United States
          Environmental Protection
            Office Of Water
August 1992
Technical Guidance Manual
For Performing Waste Load
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
                                    Printed on paper that contains
                                    at least 50% recycled fiber


                     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


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

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

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
        Organization: "Technical Guidance Manual for Performing Waste Load Allocations.
        Book III: Estuaries"
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

                        WASHINGTON, D.C. 20460
                           JUL 2 0 1992
                                                       OFFICE OF
SUBJECT:  Final Technical Guidance Manual For Performing Wasteload
          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.
     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.  !

                                                Printed on Recycled Paper

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-

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.


     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-
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

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.           '

                                      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,

  •  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.







* 	 \.._.
1 \ 	 V-.V.-.V.--V.V? 1

f ,_L_ 	 t 	 I 	







 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
 •  Saturation kinetics for phytoplankton decomposi-

 •  An  advective-dispersive model for  transport of
    chloride used  to determine  water exchange
    among the segments.
 •  Wind-dependent resuspension  for sediment

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

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.

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
f(T)          phytoplankton temperature reduction factor
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
PCM, NCM   concentrations of available nutrients (phosphorus,
             SCM nitrogen, silicon)  in water column in moles/L
PDETH       maximum predatory death rate for zooplankton in
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
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
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
                                                  TOP, TON,



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
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
                                                              The addition of the suffix "BD" to a variable refers
                                                              to the boundary value of the variable.

Table 10-2. Summary of Selected Mode! Coefficients [Bierman and Dolan (1 981)].
a. Summary of Phytoplankton Coefficients ;


mole P/cell

mole P/cell
mole N/Cell

mole N/cell
mole Si/mg
mole Si/liter
b. Summary of Zooplankton




0.724x1 0'13
0801 X10'11
0.357X1 0'5

o.5oo :
0.345x1 0'12
0.208x1 07
0.345x1 0'12

0.004 !
100 i

0.1 58x 10s
0.250x1 06
0.1 00x1 07
0.208x1 07

0.200X 107
0.566x1 0'14
0.438x1 0'12

0.246. 10"6
c. Summary of Coefficients for
1.0 ,




0.356x1 08

Unavailable Nutrients


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-

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

^ 100
o 75
< 50
           8 100
                                     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

          O  75
                                    SEGMENT 2
                                    SEGMENT 3
                     I"..I .I..I
                             I ..I
                                                     SAGINAW BAY 1975
                                                     TOTAL PHOSPHORUS
                                                                   SEGMENT A
                                         i  i  i  i—r—i—r—r
                                                                   SEGMENT 5
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].

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].
Sample Year
1974 1975

Saglnaw River Loadings (Metric Tons): '
Phosphorus ;
Nitrogen ;
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

Annual Average Water Temperature (°C)

Segment 2
Segment 4
12.0 13.9
9.8 11.1
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

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)].
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.

                    z in

                   u i:
                   o _j
                   CC UI
                     ^ 60

73 z






^i *


-* x





1 1


10 O IOO
95 T


j \
'1 J
-« N

^ J
14 ^I-




, -1


• *?


' s'
"• -.




                                                          SPRING    FALL    THRESHOLD
                                                         DIATOMS BLUE GREENS   OOOR
 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.


               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-
Figure 11-1. Location map of Potomac Estuary [USGS (1985)].

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

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

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.


                                                                  —  Modal Predictions

                                                                   X   Tide Table Data
                                        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.

                                                   Miles Below Chain Bridge
Figure 11-3. Potomac Estuary chloride verification, time period: September-October, 1969 [Adapted from Clark (1982)].

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 -

                                                                                     —   Predicted

                                                                                      O  Max. Observed

                                                                                      X  Win. Observed
                                                    !  12

                                                   Miles Below Chain Bridge
Figure 11-4. DEM calibration results for ammonia, time period: August 31-September 16,1965 [Adapted from Clark (1982)].

   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 --
                                                                                    —    Predicted
                                                                                    *    Max. Observed
                                                                                    X    Min. Observed
                                                    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)].
 5 •

 3 -
 2 -
 1  -
                               O   Max. Observed
                               X   Mm. Observed
                                                 X.	X_
                                 -ii	1	1	1	r
                                  12         16         20
                               Miles Below Chain Bridge
1	1	T
Figure 11 -6. DEM calibration results for dissolved oxygen, time period: August 31 -September 16,1965
            [Adapted from Clark (1982)].

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

  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-

 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.

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.

                                               • Nay 26. June 1
                                              — May 30
                                40      60

                                RIVER MILE

^ 100-
5. '
" 50-

• Hoy 26, June 1
— Hay JO

• * j ' 1 I, |

• i i i . . , ,-,- iTpTTr, , , , ,
                                                                                                 • Sept 7, Sept 8
                                                                                                — Sept 7
        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
                             40      6O
                             RIVER MILE
                     • Hay 26. June 1
                    — Hey 30

  I   I  I  I
80      100
Figure 11-8. PEM calibration for BODS and dissolved oxygen, May and September, 1966 [Hydroqual (1982)].

-> O.80-
j= 0.60-
| 0.40-
y o.oo
g 1.00-
| oJ

. M.y 26. June Z
~~N*yI° ___ 0.2O-
f 0.90-
': V)
g 0.60
• 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-
*- O.90
1 O.60
rl\» ' t : °-30
^"n^Lj^xj-i : : J 	
0 ' ' ' 20 " ' ' 40 ' ' ' 60 80 '100
• HIV 26. June 2
— H«y 30


1 1 1 | 1 ' I | ' | 1 I
. . • sept 7, sept 8
A '. — Sept 7
•\ t .
A ^~^-.. !rJ = ~~ ' ~7
0 ' ' ' 20 ' ' ' 40 ' 60 80 100
Flauro 11-9. PEM calibration for total organic and total inorganic phosphorus (mg P/L), May and September, 1966

            [Hydroqual (1982)].


        f, 2.40
        a 1.60
        Z  0.80

        §  0.0&
 May 26, June 2

• Niy 30


^ 2.40

S,  1.60
 Hoy 26, June 2

• H.y 30

1 • Sept 7, Sept 8
M 	 Sept 7
\ :
1 '
. 1:. \
^•1--V-^. 1 :: : . . i


                              RIVER MILE
                  " 4.00-1
                  5 3aoH
                  t 2.40-
                                                                o-oo^-f, ^0'-^40

                                       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)].

                                                                 i °-32
                                                                 « 0.16
                                                                                                  . July 18, July 20
                                                                                                 — July 19

Jl   200

-J    ISO

I    'OCH
S    50
                                            .  July IB,  July 20
                                           — July 19
                                40      60
                                RIVER MILE
1 '
? 0.48-
I 0.32-
1 0.'6-
• July It, July 20
	 July 19

X"\C_._^ : * 3
d •• :•: - • ^
                                                                             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-
J 2.4-
C ~
| 1.8-
g 1.8-
i 1.2-
z I

. July IB, July 20
— July 19

1 1 ' • | • • i j i • . |
• July 18, July 20
— July 19

/ \ '
-J- V'1" :
> 20 40' ' ' 60' ' 80' ' ' IOC
                                                               a" 6
                                                                                                  • July IB, July 20
                                                                                                  — July 19


• July IB. July 20
— July 19
-.-• ^LL^
' • • * .
'•- : ' :*•••:

                                                                           40      60
                                                                           RIVER MILE
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)].

Table 11-1.  Unear Regression Statistics
                                         Chlorophyll a
Standard Error
(uall) I
10.8 i
10.8 '
Intercept fag/l)
                                       Dissolved Oxygen
Standard Error
(mall) '•
0.74 !
Intercept (mg/l)
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

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

 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

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
 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

 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

during computation and to provide accurate model
predictions within fairly short distances of discharge

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
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)].

 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

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

 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

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.

9.O -

8.0 -

7.0 -

&.O -

5.0 -

4.0 -


2.0 -

1.O -

                                           11 — 16 August 198O
 Figure 11-14. August 1980 dye survey calibration at stations B and C [LTI (1987)].

                                                                                         - i,.,  Time Averaged
                                                                                              38.00 to 53.50 hrs
,a I
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

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)].

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-

 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-

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.

-*-•    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)].

                                                                                    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)].

                                                                                    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)].

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,

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

                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-

 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

 The following state variables are included in this ver-
 sion of MIT-DNM:
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.

                            Bargenot Bay

                        Bay Head Harbor

              (^ Boundary Nodes
                                                                               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





95% removal

2mg/l                 ;

6 mg/I


7 mg/1

None detectable by EPA-
approved methods of analysis.
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.

 The measured water quality parameters were as fol-

 •  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:
+  mEj
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

                              8/25       8/26
            8/27       8/28       8/29
                   ~ 20
  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-

 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-

 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-

 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

Table 12-2.  Model-Governing Rate Parameter* [Najarlan et al. (1981)]
Ki: Ammonlflcatlon of detritus-N
Ka: Diatom uptake of ammonlum-N
Ya (half saturation constant)
1. (Optimum intensity of solar radiation)
Topi (Optimum reaction temperature)


da/1 I
Tm«x (Maximum temperature beyond 30 °C
which denaturatlon of cell protein occurs)
Ka: Nanoplankton uptake to ammonlum-N
K<: Diatom mortality rate
KB: Nanoplankton mortality rate
KB: Copapod uptake of dlatom-N
Ky: Copepod mortality rate
KB: Copepod excretion rate
KB: Nitrification rate
Kio:DIatom uptake of (NO2+NOa)-N

Kn:Nanoplankton uptake of (NOa+NO3)-N

97 "?
£>< »O






da/1 ;

°c ;
°c : ;


Ki2:Copepod uptake of detritus-N
KiaiCopepod uptake of nanoplankton-N
Ki4:Dinoflagellate uptake of ammonium-N

Kis:Dinoflagellate uptake of (NO2 + NOa)-N
KietCopepod uptake of dinoflagellate-N
Kieopt ; /

Ki7:Dinoflagellate mortality rate
Kis:Sedimentation of detritus
Kig:Sedimentation of (NOa+NOs)

Kgo: Sediment release of ammonium-N




• '



















        • Dttritui-H
        4- (Httrltt * Nftrate>-N
        & Amoflfwll
        © Predlcttd D«trltut-«
        © Pr«dlct«d (Mitrit* * »itr«t«)-U
        O Predicted Amonlm-N
                  240          242
                  TIME IN DAYS


                                                                   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
   S  6
        238            240            242
                      TIME  IN  DAYS

Figure 12-6.  Temporal variation of CBOD at Chapman's
            Wharf, August 25-28,1980 [Najarian et al.
                                                                     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)].

          238         24O          242

                     TIKE IN DAYS

Figure 12-8. Temporal variation of NBOD at Chapman's
           Wharf, August 25-28,1980 [Najarian et al.

                                                           2:  to
                                                    UJ  g

                                                           (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)].


   UJ   p
   o   °


   s   6

   o   4

         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)].
Hj - Diatoms
N, - Nnnoplankton
N- - Dinof lagellates

36 ' 240 ' 242


<0 (0

35 O

M, - Diatoms
N^ - Nanoplankton
»7 - Dinof lagellates

58 ' 240 242


            203          205           207
                        TIME IN  DAYS

                                   TINE IN DAYS
Figure 12-14. Hydraulic verification: calculated vs. observed
            elevations at Chapman's Wharf [Najarian et al.
         Figure 12-15. Salinity verification: calculated vs. observed
                     salinities at Clark's Landing [Najarian et al.
                             204         206
                         TIME IN DAYS
                                                                                   • Detritus-N
                                                                                   -I- (Nitrite t Nitrato-N
                                                                                   A Aneontua-N
                                                                                  ® Predicted Detrttus-H
                                                                                   Predicted Ammoniu«-N
                                                                                   205        207
                                                                                   TIME IN DAYS
                                                                                                        6 fc
Figure 12-16. Salinity verification: calculated vs. observed
            salinities at Chapman's Wharf [Najarian et al.
         Figure 12-17. Temporal variation of Detritus-N, Ammonium-N,
                     and Nitrite+Nitrate-N at Clark's Landing
                     [Najarian etal. (1981)].

                  TIME IN DAYS

       <0 3
       O DC
                                         . 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

 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.


                    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-

 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)].

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

•   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

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

                                 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

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
Manning's n

Ammonia Oxidation

Nitrite Oxidation

BOD Oxidation

Benthic Oxygen Demand

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

                               Soli Water Sorrier
               I f jj,
         1   A  f
        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.
                                                            (no measured data)
                                                            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

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

•  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




                       | 10



                       ^ 6


— Meotured

— Simulated
                                   • Measured

                                   • Simulated
                        Q 8
                        tn -
                        UJ o

                        % 4
                                                                  ,-•     -M    ,;
                                         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



                              _ i.a








-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)].

                               — Mtaiurtd
                                - Mtaiurtd
                                           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

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.

13.5. 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

NOAA, "National Estuarine Inventory Date Atlas, Physi-
cal and Hydrologic Characteristics", Strategic Assess-
ments Branch, Ocean Assessments Division, 1985[

                        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 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

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

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-

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

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:
              IS ABOUT 1-2 YEARS.

                                             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-

 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

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

 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.
           "LARGE" SPACE SCALES,
               THEN SEASONAL,
 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

 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:
              WHICH DATA EXISTS;
              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

 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-

    Analyticol Solutions
     Numerical Model
      Steady State

 Temporol/Spatiol Comparisons
 Statistical Tests/Comparisons
                         INCREASE MODEL COMPLEXITY
                                State Variables
                              Kinetic Mechanisms
                              Temporal/Spatial Detail
   Critical Conditions
  Sensitivity Analysis
    Load Projection.
 Components Analysis
                                                      ALLOWABLE LOAD
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        ;
       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-

 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.
 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.


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-

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


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

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.


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-


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

      this case study should be borne in mind by any
   •  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.

14.2. 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

 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

 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

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

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

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

 (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

 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,

 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.

14.3.  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
 •  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.                     ;
 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

 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-

 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

 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
 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

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

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.                      [

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.

 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

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.

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

 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

 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
 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

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.