model. Draft Report. U.S. EPA Annapolis Field Office,
Annapolis, Maryland.
Fischer, H.B., List, J.E., Koh, R.C.Y., Imberger, J., and
N.H. Brooks, 1979. Mixing in Inland and Coastal
Waters. Academic Press, New York.
Greeley and Hansen Engineers. 1984. Blue Plains
Feasibility Study Final Report. D.C. Department of
Public Works, Water and Sewer Administration,
Washington, D.C.
11-18
-------
hr»
-*-• Model Contours
ug/l TRC
4 Discharge Site
Blue Plains Effluent TRC « 0.40 mg/1
Model Contours
ug/l TRC
Discharge Site
Blue Plains Effluent TRC « 0.02 mg/l
Figure 11-17. TRC model projection for 7Q10 low flow conditions [LTI (1987)].
11-19
-------
Tim* • 62.50 hrs
, • i • i • i • i •
0.0 miles 1.0
-j- Model Contours
ug/l TRC
4 Dischorge Site
Blue Plains Effluent TRC = 0.40 mg/1
Model Contours
ug/l TRC
Discharge Site
Blue Plains Effluent TRC * 0.02 mg/I
Figure 11 -18. TRC model projection for average April flow conditions [LTI (1987)].
11-20
-------
Tim* - 62.50 hrt
-»- Model Contours
ug/l TRC
4 Discharge Site
Blue Plains Effluent TRC = 0.40 mg/1
-»- Model Contours
ug/l TRC
\ Dischorge Site
Blue Plains Effluent TRC » 0.02 mg/1
Figure 11-19. TRC model projection for lowest April flow conditions [LTI (1987)].
11-21
-------
Hearn, Paul P. Jr., 1984. Controls on Phosphorus
Mobility In the Potomac River Near the Blue Plains
Wastewater Treatment Plant, USGS Water Supply
Paper 2231.
Hydroqual, 1982. Calibration and Verification of Math-
ematical Model of Eutrophication of the Potomac Es-
tuary. District of Columbia Department of
Environmental Services, Washington, D.C.
Hydroqual, 1989. Re-Calibration of the Potomac
Eutrophication Model to the 1983 Algal Bloom.
Metropolitan Washington Council of Governments,
Washington, D.C.
Katapodes, N. D., 1987. Finite Element Model for
Hydrodynamics and Mass Transport in the Uppkr
Potomac Estuary, in "Dissolved Oxygen Study of the
Upper Potomac Estuary," Volume 1 , Technical Appen-
dix A-7, Metropolitan Washington Council of Goverji-
ments, Washington, D.C. :
Limno-Tech and Metropolitan Washington Council pf
Governments, 1987. Summary Report: Potomac River
Residual Chlorine Study. Department of Consumer and
Regulatory Affairs, Washington, D.C.
and Estuary, Maryland and Virginia, May 1978 to
November 1981", Water Supply Paper 2234-A.
LTI, Limno-Tech, Inc. 1987. Validation of DEM to
and 1986 Upper Potomac Estuary Data. Metropolitan
Washington Council of Governments, Washingtoh,
D.C.
McDowell, D.M., and B.A. O'Connor, 1977. Hydraulic
Behaviour of Estuaries. Halstead Press, John Wilpy
and Sons, New York. !
Metropolitan Washington Council of Governments,
1987. A Dissolved Oxygen Study of the Upppr
Potomac Estuary, Final Report. Washington, D.C.
MWWRPB (Metropolitan Washington Water Resour-
ces Planning Board), 1986. "Metropolitan Washington
Water Quality Management Plan, 1986 Plan Supple-
ment,", Metropolitan Washington Council of Govern-
ments, Washington, D.C. i
NOAA, 1984. Observed Hourly Potomac River Tidal
Heights for for 1984 at Washington, D.C., provided by
the National Ocean Survey.
Thomann, R.V., N.J. Jaworski, S. W. Nixon, H. W. Paerl
and Jay Taft, 1985. The 1983 Algal Bloom in the
Potomac Estuary. Prepared for the Potomac Strategy
State/EPA Management Committee '
USGS, 1985. "Distribution and Abundance of Sub-
mersed Aquatic Vegetation in the Tidal Potomac.River
11-22
-------
12. Manasquan Estuary Real Time Modeling
12.1. Background
This study of the MIT-Dynamic Network Model (MIT-
DNM) demonstrates the successful calibration and
verification of a real-time estuary model. Unlike tidally-
averaged or steady-state models, real time models
simulate changes in flow and water quality constituents
on an hour to hour basis. MIT-DNM was selected by
the Manasquan River Regional Sewerage Authority to
predict the effect that the discharge from a proposed
wastewater treatment plant would have on the water
quality and ecology of the Manasquan Estuary
(Naj'arian et al., 1981). The Authority was primarily
concerned with nutrient enrichment and primary
productivity in the estuary. A real time model was
selected to predict photosynthesis effects on diurnal
DO concentrations and investigate the transient im-
pacts of nonpoint source pollution and salt water in-
trusion.
The hydrodynamic submodel of MIT-DNM uses a finite
element approach to solve the one-dimensional con-
tinuity and momentum equations for unsteady flow in
a variable area channel. Dispersion is defined by the
degree of stratification and the non-dimensional lon-
gitudinal salinity gradient using the relationship formu-
lated by Thatcher and Harleman (1972, 1981). The
flows and velocities calculated by this submodel are
used in another submodel in which a sequence of
conservation of mass equations calculates the tem-
poral and spatial variation in the water quality
parameters.
The following state variables are included in this ver-
sion of MIT-DNM:
Diatoms
Nitrite and nitrate
• Nanoplankton • Carbonaceous BOD
• Dinoflagellates • Dissolved Oxygen
• Organic detritus N • Chlorides
• Ammonium-N • Fecal coliform
• Herbivorous zooplankton
The model assumes that the dominant activity in the
estuary is aerobic and that nitrogen is the only nutrient
that limits the growth of algae. Water quality processes
represented in the model include phytoplankton
growth, mortality, and sinking; zooplankton grazing,
mortality, and excretion; nitrogen cycling and fluxes at
the sediment/water interface.
12.2. Problem Setting
The Manasquan Estuary is approximately 7.6 miles
long, extending from the Atlantic Ocean to Brick
Township in east central New Jersey. The estuary
receives inflow from the Atlantic Ocean, the
Manasquan River, and Barnegat Bay, which is con-
nected to the estuary by Point Pleasant Canal. The
landward reaches of the estuary are very shallow, with
large embayment and marsh areas. Figure 12-1 shows
the study area with sampling stations.
Flow records for this area came from USGS gage data
at Squankum on the Manasquan River. The freshwater
low flow was 17.0 cfs, which included 6.2 cfs from
wastewater treatment plants discharging upstream of
the gage. At the time of the study, no other major point
sources discharged into the river or the estuary. The
Manasquan River Regional Sewerage Authority, how-
ever, proposed the construction of a regional ad-
vanced wastewater treatment facility that would
discharge 9.4 cfs of effluent at the head of tide of the
estuary. The plant would obviously be a major con-
tributor to the freshwater flow into the estuary under
low flow conditions.
Figure 12-1. Manasquan Estuary and Inlet [Najarian et al.
(1981)].
12-1
-------
Bargenot Bay
Bay Head Harbor
(3
(^ Boundary Nodes
A)
Head of Tide
Manasquan Inlet
Sourcei Nojarian ilol.. 1981
Figure 12-2. Model conceptualization of the Manasquan Estuary and the Point Pleasant Canal System [Najarian et al. (1981)].
Effluent standards to be met by the proposed plahr
were established by the Authority and are shown in
Table 12-1.
The flow and salinity dynamics in the Manasquan Es-
tuary system are forced by two tidal boundaries at
Barnegat Bay and the Atlantic Ocean and by the fresh-
water inflow from the Manasquan River. Differences in
Table 12.1. Effluent Quality Standards
Parameter
Standard
BODs
NHa-N
DO
PH
NOa-N
Others
95% removal
2mg/l ;
6 mg/I
5.5-7.5
7 mg/1
None detectable by EPA-
approved methods of analysis.
i
i
Such that New Jersey Surface
Water Quality Standards for !
R/V-2 Trout Maintenance
Streams will be met. '.
tidal amplitudes and phases between the ocean and
the bay cause a complex flow regime in the estuary.
The tidal boundaries also differ in water quality. While
the constituent concentrations at the ocean boundary
are relatively constant, the concentrations at the bay
boundary are much more variable due to the mixing of
bay waters with Manasquan water.
FigUre 12-2 shows a schematic of the MIT-DNM reach
system established for the estuary. The first reach
extends 2.26 miles landward from Osborn Island. The
second reach extends from Osborn Island to the Atlan-
tic Ocean, and is 5.32 miles long. The third reach, 1.78
mile's long, represents Point Pleasant Canal. Each
reach is represented by geometrically irregular cross-
sections, with embayment volumes specified for Lake
Stockton and Sawmill Creek. Tidal boundaries are
specified for nodes 1 and 3, and an inflow boundary is
specified at node 4.
12.3. Model Calibration
Figure 12-1 showed the location of stations for model
calibration sampling performed in July and August,
1980, byElsonT. Killam Associates, Inc. Two sampling
events were conducted over a four-day period - July
21-24 and August 25-28, 1980. Salinity and nutrient
concentrations at each station were measured during
daylight hours at frequencies of 3-4 observations per
tidal cycle.
12-2
-------
The measured water quality parameters were as fol-
lows:
• Temperature • Nitrite
• Dissolved Oxygen • Nitrate
• Salinity • Ortho-phosphate
0 Secchi Depth • Silicate
• Dissolved Org.-N • Chlorophylla
• Paniculate Org.-N • BOD
• Ammonia
The zooplankton and phytoplankton species present
during each sampling event were identified. In addition,
synoptic data on the tides and the freshwater inflow at
the boundaries and at three instream stations in the
Manasquan Estuary were also collected. Freshwater
inflows at the head of the estuary in July and August
persisted at about 30 to 40 cfs without any dramatic
increases between the two sampling events. The
August data set was selected for model calibration
because all three algal species represented in the
model (diatoms, nanoplankters, and dinoflagellates)
were present during this month. The July data set was
used for the purpose of model verification.
723.1. Hydrodynamic Submodel Calibration
Calibration of model hydrodynamics must precede
water quality calibration. In the Manasquan study, it
was imperative to start the model with realistic initial
hydraulic and salinity conditions since field observa-
tions only covered a 4 day or 9 tidal cycle period of
time. To establish realistic initial conditions for the
August 24-28 sampling event, the model was run for an
antecedent period of three tidal cycles that damped out
transients resulting from unrealistic initial conditions.
Repeating tides of 12.42 hour periodicity were imposed
at the inlet and at the Bay Head Harbor boundaries.
These tides were extracted from the tides observed
during the first day of sampling. The necessary adjust-
ment to tidal records at the Bay Head Harbor boundary
was made to reflect the differences in the MSL elevation
between the inlet and the head of Barnegat Bay. The
surface elevations and velocities computed during the
last time step of repeating tide simulation were then
taken to be the initial hydraulic conditions for 12:30
p.m. on August 24,1980.
Three sets of hydraulic boundary conditions were
specified for each day of the simulation using observed
data at the head of tide, the inlet, and Bay Head Harbor.
The hydraulic calibration of the model was ac-
complished by matching the observed tidal ranges and
the phases measured at Clark's Landing and
Chapman's Wharf by calibrating Manning's friction
coefficient. The model accurately, simulated the ob-
served hydraulics. The maximum difference in ob-
served and computed data was approximately 7% of
the tidal elevation range.
Salinity was calibrated next. Salinity observations did
not begin until August 25, 1980. As done for tidal
elevations, initial conditions for salinity were calculated
by running the model for an antecedent period. The
observed salinities at sampling stations on August 25
were averaged for that day and assumed as concentra-
tions at these stations, initial salinity concentrations at
computational points between stations were
generated by linear interpolation. At the two tidal
boundary stations, the extreme observed salinities
were assigned during the end of flooding flows and the
model computed salinity concentrations during the
ebbing flows. The model requires specification of the
time it takes for the boundary salinities to reach the
extreme observed salinities after the flood flows begin.
The freshwater inflow boundary condition is assumed
to have a background salinity of 0.09 percent. To
calibrate the mass transport of chlorides in the estuary,
dispersion must be represented adequately. This re-
quires calibration of the stratification parameter K and
Taylor's dispersion multiplier m, both of which are used
in the following dispersion equation:
KdS
dX
+ mEj
where,
E(x,t) = Temporally and spatially varying dispersion
coefficient (fr/sec)
S = s/So, where s(x,t) is the spatial and temporal
distribution of salinity (dimensionless)
So = Ocean salinity (ppm)
X = x/L (dimensionless)
L = Length of estuary to head of tide (ft)
ET = Taylor's dispersion coefficient (f^/sec)
<= 77 unRhS/6
u = u(x,t) = tidal velocity (ft/sec)
n = Manning's friction coefficent
Rh = Hydraulic radius (ft)
K = Estuary dispersion parameter (f^/sec)
= UoL/1000
12-3
-------
(a)
35
30
<
v>
25
8/25 8/26
8/27 8/28 8/29
25
~ 20
J
(b)
15
<
-------
72.3.2 Water Quality Submodel Calibration
Once the hydraulics and mass transport within the
estuary were adequately defined, the model was
calibrated for water quality parameters. Like the tidal
elevation and salinity calibrations, this calibration re-
quired initial and boundary conditions, and also in-
volved the evaluation of transformation rate constants
based on plots of simulated versus observed data.
The system again requires the establishment of two
ocean boundaries and a time-varying boundary at the
head of tide, as well as initial conditions throughout the
system. The ocean boundaries were handled as in the
salinity calibration, where ocean concentrations were
specified at the end of flood flows and water quality
values were computed internally during ebb flows.
Observed conditions at the Squankum USGS gage
were used to define the time-varying water quality
conditions at the Manasquan head of tide. Because the
phytoplankton and zooplankton concentrations were
sampled only once during the sampling period, time-
invariant concentrations were specified at the three
boundaries. Initial conditions were estimated using
sampling data and linearly interpolated to establish
values between sampling stations.
Unlike the hydraulics and salinity calibrations, where a
combined total of three constants were calibrated on
the basis of observed versus simulated data, the water
quality calibration requires the determination of many
constants. Table 12-2 shows the values that were es-
tablished through model calibration. These are the
values that best represent the site-specific kinetic
processes in the Manasquan Estuary while still falling
within the range of values found in the technical litera-
ture.
Examples of simulated versus observed plots for the
various water quality parameters are illustrated in
Figures 12-4 to 12-10. The individual symbols indicate
observed data points, while the straight line shows the
continuous simulation model output. These plots rep-
resent the best simulation of observed data using
reasonable rate constants and coefficients. Model
goodness-of-fit was determined only through visual
observation of the plots; no statistical tests were per-
formed.
Figure 12-4 shows the simulation of detritus-N, am-
monium-N and nitrite + nitrate-N at Clark's Landing.
The ammonium-N concentrations predicted by the
calibrated model were reasonably close to observed
values. The simulation of detritus-N and nitrite+nitrate-
N was less accurate, particularly at Chapman's Wharf.
The computed nitrite+ nitrate-N concentrations were
sometimes an order of magnitude lower than the ob-
served concentrations at the station. The modelers
could find no explanation for this problem. To ade-
quately simulate detritus-N at Chapman's Wharf, a
source of 240 Ib/day was introduced as a distributed
load. Because the estuary is very shallow, the modelers
justified this input to the model by speculation that tidal
disturbances could have resuspended some of the
settled detritus.
The calibration of CBOD, NBOD, and DO at Clark's
Landing and Chapman's Wharf are shown in Figures
12-5 to 12-10. Like the detritus-N simulation, an ade-
quate CBOD simulation at Chapman's Wharf was not
possible withoutthe introduction of a distributed CBOD
load. Even after assuming a load of 3,500 Ib/day, the
observed and simulated data did not match well. The
DO simulation results proved to be confusing at both
stations. Although low concentrations of NBOD and
CBOD were observed at Clark's Landing, the observed
DO levels at this station are lower than the concentra-
tions predicted by the model. Conversely, the ob-
served DO levels at Chapman's Wharf climbed much
higher than the simulated DO concentrations, even
though large CBOD and detritai nitrogen concentra-
tions were observed there.
Because no sampling was conducted to measure
phytoplankton concentrations over time, the model
could not actually be calibrated for these water quality
parameters. The relative proportion of each algal
species was input to the model based on the observed
data gathered from the single sampling event and
species identification. Figures 12-11 and 12-12 show
the simulated concentrations of phytoplankton
nitrogen at the two stations. The plots clearly indicate
a strong tidal effect upon phytoplankton concentra-
tions.
12.3.3. Model Verification
The purpose of model calibration is to establish the
values of coefficients, such as Manning's "n" or decay
rates, which accurately represent the physical and
biochemical nature of the system. Once these values
are established, they must be verified. Using the same
values to represent the estuary, the model must be
applied to a different time period for which sampling
data are available. If the simulated concentrations ac-
curately predict observed concentrations the model
can be considered verified.
The verification data were obtained during a sampling
event in July 1980. This event, like the August 1980
sampling that provided the calibration data, was four
days long, with salinity and nutrient concentrations
measured during daylight hours at frequencies of 3-4
observations per tidal cycle. As in calibration, the ob-
served data were used to establish initial and boundary
12-5
-------
Table 12-2. Model-Governing Rate Parameter* [Najarlan et al. (1981)]
Transformation
Ki: Ammonlflcatlon of detritus-N
Ka: Diatom uptake of ammonlum-N
K2°P<
Ya (half saturation constant)
1. (Optimum intensity of solar radiation)
Topi (Optimum reaction temperature)
Rate
0.08
0.415
0.0122
144
10-15
Unit
da/1
da/1 I
mg-N/l
ly/day
°C
Tm«x (Maximum temperature beyond 30 °C
which denaturatlon of cell protein occurs)
Ka: Nanoplankton uptake to ammonlum-N
K30*
Y3
1,
•f
Topi
Tmax
K<: Diatom mortality rate
KB: Nanoplankton mortality rate
KB: Copapod uptake of dlatom-N
Kgopt
Ys
Toot
Tmix
Ky: Copepod mortality rate
KB: Copepod excretion rate
KB: Nitrification rate
Kio:DIatom uptake of (NO2+NOa)-N
KlQOp*
Y4
1.
Topi
Tmax
Kn:Nanoplankton uptake of (NOa+NO3)-N
K,,°<*
Y8
1,
Topt
_
max
0.553
0.004
64.8
97 "?
£>< »O
30
0.05
0.05
0.50
0.10
24-28
35
0.20
0.08
0.15
0.376
0.0063
144
10-15
30
0.501
0.015
64.8
24-28
SO
da/1
mg-N/l
ly/day
0^
•
°c
da/1
da/1
da/1
mg-N/A
°C
°c
da/1 ;
da/1
da/1
i
da/1
mg-N/l
ly/day
°c ;
°c : ;
da/1
mg-N/l
ly/day
°C
oc
Transformation
Ki2:Copepod uptake of detritus-N
Ki2opt
Yi
Topt
Tmax
KiaiCopepod uptake of nanoplankton-N
K13OP1
Y7
Topt
Tmax
Ki4:Dinoflagellate uptake of ammonium-N
Kl4°Pt
Ys
Is
Topt
Tmax
Kis:Dinoflagellate uptake of (NO2 + NOa)-N
Klgopt
Y9
,a
Topt
Tmax
KietCopepod uptake of dinoflagellate-N
Kieopt ; /
Yio
Topt
Tmax
Ki7:Dinoflagellate mortality rate
K17
Kis:Sedimentation of detritus
Kia
Kig:Sedimentation of (NOa+NOs)
K19
Kgo: Sediment release of ammonium-N
Kso
Rate
0.34
0.250
24-28
35
0.34
0.22
24-28
35
2.075
• '
0.021
288
20-25
32
1.880
0.042
288
20-25
32
0.5
0.1
24-28
35
0.05
0.1
100
10
Unit
da/1
mg-N/l
°C
°C
da/1
mg-N/l
°C
°C
day
mg-N/l
ly/day
°C
°C
day
mg-N/l
ly/day
°C
°C
da/1
mg-N/l"
o/>
C
°C
da/1
da/1
ug-at
N/m2/hr
ug-at
N/mz/hr
12-6
-------
• Dttritui-H
4- (Httrltt * Nftrate>-N
& Amoflfwll
© Predlcttd D«trltut-«
© Pr«dlct«d (Mitrit* * »itr«t«)-U
O Predicted Amonlm-N
0-6
0.2-
0.0
238
240 242
TIME IN DAYS
0.6
0.4
0.2
0.6
0.2
0.0
14
12
s
8
6-
2-
238 240 242
TIME IN DAYS
Figure 12-4. Temporal variation of Detritus-N, Ammonlum-N,
and Nit rite+Nit rate-N at Clark's Landing,
August 25-28,1980 [Najarian et al. (1981)].
Figure 12-5. Temporal variation of CBOD at Clark's Landing,
August 25-28,1980 [Najarian et al. i
14
12-
10
o.
a.
S 6
u
4-
2-
238 240 242
TIME IN DAYS
Figure 12-6. Temporal variation of CBOD at Chapman's
Wharf, August 25-28,1980 [Najarian et al.
(1981)].
-.30-
.20-
3
.10-
238 24O 242
TIME IN DAYS
Figure 12-7. Temporal variation of NBOD at Clark's Landing,
August 25-28,1980 [Najarian et al. (1981)].
12-7
-------
.40
.30-
a.
a.
.20-
O
m
.10
238 24O 242
TIKE IN DAYS
Figure 12-8. Temporal variation of NBOD at Chapman's
Wharf, August 25-28,1980 [Najarian et al.
(1981)].
14
12
2: to
z
UJ g
X
o
o
UJ
6
(o 4
Z38 240 242
TIME IN DAYS
Figure 12-9. Temporal variation of DO at Clark's Landing,
August 25-28,1980 [Najarian et al. (1981)].
14
12
10-
UJ p
o °
X
o
s 6
o 4
en
Q
238 240 242
TIME IN DAYS
Figure 12-10. Temporal variation of DO at Chapman's Wharf,
August 25-28,1980 [Najarian et al. (1981)].
J
2
Hj - Diatoms
N, - Nnnoplankton
N- - Dinof lagellates
A A A
Ns
36 ' 240 ' 242
o
OJ
n
a.
ID
o
-------
<0 (0
o
•
CM
35 O
Q.
Q.
Z
Ul
II-
z
1-
5
s.
o"
O
o
d-
2
M, - Diatoms
N^ - Nanoplankton
»7 - Dinof lagellates
A A A
Ns
58 ' 240 242
ci
-
-------
-3
200
202
203 205 207
TIME IN DAYS
204
TINE IN DAYS
206
208
Figure 12-14. Hydraulic verification: calculated vs. observed
elevations at Chapman's Wharf [Najarian et al.
(1981)].
Figure 12-15. Salinity verification: calculated vs. observed
salinities at Clark's Landing [Najarian et al.
(1981)].
8.
200
?.02
204 206
TIME IN DAYS
20(3
o.
CM
C>
203
• Detritus-N
-I- (Nitrite t Nitrato-N
A Aneontua-N
® Predicted Detrttus-H
Predicted Ammoniu«-N
205 207
TIME IN DAYS
O
.CO
O'
6 fc
!i
O
-O
-SI
-8
6
'ci
Figure 12-16. Salinity verification: calculated vs. observed
salinities at Chapman's Wharf [Najarian et al.
(1981)].
Figure 12-17. Temporal variation of Detritus-N, Ammonium-N,
and Nitrite+Nitrate-N at Clark's Landing
[Najarian etal. (1981)].
12-10
-------
8.
§
2O3
205
TIME IN DAYS
207
05
<0 3
O DC
. O
•o
O 5
Figure 12-18. Temporal variation of Detritus-N, Ammonium-N,
and Nitrite+Nitrate-N at Chapman's Wharf
[Najarianetal. (1981)].
report), the water quality simulation was considered
satisfactory.
12.3.4. Model Projections
The original goal of the modeling effort was to deter-
mine the impact that a proposed wastewater treatment
plant effluent would have upon water quality in the
Manasquan Estuary. However, the plans for the new
wastewater treatment plant were abandoned before
the calibration and verification studies were com-
pleted. Consequently, no production runs of the model
were conducted to assess discharge quality alterna-
tives for the proposed plant.
Though the developed model was not used to achieve
the original goal of the study, several important recom-
mendations were made regarding future model use.
These were:
1) External sources and sinks of nutrients should be
better defined,
2) Additional phytoplankton sampling should be done
to verify the model, and
3) Once these two steps are completed, the model
should be applied to the Manasquan Estuary.
The problems in calibrating detrital-N and CBOD at
Chapman's Wharf illustrated the need to better define
external sources and sinks of nutrients. Potential sour-
ces and sinks would include non-point source dischar-
ges, sediment-water exchanges, and marsh-estuary
exchanges. This last potential source/sink could have
been significant in the upper portion of the estuary,
where the estuary is shallow and the tidal portions
include marshlands. The other major observation
made by the modelers was that a more complete set
of data would increase confidence in the model. With
additional phytoplankton sampling, model simulation
of the algal species could be verified, and the model's
simulation of nighttime estuary activity could be
evaluated with round-the-clock sampling data. Once
the additional data were obtained, the recommenda-
tions were made that the model be used to:
(1) Determine the existing and potential impact of
nonpoint source pollution within the Manasquan River
Basin and
(2) Evaluate the potential impacts of proposed reser-
voir development within the basin on the downstream
Manasquan Estuary.
12.4 References
Najarian,T.O., Kenata, P. J. and Thatcher, M.L Decem-
ber, 1981. Manasquan Estuary Study. Manasquan
River Regional Sewerage Authority.
Thatcher, M.L. and Harleman, D.R.F. February 1972.
"Mathematical Model for the Prediction of Unsteady
Salinity Intrusion in Estuaries," Technical Report No.
144, R.M. Parsons Laboratory for Water Resources and
Hydrodynamics, Department of Civil Engineering,
M.I.T., Cambridge, MA.
Thatcher, M.L, and Harleman, D.R.F., February 1981.
"Long-Term Salinity Calculation in Delaware Estuary,"
Journal of the Environment Engineering Division,
ASCE, Volume 107, No. EE1, Proc. Paper 16011, pp.
11-27.
12-11
-------
12-12
-------
13. Calcasieu River Estuary Modeling
13.1. Background
The Calcasieu River Estuary modeling study is
presented here to illustrate a time-variable waste load
allocation model applied to a complex Gulf of Mexico
estuary. The general model framework of RECEIV-II
(Raytheon, 1974) was used to model simulate a forty-
mile stretch of river from the salt water barrier near Lake
Charles, Louisiana, extending downstream to the In-
tracoastal Waterway (shaded area in Figure 13-1). The
primary water quality problems were the result of point
source discharges. There were 64 wastewater dis-
chargers to the Calcasieu River below the salt water
barrier. I n the forty-mile study area, there is a seven mile
reach (between river miles 24 and 31) characterized by
depressed dissolved oxygen concentrations, elevated
temperatures and elevated ammonia concentrations.
The water above the salt water barrier also suffers from
low dissolved oxygen.
The poor water quality and the complexity of the sys-
tem has led to a series of water quality modeling studies
on the Calcasieu. Prior to the development of this
model, four other water quality modeling studies had
been completed on the Calcasieu. The first study was
reported in January 1974 by Roy F. Weston, covering
the entire Calacasieu River basin. It used a
nomographic (graphical) technique for preliminary
waste load allocation. A1980 study was conducted by
Hydroscience as part of a state-wide water quality
planning effort. This second model was an improve-
ment over the first, but it lacked a hydrodynamic
module and relied on the modeler to specify flow
conditions. Hydrodynamic data were very limited. In
1981, AWARE Inc. completed a third water quality
model of the Calcasieu River estuary for the section
below the salt water barrier using a two-dimensional
application of the RECEIV-II model. The model was
later used by Roy F. Weston for waste load allocation
analysis. The focus of the study described herein is a
more recent use of the RECEIV-H model for the Cal-
casieu River basin (Duke, 1985). Duke built on the work
of AWARE and Weston by improving the calibration
procedure and using new estuary cross-section infor-
mation.
13.2. Problem Setting
13.2.1. Site Description
The Calcasieu River estuary is a complex system of
natural and artificial channels. From its headwaters
near Slagle, the Calcasieu River flows southward for
160 miles to the Gulf of Mexico. The study area for this
model application was the lower40 miles of river, below
the salt water barrier (Figure 13-1).
The Army Corps of Engineers constructed the barrier
and maintains a dredged ship channel to a depth of 40
feet and bottom width of 400 feet in most of the estuary.
Stretches of the natural channel not dredged for the
ship channel are referred to as "loops" or "lakes." The
system is a tidal estuary with extensive side channel
and reservoir-like storage. Side channel and tributary
hydraulics are complicated by man-made channels
and the main channel flow is complicated by the
presence of large lakes.
High flows in the Calcasieu occur in the winter and low
flows occur in the summer. There are no permanent
stream flow measuring stations in the study area, al-
though six tide gages measure water levels. A seven
day, ten year drought flow (7010) was calculated using
relative drainage area sizes and the drought flow of the
nearest upstream gage station (Kinder, LA). The
drainage area above the salt water barrier is 3,100
square miles. The nearest upsteam station has a long
Figure 13-1. Calcasieu Estuary study area [NOAA (1985)].
13-1
-------
term mean flow of 2,600 cfs and a 7Q10 of 202 cfs. Trie
7Q10 below the salt water barrier was estimated to be
375 cfs.
13.2.2. Water Quality Monitoring •
The State of Louisiana conducted six water quality
surveys at 31 stations during the following periods:
• July 1978
• October 1978
• July 1979
August 1979
July 1980
June 1984
At each station, the following ten constituents were
measured and simulated in the model:
1) Water temperature 6) Nitrites
2) Salinity 7) Nitrates
3) Dissolved Oxygen 8) Ammonia
4) BOD
5) Phosphorus
9) Total Kjeldahl Nitrogen;
10) Chlorophyll a
Vertical profiles were measured for salinity, tempera-
ture, dissolved oxygen, pH, and conductivity. The June
1984 study was the most comprehensive. It was done
in conjunction with six other studies that included a
nonpolnt source survey, a nitrogen transformation
study, a sediment oxygen demand study, a use-at-
tainability study, a series of mini-surveys for in stiu
water quality parameters, and additional
hydrodynamic studies. All studies had municipal waste
load data, although only the 1984 study included a full
set of waste load data from all industrial discharges.;
The water quality studies showed that the ship channjel
below the salt water barrier was stratified with respect
to salinity and dissolved oxygen. The channel had onqe
been thermally stratified, but this had been reduced
because of the removal of cooling water discharges.;
The estuary water quality was characteristic of water
receiving wastewater effluent — high nitrite/nitratp,
phosphorus, and BOD, and low dissolved oxygen, in
the upper half of the estuary (below the saltwater
barrier) dissolved oxygen was below the State's 4.0
mg/l standard. Phosphorus and nitrogen concentra-
tions were characteristic of eutrophic conditions.
Phosphate ranged from 0.1 to 0.3 mg/l. Ammonia
concentrations ranged from 0 to 0.6 mg/l. Much of the
degraded water quality was from loading upstream of
the saltwater barrier. !
13.3. Model Application
13.3.1. Model Framework
The model selected for the Calcasieu was RECEIV-II. It
is a time-variable model developed from the receiving
water component of U.S. EPA's SWMM model. It was
modified by Raytheon (1974) for use on 28 New
England rivers and harbors. The 13 subroutines that
form the model remain compatible with SWMM, but
can be run independently. The model has the following
general characteristics:
• Time variable water quality and hydraulics
• Eleven water quality variables (conservative and
nonconservative)
• Link-node approach (vertically homogenous)
• Multiple tidal forcing points
The model has both a hydraulic and water quality
component. For hydraulics, the model uses a link-node
approach. Each node or junction is connected via links
or channels. The equation written for each link incor-
porates fluid resistance and wind stress using the Man-
ning and Ekman equations. Both components use a
finite difference solution. The hydraulic component
requires considerably more computer time than the
water quality component because computations are
performed for the entire system for time steps of five
minutes or less, whereas the water quality component
uses a one-hour time step.
For the Calcasieu, the RECEIV-II model framework was
used without major changes from that documented by
Raytheon. Certain changes to the FORTRAN source
code were required to tailor the hydrodynamic module
for site-specific characteristics.
13.3.2. Procedures
The model was calibrated with the data set from August
1979. It was verified using July 1978, July 1980, and
June 1984 data sets. The model was recalibrated by
revising the selection of model coefficients and extend-
ing the modeled area farther upstream at each tributary
to improve the representative network of water storage
in the system (Figure 13-2). The model has 67 load
sources, 162 links, and 114 nodes. The number of
cycles of the simulation were increased because the
short simulations of earlier modelers had not achieved
steady-state.
Model simulations were conducted at steady state for
several reasons. First, insufficient data were available
to calibrate or verify dynamic conditions. Second,
model projections were to be run at 7Q10 steady state
conditions. This assumption of steady state critical
conditions is consistent with regulatory policy to use
13-2
-------
Sourct: Duke, 1985
Figure 13-2. Model segmentation diagram [Duke (1985)].
conservative assumptions protective of the environ-
ment when dealing with model uncertainty.
Before proceeding with model calibration, the model
was tested to determine if the thirteen day simulation
used in earlier studies was a sufficient amount of time
to achieve steady-state conditions. The initial salinity
concentration was set to zero and a salinity wave was
propagated upstream from the Gulf, downstream from
the barrier, land from tributary inflows. After 13 days,
salinity was still simulated near zero, indicating a 13 day
cycle was not sufficient to achieve steady state. To
ensure steady-state conditions, Duke ran the simula-
tions for more than 900 days. This required ap-
proximately 4 hours of CPU time on the Louisiana DEP
Digital VAX 11/780 computer.
Model rate coefficients were first adjusted to best simu-
late the August 1979 calibration data set. When model
output matched observations within acceptable limits,
model verification simulations were tested. In model
verification, rate coefficients were identical to the
calibration, but environmental conditions and loadings
were adjusted to reflect the specific verification sur-
veys. These changes included:
• Tributary flows and loads
• Upstream flows and loads
• Waste discharge flows and loads
• Ambient temperatures
The model Was run for each verification survey and
compared with the field data. Whenever a model
parameter was changed during the verification, all data
sets were run again to ensure the change did not
significantly change the simulation of any data set.
The major coefficients are summarized in Table 13-1.
These values were changed spatially within each sur-
vey but not changed from survey to survey. Model
inputs for forcing conditions (e.g. tides, temperatures,
flow, etc.) and loading were as measured for each
survey.
13.3.3. Calibration/Verification
The results of the model calibration/verification are
summarized in a few representative plots. The calibra-
tion/verification was described as good for
hydrodynamics and fair for water quality. Obvious dis-
crepancies between the data and model were seen for
both selected hydrodynamics and water quality
simulations, but not viewed as a serious problem.
Problems with poorly defined loads was one com-
plicating factor. In addition, model predictions were for
steady state conditions, while observed data reflected
dynamic conditions.
Comparison of the results from the calibration and
verification simulations were divided into ship channel
simulations and other stations. The other stations in-
cluded the lake and loop areas. The results of the other
simulations were not presented by the author since
they were described as similar to the main ship chan-
nel. Also, tidal water quality calculations were per-
formed but only tidally averaged results were
compared to data.
Table 13-1. Values for Major Coefficients
Coefficient
Units
Range
Manning's n
Ammonia Oxidation
Nitrite Oxidation
BOD Oxidation
Benthic Oxygen Demand
Reaeration
none 0.018-0.035
per day 0.002-0.020
per day 1.00
per day 0.001-0.050
gm/sq.m/day 0.75-1.50
per day 0.003-2.000
13-3
-------
Cemlfon
Soli Water Sorrier
I f jj,
1 A f
HockbiHy
Measured
---•Simulated
Lake Chorln
sourcat DuXfl, 1985
Figure 13-3. Tidal stage results for August 1979
hydrodynamic calibration simulation
[Duke (1985)]
Hydrodynamics: !
The model calibration results for August 19|79
hydrodynamics are summarized in Figure 13-3 for five
stations below the salt water barrier. Model perfor-
mance was measured using water elevations. The
model was considered a satisfactory match to data
since the trends and timing were well matched. Tfie
elevation differences were considered insignificant.
The model verification comparisons were similar for
July 1980 and June 1984 in that the model matched tfie
trends well but was inconsistent in matching the mag-
nitude. However, for the July 1978 data set (see Figure
13-4) the cycles and magnitudes were poorly matched.
Overall, the hydrodynamic calibration/verification was
described in the final report as good.
Water Quality:
The water quality calibration/verification simulated tfie
ten parameters described above (Figures 13-5 through
13-9). Figures 13-5 and 13-6 are selected model com-
parisons for a few parameters from the August 1979
model calibration. Figures 13-7, 13-8, and 13-9 are
selected results for the three verification simulations.
Camiron
(no measured data)
Hoekberry
Calcasieu Lock
— Measured
— Simulated
sourcei Duke, 19S5
Figure 13-4. Tidal stage results for hydrodynamic verification
simulation (July 1978 data set) [Duke (1985)].
The water quality calibration/verification match was
characterized as fair, with many discrepancies at-
tributed to poor information on loading conditions and
dynamics.
13.3.4. Model Sensitivity
An important modeling activity is sensitivity analysis.
This procedure tests the sensitivity of model calcula-
tions to changes in selected inputs. Results can be
used to:
• Refine coefficient selection
• Identify the most important processes and loads
• Identify areas in need of better data to improve
modeling
• Define model uncertainty
The model was tested for an elimination and tripling of
BOD and ammonia deoxygenation rates and elimina-
tion of algae. The results indicated that algae had the
largest effect on the water quality calculations. This
finding is common to estuaries where algal abundance
often is the major factor in controlling water quality. As
a result, success or failure in model validation to data
13-4
-------
16
I'4
t12
| 10
OT
8
^ 6
*C
4
FLOW
— Meotured
— Simulated
• Measured
• Simulated
12
Q 8
Ul
tn
tn -
UJ o
a:
a.
% 4
-------
,-• -M ,;
s--.--'
10 15 ' 20 25
DISTANCE (milts)
33 '40
Figure 13-7. Selected water quality results for July 1978 verification simulation; salinity, dissolved oxygen, and biological oxygen
demand [Duke (1985)].
--. I
2.4
22
2.0
_ i.a
1«
x"l.4
z
1.2
1,0
0.8
O.S
tt4
02
0.0
I--...
-SlmulaUd |
Alt mtaiurtd dola ttiow«d ziro
IO 15 20 25 30 35 40
i DISTANCE (mllll)
Figure 13-8. Selected water quality results for July 1978 verification simulation; phosphate, total Kjeldhal nitrogen, and ammonia
[Duke (1985)].
13-6
-------
— Mtaiurtd
—Simulated
°o
- Mtaiurtd
"Simulated
10
20
30
35
15 20 25
DISTANCE (mil««)
Figure 13-9. Selected water quality results for July 1980 and June 1984 verification simulation: salinity and dissolved oxygen
[Duke (1985)].
can depend on proper characterization and simulation
of algal dynamics.
13.4. Total Maximum Daily Loads
The purpose of all modeling efforts on the Calcasieu
was to develop total maximum daily loads (TMDL) and
wasteload allocations. In the earliest study by Weston
(1974), the TMDL for the Calcasieu River was calcu-
lated to be 31', 190 pounds ultimate oxygen demand per
day (Ibs UOD/day). Fourteen municipal and industrial
dischargers were then allocated waste loads for BOD
andNH3-N.
In 1980, Hydroscience produced general recommen-
dations on waste load allocation rather than determine
specific TMDL They emphasized the need to regulate
the area with respect to dissolved oxygen. The study
concluded that background loads were so high that
even at zero discharge below the salt water barrier, a
DO standard of 4 mg/l would not be met. Despite the
lack of a TMDL from this modeling study, the 1980
Water Quality Management Plan for the State of
Louisiana listed a TMDL for the Calcasieu River of
52,760 Ibs UOD/day based on a dissolved oxygen
standard of 4 mg/l. The second Weston study that
followed the AWARE 1981 modeling agreed with the
Hydroscience report, computing a zero TMDL because
of a violation of the standard at zero discharge. A use
attainability study (Thompson and Fitzhugh, 1986)
demonstrated that waters above the salt water barrier
are naturally dystrophic. The Duke study, in concur-
rence with the State and EPA, developed the TMDL to
protect against the oxygen sag which occurs near river
mile 26. Using the 4 mg/l DO standard and 1979 loading
pattern, the Duke study produced an estimated TMDL
of 83,130 Ibs UOD/day.
13-7
-------
13.5. References
Duke, James H., Jr., 1985. "Calcasieu River Basin,
Louisiana, Modeling Study", report prepared by James
H. Duke, Jr., Ph.D., P.E., Consulting Water Engineer,
Austin, Texas, for the Department of Environmental
Quality, State of Louisiana.
Raytheon Company, 1974. "New England River Basins
Modeling Project, Documentation Report, Volume, 1",
draft report submitted to the U.S. Environmental
Protection Agency, Office of Water Programs under
Contract No. 68-01-1890, Program Element 2BH149.
Thompson, B.A., and G.R. Fitzhugh, 1986. "A Use
Attainability Study: An Evaluation of Fish and Macrpin-
vertebrate Assemblages of the Lower Calcasieu RK/er,
Louisiana," prepared for Lousiana DEQ Office of Water
Resources.
NOAA, "National Estuarine Inventory Date Atlas, Physi-
cal and Hydrologic Characteristics", Strategic Assess-
ments Branch, Ocean Assessments Division, 1985[
13-8
-------
14. Expert Critique of Case Studies
Estuarine modeling is a complex and evolving science.
As such, there is not total agreement among experts in
the field regarding the "proper" approach to estuarine
waste load allocation modeling. This chapter presents
the opinions of three nationally recognized experts in
estuarine modeling. These experts were asked to pro-
vide their thoughts on the proper approach to estuarine
WLA modeling in general and to the case studies
provided in this guidance manual in particular.lt should
be noted that the case study critiques are based
primarily upon the studies as described in this docu-
ment. They do not necessarily consider potentially
important factors such as resources or time available
to perform the modeling.
The reader is encouraged to examine these reviews
and to compare and contrast the expert opinions.
While all three experts are in agreement with the basic
guidance provided in Parts 1 through 3 of this manual
(each having served as a technical reviewer), their
specific approach to estuarine WLA is seen to differ.
Readers should therefore be aware that while this
manual provides a general background to estuarine
modeling, the exact approach to be taken for any given
site still requires some subjective assumptions.
14.1. Robert V. Thomann, Ph.D.
Professor, Environmental Engineering and Science
Manhattan College
Riverdale, New York 10471
14.1.1. Introduction
My overall opinion on the appropriate level of estuarine
water quality model complexity can be summarized by
the observation that:
THE BEST MODELS ARE OFTEN THE SIMPLEST
The review therefore will continually display a bias
towards doing estuarine water quality modeling in as
simple a fashion as possible and only after all simplicity
has been exhausted, should increasing complexity be
introduced and then only after careful consideration is
given to the improvements in the model that might be
realized. The reasons forthis bias are: (a) most analysts
have only limited experience, time and resources avail-
able, and (b) unnecessarily complex models some-
times tend to obscure uncertainty behind a facade of
"reality."
The choice of the appropriate level of model com-
plexity is determined in large measure by the nature of
the problem under investigation. The context for my
opinion on an appropriate level of model complexity is
the establishment of a defensible analysis framework
for a Waste Load Allocation (WLA). The opinion is not
directed toward model development in a research con-
text. This is not to say that one need not pay any
attention at all to the scientific correctness of the
model. Rather, modeling for WLA purposes imposes a
separate, but related set of constraints on the model
construction and development.
The assignment of a WLA to a particular discharger or
regional group of dischargers involves a determination
of the level of treatment over and above secondary
treatment and/or Best Practical Treatment (BPT) and
Best Available Treatment (BAT) coupled with a
specification of the allowable mass loading and/or
effluent concentration. Nonpointand transient sources
may also be a part of the WLA. The primary thrust of
modeling then for WLA purposes is from a control
engineering point of view. The modeling is not neces-
sarily conducted for a detailed understanding of the
various interactive processes that may be operative
(e.g., the dynamic behavior of nitrifying bacteria), but
rather an engineering-scientific approximation to the
real estuary which will provide a firm basis for the WLA.
Therein lies the difficulty.
The analyst must make a delicate determination be-
tween the degree of complexity necessary for a defen-
sible WLA, the time frame and budget available for
completion of the WLA and the natural urge to continue
to explore various components of the problem. Be-
cause of the skill needed to make this determination
and the limited resources that are usually available, I
would generally lean in the direction of more simple
models rather than more complex models.
A. The Difference Between a Site-Specific Model and a
Generic Model
One of the more troublesome aspects of contemporary
estuarine modeling is the confusion that exists be-
tween (1) a mathematical model of a particular estuary
with its unique setting and (2) a generic non-site-
specific model embodied in a computer code that
incorporates the principal components of water quality
theory but in a completely general way. For purposes
Of this opinion, a model is defined as the application of
accepted principles of water quality fate, transport and
transformation theory, together with appropriate deter-
mination of site-specific parameters to predict water
quality under some future conditions for the given
estuary. A generic model is considered to be a general
programming framework wh|ch also incorporates the
14-1
-------
basic theoretical components, but has no utility in a
WLA until applied to a specific problem setting. J he
computer code of a generic model is transportable, a
model of a given estuary is not.
Thus, ft does not make much sense to refer to models
of Boston Harbor and Appalachicola Bay as "WASP 4"
models. The WASP 4 computing framework may have
been used in both cases, but any other suitable com-
puter program (with similar fate and transport proces-
ses) could have been used as well. The structuring of
a water quality model for Boston Harbor requires much
more than a simple choice of computer code. This
opinion on model complexity is not directed therefore
to issues associated with how to choose an Ap-
propriate computer code. Instead, my opinion is
focused on the issues associated with determining the
level of complexity for modeling a specific estuary or
coastal water body always in the context of a WLAt
B. Analytical and Numerical Models
There are fundamentally two types of water quality
models: analytical models where the solutions to a
differential equation or set of differential equations jare
available, and numerical models where approxima-
tions are made to the derivatives of the operative!dif-
ferential equations. Analytical models are available
only for relatively restrictive conditions, usually one
dimensional, constant parameters and steady state,
although solutions for some time variable inputs e^cist,
again for restrictive situations.
It Is Interesting to note that the accompanying case
studies do not indicate any use of analytical solutions
to determine initial expected responses or to check on
numerical model results. I do know, however, jthat
analytical solutions were used for Saginaw Bay as a
completely mixed bay exchanging with Lake Huron
and the results provided important initial guidance for
further mode! development. Similarly, analytical splu-
tlons were often used in the Potomac case to check on
numerical model output in the initial stages of model
construction. One wonders whether some of the
calibration difficulties of some of the case studies
would not have been alleviated by initial analytical
checks on the order of water quality response to "cjose
In" on which particular phenomena were of importance
In describing the observed data.
In spite of the, severe assumptions that must be in-
voked, ft Is strongly suggested that:
Such computations provide the first approximations to
the order of water quality response that might be ex-
pected from input loading under different hydrological
regimes and model parameters. Also, the use of
analytical models provides a first order check on more
complicated numerical models to determine whether
the numerical computations are approximately cor-
rect.
C. Model Evolution
The use of models in decision making must recognize
that, very often, the understanding of estuarine proces-
ses, and the availability of data and model frameworks
for a given estuary are always changing. Models are
not static, but rather continually evolving. Decision
makers must be apprised of this fact and must, to some
degree, be prepared for new input into the decision
process.
The Saginaw Bay and Potomac estuary case studies
are good examples of models that began at relatively
simple levels of complexity and have subsequently
progressed to more complex kinetics and spatial and
temporal detail. The progression was dictated by an
ever increasing level of complexity in the questions
being asked of the model. For example, the early
Potomac estuary models did not explicitly include
phytoplankton dynamics. But after issues of nutrient
controls (e.g. should phosphorus or nitrogen be
removed?) were raised, an expansion of existing
models was required. However, as noted below, it is
not always clear that adding additional complexity
improves credibility. Thus, for the Saginaw Bay model,
ft is not clear that the addition of an internal nutrient
pool state variable improved the model performance,
whereas the inclusion of phytoplankton functional
groups was important in predicting the occurrence of
nuisance odors.
The Calcasieu estuary case study, on the other hand,
seems to be an example of a modeling framework that
needs to be substantially restructured (e.g. inclusion of
a vertical dimension and non-steady state) in order to
provide more credible results. Yet the original model
(albeit with some updates) continued to be used with
results that were less than desirable.
It should then be clearly recognized by all concerned
(decision makers, model analysts and scientists and
engineers) that:
ANALYTICAL SOLUTIONS SHOULD BE USED
TO COMPUTE INITIAL RESPONSE AND TEST
NUMERICAL MODEL COMPUTATIONS.
ALL MODELS MUST CONTINUALLY BE
UPDATED: IF NOT, MODEL "ATROPHY" SETS
IN AND CREDIBILITY DETERIORATES.
ESTIMATED MODEL "HALF-LIFE"
IS ABOUT 1-2 YEARS.
14-2
-------
a
g
cJ
Q
o
Optimum Cred%;iily/Comple»ity
LEVEL OF MODEL COMPLEXITY
TEMPORAL SCALE
SPATIAL SCALE
KINETIC INTERACTIONS
Figure 14-1. Illustration of relationship between model credibility and model complexity.
Existing models must therefore never be "frozen" in
time and continue to be used in the face of obvious
model inadequacies. As painful as it may be, some
model frameworks need to be restructured, expanded
or even abandoned as new information becomes avail-
able.
14.1.2. Appropriate Spatial and Temporal Scales
Unfortunately, because of the ready availability of com-
puter programs that are fully time variable and three
dimensional, there is a tendency to believe that more
complexity is better since it approaches the real world
more closely. But, increasing complexity does not
usually result in increased model credibility. Figure
14-1 illustrates this opinion. In general, increasing
model complexity requires specification of more and
more parameters and state variables, both in absolute
number and over space and time. Even more impor-
tantly, increased model complexity requires a detailed
data base across all state variables and over space and
time for a complete assessment of model adequacy.
As a result, what appears to be more realistic is actually
a model that has hidden within it a large degree of
uncertainty. Because of a generally sparse data base,
the uncertainty is not visible and it is assumed that the
model is more realistic when in fact it is not.
On the other hand, the model may be so crude in
spatial, temporal or kinetic definition that key
mechanisms or issues associated with the problem are
completely missed. Thus, a representation of a lon-
gitudinal estuary as a single completely mixed body of
water is quite inappropriate since the impact of a load
over distance is lost. Similarly, a steady state ap-
proximation may be completely incorrect because of
the dynamic nature of the problem (e.g. time variable
phytoplankton behavior).
The "art" of water quality modeling in general, is to
carefully evaluate the relevant scales of the problem.
This evaluation requires an assessment of the requisite
degree of complexity as opposed to merely assuming
that fully time variable, fine space scale models with
extensive kinetic detail are always the best choice.
A. Temporal Scale Issues
Estuaries exhibit a variety of time scales: hour to hour,
tidal and diurnal fluctuations, week to week and
seasonal variations and year to year differences. From
a modeling point of view, what are the choices? One
can try to represent the entire time spectrum from short
term to long term behavior, but this is clearly impracti-
cal. A model may concentrate on the short term, intra-
tidal and diurnal variations, with a possible loss of focus
on the longer term phenomena. Conversely, a steady
state model may miss the transient effects of storm
water inputs or transient hydrologic events. The choice
of relevant model temporal scale in my opinion centers
about the use of estuarine modeling for WLA purposes.
A WLA may be a constant (over time) effluent con-
centration or a seasonal variation may be allowed (as
in seasonal nitrification). These specifications are
usually assigned to meet water quality objectives
during some critical flow and temperature period. It is
not usual to assign a WLA on a short time scale with
the exception of a probabilisitic assignment of maxi-
mum values not be exceeded. Also, WLA analyses
often need to be conducted with relatively limited data,
which are usually not of sufficient density in time and
-14-3
-------
[
space to calibrate a fully time variable intra-tidal model.
Rather, data are more frequently available at irregular
time intervals, but with some spatial definition. Finally,
developing fully time variable models at an intra-tidal
level is a complex time consuming effort with a neces-
sity to conduct extensive data analysis and output
processing in orderto display model results in a defen-
sible manner.
The case studies show a range of temporal scales, from
the steady state analyses of the Calcasieu estuary, to
the intra-tidal models of the Manasquan estuary ahd
DO and total residual chlorine in the Potomac estuary.
The intra-tidal choice for the Manasquan (over tl/vo
4-day periods) is not considered to be the correct
choice since the water quality problem under stddy
involved kinetic behavior overtime scales of weeks and
seasons. Key behavior is therefore not captured by the
temporal scale of the Manasquan model. Also, the fact
that intra-tidal computations were performed does not,
In itself, provide for an accurate representation of the
actual variability in the data. Indeed, it is not clear from
the comparisons to data presented in the case study
that the intra-tidal calculations captured the actual
variability with any substantive degree of success.
The choice of an intra-tidal scale for the total residual
chlorine in the Potomac is correct since the kinetics of
the disappearance of chlorine are quite rapid. The time
variable behavior of the chlorine state variable thus
needs to be calculated over short time intervals in order
to model the expected transient behavior.
B. Spatial Scale Issues
The choice of spatial dimensionality and scale involves
evaluation of available data (to determine significant
gradients) and the expected geographical extent of ithe
problem. The fineness of the spatial extent of the model
is to some degreecoupled to thetemporal issues nojted
above. Generally, long time scale problems may; in-
volve larger scales and less detailed spatial definition.
The chlorine model of the Potomac is an example of
where cross-estuary gradients needed to be computed
necessitating a spatially detailed model in the lateral
and longitudinal direction. The Saginaw Bay model
consisting of five segments is a good example of a
reasonable grid since a finer spatial definition would
probably not contribute to any improved model
credibility.
Finally, a remark should be made about model bpun-
daries. The extent of the model should alwaysj be
sufficiently far removed from any existing or proposed
Inputs that may be subject to a WLA. The boundaries
should be at a point where the flows and exchanges
and state variable concentrations can be specified and
are independent of the model output. For example, it
is not entirely clear from the Manasquan case study
that the model boundary is proper, i.e. the extent of the
model may have to be extended out past the inlet in
order to provide a proper independent boundary con-
dition. This may be especially true if the model had ever
been used for analysis of the proposed regional input
at the head of tide.
C. Suggested Strategy for Temporal-Spatial Scales
Since the principal reason for estuary modeling in the
context of this opinion is a WLA, the following strategy
for choosing a relevant temporal-spatial modeling
scale is offered.
TEMPORAL-SPATIAL SCALES
BEGIN WITH STEADY STATE,
"LARGE" SPACE SCALES,
THEN SEASONAL,
MORE DETAILED SPATIAL DEFINITION,
THEN INTRA-TIDAL, FINE SCALE.
It is suggested that the temporal scale of most WLA
estuarine models should begin at steady state to deter-
mine overall relationships between input loads and
resulting water quality. Steady state is suggested even
for highly reactive variables since the steady state
modeling helps to define overall response levels and
spatial extent of the input loadings.
Following steady state analyses, if time variable
analyses need to be done in estuaries (as a result, e.g.
of a need to specify phytoplankton dynamics for
nutrient control or a seasonal WLA) then a seasonal
time scale (with a model framework representing an
average over a tidal cycle) should be used.
Only if the justification is quite clear, (e.g., transient
storm water input analyses or a complicated
hydrodynamic regime as in the Potomac estuary
chlorine model) should an intra-tidal model be con-
structed. The fact that the estuary has a tidal oscillation
is in itself not justification for constructing an intra-tidal
model. The reason is threefold: (a) as noted earlier, the
focus here is on WLA problems which are normally
limited in resources, time and money, (b) most WLA
problems involve processes that have longer time
scales than tidal, and (c), there are many other sources
of temporal variability in water quality that are not
captured by intra-tidal calculations (e.g. hour to hour
and highly local changes in solar radiation, suspended
solids, wind, or velocity, among others).
It is suggested that initially a relatively crude spatial
representation (e.g. a numerical grid size of several
miles) be used for estuaries in the longitudinal direction
14-4
-------
in order to provide a rapid understanding of the ex-
pected order of water quality variations. If vertical
gradients are significant, the model should include a
vertical dimension at the outset. Only if warranted by
the problem context should a spatially detailed (e.g. on
the order of hundreds of feet) model be constructed.
14.1.3. Need for Hydrodynamic Models
Several of the case studies (e.g., Potomac, Manasquan
and Calcasieu) make use of hydrodynamic models.
Indeed, the case study reviews seem to imply rather
consistently that a water quality model is always better
when a hydrodynamic model is included. I do not agree
that this is always true. It seems that a mathematical
model of the hydrodynamics of the estuarine system is
necessary when:
(a) the transport regime is complex in space and time
and cannot be easily specified a priori,
(b) the transport regime will be changed under some
future WLA condition, such as occasioned by channel
deepening or straightening, or construction of barriers
(c) the absence of hydrodynamic model would weaken
water quality model credibility in the eyes of a peer
scientific review.
It is not clear that hydrodynamic modeling was crucial
and essential for the Potomac DO and the Manasquan
models. Indeed, the issues of water quality model
credibility for a WLA often have little to do with the
hydrodynamic calculation. Rather model credibility
centers around (a) issues of water quality model
calibration that do not depend on hydrodynamics (e.g.
parameter specification), (b) inclusion of correct
mechanisms (e.g. appropriate state variable or sedi-
ment source/sink interactions) and (c) point and non-
point input load estimates. An alternate to a full
hydrodynamic model calculation on an intra-tidal basis
is to calculate the net transport from the fresh water
flow and estimate tidal dispersion coefficients by using
salinity (or dye) as a tracer. Many estuarine WLA
models have been successfully constructed using this
type of average across tide approach.
14.1.4. Appropriate Level of Kinetic Complexity
In addition to temporal and spatial issues, one must
also consider the need to include various levels of
kinetic complexity in the model. Specifically a choice
must be made of the relevant state variables to be
included in the model and the nature of the interaction
between the state variables. For example, for a DO
model, should phytoplankton be explicitly modeled or
input? For a, phytoplankton model, should various
functional groups be modeled or should total
chlorophyll be used? Should sediment nutrient fluxes
be calculated or input?
As a general rule, I would advise to:
KEEP STATE VARIABLES TO A MINIMUM;
MODEL ONLY THOSE FOR
WHICH DATA EXISTS;
BUT ALWAYS INCLUDE THOSE STATE
VARIABLES WHICH WILL BE
IMPACTED BY A WLA,
The case studies seem to have implicitly recognized
this general rule, although there are some exceptions.
The Manasquan model is clearly over-specified with
state variables and kinetic interactions for the nature of
the problem under study and the available data sets.
The inability to calibrate to the phytoplankton state
variables severely limits the utility of the model.
On the other hand, the initial Potomac estuary DO
model did not explicitly include organic nitrogen, nor
ammonia uptake by phytoplankton. Also photosyn-
thetic DO sources and sinks were externally inputted,
but these inputs were to be extensively impacted by a
WLA for nutrients. The model could not therefore
respond to the WLA questions associated with the
affect of nutrient control on DO.
Sometimes a state variable must be included even if
data are not available. For example, for a toxic chemi-
cal model, both dissolved and paniculate chemical
must be modeled. But data may not be available for the
dissolved component because of concentrations
below a detection limit. Nevertheless, both com-
ponents need to be included in the modeling
framework.
14.1.5. Calibration and Verification Issues
Of course, all of the above only has relevance when the
model is considered to be "representative" of the ac-
tual estuary. Thus, the question of the calibration and
verification of the model must be addressed. This is an
area about which much has been written and dis-
cussed for several decades, all centered about the
issue of whether a model has adequately reproduced
the observed data.
A. When Is A Mode! "Calibrated" And "Verified?"
In my opinion, a model is considered representative of
the real estuary when the key model state variables
reproduce the observed data over a range of expected
conditions and within expected statistical variability. Of
course, this definition may not help at all. For example,
what is the "expected statistical variability?" Perhaps
the only answer is that model "unrepresentativeness"
is obvious. We know when a model is not repre-
14-5
-------
SIMPLE MODEL
Analyticol Solutions
Numerical Model
Steady State
•*"•
MODEL CREDIBILITY
CALIBRATION -VERIFICATION
Temporol/Spatiol Comparisons
Statistical Tests/Comparisons
INCREASE MODEL COMPLEXITY
State Variables
Kinetic Mechanisms
Temporal/Spatial Detail
Marginal
Sufficient
WASTE LOAD
ALLOCATION ANALYSIS
Critical Conditions
Sensitivity Analysis
Load Projection.
Components Analysis
ALLOWABLE LOAD
PERMIT
Rgure 14-2. Suggested strategy for determining appropriate level of model complexity.
sentative. The Calcasieu case study is offered as an
example of a model that is claimed bythe analyst tp be
"good" for the hydrodynamic model and "fair" for the
water quality model. But even a casual examination of
the model comparisons to data indicate severe
problems. The DO profile is not captured and a sag is
calculated where it does not exist. This, in my opirjiion,
is "unrepresentative" and outside the bounds of statis-
tical variability.
Similarly, the Manasquan model simply fails in several
state variables to bound the data. Also, the spatial
profiles for this case are not presented so one cajinot
judge the adequacy of the model in reproducing I lon-
gitudinal variability.
The Potomac estuary DO model was compared to
various data sets by readjusting the model parameters
for each calibration. This is unacceptable. The purpose
of model calibration and verification is not to "force fit"
the model to the data. Rather, the model parameter
numerical assignment should obey the ;
PRINCIPLE OF PARSIMONY: L
BE "STINGY" WITH THE SPATIAL AND TEM-
PORAL SPECIFICATION OF MODEL 7
PARAMETERS AND HAVE A REASON [
FOR ALL ASSIGNMENTS.
The Potomac estuary phytoplankton and Saginaw Bay
models offer extensive calibration and verification
analyses, including various statistical measures of
comparisons. Both spatial and temporal comparisons
and statistics of comparisons are given. These case
studies provide some measure then of an adequate
representation of the data by the model and can be
profitably used as a "model" of a model calibration.
Two caveats are in order, however: (1) extensive data
sets and resources were available in both cases, and
(2) even with the extensive calibration and verification
of the Potomac eutrophication model, a bloom in 1983
was not captured because of presumed pH mediated
sediment phosphorus release, a mechanism not pre-
viously included in the model.
14.1.6. Summary
Figure 14-2 summarizes all of the above comments.
As indicated, the suggested procedure is to begin with
simple representations of the estuarine system. This
should always include some investigation of the es-
tuary water quality problem with analytical solutions.
This is true for all problem contexts. For DO, simple
steady state solutions should be used to provide es-
timates of the impact on carbonaceous and
nitrogenous loads, sediment oxygen demand, and
photosynthesis and respiration on the DO. For nutrient
problems, total nutrient calculations should be per-
14-6
-------
formed to determine importance of sediment fluxes
and net loss from the water column. For toxics
problems, total, paniculate and dissolved chemical
can be easily estimated.
If the estuarine system is too complex for initial analyti-
cal solutions (e.g. when vertical and lateral gradients
must be defined) then a steady state numerical model
is recommended. The spatial definition is determined
from the gradients that need to be captured.
Following the structuring of the simple model, initial
determinations should be made of the model
credibility. Comparisons to data should be presented
over the spatial dimensions of the model. Where ap-
propriate, statistical measure of model adequacy
should be computed.
The degree of model credibility should then be as-
sessed in the, light of the WLA.
THE SIMPLE MODEL MAY TURN OUT TO BE
ENTIRELY SUFFICIENT FOR WLA PURPOSES.
If a determination is made that the simple model
provides only "marginal" model credibility, then model
complexity should be increased. This increase in
model complexity often needs to proceed in the follow-
ing order: (a), additional state variables, (b) additional
kinetic interactions, (c) increased temporal and spatial
definition. It is in the latter that hydrodynamic modeling
may be necessary.
Additional calibration and verification is then con-
ducted with the hope that model credibility is in-
creased. This step should include, whenever possible,
comparisons to data sets collected over a range of
environmental and input loading conditions.
After a determination has been made, then a full WLA
analysis can be conducted. This analysis should in-
clude evaluation of water quality response under criti-
cal design conditions, sensitivity analysis, projection of
expected loads in the future and components analysis
of individual inputs. This latter analysis is aimed at
describing the relative contribution to the calculated
response from individual components, e.g. particular
point source inputs, and distributed sources (such as
sediment sources). The analysis often provides key
insights into which inputs and mechanisms are most
important in the WLA. (None of the case studies dis-
played any components analysis).
The final outcome is then the recommended WLA for
an input or region with associated permit specifica-
tions. It is this final outcome that should always be kept
in perspective when assessing the need for various
levels of model complexity. Ultimately, of course, the
measure of success of the model is the degree to which
the model projections are actually realized after the
WLA has been implemented. But that is a topic for
another opinion at a different time.
74.17. Case Study Review
Case Study 1 - Saginaw Bay
This case study is a very good example of a proper mix
of spatial and temporal specification together with
proper representation of kinetic detail. Illustrations of
the extensive calibration of the model are given and the
post audit of the model is unique. The statistical com-
parison between model output and data as shown in
Table 10-4 is a very good example of what should be
expected from a water quality model.
The use of a five segment model is entirely appropriate
since the proper exchanges and transport were deter-
mined from measured salt concentrations. In this
reviewer's opinion, a representation of the system with
a finer grid operating at finer time and space scales
would not improve the model performance and indeed
may have considerably delayed and obscured the
interpretation of model output.
It is concluded that the overall analysis of Saginaw Bay
eutrophication as given in this case study is a paradigm
analysis for water quality modeling. The modeling
provided considerable insight into the dynamic be-
havior of phytoplankton functional groups, incor-
porated a detailed calibration and verification analysis
and uniquely conducted a post-audit analysis after
nutrient controls were implemented.
Case Study 2 - Potomac Estuary
This case study, a summary of three efforts on the
Potomac estuary, illustrates a range of modeling ap-
proaches to estuarine water quality.
DYNAMIC ESTUARY MODEL
The first effort, the use of the Dynamic Estuary Model
(DEM) examined the DO resources of the estuary. The
one dimensional hydrodynamic model was used to
provide the transport and was calibrated to hydraulic
properties as well as the longitudinal extent of chloride
concentration in the estuary. This effort is a good
example of calibration of the model to observed data,
but also indicates the hazards of calibration where the
underlying kinetic structure is too simple. The DO
calibration reset initial conditions for each survey. This
is not considered a proper calibration method. As
inidicated during a post-audit, the DEM failed to
properly account for nitrification phenomena by as-
suming that all ammonia that was lost was due to
nitrification, rather than through some measure of up-
14-7
-------
take by the phytoplankton. The intratidal
hydrodynamic model, while initially appearing to pro-
vide a more realistic "real time" modeling framework,
In actuality added little to understanding of the overall
water quality behavior of the estuary.
The history of the DEM is a useful example of model
evolution in the midst of decision making. With initial
emphasis on intra-tidal calculations to a shift towards
more detailed kinetic evaluations during the post audit
stage, the DEM illustrates the need to properly include
necessary phenomena that link various water quality
constituents.
POTOMAC EUTROPHICATION MODEL
The Potomac Eutrophication Model (PEM) is an ex-
ample of an intermediate scale of estuarine water
quality model. The use of a coarse grid in the lower
estuary was justified on the basis of the lack of; any
significant gradients in water quality constituents of
interest. Vertical homogeneity is a key assumptio^and
undoubtedly influenced the ability of the model to
properly calculate water quality in the region of the
turbidity maximum. This time variable model (on a time
scale of weeks to seasons) properly did not rely on a
detailed intra-tidal hydrodynamic calculation on a: fine
time and space scale. Emphasis was rather placed on
the role of the sediment on the overlying nutrient con-
centrations and the interactions of the various nutrient
forms with phytoplankton and DO.
The PEM study is a good example of extensive calibra-
tion and verification analyses, illustrations of which are
shown in the case summary. Also, the PEM analysis
made use of extensive statistical comparisons '(see
11.4.6) between the model output and the observed
data.
Like the DEM, the PEM was subjected to a post audit
analysis. The analysis was prompted by a major algal
bloom In the summer of 1983. As noted in the case
study summary, pg. 11 -12 ff., the PEM was not able to
predict the full extent of the observed bloom, due in
some measure to a significant source of phosphbrus
that was not incorporated in PEM. Subsequent work
indicated that such a source may have been from k pH
mediated release of sediment phosphorus. Additional
input may have resulted from upstream transport of
phosphorus from downstream bottom waters. 'This
latter effect was also not included in PEM because of
the vertically homogeneous nature of the model.
Overall, the PEM is a good example of calibration and
verification of a time variable eutrophication estuarine
system. It also illustrates the hazards of apparently
"best" calibration of the model that misses a
phenomena which only appears after certain condi-
tions ensue. Nevertheless, the PEM proved useful in a
variety of decision making contexts, not the least of
which was to assess the reasons for the major 1983
algal bloom.
FINITE ELEMENT CHLORINE MODEL
This model is a very good example of the proper choice
of time and space scales. Because the decay rate of
chlorine is so rapid, the zone of influence of the chlorine
residual would be expected to be highly local. As a
result, this model has as its spatial focus a region of
about five miles centered at the location of the major
input. Detailed lateral specification is required because
of the need to calculate lateral movement of the
chlorine. Model calibration of transport and dispersion
was first accomplished by comparisons to dye study
results. The results shown in Figure 11-14 are a good
example of what one can expect. The general shape is
captured, but not all of the details even though the grid
is relatively fine. As noted in the text, further work using
dye decay would be necessary to improve the calibra-
tion. It was concluded however, that the dispersion was
properly captured in general.
That conclusion is a good example of a judgement
made by the analysts on the suitability of a model
calculation. This reviewer believes that the judgement
made here is correct, but only because of the calibra-
tion analysis to the observed dye data.
A similar conclusion can be drawn with respect to the
calibration of the total residual chlorine model to survey
data. What was required here was approximate repre-
sentation of the general field of the chlorine, to ap-
proximate order of magnitude. This was achieved.
More importantly, the sensitivity analysis indicated the
degree of model uncertainty and this is clearly dis-
cussed. That uncertainty did not affect the basic con-
clusions.
SUMMARY
These three modeling efforts of the Potomac estuary
water quality illustrate a good range of spatial and
temporal scales, level of model complexity and the
need for extensive calibration and verification to ob-
served data. Two major points seem to emerge:
• Uncertainty in the model coefficients sometimes
does not affect the conclusions, i.e., the
decisions that are reached. But under certain
situations, a model that is believed to be proper-
ly calibrated can miss entire phenomena or
linkages. Such a model may then fail in varying
degrees during a post audit. The experience of
the Potomac esturary models summarized in
14-8
-------
this case study should be borne in mind by any
analyst.
• Each problem requires its own spatial, temporal
and kinetic level of detail. Finer spatial and tem-
poral resolution is often not the issue especially
when the problem context is over a larger time
and space scale. Funding and project comple-
tion times are realities that must be faced in any
modeling effort. Such constraints must be
balanced against more and more detail in the
modeling framework with perhaps less than
desired return in improving the certainty of
decision making.
Case Study 3 - Manasquan Estuary
This model illustrates the use of an intra-tidal calcula-
tion to describe estuarine water quality. This reviewer
believes that the proper temporal scale was not used.
By focusing in on two 4 day periods as examples of
"calibration" and "verification," the model does not
capture the longer term, i.e., week to month, behavior
of the water quality constituents of interest. Further, the
analysis is flawed in several ways. The August 1980
period is used as a calibration set and July 1980 is used
as a verification data set. What would be much more
convincing is to use the model in one complete calcula-
tion extending from prior to July 1980 through the
August 1980 data. By restarting the calculation each
time before August and July and then extending the
calculation for only four days, the credibility of the
model is severely compromised.
Also, this model is presented as a demonstration of a
"successful calibration and verification of a real-time
estuary model." This reviewer does not agree that this
model is successfully calibrated and verified even fora
brief period of four days. The "real time" model is
presented in a fashion that seems to indicate that
because the model calculated at a time scale of hours
or less that it is more realistic than averaged models.
Ostensibly, the "real time model was selected to
predict photosynthesis effects on diurnal DO." But the
model fails to reproduce the observed DO range (see
e.g., Figure 12-9 and 12-10). Also, the CBOD, NBOD
and nitrogen forms are not calibrated. For example,
Figure 12-18 shows comparisons of computed
nitrogen forms to observed data. The computed forms
vary approximately sinusoidally with an apparent look
of reality and,certainty. But the comparison to the
observed ammonium data, for example, show some
significant over-calculation of the data. One wonders
how well the model would have done if the model were
not restarted for the July 1980 data set but rather was
run for a several month period.
It is recognized that this model was apparently con-
structed with only limited data and under apparently
tight constraints. As such, the exercise is useful in
showing how a model can be used to delineate data
and input load deficiencies. However, the modeling
framework is not considered to be adequately
calibrated and verified over the time scales necessary
for the water quality constituents under investigation.
The model spatial extent may also be inadequate for
evaluating certain alternatives and may have to be
extended into the ocean.
Case Study 4-Calcasieu River Estuary
This case study is adequately presented as an example
of a modeling context with problems in credibility and
in application. The modeling structure is flawed in not
adequately representing phytoplankton interactions
on the DO, no settling of paniculate forms and a lack
of vertical detail. (No data are presented however to
indicate the extent of any vertical stratification in salinity
or DO). The model is not considered to be adequately
calibrated and verified because of a failure to capture
the salinity and DO profiles on several occasions. More
critically, the conclusion on a total maximum daily load
of 83,130 Ibs UOD/day is not justified by the model
analysis. Since the data already indicate DO violations
below a standard of 4 mg/L, it is hard to see how the
stated allowable load was determined.
This case study should be seen as an example of model
evolution under different analysts with final results that
are marginal at best. The difficulty stems from differen-
ces in the opinions of analysts as to what constitutes a
satisfactorily calibrated and verified model. One
analyst described the hydrodynamic calibration and
verification as good, but this reviewer sees a very poor
comparison. At several of the stations, the computed
stage differs from the observed stage by several feet,
an apparent clear inability of the model to properly
represent the easiest of hydrodynamic variables. The
adequacy of the hydrodynamic model can also be
judged by examination of the salinity profiles which are
erratic in comparison to observed data. For example,
the July 1978 salinity profile is adequately captured, but
the computed July 1980 profile is significantly below
the observed data. A zero DO concentration is calcu-
lated in this vicinity that is not representative of the
observed data.
In general, this case study indicates a modeling
framework that is not entirely credible and as such, the
application to a waste load allocation is somewhat
problematical. The inconsistency of the computed al-
lowable UOD load with the observed data, as noted
above, is illustrative of the tenuous nature of the model
for use in decision making.
14-9
-------
14.2. Donald R.R Harleman, Ph.D.
Professor, Department of Civil Engineering
Massachusetts Institute of Technology
Cambridge, Massachusetts '
14.21. Introduction
The concept of a technical manual for performing
wasteload allocation in estuaries is an excellent o^e.
Part 4 of the manual is intended to be a "critical review
of estuarine wasteload allocation modeling." It consjsts
of four case studies "representing various levels of
complexity.11 The task assigned to the reviewer is to
provide a general discussion of the "appropriate level
of estuarine model complexity" and to comment on ;the
case studies within the context of the reviewer's
philosophy of environmental modeling.
14.2.2. Statement of Ihe Problem
Many environmental problems require the develpp-
ment of models in order to answer management ques-
tions related to the effectiveness of various control
scenarios. Such an effort requires a careful statement
of the problem and applicable regulatory constraints.
Decision makers need to be able to assess the impor-
tance of controlling point or non-point sources of pol-
lutants, and they usually need to know the time scale
at which the estuary can be expected to respond tothe
Implementation of source controls. Effective environ-
mental modeling would avoid the Lake Mead fiasco
where the City of Las Vegas operates a tertiary waste
treatment plant designed to minimize phosphorus dis-
charges to Lake Mead while the Fish and Wildlife Ser-
vice periodically adds phosphorus to the lake to
promote the growth of fish.
14.2.3. Data
Available data, both hydrodynamic and water quality,
must be studied in order to understand the spatial
complexity of the problem. In the hydrodynamic aj-ea,
ft is Important to understand the factors influencing the
currents and circulation pattern. These include: [the
degree of vertical stratification within the salinity in-
trusion zone, the extent of changes in longitudinal
salinity intrusion due to tides, wind and seasonal chan-
ges in fresh water inflows, and the degree of lateral
stratification due to fresh water inputs from tributaries
located on one side of the estuary. Temperature
stratification may also influence the vertical mixing iand
circulation pattern. The main stem of Chesapeake Bay
Is an excellent example of an estuary with distinctly
three-dimensional characteristics.
In the water quality area, vertical stratification jsig-
nlflcantly affects the vertical flux of nutrients to and from
the bottom sediments and may contribute to the [for-
mation of anoxic regions along the bottom of the
estuary. The objective of the data analysis is a decision
on the dimensionality of the model. It would obviously
be inappropriate to use a two-dimensional depth
averaged model in an estuary having a history of bot-
tom anoxia.
14.2.4. Spatial Resolution of Models
In terms of spatial resolution, environmental models
may be classified as box models or as one-, two-, or
three-dimensional hydrodynamic models. The distinc-
tion between box models and the hierarchy of dimen-
sional hydrodynamic models is an important one that
is not clear in the presentation of the four case studies
of Part 4.
A. Box Models
Box models require an empirical, ratherthan an analyti-
cal (or numerical), specification of the flow field. Thus
there is no hydrodynamic model component in a box
type model. Box models may be arranged in a lon-
gitudinal, lineal array or boxes may be arranged in
pseudo two-dimensional depth-averaged arrays. Two
examples of this are contained in the case studies.
Case study 10.0 of Saginaw Bay on Lake Huron shows
the entire Bay represented by five boxes (see Fig. 10.2).
Case study 11.4, Potomac Eutrophication Model
(PEM), uses a box network consisting of 23 main
channel longitudinal segments and 15 lateral tidal em-
bayment segments. In the lower saline portions of the
estuary, these box segments are as much as 15 miles
in length. This is mind-boggling when it is realized that,
by definition, each box is a fully-mixed compartment.
Case study 10.0 (Saginaw Bay, Lake Huron) contains
no information on how the flow between boxes (the
largest of which has as surface area of about 400
square miles) or how the dispersive mixing parameters
are determined. In addition, there is no information on
the sensitivity of model results to these important
transport quantities. The time scale of the model is
seasonal, that is, it deals with monthly variations in
water quality parameters. In terms of spatial and tem-
poral resolution, it is difficult to see how this model
would be applicable to estuary studies.
Case study 11.4 (Potomac Eutrophication Model) is
similar to the Saginaw Bay study in that there is no
information on how the daily averaged flow and disper-
sion between boxes is obtained.
In this reviewer's opinion, box models represent a
"black art." Specification of empirical advective and
dispersive transport between boxes can only be ac-
complished reliably by using a conservative substance
such as salinity. Determining the spatial distribution of
advection and dispersion for each box segment that
14-10
-------
satisfies a given salinity distribution requires the solu-
tion of an inverse problem for which there is no unique
solution. Furthermore, the spatial distribution of advec-
tion and dispersion coefficients will change in time in
relation to factors such as fresh water inflow, which
change the longitudinal and vertical distribution of
salinity.
B. Hydrodynamic models
The state-of-the-art of numerical hydrodynamic model-
ing is extremely well advanced in two-dimensional,
both laterally averaged and depth averaged, applica-
tions. Limited, but reasonably good experience exists
at the three-dimensional level. An excellent review of
the status of two- and three-dimensional
hydrodynamic modeling has been prepared by the
ASCE Task Committee on Turbulence Models in
Hydraulic Computation (ASCE, 1988). The review con-
tains a discussion of various turbulence closure
models, lists of available two- and three-dimensional
hydrodynamic computer codes, code selection guides
and case study examples.
The only case study in Part 4 which falls within the realm
of multi-dimensional hydrodynamic modeling is 11.5
Neleus (Potomac Residual Chlorine Model). This is a
two-dimensional, depth-averaged, finite element
model of the upper, fresh water, tidal portion of the
Potomac. The case study is deficient in not providing
a list of references. The two-dimensional model grid
shown in Fig. 11-13 consists of more than 1,100 ele-
ments covering a 15-mile portion of the river. Calibra-
tion of the model for the 1980 dye study (Fig. 11-14)
shows reasonably good agreement. In general, the
model is well-suited to provide information on residual
chlorine levels.
The group of case studies in Part 4 is deficient in not
providing an example of a two-dimensional, laterally-
averaged hydrodynamic model. This type of model is
well-suited to estuaries that exhibit some degree of
salinity of temperature stratification over the depth.
Bloss et al (1988) describes the application of a two-
dimensional, laterally-averaged hydrodynamic and
salinity model to the Trave estuary in Germany. A
long-term simulation of 85 days reproduced total
mixing events and strong stratification. The model
showed good agreement with extensive field data. A
similar 2-D model study of stratification and wind-in-
duced destratification in Chesapeake Bay has been
reported by Blumberg and Goodrich (1990).
One-dimensional hydrodynamic and salinity models
are in an advanced state of development. These cross-
sectionally averaged models are applicable to well-
mixed estuaries - those having strong tidal regimes and
relatively small fresh water inflows. The Delaware and
Hudson estuaries are examples of reasonably well-
mixed estuaries.
Case study 12.0 (MIT-Dynamic Network Model) ap-
plied to the Manasquan River in New Jersey is a good
example of a one-dimensional hydrodynamic and
salinity model. Longitudinal dispersion is modeled as
a function of magnitude of the local salinity gradient
and the degree of vertical stratification. Thus this model
is able to track longitudinal salinity changes due to
variations in fresh water inflow. (Thatcher and Har-
leman, 1981).
The remaining case studies of Part 4 are 11.3 (Dynamic
Estuary Model) applied to the upper portion of the
Potomac estuary and 13.0 (RECEIV-II-EPA) applied to
the Calcasieu Estuary. These models are pseudo one-
dimensional tidal models employing a link-mode
schematization. Tidal motion is represented, but the
models do not include hydrodynamic and salinity inter-
actions. The primary disadvantage of this class of
models is that dispersion effects are not modeled and
therefore must be calibrated using conservative
tracers. A characteristic of this class of model is their
inability to simulate the steepest portion of the lon-
gitudinal salinity gradient due to excessive longitudinal
numerical dispersion (See Fig. 11-3).
The Calcasieu estuary case study (13.0) states that the
model contains no dispersion. The so-called
hydrodynamic verification for tidal stages is very poor
(See Fig 13.4). This reviewer would not recommend
further use of this model for estuarine studies.
74.2.5. Temporal Resolution of Models
The prevalent modeling philosophy throughout this
manual (and one that is widely held) is that the temporal
resolution of a model should be determined by the time
scale of interest to the user of the model output. This
usually leads to the conclusion that time steps
averaged over a tidal period or longer are desirable.
The result is a model far removed from the physics
(fluid mechanics) of the relevant transport and mixing
process. Thus the modeler is required to "select" multi-
dimensional dispersion coefficients which must be "ad-
justed" by calibration to inadequate data. This
approach is based on the mistaken assumption that
there is some inherent law stating that there must a
correspondence between the time scale of the model
input (and the computational time scale) and that of the
output.
An alternative approach is to take advantage of the
powerful hydrodynamic computational tools that are
available in one, two or three dimensions. These re-
quire temporal resolution at the intratidal level. The
question then arises as to how to interact the small time
14-11
-------
step hydrodynamic model with the longer time s^tep
water quality model (This is the subject of a separate
discussion below). The philosophical point is t^iat
model output can be averaged temporally in any way
that is desired to produce a result at the time scale of
interest to the user. In other words, one should |not
average the input in order to produce an averaged
output.
14.2.6. Time and Space Scales for Interlacing
Hydrodynamic and Water Quality Mod&ls
The rational methodology for waste load allocation
makes use of water quality models that are capable of
predicting the response of a water body to various
loading scenarios. We are increasingly called uport to
model water bodies that have high degrees of temporal
and spatial complexity. Examples are unsteady,
strongly advective flow systems with density stratifica-
tion due to temperature, suspended and/or dissolved
substances. Such systems are at least two-dimen-
sional and more often three-dimensional in nature.
There exists, on one hand, a number of 2- and 3-dimen-
slonal models that include baroclinic (i.e., stratification)
effects and sophisticated hydrodynamic turbulence
closure components. On the other hand, there are a
number of ecologically sound, multi-parameter water
quality models. These two types of models have
evolved Independently of one another through [the
efforts of hydrodynamicists and aquatic scientists. A
great deal of research support has gone into th^se
separate model development efforts. However, there
has been little effort directed to the crucial problem of
Interfacing or coupling of hydrodynamic transport and
water quality models. The coupling problem arises
because of the following dichotomy.
The dynamic nature of multi-dimensional
hydrodynamic models and the associated numerical
stability requirements usually dictate a small spatial
grid and a computational time step of the ordej- of
minutes. Water quality models typically involve longer
time scales ranging from a day in the case of nutrient
recycling in the water column to months or years for
sediment-water column interactions. There is an ;ob-
vious disinclination to interface water quality mode(s at
the small spatial and temporal resolution of the
hydrodynamic model because of the enormous com-
putational burden. Yet there do not exist generic
guidelines for interfacing water quality models at larger
spatial and temporal increments.
The state of the art of interfacing hydrodynamic and
water quality models has evolved in two directions. The
first makes use of large boxes where, because of jthe
large spatial grid, it is impossible to apply numerical
hydrodynamic models. The user is then faced with the
problem of empirically calibrating transport and mixing
to an observed distribution such as temperature or
dissolved solids. It is impossible to carry out such
"large box" calibrations in stratified, strongly time-vary-
ing, multidimensional systems.
The second interfacing approach averages the advec-
tive and diffusive output of the short time step
hydrodynamic model over larger spatial grids and time
periods (e.g., 24 hours) that are thought to be ap-
propriate to the water quality model. The problem is
that important advective and diffusive information from
the hydrodynamic model is lost in direct proportion to
the length of the spatial and temporal averaging period.
There are no quantitative guidelines for multi-dimen-
sional models to indicate the extent of information loss
by averaging. Therefore, we are again faced with the
necessity of difficult empirical calibration procedures.
A number of studies have addressed the
hydrodynamic-water quality interfacing problem in the
context of one-dimensional lake and reservoir models:
Ford and Thornton (1979), Imboden et al. (1983), Wang
and Harleman (1983), and Shanahan and Harleman
(1984). Systematic studies of interfacing for time-vary-
ing, multi-dimensional, stratified water bodies are now
underway in connection with the EPA/COE
Chesapeake Bay modeling program.
74.27. Water Quality-Eutrophication Models
The water quality components of most of the waste
load allocation models in Part 4 are fairly similar in that
they model BOD, DO, ammonia, nitrite, nitrate or-
thophosphate and chlorophyll. In some case, more
than one class of algae are included, and some models
include zooplankton although data for this component
is usually sparse or non-existent.
Three important waste load allocation and manage-
ment issues are virtually ignored by the water quality-
eutrophication models presented in Part 4. They are:
(a) The question of nitrogen or phosphorus limitation
in the eutrophication process together with the role of
point versus non-point sources as sources of N and P
is of crucial importance in waste load allocation. The
issue of major investments in advanced waste treat-
ment plants as opposed to control of agricultural fer-
tilizer runoff depends on the model's ability to deal with
nutrient limitation kinetics. The problem is complicated
by the fact that most estuaries include upstream fresh
water portions as well as the downstream salinity in-
trusion region. Algal species and nutrient preferences
may shift between the fresh-salt water zones of an
estuary.
14-12
-------
(b) A significant number of estuaries experience sum-
mer anoxic conditions in deep bottom zones. Very low
or zero dissolved oxygen is known to trigger major
increases in the release of nutrients from the bottom
sediment to ;the overlying water. Eutrophication
models applied to estuaries having low DO problems
must have the ability to simulate the vertical stratifica-
tion and vertical mixing processes that affect vertical
oxygen transport and dissolved oxygen gradients and
benthic nutrient fluxes.
(c) The determination of the time scale at which an
estuary responds to changes in waste load inputs
depends on how sediment-water column interactions
are modeled. Waste load (nutrient) inputs generally
result in algal production in the upper euphotic zone.
Dead algae sink and are incorporated as organic
material into bottom sediment. Sediment diagenesis
occurs in the sediment and results in nutrient fluxes and
sediment oxygen demand. The rate at which the sedi-
ment diagenesis occurs controls the rate at which the
estuary responds to loading changes. Important
papers in this modeling area are contained in Hatcher
(1986).
Attention should be given in this document to a report
prepared by ASCE Task Committee on the Verification
of Models of Hydrologic Transport and Dispersion
(Ditmars et al.1987). The objective of the report is to
identify, collate, and define the procedures required for
evaluation of performance of an analytical or numerical
surface water model. The essential elements are: iden-
tification of the problem; relationship of the model to
the problem; solution scheme examination, model
response studies, model calibration; and model valida-
tion. Literature examples are used to define the techni-
ques that have been used to address each of the
elements above. Emphasis in the six elements is
placed on moving the evaluation of models, particularly
those in journal publication, towards more quantitative
or objective measures of calibration and validation.
74.2.8. References
ASCE Task Committee on Turbulence Models in
Hydraulic Computations, ASCE Journal of Hydraulic
Engineering 114(9), September, 1988.
Bloss, Siegried; Rainer Lehfeldt and John C. Patterson.
Modeling turbulent transport in stratified estuary. ASCE
Journal of Hydraulic Engineering 114(9), September,
1988.
Blumberg, Alan F. and David M. Goodrich. Modeling of
wind-induced destratification in Chesapeake Bay, Es-
tuaries, March, 1990.
Ditmars, J.D., E. Eric Adams, Keith W. Bedford, and
Dennis E. Ford. Performance evaluation of surface
water transport an dispersion models. ASCE Journal
of Hydraulic Engineering 113(8), August, 1987.
Ford, D. and K.W. Thornton. Time and length scales for
one-dimensional assumptions and its relation to
ecological models. Water Resources Research 15 (1)
February 1979.
Hatcher, Kathryn J. (editor). Sediment Oxygen
Demand: Processes, Modeling & Measurement.
Athens, Georgia, Institute of Natural Resources,
University of Georgia, 1986.
Imboden, D.M. et al., Mixing processes in lakes:
Mechanisms and ecological relevance. Schwerz. Z.
Hydrol., 45(1), 1983.
Shanahan, P., and D. R. F. Harieman. Transport in lake
water quality modeling. Proceedings, ASCE Env. Eng.
110(1), February 1984.
Thatcher, M. Llewellyn and Donald R.F. Harieman.
Long-term salinity calculation in Delaware estuary,
ASCE Journal of the Environmental Engineering
Division 107(EE1), February, 1981.
Wang, M. and D.R. F. Harieman. Modelling
phytoplankton concentrations in a stratified lake.
Proceedings, Ecology Modeling Conference,
Colorado State University, 1983.
14-13
-------
14.3. Gerald T. Ortob, Ph.D., RE.
Professor, Department of Civil Engineering
University of California
Davis, California \
14.3.1. Introduction
This assessment of selected case studies of estuarine
modeling was prefaced by an opportunity afforded by
the Environmental Protection Agency to review drafts
of the proposed technical guidance manual for waste
load allocation In estuarine systems. It draws on .this
background to some extent, but is based primarily on
the experience of the reviewer in developing and' ap-
plying mathematical models as tools in support of
decision making In water quality management, much
of which has been concerned with estuaries. Naturjally,
the views expressed here are reflective of thisj ex-
perience and are uniquely those of the reviewer, for
which he takes full responsibility.
Before examining the specifics of the selected case
studies, It is appropriate to identify a few of the charac-
teristics or attributes to the modeling process that need
special attention in bringing models to a level of effec-
tive application as decision support tools. Among the
more Important of these are the following: ;
• Specific goals or objectives of users to be met by
the use of models and related decision support
capabilities.
• Basic data and information required for construc-
tion, calibration, and verification of model(s). i
• Temporal and spatial scales appropriate to! the
intended use of model(s).
• Hydrodynamics input to water quality models.
• Model structure and complexity. ;
• Calibration and verification.
• Some brief comments concerning each of these
will provide a reference for the succeeding pase
study critiques.
• Goals and Objectives. ;
i
In the present context, models are to serve as useful
tools in the decision process, i.e. they are to enable a
decision maker to make better, more defensible
choices among alternatives for waste load allocation.
Although some users would like to use estuarine
models in predictive modes, this is rarely feasible at the
present state of the art. Most of models currently avail-
able for water quality simulation are inherently uncer-
tain to a degree that absolute prediction is exceedingly
risky. However, after careful calibration and verification
estuarine water quality models can usually be applied
with confidence In assessment of incremental changes
between simulated solutions for different structural or
operational alternatives.
14.3.2. Basic Data and Information
The weakest aspect of most modeling projects is the
data base. Most often data are gathered without a view
to future development or application of a model, and
the modeler is forced to adapt to an existing but inade-
quate body of data. This has prompted some modelers
to resort to construction of simple box or statistical
models rather than design and implement a data base
to serve model application. A well designed data col-
lection program is the best confidence builder for
modeling. It should be a continuing activity in any
situation, where models are to serve future manage-
ment of estuarine water quality.
14.3.3. Temporal and Spatial Scales
Selection of time and space scales for modeling is an
activity that is closely related to definition of objectives.
If decisions are to be based on long term (monthly or
more) means, then the dynamics of water
quality/ecologic processes on a diurnal or tidal basis
may not be necessary, although it may still be risky to
smooth short term data on these processes, thereby
eliminating important information on extremes. Often it
is the extreme values, occurring during daily or tidal
periods, that are of greatest importance in waste load
allocation. Temporal or spatial averaging may be jus-
tified in cases where the data are sparse or where the
decision process does not require great detail. In
today's world of computers the cost of simulation is
fast becoming a non issue, that is, the degree of tem-
poral or spatial discretization is virtually at the discre-
tion of the user. If model detail is required, it is more
likely to be controlled by the availability of data for
calibration and verification than computation cost.
14.3.4. Hydrodynamics
In the judgement of this reviewer inadequate descrip-
tion of advective transport is probably the most com-
mon cause of poor calibration and verification in water
quality models. This need not be the case, however,
since good hydrodynamic models exist for virtually all
types of estuarine systems, from simple one-dimen-
sional channel networks to complex stratified estuaries
of broad lateral extent where three-dimensional repre-
sentation is required. Most of these models are relative-
ly easy to calibrate and verify, compared to their water
quality companions, and produce descriptions of
water levels and current structure that are useful for
"driving" water quality simulators.
It is good to recognize in this connection that there is
an important trade off between improving simulation of
advective processes, which entails additional spatial
14-14
-------
and temporal detail, and depending on empirically
derived dispersive fluxes to describe transport of pol-
lutants. Model simplification usually means greater un-
certainty, which in water quality simulation is usually
manifest in empirical dispersion coefficients which rep-
resent the aggregate effects of many ill-defined quan-
tities.
74.3.5. Model Structure and Complexity
The trend in water quality modeling has generally been
toward increasing complexity, i.e. more state variables
and an even greater number of additional rate con-
stants, coefficients, etc. While this is a commendable
trend in the sense of improved understanding of the
aquatic system, it also introduces increased uncertain-
ty in model output, due in major part to the inherent
uncertainties in the parameters that have to be es-
timated or empirically determined. There is probably a
level of detail that is "best" for a given situation, some-
where between a simple black box and the detailed
model, which produces the most reliable result from
the decision maker's viewpoint. Uncertainty analysis,
e.g. first order error analysis, Monte Carlo simulation,
etc., may provide some guidance as to best structure
for the model in relation to decision goals.
14.3.6. Calibration and Verification
Although modeling implicitly requires comparison of
simulation and prototype observations, and most
modelers comply with the two step process, the prac-
tice is still largely judgmental. There are comparatively
few examples of rigorous objective assessment of
model reliability. There is a need for formalizing calibra-
tion and verification procedures, perhaps along the
lines of the uncertainty analysis approach suggested
above.
74.3.7. Case Study Review
The five case studies were ostensibly selected for the
diversity of modeling approaches to characterization
of estuarine water quality. They were chosen also, so
it appears, to represent a range of difficulties en-
countered in applying existing models to actual es-
tuaries and further to illustrate varying degrees of
success in overcoming these difficulties. No ideal ex-
amples are provided, since none actually exist. How-
ever, while those chosen for this review can fairly be
regarded as instructive of some of the problems en-
countered in the real world of water quality modeling,
they may not be as exemplary of estuarine modeling
per se as one would like. In two of these cases, as we
shall see, there is reasonable doubt that the systems
modeled can even be categorized as estuaries within
the definitions provided in the technical guidance
manual.
Yet, the several case studies do represent applications
of a number of different models and it is useful to
examine these for comparative purposes. Because the
type of estuary often dictates the structure of the model
most suitable for simulation of its hydrodynamic and
water quality behavior, this reviewer has chosen to
organize his critique according to the specific
geographic situation.
Case Study 1 - Saginaw Bay
This is a non-estuary, at least in so far as classical
definitions apply. The major difficulties here appear to
be most likely associated with characterization of
transport rates, both advective and dispersive. Al-
though it is acknowledged that "water levels and flow
directions of the Bay change" there is no explicit treat-
ment of the hydromechanical behavior of the water
body. Admittedly, this is not a trivial undertaking from
a modeling viewpoint, although there are some excel-
lent examples of two- and three-dimensional circula-
tion models for the Great Lakes that would probably be
suitable for Saginaw Bay.
The complexity of the ecosystem dynamics repre-
sented in this model and the rough "box" configuration
of the embayment, both suggest a greater interest in
ecosystem kinetics than in the practical problem of
waste load allocation. Both of these aspects, simplicity
in the one extreme (5 boxes) and complexity in the
other (18 plus compartments), lead to increased uncer-
tainty in the results of simulations. This reviewer sug-
gests that perhaps a better result from the point of view
of the decision maker might have been obtained with
a somewhat more rigorous description of lake circula-
tion and some aggregation in ecosystem compart-
ments. The trends indicated by the results shown seem
hardly sufficient for decision purposed in light of ap-
parent uncertainties in model parameters and field
data.
This is a case where it seems that the third spatial
dimension could be especially important in the model.
To what degree does stratification of water quality
variables play a role in determining primary produc-
tivity? What about vertical advection and dispersion? It
is not clear that changes along the vertical axis are
important considerations in this case study, although
they should be.
In summary, Saginaw Bay is not an estuary, so as a
case study of estuarine modeling this example leaves
much to be desired. Nevertheless, it is instructive in that
it illustrates the tradeoff between hydrodynamic cir-
culation and dispersion as driving forces in water
quality modeling, as opposed to increased complexity
in ecosystem description. However, because of the
greatly increased data requirement that accompanies
14-15
-------
the Introduction of more state variables, such a mbdel
may not be the most cost effective from the decision
viewpoint. Notwithstanding this argument, it may still
be a great learning tool. This is probably its rnost
Important attribute.
Case Study 2 - Potomac Estuary
Here we deal with a real estuary, but only partially. The
focus in this example of the application of the Dynamic
Estuary Model (DEM) and the Potomac Eutrophication
Model (PEM) is on the upper "fresh water" section of
the estuary, where the effects of tidal oscillation are
minimal. In this region dispersive effects induced by the
tide (which is of rather small amplitude, anyway) are
probably negligible and stratification is unlikely to be a
significant consideration. [
DYNAMIC ESTUARY MODEL
A redeeming feature of the DEM application is tfjiat it
does address directly the hydrodynamics of the es-
tuarine system, producing time-variant water levels,
velocities, and discharges as output of a hydrodynamic
model, which are, in turn, utilized in a separate water
quality model to describe the fate of pollutants it? the
estuary. A limitation of the model(s) is that the basic
configuration is one-dimensional, that is, flaws are
directionally constrained. Pseudo two-dimensional
representations are possible for shallow vertipally
mixed embayments, but circulations for such systems
should be regarded as rough approximations.
Calibration and verification of the hydrodynamic model
was achieved in a straight forward manner. Extensive
experience with this model in branching channel es-
tuarine systems, like the Sacramento-San Joaquin
Deltaforwhich it was originally developed, indicate that
it Is easy to calibrate and gives a good account of| tidal
effects over a wide range of boundary conditions;.
The problem of calibrating the DEM for chlorides along
the axis of the estuary is attributed to numerical mixing,
a consequence of the solution procedure. Despite this
difficulty the model appears to give fair results at the
farfleld level. The practice of varying model coefficients
from one survey to the next in an arbitrary manner in
order to assure the "best fit" is purely subjective and
should not be encouraged. If such as procedure is
employed, a rational basis for parameter adjustment
must be provided. After calibration and verification for
the Potomac study the DEM model appears to 'have
been provided with most of the attributes of a Useful
decision support tool.
POTOMAC EirTROPHICATlON MODEL
The PEM has many of the characteristics of QUAL 2E
or WASP 4, in that it is essentially a box model for which
the boundary fluxes are governed by either a simple
hydrologic mass balance or are generated by an exter-
nal hydrodynamic model like that in DEM, averaged
over a tidal cycle. The contention that the "PEM was
developed because the existing DEM model focused
more on spatial resolution than on the kinetic com-
plexities of eutrophication" implies that spatial resolu-
tion is not of consequence in eutrophication and that
kinetic complexities could not be accommodated in a
modified DEM. This reviewer believes that spatial
resolution of the degree afforded by the DEM, as well
as the hydrodynamic information such a model
provides, are indeed desirable for a eutrophication
study such as exemplified by this case. The more
detailed kinetics of PEM are, of course, appropriate.
However, experience has shown (and another of these
case studies illustrates) that the attributes of more
complex kinetics need not be at the expense of realistic
hydrodynamics.
Spatial resolution and temporal resolution may be dic-
tated in part by the structure of the basic data used to
calibrate and verify the model. The practice of ag-
gregating data from several stations and smoothing
over time seems in this case to be consistent with a
"regional and seasonal focus," but it tends to ignore
local and short term events which are often of major
concern in setting goals for wastewater management.
It also presents problems in calibration and verification,
as evidenced in some of the examples given.
The post audit experience, in which the model was
unable to predict the magnitude or spatial extent of the
1983 blue-green algae bloom, appears to confirm a
need for improved resolution and extension of the
model. It is credit to the model developers that the
model has been periodically revised to improve its
capability as a management tool.
NELEUS - CHLORINE MODEL
The problem presented in modeling the fate of chlorine
in the Potomac Estuary is properly addressed with a
two-dimensional finite element model, capable of rep-
resenting the irregular configuration of the water body
and providing the essential spatial detail. It is unfor-
tunate that field data were insufficient for thorough
calibration, but experience with such models has
shown that hydrodynamics can be closely simulated,
even for very complex geometries and unsteady
boundary conditions.
The water quality model in this package is driven by the
hydrodynamic model, but with the added requirement
of estimating lateral and longitudinal dispersion coeffi-
cients. Again, model calibration was not carried to a
satisfactory level, due in major part to inadequate field
data. There is insufficient foundation for selection of
either dispersion coefficients or the decay rate for
14-16
-------
chlorine, henee the models at this stage are of ques-
tionable use for decision purposes, despite their intrin-
sic potentials.
The importantlesson of this case study is to provide an
adequate data base for complete calibration and
verification of both hydrodynamic and water quality
models.
Case Study - Manasquan Estuary
This case, ampng all those presented, is probably the
most balanced in the treatment of hydrodynamics and
water quality, and in calibration and verification
methodology. Unfortunately, the MIT-Dynamic Net-
work Model (MIT-DNM) did not reach the stage of
actual application as a management tool, so its perfor-
mance cannot be fully assessed.
The calibration-verification sequence of
hydrodynamics/conservative tracer (salinity)/noncon-
servative water quality is representative of good model-
ing practice. Because the model is one-dimensional
and only a rough approximation of the estuary, it is
necessary to utilize an empirically derived dispersion
coefficient as a calibration parameter. While the func-
tional relationship between this parameter and
geometric and hydraulic properties of the estuary ap-
pears well founded in theory and experiment it is never-
theless unique for a particular estuarine system, e.g.
constant K and m. It purports to account for factors that
cannot be adequately represented in a one dimen-
sional model with such a coarse segmentation, e.g.
advective dispersion and stratification. Dependence
on this uncertain calibration parameter could probably
be reduced by:some additional detail in spatial charac-
terization of the estuary.
The relatively unsatisfactory results of water quality
calibration point to a need for improving the data base,
particularly the pattern of nutrient loading on the es-
tuary. It seems unlikely that the model will become a
useful tool for waste load allocation to the Manasquan
Estuary until this additional data is developed.
Case Study - Calcasieu River Estuary
This study was allegedly selected in part because it
represents the application of a so-called "canned"
model supported by EPA. This reviewer disagrees with
the implication that such models, exemplified also by
such well documented and supported models as
QUAL 2E, SWMM, DEM, WASP 4, HEC 5Q, TABS II,
etc. are likely to lead to the kind of difficulties en-
countered in modeling the Calcasieu River Estuary. It
is the responsibility of the modeler to select the most
appropriate modeling approach forthe particular situa-
tion. Most often the modeler is well advised to begin
with a package that is well documented (as are those
cited above) and for which there is a considerable body
of experience in adapting to new conditions. If what is
available proves to be unsuitable it can be modified, as
in this case, or a completely new model can be devised.
The test of its capability will be in the processes of
calibration and verification.
The principal difficulty with the Calcasieu estuary is that
it is so complex that virtually no model existing at the
time of the study was fully equal to the task. The
tortuous looping and branching channel configuration
might at first appear to be a candidate for RECEIV-II,
since the model was designed originally for such sys-
tems. However, this model assumes vertical
homogeneity where the Calcasieu system includes
many sections which are stratified. The system also
includes very broad channel reaches and embay-
ments, even lakes, which are not well represented
hydrodynamically by the pseudo two-dimensional net-
work approximation possible with RECEIV-II. The ex-
istence of stratified lakes within the system suggests
the need for a model capable of dealing with
hydrodynamics in one, two or three dimensions,
depending on the local conditions. A finite element
approach is probably the most feasible at present,
although in fairness to the modelers of the Calcasieu
estuary it is acknowledged that such a model was not
available at the time of the study.
Hydrodynamic calibration/verification for this study
was described as "good," although in certain instances
elevation differences between model and prototype
were large enough to indicate that system storage was
not well simulated, e.g. 1978. Water quality calibra-
tion/verification was fair at best, a result attributed by
the modeler to inadequate input information and
dynamics. Here again the complexity of the system and
the water quality model, with its large number of
parameters, probably preclude a good result. Future
modeling efforts for this estuary should be directed to
improving hydrodynamic simulation and estimates of
waste loads.
14.3.8. Concluding Comment
This selection of case studies illustrates most of the
problems encountered in modeling of water quality in
estuarine systems. Among the lessons to be learned
from these experiences, the following appear to this
reviewer to be the more significant in directing future
modeling efforts.
1. There is no substitute for hard data from the field.
Data collection programs should be designed with
model requirements in mind.
2. Water quality models of estuarine systems are driven
by hydrodynamics. More attention needs to be given
14-17
-------
to the hydrodynamic driver as an integral part of jthe
modeling package. In particular, effects of stratification
should be explicitly modeled. I
3. Complexity may lead to more uncertainty in mcjdel
results. Adding more compartments may improve fun-
damental understanding of important mechanisms,
but tt requires more data and does not necessarily lead
to better decisions.
4. Models should be designed and applied as tools to
support decisions by non-modelers. Output should be
readily Interpretable by decision makers.
5. Calibration/verification is still largely a subjective
process. Criteria for acceptance of a verified mddel
should be developed and related to the intended use
of the model in the decision process.
14-18
-------
-------
------- |