v/EPA
United States Office of Water (4305) EPA- 820-R-12-014
Environmental Protection August 2012
Agency
AQUATOX (RELEASE 3.1)
MODELING ENVIRONMENTAL FATE
AND ECOLOGICAL EFFECTS IN
AQUATIC ECOSYSTEMS
TECHNICAL NOTE 2: REQUIREMENTS,
SOURCES, AND CONDITIONING
OF DATA FOR AQUATOX
-------
AQUATOX Technical Note 2
Requirements, Sources, and Conditioning
of Data for AQUATOX
Prepared by
Richard A. Park1
and
Jonathan S. Clough2
Performed under
EPA Contract EP-C-06-029
Work Assignments No. 4-11,5-11
with
AQUA TERRA Consultants
and
EPA Contract EP-C-12-006
Work Assignment No. B-01
with
Horsley Witten Group
Submitted to
Marjorie Coombs Wellman
Office of Water/Office of Science and Technology
Standards and Health Protection Division
U.S. Environmental Protection Agency
Washington, DC 20460
August 2012
Modeling, Diamondhead MS 39525
2Warren Pinnacle Consulting, Inc., Warren VT 05674
-------
Contents
Acknowledgments 1
Disclaimer 1
Introduction 2
Site Characteristics 3
Length 3
Width 3
Depth 4
Thermocline Depth 6
Boundary Conditions 7
Inflow and Discharge 7
Linked-Segment Flow 10
Nutrient Loadings 11
Detrital Loadings 12
Total Suspended Solids 14
Dissolved Oxygen 15
pH 17
Light 17
Temperature 19
Organic Chemicals 20
Biotic Loadings 21
Initial Conditions 22
Parameters 23
Calibration and Validation Data 28
Nutrients 29
Dissolved Oxygen 31
Biomass 32
Density (Numbers of Individuals) 35
Percent Composition and Other Metrics 36
Organic Chemicals 38
Sensitivity Analysis 40
Summary 42
References 44
-------
Table of Figures
Figure 1. Channel width of Withlacoochee River FL measured using Google Earth 4
Figure 2. Comparison of calibrated and observed depths of water, Lower Boise River near
Parma, Idaho 5
Figure 3. Depths to Lake Onondaga thermocline estimated from the observed site length of
7.6 km, and using a time-series of observed depths and a constant depth imposed by
changing the length 6
Figure 4. Drainage pattern for Holmes Creek, FL; note tributary from the east below the site
being modeled and above the gage 8
Figure 5. Blue Earth MN watershed showing segments, USGS gage (BE-18), and
meteorological stations (Donigian et al. 2005). BE-54 is the site modeled with
AQUATOX 9
Figure 6. Discharge on the Little Withlacoochee River, FL; note effect of prolonged drought. ...10
Figure 7. Time-varying segment volumes in Tenkiller Lake, including Lacustrine A, B, and
C epilimnion and hypolimnion segments 11
Figure 8. Schematic showing segmentation and major linkages in the Lower Boise River ID
model (CH2M HILL etal. 2008) 13
Figure 9. Regression relationship between color and dissolved organic carbon, based on
Florida Department of Environmental Protection (DEP) data 14
Figure 10. Predicted and observed DO (Yl axis) and discharge (Y2 axis) in Reach 3, Lower
Boise River ID. Black error bars represent daily minimum and maximum oxygen
observations at Glen wood 16
Figure 11. Predicted rates for dissolved oxygen in Reach 3, Lower Boise River ID. Note that
Washin and Washout use the Y2 scale 16
Figure 12. Predicted pH at Upper Suwannee River FL site, and inverse relationship to
dissolved organic matter 17
Figure 13. Google Earth aerial photo of Santa Fe River FL site with accompanying
photograph, useful in estimating fraction of canopy (0.5) 19
Figure 14. Temperature in Blue Earth River MN based on HSPF predictions and annual
mean and range 20
Figure 15. Concentration of chlorpyrifos in the Ohio stream with pulsed loadings 21
Figure 16. Simulated algal biomass in Riverine segment of Tenkiller Lake OK. It is driven
almost entirely by loadings based on splitting of chlorophyll a data from the Illinois River
among four groups 22
Figure 17. Stocking hybrid striped bass in DeGray Lake 22
Figure 18. Typical species page in FishBase 26
Figure 19. ECOTOX Web site 27
Figure 20. Part of a record on chlorpyrifos in the ARS Pesticide Properties Database 28
Figure 21. Predicted and observed TSP and TP at Glenwood Bridge, Boise ID (Lower Boise
River). Note the predicted TSP and TP are superimposed most of the time, implying that
the model predicts that almost all P is inorganic 30
Figure 22. Predicted nitrate-nitrite and observed nitrate at Glenwood Bridge, Boise ID
(Lower Boise River); observed data are from two different sources 30
-------
Figure 23. Predicted and observed ammonia at Glenwood Bridge, Boise ID (Lower Boise
River). The symbol with a down arrow indicates a data point that is below detection limit...31
Figure 24. Observed and predicted dissolved oxygen in the hypolimnion of Lake Onondaga
NY 32
Figure 25. Predicted and observed periphytic chlorophyll a at Parma site, Lower Boise River
ID 33
Figure 26. Predicted and observed AFDW biomass of periphyton and moss, Cahaba River
AL 34
Figure 27. Regression of AFDW on diatom thickness rank, based on data of Stevenson et al.
(2007); R2 = 0.77 35
Figure 28. Predicted biomass of chironomids compared with observed numbers/sample (Y2
axis) in pond microcosm atDuluth MN dosed with 6.3 |ig/L chlorpyrifos 36
Figure 29. Predicted and observed percent blue-greens plotted with algal biomass, especially
diatoms 37
Figure 30. Predicted and observed percent chironomids in Upatoi Creek GA. The steep drops
are predicted due to a combination of emergence and, much more important, scour from
high flow 38
Figure 31. Predicted and observed gross primary productivity (GPP) in Sally Branch
Tributary #4, Fort Benning GA. Statistics (mean ± 1 standard deviation) are for the entire
three-year period, and therefore their position on the horizontal axis is arbitrary 38
Figure 32. Predicted concentrations of PCBs in Galveston Bay TX animals using exposure
data and observed concentrations from a Massachusetts bay 39
Figure 33. Predicted and observed BAFs for PCB 118 in Lake Ontario; complexed PCBs
were used in the computation of BAFs 40
Figure 34. Sensitivity of Phytoplanktonic Chlorophyll a in Lake Onondaga NY 41
Figure 3 5. Sensitivity of chlorophyll a to diatom PMax in Lake Onondaga NY 41
Figure 36. Original calibration of chlorophyll a with diatom PMax = 3.4 in Onondaga Lake
NY 42
Figure 37. Chlorophyll a result with diatom PMax = 1.6 in Onondaga Lake NY. Note better
fit in 1990 42
-------
Acknowledgments
We wish to thank Marjorie Coombs Wellman and Brenda Rashleigh of the US EPA and Mehmet
Umit Taner, presently at the University of Massachusetts Amherst, for their constructive
criticism and useful suggestions for additional material, which definitely improved this Note.
Work was performed under EPA Contract EP-C-06-029, Work Assignments No. 4-11 and 5-11
with AQUA TERRA Consultants; Tony Donigian's help is gratefully acknowledged. Some
examples were also taken from work done with AQUA TERRA under EPA Work Assignment 2-
11 (Tenkiller Lake) and SERDP Contract No. W912HQ-07-0026 (Fort Benning). This Note was
completed under Contract No. EP-C-12-006 to The Horsley Witten Group, Inc., Nigel Pickering,
Work Assignment Leader.
Disclaimer
This document describes ways of converting, conditioning, and using available data in the
AQUATOX model. Anticipated users of this document include persons who are interested in
using the model for various purposes, including but not limited to researchers and regulators.
The model described in this document is not required, and the document does not change any
legal requirements or impose legally binding requirements on EPA, states, tribes or the regulated
community. This document has been approved for publication by the Office of Science and
Technology, Office of Water, U.S. Environmental Protection Agency. Mention of trade names,
commercial products or organizations does not imply endorsement or recommendation for use.
-------
Introduction
The aquatic ecosystem model AQUATOX has been designed to be extremely flexible, allowing
many potential types of applications. Applications can span a wide range of time scale,
geographic scale, stressor type, ecosystem or water body type, as well as a myriad of research or
management questions that can be examined. AQUATOX is mechanistic, simulating the many
physical, chemical and biological processes, with the result that it is also data intensive.
Because of this flexibility, data requirements for setting up an AQUATOX simulation depend
considerably on the site type that is being modeled and the goal of the modeling study. For
example, precise accounting of time-series loadings is especially important in rivers and streams;
due to low retention time, inflow loadings can be the dominant factor affecting water column
conditions. For standing water the importance of daily loadings follows a continuum from low-
retention, run-of-the-river reservoirs, where daily loadings are important, to long-retention lakes
and reservoirs that can be driven with seasonal data.
AQUATOX can accommodate a variety of modeling goals. A modeling study to assess the
effects of changing nutrient levels in a particular water body (such as a site specific nutrient
criteria study or a total maximum daily load study) would benefit from an excellent accounting
of nutrients in the water column and nutrient boundary conditions, including loadings and initial
concentrations in bottom sediments. On the other hand, a study that is primarily to evaluate fate
and effects of new chemicals on a national or regional level may not require such detailed
nutrient information. In that case, a user may wish to adjust the nutrient loadings in a
representative or "canonical" environment to obtain a stable food web and then apply detailed
information about chemical loadings in the simulation to assess fate and effects. Most major
projects involving changes in land use require an environmental assessment, which could benefit
from linking a hydrologic or watershed model and a receiving water model, such as AQUATOX,
with the boundary conditions being defined by the hydrologic model.
Identification of environmental stressors that confound an analysis may involve a combination of
well-defined drivers and calibration data that span several trophic levels. An emerging use of
AQUATOX is in forecasting the effects of climate change, which is facilitated by setting up the
model for probabilistic loadings, given the uncertainty of future conditions. Perhaps most
demanding is modeling the time-to and extent-of recovery of ecosystems from pollution, which
can entail accurate initial conditions, complex sediment-water relations, and repeating time series
of environmental drivers for simulations that may represent decades.
The specificity and robustness of model formulations and associated parameter values should
also be considered. It has been stated that models cannot be realistic, general, and precise at the
same time (Levins 1966); one must choose two out of the three characteristics. While precision is
inherently desirable, for many applications it is more important to have generality and realism
(Park and Collins 1982). Precision is usually achieved by careful calibration to an extensive set
of observed data; however, without generality there is no assurance that the calibration domain
won't be exceeded under changing conditions—and most models are expected to forecast
behavior under changing conditions. Depending on the goal, the tradeoff between precision and
generality should be carefully considered.
-------
Seldom do analysts have the luxury of being able to dictate what data should be collected. Often
the model application comes after the original study and must rely on incomplete sets of data
collected for other purposes. A prime objective of this Note is to describe ways of converting,
conditioning, and using available data and of filling in the gaps with other sources. There is a
logical progression in the use of data for any implementation of the model, and this progression
is described in detail in the sections that follow. Briefly, the site is characterized by
morphometric data, and the loadings at the site boundaries are specified, usually as time series.
Initial conditions are given for each of the state variables, and the differential equations are
populated with process-level parameter values. The parameter values are then calibrated—
adjusted within reasonable limits to obtain simulation output that fits available site data, which
may or may not need to be conditioned for purposes of comparison. Finally, sensitivity analyses
may be performed to identify key parameters that might be adjusted further to improve the fit;
and the model may be validated with data from another site or time period. To illustrate this
process, in this Note we cite numerous examples in which data were obtained and conditioned
for a variety of studies, in lieu of data collected for the specific applications.
Site Characteristics
Site characteristics and loadings are numerous, and setting them up for an application can be
time-consuming. It is recommended that, whenever possible, existing studies be used as
templates for new studies; one advantage is that the inputs can be viewed as a checklist and if
site data are missing the data in the existing study can be used as defaults. Obviously, one should
be careful to choose a comparable study and to note how the new study might differ from that
"template." A brief description of each of the example studies is distributed with the AQUATOX
software. As new studies are developed, they will be added to the set of example studies
available to users.
Length
The model uses site length, which is measured in km, in several ways. Depending on the site
type, it can be well defined or arbitrary. If modeling linked segments, it is the length of an
individual segment. If modeling a lake or reservoir, it is the fetch or distance across which the
wind blows and is used to calculate the depth of the thermocline, unless that estimated depth is
defined by the user. If modeling a stream reach that is not linked, it is taken as an arbitrary
length. In this case, the reach can be considered to be represented by a point model where the
longer the length the greater the retention time. Very short reaches can result in stiff equations,
where the step size has to be decreased to simulate the fast through-flow. This can be avoided by
arbitrarily increasing the reach length, resulting in a longer retention time and a considerably
faster simulation. We quite often use an arbitrary length of 2 km.
Width
A site characteristic with low sensitivity is channel width, measured in m. It is used in
computing dynamic depth (see below). In the absence of data it may be estimated by zooming in
on an aerial photograph, for example with Google Earth, and using the measuring tool on a
representative traverse (Figure 1).
-------
Figure 1. Channel width of Withlacoochee River FL measured using Google Earth.
Depth
Depth, measured in m, is input as "mean depth" and "maximum depth." The distinction between
the two is important for lakes, reservoirs, and estuaries because they are used to define general
bathymetric relationships and calculate entities such as the area of the euphotic zone. In general,
these two parameters are important for computing the light climate at depth. Shallow streams and
ponds are not very sensitive to the difference in these two variables, and we often arbitrarily take
the mean depth to be one-half the maximum depth.
Time-varying water depth in streams is a function of the flow rate, channel roughness, slope, and
channel width using Manning's equation, which is rearranged to yield (U.S. Environmental
Protection Agency 2012):
where:
7 =
Q =
Manning =
Slope =
dynamic mean depth (m),
flow rate (m3/s);
Manning' s roughness coefficient (s/m1/3);
slope of channel (m/m); and
-------
Width = channel width (m).
The Manning's roughness coefficient represents frictional loss; it is not subject to direct
measurement but can be used as a calibration parameter to estimate dynamic depth using
observed depth and flow rate (Figure 2). In an example taken from the Parma site on the Lower
Boise River, Idaho, time-varying depths were estimated using both default and calibrated
roughness coefficients and were compared with observed depths. Although visual inspection
indicates that the calibrated depth trend is a better fit, the default trend appears to be acceptable.
The default value for the roughness coefficient for natural streams is 0.040; however, you may
wish to choose a different value from Table 1 based on site characteristics. The calibrated value
in this example is 0.055. A spreadsheet entitled Depth_estimated_from_flow.xls is supplied
with the example studies as a part of the Release 3.1 distribution.
Calibration of Depth as Function of Flow and Roughness
Depth (m)
Obs Depth (m) Default Depth (m)
Figure 2. Comparison of calibrated and observed depths of water, Lower Boise River near
Parma, Idaho.
Table 1. Values of Manning's roughness coefficient (s/m13). Based on (Cowan 1956).
Type of Channel
Minor alluvial stream
Clean, straight
Same, but more stones or weeds
Clean, winding, some pools & shoals
Same, but some weeds or stones
Same, but more stones
Sluggish reaches, weedy, deep pools
Very weedy reaches, deep pools
Minor mountain stream
Bottom: gravel, cobbles, and few boulders
Bottom: cobbles with large boulders
Minimum
0.025
0.030
0.033
0.035
0.045
0.050
0.075
0.030
0.040
Normal
0.030
0.035
0.040
0.045
0.050
0.070
0.100
0.040
0.050
Maximum
0.033
0.040
0.045
0.050
0.060
0.080
0.150
0.050
0.070
-------
Major river
Regular section with no boulders or brush
Irregular and rough section
0.025
0.035
n/a
n/a
0.060
0.100
Thermocline Depth
The depth to the thermocline in lakes and reservoirs is normally estimated by AQUATOX using
by a regression relationship based on length (U.S. Environmental Protection Agency 2012):
\og(MaxZMix) = 0.336- \og(Length) - 0.245
where:
MaxZMix = maximum mixing depth under stratified conditions (thermocline depth) for
lake (m); and
Length = maximum effective length for wave setup (m, converted from user-
supplied km).
AQUATOX Release 3 also accepts a constant or time series of user-specified thermocline
depths; this allows one to specify varying depths based on interpretations of available
temperature and dissolved-oxygen isopleths. Or one can force a particular thermocline depth by
varying the length of the lake or reservoir. This was done in an earlier implementation of the
model for Lake Onondaga NY, in which higher salinity water limits the mixing depth; in that
case changing the length from 7.6 km (the true length) to 2.376 km decreased the depth
appreciably to that observed for part of the stratified period. The results of all three options are
shown in Figure 3.
Depth (m)
-6
-7
-8
-9
'i n
1 1
Alternate Depths to Thermocline
J v
^N^
\
V
4/2/1989 5/22/1989 7/11/1989 8/30/1989 10/19/1989
T-series Length = 7.6 km —Length = 2.376km
Figure 3. Depths to Lake Onondaga thermocline estimated from the observed site length of 7.6
km, and using a time-series of observed depths and a constant depth imposed by changing the
length.
-------
Boundary Conditions
Inflow and Discharge
Probably the most important environmental loading is water flow because it is the carrier for
nutrients, suspended sediments, plankton, and organic pollutants. Depending on how it is used
with respect to volume, it may be input as inflow, outflow, or both. For any site other than a
long-retention lake, it should be input as daily values if possible; that is especially important with
respect to streams. In the U.S. observed flow data are often taken from a USGS gage, with the
values in cubic feet per second converted to cubic meters per day. If the gage is not located at
the site then the flow should be adjusted for the difference in the drainage area. For example, in
a Florida study the USGS gage on Holmes Creek is 10 km downstream from the site being
modeled. Based on the drainage, about one-third of the gaged watershed is below the site (Figure
4); therefore, the flow loading was reduced to two-thirds of that observed at the gage. Of course,
where possible it is better to estimate the flow using a hydrologic model such as HSPF calibrated
to the watershed (including both flow and precipitation data), rather than using an approximation
of the type just described. As an example, in a Minnesota study HSPF was used to obtain
estimated flow at a site at river mile 54 on the Blue Earth River, 36 miles above the USGS gage
(Figure 5).
-------
Holmes Cr
calibration
site
Figure 4. Drainage pattern for Holmes Creek, FL; note tributary from the east below the site
being modeled and above the gage.
-------
-
Watonwan River
Le Sueur River
Watershed
egend
Hydrography
A Nutrient/Biological Sites
• Met Stations
O Hourly Precipitation
• USGS nr Rapidan 05320000
Le Sueur River
Watonwan River
Blue Earth
670 Segment Number
Figure 5. Blue Earth MN watershed showing segments, USGS gage (BE-18), and meteorological
stations (Donigian et al. 2005). BE-54 is the site modeled with AQUATOX.
There is a spectrum of reliability for flow data: best is data from an adjacent gage, next best is
calibrated flow estimate from a hydrologic model, followed by data adjusted for distance to gage,
and finally data on a "comparable" watershed. The latter approach can be useful, for example, if
the site of interest has a very small watershed with response times that cannot be represented
well by taking a downstream gage and scaling back the flow. A small watershed will be "flashy"
in comparison to a larger watershed, so picking a gage on a nearby small watershed may be the
best approach.
Flow in a linked-segment reservoir is a special case. The model can accommodate main stem
and tributary flow into the segments, flow from one segment to another, overland flow into
segments (best computed by a hydrologic model), and withdrawals from the reservoir. Flow data
for the linked-segment version will be discussed in further detail below.
Also of concern is the time period being simulated. Preferably a simulation can include a wet
year, a dry year, and a normal hydrologic year. The Little Withlacoochee River in Florida is an
extreme case, where the river dried up during a couple of droughts (Figure 6). AQUATOX is
able to step through dry periods with the implicit assumptions that the time period is short and
that there are refuges so that the state variables do not change during that period. A word of
caution: if the time series shown in Figure 6 were extended to a longer period of time,
AQUATOX would "wrap around" the last year of loadings so that the drought would be
repeated. The best way to avoid this, given the example, would be to obtain a longer hydrologic
9
-------
record; otherwise, one could generate a synthetic time series offline based on a typical year of
data.
1.80E+06
1.60E+06
1.40E+06
91 1.20E+06
| l.OOE+06
~ 8.00E+05
iZ 6.00E+05
4.00E+05
2.00E+05
O.OOE+00
*%
\w
Figure 6. Discharge on the Little Withlacoochee River, FL; note effect of prolonged drought.
Velocity is usually computed by the model based on flow and cross-sectional area, but it can be
entered by the user in the site loading screen. The model can be sensitive to velocity, which
affects:
• washout and deposition of phytoplankton and scour of periphyton;
• breakage of macrophytes;
• entrainment of zooplankton;
• deposition or scour of organic matter;
• scour and deposition in the sand-silt-clay model; and
• oxygen reaeration.
Linked-Segment Flow
Flow among segments in a lake, reservoir, or estuary is best obtained from a hydrodynamic
model such as EFDC (Environmental Fluid Dynamic Code) (Tetra Tech Inc. 2002). In an
application to Tenkiller Lake OK, EFDC was used to obtain flow rates between horizontal
segments of the reservoir, and exchanges between vertical segments (i.e., epilimnion and
hypolimnion) were calculated from observed temperature differences. These flows are reflected
in the time-varying volumes of the reservoir segments (Figure 7). Note that the riverine segment
is run-of-the-river with low volume and retention time.
10
-------
Tenkiller Lake OK Segment Volumes
o
300,000,000
250,000.000
200,000.000
150,000,000
100,000,000
50,000,000
Sep-91 Dec-91 Apr-92 Jul-92 Oct-92 Jan-93 May-93 Aug-93 Nov-93 Mar-94
Figure 7. Time-varying segment volumes in Tenkiller Lake, including Lacustrine A, B, and C
epilimnion and hypolimnion segments.
For a river, flow among segments can be obtained from a model such as EFDC or, better yet,
from a hydrologic model, such as HSPF (Hydrologic Simulation Program-Fortran) (Bicknell et
al. 1991) that accounts for time-varying differences in runoff from adjacent watersheds. In an
application to the Lower Boise River ID, which is an intensely managed river with diversion
canals and return drains for agricultural irrigation, a flow-routing spreadsheet model was
developed to keep track of the complexities.
Nutrient Loadings
Nutrients are important drivers of primary productivity and therefore the basis of the aquatic
food web. In fact, AQUATOX requires nutrient loadings in order to run. Nutrient loadings can
be input in several ways:
• Concentrations (mg/L) in inflow waters (the most common loading)
• Non-point source loadings (g/d)
• Point-source loadings (g/d)
• Direct precipitation, which includes dry fall (g/m2 d)
• Concentrations (mg/L) in linked tributaries (linked segment version)
Nutrient loadings can be derived from observed data collected at the site over time. Ideally these
would be daily values; however, in practice nutrient data are often weekly, monthly, or sporadic
11
-------
observations; in temperate zones nutrient data in the winter are rare. In a project on multiple
streams in Florida quarterly averages were used for many sites, which added a degree of
uncertainty and decreased the temporal resolution of the ecosystem responses. AQUATOX
calculates nutrient loadings from streams by interpolating the input concentrations and
multiplying by the daily discharge. A more flexible and rigorous method is to use the FLUX
program (Walker 1999), available from the U.S. Army Corps of Engineers Environmental
Laboratory, to calculate the loading based on any one of six methods. Developed for calculating
loadings for reservoirs, the procedure can also be used for streams and lakes. Daily estimates of
nutrient loadings also can be obtained from a linked hydrologic model such as HSPF in much the
same way that water flows can be obtained. In fact, HSPF was used to drive both the Blue Earth
River MN (Donigian et al. 2005) and the Tenkiller Lake OK (Donigian et al. 2009)
implementations. These linkages have the advantage of facilitating "what if scenarios where
the potential effects of land-use changes can be examined (Carleton et al. 2009, Donigian et al.
2009). The effects of climate changes could also be modeled using these linked models.
Linked segments present almost endless possibilities for disaggregating and loading nutrients
from diverse sources. In the Lower Boise River ID study, the river was divided into 13
segments, including channels on opposite sides of Eagle Island (CH2M HILL et al. 2008).
Nutrient loadings came from wastewater treatment plants, groundwater, fish hatcheries,
tributaries, and drains (return flow from irrigation) (Figure 8). Each of these sources can be
subject to perturbations and analysis of impacts in subsequent simulations.
Detrital Loadings
One of the potentially important loads for nutrient and eutrophication analysis, but one which is
often missing or in a different form than that used in AQUATOX, is organic matter, or detritus.
As stated in Section 5.1 of the Technical Documentation (U.S. Environmental Protection Agency
2012): "For the purposes of AQUATOX, the term 'detritus' is used to include all non-living
organic material and associated decomposers (bacteria and fungi).... Detritus is modeled as
eight compartments: refractory (resistant) dissolved, suspended, sedimented, and buried detritus;
and labile (readily decomposed) dissolved, suspended, sedimented, and buried detritus. This
degree of disaggregation is considered necessary to provide more realistic simulations of the
detrital food web; the bioavailability of toxicants, with orders-of-magnitude differences in
partitioning; and biochemical oxygen demand, which depends largely on the decomposition
rates.... In general, dissolved organic material is about ten times that of suspended particulate
matter in lakes and streams (Saunders 1980), and refractory compounds usually predominate;
however, the proportions are modeled dynamically."
The effort necessary to obtain these loadings depends on how important the detrital components
are in a particular study. Oftentimes a general value of 10% Particulate is used (see above
paragraph); % Refractory is more dependent upon the site. An extreme case where detrital
dynamics are important is the Upper Suwannee River FL—a blackwater river that drains the
Okefenokee Swamp. Three different measures of organic matter were available for the river, and
those were used to populate time-series loadings (Table 2).
12
-------
o
Model Segment
Number
R.M. = River Mile
Diversion Dam
R.M.61.2
Eckert Road
R.M. 58-2
Veterans Bridge
R.M. 50.1
Glenwood Bridge
R.M. 47,5 —
Head of Eagle
Island, R.M, 45.8 —
i —
Parma
R.M. 3.5
Segmentation of Lower
Boise River for Aquatox
Model
Figure 8. Schematic showing segmentation and major linkages in the Lower Boise River ID
model (CH2M HILL et al. 2008).
Table 2. Observed and estimated detrital characteristics, Upper Suwannee River FL.
Date
6/4/1990
COLOR*
Ptu
335
TOC*
mg/L
50
BOD 5*
mg/L
1.55
Est.
DOC
(mg/L)
37.70
Estimated
%Particulate
25
* Ob served
13
-------
The observed Total Organic Carbon (TOC) was used as the primary loading to the Upper
Suwannee River, where it was converted to organic matter by the model using a conversion
factor of 1.9 grams of organic matter per gram of organic carbon (Winberg 1971). Color was
used to estimate Dissolved Organic Carbon (DOC) using a weak regression relationship based on
data from numerous Florida sites (Figure 9). In turn, comparing observed TOC to estimated
DOC, the % Particulate can be calculated. Of course, each of these conversions should be
checked for consistency; for example, the estimated DOC might exceed observed TOC on a
particular date, in which case it would need to be constrained.
80
70
60
50
40
O 30
Q
20
10
0
-10
-20
Scatter Plot with Fit
• Linear fit (3.403 +0.1023x)
• 90% Cl
90% Prediction interval
100
200 300
COLOR
400
500
Figure 9. Regression relationship between color and dissolved organic carbon, based on Florida
Department of Environmental Protection (DEP) data.
The same guidelines apply to detrital loadings as to nutrients; that is, where possible either
observed or simulated daily loadings should be used for low-retention sites. For example, HSPF
was used to obtain estimated BOD loadings for Tenkiller Lake based on land use in the
watershed (Donigian et al. 2009). There were no calibration data for BOD, so the loadings were
used with appreciable uncertainty; however, data collected later showed that the maximum value
simulated by HSPF was remarkably close to the observed maximum for later dates.
Total Suspended Solids
TSS can be an important loading, especially as an alternative to modeling sand, silt, and clay.
Daily values are highly desirable for streams and reservoirs, where sporadic light limitation and
sedimentation can affect algae and zoobenthos. Unfortunately, TSS is often not measured or is
measured only occasionally. Two Minnesota streams that differ greatly in TSS exemplify two
14
-------
approaches to representing this loading, as described in AQUATOX Technical Note 1 (Park et al.
2009). The Blue Earth River demonstrated a close relationship between TSS and flow, and that
relationship was used to obtain daily loadings (Figure 2 in (Park et al. 2009). On the other hand,
data from the Crow Wing River, which is located on well-sorted glacial outwash sands, showed
no significant variation in TSS values with differing flow rates. In this case, TSS was based on
an HSPF calibration that incorporated more complex controls in the watershed such as upstream
lakes (see Figure 4 in (Park et al. 2009).
Dissolved Oxygen
Sufficient dissolved oxygen is, of course, vital for the survival of aquatic animals; however,
accurate simulation is important only when concentrations drop below about 3 mg/L. Oxygen
dynamics are more or less sensitive to loads depending upon the type of water body. Dissolved
oxygen loadings in a stream present a challenge. Because of low retention times, loadings may
completely dominate concentrations in a particular reach, reflecting upstream conditions rather
than processes occurring in the reach; both sources affect observed data and are usually
simulated. All loadings can be turned off in the oxygen loading screen in order to isolate in-reach
processes, providing a valuable perspective on DO concentrations. However, this is an analytical
technique and can distort the results if the oxygen loads are left "turned off.
As an example of how the influence of dissolved oxygen loads can vary, Reach 3 of the Lower
Boise River ID is 13 miles downstream from the Diversion Dam (Figure 8), which is the source
of daily loadings for the model. There are very few periphyton under conditions of high flow,
and water from the dam passes through Reach 3 quickly, accounting for most of the DO; the
daily loadings correspond closely to the mean observed in Reach 3 (Figure 10). However, with
low-flow conditions the retention time above Reach 3 increases considerably and periphyton are
able to grow; therefore, the predicted DO no longer corresponds entirely to the loadings at the
dam, and photosynthesis increases the range of predicted diel DO (Figure 11).
15
-------
Seg 3 (CONTROL)
Run on 08-15-07 2:08 PM
2,090,000.0
1,900,000.0
1,710,000.0
1,520,000.0
1,330,000.0 n
1,140,000.0-g.
950,000.0
760,000.0
570,000.0
380,000.0
Oxygen (mg/L)
Min. Oxygen (mg/L)
Max. Oxygen (mg/L)
• Obs DO at Glenwood (mg/L)
o DOatGB, City of Boise (mg/L)
• Obs DO Glenwood (mg/L)
DischH20 (cu.m/d)
9/15/2005
9/30/2005
10/15/2005
Figure 10. Predicted and observed DO (Yl axis) and discharge (Y2 axis) in Reach 3, Lower
Boise River ID. Black error bars represent daily minimum and maximum oxygen observations
at Glenwood.
Seg 3 (CONTROL)
Run on 08-15-07 2:08 PM
112
1540.0
1400.0
1260.0
1120.0
X.
- Oxygen Load (Percent)
•Oxygen Photosyn (Percent)
• Oxygen Reaer (Percent)
• Oxygen BOD (Percent)
Oxygen SOD (Percent)
- Oxygen Respiration (Percent)
• Oxygen Nitrific (Percent)
•Oxygen Washout (Percent)
Oxygen FB.Washln (Percent)
-16
-32
420.0
280.0
9/15/2005
9/30/2005
10/15/2005
Figure 11. Predicted rates for dissolved oxygen in Reach 3, Lower Boise River ID. Note that
Washin and Washout use the Y2 scale.
16
-------
On the other hand, DO loadings from tributary streams may be an important source of oxygen in
lakes and reservoirs during the winter when ice cover may impede photosynthesis and preclude
reaeration. Otherwise, DO loadings are seldom critical for a lacustrine epilimnion. The
hypolimnion may receive much-needed oxygen from inflowing water, but that may be offset by
the BOD from concurrent detrital loadings.
For smaller streams, hourly oxygen loadings may be input in conjunction with an hourly
AQUATOX simulation. This can vastly improve the simulation of oxygen fluctuations within
each day; however, hourly oxygen data are rarely available.
pH
pH is important in AQUATOX because it affects precipitation of calcium carbonate and sorbed
phosphorus and it can control hydrolysis of organic toxicants and the fraction of total ammonia
that is in its more toxic form. AQUATOX can predict time-varying pH, but the procedure
requires good estimates of alkalinity and dissolved organic matter. Perhaps the best example of
computed pH is the simulation of the Upper Suwannee River FL. Alkalinity was held constant at
220 ueq CaCOs (an input variable on the pH screen), but the principal factor contributing to
time-varying pH was the variation in dissolved refractory detritus (mainly humic acids). The
algorithm constrains the minimum pH to 3.75, which was achieved with higher concentrations of
dissolved organic matter (Figure 12). However, it is often better to use observed values of pH,
and that is our practice with most studies; it is commonly available.
Upper Suwannee River FL (PERTURBED)
Run on 12-21-09 8:19 AM
-pH(pH)
Obs pH(pH)
- Rdetr diss (mg/L dry)
12/4/2005 12/4/2006 12/4/2007 12/3/2008
Figure 12. Predicted pH at Upper Suwannee River FL site, and inverse relationship to dissolved
organic matter.
Light
Seldom are light observations available for a site. However, broad seasonal patterns of solar
radiation will usually suffice; with correction for riparian shading if necessary (see below).
AQUATOX computes time-varying light given the annual mean and range.
17
-------
A very useful source is a NASA Web page, which can be used to estimate solar radiation and
other climate characteristics worldwide. The page changes frequently, so search on "NASA,
Surface meteorology and Solar Energy." An example that uses NASA data is given for the
Upper Suwannee River site (Table 3). From the table one can obtain the annual mean, and by
scanning for the annual maximum and minimum values the range can be calculated. The values
are in kWh/m2 d, which AQUATOX will convert to Ly/d when entered in the site screen. It is
not uncommon to find values in other data sets expressed as |iEin/m2 s, which the model will
also convert.
Table 3 Monthly averaged insolation incident on a horizontal surface (kWh/m2/d) (NASA 2010).
Year
2U04
2nn5
2UU4 - 20ii5
Jan
321
3 in
3.16
Feb
2 S3
342
3.12
Mar
535
444
4 &
Apr
529
-U7
f 13
May
6 53
628
643
Jun
5.56
5 16
536
Jul
5.97
5C4
5.91
Aug
5U7
5.21
5.14
5ep
4.17
554
4 y.
'Jet
4.15
4 115
4. lu
Niv
3 36
3 65
351
Dec
z 98
2 C5
[_ 92
Annual
4 64
4 63
463
In narrow streams with dense riparian vegetation, light can be very limiting during the growing
season. The user can also specify shading as the fractional coverage by riparian canopy, either as
a constant or as a time series. In the absence of data, sources of aerial photographs such as
Google Earth provide a means to obtain an approximate canopy value by zooming in on a site,
sometimes with accompanying photographs (Figure 13).
18
-------
http: fjwww. panoramic^, com/photo/21491285
Pan©ramio
Sign up Upload Places Tags
(3 World Map > USA FL Santa Fe
Untitled
Google Earth Share on;
by SCBerry
This photo is selected for Google Earth [?] - ID: 21491285
Figure 13. Google Earth aerial photo of Santa Fe River FL site with accompanying photograph,
useful in estimating fraction of canopy (0.5).
Temperature
Temperature is an important driving variable, affecting many physical and all chemical and
biologic processes; and this is confirmed by sensitivity analysis. However, it seldom changes
very rapidly in aquatic systems, so that a sinusoidal seasonal approximation based on user-
specified annual mean and range is usually sufficient. Alternatively, time series of observed
values can be specified. Hydrodynamic-thermal stratification models can be used to compute
vertical temperature profiles based on climate and hydrological drivers. If the site is a stream,
loadings can be obtained from a hydrologic model such as HSPF. Figure 14 illustrates the
difference between simulated temperature loadings obtained from HSPF and those calculated
using the annual mean and range. The simulated temperature would appear to be preferable, but
the sinusoidal curve based on minimal data seems to be adequate.
19
-------
11/1/1998 2/9/1999 5/20/1999 8/28/1999 12/6/1999 3/15/2000
HSPFT Sinusoidal!
Figure 14. Temperature in Blue Earth River MN based on HSPF predictions and annual mean
and range.
Organic Chemicals
There are numerous applications of AQUATOX involving potentially toxic organic chemicals,
with varying levels of precision required for the chemical loadings. Simulating time to recovery
from historic pollution often involves painstaking specification of initial conditions including
concentrations (jig/kg) in buried sediment layers and associated pore waters (|ig/L). Analysis of
whether a chemical has a high potential of being of concern may just require an observed
concentration used as a constant. This was done with Blue Earth River MN, using a single
reported atrazine concentration, to make sure that the herbicide was not affecting the algal
calibrations. A common application is environmental risk analysis of pesticides and other
toxicants, often using a representative site (sometimes called a "canonical environment"). Three
types of analyses can be performed by AQUATOX on the same study with minimal effort:
• Pulsed loadings of toxicant corresponding to runoff during storm events, provided by the
analyst or predicted by a model such as PRZM (Suarez 2005) or SWAT (Arnold et al.
1998)
• Initial toxicant concentration with no loadings (uncheck "Use Dynamic Loadings" and
ensure that constant loadings are set to zero)
• Initial concentration held constant
All three can be performed with the Ohio Stream Chlorpyrifos studies that are provided as
examples in the AQUATOX installation. In the "Pulsed" study the initial condition is set to the
maximum observed (Figure 15), thus when the dynamic loadings are turned off there will be the
single initial concentration of chlorpyrifos. Setting the simulation to "keep toxicant constant"
will result in a constant concentration of 0.4 |ig/L, unless the initial condition is changed.
20
-------
Ohio Creek (PERTURBED)
Run on 06-4-10 10:23 AM
0.40
_i
D
000
1/30/1997 3/31/1997 5/30/1997 7/29/1997 9/27/1997 11/26/1997
T1 H2O(ug/L)|
Figure 15. Concentration of chlorpyrifos in the Ohio stream with pulsed loadings.
Biotic Loadings
Usually biotic loadings are set to an arbitrarily small number to serve as a "seed" to simulate
recolonization following extinction, and therefore allow the species to establish itself under
improved environmental conditions. Such a "seed" is appropriate for algae and aquatic insects.
The analyst should consider whether recolonization can occur for other invertebrates and fish;
for example, if fish are driven to extinction in a pond they may not repopulate the site. On other
hand, biotic loadings may be quite important for a reservoir, where upstream productivity in an
unmodeled tributary can provide a seasonal or continuous supply of organisms to the reservoir.
For example, the Illinois River transports significant biomass of sestonic algae (measured as
chlorophyll d) into the riverine segment of Tenkiller Lake OK. AQUATOX models individual
algal groups, and combines them for an estimate of chlorophyll; but frequently algal species data
are not available, as was the case in the Tenkiller. The chlorophyll value was converted to
biomass (mg/L) and split evenly among four phytoplankton groups (Donigian et al. 2009), as
seen in Figure 16. Stocking can also be an important factor in lakes and reservoirs managed for
fishing. In a study of DeGray Lake AR stocking records (in numbers offish per acre) were
available for hybrid striped bass (Fourt et al. 2002), and those were converted and used in
modeling that reservoir (Figure 17).
21
-------
Riverine (CONTROL)
Run on 07-8-09 8:32 AM
- Phyto, Diatom wrm (mg/L dry)
- Phyt Low-Nut Diatom (mg/L dry)
- Phyto Greens (mg/L dry)
- Phyt, Blue-Green max (mg/L dry)
Cryptomonad (mg/L dry)
Figure 16. Simulated algal biomass in Riverine segment of Tenkiller Lake OK. It is driven
almost entirely by loadings based on splitting of chlorophyll a data from the Illinois River among
four groups.
SmGameFish2: [Hybrid StrBass Juv]
Fish Stocking in grams per day
(S Use Const, Loading of [<)g/d
Use Dynamic Loadings
Initial Condition:
0 g'm2 dry
Loadings from Inflow:
• use Constant Loading of
g>'m2 dry
Use Dynamic Loadings
Multiply loading by 1
Fish Stocking in grams per sq.m
Notes: Sfingerlings per acre stocked each year starting
1975 6'73.5.'«M«.9 = 1.2, say|0.2 g/'m2-yr
Multiply loading by 1
Edit Underlying Data
-/ O.K. X Cancel
Figure 17. Stocking hybrid striped bass in DeGray Lake.
Initial Conditions
AQUATOX requires initial conditions for all state variables, but depending upon the site type,
they may not be important. In particular, initial conditions for nutrients, suspended detritus, and
phytoplankton are quickly replaced by loadings from upstream in a river. Initial conditions are
22
-------
often uncertain; or, if good values are available, they are from the growing season. Of course,
you can begin a simulation on the date with the best data, but seldom are data available for all
state variables. We prefer to begin a simulation in winter (for simulations of temperate
ecosystems) and allow the system to "spin up" as it goes into the growing season. Furthermore,
unless it represents a short-lived mesocosm, a simulation should run for at least a year, and
preferably several years. That way the initial conditions may be uncertain, but stable fluctuations
can be attained and compared to available data. To that end, one can specify "Run model in
Spin-up Mode" and the initial conditions for the biotic state variables will be set using the end
values; the same may be done with nutrient and detrital state variables. Once the model is
calibrated, the spin-up mode should be turned off; in fact, a warning is issued at the beginning of
each run if spin-up mode is being used.
Parameters
Unlike driving variables, which are characteristics of a site, model parameters are intrinsic
characteristics of the species or chemical of interest. Obtaining biotic, chemical, and toxicity
parameters can be difficult. The AQUATOX libraries of parameters cover a wide range of
species and chemicals, and should be consulted first. However, one may wish to try different
values as part of the calibration process, or one may wish to add state variables that are not in the
library. There are three compendiums of parameters that were assembled for the US Corps of
Engineers about 30 years ago. We have found them so useful that we have made them available
in pdf format on the AQUATOX Web
site: http://water.epa.gov/scitech/datait/models/aquatox/data.cfm.
The report by (Collins and Wlosinski 1983) covers phytoplankton, zooplankton, zoobenthos, and
fish. Data are in tabular form (for example, Table 4), and the user can obtain representative
values and statistics to define distributions for uncertainty analysis.
23
-------
Table 4. Part of a table giving observed values for maximum photosynthesis and temperature at
time of the observation (Collins and Wlosinski 1983).
Table 5
Gross production rates of phytoplankton (I/day)
SPECIES
DIATOMS
Asterionella formosa
Asterionella formosa
Asterionella formosa
Asterionella formosa
Asterionella formosa
Asterionella formosa
Asterionella formosa
Asterionella formosa
Asterionella formosa
Asterionella formosa
Asterionella japonica
Asterionella japonica
Asterionella japonica
Biddulphia sp.
Coscinodiscus sp.
Cyclotella meneghiniana
Cyclotella nana
Detonula confervacea
TPMAX
0.81
0.69
1.38
1.66
1.71
0.28
0.69
1.38
2.2
1.9
1.19
1.3
1.7
1.5
0.55
0.34
3.4
0.62
TEMP °C
20
10
20
25
20
4
10
20
20
18.5
22
18
25
11
18
16
20
2
REFERENCE
Holm and Armstrong 1981
Hutchinson 1957
Hutchinson 1957
Hutchinson 1957
Fogg 1969
Tailing 1955
Tailing 1955
Tailing 1955
Hoogenhout and Amesz
Hoogenhout and Amesz
Fogg 1969
Hoogenhout and Areesz
Hoogenhout and Amesz
Castenholz 1964
Fogg 1969
Hoogenhout and Amesz
Hoogenhout and Amesz
Smavda 1 969
1965
1965
1965
1965
1965
1965
Useful parameter tables for zooplankton and fish were developed under contract to the Corps of
Engineers (Leidy and Jenkins 1977, Leidy and Ploskey 1980). Much of the data are summarized
by Collins and Wlosinski (1983), but there are lots of tables on fish in particular that are very
useful. For example, maximum biomass by fish species is listed for many reservoirs (Table 5),
providing both a parameter (carrying capacity) and a reality check on AQUATOX simulations.
This particular set of data (which can be converted by g/m = pounds/acre * 0.028) is based on
recovery offish from coves or enclosed open-water areas following application of rotenone.
24
-------
Table 5. Part of a table giving observed values for standing crop offish; data are pounds/acre
(Leidy and Jenkins 1977).
Fitth Carrying Capacity
Species or Sp«ct*«
Carry tog
Groop Capacity
Appendix C
Arranged by Species and Major Reservoir Group*
Carrying Capacity iiemmss In Pounds per Acre
Supported by Each Food tomfaitmettt
Dettittt* lent has Z*»epi*aki<»
Fish Terrestrial
Gulf and South Atlantic Drainage Area
Gars
Bowf in
Gizzttd shad
Thread! in sttmA
Pickareis
Carp
Minnows
Carpsuckers
Suckers
Hog sucker*
Buffaloflabe*
Redhoraea
iuilh*«d«
Cacf iahea
Had COBS
Silveraldes
Temp«taie basses
Sunfiah
Black basses
Crappi««s
Perches
Freshwater drum
All other species
Tocal
0.6
0.5
25.5
6.4
0.8
18.1
0.7
5.2
5.2
2.S
6.8
0.8
18. S
10.0
8.2
1.4
1.2
112. S
24.2 1.3
4.5 1.9
10. § 5.4 1.8
0.1 0.6
4.2 0.3 O.S
5.2
Q.I 1.4
1.2
0.9 12. S
0.8
0.3 1.7 1.5
0.3 0.3
1.2
45.7 33. i 4.9
0.6
0.5
0.8
0.5
5.5
O.S
1.7 3.
8.6 0.
4.8 0.
0.9
24.7 4.
4
6
6
0
Another useful fish database is FishBase (note: several URLs are given in the succeeding
examples; if any link gets broken, then use Google to search on the key
words): www.fishbase.org/search.php. FishBase has 32,200 species at this time with more being
added. This database gets 33 million hits a month, so the response time is sometimes slow.
However, as of this writing, there are seven mirrors listed across the top of the Web page, so
look for one that is faster. FishBase is most useful as a source of information on feeding
preferences, but it also gives general information on distribution, size, and habitats (Figure 18).
25
-------
ft Species Summary - Microsoft Internet Explor
File Edit View Favorites Tools Help
*
' S [»] rti ^Search ^Favorites
Address @http://filaman.jni-Wel,de/Summary/SpedesSummary,cfm?geni.jsnarrie=Sander&spedesname=vitreus |v Q Go Links J> Norton AntiVirus
Sander vitreus
Walleye
You can sponsor this page:
picture (SKTt
Sander vitreus (Mtchiii,
Family: Percidae (Perches)
Order: Percifonnes (perch-likes)
Class: Actinopterygii (ray-finned fishes)
FishBase name: Walleye
Max. size: 107 cm FL (male/unsexed; Re£ 1998); max. published
weight: 11.3 kg (Ref 4699); max. reported age: 29 years ^
Envii-oiunent: demersal; freshwater; brackish , depth range - 27 m
Climate: temperate; 29.0°C; 55°N - 35°N
Importance: fisheries: commercial; aquaculture: experimental; gamefish: yes; aquanurn: public aquariums
Resilience: Low, minimum population doubling time 4.5 - 14 years (K=0.05; tm=2-4; tmax=29)
Distribution: North America: St. Lawrence-Great Lakes, Arctic, and Mississippi River basins from Quebec to
Gazetteer Northwest Territories in Canada, south to Alabama and Arkansas in the USA. Widely introduced
elsewhere in the USA, including Atlantic, Gulf, and Pacific drainages. Rarely found in brackish waters of
North America (Ref 1998).
Morphology: Dorsal spines (total): 13-17; Dorsal soft rays (total): 18-22, Anal spines: 2, Anal soft rays: 11-14;
Vertebrae: 44-48. Nuptial tubercles absent. Differentiation of sexes difficult. Branchiostegal rays 7,7 or
7,8 (Ref 1998).
Biology: Occurs in lakes, pools, backwaters, and runs of medium to large rivers. Prefers large, shallow lakes with
high turbidity (Ref 9988). Feeds at night, mainly on insects and fishes (prefers yellow perch and
freshwater drum but will take any fish available) but feeds on crayfish, snails, frogs, mudpuppies, and
small mammals when fish and insects are scarce (Ref 1998). Although not widely farmed commercially
for consumption, large numbers are hatched and raised for stocking lakes for game fishing (Ref 9988).
Utilized fresh or frozen; eaten pan-fried, broiled, microwaved and baked (Ref 9988).
Red List Status: Not in IUCN Red List , (Ref 36508)
Dangerous: harmless
Figure 18. Typical species page in FishBase.
The Wisconsin Bioenergetics Model (Hewett and Johnson 1992, Hanson et al. 1997) is the
source of the allometric parameters (i.e., those that vary with the size and weight of the fish, such
as respiration). These parameters are already incorporated into the fish records in the
AQUATOX library. However, the user may wish to refer to one of the original references.
ECOTOX is an extensive database published on CD-ROM by Elsevier (J0rgensen et al. 2000).
However, the database is out of print and may no longer be available. The database was first
published as a book twenty years earlier (J0rgensen 1979), and it may be available in libraries. It
should not be confused with the ecotoxicological database of the same name developed by the
Duluth Laboratory of the US EPA (see below).
ToxRefDB is a new US EPA portal for fate and toxicity data. The URL and the description on
the Web page follows: http://actor.epa.gov/toxrefdb/faces/Home.j sp. "ToxRefDB (Toxicity
Reference Database) captures thousands of in vivo animal toxicity studies on hundreds of
chemicals." Unfortunately, the information presented is so voluminous and organized for
purposes other than model parameterization so that it may be counterproductive for our purposes.
For example, there are 722 pages of information on chlorpyrifos, yet LCSOs for aquatic
26
-------
organisms are mingled with rat studies and are scattered over many pages. Furthermore, only one
octanol-water partition coefficient (KOW) value is given.
Far more useful for purposes of model setup is the US EPA ECOTOX database (Figure 19),
where data can be filtered and output in Excel format:
http ://cfpub. epa. gov/ecotox/
U.S. Environmental Protection Agency
Advanced
Database
Query
Recent Addrtiurif [ Cuiriact Hi | Prn it_j/et__; _ion Search:
EPA Home > ECOTOX
Quick
Database
Query
I The ECOTOX (ECOTOXicology) database provides single chemical toxicity information for aquatic and terrestrial life. ECOTOX is a useful
Itool for examining impacts of chemicals on the environment. Peer-reviewed literature is the primary source of information encoded in the
1 database. Pertinent information on the species, chemical, test methods, and results presented by the author(s) are abstracted and
I entered into the database. Another source of test results is independently compiled data files provided by various United States and
I International government agencies. Prior to using ECOTOX, you should visit the "About ECOTOX/Help" section of this Web Site. In
I addition, it is recommended that you consult the original scientific paper to ensure an understanding of the context of the data retrieved
I from the ECOTOX database.
If you use a popup blocker program, ECOTOX reports, help and browse features will not display. Please add the ECOTOX web site to
your popup browser exception list to ensure full usability,
Office of Research and Development I National Health and Environmental ETU- Laboratory | Mid-Continent Ecology Division
Figure 19. ECOTOX Web site.
The Web-ICE database, developed by US Office of Research and Development (Raimondo et al.
2007), is incorporated in AQUATOX (U.S. Environmental Protection Agency 2009, 2012) and
can be used to extend the toxicity data to species that are not in ECOTOX. Details on using
Web-ICE in AQUATOX are given in the User's Manual and the context-sensitive Help files.
The Agricultural Research Service (ARS) Pesticide Properties Database:
www.ars.usda.gov/Services/docs.htm?docid=14199 has been a useful source of chemical fate
parameters. Many of the physical-chemical parameters needed to represent older pesticides in
AQUATOX can be found in this ARS database (Figure 20). The biggest drawbacks are that it is
restricted to pesticides, and it was last updated in 2001.
27
-------
namejCHLORPYRIFOS CASRN: 2921-88-2
molecular formula: C9H11CL3NO3PS
rjD le cu la r w e 1 g r. t : 5 5 ;•. 6 2
(L=liguid; G=gas; S=3olid)
reference: 9ACH32
Key to sources: (M) Manufacturer, (R) eview, (H) andboolc, (3) xperiment,
- H •*^VvVVV^.-V^^.-V\'^.^V-.-VV-vVv '
(Calculated, (D)nknown, (P)SPA data, (W)auchope
* denotes a selected value where multiple values of a property are listed
-val^e- -medium- -temp- -pH- -source- -reference-
Boiling pj3.1ivt^deg C) :
Melting point(deg C):
» -^f^jf^jf^jf^jf^Mf *
42-43.5 H 9ACH32
Eecomposition point(deg C):
•^VVWWvWv^- J
Heat of vaporization(deg C):
•VvVV^.-V1v1v'v'v1v1v'v'v1v\-'v'v1v1'.-'v'v1v1'.-V J
—RAT2 CONSTANTS—
Hydrolysis (per day):
0.009* 25 5 M 6DOWCH
0.0236* 25 7 M 6DOWCH
0.0440* 25 9 M 6DOWCH
Photolysis (per day):
Vapor pressure (mPa):
2.^*^^^^^ m ^^Jl^,™™^
12 35 H 9ACHB2
-^svvvvvvv\\sv^^
2.3 20 R 9HRLCP
2.5(av.2.5,2.4,2.T) 25 M 6DOWCH
•*VVVVvVV*VVVVvVV *
Figure 20. Part of a record on chlorpyrifos in the ARS Pesticide Properties Database.
In addition to the sources given above, there are a number of books that can be used to obtain
parameter values for ecosystems (Goldman and Carpenter 1974, Home and Goldman 1994,
Wetzel 2001) and for organic chemicals (Verscheuren 1983, Schwarzenbach et al. 1993, Schnoor
1996). Of course, a search engine, such as Google, can query the open and "gray" scientific
literature for both historic and recent research.
Calibration and Validation Data
As stated in the Technical Documentation, (Rykiel 1996) defines calibration as "the estimation
and adjustment of model parameters and constants to improve the agreement between model
output and a data set" while "validation is a demonstration that a model within its domain of
applicability possesses a satisfactory range of accuracy consistent with the intended application
of the model." The purpose of calibration is to obtain the best goodness-of-fit to observed data
while using general, defensible parameter values. Validation is a test of the robustness of the
calibration by evaluating its application to an independent set of observed data. Data on state
variables, and indices based on state variables, while used in calibration and validation, may also
be used to specify initial conditions; but if so, they cannot be used for goodness-of-fit tests as
28
-------
well. Driving-variable data are external data that are introduced at the boundaries of the
simulation (see Boundary Conditions) and cannot be used directly for calibration and
validation.
As noted in the Introduction, models are often applied after the fact, using data that were
collected for other purposes. This especially can be a problem with data used for calibration and
validation. Data may be inadequate when sampling is infrequent and/or sporadic. For example,
samples taken every two weeks may completely miss an algal bloom. Problems also occur when
data do not specify the precise area sampled so that they can be normalized to a unit area or
volume. Furthermore, data on numbers of individuals (density) are common, but ideally they
would be paired with mean weights so that biomass can be calculated.
Nutrients
Nutrient data are usually easy to collect so that coverage is often relatively extensive, and their
use should be straightforward. However, two commonly encountered complications may be:
• the form of the nutrient data—total soluble phosphorus (TSP) should be distinguished
from total phosphorus (TP), which may include that associated with particulate and
dissolved organic matter (Figure 21); and nitrate may or may not include nitrite (Figure
22).
• data that are below the detection limit—the user may indicate such data, and AQUATOX
will plot those points with a special symbol (Figure 23). Although goodness-of-fit cannot
be determined directly, a suitable fit would be indicated if the model output is also below
the detection level.
Sometimes other forms of nutrient data are provided, which may increase uncertainty in the
model setup. If only soluble reactive phosphorus (SRP) is provided then the contributions of
soluble unreactive phosphorus (SUP) are ignored. SUP includes phosphorus temporarily tied up
in organic complexes, which are subject to cleaving by algal enzymes and UV light (Wetzel
2001). Total Kjeldahl nitrogen data are often available; these consist of organic nitrogen plus
ammonia. At present, AQUATOX does not accept data in this form; however, if the amount of
nitrogen tied up in organic compartments can be determined, the user can calculate the ammonia
concentration offline.
In the examples from the Lower Boise River, Idaho, the model was not calibrated to the nutrient
data, but rather, the data reflect concentrations in a river reach that is strongly affected by
wastewater treatment effluent (specified as loadings). Given the low retention times, the
loadings often overwhelm the in-stream processes. However, the upstream release of water
decreases considerably during the winter when no longer needed for irrigation, and some degree
of calibration may be possible with the longer retention. This is true for ammonia (Figure 23),
which was not calibrated and exhibits a poor fit during winter flow conditions. A more striking
example is given for dissolved oxygen, discussed below.
29
-------
Glenwood Bridge (PERTURBED)
Run on 11-16-073:35 PM
• Obs PO4 at Glenwood (mg/L)
Tot. Sol. P (mg/L)
TP (mg/L)
• TP at GB, City of Boise (mg/L)
2/27/1999 8/26/1999 2/26/2000 8/26/2000 2/24/2001 8/25/2001
Figure 21. Predicted and observed TSP and TP at Glenwood Bridge, Boise ID (Lower Boise
River). Note the predicted TSP and TP are superimposed most of the time, implying that the
model predicts that almost all P is inorganic.
Glenwood (PERTURBED)
Run on 10-24-07 10:40 AM
• Obs Nitrate at Glenwood (mg/L)
• Nitrate at GB, City of Boise (mg/L)
N03 (mg/L)
12/6/1999
12/5/2000
12/5/2001
Figure 22. Predicted nitrate-nitrite and observed nitrate at Glenwood Bridge, Boise ID (Lower
Boise River); observed data are from two different sources.
30
-------
Glenwood (PERTURBED)
Run on 10-24-07 10:40 AM
0.40
0.36
Obs Ammonia at Glenwood (mg/L)
Ammonia at GB, City of Boise (mg/L)
•NH3&NH4+(mg/L)
0.04
0.00
12/6/1999
12/5/2000
12/5/2001
Figure 23. Predicted and observed ammonia at Glenwood Bridge, Boise ID (Lower Boise River).
The symbol with a down arrow indicates a data point that is below detection limit.
Dissolved Oxygen
Dissolved oxygen can vary vertically within the water column, going from supersaturated near
the surface to anoxic at depth in some lakes and many reservoirs. This is a source of uncertainty
in the application of a model that only represents the epilimnion and hypolimnion as two well-
mixed layers. Ideally, observed data would be averaged for each of the two layers. In practice,
values often are arbitrarily chosen from "representative" depths. It is not surprising that the fit is
often better for an anoxic hypolimnion (Figure 24), as actual conditions would be expected to be
more uniform than in the epilimnion.
31
-------
ONONDAGA 2010 (PERTURBED) Run on 05-25-10 3:42 PM
(Hypolimnion Segment)
Oxygen (mg/L)
02HYPOBS(mg/L)
2/17/2005 4/18/2005 6/17/2005 8/16/2005 10/15/2005 12/14/2005
Figure 24. Observed and predicted dissolved oxygen in the hypolimnion of Lake Onondaga NY.
Biomass
Observed data on biomass of organisms are often difficult to obtain, especially for invertebrates
and fish. (Stevenson 1996) gives a good discussion of the different measurements of algal
biomass. The most accurate, but laborious, method for obtaining biomass of both phytoplankton
and periphyton is to compute the biovolume of each algal species. As described by (Hill and
Boston 1991), they "measured the dimensions of at least 15 individuals with an ocular
micrometer. The averages of these dimensions were used in volumetric formulae of appropriate
geometric shapes. Biovolume (|im3/cm2) for each algal species was calculated by multiplying
cell density (cells/cm2) by estimated cell volume." Biovolume can then be converted to biomass
(g/m3) for use in AQUATOX. Table 4 in (Reynolds 1984) lists biovolumes and other common
measures of biomass for common species.
The most common surrogate for algal biomass is chlorophyll a. AQUATOX assumes a constant
relationship between chlorophyll a and biomass, depending on the algal group and whether or
not it is periphyton. Biomass is output and is also converted to chlorophyll a so that the predicted
values can be compared with observations, whether expressed as biomass or chlorophyll a. It is
not unusual to have two or three replicate samples; we prefer to plot each point so that the spread
of values can be assessed (Figure 25). Of course, if sufficient samples were collected to calculate
standard deviations then those should be plotted. Care must be taken to determine the
representativeness of the sampling site, especially for periphyton in streams where the substrate
may be very heterogeneous. For example, if periphyton samples are taken from cobbles, it is
reasonable to normalize the values for the percent of the site with cobbles in order to compare
with the AQUATOX results, which are for the entire site. That was done by the user for the
32
-------
example shown in Figure 25 (the percent riffles site characteristic is not used by AQUATOX to
automatically normalize the results).
Parma (PERTURBED)
Run on 10-24-07 10:37 PM
Peri Chi a at Parma (Norm/ (mg/sq.m)
• Peri. Chlorophyll (mg/sq.m)
12/6/1999
12/5/2000
12/5/2001
Figure 25. Predicted and observed periphytic chlorophyll a at Parma site, Lower Boise River ID.
AQUATOX computes algal biomass as ash-free dry weight (AFDW). In a study of the Cahaba
River, Alabama, AFDW data were collected specifically for use in the model. Furthermore, the
aquatic moss Fontinalis was analyzed and modeled separately (Figure 26).
33
-------
Cahaba River AL (CONTROL)
Run on 10-29-084:53 PM
20.0
18.0
• Peri. Biomass (g/m2 dry)
• Fontinalis a (g/m2 dry)
Obs Peri AFDW(g/m2 dry)
Obs Moss AFDW(g/m2 dry)
0.0
1/4/2002 3/5/2002 5/4/2002 7/3/2002 9/1/2002 10/31/2002 12/30/2002
Figure 26. Predicted and observed AFDW biomass of periphyton and moss, Cahaba River AL.
A quick method for characterizing periphyton biomass is to estimate thickness as part of a rapid
bioassessment (Barbour et al. 1999). Thickness measurements are ranked: "0" indicates a rough
surface with no algae, "1" for no algae visible but surface is slimy, "2" for 0.5 mm to 1 mm, "3"
indicates an algal cover of greater than 1 mm but less than 5 mm, "4" for 5-20 mm and "5" for
algae over 20 mm (Barbour et al. 1999). These can be converted to AFDW (g/m ) values based
on polynomial regression of paired thickness and AFDW data in the literature (Stevenson et al.
2007) (Figure 27). The conversion yields estimates with a relatively large error component,
which should be made explicit in plots by plotting error bars.
34
-------
•Polynomial fit (0.08821
+0.4793X +0.09005X2)
• 95% Cl
1234
Periphyton Scaled Thickness
Figure 27. Regression of AFDW on diatom thickness rank, based on data of Stevenson et al.
(2007); R2 = 0.77.
Another challenge in using data collected for other purposes occurs then the data were taken
from periphytometers—artificial substrates placed in the stream for a given period of time.
Depending on the design, the periphytometers may or may not exclude grazers. Although natural
substrates are preferred, periphytometer data can be used if grazers were not excluded; if they
have been excluded then the data will not be usable for purposes of modeling algal biomass, as
grazers can be a significant control on periphyton biomass.
Density (Numbers of Individuals)
Invertebrate and fish data are often reported as numbers of individuals or density. Presumably
the data are collected from a unit area so that they are internally consistent, although sometimes
the area is not reported—especially for fish collected by electrofishing. If the area is given and if
mean weights are reported, then the data can be converted to biomass (g/m ) and compared
directly with model output. However, if only density is reported then one can still compare
trends indirectly. For example, in a microcosm study conducted at Duluth MN (U.S.
Environmental Protection Agency 1988), a pond enclosure was dosed with 6.3 |ig/L
chlorpyrifos. Data on chironomids were reported as numbers/sample, which were compared with
predicted biomass by plotting the latter on the Yl axis and the density on the Y2 axis (Figure
28). Of course, such comparisons are only approximate because the weights of the individuals
may change considerably over the period being simulated, especially if it involves fast-growing
aquatic insect larvae. Nevertheless, in this example the comparison was informative and
35
-------
provided a limited validation of the ecotoxicological submodel of AQUATOX. Similarly, fish
data are often reported as catch per unit effort (CPUE). Here again trends can be compared by
plotting biomass on the Yl axis and CPUE on the Y2 axis.
0.36
•a
ai
0.00
CHLORPYRIFOS 6 ug/L (PERTURBED)
Run on M 97-flQ R"I1 AM
—
6/27/1986 7/27/1986 8/26/1986
AI - - 1 * 1 « 1 »
• Obs. Chironomids (no./sample)
-900
Figure 28. Predicted biomass of chironomids compared with observed numbers/sample (Y2 axis)
in pond microcosm at Duluth MN dosed with 6.3 |ig/L chlorpyrifos.
Percent Composition and Other Metrics
Species data may be available as percent composition. AQUATOX can convert biomass of
many biotic groups to percentages, both for purposes of comparison to reported values and as
metrics (such as percent blue-greens) used by decision makers. There are a couple of pitfalls to
this approach. First, percentages involve closed-number systems that are sensitive to values of
other state variables. In Figure 29 the percent blue-greens is partly an inverse function of the
amount of diatom biomass at any given time.
Second, reported percent compositions are usually based on numbers of individuals, but
AQUATOX computes them on the basis of biomass, making comparisons between observations
and predictions difficult. Biomass-based percentages are actually more accurate because there is
a common basis; percentages based on numbers can be biased by numerous small organisms
compared to a few large organisms. In computing percentages of invertebrates, AQUATOX
excludes mussels because one large individual can exceed all the biomass of the other
invertebrates at a site.
Metrics based on predicted percent aquatic insect larvae may not coincide temporally to
observed data. AQUATOX simulates emergence of insects based on computed growth rates.
That approximation may throw off the timing and cause a dip in biomass that is out of phase with
36
-------
the actual emergence. Therefore, the magnitude of the predicted metric, such as percent
chironomids (Figure 30), is much more important than the timing.
Gross primary productivity (GPP) is a metric that represents overall primary productivity
(expressed as oxygen production) at a site and can be summarized over time by calculating the
mean and standard deviation (Figure 31). Of course, the predicted and observed statistics should
be calculated for the same period.
Lake B Epi. (Control)
Run on 07-8-09 10:17 AM
2.2
2.0
1.8
1.6
1.4
^ 1.2
f 1.0
0.8
0.6
0.4
0.2
0.0
i_/\
1
I
1 11 *
J JL
fj\l\^\ rt
!L&w]v4cJw
i
i
J
ten
12/7/1992
1
ML
w
h
sgy
•52
•47
-36
-31
26 #
•21
•16
•10
•5
Phyt Low -Nut Diatom (mg/L dry)
Phyt, Blue-Green max (mg/L dry)
• Obs blue-greens (%)
12/7/1993
Figure 29. Predicted and observed percent blue-greens plotted with algal biomass, especially
diatoms.
37
-------
Upatoi Creek Ft Benning GA (Control)
Run on 01-22-105:02 PM
100
•Pet Chironomid (%)
Obs%Chiro (percent)
•Chironomid Emergel (Percent)
Chironomid Scour_Entrain (Percent)
12/6/1999 12/5/2001
12/5/2003
Figure 30. Predicted and observed percent chironomids in Upatoi Creek GA. The steep drops are
predicted due to a combination of emergence and, much more important, scour from high flow.
Sally Br Trib 4 Ft Benning GA (PERTURBED)
Run on 06-7-10 12:17 PM
0.50
•GPP (gO2/m2d)
Obs GPP +- Std Dev (gO2/m2 d)
Pred GPP +- Std Dev (gO2/m2 d)
12/5/2001
12/5/2002
12/5/2003
Figure 31. Predicted and observed gross primary productivity (GPP) in Sally Branch Tributary
#4, Fort Benning GA. Statistics (mean ± 1 standard deviation) are for the entire three-year
period, and therefore their position on the horizontal axis is arbitrary.
Organic Chemicals
Persistent bioaccumulative organic compounds may be simulated for both short and long periods
of time. Unless the goal is to simulate acute effects, the model should be run at least to
equilibrium. A quick way to determine the length of time required for such a model run is to
check the chemical-parameter input screen where an estimate for equilibrium in fish is given
38
-------
based on the octanol-water partition coefficient. For example, according to the estimate in the
chemical-parameter screen, PCB 1254 should reach equilibrium in fish in 366 days. This result
is exemplified by a simulation of Galveston Bay TX (Figure 32). The results shown use PCB
exposure data for a Massachusetts bay in a calibrated Galveston Bay Texas AQUATOX
simulation (exposure data were not available for Galveston Bay). The only serious mismatch
between predicted and observed PCB concentrations is for polychaetes, which apparently are not
getting adequate exposure to PCBs in the sediments within this simulation. Note the
concentrations use the customary units of |ig/kg wet weight, in contrast to concentrations in
sediments that are expressed as |ig/kg dry weight.
Galveston Bay TX, PCB 1254 (PERTURBED)
(Lower Segment)
1.0E+4 =
1.0E+3 =
, 1 .OE+2 =
1 .OEM :
1.0E+0 •
• TIPolychaete Streblosp(ppb) (ug/kg wet)
TICallinectes (Crab)(ppb) (ug/kg wet)
- TIMugil (mullet)(ppb) (ug/kg wet)
Obs PCBs Polychaete (ug/kg wet)
Obs PCBs Mussel (ug/kg wet)
• TlOstrea (oyster)(ppb) (ug/kg wet)
Obs PCBs Crab (ug/kg wet)
Obs PCBs Flounder (ug/kg wet)
12/6/1999
12/5/2000
12/5/2001
Figure 32. Predicted concentrations of PCBs in Galveston Bay TX animals using exposure data
and observed concentrations from a Massachusetts bay.
A common way to express bioaccumulation is as a bioaccumulation factor (BAF), which is the
concentration in the organism divided by the concentration in the water. However, one must be
aware of what the "concentration in the water" means, because 50% or more could be bound to
colloidal detritus (Oliver and Niimi 1988). AQUATOX provides the alternative to "Include
complexed toxicant in BAF calculations" when setting up the study. Essentially, this option
allows the analyst to use legacy data collected before the relationship to colloidal organics was
recognized. The data of (Oliver and Niimi 1988) were used in a study of PCBs in Lake Ontario.
The observed data include the complexed PCBs in the calculation, so that option was chosen for
the simulated values too (Figure 33).
39
-------
8.1
LL 45
< 4.O
CO
O
^36
0
LAKE ONTARIO PCBs (CONTROL)
(Epilimnion Segment)
t.f^ _^ J •**T-J VJ •
/^^^^^^^^^^~
u
s
12/12/1972 12/12/1973 12/12/1974 12/12/1975
BAF TlOCyclotella nana
BAF TIOAIewife
BAFTIOSculpin
BAF TlOSmelt
BAF TIOLake Trout
• PCB 118 Am phipod (loglOBAF wet)
• PCB 1 18 Mysid (loglOBAF wet)
• PCB118Alewife(log10BAFwet)
o PCB 1 18 Sculpin (loglOBAF wet)
• PCB118 Smelt (loglOBAF wet)
o PCB 118 Trout (loglOBAF wet)
Figure 33. Predicted and observed BAFs for PCB 118 in Lake Ontario; complexed PCBs were
used in the computation of BAFs.
Sensitivity Analysis
AQUATOX includes a built-in nominal range sensitivity analysis (Frey and Patil 2001), which
varies each user-specified parameter and loading by plus and minus a given percentage and ranks
them according to how much each input affects the target output. Furthermore, one can perform
statistical sensitivity analysis in which parameters are tested one at a time using an appropriate
distribution of parameter values. By performing sensitivity analyses on parameters and loadings
once the model is initially calibrated important parameters deserving additional attention can be
identified, and alternate parameter values can be identified. These may then be used to improve
the agreement between the observed and predicted values of the calibration target. For example,
nominal range sensitivity and statistical sensitivity analysis were performed on Onondaga Lake
NY. One of the sensitive parameters for chlorophyll a was the maximum photosynthetic rate
(PMax) for diatoms (nominally Cyclotella) (Figure 34). This parameter was then used in a
statistical sensitivity analysis, where a normal distribution of diatom PMax values was used,
based on the compilation of values by (Collins and Wlosinski 1983); the results are shown in
Figure 35.
40
-------
Sensitivity of Phyto. Chlorophyll (ug/L) to 15% change in tested parameters
3/30/2009 12:44:31 PM
65.6% - Cryptomonad: Max Photosynthetic Rate (1/d)
52.6% - Phyt, Blue-Gre: Optimal Temperature (deg. C)
52.5% - Cyclotella nan: Max Photosynthetic Rate (1/d)
49.1% - Cryptomonad: Temp Response Slope
46.5% - Cryptomonad: Exponential Mort. Coefficient: (max / d)
43.3% - Susp&Diss Detr: Mult. Point Source Load by
37.7% - Cryptomonad: Maximum Temperature (deg. C)
26.6% - Phyt, Blue-Gre: Temp Response Slope
24.2% - Cryptomonad: Optimal Temperature (deg. C)
23.3% - Greens: Max Photosynthetic Rate (1/d)
20.6% - Cyclotella nan: Light Extinction (1/m)
19.1% - Susp&Diss Detr: Mult. Non-Point Source Load by
16.3% - Cryptomonad Resp. Rate, 20 deg C (g/g d)
14.4% - Cyclotella nan: Temp Response Slope
23 24 25
Phyto. Chlorophyll (ug/L)
26
Figure 34. Sensitivity of Phytoplanktonic Chlorophyll a in Lake Onondaga NY.
Phyto. Chlorophyll (u
4/29/2009 4:58:03 PM
90.0
80.0
- Mean
• Mean -StDev
• Mean + StDev
Deterministic
1/12/1989
5/12/1989
9/9/1989
1/7/1990
5/7/1990
9/4/1990
1/2/1991
Figure 35. Sensitivity of chlorophyll a to diatom PMax in Lake Onondaga NY.
The mean results seemed to be a better fit than the deterministic, so the model was run with
PMax =1.6 (the mean of observed values) instead of the prior value of 3.4. The results
(perturbed) did seem better than those with the original value (control), especially in the second
year when a clearing event with low observed chlorophyll a was simulated (Figure 36, Figure
41
-------
37). Of course, any new parameter value should be within a reasonable range, not just one that
gives a better fit.
ONONDAGA LAKE, NY (CONTROL) Run on 04-22-09 9:59 AM
(Epilimnion Segment)
A Epilimnion Chla(ug/L)
— Phyto. Chlorophyll (ug/L) |
1/12/1989 5/12/1989 9/9/1989 1/7/1990 5/7/1990 9/4/1990 1/2/1991
Figure 36. Original calibration of chlorophyll a with diatom PMax = 3.4 in Onondaga Lake NY.
ONONDAGA LAKE, NY (PERTURBED) Run on 04-29-09 5:08 PM
(Epilimnion Segment)
A Epilimnion Chla (ug/L)
— Phyto. Chlorophyll (ug/L)
1/12/1989 5/12/1989 9/9/1989 1/7/1990 5/7/1990 9/4/1990 1/2/1991
Figure 37. Chlorophyll a result with diatom PMax = 1.6 in Onondaga Lake NY. Note better fit in
1990.
Summary
42
-------
Seldom do modelers have the luxury of being able to dictate what data should be collected before
setting up the model. Often the model application comes after the original study and must rely on
incomplete sets of data collected for other purposes. One of the objectives of this Note is to
describe ways of using available data and of filling in the gaps with other sources.
Site characteristics include length and width, to which the model is not usually sensitive; and
depth and thermocline depth, which can be important. Site loadings are often very important and
include, roughly in order of importance, time-varying inflow and discharge, flow among linked
segments, organic chemicals, nutrients, several detrital compartments (based on organic matter,
organic carbon, or BOD), and total suspended solids; and seasonally-varying light, temperature,
dissolved oxygen, pH, and biotic loadings. The analyst may of necessity use diverse data
sources such as USGS for flow and nutrient data from a representative gage, NASA for solar
radiation estimates, and state and local laboratories for data that often require conversion to units
used in AQUATOX.
Because models are often applied using data collected for other purposes, data for calibration and
validation may be inadequate or problematic. Nutrient data are often extensive; however, certain
chemicals, such as Kjeldahl nitrogen, may be analyzed to the exclusion of other forms, such as
nitrate, making it difficult to evaluate nutrient mass balance in the model. Biomass is most
important for calibrating ecosystemic responses, yet temporal coverage is often sporadic or
confined to one season. Furthermore, biotic data are often collected as environmental indices
and cannot be converted to biomass per unit area or volume without making tenuous
assumptions. Often the analyst is left comparing simulated biomass trends with trends in
numbers of individuals (especially for invertebrates) and catch per unit effort (for fish).
Obtaining parameter values can be difficult, especially since they are scattered among many
published papers and gray-literature reports. Fortunately, there are compendiums of parameter
values assembled in support of models. Unfortunately, several of these were published decades
ago and have not been kept up to date. The exceptions are FishBase and EPA's ECOTOX, which
continue to track the literature and grow in size. Now, Internet search engines are able to find
parameter values almost instantaneously in both the open literature and in obscure reports and
other sources.
The model is usually not sensitive to initial conditions, so that one may use a spin-up period to
achieve equilibrium. Finally, it is a good idea to perform sensitivity analyses on parameters and
loadings once the model is initially calibrated; AQUATOX has several tools to facilitate this.
Important parameters deserving additional attention can be identified, and alternate parameter
values can be tentatively identified.
In conclusion, the model has extensive data requirements, and those requirements are often not
satisfied with data collected for purposes of the model. This paper describes ways of coping with
commonly encountered data issues. Data can usually be obtained from a variety of sources, and
those data can be converted and conditioned to forms acceptable to the model. Furthermore,
AQUATOX can accept various data types, facilitating application to numerous water body types
and modeling goals.
43
-------
References
Arnold, J. G., R. Srinivasan, R. S. Muttiah, and J. R. Williams. 1998. Large-area Hydrologic
Modeling and Assessment: Part I. Model Development. J. American Water Res. Assoc.
34:73-89.
Barbour, M. T., J. Gerritsen, B. D. Snyder, and J. B. Stribling. 1999. Rapid Bioassessment
Protocols for Use in Streams and Wadeable Rivers: Periphyton, Benthic
Macroinvertebrates and Fish, Second Edition. EPA 841-B-99-002, U.S. Environmental
Protection Agency; Office of Water, Washington, D.C.
Bicknell, B. R., J. C. Imhoff, J. L. Kittle, A. S. Donigian, and R. C. Johanson. 1991.
Hydrological Simulation Program - FORTRAN — User's Manual for Release 10 (Pre-
release Draft Version). U.S. EPA Technology Development and Applications Branch in
cooperation with USGS Water Resources Division, Office of Surface Water.,
Washington, D.C.
Carleton, J. N., R. A. Park, and J. S. Clough. 2009. Ecosystem Modeling Applied to Nutrient
Criteria Development in Rivers. Environmental Management 44:485-492.
CH2M HILL, Eco Modeling, Warren Pinnacle Consulting, and Boise City Public Works. 2008.
Numeric Nutrient Criteria Development for the Lower Boise River Using the
AQUATOX Model. U.S. Environmental Protection Agency, Office of Water,
Washington, DC.
Collins, C. D. and J. H. Wlosinski. 1983. Coefficients for Use in the U.S. Army Corps of
Engineers Reservoir Model, CE-QUAL-R1. Tech. Rept. E-83-15, Environmental
Laboratory, U.S. Army Engineer Waterways Experiment Station, Vicksburg, Miss.
Cowan, W. L. 1956. Estimating hydraulic roughness coefficients. Agricultural Engineering
37:473-475.
Donigian, A. S., J. T. Love, J. S. Clough, R. A. Park, J. N. Carleton, P. A. Cocca, and J. C.
Imhoff. 2005. Nutrient Criteria Development with a Linked Modeling System: Watershed
and Ecological Model Application and Linkage.in 2005 TMDL Conference. Water
Environment Federation, Philadelphia PA.
Donigian, A. S., R. A. Park, and J. S. Clough. 2009. Tenkiller Ferry Lake Modeling Case Study -
HSPF and AQUATOX. Office of Water/Office of Science and Technology, U.S.
Environmental Protection Agency, Washington DC.
Fourt, R., S. Wooldridge, and D. Brader. 2002. Striped Bass Management Plan. Arkansas Game
and Fish Commission.
Goldman, J. C. and E. J. Carpenter. 1974. A kinetic approach to the effect of temperature on
algal growth'. Limnology and Oceanography 19:756-776.
Hanson, P. C., T. B. Johnson, D. E. Schindler, and J. F. Kitchell. 1997. Fish Bioenergetics 3.0.
Center for Limnology, University of Wisconsin, Madison.
Hewett, S. W. and B. L. Johnson. 1992. Fish Bioenergetics 2 Model. University of Wisconsin
Sea Grant Institute, Madison.
Hill, W. R. and H. L. Boston. 1991. Community development alters photosynthesis-irradiance
relations in stream periphyton. Limnol. Oceanogr. 36:1375-1389.
Home, A. J. and C. R. Goldman. 1994. Limnology. Second edition. McGraw-Hill, New York.
J0rgensen, L. A., S. E. J0rgensen, and S. N. Nielsen. 2000. ECOTOX: Ecological Modelling and
Ecotoxicology. Elsevier Science B.V.
44
-------
J0rgensen, S. E., editor. 1979. Handbook of Environmental Data and Ecological Parameters.
Pergamon Press, Oxford.
Leidy, G. R. and R. M. Jenkins. 1977. The Development of Fishery Compartments and
Population Rate Coefficients for Use in Reservoir Ecosystem Modeling. CR-Y-77-1, U.S.
Army Engineer Waterways Experiment Station, Vicksburg, Miss.
Leidy, G. R. and G. R. Ploskey. 1980. Simulation Modeling of Zooplankton and Benthos in
Reservoirs: Documentation and Development of Model Constructs. E-80-4, U.S. Army
Engineer Waterways Experiment Station, Vicksburg, Miss.
Levins, R. 1966. The strategy of model building in population biology. American Scientist
54:421-431.
NASA. 2010. Surface Meteorology and Solar Energy Interannual Variability. NASA
Atmospheric Science Data Center.
Oliver, B. G. and A. J. Niimi. 1988. Trophodynamic Analysis of Polychlorinated Biphenyl
Congeners and Other Chlorinated Hydrocarbons in the Lake Ontario Ecosystem. Environ.
Sci. Technol. 22:388-397.
Park, R. A., J. N. Carleton, J. S. Clough, and M. C. Wellman. 2009. AQUATOX Technical Note
1: A Calibrated Parameter Set for Simulation of Algae in Shallow Rivers. EPA-823-R-
09-003, U.S. Environmental Protection Agency, Washington D.C.
Park, R. A. and C. D. Collins. 1982. Realism in Ecosystem Models. Perspectives in Computing
2:18-27.
Raimondo, S., D. N. Vivian, and M. G. Barren. 2007. Web-based Interspecies Correlation
Estimation (Web-ICE) for Acute Toxicity: User Manual. Version 1.1. EPA/600/R-
07/071, U.S. Environmental Protection Agency, Gulf Breeze, FL.
Reynolds, C. S. 1984. The Ecology of Freshwater Phytoplankton. Cambridge University Press,
Cambridge.
Rykiel, E. J., Jr. 1996. Testing ecological models: the meaning of validation. Ecological
Modelling 90:229-244.
Saunders, G. W. 1980. Organic Matter and Decomposers. Pages 341-392 in E. P. Le Cren and R.
H. Lowe-McConnell, editors. The Functioning of Freshwater Ecosystems. Cambridge
University Press, Cambridge.
Schnoor, J. E. 1996. Environmental Modeling: Fate and Transport of Pollutants in Water, Air,
and Soil. John Wiley & Sons, Inc., New York.
Schwarzenbach, R., P. M. Gschwend, and D. M. Imboden. 1993. Environmental Organic
Chemistry. John Wiley & Sons, New York.
Stevenson, R. J. 1996. An Introduction to Algal Ecology in Freshwater Benthic Habitats. Pages
3-30 in R. J. Stevenson, M. L. Bothwell, and R. L. Lowe, editors. Algal Ecology:
Freshwater Benthic Ecosystems. Academic Press, San Diego.
Stevenson, R. J., Y. Pan, J. P. Kociolek, and J. C. Kingston, editors. 2007. Advances in Algal
Biology: A Commemoration of the Work of Rex Lowe. Springer, Amsterdam.
Suarez, L. A. 2005. PRZM-3, A Model for Predicting Pesticide and Nitrogen Fate in the Crop
Root and Unsaturated Soil Zones: Users Manual for Release 3.12.2. EPA/600/R-05/111,
U.S. Environmental Protection Agency, Athens GA.
Tetra Tech Inc. 2002. Draft User's Manual for Environmental Fluid Dynamics Code Hydro
Version (EFDC-Hydro). U.S. Environmental Protection Agency, Atlanta, GA.
45
-------
U.S. Environmental Protection Agency. 1988. The Effects of Chlorpyrifos on a Natural Aquatic
System: A Research Design for Littoral Enclosure Studies and Final Research Report.
Duluth MN.
U.S. Environmental Protection Agency. 2012. AQUATOX (Release 3.1) Modeling
Environmental Fate and Ecological Effects in Aquatic Ecosystems, Volume 2: Technical
Documentation. EPA-820-R-12-015, U.S. Environmental Protection Agency, Office of
Water, Office of Science and Technology, Washington DC.
Verscheuren, K. 1983. Handbook of Environmental Data on Organic Chemicals. Second edition.
Van Nostrand Reinhold, New York.
Walker, W. W. 1999. Simplified Procedures for Eutrophication Assessment and Prediction: User
Manual. Instruction Report; W-96-2, U.S. Army Engineer Waterways Experiment
Station, Vicksburg, MS.
Wetzel, R. G. 2001. Limnology: Lake and River Ecosystems. Third edition. Academic Press, San
Diego.
46
------- |