DRAFT: August 1990
                      TECHNICAL GUIDANCE MANUAL
              FOR PERFORMING WASTE LOAD ALLOCATIONS
                            BOOK III: ESTUARIES


             PART 4: Critical Review of Estuarine WLA 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. Dilks, 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
                                  401 M. Street, S.W.
                                Washington, D.C. 20460

-------
841 R900O2

-------
                             Table of Contents

Preface     	v

Acknowledgements	   vii

 8. Great Lakes Embayment Seasonal Phytoplankton Model of Saginaw Bay   	8-1

     8.1. Background   	8-1
     8.2. Problem Setting    	8-1
     8.3. Model Application   	8-2
     8.4. Post-Audit    	8-8
     8.5. References   	8-9

9. Potomac Estuary Water Quality Modeling   	9-1

     9.1. Background   	9-1
     9.2. Problem Setting	9-1
     9.3. Dynamic Estuary Model (DEM) of Dissolved Oxygen   	9-2
     9.4. Potomac Eutrophication Model (PEM)   	9-6
     9.5. Finite Element Model  	9-13
     9.6. References   	9-22

10. Manasquan Estuary Real Time Modeling   	10-1

     10.1. Background  	10-1
     10.2. Problem Setting   	10-1
     10.3. Model Calibration  	10-2
     10.4 References	10-11

11. Calcasieu River Estuary Modeling   	11-1

     11.1. Background    	11-1
     11.2. Problem Setting   	11-1
     11.3. Model Application  	11-2
     11.4. Total Maximum Daily Loads    	11-4
     11.4. References   	11-7

12. Expert Critique Of Case Studies    	12-1

     12.1. Robert V. Thomann, Ph.D	12-1
     12.2. Donald R.F.  Harleman, Ph.D	12-11
     12.3. Gerald T.Orlob, Ph.D., P.E	12-14
                                       iii

-------
                                            Preface
The document is the third of a series of manuals pro-
viding information and guidance for the preparation of
waste load allocations. The first documents provided
general guidance for performing waste load allocation
(Book I), as well as guidance specifically directed to-
ward 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 summaries the important
water quality problems, estuarine  characteristics and
processes affecting those problems, and the simula-
tion models available for addressing these problems.
Part two provides a guide to monitoring and model
calibration and testing,  and a case study tutorial on
simulation of waste load allocation problems in simpli-
fied estuarine systems. The third part summarizes initial
dilution and mixing zone processes,  available models,
and their application In waste load allocation.

This part, "Part 4: Critical Review of Estuarine Waste
Load Allocation Modeling," summarizes several histor-
ical case studies, with critical review  by noted experts.
The reader should refer to the  preceding  parts for
information on model processes, available  models,
and guidance to monitoring and calibration.

The technical guidance is comprehensive and state-of-
the-art. Case studies of applications serve as the best
teacher of the proper and improper use of this technical
guidance. Therefore, included In this  part are four
summaries of actual estuarine studies where models
were used for waste load allocation. These studies
have been selected to provide a range of representa-
tive geographic areas, estuaries, and models. The
studies were not selected because they were exem-
plary but rather because they represented applications
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 prob-
lems.

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 estu-
arine modeling.

A draft version of this  document received scientific
peer review from the following modeling experts:

   Steven C. Chapra,
      University of Colorado-Boulder
   Donald R.F Harleman,
     Massachusetts Institute of Technology
   Gerald T. Orlob,
     University of California-Davis
   Robert V. Thomann,
      Manhattan College
Their comments have been incorporated Into the final
version.
       Organization: Technical Guidance Manual for Performing Waste Load Allocations.
       Book III: Estuaries"
Part
1
2
3
4
Title
Estuaries and Waste Load Allocation Models
Application of Estuarine Waste Load Allocation Models
Use of Mixing Zone Models in Estuarine Waste Load Allocation Modeling
Critical Review of Estuarine Waste Load Allocation Modeling

-------
  8. Great Lakes Embayment Seasonal Phytoplankton Model of
                                      Saginaw Bay
8.1. Background
The Saginaw Bay phytoplankton model of Biemnan 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-
sively  by the USEPA and International Joint
Commission to evaluate nutrient loading reductions for
Saginaw Bay.
          SAGIHA*
                         FLI MT
          SCALE
          to  to
      6 »  20 SO 40 50kn
   Figure 8-1. Saginaw Bay Site Map [Blerman and
   Dolan (1980)]
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. Herbh/ory, settling, and decom-
position are mechanisms of phytoplankton depletion.

82. Problem Setting
Located on the western shore of Lake Huron (Figure
8-1),  the  Saginaw Bay watershed Is  approximately
21,000 km2 (8108 ml2). It is dominated by agriculture,
forest, and four urban-industrial centers: Bay City, Rint,
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-dogging 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-audit
following the  large reductions of phosphorus (1980).
                                               8-1

-------
8.3. Model Application
Although this model's development began in a more
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 about the 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 growth-
rate limiting than nutrients. However, nutrient limitation
became important for peak phytoplankton crops. In the
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.

8.3.1 Model Description
The model developed for Saginaw Bay falls into 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 (see  Figure 8-2). Transport of chemicals,
biomass, and water between cells occurs through ad-
vective 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
is written for each cell and the resulting differential
equation solved simultaneously through time for all
cells by a numerical method.

The model incorporated three nutrients - nitrogen,
phosphorus, and silica - each with biologically avail-
able and unavailable components, and a  biomass
 LEGEND: •  BOAT STATION

        A  WATER INTAKE
   Figure 8-2. Model Segmentation of Saginaw Bay.
   [Bierman and Dolan (1980)]

component. It includes five classes of algae and two
classes of zooplankton. The interaction of the com-
ponents are shown in Figure 8-3.

The model was 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  used
phytoplankton cell biomass.

A number of mechanisms are considered in this model,
including:

•  Internal  nutrient  pool kinetics for  phosphorus,
    nitrogen, and silicon.
•  A  reaction-diffusion  mechanism for carrier-
    mediated uptake of phosphorus and nitrogen that
    includes luxury uptake of nutrients.
•  Biological-chemical kinetics, included in sediment
    compartments for total concentrations of phos-
    phorus, nitrogen, and silicon.
•  Zooplankton grazing
    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 concentration with no net flow of water.
                                                 8-2

-------
                                HIGHIH
                              PREDATORS






DIATOMS
.............
7
/


X




1
CACCNS

V
\




OTHCKS







•LUC-GMCNS
MOW m ttMt*U











•LUC-CftCfNS
in num

   Figure 8-3. Schematic Diagram of Principal Model Compartment* and Interaction Pathway*. [Blerman and Dolan
   (1986a)]
 •  Saturation  kinetics for water column nutrient
    mineralization.
 •  Saturation kinetics for phytoplankton decomposi-
    tion.
 •  An advective-dispersive model for transport of
    chloride used  to  determine  water  exchange
    among the segments.
 •  Wind-dependent resuspension for sediment
    nutrients
The internal nutrient pool  kinetics  are a noteworthy
aspect of the model because they treat cell growth as
a two-step process: 1) uptake  of nutrients and 2)
biomass growth. The common approach is a one step
use of the Monod (Michaelis-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-com-
ponent loss mechanism. It was included for diatoms,
greens, and 'other" phytoplankton, but not for blue-
greens. The zooplankton response function included
losses to higher-order predators. A constant "refuge
                                                 8-3

-------

-------
    Table 8-9. Description of Model Coefficient*. [Blerman and Dolan (1981)]
FACT        phytoplankton cell size in mg dry wt/cell

f (L)          phytoplankton light reduction factor
             (dimensionless)

f(T)          phytoplankton temperature reduction factor
             (dimensionless)

Ke           light extinction coefficient in meter'1

KNCELL     intracellular half-saturation constant for nitrogen-
             dependent growth in moles N/cell

KPCELL     intracellular half-saturation constant for phosphorus-
             dependent growth in moles P/cell

KSCM       half-saturation constant for silicon-dependent
             growth of diatoms in moles Si/L

KZSATk      half-saturation concentration of phytoplankton
             for grazing by zooplankton k

P, N         actual moles of phosphorus (nitrogen) per phyto-
             plankton cell

PCA, NCA    intracellular available phosphorus (nitrogen)
             concentrations in moles/liter cell volume

PCAMIN, NCAMIN minimum  intracellular concentrations, cor-
             responding to PSAMIN and NSAMMIN,
             respectively, for available phosphorus (nitrogen) in
             moles/liter cell volume

PCM, NCM   concentrations of available nutrients (phosphorus,
             SCM nitrogen, silicon) in water column in moles/L

PDETH      maximum predatory death rate for zooplankton in
             liter/mg-day

PHOTO      photoperiod (dimensionless)

PKI, NKI      affinity coefficient for phosphorus (nitrogen) uptake
             mechanism in liter/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, NSAMIN minimum quota of phosphorus (nitrogen) in
             phytoplankton cells in moles/mg dry wt

PSATi,       saturation concentration of zooplankton k above
             which predatory death rate remains constant

Q           water circulation rate in volume/day

RIPM,        maximum phosphorus (nitrogen) uptake rate in

RINM        day"1

RAOINC      incident solar radiation in langleys/day

RAOSAT      saturation light intensity for phytoplankton growth In
             langleys/day

RAGRZOi     rate at which phytoplankton I is ingested (grazed) by
             zooplankton in mg A/liter day

RAMAX      phytoplankton maximum growth rate at 20 C In day'1
RLYS       phytoplankton decomposition rate in liter/mg day

RRESP      phytoplankton respiration rate in day"1

RTOP, RTON, RTOS rates of transformation from available
            nutrient forms (phosphorus, nitrogen, silicon) to
            available forms in day "1
RZ
zooplankton specific growth rate in day"1
RZMAX     zooplankton maximum ingestion rate in day"1

RZPEX, RZNEX RZSEX nutrient (phosphorus, nitrogen, silicon)
            excretion by zooplankton to unavailable nutrient
            pool in moles/mg zooplankter-day

SPGR       phytoplankton specific growth rate In day "1

SSA        silicon composition of diatoms in moles/mg dry wt

T           temperature in C

CROP       total phytoplankton concentration in mg dry wt/L

TOP, TON, TOS concentration of unavailable nutrient forms
            (phosphorus, nitrogen, silicon) in moles/L

TOPSNK, TONSNK, TOSSNK sinking rates of unavailable
            nutrient forms (phosphorus, nitrogen, silicon) in
            meters/day

V           inner bay volume in liters

WPCM, WNCM, WSCM external loading rates of available
            nutrients (phosphorus, nitrogen, silicon) in
            moles/day

WTOP, WTON, WTOS external  loading rates of unavailable
            nutrients (phosphorus, nitrogen, silicon) in
            moles/day

Z           zooplankton concentration in  mg dry wt/L

ZASSIM     zooplankton assimilation efficiency
            (dimensionless)

ZEFFw       ingestion efficiency of zooplankton k for phyto-
            plankton  I (dimensionless)

ZOETH      specific zooplankton death rate in day"1

ZKDUM     effective half-saturation concentration of total
            phytoplankton for grazing  by zooplankton
ZSAFE
refuge concentration of zooplankton below which
predatory grazing does not occur
NOTE:
The addition of the suffix "BD" to a variable refers
to the boundary value of the variable.
                                                          8-5

-------
   Table 8-2. Summary of Selected Model Coefficient* [Blerman and Dolan (1981)]
a. Summary of Phytoplankton Coefficient*

Coef.
R1PM
PK1
PO
CONCP
KPCELL
R1NM
NK1
NO
CONCN
KNCELL
SSA
KSCM
RAMAX
ASINK
RLYS
FACT
RADSAT
RRESP

Units
da/1
liters/mole
mole P/cell
mole P/cell
day1
litert/mole
mole N/Cell

mole N/cell
mole Si/mg
mole Si/liter
da/1
meter/day
liters/mg/day
mg/cell
langleys/day
da/1
b. Summary of Zooplankton
Coef.

RZMAX
ZASSIM
KZSAT
AZMIN
BDETH
PDETH
ZSAFE
PSAT
Unit

da/1

mg/liter
mg/liter
da/1
da/1
mg/liter
mg/liter

Diatoms
0.500
0.518x10*
0.724X10'13
0.250x10*
0.724x10'13
0.125
0.100x107
0801x10'"
0.208x107
0.801x10'"
0.334X10"9
0.357x1 0"9
1.6
0.05
0.004
0.450x1 0"5
100
0.05
Coefficient
Faster
Ingester
0.70
0.60
1.0
0.20
0.05
0.50
0.01
1.0

Greens
0.500
0.167x107
0.312X1014
0.250S10*
0.312x10'u
0.125
0.100x107
0.345x10"12
0.208x1 07
0.345x10'12


1.4
0.05
0.004
0.194x10"*
100
0.05
Blue-Greens
Others
0.500
0.158x 10*
0.148X10'13
0.250x10*
0.148x10'13
0.125
0.100X107
0.163x10'"
0.208x1 07
0.163x10'"


1.2
0.05
0.004
0.91 8x10"*
100
0.05
Blue-Greens
(non-N2)
0.500
0.200X 107
0.566x10u
0.356x10*
0.566x10'14
0.125
0.100X107
0.438X10'12
0.208x1 07
0.438x10-12


1.0
0.05
0.012
0.246. 10*
50
0.05

(Na-Rxing)
0.500
0.518x10*
0.488x10u
0.356x10*
0.488x10'14
0.125
0.100X107
0.377X10'12
0.208X107
0.377X10'12


0.70
0.05
0.012
0.21 2x10*
50
0.05
c. Summary of coefficients for unavailable nutrients
Slow
Ingester
0.10
0.60
1.0
0.20
0.01
0.10
0.01
1.0
Coefficient

RTOP, RTON,
RTOS, TOPSNK,
TONSNK





Units

da/1
meters/day






Value

0.005
0.05






tewater treatment and non-point source reduction. The
results were presented as annual average total phos-
phorus concentration,  total phytoplankton  biomass,
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 portion of
the Bay. Segment 4 had the highest water quality in the
Bay. These segments represent 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
                                                distlnquish nutrient limitation among different types of
                                                phytoplankton and hence allows changes in composi-
                                                tion.

                                                In general, the model showed phytoplankton growth to
                                                be nitrogen-limited, but for a two month period (May
                                                and June) diatoms were silica-limited. This agreed with
                                                actual observations of nutrient depletion. In mid-
                                                August the nitrogen-fixing blue-greens capitalized on
                                                the depletion of nitrogen and proliferated. Their growth
                                                was then restricted by phosphorus limitation.
Later application of the model to 1980 data in a post-audit shows the blue-greens actually responded in a much greater
nrnnnrtinn to ohosohorus reduction.
    proportion to phosphorus reduction
                                                 8-6

-------
                                                 SAGINAW BAY  1974
                                                 TOTAL PHOSPHORUS
                      A'M'J'J'A'S'O'N'O'  'J'F'M'A'M'j1 J'A'S'O'N'D
          too
           75
           50
           25
        _l
        ^
        ^ too
        1  75
        <  50
        K
        £  25
        111
        u
        I 100
           75
           50
           25
           0
SEGMENT Z
SEGMENT 3
              J'F  M'A'MVJ'A'S'O'N'D
               SAGINAW BAY 1975
               TOTAL PHOSPHORUS
SEGMENT 4
                             SEGMENT 5
           J'F'M'A'M'J'J'A'S'O'N'O
Figure 8-4. Model Output and Field Date Comparison for Total Phoephonw (Solid Une Is Model Output; Data are
Sampling Cruise Means and Three Standard Deviations) [Blerman and Dolan (1M6a)]
                                       8-7

-------
   Table 8-3. Summary of Selected Model Input*. [Blerman and Dolan (19866)]

Parameter
Saginaw River Loadings (Metric Tons):
Total Phosphorus
Total Nitrogen
Total Silicon
Environmental Forcing Factors:
Number of Days Where Wind Speed Exceeded
Threshold for Resuspension (Annual)
Annual Average Water Temperature (°C)
Segment 2
Segment 4
Sample Year
1974 1975

1266 1470
14,100 15,290
23,000 31,000

29 40


12.0 13.9
9.8 11.1
1980

493
11,030
12,250

35


14.9
12.8
* Percent of sampling cruises for which computed mean values not significantly different from observed mean values at 95% con-
fidence level; average of 13 variables.


8.4. 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 nearly
eliminated.

The predictive capability of the model was tested  using
the 1980 data. The model was rerun using the 1974 and
1975 model coefficients but loading and environmental
conditions for 1980. The  results  are presented as a
comparison of predicted and observed percent reduc-
tions between the 1974-75 calibration  years and the
1980 resurvey year (Figure 8-5). In general, the model
overestimated the percent reduction  in total phos-
   Table 8-4. Statistical Comparison Between Model
   Results and Held Data. [Blerman and Dolan (1986b)]
Year
1974
1975
1980
Cell*
1234
72 85 65 88
64 80 72 86
57 64 52 85
5
87
87
86
* Percent of sampling cruises for which computed mean values
not significantly different from observed mean values at 95%
confidence level; average of 13 variables.
phorus, and underestimated reductions in diatoms and
blue-green algae.

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 four
was correlated with threshold  odor in  the  drinking
water intake. The model predictions for threshold odor
violations in the drinking water intake agreed with ob-
servations 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.

8.5. References
Ayers., J.C., Anderson, D.V., Chandler, D.C., and Lauff,
G.H. 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
                                                 8-8

-------
                       100
                        20
   SAGINAW BAY

 AVERAGE OF RESULTS
FOR SEGMENTS 2 AMD 4
                               64
                             PHOSPHORUS
                               LOADS
            SPRING   FALL
            TOTAL PHOSPHORUS
             CONCENTRATION
 SPRING   FALL    THRESHOLD
DIATOMS 1LUE GREENS  DOOR
               VIOLATIONS
   Figure 8-5. Change In Water Quality Constituent* Between 1974 and 1980 In Segment* 2 and 4. [Blerman and Dolan
   (1986b)]
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.

Bierman, V.J. and O.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,  V.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.
                                                 8-9

-------
                9. Potomac Estuary Water Quality Modeling
9.1. 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 and 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.

92. 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 9-1.) The estuary
                                                                       L/1 »
                       Quantli
  Figure 9-10. Location Map of Potomac Estuary  [USGS(1M5)]
                                              9-1

-------
 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 only freshwater,
 and extends from Chain Bridge to just 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 dis-
 cussed 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 least
 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
 effluent flow 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 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 been
 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 have
 dramatically declined because of increased  waste-
 water treatment.

 9.3. Dynamic Estuary Model (DEM) of
 Dissolved Oxygen
 The Potomac Estuary was regularly depleted of dis-
 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  wasteload
 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 (e.g. no lateral  variation)
with branching. DEM as configured for the Potomac
extends from  Chain Bridge (River Mile 0.0) as an
upstream boundary to Piney Point (River Mile 96.2) as
a downstream  boundary. This configuration consists
of 133 junctions and 139 channels, but the focus of the
water quality modeling was in the upper 20 miles. DEM
simulates In "real time", meaning that the model
predicts conditions as they vary through diurnal and
tidal variations.

DEM  consists  of two  separate but closely related
models. The first, a hydrodynamic  model, simulates
both the tidal and net advective movement of water.
This model  provides  predictions  of water depth,
velocity, and direction of flow based upon input infor-
mation on geometry, roughness, tributary inflows and
tidal variations  in depth at Piney Point. The results of
the hydrodynamic model are input to the second
model, which simulates water quality.

The water quality model predicts the transport and
transformation  of pollutants in the Potomac Estuary.
The model, as applied for Potomac dissolved oxygen,
simulates three  variables: dissolved oxygen,  car-
bonaceous biochemical oxygen demand (CBOD), and
ammonia.

 Dissolved oxygen concentrations are increased by
atmospheric reaeratlon and algal photosynthesis, and
are decreased  by oxidation of CBOD, nitrification of
ammonia, sediment oxygen demand, and  algal
respiration. The model does not predict algal
photosynthesis or respiration. Instead, these values
must be input by the modeler based upon observed
data or calculations performed external to the model.
CBOD concentrations are increased  by point and non-
point loadings, and are  decreased by  settling and
deoxygenation  of CBOD. Ammonia concentrations are
increased by point and nonpoint loadings, and  are
decreased by a first-order loss term defined In DEM as
nitrification.

Water quality data for model calibration and verification
consisted of both wet and dry weather surveys con-
ducted in 1965, 1966, 1967, 1968, 1970, 1977, 1978,
1979, and 1980. The Blue Plains wastewater treatment
plant, the primary point source of pollutants to the river,
implemented secondary treatment in 1977.
                                               9-2

-------
                                           Uilct Below Chain Bridg*

   Figure 9-2. Potomac Estuary Chloride Verification September - October, 1969
 9.3.1. 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 9-2. 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 9-3).  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 are
shown  in Figures 9-4 to 9-6 for the parameters am-
monia, nitrate-i-nitrite, and  dissolved  oxygen. Com-
parisons  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.
                                                9-3

-------
                  6 -
                  5 -
         £.       4-
          I)
         J=
                  2 -
                  1 -
                                                                        —   Model Predictions
                                                                         x   Tide Table  Do to
                     0             20             40            60
                                           Miles Below  Chain Bridge
Figure 9-3.  Potomac Estuary Hydraulic Calibration High Water Phasing — Mean Tide
                                                                                80
100
         n
         t
                2.8 -
                2.6 -
                2.4 -<
                2.2 -
                  2 -
                1.6 -
                1.6 -
                1.4 -
                1.2 -
                  1  -
                OJ -
                0.6 -
                0.4 -
                0.2 -
                  0
                                                                              —  Predicted
                                                                                «  Mox. Observed
                                                                                X  Win. Observed
                                                                             24
                    0         4        8         12       16        20
                                              Uiles Below Chain Bridge
Figure »-4.  DEM Calibration Results: Ammonia Time Period: August 31-September IS, 1965
                                                                                       28
                                                                                                 32
                                                    9-4

-------
           n
           O
           •f
           N
           O
  1.7 -i—
  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 H:
    0 --
                                                                               «
                                                                               x
      Predicted
      Max. Obitrved
      Win. Observed
                                                  —I—
                                                   12
                                                    —I—
                                                     20
—r~
 24
                                                                                        T—
                                                                                        28
                      0         4        B         12        16
                                                Uilei Belo« Chain Bridge
Figure 9-5. DEM Calibration RMult*: NHrata + NHrlta Tim* Period: August 31-September 16,1965
                                                                                                 32
                12
        O
        a
11 -
10 -
 9 -
 8
 7 -
 6 -
 3 -
 4 -
 3 -
 2 -
 1 -
 0
                          —   Predicted
                           *   Max. Observed
                           X   Win. Observed
                                                            —I—
                                                             16
                                                      —I—
                                                      20
             T~
             28
                   0          4         8         12        16        20         24
                                               Kites Below Chain Bridge               '
Figure 9-«. DEM Calibration Reeults: Dissolved Oxygen Time Period: August 31-September 16,1965
                                                       9-5

-------
 9.32. Model Application
 Application of DEM was conducted over the course of
 several years and modeling efforts. Initial wasteload
 allocation projections were made by U.S. EPA (dark
 1982).  A revised and updated examination was per-
 formed in 1984 but recommendations from this effort
 were deferred  when data from the mkj-1980's ap-
 peared inconsistent with model predictions (MWCOG,
 1987). The model was then revalidated In 1987 to more
 recent water quality data, and new wasteload alloca-
 tion projections performed.

 DEM was applied by Greeley and Hansen  (1984) as
 part of the  Washington D.C. Blue Plains Feasibility
 Study, to determine regional capacity treatment needs
 and establish plant allowable effluent  loads for dis-
 chargers to the Upper Estuary. Numerous alternatives
 were examined 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 en-
 vironmental conditions consisting of drought (7Q10)
 freshwater flow and a water temperature of 28°C. and
 the upper 90th  percentile temperature at summer low
 flow. Model coefficients were based on  the average of
 post-1977 simulations. Algal productivity and respira-
 tion 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 back-
 ground values due to the expected decrease in pol-
 lutant  loading.  Boundary concentrations were
 representative of the 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, en-
 gineering and other considerations.

 9.3.3. DEM Post Audit
 State and Federal regulators originally rejected the
 OEM-based  wasteload  allocation recommendations,
 due primarily to a review  of  1982-1985 dissolved
 oxygen data from the  Upper Potomac. These  data
 indicated that dissolved oxygen standards violations
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 revisit the 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.
Umno-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 slmplistical-
ly assumed all ammonia lost 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.

9.4. 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.
                                                9-6

-------
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 tidal
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, but in calibrating the model, the focus was
on matching monthly and annual trends over a regional
scale of 75-100 miles. Such spatial and temporal scales
represent the  global  responses 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.
9.4.1. 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 mBes 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 mBes apart In the
freshwater portions of the Upper Estuary. Data col-
lected during 1966, and 1968 through 1970 were used
In the calibration, and are representative of water
quality conditions prior to the Implementation of phos-
phorus removal at the major sewage treatment plants
along the estuary.

USGS data from the years 1977 through 1979 were
used to  verify PEM. These years were selected be-
cause they offered the chance to study the changes in
the estuary after institution of phosphorus removal  at
Blue Plains. Thus,  the verification period provided an
opportunity to further test the model's ability to simu-
late the  eutrophication process in the Potomac Es-
tuary.

The verification data set Involved short, intensive week-
long surveys in 1977 and 1978. The entire length of the
estuary was usually sampled twice during the 1977 and
1978 surveys, with vertical samples collected at a num-
ber of stations. In 1979, the spatial and temporal
coverage was reduced, and sampling was limited to
twice a week at five major stations.

9.42. 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 completely mixed seg-
ment 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 Uttle Falls (the upstream boundary)  and the
Blue  Plains  sewage 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 approxima-
tions of seasonal flow patterns, not actual day-to-day
fluctuation.  For  model  verifications, the  model also
included flows for the Anacostla River and Occoquan
                                               9-7

-------
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 three
categories:  1)  Point Sources, 2) Combined Sewer
Overflows, and 3) Non-Point Sources. Point source
inputs of pollutants were defined by monitoring data
and daily operating reports from the area's municipal
wastewatertreatment plants. The Blue Plains treatment
facility accounted for the 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 with a
SWMM model simulation of the D.C. sewer network.
Combined sewer overflows for Alexandria were es-
timated based on calculated stormwater runoff and the
average CSO concentrations measured  in the O.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 each
tributary was based on data from USGS gaging sta-
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 1977
to 1979 were analyzed, and a mean concentration 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.

9.4.3. Boundary Conditions
Model inputs for upstream boundary conditions were
based on data but required considerable extrapolation
and interpolation to simulate the several years of con-
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 were
smoothed to characterize seasonal trends, not day to
day transients.
9.4.4. Calibration
The  model calibration Included reaction rates for
phytoplankton growth, nitrogen and phosphorus 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 quality data
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 9-7 to 9-10 show predicted and observed water
quality data. These figures present calibration results
for chlorides, chlorophyll a, DO,  BODs, total organic
phosphorus, total Inorganic phosphorus, ammonia
nitrogen,  and nitrite-nitrate nitrogen during May and
September of 1966, the year with  the lowest recorded
flow. The model predicted the overall variation in the
data well. Of particular note is the chloride calibration,
which validated water transport. Other calibration runs
were slmBar.

9.4.5. 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 for the 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 9-11 and 9-12, 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.

9.4.6. 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.
                                                9-8

-------
                           40      60
                           RIVER  MILE
60 '  ' ',00
                                                          25O-
                                                          200-
                                                           130
                                                        = 100
                                                        ~ so-
                                                         d
                                                        i  °
                                                        t.
                                                        gzso
                                                        "200H
                                                           00
                                                           no
                                                           so
                                                            o
                           20
                                  40      CO
                                  RIVER MILE
                                                 SO      100
                                                                                  from Bydroqual (1»»J)
Figure 9-7. PEM Calibration for ChlorkJe* and Chlorophyll a, May and S^trnibw, 1966
                                                                   20     40     eo      to     wo
                                                                          RIVER MILE
                                                                               trea lydreq««l  (If(I)
Figure 9-«. PEM Calibration for BODS and DlMotvcrf Oxygen, May and September, 1966
                                                    9-9

-------
1.00-
-
= 0«0-
1 0.60-
in
| O40'
0 020
I
O M AM


• «*r 1*. jww I
-Mr*



* •
z^^*^-^ ...

                            4O     60
                            RIVE*  MILE
                                                 too
                                                         OJOi

                                                       _a»-
                                                       x.
                                                       f OJO
                                                       in
                                                       = 0.60

                                                       % aw

                                                       Oaoo-
                                                       2 130
                                                          0.6O
                                                         OOO
                                                                           • ••» M. «•• I
                                                                           -Mr M
                                                                    20
                                                               40     60      W     100
                                                               RIVEN  MILE
                                                                               froa Mr'roquil  (>*•»

Figure 9-9. PEM Calibration for Total Organic and Total Inorganic Phosphorus (mg P/L), May and Saptambar, 1966
    4.0O

    3.20

 „  2.40
 • I. GO
 x  0.80
 kj
 8
 «e  0.00-
                        40      60
                       RIVER MILE
                                               4.00

                                               3.20

                                            ^ 2.40
                                            I
                                            E 1.60-
                                            ui 080
                                            o
                                            o
 •  H«v M.
— H«V II
                                                                  40      60
                                                                 RIVER  MILE
Figura9-10.
                                                                        frea lydroqucl  (H87)
PEM Calibration (or Ammonia & NHrata-NHrtta (mg N/U, May and Saptnwbar, 1966
                                                    9-10

-------
          IOPOO
                                                          OJ2-

                                                          024
                                          i  |  i i  i  I       OOO
       -.   25°
       ^
       J.   200

        d
       J   ISO

       i   WO
 •  Mir it, Miy n
— Mly l»
                      20      40     CO     BO     IOO
                             RIVER MILE
                   J0.4S-

                     032'
 . Mly It. Mly It
	Mly »»
                               20     4O     CO     10      IOO
                                      MIVCR MILE
                                                                                  fro» lydraqual (>»»J)

Flgur* »-11. PEM Verification Chloride (mg/l), ChorophyH-a(ufl/l), D(Molv*d Inorganic PhocphonM (mg P/I) and Total
Phosphorus (mg P/I), July 1977
—
1
?
c
9
0
2
_o
'e
o
1
<
3.0n
.
2.4-
1.8-
1.2-

O.fr


. July 1*. July It
— July t*
.
A-
V
\
J v.. _ .
i • • i . . ' i j . • I . • i i i . . i
   £  00
                                    — July i»
                                                           10-

                                                        !e
                                                        f:
                                                        I  4-
                                                        ta
                                                            2'
                                                           0
                  =  15
July U, July It      "i

                  I  '2

                  I  t

                  1
                                    l: <
                                                     • J»ly W, Mly 1*
                                                     — July It
                                                                                          - 4nly II, July It
                                                                                         — «ly I*
                                                                     20      «O      «0      80
                                                                             RIVER MILE
                                                                                  (torn Hydroquil (1»«2)
Hgura 9-12. PEM Verification for Ammonia Nitrogen (mg N/I),Nltrtte-Nltrate Nitrogen (mg N/l), Bottle BODS (mg/l) and
Dissolved Oxygen (mg/l), July 1977
                                                    9-11

-------
 Table 9-5. Linear Regression Statistics
                                          Chlorophyll a
year
1968
1969
1977
1978
1979
r2
0.81
0.75
0.73
0.88
0.78
Standard Error
tag//)
10.8
10.8
17.6
4.3
2.3
Slope
0.79
0.76
0.84
0.70
0.35
Intercept fag/l)
7.7
2.5
12.99
12.26
20.00
Hypothesis
R
R
A
R
R
                                       Dissolved Oxygen
Year
1968
1969
1970
1977
1978
1979
f*
0.60
0.93
0.74
0.73
0.75
0.68
Standard Error
(ma/I)
0.74
0.38
1.41
0.93
0.59
0.56
Slope
0.83
1.16
0.58
1.21
0.68
1.08
Intercept (mg/l)
1.55
-0.76
2.45
-1.12
2.77
-0.29
Hypothesis
A
A
R
A
R
A
 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 9-1 shows that 73 to 83 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.

 The relative errors of the summer average means of the
 principal state variables were also calculated in the
 PEM study. These values indicate a large degree of
 variation among variables for any one year, as well 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 across all
 variables.

 In comparing the means, a Student's "t" test was used
 to determine the difference between the  observed
 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 percent
of the variable-segment pairs for which a comparison
could be made.
 9.4.8. Post-Audit
Despite the continued reduction in point source phos-
phorus loading and gradual improvement in water
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 f*g/\ 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.
                                               9-12

-------
 •  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 that in-
creases in water column pH could significantly in-
crease  the magnitude 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 addi-
tion of a separate submodel to simulate the equilibria
between the  multiple forms of inorganic carbon. This
pH-driven equilibrium 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 sediment quality and the flux of
nutrients across the sediment water interface. The up-
dated PEM removed these  sediment  computations
and replaced the 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  Microcystis, which was the primary
component of the observed  bloom. Re-calibration of
the model  provided an  improved description of the
observed data. This revised PEM is now available for
use in evaluating the impact of various future nutrient
control strategies in the Upper Potomac Estuary.

9.5. Finite Element Model
Chlorine has  been used extensively as a wastewater
disinfectant and as an agent to prevent biofouling in
cooling  waters. Concerns have been raised that the
discharge of chlorine in wastewater to the Upper
Potomac Estuary might pose ecological health risks.
In particular, discharges from opposing  shorelines
might result  in a cross channel  barrier that  could
prevent fish movement and migration. This study was
conducted to determine the occurrence and fate of
residual  chlorine in the Potomac and to evaluate the
likelihood of the formation of a toxic cross channel
barrier.

A comprehensive study was conducted involving field
surveys of discharge and Potomac Estuary total
residual chlorine (TRC) to document the current spatial
extent of TRC; to develop and calibrate a two dimen-
sional TRC model for testing various environmental
scenarios;  and to conduct  model analysis of the
various scenarios to establish the risk of a chlorine
barrier.

The study area of the Potomac Estuary is freshwater
but hydraulically  influenced by ocean tides. The con-
fluence with the  Anacostia River,  numerous embay-
ments, and highly variable  channel physiography
make this section of the  Potomac Estuary
hydrodynamically complex. The data available to sup-
port a TRC model were limited to grab samples in only
the longitudinal and lateral dimensions. Modeling was
therefore constrained to two dimensions. This was,
however, consistent with the purpose of the modeling
- to define the lateral and longitudinal extent of effluent
residual chlorine  plumes as a potential barrier to fish
migration.

The complex physiography of the upper Potomac Es-
tuary did not allow use of simple analytical models.
One-dimensional water quality models were of little use
for evaluating the chloride discharges because the
lateral extent of contamination could not be simulated.
Branching one-dimensional estuary models, such as
the Dynamic Estuary Model (DEM) may be configured
to run as pseudo-two-dimensional models but have
unrealistically high dispersion for localized calculations
and  poor  characterization of  two dimensional
transport. For these reasons,  a true two dimensional
hydrodynamic and water quality model was required.

The Neleus chlorine  model selected  for this study
consists of  a hydrodynamic model linked to a water
quality model. The hydrodynamic model  solves the
complete non-linear, two-dimensional, partial differen-
tial equations of fluid motion  (Katopodes, 1987;  LTI,
1987). The equations are integrated over time using a
modified Petrov-Galerkin finite element model numeri-
cal technique yielding surface elevation and velocity at
each of the  model grid nodes. The results are input as
mass transport terms to the water quality model.

The water quality  model uses the same grid framework
as the hydrodynamic model and is represented by a
two-dimensional, vertically averaged, partial differen-
tial equation of mass transport. The equation includes
terms for advective and diffusive mass transfer, mass
sources and/or  sinks, and  first-order decay.  The
numerical solution is obtained in the  same manner as
                                               9-13

-------
with the flow 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.

9.5.1. Model Inputs

The Neleus model required a finite element grid com-
prising 1171 quadrilateral elements with 1408 nodes
(element intersections) as shown in Figure 9-13. 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
locations.

After setting the model grid, model Inputs for boundary
conditions and loadings were determined. These in-
cluded tidal elevation and flows. The NOAA Tide Tables
  Figure 9-13. Chlorine Model Finite Element Grid Network
                                               9-14

-------
 provided minimum and maximum tidal elevations and
 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
 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 plant was the only source
 for which information was  known about outfall con-
 figuration and precise location. As a result, the other
 chlorine sources were treated as mass pollutant load-
 ings with no momentum effects. The impact of this
 simplification on main channel model results was min-
 imal since Arlington and Alexandria discharge to em-
 bayments and  PEPCO  discharges  chlorine
 intermittently at very low levels.

 9.52 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.

 9.53. Model Calibration/Verification
 The Neleus model involved validation for both
 hydrodynamic and  water quality models. The
 hydrodynamic model has one calibration parameter -
 Manning's n, which reflects the hydrodynamic effects
 of bottom roughness. The lack of hydrodynamic field
 data limited the calibration of the hydrodynamic model.
 However, previous work by Katopodes (1987) resulted
 in a limited calibration  of the model hydrodynamics
through comparison with DEM hydrodynamic predic-
tions. A constant Manning coefficient of  0.026 was
used by Katopodes (1987) and was chosen for  use in
the chlorine study. The water quality model has three
parameters that require calibration: longitudinal  and
lateral dispersion  coefficient, and the first-order
chlorine decay rate. The dispersion terms were ad-
justed through simulation of the August 1980 USGS
dye study, while the chlorine decay rate was selected
through simulation  of two of the  1984 chlorine field
studies.

The 1980 USGS dye study was used to calibrate the
lateral and longitudinal dispersion coefficients.  The
model simulation began on the 10th of August with the
dye release simulation starting on the 11th. Discharge
from the Blue Plains outfall was constant with a flow of
517 cfs (334 MGD) and dye concentration was 0.03446
mg/l over the release period of 24.8 hours.

Longitudinal and lateral dispersion coefficients were
first estimated from literature information (Fischer et
al.,  1979 and McDowell and O'Connor, 1977)  but
refined to values of  120 ft2/sec for longitudinal disper-
sion and 10 ft2/sec  for lateral dispersion. Figure 9-14
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  fit of 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
cnlorination occurred were not known, but  the levels
of chlorine applied  to the cooling water were low. A
constant residual chlorine concentration of 0.02 mg/l
was used to represent the likely level of discharge from
PEPCO.

For initial simulations a chlorine decay rate of 12.8 per
day was determined experimentally. A more conserva-
                                               9-15

-------
live decay rate of 6.4 per day was also tested. The
comparison of model versus data is shown in Figures
9-15 and 9-16 for averaged field data and model pre-
dictions. Averaging was used to simplify the presenta-
tion  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
              10.0
                                       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  con-
                                       sidered reasonable for both loss rates. The differences
                                       in the simulations were not dramatic and and indicated
         V

        '§
         D
        T»
         O
        (K
              o.o
             1O.O
        Q.
        a
£
£
V

O
T>
O
£
 9.O -


 8.0 -


 7.0 -


 6.0 -


 3.0 -


 4.O -


 3.0 -


 2.0 -


 1.0 -
              O.O
                                                   Model
                                                                       C
   Figure 9-14
     11                      13                      15
                             11-16 August 1960

August 1980 Dye Survey Calibration at Stations B and C (from LT1,19M)
                                                                                           17
                                                9-16

-------
^•-*—
                                                                                  Tim. Awaatd
                                                                                  M.OO lo SJJO
vo*
&
X,U)*miL
-ft.
f.
4
Model Contours
ug/l TRC
Held Data
ug/l TRC
Discharge Site
  Figure »-15. July 11,1984 TRC Survey Calibration at 12.8/Oay LOM Rat*
                                                                                    Tim* Awoatd
                                                                                    36.00 to 53.50
^tOCTM^
-«-
f.
4
Model Contour*
ug/l TRC
Field Data
ug/l TRC
Discharge Site
Figure 9-16. Jury 11,1984 TRC Survey Calibration at 6.4/Day LOM Rate
                                                 9-17

-------
 that physical transport was dominant. The model char-
 acterized the dissipation of TRC especially when con-
 sidering the data can only be best relied on as an order
 of magnitude indicator. The value of 96 /
-------
ZvavnJ.
-!-•
1
Model Contours
ug/l TRC
Dischorge Site
fFVr
Blue Pliins Effluent TRC • 0.40 mg/l

1
&O4/rv(/
Model Contours
ug/l TRC
Dischorge Site
y?*<*
Blue Plains Effluent TRC « 0.02 mg/l
Flgur. *17. TRC Model Projection for 7Q10 Low Flow Condition.
                                                  9-19

-------
                                                                              TVn« - W.SO Kn
-i-    Model Contours
       ug/l TRC

 4    Dischorge Site
                                    Blue Pliini Effluent TRC • 0.40 mg/l
 _»-    Model Contours
        ug/l TRC

 4    Oischorgc  Site
                                    Blue Plains Effluent TRC - 0.02 mg/l
Figure 9-18. TRC Model Projection for Average April Row Condition.
                                             9-20

-------
                                                                                 TVn« - (2.90 hr»
y«»«*w/
-»-•
Model Contours
ug/l TRC
Disehorge Site
^•«JP*
Blue Plains Effluent TRC • 0.40 mg/l
       Model  Contours
        ug/l  TRC
       Disehorge Site
Blue Plains Effluent TRC  « 0.02 mg/l
Figure 9-19. TRC Model Projection for Lowe.1 April Flow Condition.
                                               9-21

-------
9.6. References
Thomann, R.V., N.J. Jaworski. S. W. Nixon, H. W. Paeri
      _	,^_                                  j | luillailfl, n. V., I^I.U. \fGLWW\Jl OIM| W. w ». i ^t**w«», i i. » » . i u^»<
Clark, L J.  1982. A Modeling Study of the  Upper   and Ja* Taft'  1985' The  1983uAlial Bloonl in  the
Potomac Estuary: Application of the Dynamic Estuarv   Potomac Estuary" Prepared for the P°tOmaC Strategy
Model. Dran Report. U.S. EPA Annapolis Field  Office    State/EPA Management Committee.
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.

Heam, 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  Upper
Potomac Estuary, in "Dissolved Oxygen Study of the
Upper Potomac Estuary," Volume 1, Technical Appen-
dix A-7, Metropolitan Washington Council of Govern-
ments, Washington, D.C.

LJmno-Tech  and Metropolitan Washington Council of
Governments, 1987. Summary Report: Potomac River
Residual Chlorine Study. Department of Consumer and
Regulatory Affairs, Washington, D.C.

LTI, Limno-Tech, Inc. 1987. Validation of DEM to 1985
and 1986 Upper Potomac Estuary  Data. Metropolitan
Washington  Council of Governments, Washington,
D.C.

McDowell, D.M., and B.A. O'Connor, 1977.  Hydraulic
Behaviour of Estuaries. Halstead  Press, John Wiley
and Sons, New York.

Metropolitan Washington Council of  Goverments,
1987.  A Dissolved Oxygen Study of the  Upper
Potomac Estuary, Final Report. Washington, D.C.

NOAA, 1984. Observed Hourly Potomac River Tidal
Heights for for 1984 at Washington, D.C., provided by
the National Ocean Survey.
                                              9-22

-------
               10. Manasquan Estuary Real Time Modeling
10.1. 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
(Najarian et al.,  1981). The Authority was primarily
concerned with nutrient enrichment and primary
productivity in the estuary.  A real time model was
selected to predict photosynthesis effects on  diurnal
DO concentrations and investigate the transient im-
pacts of nonpoint source pollution and salt water in-
trusion.

The hydrodynamic submodel of MIT-DNM uses a finite
element approach to solve the one-dimensional con-
tinuity and momentum equations for unsteady flow in
a variable area channel. Dispersion is defined by the
degree of stratification and the non-dimensional lon-
gitudinal salinity gradient using the relationship formu-
lated by Thatcher and Harleman (1972, 1981). The
flows and velocities calculated by this submodel are
used in another submodel  in which  a sequence of
conservation of mass  equations calculates the tem-
poral and  spatial  variation in the water  quality
parameters.

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

 •  Diatoms           •  Nitrite and nitrate

 •  Nanoplankton      •  Carbonaceous BOD

 •  Dinoflagellates     •  Dissolved Oxygen

 •  Organic detritus N  •  Chlorides

 •  Ammonium - N     •  Fecal coliform

 •  Herbivorous zooplankton

The model  assumes that the dominant activity in the
estuary is aerobic and that nitrogen is the only  nutrient
that limits the growth of algae. Water quality processes
represented  in the model include  phytoplankton
growth, mortality, and sinking;  zooplankton grazing,
mortality, and excretion; nitrogen cycling and fluxes at
the sediment/water interface.
102. 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 Bamegat 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 10-1 shows
the study area with sampling stations.

Row 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-
                                     D«c«*b«r ltd
    Rgur* 10-1. Manatquan Estuary and Inlet
                                               10-1

-------
 tributor to the freshwater flow into the estuary under
 low flow conditions.

 Effluent  standards to be met by the proposed plant

                    thS AUthri                 in
 Table 10.1 .Effluent Quality Standard*
Parameter
BODs
NH3-N
DO
pH
NOa-N
Cl2
Standard
95% removal
2mg/l
6 mg/l
5.5-7.5
7 mg/I
None detectable bv EPA-
 Others
approved methods of
analysis.

Such that New Jersey
Surface Water Quality
Standards for FW-2
Trout Maintenance
Streams will be met.
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
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 10-2 shows a schematic of the MIT-DNM reach
system established for the  estuary. The first reach
extends 2.26 miles landward from Osbom Island. The
second reach extends from Osbom Island to the Atlan-
tic Ocean, and is 5.32 miles long. The third reach, 1.78
miles 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.
 10.a Model Calibration
 Figure 10-1 shows the location of stations for model
 calibration sampling performed  in July and August,
 1980, by Elson T. 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-
 lows:

 •  Temperature        •   Nitrite

 •  Dissolved Oxygen   •   Nitrate

 •  Salinity             •   Ortho-phosphate

 •  Secchi Depth       •   Silicate

 •  Dissolved Org.-N     •   Chlorophyll-a

 •  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 dataset was
 used for the purpose of model verification.

 10.3.1. 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
                                                10-2

-------
                             Borgenat Bay

                          Boy H«ad Harbor
                                                                      Boundary Nodtt
               Manasquan Inlet
               Saurct:  Naiarion il ol , 1991
    Figure 10-2. Model Conceptualization of Manacquan Estuary

 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 (Chapman's
 Wharf) by calibrating Manning's friction coefficient. The
 model accurately simulated the observed  hydraulics.
 The maximum difference in observed and computed
 data was approximately 7% of the tidal elevation range.

 Salinity was calibrated next. Salinity observations did
 not begin until August 25,1980. Just 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 computer salinity concentrations during the eb-
bing 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:
              E(x,t)
                       KdS
                        dX
where,
E(x,t) = Temporarily and spatially varying disper-
       sion coefficient (fnsec)

S   =  s/So, where s(x,t) is the spatial and temporal
       distribution of salinity in ppm)
So

X

L

ET


u
    =  Ocean salinity (ppm)

    =  x/L

    =  Length of estuary to head of tide (ft)

    =  Taylor's dispersion coefficient (f^/sec)
        = 77     *6
       u(x,t) = tidal velocity (ft/sec)
                                                 10-3

-------
                        40
                         35
                        30
                      <
                      (ft
                        25
                                                    (a)
                                8/25      8/26        8/27       8/28      8/29
                        25
                      .-. 20
                         15
                         10
                                                   (b)
                                8/25
8/26
8/27
8/28
8/29
                           faurc«t »«J«ri«o,  «t «1.. D*c»b«r Ittl


   Figure 10-3. Time Varying Salinity Concentrations During and th« Point Pleasant Canal System
n    =  Manning's friction coefficent

Rh   =  Hydraulic radius (ft)

K    =  Estuary dispersion parameter (ft^sec)
        =  UoL/1000

Uo   =  Maximum ocean velocity at ocean entrance
        (ft/s)

m    =  Multiplying factor for bends and channel ir-
        regularities
           In this case, values of K = 58.5 f^/sec and m = 25
           were used. More information regarding the develop-
           ment of this equation can be found in Thatcher and
           Harteman (1972,1981).

           The results of the simulation of salinity concentrations
           at Clark's Landing and Chapman's Wharf are shown in
           Figure  10-3. The observed and computed salinities
           matched well at Clark's Landing, except for two un-
           usually high observed ocean salinity concentrations.
           Because these observed values exceeded the ob-
           served  ocean salinity concentrations on  those days,
           the  modelers concluded  that the data  points  were
                                                 10-4

-------
Table 10-2.  Model-Governlng Rate Parameter*
Transformation

Ki: Ammonification of detritus-N
Ka: Diatom uptake of ammonium-N
    fc0*
    Yz (half saturation constant)
    l( (Optimum intensity of solar radiation)  144
    Topt (Optimum reaction temperature)    10-15
    Tm«x (Maximum temperature beyond    30
    which denaturation of cell protein occurs)
Ks  Nanoplankton uptake to ammonium-N
    K30pt
    Y3
    It
    Topi
    Tmax
Ka: Diatom mortality rate
Xs: Nanoplankton mortality rate
Ke: Copepod uptake of diatom-N
    Keopt
    Y5
    Topi
    Tmax
Kr. Copepod mortality rate
Kg: Copepod excretion rate
«s: Nitrification rate
Ki0:Diatom uptake of (NO2 + NOa)-N
    K,oopt
    Y<
    I.
    Top,
    Tmax
KH: Nanoplankton uptake of (NOz + NOa)-N
    K,iopt
   Ye
   U
   Top,
Rate
0.08

0.415
0.0122
144
10-15
30
\
)

0.553

0.004

64.8

27.5

30

0.05

0.05


0.50
0.10

24-28

35

0.20

0.08
0.15

0.376
0.0063
144
10-15

30


0.501
0.015
64.8
24-28
30

Unit
da/1

da/1
mg-N/1
1y/day
°C
°C


da/1

mg-N/l

ly/day

°C

°C

da/1

da/1


da/1
mg-N/A

°C

°C

da/1

da/1
da/1

da/1
mg-N/l
l/y/day
°C

°C


da/1
mg-N/l
ly/day
C
°C

Transformation
K,2:Copepod uptake of detritus-N
K«*
Y,
Top,
T™,
Ki3:Copepod uptake of nanoplankton-N
Kw*
Y7
Top,

Tm«

Ki4:Dinoflagellate uptake of ammonium-N

K^0"1

Y8

It

Top,

Tmax
K,5:0inoflagellate uptake of (NOz + NOa)-N
K,sopl

Ye

1.

Top,

Tmax
Kte: Copepod uptake of dinoflagellate-N
K,.0"1
Y,o
Top,
Tm«
K,7:Dinoflagellate mortality rate

K,7

Ki«: Sedimentation of detritus
K,8
K,8: Sedimentation of (NOs + NOa)
Kl9

Kio: Sediment release of ammonium-N
KJO
Source: Najarian, etal., December 1981
Rate

0.34
0.250
24-28
35

0.34
0.22
24-28

35



2.075

0.021

288

20-25

32

1.880

0.042

288

20-25

32

0.5
0.1
24-28
35


0.05


0.1

100


10

Unit

day'1
mg-N/l
°c
°C

da/1
mg-N/l
°C

°C



day

mg-N/l

ly da/1

°C

°C

day

mg-N/l

ly da/1

°C
o/*
C

da/1
mg-N/l'1
°C
°C


da/1


da/1

ug-at
N/m2/hr

ug-at
Nm2/hr

                                                         10-5

-------
unrealistic. Based on the plots of observed vs simu-
lated salinity concentration, the calculated dispersion
coefficients were considered adequate.

70.3.2 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 vs. 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  vs. simulated data, the water
quality calibration requires the determination of many
constants. Table 10-2 shows the values that were es-
tablished  through model calibration. These are the
values that best represent  the site-specific kinetic
processes in the Manasquan Estuary while still falling
within the range  of values found in the technical litera-
ture.

Examples  of simulated vs.  observed  plots  for the
various water quality  parameters are  illustrated in
Figures 10-4 to 10-10. The individual symbols indicate
observed datapoints, while the straight line shows the
continuous simulatbn model output. These plots rep-
resent the best  simulation of observed data using
reasonable rate constants and  coefficients. Model
goodness-of-fit was determined only through visual
observation of the plots; no statistical tests were per-
formed.

Figure 10-4 and shows the  simulation of detritus-N,
ammonium-N and nitrite + nitrate-N at Clark's Land-
ing and Chapman's Wharf.  At both sites, the am-
monium-N concentrations predicated by the calibrated
model were reasonably close to observed values. The
simulation of detritus-N and (nitrite and nitrate)-N was
less accurate,  particularly at Chapman's Wharf. The
computed (nitrite and nitrate) 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 DO, NBOD, and Clark's Landing and
Chapman's Wharf is shown  in Figures 10-6 to 10-10.
Like the detritus-N  simulation, an adequate  CBOD
simulation at  Chapman's Wharf was not  possible
without the introduction of a distributed CBOD load.
Even after assuming a load  of 3,500 Ib/day, the ob-
served and simulated data did not match well. The DO
simulation results proved to be confusing at both sta-
tions. Although low concentrations of NBOD  and
CBOD were observed at Dark'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 detrital 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 10-11  and 10-12 show
the simulated concentrations  of  phytoplankton
nitrogen at the two stations. The plots clearly indicate
a strong tidal effect upon phytoplankton concentra-
tions.

10.3.3. 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
                                                10-6

-------
                                                                14
  0.6
 i 0.4
 02
   238
                240         242
               TIME  IN  DAYS
                                    0.6
                                    0.4
                                     0.2
                                             06
                                             0.4
                                             0.2
     12-

     10

     8
  i   •
       238          240          242
                   TIME  IN DAYS
    fourct: Hj«rl.n. tt «l .  D«cM»r mi

 Figure 10-4. Temporal Variation of Detrltus-N,
 Ammonlum-N
     14
     12
     10-
  t  8
  O  e
  ca  5
  u
       238           240           242
                    TIME IN  DAYS
        Source:  Naj«ri«n, et •(..  Beee^>er  1981

Figure 10-6. Temporal Variation of CBOD Variation at
Landing, August 25-28,1980
        Sourc*:  Ntjiricn, «t it.. December  1981

Rgure 10-5. Temporal Variation of C8OO at Clark'*
and Nltrrie-Nitrate-N at Clark's Landing, August
25-28,1980
                                                                  40
                                                                  20
                                                                  .10
          238        240         242
                   TINE  III DAYS

          feurc*: ••jari.n, tt il.. Oteeeber 19S1

Figure 10-7. Temporal Variation of NBOD at Clark's
August 25-28,1980
                                                   10-7

-------
      40
     .30
    8
      .20
      .10
         238         240         242

                   TIME  IN  DAYS
         Sourct: N«j»H»n, tt  »!..  Dtcinb«r 1981

Figure 10-8. Temporal Variation of NBOD at
Chapman's Chapman's Wharf, August 25-28,1980
                                                               14
                                           12



                                        i  10
                                                               a
                                                           X
                                                           o
                                                           s  6
                                                           3
                                                           8  4
                                                             +*
                                             238          24O         242

                                                       TIME  IN  DAYS


                                              Sourer: NiJ»rl«n, n mi.. D*ct«b*r 1981


                                      Figure 10-9. Temporal Variation of DO at Clark's
                                      LandlngWharf, August 25-28,1980
     14
      12
      IO
  x
  O
  O
  OT
  OT
       296
  240          242

TIME  IN DAYS
t«> l*ti«








A A A
Ns
o

CM
bs
a.
a.
O.O8
ITROGEN
Z
O
z
z
s
b
8
238         240         242

           TIME  IN  DATS
                                                                                 °a
                                                                                   a.
                                                                                   a.
                                                                                   So
                                                                                   o
                                                                                 o«
        Soure.: Haj.ri.n.  ft  ft.. 0*e««b«r 1981


Figure 10-10. Temporal Variation of DO at Chapman's

Wharf August 25-28,1980
                                          Source:  lijirtan. tt.it.. D«CMb*r 19(1


                                      Figure 10-11. Temporal Variation of Dlatom-N,
                                      Nanoplankton Observed Elevations at Chapman's
                                      Wharf
                                                  10-8

-------


•
CNJ

I5'
Q.
^
NITROGEI
008
t-
4
5
s.
6
8

• j • Diatom
*l • Hanoptanktoo
• 7 Ofnofl.g.iut.i






tf$Q^
238 ' 24O ' 242
r-
O
•


'65
fc
O.08
>< ITROOEN
O
.5*
6
o
6
»
6
•

CJ
o —
fc
0.08
ITROOEN
Z
1
1°
8
o
               TIME IN OATS
     tourc*: NijiMtn,  «t «1..  Dtctnbir 19«1

 Figure 10-12. Temporal Variation of Dlatom-N,
 Nanoplankton Landing, August 25-28,1980
    u 0
       -2
      -3
        203
         Souret:
  205        207
TIME  IN DAYS
                                     1981
Figure 10-13. Hydraulic Verification: Calculated v».-N
and Dlnoflagellate-N at Chapman's Wharf, August
25-28,1980
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
conditions, and to evaluate the adequacy of the simula-
tion.

The simulation of tidal elevations was investigated first.
The initial water surface elevations, the time-varying
water surface elevations at the tidal boundaries, and
the time-varying freshwater inflow at the head of tide
were established from observed data. The Manning's
"n" values, 0.018 downstream and  0.022 upstream of
Chapman's Wharf, were used without change from the
calibration study.

The simulation of tidal elevations at  Clark's Landing
and Chapman's  Wharf is shown in  Figures 10-13 and
10-14.  The largest difference In predicted and calcu-
lated water surface elevation is approximately 0.2 feet
at Clark's Landing.

The simulation of salinity was the next step in the
verification procedure. The initial and boundary condi-
tions were established in a manner consistent with the
calibration study, using observed salinity concentra-
tions at the sampling stations.  The stratification
parameter K and the calibration multiplier m were set
equal to the values used  In  the  calibration study.
Analysis of Figures  10-15 and  10-16 show again that
observed and simulated values were more similar for
the verification than the calibration. The modelers were
particularly pleased that the furthest inland sampling
station in the estuary, Chapman's Wharf, gave the best
comparison between predicted and observed values.
Based on these results, they concluded that the advec-
tive and dispersive processes throughout the estuary
were well represented in the model.

Finally, the model was verified for water quality proces-
ses. Again, initial and boundary conditions were estab-
lished using observed data and an  antecedent period
simulation, and  all  the  constants evaluated in the
calibration study were used without modification in the
verification study. Because of project constraints, the
simulation of CBOD,  NBOD, and  DO was not  per-
formed. The results  of the water quality comparisons
are shown in Rgures 10-17 and 10-18. As in the calibra-
tion study, the detritus-N concentration at Chapman's
Wharf could not be accurately simulated without the
specification of a distributed detrital load.  The
modelers found that a load of 360 Ibs/day resulted in
an adequate simulation; however, this load was set at
240 Ibs/day  in the calibration study. Because the
detritus-N concentrations calculated in the calibration
study were lower than the observed concentrations,
the modelers concluded that the 360 Ib/day load would
be valid for the calibration as well  as the verification
                                               10-9

-------
          203          205          207
                      TIME IN DAYS
           Sourct: mill-tin, et 11.. Dictator 1981

Figure 10-14. Hydraulic Verification: Calculated v«.
Observed Elevations at Clark's Landing
                                                             I


                                                             58
                                                             •
                                                             ^
                                                             «
                                                             m


                                                               s
                                                           200       202        204        206         208
                                                                             TIM IN OAVf



                                                            l«urc«t H.itM.n. tl  «!.. 0«CMk»r I»i1
                                                        Figure 10-15. Salinity Verification: Calculated vs.
                                                        Observed at Chapman's Wharf
        202         204         206        208

                TINE IN OATS



ircit lljiriin. .t it..  Bic«b.r 1911
Figure 10-16. Salinity Verification Calculated vs.-N
and Dlnoflagellate-N at Clark's Landing, August
25-28,1980
                                                                  l§
                                                                  . o
                                                                   o.
                                                                   ««
                                                                   C
                                                                     203
                                                                                205        207
                                                                               TIM II MTI
                                                        Figure 10-17. Temporal Variation of Detrltus-N,
                                                        Ammonlum-N Observed at Chapman's Wharf
                                                     10-10

-------
       §
     is
     S o
       o
       rg
       6
       8
       cs
        203
                   205
                  TIME II 0*Tt
207
       if
             O
             O
,81
 o ~
   Figure 10-18. Temporal Variation of Oetrltue-N,
   Ammonlum-N and NIlrlte-Nltrale-N at Clark's Landing
period.  Other than the ammonium-N concentrations
predicted at Chapman's Whan" and Osborn Island, (a
station that is not included in the modeling report), the
water quality simulation was considered satisfactory.

10.3.4. 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-
tion was made that the model be used to:

(1)  Determine the existing and  potential impact  of
non-point 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.

10.4 References
Najarian.T.O., Kenata, P.J. and Thatcher, M.L Decem-
ber, 1981. Manasquan Estuary Study. Manasquan
River Regional Sewerage Authority.

Thatcher, M.L and Harleman, D.R.F. February 1972.
"Mathematical Model for the Prediction of Unsteady
Salinity Intrusion in Estuaries," Technical Report No.
144, R.M. Parsons Laboratory for Water Resources and
Hydrodynamics, Department of  Civil Engineering,
M.I.T., Cambridge. MA.

Thatcher, M.L, and Harleman, D.R.F., February 1981.
"Long-Term Salinity Calculation in Delaware Estuary,"
Journal of the Environment Engineering Division,
ASCE. Volume 107, No. EE1, Proc. Paper 16011,, pp.
11-27.
                                               10-11

-------
                    11.  Calcasieu River Estuary Modeling
11.1. Background
The Calcasieu River Estuary modeling  study is
presented here to illustrate a time-variable waste load
allocation model applied to a complex Gulf 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 St. Charles,
Louisiana, extending  downstream to the Intracoastal
Waterway (shaded area in Figure 11-1). The primary
water quality problems were the result of point source
discharges. There were 64 wastewater dischargers to
the Calcasieu River below the salt water barrier.  In the
forty-mile study area,  there is a seven mile reach (be-
tween 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-II model for the Cal-
casieu River basin (Duke, 1985). Duke built on the work
of AWARE and  Weston by improving the calibration
procedure and using  new estuary cross-section infor-
mation.

11.2. Problem Setting

11.2.1. 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 lower 40 miles of river, below
the salt water barrier (Figure 11-1).

The Army Corps of Engineers constructed the barrier
and maintains a dredged a 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
 M»n>. no**, IIM
                                                     Rgura 11-1. Calcasieu Estuary Study Area
                                               11-1

-------
term mean flow of 2,600 cfs and a 7Q10 of 202 cfs. The
7010 below the salt water barrier was estimated to be
375 cfs.

11.2.2. 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              9)  Total Kjeldahl Nitrogen

5)  Phosphorus       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
nonpoint source survey, a nitrogen  transformation
study, a sediment oxygen demand study,  a use-at-
tainability study, a series of mini-surveys for in situ
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 channel
below the salt water barrier was stratified with respect
to salinity and dissolved oxygen. The channel had once
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/nitrate,
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.
 11.3. Model Application

 11.3.1. 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
    nonconservatK/e)
 •  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. The one  exception was a change in the
 hydrodynamic module that allows simulation through
 numerous tidal cycles.

 11.3.2. 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 11-2). The number of cycles of
the simulation were increased because the  short
 simulations of earlier modelers had not achieved
 steady-state.

 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 guff, downstream from
the barrier, and  from tributary  inflows. After 13  days,
                                               11-2

-------
                                  Somc« Putt, l»«3
Figure 11-2. Modal Segmentation Diagram

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.

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 11-1.
These values were changed spatially within each sur-
vey but not changed from  survey to survey.  Model
inputs for forcing conditions (e.g. tides, temperatures,
flow,  etc.) and  loading were as measured for each
survey.

11.3.3. 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 and system tran-
sients were complicating factors.

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 tidal averaged results were compared
to data.

Hydrodynamics:

The model  calibration  results  for August 1979
hydrodynamics  are summarized in  Figure 11 -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. The
elevation differences were  considered insignificant.
The model verification comparisons  were similar for
July 1980 and June 1984 in that the model matched the
trends well but was inconsistent in matching the mag-
nitude. However, for the July 1978 data set (see Figure
Table 11-1.  Major Rates and Parameter* Uaad In Mod*)
Coefficient
                      Units
           Rang*
Manning's n

Ammonia Oxidation

Nitrite Oxidation

BOO Oxidation

Benthic Oxygen Demand

Reaeration
none

per day

per day

per day
0.018-0.035

0.002-0.020

1.00

0.001-0.050
gm/sq.m/day 0.75-1.50

per day     0.003-2.000
                                                 11-3

-------
         .  /"X   ./
                                   •oure«i Duk*. 1MI
    Flgur* 11 -3. Tidal Stag* RMUIU for August 1979
    Hydrodynamic Calibration Simulation
                                                                                  MIWMv taRM>
                                                            ClttMlM Ltck
                                                                                       fovreai BMk«, ItM
   Flgur* 11 -4. Tidal Stag* RMulta Hydrodynamic
   Verification Simulation (July 1978 Data S«1)
 11 -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 the
 ten parameters described above (Figures 11 -5 through
 11-9). Figures 11 -5 and 11 -6 are selected  model com-
 parisons for a few parameters from the August 1979
 model calibration. Figures 11-7,  11-8, and  11-9 are
 selected results for the three verification  simulations.
 The water quality calibration/verification  match  was
 characterized as fair, with  many discrepancies at-
 tributed to poor information on loading conditions and
 dynamics.

 11.3.4. Model Sensitivity
An important modeling activity  is sensitivity analysis.
This procedure tests the sensitivity of model calcula-
tions to changes in selected inputs. Results can be
used to:

 •  Refine coefficient selection
 •  Identify the most Important processes and loads
 •  Identify areas In need of better data to improve
    modeling
 •  Define model uncertainty
The model was tested for an elimination and tripling of
BOD and ammonia deoxygenation rates and elimina-
tion of algae. The results indicated that algae had the
largest effect on the water quality  calculations.  This
finding is common to estuaries where algal abundance
often is the major factor in controlling water quality. As
a result, success or failure in model validation to data
can depend on proper characterization and simulation
of algal dynamics.

11.4. Total Maximum Daily Loads
The purpose of all modeling efforts on the Calcasieu
was to develop total maximum daily loads fTMDL) 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 wasteloads for BOD
and NHa-N.
                                                 11-4

-------
                      10
                   <
                   tn
                   „ 6
                                                                          Soureci Duk«,  19tS
-M«««uf«d
- Slmuloltd
                              — Mtasurtd
                              "Simulaltd
                    o  8
                    I4
                     82
                    o
                    o
                    «  n
                              "Simulaltd ultima)*
                                          K>
                     IS       20       23
                      DISTANCE (milli)
                                                                                     35
 Figure 11 -5. S«lcc1ed Water Quality Results for Verification Simulation (July 1978 Data Set)
                             03
                             "
                             (2

                           - ID

                           lo,
                           m
                           z M

                             04

                             O.Z
                                                                                    144
                                                                                ••••111
                                                                             fftmw>M*» :I2I
                                                  ,.
                                                                    h."-iT-l-
                                         •
                     TS"	£	«
                       DISTANCE (•!!••)
                                                                               U
Figure 11-6. Selected Water Quality Reeutt* for  Hydrodynamlc Calibration Simulation
                                                       11-5

-------
                       12
                     2  2
                        0
                     _ K)
                     %n
                     I *

                     3 «
--Simulated ultimo!*
                                                 15       20      23
                                                   DISTANCE (milll)
                                                                         30
                                                                                        40
Figur* 11-7. Selected Water Quality Results for Calibration Simulation (Auguct 1979 Data Set):  Phoephate, Total
KJeldahl Nitrogen, and Ammonia
                             I.,
     ..I
                                      1	\
I.'.
                                       All IM««wr«4 4«<4
                                                   IS     20    »
                                                    DISTANCE (*I!M)
       11^ Selected Water Quality Reaultt forCallbratlon Simulation (July 1978 Data Set):  Salinity, Dissolved Oxygen,
and Biological Oxygen Demand
                                                       11-6

-------
                                                 20    29
                                              DISTANCE (milti)
    Figure 11-9. Selected Water Quality Re*ult«for Verification Simulation (July 1978 Data Set): Phosphate, Total Kjeldahl
    Nitrogen, and Ammonia
 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/I 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/I. 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. Using
the 4 mg/I DO standard and 1979 loading pattern, the
Duke study produced an estimated TMDL of 83,130 Ibs
UOD/day.
11.4. References
Duke, James H., Jr., "Calcasieu River Basin, Louisiana,
Modeling Stud/1, 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,  14 August 1985.

Raytheon Company, "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,
December 1974.
                                               11-7

-------
                      12. 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.

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 Books I through III 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.

12.1. Robert V. Thomann, Ph.D.
Professor, Environmental Engineering and Science
Manhattan College
Riverdale, New York 10471

121.1. INTRODUCTION
My overall opinion on the appropriate level of estuarine
water quality model complexity can be summarized by
the observation that:
  THE BEST MODELS ARE OFTEN THE SIMPLEST
The review therefore will continually display a  bias
towards doing estuarine water quality modeling in as
simple a fashion as possible and only after all simplicity
has been exhausted, should increasing complexity be
introduced and then only after careful consideration is
given to the improvements in the model that might be
realized. The reasons for this bias are: (a) most analysts
have only limited experience, time and resources avail-
able, and (b) unnecessarily complex models some-
times tend to obscure uncertainty behind a facade of
"reality".

The choice of the appropriate level of model com-
plexity is determined in large measure by the nature of
the problem under investigation. The context for my
opinion on an appropriate level of model complexrty 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. Nonpoint and 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
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 which also  incorporates the
basic theoretical components, but has no utility in a
WLA until applied to a specific  problem setting. The
                                               12-1

-------
computer code of a generic model is transportable, a
model of a given estuary is not.

Thus, it 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 WLA

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 are
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 contitions, usually one
dimensional, constant parameters and steady state,
although solutions for some time variable inputs exist,
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 reasons or to check on
numerical model  results. I do know, however,  that
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 model development. Similarly, analytical solu-
tions 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 "dose
in" on which particular phenomena were of importance
in describing the observed data.

In spite of the severe  assumptions that must be in-
voked, it is strongly suggested that:
   ANALYTICAL SOLUTIONS SHOULD BE USED
   TO COMPUTE INITIAL RESPONSE AND TEST
      NUMERICAL MODEL COMPUTATIONS.
Such computations provide the first approximations to
the order of water quality response that might be ex-
pected from input loading under different hydrological
regimes and model parameters. Also,  the use of
analytical models provides a first order check on more
complicated numerical models to determine whether
the numerical computations are approximately cor-
rect

C. Model Evolution
The use of models in decision making must recognize
that, very often, the understanding of estuarine proces-
ses, and the availability of data and model frameworks
for a given estuary are always changing.  Models are
not static, but rather continually evolving. Decision
makers must be apprised of this fact and must, to some
degree, be prepared for new input into the decision
process.

The Saginaw Bay and Potomac estuary case studies
are good examples of models that began at relatively
simple levels of complexity  and have  subsequently
progressed to more complex kinetics and spatial and
temporal detail. The progression was dictated by an
ever increasing level of complexity In the questions
being asked of the  model.  For example, the early
Potomac estuary  models did not explicitly include
phytoplankton dynamics. But after issues of nutrient
controls (e.g. should  phosphorus  or nitrogen be
removed?) were raised, an expansion  of existing
models was required. However, as noted below, It is
not always clear that adding additional  complexity
improves credibility. Thus, for the Saginaw Bay model,
it 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:
 ALL MODELS MUST CONTINUALLY BE UPDATED:
    IF NOT, MODEL "ATROPHY" SETS IN AND
    CREDIBILITY DETERIORATES. ESTIMATED
 MODEL "HALF-LIFE" IS ABOUT ONE-TWO YEARS.
                                               12-2

-------
                             mJ
                             CD
                             O
                             Ul
                            §
                                            LEVEL OF MOOa COMPLEXITY

                                                1EW3RM. SCALE
                                                 SPA.TIH. SCALE
                                                KMTTC WTIRACTKJNS
 Figure 12-1. Illustration of Relationship Between Model Credibility and Model Complexity
 Existing models must therefore never be "frozen" in
 time and continue to be used in the face of obvious
 model inadequacies. As painful as it may be, some
 model frameworks need to be restructured, expanded
 or even abandoned as new information becomes avail-
 able.

 12.1.2. 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
 12-1 illustrates this opinion. In general, increasing vari-
 ables both in absolute number and over space and
 time.  Even more importantly, increased model com-
 plexity requires a detailed data base across all state
 variables and over space time for a complete assess-
 ment 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 probabilistic 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.
                                                12-3

-------
Rather, data are more frequently available at irregular
time intervals, but with some spatial definition. Finally,
developing fully time variable models at an intra-tidai
level is a complex time consuming effort with a neces-
sity to  conduct extensive data  analysis and output
processing in order to display model results in a defen-
sible manner.

The case studies showa range of temporal scales, from
the steady state analyses of the  Calcasieu estuary to
the intra-tidal models of the Manasquan estuary and
DO and Total residual chlorine in the Potomac estuary.

The intra-tidal choice for the  Manasquan (over two
4-day  periods) is  not considered  to be the correct
choice  since the water quality problem under study
involved kinetic behavior over time 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
need 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 the
problem. The fineness of the spatial extent of the model
is to some degree coupled to the temporal issues noted
above.  Generally,  long time scale problems may in-
volve larger scales and less detailed spatial definition.

The chlorine model of the Potomac is an example of
where cross-estuary gradients needed to be computed
necessitating a spatially detailed  model in the lateral
and longitudinal direction. The Saginaw Bay  model
consisting of five segments  is a  good  example of a
reasonable grid since a finer spatial definition would
probably not contribute to any  improved model
credibility.

Finally  a remark should be made about model boun-
daries.  The  extent of the model should always be
sufficiently far removed from any existing or proposed
inputs that may be subject to a WLA The boundaries
should be at a point where the flows and exchanges
and state variable concentrations can be specified and
 are independent of the model output For example, it
 is not entirely clear from the Manasquan case study
 that the model boundary is proper, i.e. the extent of the
 model may have to be extended out past the Inlet In
 order to provide a proper independent boundary con-
 dition. This may be especially true if the model had ever
 been used for analysis of the proposed regional Input
 at the head of tide.

 C. Suggested Strategy for Temporal-Spatial Scales
 Since the principal reason for estuary modeling in the
 context of this opinion is a WLA, the following strategy
 for choosing a relevant  temporal-spatial modeling
 scale is offered.
         TEMPORAL-SPATIAL SCALES
          BEGIN WITH STEADY STATE,
            "LARGE" SPACE SCALES,
               THEN SEASONAL,
      MORE DETAILED SPATIAL DEFINITION,
         THEN INTRA-TIDAL. FINE SCALE.
It is suggested that the temporal scale of most WLA
estuarine models should begin at steady state to deter-
mine  overall relationships between Input loads and
resulting water quality. Steady state Is suggested even
for highly reactive variables since the steady state
modeling helps to define overall response levels and
spatial extent of the input loadings.

Following steady state analyses, If time variable
analyses need to be done in estuaries (as a result, e.g.
of a need  to  specify phytoplankton dynamics for
nutrient control or a seasonal WLA) then a seasonal
time scale (with a model framework representing an
average over a tidal cycle) should be used.

Only if the justification is quite clear, (e.g., transient
storm water  input  analyses or a complicated
hydrodynamic regime  as  in the  Potomac estuary
chlorine model) should an intra-tidal model be con-
structed. The fact that the estuary has a tidal oscillation
Is in itself not justification for constructing an intra-tidal
model. The reason is threefold: (a) as noted earlier, the
focus here Is on WLA problems which are normally
limited in resources, time and money, (b) most WLA
problems involve processes that  have longer  time
scales than tidal, and (c), there are many other sources
of  temporal  variability in water quality that are not
captured by intra-tidal calculations  (e.g. hour to hour
and highly local changes in solar radiation, suspended
solids, wind, or velocity,-among others).

It is suggested that initially a relatively crude spatial
representation  (e.g. a numerical grid size  of several
                                                12-4

-------
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.

12.1.3. 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  seem that a mathematical
model of the hydrodynamics of the estuarine system If
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.

721.4. 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
inputted? For a phytoplankton model,  should vanous
functional groups be modeled or should total
chlorophyll be used? Should sediment nutrient fluxes
be calculated or inputted?

As a general rule, I would advise to:
     KEEP STATE VARIABLES TO A MINIMUM;
  MODEL ONLY THOSE FOR WHICH DATA EXISTS;
    BUT ALWAYS INCLUDE THOSE STATE VARI-
   ABLES WHICH WILL BE IMPACTED BY A WLA.
The case studies seem to have implicitly recognized
this general rule, although there are some exceptions.
The Manasquan model is clearly over-specified with
state variables and kinetic interactions for the nature of
the problem under study and the available data sets.
The inability to calibrate to the phytoplankton  state
variables severely limits the utility of the model.

On the other hand,  the initial  Potomac  estuary DO
model did not explicitly include organic nitrogen, nor
ammonia uptake by phytoplankton. Also photosyn-
thetic DO sources and sinks were externally inputted,
but these inputs were to be extensively impacted by a
WLA  for nutrients. The model could not therefore
respond to the WLA questions associated with the
affect of nutrient control on DO.

Sometimes a state variable must be included even If
data are not available, for example, for a toxic chemical
model, both dissolved and particulate chemical  must
be  modeled. But data  may not be available for the
dissolved component because of concentrations
below a  detection limit.  Nevertheless, both com-
ponents need to be included  in  the modeling
framework.

121.5. 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 Model "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 "unrepresentatrveness"
                                               12-5

-------
      SIMPLE MODEL
     Anolytical Solutions
      Numerical Model
       Steady State
     MODEL CREDI8LITY
CALIBRATION - VERIFICATION

 Temporal/Spatial Comparisons
 Statistical Tests/Comparisons
                          INCREASE  MODEL COMPLEXITY
                                 State Variables
                               Kinetic Mechanisms
                               Temporal/Spatial Detail
     Marginal
Sufficient
   WASTE LOAD
ALLOCATION ANALYSIS
   Critical Conditions
  Sensitivity Analysis
    Load Projection
 Components Analysis
                                                       ALLOWABLE LOAD
                                                           PERMIT
Figure 12-2. Suggested Strategy for Determining Appropriate Level of Model Complexity
is obvious.  We  know  when a  model is  not  repre-
sentative. The Calcasieu case study is offered as an
example of a model that is claimed by the analyst to 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 opinion,
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 cannot
judge the adequacy of the model in reproducing lon-
gitudinal variability.

The  Potomac estuary DO model was compared to
various data sets by readjusting the model parameters
for each calibration. This is unacceptable. The purpose
of model calibration and verification is not to "force fit"
the model to the data.  Rather, the model  parameter
numerical assignment should obey the
            PRINCIPLE OF PARSIMONY
   BE "STINGY" WITH THE SPATIAL AND TEMPORAL
   SPECIFICATION OF MODEL PARAMETERS. HAVE
  A REASON FOR ALL ASSIGNMENTS
  The Potomac estuary phytoplankton and Saginaw Bay
  models offer  extensive calibration and  verification
  analyses, including various statistical measures of
  comparisons. Both spatial and temporal comparisons
  and statistics of comparisons are given. These case
  studies provide some measure then of an adequate
  representation of the data by the model and can be
  profitably used as a "model" of a model calibration.
  Two caveats are in order, however: (1) extensive data
  sets and resources were available in both cases, and
  (2) even with the extensive calibration and verification
  of the Potomac eutrophication model, a bloom in 1983
  was not captured because of presumed pH mediated
  sediment phosphorus release, a mechanism not pre-
  viously included in the model.

  121. & Summary
  Figure 12-2 summarizes all of the above comments.

  As indicated, the suggested procedure is to begin with
  simple representations of the estuarine system. This
                                               12-6

-------
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.
    THE SIMPLE MODEL MAY TURN OUT TO BE
   ENTIRELY SUFFICIENT FOR WLA PURPOSES.
If a  determination is made that  the  simple model
provides only "marginal" model credibility, then model
complexity  should be  increased. This increase in
model complexity often needs to proceed in the follow-
ing order: (a) additional state variables, (b) additional
kinetic interactions, (c) increased temporal and spatial
definition. It is in the latter that hydrodynamic modeling
may be  necessary.

Additional calibration and verification is  then  con-
ducted  with the hope that  model credibility is  in-
creased. This step should include, whenever possible,
comparisons to data sets collected over a range of
environmental and input  loading conditions.

After a determination has been made, then a full WLA
analysis can be conducted.  This analysis  should  in-
clude evaluation of water quality response under criti-
cal design conditions, sensitivity analysis, projection of
expected loads in the future and components analysis
of individual inputs. This latter analysis is  aimed at
describing the relative contribution to the calculated
response from individual components,  e.g. particular
point source inputs, and  distributed sources (such as
 sediment sources). The analysis often provides key
 insights into which inputs and mechanisms are most
 important in the WLA. (None of the case studies dis-
 played any components analysis).

 The final outcome is then the recommended WLA for
 an input or region with associated permit specifica-
 tions. It is this final outcome that should always be kept
 in perspective when assessing the need for various
 levels of model complexity. Ultimately, of course, the
 measure of success of the model Is the degree to which
 the model projections are actually realized after the
 WLA has been implemented. But that is a topic for
 another opinion at a different time.

 12.1.7. 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 8-3 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, Dlustrates a range of modeling ap-
proaches to estuarine water quality.

DYNAMIC ESTUARY MODEL

The first effort, the use of the Dynamic Estuary Model
 (DEM) examined the DO resources of the estuary. The
                                                12-7

-------
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
constituents.

POTOMAC EUTROPHICATIQN  MODEL

The Potomac  Eutrophication Model (PEM)  is an ex-
ample of an  intermediate scale of  estuarine water
quality model.  The use of a coarse grid in the lower
estuary was justified on the basis of the lack of any
significant gradients in water quality constituents of
interest. Vertical homogeneity is a key assumption 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 compansons (see
9.4.6) between the model output and the observed
data.

Like the DEM.  the PEM was subjected to a post audit
analysis. The analysis was prompted by a major algal
bloom in the summer of 1983.  As noted « the case
studv summary, pg. 9-25 ff., the PEM was not able to
                    of the observed bloom, due in
 some measure to a significant source of phosphorus
 that was not incorporated in PEM. Subsequent work
 indicated that such a source may have been from a 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 eutrophicatlon estuarine
 system.  It also illustrates the hazards of apparently
 "best"  calibration of the model  that misses a
 phenomena which only appears after certain condi-
 tions ensue. Nevertheless, the PEM proved useful in a
 variety of decision making contexts, not the least of
 which was to assess the reasons for the major 1983
 algal bloom.

 FINITE ELEMENT CHLORINE MODEL

 This model is a very good example of the proper choice
 of time and space scales. Because the decay rate of
 chlorine is so rapid, the zone of influence of the chlorine
 residual would be expected to be highly local. As a
 result, this model has as it's spatial focus, a region of
 about five miles centered at the location of the major
 input.  Detailed lateral  specification Is required  be-
 cause 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 9-14 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
 calibration.  It was concluded however, that the disper-
 sion  was property captured In general.

That conclusion Is a good example  of a judgement
 made by the  analysts  on the suitability of a model
calculation. This reviewer believes that the judgement
made here is correct, but only because of the calibra-
tion analysis to the observed dye data.

A similar conclusion can be drawn with respect to the
calibration of the total residual chlorine model to survey
data. What was required here was approximate repre-
sentation of the general field of the  chlorine, to  ap-
proximate order of magnitude.  This was achieved.
More importantly, the sensitivity analysis indicated the
degree of model uncertainty and this is clearly dis-
cussed. That uncertainty did not affect the basic con-
clusions.
                                                12-8

-------
SUMMARY

These three modeling efforts of the Potomac estuary
water quality Illustrate a good range of spatial and
temporal scales, level  of model complexity and the
need for extensive calibration and verification to ob-
served data. Two major points seem to emerge:

  •  Uncertainty in the model coefficients sometimes
     does not affect the conclusions, i.e., the
     decisions that are reached.  But under certain
     situations, a model that is believed to be proper-
     ly calibrated can miss entire phenomena or
     linkages.  Such a model may then fail in varying
     degrees during a post audit.  The experience of
     the Potomac esturary models summarized in
     this case study should be borne in mind by any
     analyst.
  •  Each problem requires its own spatial, temporal
     and kinetic level of detail.  Finer spatial and tem-
     poral resolution is often not the issue especially
     when the problem context is over a larger time
     and space scale.  Funding and project comple-
     tion times are realities that must be faced in any
     modeling effort. Such constraints must be
     balanced against more and more detail in the
     modeling framework with perhaps less than
     desired return in improving the certainty of
     decision making.

Case Study 3 - Manasquan Estuary
This model illustrates the use of an intra-tidal calcula-
tion to describe estuarine water quality.  This reviewer
believes that the proper temporal scale was not used.
By focusing in on two  4 day periods as examples of
"calibration" and "verification", the model does not
capture the longer term, i.e., week to month, behavior
of the water quality constituents of interest. Further,
the analysis is flawed in several ways. The August 1980
period is used as a calibration set and July 1980 is used
as a verification data set. What would be much more
convincing is to use the model in one complete calcula-
tion  extending  from prior to July 1980 through the
August  1980 data.  By restarting the calculation each
time before August  and July and then extending the
calculation for only four days, the credibility  of the
model is severely compromised.

Also, this model is presented as a demonstration of a
"successful calibration  and verification of a real -time
estuary model". This reviewer does not agree that this
model is successfully calibrated and verified even for a
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 10.9 and 10.10).  Also, the CBOD, NBOD
 and nitrogen forms are not calibrated.  For example,
 Figure 10-18 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-Cakasieu 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 dear 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,
                                                12-9

-------
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
allowable UOD load with the observed data, as noted
above, is illustrative of the tenuous nature of the model
for use in decision making.
                                               12-10

-------
 122. Donald R.F. Harteman, Ph.D.
Professor, Department of Civil Engineering
Massachusetts Institute of Technology
Cambridge, Massachusetts

12,2,1. Introduction
The concept of a technical manual  for performing
wasteload allocation in estuaries is an excellent one.
Part IV of the manual is intended to be a "critical review
of estuarine wasteload allocation modeling." It consists
of four case studies "representing various levels of
complexity." 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.

7222 Statement of the Problem
Many environmental problems require the develop-
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 to the
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.

1223. Data
Available data, both hydrodynamic and water quality,
must be studied in order to understand the spatial
complexity of the problem. In the hydrodynamic area,
it 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 and
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 sig-
nificantly 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.

7224. 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 dear in the presentation of the four case studies
of part IV.

A. Box Models
Box models require an empirical, rather than 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 8.0 of Saginaw Bay on Lake Huron shows
the entire Bay represented by five boxes (see Fig. 8.2).
Case study 9.4, Potomac Eutrophication Model (PEM),
uses a box network consisting  of 23 main channel
longitudinal segments and 15 lateral tidal embayment
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 defini-
tion, each  box is a fully-mixed compartment.

Case study 8.0 (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 deter-
mined. In addition, there is no information on the sen-
sitivity 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 temporal resolu-
tion,  it is  difficult to see  how this  model would be
applicable to estuary studies.

Case  study 9.4  (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  pf 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
                                               12-11

-------
satisfies a given salinity distribution requires the solu-
tion of an inverse problem for which there is not unique
solution. Furthermore, the spatial distribution of advec-
tion and dispersion coefficients will change  in time in
relation to factors such as fresh water inflow, which
change the longitudinal and vertical distribution of
salinity.

B. Hydrodynamfc 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 IV which falls within the
realm of multi-dimensional hydrodynamic modeling is
9.5 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. 9-13 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. 9-14)
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 IV 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 ad
Hudson estuaries are examples of reasonably well-
mixed estuaries.

Case study 10.0 (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 IV are 9.3 (Dynamic
Estuary Model) applied to the upper  portion of the
Potomac estuary and 11.0 (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
there must be calibrated using conservative tracers. A
characteristic of this class of model is their inability to
simulate the steepest portion of the longitudinal salinity
gradient due to excessive longitudinal  numerical dis-
persion (See Fig. 9-3).

The Calcasieu estuary case study (11.0) states that the
model contains no dispersion. The so-called
hydrodynamic verification for tidal stages is very poor
(See Fig 11.4). This  reviewer would not recommend
further use of this model for estuarine studies.

7225. Temporal Resolution of Models
The question of the  temporal resolution of estuary
models - intra-tkJal or tidal-averaged - is discussed  in
this reviewer's comments in Parts I and II.

7226. Water Quality-Eutrophication Models
The water quality components of most of the waste
load allocation models in Part IV 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-
eutrophicatlon models presented in Part IV. 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-
                                               12-12

-------
ment plants as opposed to control of agricultural fer-
tilizer runoff depends on the model's ability to deal with
nutrient limitation kinetics. The problem is complicated
by the fact that most estuaries include upstream fresh
water portions as well as the downstream salinity in-
trusion region. Algal species and nutrient preferences
may shift between the  fresh-salt water zones of an
estuary.

(b) A significant number of estuaries experience sum-
mer anoxic conditions in deep bottom zones. Very low
or zero dissolved  oxygen is known to trigger major
increases in the release of nutrients from the bottom
sediment  to  the  overlying water. Eutrophication
models applied to estuaries having low DO problems
must have the ability to simulate the vertical stratifica-
tion and vertical mixing processes that affect vertical
oxygen transport and dissolved oxygen gradients and
benthic nutrient fluxes.

(c)  The determination of the time scale at which an
estuary responds to changes  in waste load inputs
depends on how sediment-water column interactions
are modeled. Waste  load (nutrient) inputs generally
result in algal production in the upper euphotic zone.
Dead algae  sink  and are incorporated as  organic
material into bottom  sediment. Sediment diagenesis
occurs in the sediment and results in nutrient fluxes and
sediment oxygen demand, the rate at which the sedi-
ment diagenesis occurs controls the rate at which the
estuary responds to loading changes. Important
papers in this modeling area are contained in Hatcher
(1986).

Attention should be given in this document to a report
prepared by ASCE Task Committee on the Verification
of Models of Hydrologic Transport and Dispersion
(Ditmars et al,1987). The objective of the report is to
identify, collate, and define the procedures required to
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.

1227 References
ASCE  Task  Committee on Turbulence Models in
Hydraulic Computations, ASCE Journal of Hydraulic
Engineering 114(9), September, 1988.
Bloss, Siegried; RalnerLehfeldtand John C. Patterson.
Modeling turbulent transport in stratified estuary. ASCE
Journal of Hydraulic Engineering 114(9). September.
1988.

Blumberg. Alan F. and David M. Goodrich. Modeling of
wind-induced destratification in Chesapeake Bay, Es-
tuaries, March, 1990.

Ditmars. J.D., E. Eric Adams, Keith W. Bedford, and
Dennis E. Ford. Performance evaluation of surface
water transport an dispersion models. ASCE Journal
of Hydraulic Engineering 113(8), August, 1987.

Hatcher.  Kathryn  J.  (editor). Sediment  Oxygen
Demand:  Processes,  Modeling &  Measurement.
Athens, Georgia, Institute  of Natural  Resources.
University of Georgia, 1986.

Thatcher,  M. Llewellyn and  Donald R.F. Harleman.
Long-term  salinity calculation in Delaware estuary,
ASCE Journal of the  Environmental  Engineering
Division 107(EE1), February,  1981.
                                               12-13

-------
12.3 Gerald T. Orlob, Ph.D., RE.
Professor, Department of CMI Engineering
University of California
Davis, California

12.3.1. 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. Naturally,
the views expressed here are reflective of this ex-
perience and are uniquely those of the reviewer, for
which he takes full responsibility.

Before examining the  specifics of the selected case
studies, it is appropriate to identify a few of the charac-
teristics or attributes to the modeling process that need
special attention in bringing models to a level of effec-
tive application as decision support tools. Among the
more important of these are the following:

•  Specific goals or objectives of users to be met by
    the use of models and  related decision support
    capabilities.
•  Basic data and information required for construc-
    tion, calibration, and verification of model(s).
•  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 case
    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.

12.3.2. 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.

12.3.3. 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/ecdogic 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.

723.4. 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
                                                12-14

-------
and temporal detail, and depending on empirically
derived dispersive fluxes to describe transport of pol-
lutants. Model simplification usually means greater un-
certainty, which in water quality simulation is usually
manifest In empirical dispersion coefficients which rep-
resent the aggregate effects of many Hi-defined quan-
tities.

123.5. 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 of rate con-
stants, coefficients, etc. While this is a commendable
trend  In the sense of improved understanding of the
aquatic system, ft 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.

12.3.6. Calibration and Verification
Although modeling implicitly requires comparison  of
simulation and prototype  observations, and  most
modelers comply with the two step process, the prac-
tice is still largely judgmental. There are comparatively
few examples of rigorous  objective assessment  of
model reliability. There is a need for formalizing calibra-
tion and verification procedures, perhaps  along the
lines of the uncertainty analysis  approach suggested
above.

123.7. Case Study Review
The five case studies were ostensibly selected for the
diversity of modeling approaches to characterization
of estuarine water quality. They were chose also, so it
appears, to represent  a range of  difficulties en-
countered in applying existing models  to actual es-
tuaries and  further to  Illustrate varying degrees  of
success in overcoming these difficulties. No ideal ex-
amples are provided, since none actually exist. How-
ever, while those chosen for this review can fairly be
regarded as instructive  of some of the problems en-
countered in the real world of water quality modeling,
they may not be as exemplary of estuarine modeling
per se as one would like. In two of these cases, as we
shall see, there is reasonable doubt that the systems
modeled can even be categorized as estuaries within
the definitions provided  in the technical  guidance
manual.
Yet, the several case studies do represent applications
of a number of different  models and  it Is useful to
examine these for comparative purposes. Because the
type of estuary often dictates the structure of the model
most suitable for simulation of Its hydrodynamic and
water quality behavior, this reviewer has  chosen to
organize his critique according to the specific
geographic situation.

Case Study 1 - Saginaw Bay
This is a non-estuary, at  least in so far as classical
definitions apply. The major difficulties here appear to
be most likely associated with characterization of
transport rates, both advective and dispersive. Al-
though It Is acknowledged that "water levels and flow
directions of the Bay change" there is no explicit treat-
ment of the hydromechanical behavior of the water
body. Admittedly, this is not a trivial undertaking from
a modeling viewpoint, although there are some excel-
lent examples of two- and three-dimensional circula-
tion models for the Great Lakes that would probably be
suitable for Saginaw Bay.

The complexity of the ecosystem dynamics repre-
sented in this model and the rough "box" configuration
of the embayment, both suggest a greater interest in
ecosystem kinetics than in the practical problem of
waste load allocation. Both of these aspects, simplicity
in the one extreme (5 boxes) and complexity in the
other (18 plus compartments), lead to increased uncer-
tainty in the results of simulations. This reviewer sug-
gests that perhaps a better result from the point of view
of the decision maker might have been obtained with
a somewhat more rigorous description of lake circula-
tion and some aggregation  in ecosystem compart-
ments. The trends indicated by the results shown seem
hardly sufficient for decision purposed  in light of ap-
parent uncertainties  in model  parameters and field
data.

This is a case where it seems that the third spatial
dimension could be especially important in the model.
To what degree does stratification of water quality
variables play a role  in determining primary produc-
tivity? What about vertical advection and dispersion? It
is not dear 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
                                                12-15

-------
the introduction  of more state variables, such as a
model 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
most important attribute.

Case Study 2 - Potomac Estuary
Here we deal with a real estuary, but only partially. The
focus in this example of the application of the Dynamic
Estuary Model (DEM) and the Potomac Eutrophication
Model (PEM) is on the upper "fresh water" section of
the estuary,  where the effects of tidal oscillation are
minimal. In this region dispersive effects Induced by the
tide (which is of  rather small amplitude, anyway) are
probably negligible and stratification is unlikely to be a
significant consideration.

DYNAMIC ESTUARY MODEL
A redeeming feature of the DEM application is that 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 in the
estuary. A limitation of the model(s) is that the basic
configuration is one-dimensional, that  Is , flows are
directionally constrained. Pseudo two-dimensional
representations  are possible for shallow vertically
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
Delta for which 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
far field level. The practice of varying model coefficients
from one survey to the  next in an arbitrary manner in
order to assure the "best fit" is purely subjective and
should not be encouraged. If such as procedure is
employed, a rational basis for parameter adjustment
must be provided. After calibration and verification for
the Potomac study the DEM model appears to have
been provided with most of the attributes of a useful
decision support tool.

POTOMAC EITTBOPHICATION MQDEL
The PEM has many of the characteristics of QUAL 2E
or WASP 4, in that it is essentially a box model for which
the boundary fluxes are governed by either a simple
hydrologic mass balance or are generated by an exter-
nal hydrodynamic model like that in DEM, averaged
over a tidal cycle. The contention that the "PEM was
developed because the existing DEM model focused
more on spatial resolution than on the kinetic com-
plexities of eutrophication" Implies that spatial resolu-
tion is not of consequence in eutrophication and that
kinetic complexities could not be accommodated In a
modified DEM. This reviewer believes that spatial
resolution of the degree afforded by the DEM, as well
as  the hydrodynamic Information such a model
provides, are indeed desirable for a eutrophication
study such as exemplified  by this case. The more
detailed kinetics of PEM are, of course, appropriate.
However, experience has shown (and another of these
case  studies  illustrates) that the attributes of more
complex kinetics need not be at the expense of realistic
hydrodynamics.

Spatial resolution and temporal resolution may be dic-
tated  in part by the structure of the basic data used to
calibrate and verify  the model. The practice of  ag-
gregating data from several stations and  smoothing
over time seems in this case to be consistent with a
"regional and seasonal focus," but It tends to ignore
local and short term events which are often of major
concern in setting goals for wastewater management.
It also presents problems in calibration and verification,
as evidenced  in some of the examples given.

The post audit experience,  in which the model was
unable to predict the magnitude or spatial extent of the
1983  blue-green algae bloom, appears to confirm a
need  for Improved resolution and extension of  the
model. It is credit to the model developers that  the
model has  been periodically revised to improve its
capability as a management tool.

NELEUS - CHLORINE MODEL
The problem presented in modeling the fate of chlorine
in the Potomac Estuary is property 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 earned 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
                                               12-16

-------
chlorine, hence the models at this stage are of ques-
tionable use for decision purposes, despite their intrin-
sic potentials.

The important lesson of this case study is to provide an
adequate  data base  for complete  calibration and
verification of both hydrodynamic and water quality
models.

Case Study - Manasquan Estuary
This case, among 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 tod, so its perfor-
mance cannot be fully  assessed.

The  calibration-verification  sequence   of
hydrodynamics/conservative  tracer (salinity)/noncon-
servative water quality is representative of good model-
ing practice. Because  the model is one-dimensional
and  only a rough approximation of the estuary, it is
necessary to utilize an empirically derived dispersion
coefficient as a calibration parameter. While the func-
tional relationship between  this  parameter and
geometric and hydraulic properties of the estuary ap-
pears well founded in theory and experiment it is never-
theless unique for a particular estuarine system, e.g.
constant K and m. It purports to account for factors that
cannot be adequately represented in a one  dimen-
sional model with such a coarse segmentation, e.g.
advective dispersion and stratification. Dependence
on this uncertain calibration parameter could probably
be reduced by some additional detail in spatial charac-
terization of the estuary.

The  relatively unsatisfactory  results of water quality
calibration point to a need for improving the data base,
particularly the pattern of nutrient loading on  the es-
tuary. It  seems unlikely that the model will become a
useful tool for waste load allocation to the Manasquan
Estuary  until this additional data is developed.

Case Study - Calcasieu River Estuary
This study was allegedly  selected in part because It
represents  the application of a so-called "canned"
model supported by EPA. This reviewer disagrees with
the implication that such models, exemplified also by
such well documented and supported  models as
QUAL 2E,  SWMM, DEM, WASP 4, HEC 5Q, TABS II,
etc.  are likely  to lead  to the kind of difficulties en-
countered in modeling the Calcasieu  River Estuary. It
is the responsibility of the modeler to select the most
appropriate modeling approach for the 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.

12.3.8. 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
                                               12-17

-------
to the hydrodynamic driver as an integral part of the
modeling package. In particular, effects of stratification
should be explicitly modeled.

3. Complexity may lead to more uncertainty in model
results. Adding more compartments may improve fun-
damental understanding  of important mechanisms,
but It 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 model
should be developed and related to the Intended use
of the model in the decision process.
                                               12-18

-------