United States
Environmental Protection
Agency
MICHTOX: A Mass Balance
and Bioaccumulation Model
for Toxic Chemicals in
Lake Michigan
RESEARCH AND DEVELOPMENT
-------
EPA/600/R-05/158
December 2005
MICHTOX: A Mass Balance and
Bioaccumulation Model for
Toxic Chemicals in
Lake Michigan
Ronald Rossmann, Editor
Large Lakes and Rivers Forecasting Research Branch
Mid-Continent Ecology Division
National Health and Environmental Effects Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Large Lakes Research Station
9311 Groh Road
Grosse He, Michigan 48138
Printed on Recycled Paper
-------
Notice
The information in this document has been obtained primarily through funding by the United Sta~es
Environmental Protection Agency (USEPA) under the auspices of the Great Lakes National Program Office
(GLNPO) and the Office of Research and Development (ORO). The report has been subjected to the
Agency's peer and administrative review and it has been approved for publication as a USEPA document.
Mention of trade names or commercial products does not constitute endorsement or recommendation for use.
Because the purpose of this report is to document the development of the MICHTOX model, conclusions are
historical and should not be considered current.
ii
-------
Foreword
Federal and contractor staff at the United States Environmental Protection Agency's Large Lakes Research
Station have been involved with the development of mass balance models for the Great Lakes since the early
1970s. MICHTOX is a mass balance model developed to predict chemical concentrations in water and
sediments of Lake Michigan in response to chemical loads to the lake. The model was adapted from the
general water quality model WASP4. The MICHTOX bioaccumulation model was based upon the WASTOXv4
food chain model. Development of MICHTOX began in the early 1990s. The model was developed as a
planning tool for the Lake Michigan Mass Balance Project (LMMBP) (U.S. Environmental Protection Agency,
1997). This work was documented in an in-house report in 1992 (Part 1). The model was applied as a
screening-level model for atrazine in Lake Michigan in support of the LMMBP (Rygwelski et al., 1999). The
model was slightly revised and applied to polychlorinated biphenyls (PCBs) in Lake Michigan to confirm model
results with the LMMBP project data and to provide preliminary modeling results for inclusion in the 2002 Lake-
wide Management Plan (LaMP) report (Lake Michigan Technical Committee, 2002). These were reported in
a 2002 contractor report (Part 2). The purpose of this report is to document through 2002 the progression of
MICHTOX model development and application of the model to describing the behavior of contaminants,
especially PCBs, in Lake Michigan. Both parts of this report have been cited numerous times in the literature.
This report provides ready access to these for interested parties. For PCBs, results from application of the
model have been superceded by more recent results.
Lake Michigan Technical Committee. 2002. Lake Michigan Lakewide Management Plan (LaMP), 2002. U.S.
Environmental Protection Agency, Great Lakes National Program Office, Chicago, Illinois. 102 pp.
Rygwelski, K.R., W.L. Richardson, and D.D. Endicott. 1999. A Screening-Level Model Evaluation of Atrazine
in the Lake Michigan Basin. J. Great Lakes Res., 25(1 ):94-106.
U.S. Environmental Protection Agency. 1997. Lake Michigan Mass Budget/Mass Balance Work Plan. U.S.
Environmental Protection Agency, Great Lakes National Program Office, Chicago, Illinois. EPA/905/R-97/018,
155 pp.
iii
-------
Abstract
MICHTOX is a toxic chemical mass balance and bioaccumulation model for Lake Michigan. It was developed
for the United States Environmental Protection Agency's Region V in support of the Lake Michigan Lake-wide
Management Plan (LaMP) to provide guidance on expected water quality improvements in response to critical
pollutant loading reductions. The 11 critical pollutants modeled were benzo(a)pyrene, chlordane, total
dichlorodiphenyltrichloroethane (DOT), dieldrin, heptachlor epoxide, hexachlorobenzene, lead, total
polychlorinated biphenyls (PCBs), 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), 2,3,7,8-tetrachlorodibenzofuran
(TCDF), and toxaphene. Concentrations of these were predicted in 17 water and sediment segments in
response to atmospheric and tributary loadings. The bioaccumulation model was coupled to the mass balance
model to predict chemical accumulation in lake trout and bloater through pelagic and benthic food chains.
Mass balance predictions were validated using plutonium, lead, and PCBs data; and bioaccumulation
predictions were validated with PCBs data. The model was later applied to provide preliminary PCBs model
results for the Lake Michigan Mass Balance Project. Results from this application were used to guide the
development of a more resolute model for PCBs. Results for PCBs described in Part 1 are superceded by
results in Part 2. Part 2 results have been replaced by a more recent application of MICHTOX that has been
presented at various meetings and will be published at a future date. This document is meant to provide a
historical perspective of MICHTOX development and application.
iv
-------
Table of Contents
Notice. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Foreword. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . '. . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . .
Abstract. . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .'. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Table of Contents. . . . . . . . . . . '. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
List of Figures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
List of Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Abbreviations. . . . . . . . . . .'. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Executive Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . " . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Part 1 1992 MICHTOX: A Mass Balance and Bioaccumulation Model for Toxic
Chemicals in Lake Michigan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . , " . . . . . . . . . . . . . . .
1.1
1.2
Executive Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Recommendations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1 Verification of Model Predictions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1.1 Air Concentrations/Deposition Fluxes. . . . . . . . . . . . . . . . . . . . .
1.2.1.2 Surficial Sediment Concentrations. . . . . . . . . . . . . . . . . . . . . . . .
1.2.1.3 Lake Trout. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1.4 Major Tributaries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1.5 Water..............................................
Extend Model to Other Critical Pollutants and Target Organisms. . . . . . . .
Further Model Development. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.3.1 Circulation..........................................
1.2.3.2 Sediment Transport. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1 .2.3.3 Organic Carbon Dynamics. . . . . ~ . . . . . . . . . . . . . . . . . . . . . . . .
1.2.3.4 Food Chain Variability and Dynamics. . . . . . . . . . . . . . . . . . . . . .
1.2.4. Establish Linkages to Atmospheric and Watershed Models. . . . . . . . . . . .
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.1 Project Objectives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.2 Lake Michigan Toxics Problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Model Description. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.1 Model Framework. . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . .
1.4.2 Segmentation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . '. ..
1 .4.3 Circulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . : . . . . . . . . . . .
1.4.4 Solids Balance. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . .
1.2.2
1.2.3
1.3
1.4
v
ii
iii
iv
v
viii
xiv
xvi
xvii
xviii
1
1
3'
3
3
3
3
3
3
3
4
4
4
5
5
5
6
6
7
8
8
10
12
13
-------
1.4.5
1.5
Chemical Partitioning and Loss. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.5.1 Partitioning..........................................
1.4.5.2 Volatilization.........................................
1 .4.5.3 Photolysis...........................................
1.4.5.4 Sediment-Water Diffusive Exchange. . . . . . . . . . . . . . . . . . . . . .
Chemical Loads and Boundary Conditions. . . . . . . . . . . . . . . . . . . . . . . . .
1 .4.6.1 Atmospheric Deposition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.6.2 Tributary Loads. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.6.3 Loading Histories. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1 .4.6.3.1 Plutonium..................................
1.4.6.3.2 Lead......................................
1.4.6.3.3 PCBs.....................................
1.4.6.4 Lake Huron Boundary Conditions. . . . . . . . . . . . . . . . . . . . . . . .
Chemical Bioaccumulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.7.1 Food Chain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.7.2 Uptake Rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.7.3 Elimination Rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.7.4 Dietary Accumulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.7.5 Modeling the Base of the Food Chain. . . . . . . . . . . . . . . . . . . . .
1.4.8 Steady-State Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.9 Chemical-Specific Parameterization.... . .. . .. . ... .. ... .,. . ... . ...
Model Validation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5.1 Plutonium. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5.2 Lead. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5.3 PCBs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5.3.1 Water..............................................
1.5.3.2 Sediment...........................................
1.5.3.3 Biota...............................................
1.5.3.4 Bioaccumulation......................................
Steady-State Model Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.1 Steady-State Load-Response Predictions. . . . . . . . . . . . . . . . . . . . . . . . .
1.6.2 Mass Fate and Transport. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.3 Sensitivity Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.4 Uncertainty Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.4.1 Analysis of Model Uncertainty. . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.4.2 Results.............................................
1.6.4.3 Critical Parameterization Uncertainty. . . . . . . . . . . . . . . . . . . . . .
Dynamic Model Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.1 Toxic Chemical Lag Time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.72 PCBs Fate and Transport Fluxes in Dynamic Simulations. . . . . . . . . . . . .
1.7.3 Additional Dynamic Simulations for PCBs. . . . . . . . . . . . . . . . . . . . . . . . .
1.7.3.1 PCBs Control Scenarios. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.3.2 Severe Storm Event. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.4 Uncertainty in Dynamic Simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Appendix: Steady-State Model Output for Each Toxic Chemical. . . . . . . . . . . . . . . . .
1.4.6
1.4.7
1.6
1.7
1.8
1.9
vi
15
15
16
18
18
18
18
19
20
20
20
20
21
21
23
23
23
25
25
26
28
30
31
32
33
33
34
35
38
38
41
41
44
50
50
51
54
57
57
60
60
61
61
62
64
72
-------
Part 2 2002 Lake Michigan Mass Balance Project: Modeling Total Polychlorinated Biphenyls
Using the MICHTOX Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1
2.2
2.3
2.4
2.5
2.6
Executive Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Recommendations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Description of Model, Data, and Simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.1 MICHTOX Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.2 Mass Balance Data for PCBs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.3 Revised MICHTOX and LMMBP Forcing Functions. . . . . . . . . . . . . . . . . .
Results and Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5.1 Confirmation of MICHTOX PCBs Bioaccumulation Predictions. . . . . . . . .
2.5.2 Comparison of Original MICHTOX PCBs Simulations to the
LMMBP Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Updating Parameterization for Henry's Constants. . . . . . . . . . . . . . . . . . .
Updating the Chemical Volatilization Rate Formulations. . . . . . . . . . . . . .
Long-Term HindcasVForecast Simulations. . . . . . . . . . . . . . . . . . . . . . . . .
Toxic Chemical Management Forecasts. . . . . . . . . . . . . . . . . . . . . . . . . . .
Uncertainty of MICHTOX Model Predictions. . . . . . . . . . . . . . . . . . . . . . . .
Are Lake Michigan Total PCBs Concentrations in Equilibrium With
Atmospheric Vapor Concentrations? . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sensitivity of Bioaccumulation Predictions to Initial Total PCBs
Concentrations in Fish. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sensitivity of Bioaccumulation Predictions to Food Chain Model
Parameterization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Are Total PCBs Bioaccumulation Factors Constant for Lake Trout
in Lake Michigan? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5.3
2.5.4
2.5.5
2.5.6
2.5.7
2.5.8
2.5.9
2.5.10
2.5.11
vii
84
84
85
85
86
86
90
94
98
98
104
109
110
110
129
131
131
132
132
136
136
-------
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
1.10
1.11
1.12
1.13
1.14
1.15
1.16
1.17
1.18
List of Figures
Long-term concentration trends for toxic chemicals in small Lake Michigan fish. . . . . . . . . . . .
7
MICHTOX mass balance schematic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9
Spatial segmentation for the 17 segment MICHTOX model. . . . . . . . . . . . . . . . . . . . . . . . . . . .
11
Results of chloride calibration of Green Bay dispersive exchange. . . . . . . . . . . . . . . . . . . . . . .
13
Suspended particle calibration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14
Organic carbon fraction of surface water particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
Partitioning model: Sensitivity of dissolved chemical fraction to non-settling organic
maner binding efficiency. . . . . . . . . . . . . . . . . . . . . . . '.' . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16
Calibration of partitioning model: Comparison to distribution coefficient data for
HOCs in the Great Lakes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
Sensitivity of computed volatilization rate to wind speed and temperature
(pentachlorobiphenyl) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
Plutonium deposition to Lake Michigan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 20
Total loading of lead to Lake Michigan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 20
PCBs loading time function and comparison to reported PCBs load estimates. . . . . . . . . . . .. 22
MICHTOX food chain structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ., 24
Chemical assimilation efficiency for Mysis calculated from Lake Ontario PCBs data. . . . . . . .. 26
Chemical assimilation efficiency for alewife calculated from Lake Ontario HOCs data. . . . . . .. 26
Chemical assimilation efficiency for Diporeia calculated from Lake Ontario PCBs data. . . . . .. 26
Steady-state spreadsheet model for pentachlorobiphenyl (PCBS) . . . . . . . . . . . . . . . . . . . . . .. 27
Simplified Lake Michigan lake trout food chain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
viii
-------
1.19
1.20
1.21
1.22
1.23
1.24
1.25
1.26
1.27
1.28
1.29
1.30
1.31
1.32
1.33
1.34
1.35
1.36
1.37
1.38
1.39
1.40
1.41
1.42
MICHTOX simulation of plutonium in southern Lake Michigan (epilimnion, hypolimnion,
overturn data) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
MICHTOX simulation of plutonium in southern Lake Michigan (epilimnion, hypolimnion,
and seasonal epilimnetic data) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
31
MICHTOX simulation of plutonium in southern Lake Michigan - sensitivity to vertical
segmentation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
32
Simulation of annual-averaged lead concentrations in southern Lake Michigan. . . . . . . . . . . . .
32
Lead simulation in southern Lake Michigan surficial sediment. . . . . . . . . . . . . . . . . . . . . . . . . .
32
Simulation of PCBs in southern Lake Michigan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
Simulation of PCBs in Green Bay. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
PCBs simulations in central Lake Michigan and outer Green Bay. . . . . . . . . . . . . . . . . . . . . . . .
34
PCBs simulations in Lake Michigan and outer Green Bay surficial sediments. . . . . . . . . . . . . .
34
PCBs simulations in Green Bay sediments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
34
Simulated PCBs distribution in southern Lake Michigan sediments. . . . . . . . . . . . . . . . . . . . . .
35
PCBs concentrations in Lake Michigan lake trout. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
Verification of PCBs accumulation in age seven lake trout. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
Sensitivity of trout PCBs predictions to the food chain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
Verification of PCBs accumulation predictions in lake trout age classes 2-9 ...............
37
Simulation of PCBs concentrations in age 2-12 trout, 1980-1989. . . . . . . . . . . . . . . . . . . . . . . .
37
Simulation of PCBs concentrations in age 2-12 trout, 1980-1990. . . . . . . . . . . . . . . . . . . . . . . .
37
Verification of PCBs concentration predictions for bloater. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
38
Validation of bioaccumulation predictions for lower trophic levels in Lake Michigan. . . . . . . . . .
39
MICHTOX predicted trout bioaccumulation and data from Lake Ontario. . . . . . . . . . . . . . . . . .
40
MICHTOX predicted trout biota-to-sediment ratio and data from Lake Ontario. . . . . . . . . . . . . .
41
Load-concentration relationship for PCBs in southern Lake Michigan. . . . . . . . . . . . . . . . . . . .
42
Relative magnitude of PCBs fate and transport fluxes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
43
PCBs air-water fluxes at steady-state in southern Lake Michigan. . . . . . . . . . . . . . . . . . . . . . .
44
ix
-------
1.43
1.44
1.45
1.46
1.47
1.48
1.49
1.50
1.51
1.52
1.53
1.54
1.55
1.56
1.57
1.58
1.59
1.60
1.61
1.62
1.63
1.64
1.65
S .t' .t f . 44
ensl IVI y 0 water concentrations to I
-------
1.66
1.67
1.68
1.69
1.70
1.71
1.72
1.73
1.74
1.75
1.76
1.77
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
Simulation 01 chlordane in Lake Michigan trout (Michigan Department 01 Natural
Resources 1990) . . . . . . . . . . . . . . . . . . . . . . . . . . .. 59
, . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . .
Simulation 01 DOT in Lake Michigan trout (DeVault et al., 1986; Michigan Department
01 Natural Resources, 1990). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 59
Simulation 01 dieldrin in Lake Michigan trout (DeVault et al., 1986; Michigan Department
of Natural Resources, 1990). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 59
Load cutoff simulation of PCBs in Lake Michigan trout. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 59
Tetrachlorodibenzo-p-dioxin simulation in Lake Michigan trout (U.S. Environmental
Protection Agency, 1989b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . " 60
Tetrachlorodibenzofuran simulation in Lake Michigan trout (U.S. Environmental
Protection Agency, 1989b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 60
Predicted effectiveness of PCBs load reductions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 61
Simulation of storm event in southern Lake Michigan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 62
Effect of storm event on PCBs in southern Lake Michigan sediment. . . . . . . . . . . . . . . . . . . .. 62
Sensitivity of PCBs concentrations in trout to thin (1.1 cm) surficial sediment layer
thickness. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 63
Predicted water PCBs concentrations for ten realizations of dynamic model. . . . . . . . . . . . . " 63
Predicted trout PCBs concentrations for ten realizations of dynamic model. . . . . . . . . . . . . . .. 64
MICHTOX mass balance schematic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 87
Spatial segmentation for the 17 segment MICHTOX model. . . . . . . . . . . . . . . . . . . . . . . . . . .. 88
The LMMBP estimates of total PCBs atmospheric vapor concentrations processed
as monthly values for MICHTOX Segments 1-3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 91
The LMMBP estimates of total PCBs atmospheric wet and dry deposition processed
as monthly values for MICHTOX Segment 1 (southern Lake Michigan) . . . . . . . . . . . . . . . . . .. 92
The LMMBP estimates of total PCBs tributary loading processed as monthly values
for MICHTOX Segment 1 (southern Lake Michigan) """""""""""""""'" 93
Long-term estimates of Lake Michigan total PCBs vapor concentrations. . . . . . . . . . . . . . . . .. 99
Long-term estimates of Lake Michigan total PCBs atmospheric deposition loading. . . . . . . . .. 100
Long-term estimates of Lake Michigan total PCBs tributary loading. . . . . . . . . . . . . . . . . . . . .. 101
xi
-------
2.9
2.10
2.11
2.12
2.13
2.14
2.15
2.16
2.17
2.18
2.19
2.20
2.21
2.22
2.23
2.24
2.25
2.26
2.27
Comparison of MICHTOX steady-state total PCBs concentrations to Saugatuck fish
data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ., 102
Comparison of MICHTOX steady-state total PCBs concentrations to Sheboygan Reef
fish data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 103
Original MICHTOX predictions and data for main lake total PCBs concentrations. . . . . . . . . .. 105
Original MICHTOX predictions and data for main lake sediment total PCBs
concentrations (sediment cores collected in 1991-1992) ........ . . . . . . . . . . . . . . . . . . . . .. 106
Original MICHTOX predictions of total PCBs concentrations in fish and comparison
to Sheboygan Reef zone data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 107
Original MICHTOX predictions of total PCBs concentrations in lake trout and
comparison to DeVault et al. (1986) data. . . . . . . . . . . . . . . . . . . . . .. ................... 108
Long-term Scenario A predictions of main lake total PCBs concentrations. . . . . . . . . . . . . . .. 111
MICHTOX southern Lake Michigan total PCBs predictions. Comparison of long-term
Scenario A to original model predictions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 112
Comparison of long-term Scenario A predictions to main lake sediment total PCBs
concentrations (sediment cores collected in 1991-1992) .............................. 113
Comparison of long-term Scenario A predictions to DeVault et al. (1986) lake trout data. . . . .. 114
Comparison of MICHTOX Scenario A total PCBs concentrations to Sheboygan Reef data. . .. 115
Long-term Scenario B predictions of main lake total PCBs concentrations. . . . . . . . . . . . . . .. 117
MICHTOX southern Lake Michigan total PCBs predictions. Comparison of long-term
Scenario B to original model predictions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 118
Comparison of long-term Scenario B predictions to the LMMBP deepwater dissolved
total PCBs concentrations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 119
Comparison of Scenario B predictions to main lake sediment total PCBs concentrations
(sediment cores collected in 1991-1992) .................... . . . . . . . . . . . . . . . . . . . . ., 120
Comparison of long-term Scenario B predictions to average total PCBs sediment
concentrations (LMMBP and GBMBP box core samples) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 121
Comparison of long-term Scenario B predictions to DeVault et al. (1986) lake trout
data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 122
Comparison of long-term Scenario B total PCBs concentrations to Sheboygan Reef
fish data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ., 123
Long-term Scenario C predictions of main lake total PCBs concentrations. . . . . . . . . . . . . . . .. 124
xii
-------
2.28
2.29
2.30
2.31
2.32
2.33
2.34
2.35
2.36
2.37
MICHTOX southern Lake Michigan total PCBs predictions. Comparison of long-term
Scenario C to original model predictions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 125
Comparison of Scenario C predictions to main lake sediment total PCBs concentrations
(sediment cores collected in 1991-1992) .......................................... 126
Comparison of long-term Scenario C predictions to DeVault et al. (1986) lake trout
data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 127
Comparison of MICHTOX Scenario C total PCBs concentrations to Sheboygan Reef
fish data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 128
Toxic chemical management alternatives. Comparison of forecast simulation for age
seven lake trout in southern Lake Michigan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 130
Total PCBs concentrations in Lake Michigan air and water. . . . . . . . . . . . . . . . . . . . . . . . . . .. 133
Ratio of total PCBs concentrations between dissolved water and vapor (MICHTOX
No-Action forecast) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 134
Sensitivity of food chain model to initial conditions. Age seven lake trout: No-Change
forecast. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 135
Sensitivity of MICHTOX steady-state PCBs concentrations to chemical assimilation
efficiency and comparison to Saugatuck biota zone data. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 137
Predicted bioaccumulation factors for total PCBs in age seven lake trout (No-Change
forecast simulation). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 138
xiii
-------
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
1.10
1.11
1.12
1.13
1.14
List 01 Tables
Sediment Segment Parameterization. . . . . . . . . . . . . . . . . . . . . . . . . . . .. ..............
13
Particle Flux Parameterization
14
. . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Selected Air Concentrations and Calculated Atmospheric Deposition Loadings for
Lake Michigan Priority Pollutants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
Octanol-Water Partition Coefficient (Kow) and Organic Carbon Partition Coefficient
(I
-------
1.15
1.16
2.1
2.2
2.3
2.4a
2.4b
2.4c
2.5
2.6
2.7
2.8
2.9
2.10
2.11
Summary of Results for Third Test of Model Uncertainty. Predicted Steady-State PCBs
Concentrations for Varying Tributary Loads and Expected Air Concentrations. . . . . . . . . . .. 54
Predicted Long-Term Chemical Loss Rates for Lake Michigan Critical Pollutants
Following Loading Reduction. First-Order Loss Rates in Units of 1/Year in Southern
Lake Michigan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
Cruise- and Segment-Specific Dissolved Fraction of Total PCBs Concentrations (ng/L) . . . . .
94
Cruise- and Segment-Specific Average Particulate Total PCBs Concentrations (ng/L) . . . . .. 94
Segment-Specific Average Surficial Sediment Total PCBs Concentrations (ng/g). . . . . . . . . .
95
Average Total PCBs Concentrations in Fish in the Saugatuck Biota Zone. . . . . . . . . . . . . . .. 95
Average Total PCBs Concentrations in Fish in the Sheboygan Reef Biota Zone. . . . . . . . . " 96
Average Total PCBs Concentrations in Fish in the Sturgeon Bay Biota Zone. . . . . . . . . . . .. 96
Comparison of Original MICHTOX Total PCBs Forcing Functions for 1994-1995 to
the LMMBP Estimates (Whole-Lake Average) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 104
Mass Balance Diagnostics for Total PCBs in the Original MICHTOX Simulation
(Year 1994-1995) [[[ 109
Comparison of Original and Revised Henry's Constant (atm m3/mol) Parameterization. . . .. 110
Monthly PCBs Volatilization Rates (mid) Calculated by Original and Revised
MICHTOX Formulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 111
Mass Balance Diagnostics for Total PCBs in MICHTOX Scenario B Simulation
(Year 1994-1995) """"""""""""""""""""""""""""" 127
MICHTOX Predictions of Total PCBs Concentrations (JIg/g) in Lake Trout for Toxic
Chemical Management Alternatives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 131
-------
AOCs
BAF
BaP
BCF
BMC
BSF
BSR
CV
DDD
DDE
DDT
EPRI
ERL
GBMBP
GLEC
GLNPO
HCB
HOCs
LaMP
LCL
LMMBP
LLRFRB
LLRS
InCV
MDNR
MED
NSOM
PCBs
PCB4
PCBS
PCDD
PCDF
QAPP
TCDD
TCDF
UCL
USEPA
USFWS
VWA
Abbreviations
Areas of Concern
Bioaccumulation factor
Benzo(a)pyrene
Bioconcentration factor
Bayesian Monte Carlo
Biota-to-sediment factor
Biota-to-sediment ratio
Coefficient of variance
Dichlorodiphenyldichloroethane
Dichlorodiphenyldichloroethylene
Dichlorodiphenyltrichloroethane
Electric Power Research Institute
Environmental Research Laboratory
Green Bay Mass Balance Project
Great Lakes Environmental Center
Great Lakes National Program Office
Hexachlorobenzene
Hydrophobic organic chemicals
Lake-wide Management Plan
Lower confidence limit
Lake Michigan Mass Balance Project
Large Lakes and Rivers Forecasting Research Branch
Large Lakes Research Station
Lognormal coefficient of variance
Michigan Department of Natural Resources
Mid-Continent Ecology Division
Non-settling organic matter
Polychlorinated biphenyls
Tetrachlorobiphenyl
Pentachlorobiphenyl
Pentachlorodibenzo-p-dioxin
Pentachlorodibenzofuran
Quality Assurance Project Plan
T etrachlorodibenzo-p-dioxin
T etrachlorodibenzofuran
Upper confidence limit
United States Environmental Protection Agency
United States Fish and Wildlife Service
Volume-weighted average
xvi
-------
Acknowledgments
Special thanks for the assistance and cooperation of the federal and on-site contractor staff at the U.S.
Environmental Protection Agency, Office of Research and Development, National Health and Environmental
Effects Research laboratory, Mid-Continent Ecology Division-Duluth, large lakes Research Station, Grosse
lie, Michigan. David Griesmer, Xiangsheng Xia, Katie Taunt, Xiaomi Zhang, and Xin Zhang provided raw and
processed lake Michigan Mass Balance Project and historical data used for the modeling. Wilson Melendez
provided programming support for the necessary modifications of MICHTOX mass balance, food web
bioaccumulation, and post-processing programs. Thanks to Kay Morrison for the graphic renditions and
figures and to Debra Caudill for formatting and word processing.
xvii
-------
Executive Summary
MICHTOX is a toxic chemical mass balance and food chain bioaccumulation model that was first developed
in the early 1990s. A Bayesian Monte Carlo uncertainty analysis demonstrated that MICHTOX predicted
polychlorinated biphenyl (PCBs) concentrations should be within a factor of two of the measured data. During
the early part of the Lake Michigan Mass Balance Project (LMMBP), MICHTOX was updated and used as a
preliminary assessment tool of the LMMBP PCBs data and to provide a screening-level analysis of the
potential future trends in total PCBs concentrations in Lake Michigan water, sediment, and fish under a variety
of contaminant load scenarios.
As reported in 1992, the model predicted the response of Lake Michigan, and with additional resolution, Green
Bay to atmospheric and tributary loadings. With its bioaccumulation component, chemical accumulation in
biota was predicted in response to the loadings. The model is capable of either dynamic or steady-state
simulations. Dynamic model predictions were used to predict the long-term rate of concentration decline
following load reduction for each toxic chemical. Significant reductions of PCBs in lake trout were predicted
for 2000 with no additional loading reductions. Additional reductions of PCBs concentrations could only be
achieved with significant reductions in atmospheric sources. These results were uncertain because PCBs
loading history is poorly defined and because of potential error in the parameterization of the surficial sediment
layer thickness. The thickness of this layer was demonstrated to be a critical factor in model uncertainty.
Additional factors leading to model uncertainty included uncertainty in initial concentrations and loading history
and dynamics of the Lake Michigan trophic structure.
As reported in 2002, MICHTOX was used to provide a preliminary mass balance modeling assessment of
PCBs in Lake Michigan. Because PCBs vapor concentrations from the LMMBP were significantly higher than
estimated in the original model, total PCBs forcing functions were recalculated using the LMMBP estimates.
Recommended changes to the model increased the volatilization mass transport rates, resulting in the PCBs
equilibrium shifting significantly towards the atmospheric vapor phase quicker than previously predicted. This
demonstrated that air-water fluxes predominated the transport pathways for PCBs in Lake Michigan. The best
prediction of PCBs concentrations in water, sediment, and fish were obtained with the forcing function peaking
in 1961-1963. This was different than the original model simulation reported in 1992. The model was used
to forecast total PCBs concentrations in lake trout for a variety of scenarios representing alternative strategies
for managing PCBs in Lake Michigan. Because of model uncertainty, observed average total PCBs
concentrations should be within a factor of two of predicted values. The bioaccumulation predictions were not
sensitive to initial conditions but were sensitive to model parameterization. The PCB predictions of this model
are historic and have been replaced by the predictions derived from the improved models used for the LMMBP.
xviii
-------
PART 1
1992 MICHTOX: A MASS BALANCE AND BIOACCUMULATION MODEL
FOR TOXIC CHEMICALS IN LAKE MICHIGAN
Douglas D. Endicott and William L. Richardson
U.S. Environmental Protection Agency
Office of Research and Development
Mid-Continent Ecology Division
and
Dean J. Kandt
ASci Corporation
Large Lakes Research Station
9311 Groh Road
Grosse lie, Michigan 48138
1.1 Executive Summary
A mass balance model has been developed for
critical pollutants (except mercury) in Lake Michigan.
The model predicts chemical concentrations in 17
water and sediment segments in response to
atmospheric and tributary loadings. It was designed
to predict chemical concentrations in the open waters
of Lake Michigan, with additional nearshore
resolution in Green Bay. The model was calibrated
using existing information to define water circulation
and the particle balance.
A bioaccumulation model has also been developed
for critical pollutants in Lake Michigan. The model
predicts chemical accumulation in lake trout and
bloater by pelagic and benthic food chains. The
bioaccumulation model is coupled to the mass
balance model for exposure to chemicals in water
and sediment in southern Lake Michigan. By this
coupling, chemical accumulation in biota is predicted
in response to loadings.
The model is capable of either dynamic (time
variable) or steady-state simulations. A simplified
steady-state model was developed for performing
sensitivity and uncertainty analysis. Chemical-
specific partitioning, fate, and bioaccumulation
processes were parameterized solely by
physicochemical properties including octanol-water
partition coefficient, vapor pressure, Henry's
constant, photolysis rate, and rate of metabolism in
fish. Atmospheric input of a chemical was based
upon the ambient air concentration of each chemical.
Model predictions were validated using data for
plutonium, lead, and polychlorinated biphenyls
(PCBs). Because MICHTOX was not calibrated
using toxic chemical data for Lake Michigan,
validation represents an independent test of the
model predictions as well as the assumptions,
simplifications, and aggregations that were used to
build the model. Plutonium data validated the particle
transport calibration. Seasonal stratification was
demonstrated to be an important process
determining long-term chemical dynamics. Toxic
chemical fate, transport, and bioaccumulation were
validated using PCBs data. A time-series of PCBs
loading to Lake Michigan was developed based upon
a review of estimates made by other researchers.
The model was run for the period 1940-1990 using
these loadings, and the resulting predictions were
compared to available data. Predictions of PCBs
water concentrations in southern Lake Michigan were
in good agreement with limited data. Southern Lake
Michigan sediment concentration and water and
sediment concentration predictions in Green Bay also
1
-------
agreed qualitatively with available data. Predictions
of lake trout PCBs concentrations did not exactly
match details of the concentration trends in the data.
However, the prediction was acceptable considering
the accuracy of the PCBs loadings and possible data
quality concerns. Predicted PCBs concentrations in
bloater were also in agreement with available data,
further validating the bioaccumulation predictions.
The results of model validation using PCBs were
generally better thcin expected considering the
preliminary nature of this model.
The steady-state model solution was used to
generate load-response predictions and mass fate
and transport fluxes, to examine model sensitivity,
and to estimate predictive uncertainty. The mass
balance for toxic chemicals was dominated by
internal fluxes associated with particles (settling and
resuspension) and particle burial, volatilization, and
photolysis. The relative magnitude of these fluxes,
and hence, the model sensitivity to parameterization,
were found to vary between chemicals and model
segments. Transport of the critical pollutants
between lake basins or from Green Bay to the main
lake was not significant. Volatilization and
redeposition is likely to be a much more significant
transport mechanism.
At steady-state, a linear relationship between total
,G~ ding and concentrations was predicted. Monte
Carlo analysis was used to estimate the uncertainty
in model predictions. In terms of the width of the
95% confidence interval, predictive uncertainty was
on the order of 10 times for concentrations in water,
and 30-300 times for concentrations in trout. The
confidence intervals "bracket" the load-concentration
predictions and define the expected bounds of model
error due to parameterization. The magnitude of
predictive uncertainty suggests that model results
may be qualitatively useful, for instance, in
comparing alternative simulations, but quantitative
use of the results would be inappropriate.
Uncertainty analysis was also used to identify
"critically" uncertain model parameters to be
prioritized for process research.
Several factors complicate the relationship between
loads and concentrations. First, reducing the load
from only one source category (tributaries, for
instance) will have a less-than-proportional effect
upon concentrations. PCBs concentrations, for
example, were demonstrated to be largely insensitive
to reduced tributary loads. Another factor was ~he
"lag time" in concentration response to. loading
reduction. For the critical pollutants In Lake
Michigan, lag time reflects the large in~entory of
chemicals in the sediments. The effectiveness of
loading reduction (even reductions to zero load) at a
given time is ultimately constrained by the system lag
time. Dynamic model predictions were used to
predict the long-term rate of concentra~lon decline
following loading reduction for each toxic chemical.
The model was used to predict the effect of
eliminating tributary and total PCBs loadings,
compared to a "No-Action" scenario of constant
PCBs loading after 1990. The results suggest that
significant reductions in trout PCBs concentrations
will occur in the next 10 years, even if no additional
loading reductions are made. Additional reductions
of PCBs concentrations in Lake Michigan will be
achieved only if significant reductions in atmospheric
sources are made. These results are, however,
uncertain because the p.::-- ~'s loading history is poorly
defined and because u potential errors in the
parameterization of the surficial sediment layer
thickness.
The model was also used to predict the potential
impact of a severe storm "event" on the
remobilization of toxic chemical from the lake
sediments. A two-day storm resuspending the entire
surficial sediment layer in southern Lake Michigan
resulted in elevated total water column PCBs for
almost a year. However, because most of the
resuspended PCBs remain sorbed to :Jarticles, the
effect of the storm upon biota conce' ,'::ltions was
negligible.
Finally, uncertainty in the dynamic model predictions
was evaluated. The thickness of the surficial
sediment layer, which determines the residence time
of particles and particle-associated contaminants in
the lake, was demonstrated to be a critical
parameter. Uncertainty in initial concentrations,
loading history, and dynamics of the Lake Michigan
trophic structure were considered as additional
factors leading to uncertainty in model predictions.
2
-------
1.2 Recommendations
1.2. 1 Verification of Model Predictions
A high priority should be placed upon generating data
of known quality and consistency for the purpose of
model verification. Further verification to
contemporary data is necessary to quantitatively
demonstrate the predictive ability of the model. Such
a demonstration is the fundamental test of a model's
adequacy as a predictive tool. Data should be
collected for all chemicals that are identified as the
highest priority toxics. (For critical pollutant
"mixtures," chemical-specific representatives of the
mixture should be quantified.) These data would
include representative measurements of the
following:
1.2.1.1 Air Concentrations/Deposition Fluxes
Air concentrations and deposition fluxes should be
measured, on at least a seasonal basis, over each
lake basin with additional measurements near likely
emission sources (urban/industrial areas). If based
upon shore station measurements, then methods for
over-lake extrapolation must be devised.
1.2.1.2 Surficial Sediment Concentrations
The distribution of chemical concentrations and
organic carbon in surficial sediment should be
characterized similar to the 1975 sediment survey
(Frank et al., 1979). This should be accompanied by
more limited sediment core sampling in focusing
zones to measure vertical distributions of
contaminants and particle tracers. Near-surface
distribution should be resolved on 1 cm or finer
intervals.
1.2.1.3 Lake Trout
Lake trout should be sampled from different lake
regions, including the nearshore and reef zones.
Chemical concentrations and lipid should be
measured in age seven male and female trout. It
would be preferable to analyze individual fish instead
of composites. Analysis of whole fish would be
preferred.
1.2.1.4 Major Tributaries
Chemical loads should be determined for all major
tributaries to Lake Michigan. Monitoring should be
conducted at a location on each tributary where lake
water inflow does not persist except during low flow.
Point sources below the monitoring location should
be sampled separately. Total chemical
concentration, particulate and dissolved organic
carbon, total suspended solids, chlorophyll, and
chloride should be measured based upon flow-
proportioned sampling. Flow, conductivity,
transmissivity, and temperature should be measured
daily (or hourly, for event responsive tributaries).
Chemical monitoring should be proceeded or at least
accompanied by evaluation of in-place sediment
contamination. In-place pollutants may be mobilized
only under extremely high flows; their contribution to
tributary loading is, therefore, unlikely to be detected
by conventional monitoring.
1.2.1.5 Water
Chemical concentrations (dissolved and particulate)
should be measured in the main lake basins.
Particulate and dissolved organic carbon, total
suspended solids, chlorophyll, and other standard
limnological parameters should also be measured.
Sampling should be conducted during both stratified
and unstratified periods, with mid-epilimnion,
metalimnion, mid-hypolimnion, and benthic nephloid
sampling during stratification. A true field blank for
dissolved and particulate chemicals must also be
obtained to validate these measurements.
Although the effort and cost associated with such an
undertaking would be substantial, it may be argued
that most of these data would be necessary to justify
additional priority toxics load reductions regardless of
modeling effort.
1.2.2 Extend Model to Other Critical
Pollutants and Target Organisms
It may be necessary to extend the modeling effort to
other critical pollutants. Mercury is one such
chemical that is not addressed by the present model.
Existing mercury models, most notably the Electric
Power Research Institute (EPRI) Mercury Cycling
Model (Hudson et al., 1991) lack the flexibility to
3
-------
simulate site-specific conditions for the Great Lakes.
A substantial process research effort will be
necessary before management-level simulation of
mercury mass balance, transformation, and
bioaccumulation can be made. Planning for such
efforts by organizations such as the Mid-Continent
Ecology Division (MED)-Duluth are underway; rapid
progress is not expected, given the many
fundamental unknowns regarding mercury's behavior
in the ecosystem. Extensive monitoring of loads and
ambient concentrations in Lake Michigan will be
necessary as well, as virtually no mercury data exist
for this system. This will require the development of
analytical capabilities that presently do not exist.
Thus, the process of developing models for other
critical pollutants may be lengthy.
It may also be necessary to extend the modeling
effort to other target organisms. These could include
a variety of birds and wildlife which consume fish
from the lake: mink, otter, heron, cormorants,
eagles, gulls, terns, and turtles. Although
toxicokinetics of hydrophobic organic chemicals
(HOCs) in herring gulls have been studied (Clark et
al., 1987), top predators other than fish have not
been incorporated in bioaccumulation models. This
would, again, necessitate a developmental effort.
1.2.3 Further Model Development
Further development of mass balance and
bioaccumulation models for priority toxics in Lake
Michigan may be justified for at least two reasons: to
improve the scientific credibility of the model and to
improve the accuracy and resolution of model
predictions. Because the development of the model
was based largely upon existing information,
numerous aspects of the structure and
parameterization of MICHTOX lack adequate
justification to establish scientific credibility.
Furthermore, a variety of assumptions which are
critical to model performance have not been
validated. To go beyond a management-level
application, which has limited acceptance and/or
utility, will require the resolution of these issues. In
part, this resolution may be achieved by obtaining the
proper calibration/verification data for toxic chemicals
in Lake Michigan. However, specific process
research to improve process descriptions and
parameterization will also be necessary. These
processes include:
. Chemical properties: octanol-water partition
coefficient, Henry's constant, photolysis rate
. Particle transport at the water-sediment
interface
. Chemical partitioning to organic carbon and
plankton
. Atmospheric deposition fluxes
. Chemical metabolism
. Chemical assimilation efficiency, particularly in
benthos
The second aspect of further model development is
improving the predictive ability of the model.
Predictive ability, in terms of both accuracy and
resolution of predictions, may be improved by
incorporating more fundamental and realistic process
descriptions and linkages in the simulation. This, in
turn, will allow for finer spatial and temporal
resolution of predictions. Specific areas for further
development include the following:
1.2.3.1 Circulation
In the WASP models, circulation is specified as
advective and dispersive transport functions. This
approach ha~ the disadvantages that calibration of
the transport function requires extensive tracer data,
circulation is not predicted by meteorologic forcing
functions, and (practically) the model loses resolution
because of the difficulty in measuring/calibrating fine.
scale transport variability. The alternative approach
is to predict circulation using hydrodynamic models,
which are based upon momentum and continuity
balances in two or preferably, three dimensions.
Hydrodynamic predictions are driven by meteorologic
forcing functions and provide fine-scale spatial and
temporal resolution. Basing circulation simulations
on hydrodynamics will be necessary for accurate
mass balance modeling in nearshore zones.
1.2.3.2 Sediment Transport
Particle transport is specified as velocity fields in the
W ASP models, and is based upon calibration. This
approach is descriptive rather than predictive. It
does not relate particle transport to actual forcing
4
-------
functions, and its accuracy and resolution are limited
by measurement availability. Sediment transport
models, which predict the settling and resuspension
of particles as functions of shear stress,
aggregation/disaggregation, and compaction in the
sediment bed, could be coupled to the mass balance
to overcome these limitations. As was the case for
hydrodynamics, this coupling would be particularly
important in shallow, nearshore zones where
sediment resuspension is highly episodic. Sediment
transport coupling would also be very useful in
predicting particle redistribution and focusing, which
produce complex contaminant distributions in Lake
Michigan sediments.
1.2.3.3 Organic Carbon Dynamics
Present toxic chemical mass balance models treat
suspended particles as fundamentally abiotic. Yet
most of the particulate matter in the Great Lakes
water column is phytoplankton, at least during the
growing season. Organic carbon, the principal
sorbent for HOCs, is therefore, cycled largely by
biotic processes - production, grazing, respiration,
and decay. Because phytoplankton and toxic
chemicals are states related through organic carbon,
their behavior is expected to be coupled. Building a
model with this coupling is a principal research
objective of the Green Bay Mass Balance Project
(GBMBP). Not only would this coupling improve
model realism and accuracy, but it would also allow
the model to simulate how nutrient control may
impact toxic chemicals in the ecosystem.
1.2.3.4 Food Chain Variability and Dynamics
The W ASTOX food chain model, which was adapted
for use in MICHTOX, predicts bioaccumulation for
"representative" (Le., average) organisms. This
population-based model has been validated in a
number of ecosystem/contaminant applications. Yet
it is unclear whether this model is capable of
describing the variability in bioaccumulation observed
for organisms in many data sets. Individual-based
bioaccumulation models have been proposed
(Hallam et al., 1990) as an alternative to the
population-based model, offering the advantage of
treating bioenergetic parameters and exposure
histories as functions of the individual organism.
Bioaccumulation predicted for many individuals then
defines the probability distribution for the population
or species, which in some instances (Le., risk
assessment) may be important to predict. This
approach to modeling may be particularly
advantageous at the point of modeling effects of
bioaccumulating chemicals, where a given body
burden may affect only a portion of the population.
Another weakness of the present bioaccumulation
model is the static structure of the food chain. This
limitation is discussed in Section 1.7.4. Just as toxic
chemical dynamics are coupled to those of
phytoplankton, there may be linkages to higher
trophic levels as well. In particular, as the structure
of the food chain changes, so may bioaccumulation
in higher trophic levels. Ecosystem models capable
of simulating the dynamics of trophic structure have
been developed and proposed (DePinto, 1990). Their
data requirements are extensive, however, and their
predictive ability in systems as large as the Great
Lakes is unknown. The linkage of bioaccumulation
and ecosystem models is probably the most
ambitious modeling recommendation, but it would
provide the ability to predict the consequences of
stresses such as fisheries management and exotic
species in terms of population diversity and
bioaccumulation.
1.2.4 Establish Linkages to Atmospheric
and Watershed Models
A final aspect of model development that would be
particularly useful for management, as well as
scientific applications, is the linkage of the water
quality model to mass balance simulations in the
atmosphere and watershed. This linkage is
necessary to relate actual sources of toxics to their
delivery to (and removal from) the lake, instead of
measuring this delivery as loads. The linkage to
atmospheric mass balance is particularly critical,
because it is not at all apparent that present
measurements of atmospheric concentrations and
deposition fluxes are free of the "boundary effects"
due to volatilizing chemicals. Furthermore, the
atmospheric transport and subsequent redeposition
of volatilizing chemicals can only be simulated by
coupling air and water mass balances.
5
-------
1.3 Introduction
1.3. 1 Project Objectives
This report describes the development and
application of MICHTOX, a toxic chemical mass
balance and bioaccumulation model for Lake
Michigan. This work was supported by the United
States Environmental Protection Agency (USEPA)
Region V, which requested the MED-Duluth/Large
Lakes and Rivers Forecasting Research Branch
(LLRFRB) to develop a mathematical model for Lake
Michigan. The primary objective of modeling was to
provide guidance to Region V and the Lake Michigan
Lake-wide Management Plan (LaMP) as to expected
water quality improvements in response to critical
pollutant loading reductions. A secondary objective
was to demonstrate the potential utility of the mass
balance modeling approach. The model addressed
two primary questions related to the LaMP.
1. For ~hemicals identified as critical pollutants,
what IS the relationship between loads from the
atmosphere, tributaries, point sources, and
ground water to concentrations in water
sediment, and biota in Lake Michigan? '
2. If the loads of these chemicals to Lake Michigan
were reduced, how rapidly would concentrations
change?
In response, a management-level model has been
developed which, within expected confidence limits
addresses these management questions. Th~
loading-concentration relationship was modeled for
11 critical pollutants:
Benzo(a)pyrene (BaP)
Chlordane (total chlordane and nonachlor isomers)
Total dichlorodiphenyltrichloroethane (DDT)
[p, p' -DDT, -dichlorodiphenyldichloroethylene
(DDE) and -dichlorodiphenyldichloroethane
(DDD)]
Dieldrin
Heptachlor epoxide
Hexachlorobenzene (HCB)
Lead
Total PCBs
2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD)
2,3,7,8-tetrachlorodibenzofuran (TCDF)
Toxaphene
The scope of this project was ambitious; indeed, no
~ater quality model intended to simulate as many
different tOXIC chemicals in such a large system had
been previously developed.
The model is based upon available theory and data
characterizing the sources, transport, fate, and
bioaccumulation of toxic chemicals in Lake Michigan
and the other Great Lakes. It builds upon 15 years
of Great Lakes modeling research sponsored and
conducted by the USEPA. The results of MICHTOX
should be considered preliminary because the model
is based upon a variety of assumptions which have
not been validated for either Lake Michigan or
individual toxic chemicals. In addition, significant
compromise was made regarding model calibration.
The customary application of mass balance models,
such as MICHTOX, includes extensive calibration of
model parameters to site-specific load, chemical
concentration, and process variable data. Calibration
is a step of model development necessary for
accurate parameterization and simulation. Because
only some of the necessary data for calibration were
available for toxic chemicals in Lake Michigan, this
step of r:nodel development could only be partially
acco~phshed. Consequently, it is necessary to
c~nslder the magnitude of uncertainty associated
Y"lth MICHTOX predictions, particularly if the model
IS to be useful to the LaMP. Past experience has
~emonstr~ted that quantifying predictive uncertainty
IS essential to credible model application for
management purposes. To this end, extensive
analysis of model uncertainty was applied to the
model. MICHTOX is intended to be a valid
repre~entation of current understanding of toxic
chemical transport, fate, and bioaccumulation
processes in the Lake Michigan ecosystem; however,
It also reflects the significant limitations of this
knowledge.
Besides addressing the two primary questions above,
the process of applying the model reveals further
research ~nd data needs. These include identifying
shortcomings and potential inconsistencies in the
dat~b~se. for toxic chemicals in Lake Michigan as well
as hmltatlons ~f the modeling approach due to poor
model resolution, uncertainty in model structure
process formulations, and parameterization. Such
6
-------
identification of research and data needs seNes as
a useful planning exercise prior to large-scale
project(s) designed to improve our understanding of
toxic chemical behavior in the ecosystem.
1.3.2 Lake Michigan Toxics Problem
The long-term trend of toxic chemical contamination
of Lake Michigan is well illustrated by merging two
United States Fish and Wildlife SeNice (USFWS)
data sets for chemical concentrations in small fish
(Neidermeyer and Hickey, 1976; Hesselberg et al.,
1990). The result shows an abrupt rise in PCBs and
DOT concentrations beginning around 1950 (Figure
1.1). Concentrations peak somewhere between
1960 and 1970, then decline almost as abruptly.
Similar trends throughout the Great Lakes have been
correlated to chemical production and usage rates
(Oliver et al., 1989; Eisenreich et al., 1989).
However, concentrations during the last 10 years
appear to be nearly constant at values elevated
above pre-contamination conditions. Dieldrin
90
80
o LPCB
1:1 LOOT
o dieldrin
70
-
:g
a.
.- 60
C)
-
~ 50
'-'
c
o
:;:: 40
ro
....
-
c
~ 30
c
o
u
A
1\
,
I
/~
/
/
/
20
10
o
1930
1940
1950
concentrations, on the other hand, are much lower
but have been steadily increasing since first detected
in fish collected around 1950. In general, monitoring
of toxic chemicals in Lake Michigan in the past
decade suggests little trend in concentrations. Prior
to this, concentrations of several toxic chemicals,
including chlordane, as well as PCBs and DOT,
declined dramatically as bans on chemical production
were initiated. What has changed is unclear: has the
lake attained equilibrium with continuing loads such
as atmospheric deposition, or does variability in the
data mask continuing slow declines in toxic chemical
concentrations? Resolving this issue is important
because further reductions in toxic chemical
concentrations appear necessary to protect human
health and the ecosystem. Yet, further reductions in
loading mayor may not be necessary to achieve
those concentration reductions. The approach to
modeling toxics is strongly related to the desired
scale and resolution of the analysis. The Lake
Michigan LaMP addresses water quality in the open
lake waters, defined to include all waters within the
1960
1970
1990
1980
Figure 1.1. Long-term concentration trends for toxic chemicals in small Lake Michigan fish.
7
-------
lake including all bays, harbors, and inlets. The
spatial scale of MICHTOX is on the order of the well-
mixed lake basins, with additional model resolution in
Green Bay. This specifically excludes the simulation
of toxic chemicals in other nearshore regions,
because MICHTOX is not predictive at the scale of
variability exhibited in the nearshore. Modeling water
quality in nearshore regions, such as bays, harbors,
and inlets, would require site-specific data collection
efforts and potentially different modeling techniques.
Of course, this excludes all except one of the 10
Areas of Concern (AOCs) around Lake Michigan.
These AOCs should be treated as source
components to the mass balance; this remains to be
accomplished. As sources, the AOCs may be highly
significant. Marti and Armstrong (1990) estimate that
half of the tributary loading of PCBs in the early
1980s was discharged from the Fox River.
According to Thomann and Kontaxis (1981), the
source of 50 to 90% of peak PCBs concentrations in
Lake Michigan may have been a single AOC:
Waukegan Harbor.
1.4 Model Description
1.4.1 Model Framework
The MICHTOX mass balance model was used to
predict chemical concentrations in the water and
surficial sediment in response to chemical loads to
Lake Michigan. The mass balance model was
adapted from the general water quality model
WA8P4 (Ambrose et al., 1988). The model
implements mass balance equations describing the
input, transport, and fate of hydrophobic toxic
chemicals in the Great Lakes. A schematic of the
mass balance model for a vertically-segmented lake
basin is presented in Figure 1.2. Chemical
concentrations in the epilimnion (cE), hypolimnion
(cH) and surficial sediment (cs) are calculated by
solving the coupled mass balance equations.
The mass balance in the epilimnion is:
d(VE CE)
dt
... accumulation of chemical mass
n
= I
1=;
WE,n
chemical loads
- EEH (CE - CH)
dispersive exchange with hypolimnion
- I EEB (CE - cs)
dispersive exchange with adjacent epilimnion
segments
- QES CE + QBE Cs
advective transport to and from adjacent
epilimnion segments
- vS,E A 'sE cE
settling from epilimnion to hypolimnion
+ v"H A 'sH cH
resuspension from hypolimnion to epilimnion
+ A (Wv G, + Vdry 'pA CA)
vapor exchange with atmosphere
- kp 'dE VE cE
photolysis
Variables introduced in this equation are defined as
follows:
Vs
Cs
WS.i
8
= volume of segment 8 [L 3]
=
total chemical concentration in
segment 8 [M/L3]
= incremental chemical loading to
segment 8 [M/T]
E
S1S2
= bulk dispersive exchange coefficient
between segments 81 and 82 [L3/T]
-------
Watershed
Atmosphere
Deposition
Volatilization!
Absorption
Epilimnion
sorbed
chemical
Transport and
Exchange
bound
chemical
NSOM
~ ~
" /
',- Partitioning /'
-....... -.... -............ -........
.................. --................
Hypolimnion
P~C
Resuspension
Settling
bound
chemical
Bioconcentration
:ft
NSOM
dissolved
chemical
sorbed
chemical
Exchange
------.-.---------
-------.------.--.
P~C
Diffusion
Resuspension
Settling
Surficial
Sediment
Deep
Sediment
Figure 1.2. MICHTOX mass balance schematic.
Q
Sl S2
=
flow from segments 8, to 82 [L3/T]
Wv
=
volumetric washout ratio
fds' fss, fbS =
dissolved, sorbed (to particles), and
bound (to non-settling organic
matter) chemical fractions in
segment S
Gr
=
rainfall [UT]
Vdry
=
dry deposition velocity [LIT]
vs.s = particle settling velocity from
segments S [Uf]
kv
vr.s = particle resuspension velocity from
segment S [Uf] H'
A = surface area [L2] kp
cA = total chemical concentration in air
[M/L3]
fvA' fpA
=
vapor and particulate chemical
fractions in air
=
volatilization rate [LIT]
=
dimensionless Henry's constant
=
photolysis rate [UT]
9
-------
The mass balance in the hypolimnion is:
d(VH CH)
dt
... accumulation of chemical mass
- EEH (CE - CH)
dispersive exchange with epilimnion
- I EHB (CH - CB)
dispersive exchange with adjacent
hypolimnetic segments
+ vs,E A 'sE cE
settling from epilimnion to hypolimnion
- vs,H A 'sH cH
settling from hypolimnion to surficial
sediment
- v"H A 'sH CH
resuspension from hypolimnion to epilimnion
+ v"s A 'ss Cs
resuspension from surficial sediment to
hypolimnion
Cs
+ K,Ad ['dS + 'bS - - ('dH + 'bH) cHI
ns
sediment-water diffusion
Additional variables introduced in this equation are:
K,
= diffusive exchange coefficient [Lfll
= sediment deposition area [L2]
Ad
ns
= surficial sediment porosity
The simplification of the mass balance equation for
a vertically-integrated (Le., unstratified) water column
has been previously reported (Oi Toro, 1987). The
mass balance in the surficial sediment is:
dcS
Vs-
dt
... accumulation of chemical mass
= vs,H A 'sH CH
settling from hypolimnion to surficial sediment
- v"s A 'ss Cs
resuspension from surficial sediment to
hypolimnion
- K, Ad [('dS + 'bS) Cs - ('dH + 'bH) CH]
ns
sediment-water diffusion
- vb Ad 'sS Cs
burial to deeper sediment layers
The only additional variable introduced in the surficial
sediment mass balance is Vb' the sediment burial (or
sedimentation) velocity [LIT]. A single surficial
sediment layer has been used in MICHTOX; this is
assumed to be adequate to simulate sediment-water
chemical exchange processes. Vertical resolution of
chemical concentrations in deeper sediment layers
may be obtained from the model by transforming the
time scale to sediment depth using the burial
(sedimentation) rate.
1.4.2 Segmentation
Lake Michigan is spatially divided into 17 segments
in the MICHTOX model, as depicted in Figure 1.3.
This segmentation represents an intermediate level
of spatial resolution, which balances the desire to
predict concentration response to spatially non.
uniform loads against a lack of data to either
implement or validate a greater level of resolution.
The main lake is divided into southern, central, and
northern basins, according to the large-scale
circulation and sedimentation patterns observed in
Lake Michigan. The circulation of the southern and
central basins of the lake are characterized by
distinct counter-rotating gyres (Schwab, 1983). The
shallow northern basin has little depositional
sediment and no apparent large-scale circulation, but
10
-------
\
Straits of
Mackinac
Us seasonally-stratified
L.i!LJ water column
rriI completely-mixed
L..:.:...J water column
~ surficial sediment
IR""'I grey scale intensity
.~N Indicates lake depth
Lake
Michigan
~
,~
Chicago
Diversion
Figure 1.3. Spatial segmentation for the 17 segment MICHTOX model.
11
-------
undergoes considerable exchange with northern
Lake Huron water across the Straits of Mackinac.
Each main lake basin is segmented vertically into
epilimnion, hypolimnion, and surficial sediment.
Green Bay is divided into three completely-mixed
water column and sediment segments. The lower
Fox River is represented as a final water/sediment
segment pair in the model. Only in these segments
does MICHTOX represent nearshore-to-open water
gradients. Additional model resolution is provided in
Green Bay because (1) significant past and present
PCBs loading from the Fox River has resulted in
persistent water and sediment concentration
gradients, and (2) results here may be compared to
those of the calibrated models under development for
the GBMBP. In fact, the MICHTOX Green Bay
segments are a superset of the segments chosen for
the GBMBP model. Segment geometry (volumes,
interfacial and surface areas, and depths) were
based upon the 2 km digitized bathymetry data of
Schwab and Sellers (1980).
1.4.3 Circulation
Lake circulation in MICHTOX is specified as inflows
from tributaries, flows and dispersive exchanges
between the water column segments, and outflow
and exchange across the Straits of Mackinac. Flows
were based upon the whole-lake water balance of
Quinn (1977), which provided monthly average
changes in storage, tributary flow, outflow and
diversion, precipitation, and evaporation. Tributary
flow was apportioned to the various surface water
segments according to the river mouth location and
mean flow of the 14 largest Lake Michigan tributaries.
Changes in storage, precipitation, and evaporation
were apportioned based upon segment surface area.
Based upon this water balance, flows between all
surface water segments were calculated for each
month. Annual hydraulic residence times
(volume/outflow) for the main lake segments range
from 110 years in the southern basin to seven years
in the north. In comparison, the hydraulic residence
time for the lake as a whole is 100 years
(Winchester, 1969). Hydraulic residence times in
Green Bay segments are much shorter: from 0.4
years in the inner bay to four years in the outer bay.
Vertical exchange coefficients, which quantify the
extent of mixing between epilimnetic and
hypolimnetic segments in the main lake were taken
from the Lake Michigan WASP eutrophication model
MICH1 (Rodgers and Salisbury, 1981). These
coefficients vary seasonally from minimum values of
essentially zero during stratification (from
approximately May to October) to maximum values of
15 and 40 cm2/s (southern and northern lake basins,
respectively) during unstratified periods. While
vertical exchange was curtailed during the summer,
entrainment of hypolimnetic water due to the
deepening of the thermocline was simulated.
Entrainment "mixes" hypolimnetic water into the
epilimnion; however, the mass balance of the
hypolimnion is not affected. The thermocline in
southern and central lake basins is simulated to
deepen from an initial depth of 10m at the onset of
stratification to a maximum depth of 50 m before
overturn, based upon data from Robbins and Eadie
(1991). In the northern basin, the thermocline was
simulated to deepen from 1 0 to 15 m based upon
temperature profile data of Ayers et al. (1958).
Horizontal exchange is the mixing of water from
adjacent segments due to fluctuations in flow in
response to surface shear stress from st'" "'-~s and to
rapid surface heating and cooling. rizontal
exchange coefficients in Green Bay we' dbrated
to reproduce observed chloride grac The
calibration achieved using 1982 chlorid~ ,': (Auer,
1989) is shown in Figure 1.4. Thf'.rizontal
exchange between Green Bay and La' ;lichigan
was verified by comparison to the bi-dire' ')al flows
measured in the bay-lake passages in 1 \:1 Eadie et
al., 1991) and 1989 (Gottleib et al., 1990). The
calibrated exchange between bay and lake reaches
a maximum during July of 7800 m3/s, som" ') times
greater than the combined tributary inflo~!e bay.
Horizontal exchange coefficients were m0 a difficult
to define in the maw lake, where no similar "tracer"
gradients were observed. Values of 100 and 1000
cm2/s were used in MICH1 for hypolimnetic and
epilimnetic horizontal exchange, respectively. These
values are similar to the exchange coefficients used
by Thomann et al. (1979) for Lake Ontario, but they
are considerably smaller than the 1000 m2/s
suggested by Prospero (1978) for horizontal scales
of 100 km. Current velocities measured in Lake
Mic.higan (Mortimer, 1971) also suggest that larger
hOrizontal exchange coefficients may be appropriate.
Perhaps recognizing this, MICH1 was calibrated with
12
-------
30
. data
o model
;J
.[ 20
c:
o
~
c:
!'!
8
..
:g 10
o
E
u
o
sag 5
seg 6
seg 7
Figure 1.4. Results 01 chloride calibration 01
Green Bay dispersive exchange.
a large epilimnetic bi-directional flow. This flow,
about eight times the outflow, was converted to an
equivalent exchange of 1200 m2/s in MICHTOX.
Hypolimnetic horizontal exchange was increased to
40 m2/s in MICHTOX, equivalent to outflow.
Exchange across the Straits of Mackinac was defined
according to Quinn (1977), who estimated that bi-
directional flow during stratification approximately
doubled the outflow of Lake Michigan water. More
recent measurements (Quinn, personal
communication) indicate that the exchange may be
considerably greater. Consideration of the bi-
directional Straits of Mackinac flow, which produces
a maximal outflow component of seven times the
Table 1.1. Sediment Segment Parameterization
average discharge, reduces the northern Lake
Michigan hydraulic residence time from seven years
to less than four years.
1.4.4 Solids Balance
A solids mass balance was constructed for
MICHTOX to simulate the particle transport fluxes of
settling, resuspension, and burial in Lake Michigan.
These appear as particle velocities in the toxic
chemical mass balance equations. Although Great
Lakes sediments represent a diverse particle
assemblage, a single solid class representing fine-
grained sediments was simulated in MICHTOX.
Thus, no distinction between biotic and abiotic
particles nor representation of particle aggregation
were considered in this model. Sediment focusing
was simulated by defining the area of each surficial
sediment segment according to the extent of
deposition zones in that portion of the lake.
The calibration of the solids balance was achieved by
varying solids loading and resuspension velocities to
match data for monthly average suspended particle
concentrations M [MIL 3] in the surface water
segments. Surficial sediment burial rates, thickness,
and depositional fractions were based upon
measured values listed, along with their sources, in
Table 1.1. A constant settling velocity of 1.5 mId was
assumed, and sediment resuspension fluxes were
constrained to maintain a constant sediment particle
concentration of 240 kg/m3. Finally, the epilimnion
Sediment Depositional Burial Velocity Mixed Layer Mixed Layer Residence
Segment Fraction (mm/y) Thickness (cm) Time (y)
11 0.68a 1.7c 3.3c 20
12 0.59a 1.7c 3.3c 20
13 0.2a 0.63d 3.3c 53
14 0.32 20 10 5
15 0.5b 0.25b 9b 350
16 0.74b 0.25b 4b 160
17 0.31b 0.25b 4b 160
Sources: aCahill, 1981; bEdgington, 1991; cRobbins and Eadie, 1991; dRobbins and Edgington, 1975
13
-------
was isolated from resuspended particles during
stratification. The resulting particle loads and
resuspension fluxes, expressed as annual averages,
are presented in Table 1.2. The resuspension fluxes
vary inversely with water column depth, as expected
for a process responding predominantly to wave
action. The suspended particle calibration, for all
water column segments, is displayed in Figure 1.5.
Surface water suspended solids concentrations for
Lake Michigan were reported by Robbins and Eadie
(1991); concentrations in Green Bay were based
upon the GBMBP cruise data (U.S. Environmental
Protection Agency, 1989a). The "build-up" of
suspended particle concentrations in the hypolimnion
during stratification is consistent with the observed
development of a nephloid layer near the lake bottom
(Eadie et al., 1983).
Table 1.2. Particle Flux Parameterization
Segment
1
2
3
4
5
6
7
Resuspension
Flux (kg/m2/d)
2.2
1.7
5
38
13
3.6
2.2
Solids
Load
(kg/d)
2.5e7
2.2e7
4.2e6
4.3e5
-1.ge5
1.3e5
4.4e5
Depth
75
55
36
2.2
5.9
13
16
Because the surficial sediment is treated as
completely-mixed in the mass balance, its volume Vs
is properly defined by the thickness of the mixed
sediment layer and the segment depositional area.
The mixing of the surficial sediment is primarily the
result of bioturbation. Dividing the mixed sediment
layer thickness by the burial velocity determines the
residence time of particles, and presumably of
particulate chemical, in this well-mixed layer. The
sediment segment also represents the reservoir of
particles and particulate chemical available for
resuspension. The sediment residence time controls
the accumulation rate in the sediment mass balance,
and resuspension of particulate chemical ties the
long-term water column accumulation to this rate as
~ 5
g
.~ 4
~
c
2! 3
c:
o
<)
~ 2
1:
'"
0.
::J'
C.
g 15
c:
.2
~
C
Q)
g 10
o
<)
Q)
u
1:
~ 5
6
main lake
- epilimnion (seg.1)
o epilimnion data
- - - - - hypolimnion (seg.8)
,
-----'
, -
,
,
,
,
, ,
. ,
.' ,
, '.
,
I
I
I
,
I
,
I
I
,
I
,
o
o
240
300
360
120
180
Juhan day
60
20
Green Bay
IJ Inner GB data
- inner GB (seg.5)
~ mid GB data
...... mid GB data (seg.6)
. outer GB data
- - outer GB data (seg.7)
o
o
60
120
180
julian day
240
300
360
Figure 1.5. Suspended particle calibration.
well. Based upon the observed decline of particle-
reactive radioisotopes in the Great Lakes, the mixed-
layer residence time is approximately 20 years. The
main lake surficial sediment thickness, 3.3 em, was
based upon this residence time. Radioisotope
distribution in sediment cores from Lake Michigan
suggest a mixed-layer thickness ranging from 1 to 2
cm based upon lead-210 (Edgington and Robbins,
197~) to 4 cm based upon cesium-137 (Robbins and
Ed~lngton, 1975~. Extensive sampling of Green Bay
s~dlments (Edgington, 1991) provide mixed-layer
thicknesses of 9 ?~ (inner bay) and 4 cm (mid- and
outer-ba~). Surfl~lal sediment mixed-layer depths
and re~lden~e times used in the model are
summanzed In Table 1.1.
14
-------
The model for HOC partitioning described below
requires specification of the organic carbon fraction
(foe) of particles, because organic carbon is
considered to be the active sorbent for these
chemicals. Based upon the data of Robbins and
Eadie (1991), foe for particles in surface water was
specified monthly (Figure 1.6) after correction to
remove CaC03. Particle foe in the hypolimnion was
specified as 30% of epilimnion values, based upon
limited data from the same source. Considerable
decomposition of particulate organic carbon occurs
in the hypolimnion of Lake Michigan (Eadie et a/.,
1984). The foe for surficial sediment particles was
assumed to be 3%; this value was based upon
extensive surficial sediment characterization in Lake
Ontario (Thomas et a/., 1972). .
0.4
0.3
u 0.2
.£!
0.1
0.0
f m a m
o n
d
a
5
month
Figure 1.6. Organic carbon fraction of surface
water particles.
1.4.5 Chemical Partitioning and Loss
1.4.5.1 Partitioning
Partitioning, which defines the distribution of
chemical between dissolved and sorbent
compartments, is a process of fundamental
importance in determining the transport and fate of
hydrophobic toxic chemicals. Model sorbent
compartments for HOCs include the organic fraction
of sediment particles and non-settling organic matter
(NSOM), also referred to as colloidal organic carbon.
Partitioning between phases is treated as a linear,
reversible, and rapidly equilibrating process.
Partitioning is represented in the mass balance
formulations as fractions (fd: dissolved; fs: particle-
sorbed; fb: NSOM-bound) of the total chemical in
each phase. Examination of the mass balance
equations reveals that partitioning fractions appear
throughout the model. Partitioning affects nearly all
other processes in the mass balance model by (1)
defining chemical fractions transported by particles,
(2) defining the dissolved chemical fraction subject to
volatilization, photolysis, and available for direct
uptake by biota; and (3) defining mobile chemical
fractions in sediment pore water.
Chemical fractions are determined by organic carbon
sorbent concentrations and partition coefficients
defining equilibrium chemical distribution between
phases. The three-phase model developed to
describe PCBs partitioning (Baker et a/., 1986) was
simplified and used in conjunction with a correlation
relating the particulate organic carbon (POC) partition
coefficient Koe to the octanol-water partition
coefficient Kow (Eadie et a/., 1990):
log Koe = 1.94 + 0.72 log Kow
The partition coefficient Kp is calculated from Koe by:
Kp = r/cd = foe Koe
where:
r = fsjM = sorbed (particulate) concentration of
chemical [Mehe,,/Msed]
Cd = fd C = dissolved concentration of chemical
[Mehe,,/L 3]
foe
= fraction organic carbon of particles
The three-phase model predicts that the observed
distribution coefficient Kd for HOCs will be lower than
the partition coefficient Kp due to the influence of
NSOM binding:
r
Kd = = Kp/(1 + fe MKp)
cd + Ch
This formulation produces results comparable to the
"particle effect" model of Oi Toro (1985). The foe is
an empirical parameter relating the efficiency of
NSOM binding to that of particle sorption. Figure
1.7 displays how the dissolved chemical fraction
15
-------
10
c
£!
ti
g
iii
u
'e
'"
~
u
"-
........., '. ""'... .~.
"
"
"
"
"'
\
\
"'
"'
"'
"'
"
"
"
"
"
".
'.
"""'-.-.
M=2 mg/L
foc=0.25
06
08
- fc=0.5
....--.-- fe=1.0
-.-.- fc=50
~
2
o
..
..
'5
.-...............--.-.-
04
02
4
7
8
5
6
IogKow
Figure 1.7. Partitioning model: Sensitivity of
dissolved chemical fraction to non-settling
organic matter binding efficiency.
varies according to Kow for different values of fc. The
larger the value of fc' the greater will be the reduction
in dissolved fraction as hydrophobicity increases; if fc
equals one, a chemical will partition equally between
particles and NSOM. This model has been calibrated
to several sets of water column partitioning data from
the Great Lakes; the results are shown in Figure 1.8.
A fc of 0.5 was used in MICHTOX. A similar
approach was used to model partitioning in the
surficial sediment. There, however, a NSOM binding
efficiency based upon organic carbon was
parameterized, based upon the data of Capel and
Eisenreich (1990).
Lead and plutonium partitioning was defined for
dissolved and particulate phases only. Data for lead-
210 in Lake Michigan (Van Hoof and Andren, 1989)
suggested a lead 10gKp of 6.3. For plutonium, Kp
was parameterized to reproduce a dissolved water
column fraction of 80%, as reported for Great Lakes
waters by Alberts and Wahlgren (1981). The particle
effect model (Di Toro, 1985) was used to predict the
expected decline in lead and plutonium partition
coefficients with increasing suspended particle
concentration.
1.4.5.2 Volatilization
Chemical exchange between air and water occurs by
rainfall washout, dry deposition, absorption, and
volatilization. Washout and deposition will be
considered as loading terms to MICHTOX and are
described later in this report. Absorption and
volatilization may be combined as an expression for
net volatile exchange:
kv('vA cAIHI - 'dE CE}
the product of a volatilization rate kv [UT] and the
gradient between atmospheric (feA ciH') and water
column (fdE cE) chemical concentrations. Depending
upon the direction of this gradient, net volatilization
may represent either a source or sink of chemical.
Applying the two-film theory (Whitman, 1923),
volatilization rate becomes a function of serial mass
transfer resistances in liquid and gas films at the
air-water interface, with the overall rate constant
given as:
k =
v
1
1
- +
K1
1
K HI
g
where:
K1
= the liquid film mass transfer coefficient
[LIT]
Kg
= gas film mass transfer coefficient [LIT]
H'
= dimensionless Henry's constant defining
chemical equilibrium between vapor and
dissolved phases.
Henry's constant for each toxic chemical was either
based upon direct measurement or, more often,
calculated from solubility and vapor pressure data.
Liquid and gas film transfer coefficients were
extrapolated from reaeration and evaporation rates,
which can be reliably estimated from correlations with
environmental factors. The correlations of O'Connor
(1983) and Liss (1973) were used in MICHTOX.
Details of the volatilization rate computation are
provided in Endicott et al. (1990). The model
incorporated the effects of both spatial and temporal
variation of water and air temperature, wind speed,
and ice cover upon volatilization rate. The simulated
variability of ~ with temperature and wind speed for
pe~tac~loroblp~~nyl (PCBS) is plotted in Figure 1.9.
kv IS fairly sensitive to both of these environmental
factors. The higher August water temperature
16
-------
8
7
7
o
6
o
"0 6
~
CJ)
o 5
"0
~ 5
CJ)
o
4
4
o data: average log Kd :f: 1 sd
(10/89 GB stations 24-27)
3
3
3
3
4
6 7
log Kow
Outer Green Bay
5
8
9
o PCB homo logs
. other HOCs
5 6
log Kow
Lake Ontario
(Niimi & Oliver, 1983;
Oliver et aI., 1989)
4
7
8
6 7
6
5
"0 "0 5
~ ~
CJ) CJ)
0 0 4
4
, 0 PCB homologs
3 . other HOCs
! (data :t 1 standard deviation)
3 2
0 5.0 5.5 6.0 6.5 7.0 7.5 3 4 5 6 7 8
log Kow log Kow
Lake Superior Connecting Channels
(Baker et aI., 1986) (Oliver, 1987)
Figure 1.8. Calibration of partitioning model: Comparison to distribution coefficient data for HOCs
in the Great Lakes.
17
-------
25
- February kv (H=9 67E~)
- - - -. August kv (H=4 85E-3)
2.0
1 5
:c
1
>
x
.
.
.
.
.
.
,
.
.
.
.
10
0.5
0.0
o
3
15
18
6 9
wind speed (m/s)
12
Figure 1.9. Sensitivity of computed volatilization
rate to wind speed and temperature
(pentachlorobiphenyl).
increases Henry's constant as well as the sensitivity
of kv to wind speed. Monthly average surface w~ter
temperature, over-lake air temperature, and wind
speed were taken from Quinn (1977). Segment-
specific ice cover was estimated from maps in the
Great Lakes Ice Atlas (Assel et al., 1983).
1.4.5.3 Photolysis
Photolysis, a chemical reaction caused by the
absorbance of light, was the only transformation
process included in MICHTOX. Seasonal photolysis
rate constants were calculated by the method of
Zepp and Cline (1977) for BaP, HCB, and TCDD,
while rates found in the literature were used for DOE
and dieldrin. Photolysis was assumed to be
insignificant for the other toxic chemicals exc~pt
TCDF, based upon limited data (Mabey and Smith,
1982). No information regarding the photolysis of
TCDF or other furans could be found in the literature;
again, photolysis was assumed to be negligible.
1.4.5.4 Sediment-Water Diffusive Exchange
The final chemical transport process considered in
MICHTOX is pore water diffusion between the
surficial sediment and overlying water column. The
diffusion process is considered to be a minor
component of chemical sediment-water exchange,
although data to confirm this are lackin~. The
diffusive exchange coefficient, Kf' is often e~t~mated
by the chemical free liquid diffusivity modified f~r
pathlength tortuosity (Eisenreich et al., 1989). This
produces a Kf of 0.1 to 1 cm/d, a range of values also
suggested by Thomann and Mueller (1987). A I
-------
The aerosol partition coefficient is, in turn, calculated
from the chemical's liquid sub-cooled vapor pressure
(psd:
K - 6.0e + 6
pA - S
PL
Deposition parameterization followed values
suggested as appropriate for the Great Lakes by
Mackay (1989).
Air concentrations and deposition fluxes were treated
as spatially uniform and (except for PCBs) constant
in MICHTOX. The apparent spatial and temporal
variability reported for both air concentrations and
deposition processes suggests that this may be a
poor assumption (Eisenreich et al., 1981; Hoff et al.,
1992). Coupling MICHTOX to simulations of
atmospheric chemical transport may be particularly
valuable as a means to improve the realism and
accuracy of this aspect of the model. For example,
coupling water and air toxics models would allow
simulation of the migration of PCBs from Green Bay
to Lake Michigan or elsewhere in the Great Lakes via
air transport.
Expected air concentrations and atmospheric
deposition loads are presented in Table 1.3. These
values and estimates of their uncertainty in terms of
the lognormal coefficient of variation (InCV), were
selected based upon review of the literature and air
monitoring data from the Michigan Department of
Natural Resources (MDNR) (Moon, personal
communication). TCDD concentrations have not
been reported for ambient air; the expected value is
one-third of the detection limit reported by Smith et
al. (1990). It should be noted that the atmospheric
deposition loads in Table 1.3 do not include
absorption, which for some chemicals represents a
large flux to the lake.
1.4.6.2 Tributary Loads
Tributaries convey toxic chemicals to the lake from a
variety of sources including runoff, in-place
pollutants, point source discharges, and ground
water inflow. Present methods of estimating tributary
loads rely upon frequent monitoring of flow and
concentrations near the tributary discharge, an
expensive and logistically-complicated effort.
Tributary loading, particularly of hydrophobic
chemicals including PCBs and lead, appears to occur
predominantly during flood events. Such events
must be sampled in order to accurately estimate in-
place pollutants from tributaries into the lake; the
likelihood of monitoring such events is, however,
loads. Extreme events (such as a 50- or 100-year
flood) could potentially transport huge quantities of
extremely small. The data collection necessary to
make reliable estimates of tributary loading of toxic
chemicals to Lake Michigan has begun only recently
Table 1.3. Selected Air Concentrations and Calculated Atmospheric Deposition Loadings for Lake
Michigan Priority Pollutants
Atmospheric Deposition
Chemical cA (ng/m3) InCV Source Loading (kg/y)
BaP S.Oe-3 0.41 Baker and Eisenreich, 1990 67
Chlordane 0.039 0.64 Hoff et al., 1992 47
DDT 0.030 0.61 Eisenreich et al., 1981 220
Dieldrin 0.032 0.64 Eisenreich et al., 1981 210
Heptachlor epoxide 0.016 0.64 Hoff et al., 1992 2.3
HCB 0.063 0.64 Hoff et al., 1992 4.4
PCB4 0.12 0.18 Hoff et al., 1992 83
PCBS 0.12 0.18 Hoff et al., 1992 340
TCDD 3.2e-S 0.64 Smith et al., 1990 0.38
TCDF 3.4e-4 0.64 Smith et al., 1990 3.6
Toxaphene 0.18 1.7 Rice et al., 1986 860
19
-------
with the GBMBP. Approximately 300 samples were
collected over a 17 -month period to estimate loading
of PCBs and lead from five tributaries to Green Bay.
1.4.6.3 Loading Histories
Because toxic chemical load estimation has not been
a part of water quality surveillance efforts in Lake
Michigan. the time history of loadings for only two
toxic chemicals, plutonium and lead, could be reliably
estimated. A more speculative loading history for
PCBs was also estimated. based upon limited
information.
1.4.6.3.1 Plutonium
Radioactive plutonium-239/240. a product of
atmospheric bomb testing. has been monitored in
the Great Lakes since 1970. A remarkable feature of
plutonium is that its loading to the Great Lakes via
atmospheric deposition is well-known (Robbins.
1985) due to measurements made at the Argonne
National Laboratory. Additionally, the extent of
plutonium partitioning to fine-grained sediments is
similar to that for other hydrophobic toxic chemicals,
and the only significant loss process for plutonium is
sediment burial (radioactive decay may be neglected
for the time scale of interest). These factors make
plutonium an excellent state variable for calibration of
MICHTOX. Robbins' plutonium deposition flux
history is plotted in Figure 1.10. Plutonium
deposition peaked in 1958-1959 and again in 1962-
1964; values since 1980 have remained essentially
zero.
1.4.6.3.2 Lead
Lead has also entered the Great Lakes largely by
atmospheric deposition. Regional lead deposition
fluxes back to the 19th century were reconstructed
from sediment records and coupled to recent
atmospheric measurements (Edgington and
Robbins. 1976). Tributary and other non-
atmospheric sources were neglected in the
estimation of lead loading. The resulting lead
deposition history. converted to total Lake Michigan
load and updated to the mid-1980s (Robbins.
personal communication). is plotted in Figure 1.11.
The decline in lead loading after 1970 coincides with
the introduction of unleaded gasoline.
1.0
0.9
~ 0.8
N
E
o!!:
U 0.8
.s
g 0.6
..
'in
&. 0.5
CD
"C
E 0.4
'"
'2
.9 0.3
'"
Q.
Figure 1.10.
Michigan.
20e+6
1.6e+6
~
l. 1.2e+6
"C
-------
Mackay's estimates for Lake Ontario are unique for
they provide a continuous history of atmospheric,
tributary, and point source loadings for PCBs. To
adapt these estimates for use in MICHTOX, the
magnitude of tributary loads was scaled to match the
650 kg/y estimated by Marti and Armstrong (1990)
for Lake Michigan tributaries in 1980-1983.
Atmospheric concentrations and deposition fluxes
were also scaled to match the estimated average
PCBs air concentration of 0.24 ng/m3 in 1989. The
resulting historical PCBs loading time-series is
plotted in Figure 1.12 (point source loading was
neglected in calculating total Lake Michigan PCBs
load). The peak total loading in 1968 was estimated
to be 7000 kg/y; 80% was contributed by
atmospheric deposition. By 1990, the estimated total
loading has declined to 640 kg/y, with the
contribution of atmospheric deposition reduced to
40%.
This PCBs loading time-series was found to be in
general agreement with other PCBs loading
estimates for Lake Michigan (Figure 1.12), although
the atmospheric deposition loads may be somewhat
high. Further confirmation of the tributary loading
function was provided by data analysis and mass
balance model development for the lower Fox River
as part of the GBMBP. The calibration of that model
for the period 1989-1990 suggests a Fox River PCBs
load of 160 kg/yo If the Fox River provides 50% of
the total Lake Michigan tributary load of PCBs (as
suggested by Marti and Armstrong, 1990) then the
tributary load to the lake would be about 320 kg/yo
This value is in acceptable agreement with the 1990
MICHTOX tributary loading value of 370 kg/yo
Atmospheric deposition of PCBs was also measured
as part of the GBMBP. Based upon these
measurements, deposition to Green Bay was
estimated as 2.5 to 22 kg/y (Franz and Eisenreich,
1991; Sweet and Murphy, 1991) with a best estimate
of 11 kg/yo When extrapolated to all of Lake
Michigan, this depositional load (32-280 kg/y) agreed
fairly well with the loading time-series value of 260
kg/yo
1.4.6.4 Lake Huron Boundary Conditions
Toxic chemical concentrations in Lake Huron, a
boundary condition to MICHTOX, were based upon
1986 average concentrations reported by Stevens
and Neilson (1989). Because model results were
found to be generally insensitive to this boundary
condition, further resolution of Lake Huron
concentrations was considered unnecessary.
1.4.7 Chemical Bioaccumulation
The MICHTOX bioaccumulation model was used to
predict chemical accumulation up to lake trout in
Lake Michigan. The bioaccumulation model was
based upon the W ASTOXv4 food chain model
(Connolly and Thomann, 1985; Connolly, 1991). The
model treats bioaccumulation as a chemical mass
balance within individual organisms. The
fundamental bioaccumulation equation for organism
I of the food chain, consuming organisms j, is:
dv. n
--.!. = ku; c fd + E PijaijCijvj - Kjv;
dt j=1
where the rate of chemical accumulation in the
organism dV/dt equals the sum of direct uptake of
chemical by the organism from water (kui c fd) and the
flux of chemical into the animal through feeding (Pij aij
Cij Vj)' balanced by chemical elimination (K'i Vi)' The
parameters in the bioaccumulation equation are:
v.
I
= chemical concentration in organism I
[McheJMwet]
kUi
= uptake rate [L3fT/Mwet]
Pij
n
= feeding preference factor ( I Py = 1) of
organism I for organism j i = 1
Qij
= chemical assimilation efficiency across gut
C.
IJ
= food consumption rate [Mprey,we/Mpred,weJ
K'.
I
= chemical elimination rate [1fT]
The bioaccumulation equation is solved for time
variable chemical concentration in individual age
classes of each organism in the model.
Bioaccumulation simulations were made for
organisms residing in the southern Lake Michigan
hypolimnion and sediment. Migration between
segments was not considered in MICHTOX, although
it could be added to the simulation. Migration could
significantly impact bioaccumulation predictions for
fish moving across large exposure concentration
21
-------
10000
8Ci
-
>-
-
~ 6000
-
"C
co
..Q
m 4000
u
a..
2- )
10000
-
>-
-
C)
~
-
i 1000
..Q
m
U
a..
100
1970
total load
atrr. : spheric deposition
...........
tributary load
----
.........
.8 8.
.8 e.
, .
.8 e.
, .
, .
, ,
, .
...- -..
, ,
.8 -..
.' ,
~ ~
, .
, ,
, ,
~ .
. ~
....- -...
/ ~~---~~ ,
.- ., ..... ..... -.
..-." ...... ... -..
.....- "". -- .., ..... -- -.8:.:"
... ... -- ......~ - -
- ... -- - ... ... .............
o
1940
1950
1990
2000
1960
1970
year
1980
total
atmospheric
.
/).
...... 8
.....
R&S ...........
......
.......
'.
-...
-...
'.
'.
'.
M&R 0 "lI.J&O
... - - .... .... -....
.... ... e.
... ..... 8.
- - S&E ""
- - - - "::.. "'"
T~ "'tJ- - - e - - H"'""""
M&A -4.. - "'"
--... -.
S&A /). -"'''=':---
-. ........
-.... [J
GBMBS"""
'.
'.
...........
tributary - - ~ -
(refer to text
for data sources)
[J S&A
S&A/).
1975
1980
year
1985
1990
Figure 1.12. PCBs loading time function and comparison to reported PCBs load estimates.
22
-------
gradients, most likely between open water and
nearshore zones.
1.4.7.1 Food Chain
Trophic transfer (of both energy and chemical) was
simulated by defining the feeding preferences
expressed as diet fractions, Pij' of all organisms in the
food chain. Two alternative food chain structures,
displayed in Figure 1.13, were defined for Lake
Michigan lake trout. The first was the food chain
developed by Thomann and Connolly (1984) to
model PCBs accumulation in lake trout. This pelagic
food chain of plankton, Mysis, alewife, and lake trout
was constructed based upon extensive gut analysis
data from the early 1970s. An alternative food chain
incorporating linkage to benthos, in which trout
consume bloater, was also constructed. This food
chain structure was based upon the observation that
bloater may have substantially replaced alewife as
the major component of the adult lake trout diet in the
deep, mid-lake reefs of Lake Michigan (Eck and
Brown, 1990; Miller and Holey, 1991). This change
in diet has apparently accompanied the reported
decline of alewife since the late 1970s. Bloater were
assumed to consume the benthos Diporeia and
Mysis. This latter food chain may more accurately
represent the current Lake Michigan lake trout trophic
structure, although the data necessary to define all
the feeding preferences in this food chain are
lacking. For instance, the relative contribution of
benthos to the bloater diet is unknown. However,
simulation of the benthic-coupled food chain was
considered important because benthos may
accumulate significant concentrations of sediment-
associated chemicals.
1.4.7.2 Uptake Rate
A number of assumptions must be made to relate the
parameters in the bioaccumulation equation to
properties of either the organism or the chemical.
Particularly critical is the assumption that the
variability in bioaccumulation between chemicals can
be adequately parameterized as a function of Kow
(Thomann, 1989). This has been criticized because
some data suggest that bioaccumulation varies
according to chemical class as well as
hydrophobicity. In MICHTOX, this assumption was
modified to incorporate another chemical-specific
bioaccumulation parameter, metabolization.
The rate of chemical uptake ku' which parameterizes
the transport of chemical across the gill, may be
related to the respiration rate of the organism (R'):
k = E R1
u [02]
where E is the efficiency of chemical transfer across
the gill relative to oxygen, and [02] is dissolved
oxygen concentration [M/L3]. Respiration was
calculated by standard allometric relationships
(Thomann and Connolly, 1984). As suggested by
Thomann (1989), E was treated as a function of Kow
and organism size. E increases with chemical
hydrophobicity, reaches a constant, maximum value
at log Kow of six, and then apparently declines.
1.4.7.3 Elimination Rate
Elimination represents the net loss of chemical from
the organism by excretion, dilution by growth, and
chemical metabolization:
K1 = K + G + Me
The chemical excretion rate K [1fT] can be calculated
from the uptake rate and the bioconcentration factor
(BCF) [L3/M]:
K=~
BCF
BCF, normalized by organism lipid content f,
[MhPic/Morganism] is approximately equal to K ow at least
up to log Kow of six (Thomann, 1989):
BCF = " Kow
so that excretion rate may
combining these relationships:
ku
K=-
" Kow
be calculated by
The accuracy of this relationship for chemicals with
log Kow greater than six (including BaP, PCBS, and
TCDD) is uncertain, as bioaccumulation data for such
superhydrophobic chemicals are limited and often
conflicting.
23
-------
lake Trout I
0-2yr. . ~ L:.~~yr)
...../ :
/ :
~/ !
.//
1 yr. 'I i
.- . ',,+;<7 ,..~ /
l_o~~~~J \9-12mO; ..
~\
\ ~
\
6-8yr. . ~
~.~ . ',.., ~...
~ /
Alewife I
2 yr.
3 - 7 yr.
. . ~--~
Afysis
113 - 16 mo. I.
~ - ,-~- ,.:..-iA.-= .....:l<. .~.l..q" -
PLANKTON,
A. Pelagic Food Chain
D;pore;1I
1~1~~ ;
\ : /
\ i ,
I DETRITUS t
7-12yr.1
. ~ ,;..~:.. . r,,-~'. .,' ,
I
/)'f
/
lake Trout
Bloater
Mysis
B. Benthic I Pelagic Food Chain
Figure 1.13. MICHTOX food chain structure.
24
-------
Growth rates G [1 fT] were calculated from age-
weight data of the individual species (Thomann and
Connolly, 1984; Jobes, 1949; Evans and Landrum,
1989). The rate of chemical metabolism, Me [1fT],
was estimated for the three toxic chemicals
metabolized by fish: BaP, TCDF, and TCDD. Lower
trophic levels apparently do not metabolize BaP
(Evans and Landrum, 1989); it was assumed they did
not metabolize TCDF or TCDD either. Although Me
may be expected to vary with organism as well as
chemical, adequate data for such specific model
parameterization do not exist. Instead, constant
rates of metabolism were parameterized for each
chemical in fish. For BaP, a metabolization rate of
0.023/d was estimated from the bluegill sunfish data
of McCarthy and Jiminez (1985). For TCDF and
TCDD, the calibration of a bioaccumulation model to
Lake Ontario trout data suggested a metabolization
rate of 0.0035/d (Endicott et al., 1991). These
metabolism rates correspond to chemical half-lives of
30 days for BaP and 200 days for TCDF and TCDD.
This calibration was based upon comparing the biota-
to-sediment factor (BSF) for polychlorinated dibenzo-
p-dioxins (PCDDs) and dibenzofurans (PCDFs) to
that for dther HOCs. The order-of-magnitude
decrease in BSFs for dioxins and furans were
attributed to metabolization in fish, although other
explanations for reduced accumulation of PCDDs
and PCDFs have been offered (Oppenhuizen and
Sijm, 1990). Reduced bioaccumulation of highly
hydrophobic chemicals has also been suggested as
evidence for metabolism of organophosphate
pesticides (de Wolf et al., 1992). This procedure
represents only a tentative calibration of
metabolization; Me would preferably be based upon
direct measurement instead of inferred from an
observed reduction in bioaccumulation relative to
other chemicals of similar hydrophobicity.
1.4.7.4 Dietary Accumulation
Accumulation of chemical from food depends upon
feeding preference, consumption rate, and chemical
assimilation efficiency, the fraction of ingested
chemical transferred through the gut to the organism.
The rate of food consumption was calculated by the
organism's energy requirements for growth and
respiration, estimated by standard allometric
relationships. Equating the caloric density to dry
weight fraction (fdry) of food consumption rates was
calculated as:
fd . G.+R.
Cij = (~) I I
fdty,j aj
where aj is the food assimilation efficiency
[MjngesteJMeonsumed]. A food assimilation efficiency of
0.8 was used for alewife and trout (Thomann and
Connolly, 1984), while a lower value characteristic of
herbivores, 0.4, was used for Mysis. Bloater food
assimilation efficiency was 0.68, according to
Rudstam et al. (1992). An aj of 0.072 was selected
for Diporeia, based upon consumption data (Dermott
and Corning, 1988; Landrum and Robbins,1990).
Chemical assimilation efficiency was treated as a
function of both species and chemical hydrophobicity.
A chemical assimilation efficiency of 0.6 was used for
trout and bloater based upon experimental data for
PCBs assimilation in trout (Niimi and Oliver, 1983).
For Mysis and alewife, the log Kow-E relationship was
found to also describe ajj computed from HOC data
for Lake Ontario (Oliver and Niimi, 1988), as
displayed in Figures 1.14 and 1.15. The somewhat
poorer fit for Mysis aij may be due to scatter in the
plankton concentration data. A similar treatment of
Lake Ontario Diporeia data was used to define a
10gKow-ajj relationship for that species; the result is
presented in Figure 1.16. Regressing the PCBs
congener data only, the following relationship was
obtained for use in the model:
log aij = 5.49 - 109 Kow for 109 Kow > 5.49
(aij = 1.0 for 109 Kow ~ 5.49
Better parameterization of the chemical assimilation
efficiency would be desirable, especially for benthos,
because significant unexplained variability is
apparent for this parameter.
1.4.7.5 Modeling the Base 01 the Food Chain
Bioaccumulating chemicals enter the pelagic and
benthic food chains at the plankton and detritus,
respectively. For plankton, chemical accumulation
was assumed to be a partitioning process, so the
plankton BCF, was calculated from Koe assuming 2%
25
-------
10
08-
'0; 06-
.S!
i
Q)
'iij
0.4-
02-
,'..................Q..
" ".
.../ ;t..\..
,.: Q) '9
../ 0.- a...
! 8 \ .
.... 8:> ""
.~/ 8b~ .
. 0 ~~~
W""
..............
. PCB congeners
o other HOCs
0.0
4
I
I
7
I
5
I
6
log KOIN
8
Figure 1.14. Chemical assimilation efficiency for
Mysis calculated from Lake Ontario PCBs data.
10
08-
0.6-
'0;
.S!
UJ
,.,
E
0.4-
0.2 -
..'..'..'.'...'.'..".
/ 0 \~
./ 0 \. C\::) -D
. °0 0-
...." 0 \"
..."" 0 CR.
.' "'Q
o """"
............
o
0.0
4
o
I
I
7
8
I
5
I
6
log KOIN
Figure 1.15. Chemical assimilation efficiency for
alewife calculated from Lake Ontario HOCs
data.
organic carbon (wet weight basis). The chemical
concentration in detritus, the benthic food source,
was assumed to be equal to that of the surficial
sediment.
Neither plankton nor benthos accumulation is
particularly well described by this model. Evidence
that plankton accumulation is not simply a partitioning
process has been presented by Skoglund and
2
y = 54905 - 0 99886x RA2 = 0.705
o
'0;
to
a;
o
Q,
'C
'"
.2
o
o
-1
-2
-3
4
o
7
6
log KOIN
8
5
Figure 1.16. Chemical assimilation efficiency for
Oiporeia calculated from Lake Ontario PCBs
data.
Swackhamer (1991). Also, the feeding of Diporeia
is highly selective for fine, high organic carbon
sediment (Landrum and Robbins, 1990) which may
be enriched in HOC, thereby increasing contaminant
accumulation above that being predicted by the
model.
1.4.8 Steady-State Model
The solution of the mass balance and
bioaccumulation equations simplifies considerably if
the time derivative terms (d/dt) are eliminated. This
steady-state solution of the model equations
produces results which are adequate for many
applications, except during periods of substantial
concentration change. A steady-state version of
MICHTOX was developed to validate the numerical
computations in the dynamic model, and to facilitate
model uncertainty analysis. The steady-state
MICHTOX was implemented as a spreadsheet as
well as a FORTRAN program; a sample of the
spreadsheet output is shown in Figure 1.17.
Solution of the steady-state mass balance equations
is obtained by simultaneous solution of 14 linear
equations, requiring the inversion of a matrix M of
coefficients (Thomann and Mueller, 1987).
Several additional simplifications to the model were
necessary to directly obtain the steady-state solution.
Time variable model parameters were replaced by
26
-------
:!!
cc
c::
...
(I)
~ ~,""._.'C,.ci'ic i.p_' .......i...,.,I-c......;'i... i.p_' ...",1....1::.1." ,,.,r,.,.....,,
~ ~
~ Q....IH' I09-Kow 628 109-l(oc ..6
divu:::iOft '" Hie 5 38
Q...2 1036 Vp "65E-04 'dw_' 0.=
en 1)..3 1133 1<1 280 fdw_2 bM.?
- Q....i 16$ 1<9 354 fdw_3 0.=
(I) Q...5 '6' fdw_. (ltJ,u2
AI Q...6 310 .....,.,... IIi"... ~./d" fdw_5 u./J.$
Co GLJ 383 ~P 0 Idw_6 o.~$
'< ."''''':110.,.(_~/c)- U 3 ODE -0" Idw_J o.~66
I [..122110 Ipw_' o..2l'2
en L23 2110 $
(I) ......,Ii.n'.Id'll- C_!&I 0120 Id: 2 'ME-<>$
- y~_1 15 Id:_J . ME -<>$
3 ,'_2 15 1II..6~n..,lr 61111......5:- Id~_4 ~=-Q$
y,_3 1'.) ~uo~oU 200E-t1 fd,_S ~ ME -Q$
o y~_4 15 ,uv 2.00E-05 'd:;_' ~ ME -<>$
Co y~_5 15 uin 2.25[-03 'd~_1 ~=-<>$
~ y~_6 15 ddc.p 11' Ip~_' 0.-"9
y:;_115 RT 2353 Ip:_2 0.-"9
- ".6ft~I.lft.'.I"J' Ip:_' 0.-"9
o ' ,,_5 f~.sc..Q!
:T ~ ,,_6 IfOE-<>$
(I) loe....1 0118 [AD 190 ",_1 30SUJ6
::1 '0(.....2 on8
~ '0e....3 0 n8
loc....4 0118
- foc....S 0 n8
"'C 'oL6 0 118
0 10<....10118
m loc:;cd 0032
....1.... nor. D61.d.,.
U1 pwdo< 10
- poro~ 0,9
D:IO.ti.ID.I.a D..:...t...-
Ie 0 S
b2c11 061
i8.else 0' "M"
wtet., av.
6£./1 4£./1 ...~-1.2 -£.1.2 _"£-/.$ ....~.Jj /C.I$ -:J£.U /C.I$ -/C./.$ .'t: 1$ ....~-/.$ :JE.I$ -:J£.I$ I
6£-tJ6 oX-tJ6 IE.IN ~.(16 .'t:.(16 -/C.(16 i'£-IM -6£ -1M i'£-IM -i'E-M 1£.(16 .1£.(16 6£.(16 -6£.(16 tJ
-pC-1.2 -1£-1.2 /C-1tJ -$£./1 IE./I .1£./1 1£-12 ~£-1.? $E-!.? .1£-1.P ..!IE-I,? ~-!.? ~E.II 4£./1 I
...,£..qi" ~.(16 i'£-tJ6 -1£ -Q$ $E-Di" 'IY-<~" J:E.o,.. ..$£.0,.. ~£-Dt 4£.(Ii" 6E'()i" -6£-tJi" X-t>< -X-tJ6 0
X.I$ ....¥:-Ij £-1/ N.t.! ~£-IQ ~£.1fJ ,y-Ij -6£.1$ i'£.1j -;'£-/.$ /E.I..? ./E.I..? M-I..? -6£-1.! IU
:JE.(16 -1£.(16 1C-tJ6 .6£ -OJ" :JE-<>$ .$£-<>$ $£.(16 .$£.(16 $£.(16 .$£.(16 :JE.(16 -:J£.(16 1£ -Dt ~E.oi" 0
1"£-/16 -X." X." ....¥:.U X." -X.I! 6£.(16 -6£.(16 _1>£-10 ....¥./ItJ 1£11 .1£.1/ $£.1$ .1£./.1 0
~£-I..? ....'E.!'? ...':>£-10 ...K-/I .:¥" ...."£"-/1 '£-IM .IC-M 1£""" .1C-tJ6 $£.(16 -;'£.os :JE-IM ~'t:-IM 0
$£." ....~.u ,2E.f.2 -1£-/,2 ...¥.I$ ....-;C.IS 1C.tJ$ -1£.(16 1£.(16 -1£.(16 1£ -1M -:J£.1tJ X.II ..$£-If D.Q/16
4£.1tJ 4£.NJ :JE.tJ$ ~,£.(16 ~E-IM -X -1M .'t:-IM ~'t:-IM K-IM -X -1M .'£ -Q$ ~'t:-<>$ JE-bt .J£.Qi" "
/C.I$ -6£." JE-t.! ..x-I,! $£-1$ .fE.1S 1£ -1M -IE -D3 1£ -1M .1£ -D3 .'£ -D3 ~'t:-IM 6£-1/ N'II 0
$E -D3 ~'t:-D3 2£..Qt .I£.ot .'t:.(16 ~'t:.(16 $E-<>$ .$£-<>$ 6£-<>$ -6E -Q$ 1£ -1M .1£ -1M :JE-tJ6 .:JE-tJ6 0
IE-I.? .1£./.$ ~£.II ....¥:-II ~£.I.? -.lE.1.? ;r.1I -6£-" l'£-/I -;r-II 1£." ./E.1O ;'£-10 -6£.1tJ o I.!
6£.(16 -X.(16 .'t:-Q6 .1C-tJ6 ...?£.Qt ....?£.Qt 'E-tJ6 -<£-tJ6 ~£-tJ6 -<£.1M ,y""" .,Y.Q6 '£-<>$ .i"C.Q$ 0
e'e8in,1 eoaee.a"a.ioa
sol.U08 weeto. . II
C_wl $s...~.Qf
C_~I J 90
C_w2 $.?l£.Qf
C ~2 J~
C_w3 /...~~
C_~3 "$6
C_w.t 0016t
C ~.t "1/
C_wS SJIX-Q..5
C_~'S J!9
Cw6 J$J£..{J.I
e ~6 .2~-J
C_w1 I$.!I£-OJ
C_~T I~-'t
Cd '1
1 J,:1.$
1 ,?,s"
2 ~"Q
2 .?,J,:/
:3 t6J
3 Si".!1
4 ,"'$6
, =<
, '$$
5 .?..."6D
6 .'DO
6 ....'"
T 6JO
7 .s.!.?
, . ,
.".
2~6
."$
.?J J
$N
$$0.
H-'
.5P.1
2$$
2$$
m
=
nf
n2
18 .t
10g(Kow)= 6.?$
w~tcl(Of'l«(\l9f1)= ~9$£.(Jf
pore w~tel cone (\19")" 2.~-OJ
~edl'Ml'\t COI'\C (flglg) = .2'/6
dl~~olyed o-ygcf'l (mg/I) = 100
pcl~gle diet Inct,on-;:,Iewile = 0620
lipid"~ction=
chell'> ~~dll'>ib,ion {If =
ehm tr~fI~f{r eff :
'ood ;:'~~11f'I dlldc.flcy =
con~ulf'lpt rt (g/g/d)"
growth nte (Ud) =
le~pir~llon rt (g/g/d) =
II'Ic.hboll~1f'I nte (11d)
rt~p rt. (g 02/g/d) =
upt~ke nte ('!kg/d):
qlld)~
Bef(t/kg):
'ood Chlfl tr~n: nllo.1 =
BAf (11'9) ~
log (BAF):
cOI'\(tl'llr~tiolll(ng/g)=
Cd (P911) b, u.,.ut
1 (."'.. b, st._e..
:~:~
t I I t I I ,
~:~
. .
biota co.entratio. (U98.r:.t 1; 89'9)
':~b..d
I 3
229
, "
4 118
5 1121
M sis
o 0400
o.J;"J
Q.JI~
0.400
Plu.to.
lipId h ~ctlon : 0 0200
eCf (l/kg"l) ~ .?$X..()t6
BCf 1''''9)~ JI-X~
(onctl'llr~tlon (ng/g) : 2M
LahT...t
a 0300 0 0800 aOMO 011.t
0165 05tJ Ofl'! 0600
Q.S,""! OJl'! OJ,., Q~~i"
00120 0800 oeoo o.eoo
ow .1 JX..()S JS/E-IJ.$ 66;'£-DS
200E.03 e .0[-0. 1340E-04 498E.04
00340 001t4 00". SOOE.03
000 000 000
S.f,Y..Q.S .?U£..Q.S .?/"'J£.()$ l...~.()$
.iV IJ6 IJ6 f($
$.2..~.()$ I():JE.()$ I():JE.()$ 1 HE.Qi
~.6Q£..tJ4 641£.QJ $'/!£-Q.t ,.. $K.Q4
161 .?.11 Ii"" 6~
.?!...~...os /"';"£..06 :i,?!tE..{J'"
SoJ 6." ;$6
S06 6; !ItJl ft:?i"
T ohl Aheih = !tS
0.?.$1
1.3OE.03
o 0.00
961£.fJ.S
fM
t.?6£ .()$
$ ME.04
3.55
J,"'X,.f)$
$;6
.?SJ
-------
annual-average constant parameters, and the
seasonal stratification of the main lake water column
was eliminated in favor of vertically-integrated
segments. Volatilization and photolysis rates were
adjusted to compensate for the missing influence of
stratification on water column loss rates. The
bioaccumulation model was also simplified, by
adopting a simpler food chain structure (Figure 1.18).
The alewife were assumed to consume 60% Mysis,
a value based upon Flint's (1986) assessment of
carbon flow in the Lake Ontario lake trout food chain.
Alewife serves as a "generic" forage fish in this food
chain, including other fish such as bloater, sculpin,
and smelt.
Steady-state model solutions were compared to the
equivalent dynamic model results to ensure
consistency between models. For all chemicals, the
steady-state model results were found to be within a
factor of two of their dynamic model counterparts.
Given the rather substantial simplifications to the
dynamic model, this agreement with steady-state
model results was considered acceptable. The
alternative to developing a simplified steady-state
model would be to run the dynamic model to steady-
state, which was not feasible due to constraints upon
computer resources.
1.4.9 Chemical-Specific Parameterization
Chemical-specific input parameters to MICHTOX
include the octanol-water partition coefficient, vapor
pressure, Henry's constant, the photolysis rate, and
the rate of metabolism. These parameters, including
sources of data, are summarized in Tables 1.4
through 1 .7. Significant derived chemical
parameters are tabulated as well. These tabulations
also contain estimates of the uncertainty associated
with the parameter values for use in model
uncertainty analysis.
Parameterization was particularly difficult for the
priority toxics representing chemical mixtures:
chlordane, DOT, PCBs, and toxaphene. This
difficulty arises because the parameterization must
Lake Trout
Alewife
p=O.38/
/
'~=O.62
Diporeia
.~, - -. -, ;.:.. '
-
Detritus i
~
'~.!:""'7 ;~.-r
Mysis
". '~,.,,'~:.:l!'-
Plankton
.' ':~I!>' '~'{i
"''r'1
Figure 1.18. Simplified Lake Michigan lake trout food chain.
28
-------
Table 1.4. Octanol-Water Partition Coefficient (Kow) and Organic Carbon Partition Coefficient (Koc)
for Lake Michigan Priority Toxics
Chemical log Kow InCV of Kow Source log Koc
SaP 6.14 0.28 Endicott et al., 1990 6.4
Chlordane 6.00 0.69 Endicott et al., 1990 6.3
~DDT 6.00 0.49 Endicott et al., 1990 6.3
Dieldrin 5.50 1.0 Endicott et al., 1990 5.9
Heptachlor epoxide 5.40 0.66 Veith et al., 1979 5.8
HCB 5.84 0.49 Endicott et al., 1990 6.1
PCB4 5.89 0.28 Endicott et al., 1990 6.2
PCB5 6.28 0.31 Endicott et al., 1990 6.5
TCDD 7.02 0.31 Endicott et al., 1990 7.0
TCDF 5.82 0.66 Burkhard and Kuehl, 1986 6.1 .
Toxaphene 4.82 0.66 Lyman et al., 1982 5.4
Table 1.5. Physicochemical Parameters Used in Volatilization Parameterization of Lake Michigan
Critical Pollutants
Py Py Hhl
Chemical (Pa @ 10CC) InCV (PalM @ 10CC) HIC InCV Source
SaP 2.4e-6 0.64 0.013 0.64 Endicott et al., 1990
Chlordane 1 .3e-3 0.64 1.3 0.64 Endicott et al., 1990
~DDT 1 . 1 e-4 0.50 0.53 0.50 Endicott et al., 1990
Dieldrin 3.3e-4 1.0 0.028 0.98 Endicott et al., 1990
Heptachlor epoxide 5.1 e-2 0.89 1.0 1.2 SCR*, 1988
HCS 2.8e-2 0.37 13 0.088 Endicott et al., 1990
PCB4 2.4e-3 1.1 6.3 0.78 Endicott et al., 1990
PCB5 4.7e-3 1.3 5.4 0.93 Endicott et al., 1990
TCDD 1 .4e-5 0.78 0.45 0.40 Endicott et al., 1990
TCDF 3.5e-5 0.78 0.25 0.64 Rordorf, 1989
Toxaphene 3.3e-4 1.3 0.075 0.64 Sunito et al., 1988
*Syracuse Research Corporation
29
-------
Table 1.6. Volatilization Rate Parameters for Lake Michigan Critical Pollutants
Chemical
BaP
Chlordane
IDDT
Dieldrin
Heptachlor epoxide
HCB
PCB4
PCB5
TCDD
TCDF
Toxaphene
K, (mid)
3.0
2.7
2.7
2.7
2.8
3.1
2.9
2.8
2.8
2.8
2.6
Kg (mid)
360
340
340
330
350
370
370
360
330
330
330
Ky (mid)
2.0e-3
0.17
0.074
4.0e-3
0.14
1.2
0.73
0.63
0.062
0.034
0.011
Table 1.7. Photolysis Rate for Lake Michigan
Critical Pollutants
Chemical kp (mid) kp InCV
BaP 1.4 (log -uniform over
0.64-3.0)
IDDT 0.068 1.7*
Dieldrin 3.6e-4 1.7
HCB 1 .8e-3 1.7
TCDD 0.02 (log-uniform over
0.020 - 1.1)
*The 95% confidence limits are :t factor of 10.
properly average the properties of the chemicals in
the mixture. Even if the constituent properties are
known, averaging is difficult because the composition
of the mixture may be variable or unknown. Model
predictions for mixtures are particularly uncertain
because the probability density functions for
chemical-specific model parameters must account
for variation amongst the properties of the
constituents.
For DOT, the parameters determined for p,p'-DDD,
-ODE, and -DOT were averaged; photolysis rate was
based upon data for DOE. PCBs was modeled as
two homologs, parameterized as average
tetrachlorobiphenyl (PCB4) and PCB5. PCBs loads
and air concentrations were assumed to be equally
distributed between the two homologs, and PCBs
model predictions were obtained by summing the
PCB4 and PCB5 results. This approach was found
to yield results almost identical to a more complicated
procedure of modeling PCBs as six homologs, with
loads defined according to an Aroclor-1248 homolog
distribution. The mixture of polychlorinated terpenes
that make up toxaphene probably have a range of
parameters more variable than PCBs, yet data are
available only for the technical mixture. Similarly,
only limited data are available to parameterize
chlordane, a mixture of cis- and trans-chlordane
isomers, trans-nonachlor, and a variety of other
chlordane-related chemicals.
1.5 Model Validation
Validation of predictions is necessary to judge the
overall model performance and provides one
indication of expected model accuracy. MICHTOX
was validated by comparing model predictions to
existing data for plutonium, lead, and PCBs in Lake
Michigan. Additional validation of the
bioaccumulation model was also performed. It
should be noted that model parameterization
included no direct calibration to toxic chemical data
for Lake Michigan. Therefore, validation represents
a "fair test" of the model's predictive abilities given
the constraints of the data set.
30
-------
1.5. 1 Plutonium
The long-term prediction of plutonium concentrations
in southern Lake Michigan is plotted in Figure 1.19.
Also plotted in that figure is a long-term monitoring
record for plutonium concentrations in the lake during
unstratified periods (Robbins, personal
communication). Plutonium concentrations are given
in femtocuries per liter (fCi/L), a measure of
radioactivity. The predicted seasonal divergence of
epilimnion and hypolimnion simulations will be
considered further. However, the simulated
plutonium concentrations converge at fall overturn,
and the agreement between model predictions at
such a time with the long-term data is excellent.
Because the singular loss mechanism for plutonium
is sediment burial, this long-term agreement validates
the main lake particle flux parameterization of
MICHTOX.
Because the hypolimnion is seasonally isolated from
the epilimnion due to stratification, concentrations in
the two water column segments diverge during the
summer of each year. This divergence is particularly
apparent in the MICHTOX simulation after 1965,
when plutonium loading had significantly declined.
This portion of the simulation is plotted in greater
detail in Figure 1.20 along with seasonal epilimnetic
5
~ 4
~
g
. -- epilimnion
- hypolimnion
. overturn data
c::
.2
~
c:
Q)
u
c::
o
u
E
:::I
.~
o
'5
Q.
3
2
o
1950
1960
1970
1980
1990
year
Figure 1.19. MICHTOX simulation of plutonium in
southern lake Michigan (epilimnion,
hypolimnion, and overturn data).
2.0
~
g
c
.2
~
C
Q)
t)
c
o
t)
E
:::J
'c
o
:;
a.
.... epilimnlon
- hypolimnion
. seasonal epllimnetlc data
1.5
1.0
0.5
00
1972
.,
1975 1976 1977
1978 1979
1973 1974
year
Figure 1.20. MICHTOX simulation of plutonium in
southern lake Michigan (epilimnion,
hypolimnion, and seasonal epilimnetic data).
data (Wahlgren et al., 1977). From this period
through the end of the simulation, the major source
of plutonium entering the water column is
resuspended sediment. Plutonium "builds up" in the
hypolimnion during stratification as sediments
resuspend and particle concentrations increase. At
the same time, plutonium is depleted in the
epilimnion because resuspended particles are
trapped in the segment below. Residual atmospheric
loading (which peaks each summer) and entrainment
prevent plutonium from disappearing entirely from
the epilimnion, however (Robbins and Eadie, 1991).
At fall overturn, vertical particle fluxes quickly re-
establish uniform water column plutonium
concentrations. MICHTOX adequately simulates the
magnitude of the epilimnetic depletion, although the
prediction is out of phase in some years. This is
because the seasonal representation of stratification
in MICHTOX is somewhat inaccurate and does not
vary year-to-year as does the lake. However, this
agreement generally validates the simulation of the
stratification's impact upon particle and particle-
associated contaminant fluxes.
The variation of chemical concentrations in the lake
water exhibited for plutonium should be expected for
the other toxic chemicals as well, although
substantiating data are not known to exist. Previous
mass balance models for toxic chemicals in the Great
31
-------
Lakes have utilized a vertically-integrated
(completely-mixed) water column, assuming that the
impact of stratification upon, at least, long-term
simulations would be negligible. MICHTOX
plutonium simulations suggest this may not be
correct. Figure 1.21 displays a comparison of
plutonium predictions made both with and without a
stratifying water column. All other parameters and
forcing functions were common to both simulations.
The effect of stratifi- "'tion is seen to be a persistent
elevation in wate Tin concentrations following
load reduction. B. this increases the simulated
persistence of thE: :al, it appears important to
incorporate stratifl' n fate and transport models
for, at least, Lake ran.
5
:J 4
:::.
g
c
,2 3
g
c
~
c
8 2
E
::I
'2
S
::I
0.
0
1950 1960
- epilimnion
- hypolimnion
..... vertically-integrated model
. overturn data
1970
1980
1990
year
Figure 1.21. MICHTOX simulation 01 plutonium in
southern Lake Michigan - sensitivity to vertical
segmentation.
1.5.2 Lead
Lead, like plutonium, is a toxic chemical subject to
loss by burial. The prediction of annually-averaged
lead concentrations in southern Lake Michigan is
plotted in Figure 1.22. Problems with analytical
detection limits and sample contamination have
confounded efforts to monitor lead as well as other
trace metals in Great Lakes water; only the data
reported by Rossmann and Barres (1988) is
considered reliable. The reported concentration of
30.3
C,
2-
c:
~
!!
C 0.2
!!!
c:
8
"C
III
J! 0.1
0.4
- - - - - epilimnlOn
- hypolimnion
o data
......, ..., ......,...... ". "'.
.'
.'
......-
..'
-..........
0.0
1940
1950
1960
1970
1980
1990
year
Figure 1.22. Simulation of annual-averaged lead
concentrations in southern Lake Michigan.
lead in Lake Michigan, 0.25 pg/L, is in good
agreement with the MICHTOX prediction. The
simulation indicates essentially a "plateau" in lead
concentrations since the early 1980s. The prediction
for lead in southern Lake Michigan surficial sediment
is plotted in Figure 1.23. Mudroch and Williams
(1989) reports lead concentrations in Lake Michigan
surficial sediment of 10 to 130 ng/g, with a mean
value of 40 ng/g. The MICHTOX simulation is in the
range of this data; however, the model overpredicts
the mean value.
c;;
c,
.s
.~ 60
~
c
8
8 40
"C
III
j!
100
- segment 11
o data
80
20
)
1
o
1940
1950
1960
1970
1980
1990
year
Figure 1.23. Lead simulation in southern Lake
Michigan surficial sediment.
32
-------
1.5.3 PCBs
A relative wealth of concentration data exists for
PCBs in Lake Michigan, at least in comparison to the
other priority toxics. However, this is an insufficient
database to quantitatively compare with model
predictions, which represent average concentrations
over relatively large areas of the lake. Thus, only
qualitative comparison to data was used to validate
model predictions in water, sediment, and biota.
1.5.3.1 Water
The verification simulation for PCBs was made with
the PCBs loading history described previously;
unless otherwise noted, results are annual average
predictions. Water column PCBs predictions are
plotted in Figures 1.24 and 1.25 along with data for
comparison. The southern lake simulation in Figure
1.24 indicates peak water column concentrations of
3.4 (hypolimnion) and 2.2 ng/L (epilimnion) in 1970,
with values dropping to less than 0.6 ng/L by 1990.
These predictions are in agreement with the data of
Rodgers and Swain (1983; <10 ng/L in 1970, 3-9
ng/L in 1976) as well as the 1980 data of
Swackhamer and Armstrong (1987). This latter data
suggests a variance between epilimnetic and
hypolimnetic PCBs concentrations (1.2 versus 1.7
ng/L) similar to the MICHTOX predictions (1.0 versus
1.8 ng/L). Central lake simulations are essentially
the same; however, simulated PCBs concentrations
in the northern lake segments are higher. At the
1970 peak values, northern lake concentrations are
30% higher than in the southern and central lake
segments. Swackhamer and Armstrong's data show
no such PCBs concentration increase in Lake
Michigan, although none of the stations sampled
were actually in the northern basin as defined in
MICHTOX. Several factors could contribute to a
problem in the northern basin, including insufficient
data to characterize particle fluxes and
concentrations, and overestimation of air
concentrations and atmospheric fluxes over the
northern lake.
Predicted PCBs concentrations in the three Green
Bay segments are plotted in Figure 1.25, indicating
a strong and persistent concentration gradient
between inner- and mid-outer segments.
33
5
«10)
.......... epilimnion (seg.1)
- hypolimnion (sag.8)
4
(3-9)
:J'
C>
.s
c:
o
~
E
OJ
()
c:
o
()
cc
()
a.
3
2
..~.//m'\H
"".
.........~.
",
.............
o
1940
1960
1970
1980
1990
1950
year
Figure 1.24. Simulation of PCBs in southern Lake
Michigan.
40
(121 )
:J 30
C>
.s
c:
o
~
E
~
c:
8
cc
()
a.
~ inner GB (seg.5)
....!.... mid GB (seg.6)
_!: _. outer GB (seg.7)
20
10
o
1940
------.."""cc.,.,:.:.:.:.:.:.:L:.::.:.. ..
1980
1960
1970
1990
1950
year
Figure 1.25. Simulation of PCBs in Green Bay.
Again, the PCBs predictions are in good agreement
with water column data for 1980 (Swackhamer and
Armstrong, 1987) and 1989 (GBMBP, October 1989
cruise). Simulated PCBs concentrations in the outer
bay are compared to the main lake predictions in
Figure 1.26, indicating a PCBs water concentration
gradient of 0.5 ng/L between the two segments.
-------
5
4
- outer GB (seg.7)
.......... central LM hypolimnion (seg.9)
:J
C,
.s
c:
o
.+:
~
C
Q)
(.)
c:
o
(.)
II)
<.)
Q.
- - -. central LM epilimnion (se9.2)
3
2
o
1940
1970
1980
1990
1950
1960
year
Figure 1.26. PCBs simulations in central Lake
Michigan and outer Green Bay.
1.5.3.2 Sediment
Verification of sediment concentration simulations is
made difficult by spatial sediment variability which
has generally not been adequately sampled. This
should be partially resolved by the sediment
sampling conducted during the GBMBP.
Unfortunately, these data were not available at the
time of this report. Simulated PCBs concentrations
in surficial sediment segments are plotted in Figures
1.27 and 1.28. The southern lake simulation in
Figure 1.27 indicates peak surficial sediment
concentrations of 260 ng/g in 1974, with values
dropping to 100 ng/g by 1990. Concentrations
predicted for 1980 are consistent with the <200 ng/g
reported by Sonzogni and Simmons (1981), Strachan
and Eisenreich (1988), and Weinenger et a/. (1983).
Predicted sediment concentrations in inner- and mid-
Green Bay (Figure 1.28) are much higher; peak inner
bay sediment concentrations of 1900 ng/g are
predicted for 1978. The predictions are again similar
to sediment concentrations measured in the early
1980s (Swackhamer and Armstrong, 1988;
Hermanson et a/., 1991).
The vertical distribution of PCBs concentrations in
sediment beneath the surficial mixed layer can be
obtained from the time-series of surficial
concentrations, by transforming the time scale to the
600
500
~
~
C)
.s 400
c:
.Q
iii
~ 300
~
c:
8
II) 200
<.)
Q.
100
.
- southern LM (se9.11)
........., outer GB (seg.17)
o
1940
............
,..,
.................
1980
1990
1950
1960
1970
year
Figure 1.27. PCBs simulations in Lake Michigan
and outer Green Bay surficial sediments.
--
~
C)
.s 2000
c:
o
~
~ 1500
Q)
(.)
c:
o
(.)
II) 1 000
<.)
Q.
3000
2500
-.!..... inner GB (seg.15)
....?.... central GB (seg.16)
.
.
500
o
....,/.....}.....
o
1940
1970
1980
1990
1950
1960
year
Figure 1.28. PCBs simulations in Green Bay
sediments.
sediment depth using the particle burial velocity.
This transformation assumes that particles are buried
without mixing beneath the surficial layer, and burial
velocity (Le. sedimentation rate) is constant. While
these assumptions may be questionable, they are
routinely made when sediment cores are
radiometrically dated. A simulated sediment PCBs
34
-------
profile for southern Lake Michigan, corresponding to
an "average" depositional zone sediment core
collected at 1990, is shown in Figure 1.29. A
pronounced concentration maximum beneath the
surficial mixed layer is predicted; such a PCBs
concentration distribution has been reported for
several sediment cores collected in Lake Michigan
(Hermanson and Christensen, 1991). However,
comparison of other aspects of the sediment
distribution to data is difficult. For instance, PCBs
concentration profiles have apparently not been
reported for sediment cores with sedimentation rates
as high as the 400 g/m2/y parameterized for southern
Lake Michigan.
-0.5
-1
-1.5
-2
-2.5
-3
E -3.5
.8. -4
~
c.. -4.5
Q)
'0 -5
-5.5
6
-6.5
-7
-7.5
-8
o
50
100
150
200
250
300
PCB concentration (ng/g)
Figure 1.29. Simulated PCBs distribution in
southern Lake Michigan sediments.
1.5.3.3 Biota
MICHTOX bioaccumulation model predictions for
PCBs were validated using lake trout and bloater
data. Lake trout monitoring has taken place since
1971, and a number of data sets have been compiled
which document PCBs and other HOC levels in Lake
Michigan fish. Several of these data sets are
presented in Figure 1.30. The (age class) compiled
age seven data were based upon USEPA STORET
and other sources (Thomann and Connolly, 1984);
the USEPA/USFWS data is from an ongoing federal
monitoring effort (DeVault et al., 1986). These data
both show peak PCBs concentrations of 18 to 23
30
25
-+- compiled data
"",0"" age 7 only
--0-- EPNFWS (:l:1sd)
--l::r- MDNR (:l:1sd)
c;
0>
2: 20
c:
o
~
~ 15
Q)
o
c:
o
o
II) 10
o
c..
\~u ~
5
o
o
1970
1975
1980
1985
1990
year
Figure 1.30. PCBs concentrations in Lake
Michigan lake trout.
Jig/g (ppm) occurring in 1974 followed by a fairly
rapid decline; however, details of the time-series vary
between data. More limited, recent data were
obtained from the State of Michigan (Michigan
Department of Natural Resources, 1990) which
indicate that PCBs concentrations have leveled off at
about 3 Jig/g. The MICHTOX predictions of age
seven lake trout PCBs concentrations are plotted
with these data in Figure 1.31. Age seven was
chosen because this age class is represented in all
data sets. The model predictions are similar to the
magnitude of PCBs concentrations measured in fish
through 1980, although the distinct "peak" in PCBs
concentrations around 1974 (particularly evident in
the USEPA/USFWS data) is not reproduced. After
1980, the predicted PCBs concentrations, although
declining, are higher than the data by a factor of two.
In general, the predicted PCBs concentration trend is
considerably less "dynamic" than that of the data.
Three factors may be suspected as causes of the
lack of fit of the trout PCBs predictions. First, errors
in the model structure and/or parameterization may
be responsible. The dynamics of the trout PCBs
predictions follow the concentration change in the
sediment; this aspect of the model simulation will be
considered later in the report. Alternatively, the
loading time-series for PCBs to Lake Michigan may
be in error. If PCBs loading declined more rapidly
35
-------
30
IB1 0 EPAlFWS
:e; 25 . MDNR
C) - prediction
2:
c 20
.Q
~
'E
Q)
u 15
c
0
u
II)
u 10
Q.
'S fn
g
5
! i .
o .
1970 1975 1980 1985 1990
year
Figure 1.31. Verification of PCBs accumulation In
age seven lake trout.
than the estimated time-series after 1970, then PCBs
concentrations throughout the model, including trout,
would also decline at a rate more consistent with the
data. Presently, it is not possible to distinguish
between these possibilities as the source of lack of
fit. Finally, the quality of the data may be questioned.
The reliability of PCBs concentration data generated
during the early 1970s has been repeatedly
questioned because of concerns with quantification
errors due to interference with other chemicals such
as DOT and toxaphene which coelute in GC
chromatograms (Swackhamer and Armstrong,
1987). Both data sets used packed column GC for
analysis, with results quantified as Aroclors; however,
comparability of the analytical results cannot be
directly evaluated. The two data sets plotted
together may not be homogeneous, either. The
USEPAlUSFWS data were based upon fish collected
from a single Lake Michigan location (off Saugatuck,
Michigan). In comparison, the MONR data are for
relatively small sample sizes collected at a number of
southern Lake Michigan locations.
Simulated trout concentrations for both pelagic and
benthic/pelagic food chains are plotted in Figure
1.32. Predicted trout bioaccumulation for the
coupled benthic-pelagic food chain lags the
prediction for the pelagic food chain, and benthic
coupling lowers trout concentrations. This is contrary
25
c; 20
CI
2:
c;
o
~ 15
i:
Qj
u
c;
o
u 10
III
(J
a.
- pelagic food chain
.......... coupled benthic/pelagic food chain
"5
g 5
o
1940
1970
1980
1990
1950
1960
year
Figure 1.32. Sensitivity of trout PCBs predictions
to the food chain.
to the expectation that a benthic food chain linkage
would result in greater bioaccumulation. At steady-
state, higher trout concentrations are predicted for
the benthic-coupled food chain; however, this
condition does not occur in the dynamic simulation.
This discrepancy relates, in part, to the dynamics of
benthic versus pelagic exposure concentrations.
Sediment PCBs concentrations, which provide the
additional chemical exposure to the benthic-coupled
food chain, lag significantly behind the water column
concentrations. Polychlorinated biphenyl loading
significantly declines before sediment concentrations
approach steady-state with the maximum load; this
"hysteresis" is reflected in the bioaccumulation
predictions. Because the pelagic-based food chain
results better match the data, they will be presented
as dynamic model results for the remainder of this
report.
Simulated PCBs concentrations in other age classes
of lake trout were also verified. Age class
simulations are plotted with data for 1971 (Thomann
and Connolly, 1984) in Figure 1.33, indicating very
good agreement except for ages two and three. A
number of explanations have been offered for the
lack of fit for young trout, including problems with age
classification of fish based upon age-weight
relationships, the impact of reproduction on chemical
concentrations, and different (higher) exposure
36
-------
30
25
. 1971 data (:t1sd)
o model
a;
e;,
~ 20
c
o
~
~ 15
(J
c
o
(J
CD 10
u
Cl.
5
o
2
3
4
5
6
7
8
9
age class
Figure 1.33. Verification of PCBs accumulation
predictions in lake trout age classes 2-9.
environments for young versus adult trout. The
annual simulation of age two through 12 trout for the
period 1980-1990 is plotted in three dimensions in
Figure 1.34, showing the decline in PCBs
concentrations for all age classes of fish over that
period. Figure 1.35 displays the same model
simulation on a finer time scale; although the trend in
PCBs age class concentrations is downward, the
concentration for an individual trout cohort over this
period is still increasing. According to Thomann
(1989), the variation between age classes of fish
'" ~
'"
'"
o 2:
25 '" c:
.9
20 ~ ~
c:
52 Q)
15 u
c:
o
'" u
10 aJ
()
o a..
5
trout age class
year
Figure 1.34. Simulation of PCBs concentrations
in age 2-12 trout, 1980-1989.
30
25
~ 20
-=
c
.2
~
C
Q)
u
c
o
u
en
U
Q.
o
1980
1986
1988
1990
1982
1984
year
Figure 1.35. Simulation of PCBs concentrations
in age 2-12 trout, 1980-1990.
largely disappears when chemical concentrations are
normalized by lipid content. This was the justification
used to simplify the steady-state model to predict
only a single concentration in trout.
The predicted PCBs concentration in bloater is
plotted with available data (Hesselberg et al., 1990)
in Figure 1.36. Because both lipid content and size
(and presumably age) of the collected bloater have
substantially declined through time, the
concentrations were normalized for lipid content. The
agreement of model predictions with these data is
good; in fact, the prediction matches the bloater
PCBs concentrations exactly in the early 1970s, and
again in the mid-1980s. Between these periods,
predicted concentrations decline gradually, while the
data show a rapid concentration drop followed by a
gradual increase after 1980. As discussed by
Hesselberg, the dynamics of PCBs bioaccumulation
in bloater may be driven more by ecological stress
than by chemical exposure. Given that MICHTOX
simulates a fixed food chain structure and constant
parameterization of consumption and growth, such
factors cannot be accommodated in the simulation.
However, the general agreement with the bloater
PCBs data validates the benthic coupling of the
bioaccumulation model because bloater consume
Diporeia as well as Mysis.
37
-------
40
~
:9-
~
en 30
2:
c
.2
~
c:
~ 20
c
o
u
IX)
u
a..
4i
(;j
o
:0
- age 3 prediction
o
data (:t1sd)
2 2
Q
10
o
1970
1985
1990
1975
1980
year
Figure 1.36. Verification of PCBs concentration
predictions for bloater.
The validation of MICHTOX using PCBs data was
considered to be successful based upon the
qualitative agreement between model and data.
Because the available data were inadequate to
calculate average concentrations consistent with
model state variables, no quantitative or statistical
validation was possible. In particular, differences in
concentration trajectories for trout suggest a need for
improvement of this aspect of model predictions.
This suggestion will be pursued later in the report.
1.5.3.4 Bioaccumulation
Toxic chemical data to validate bioaccumulation
simulations at all trophic levels and for chemicals
other than PCBs, were generally unavailable for Lake
Michigan. Only limited measurements of PCBs,
toxaphene, and DDT in lower trophic levels could be
found (Evans et al., 1991). These data are
compared to model predictions in Figure 1.37.
Bioaccumulation model predictions were scaled to
match the concentrations reported for plankton; thus,
the comparison is of relative accumulation above that
trophic level. The model predicts PCBs accumulation
very well. The model is not so successful in
predicting accumulation of toxaphene, which is
underpredicted in fish, and of DDT, where all trophic
levels are underpredicted. The quality and
representativeness of these data are questionable,
however, because sample sizes were apparently
small and variability in concentration measurements
were not reported.
The most useful available data set for validating toxic
chemical bioaccumulation simulations in the Great
Lakes was developed for Lake Ontario (Oliver and
Niimi, 1988; Oliver et al., 1989). These data included
concentrations of PCBs and other HOCs in water,
sediment, and biota, including all trophic levels.
When normalized for the difference in exposure
(water and sediment) concentrations between the
two lakes, these data can be used to validate the
M ICHTOX steady-state bioaccumulation predictions.
The MICHTOX predictions of lake trout
bioaccumulation, expressed as bioaccumulation
factors (BAF = v/(cd» for each toxic chemical are
plotted with the Lake Ontario data in Figure 1.38.
The predicted BAF for each toxic chemical is also
tabulated in Table 1.8. Bioaccumulation factors are
plotted as a function of log Kow. the chemical-specific
parameter used to define toxicokinetic parameters in
the model. The predicted trout bioaccumulation
factors increase with log Kow in agreement with the
Lake Ontario data. The three toxic chemicals which
diverge from this pattern, TCDF, BaP, and TCDD,
are those which metabolize in fish. Although no data
could be found to verify the BaP bioaccumulation
simulation, the predictions for TCDF and TCDD can
be verified. This requires that biota concentrations
be normalized to those in sediment instead of water,
because water concentration data for PCDFs and
PCDDs are not available. This normalization is the
biota-to-sediment ratio (BSR = v/(r s»' BSR data and
predictions are plotted as a function of 109 Kow in
Figure 1.39; PCDD/PCDF data are from Carey at al.
(1990) and DeVault et al. (1989). The distinctly lower
accumulation of PCDFs and PCDDs apparent in the
Lake Ontario data is reflected in the BSR predictions
for TCDF and TCDD.
1.6 Steady-State Model Applications
The steady-state model was used for four
applications which explore different aspects of
MICHTOX. First was to predict the concentrations
expected in response to a unit load of each critical
pollutant to the lake. Second, the steady-state model
was used to quantify the magnitude of fate and
transport fluxes of each toxic chemical in the mass
balance model. This serves as a starting point
38
-------
1.0
. data (Evans et ai., 1991)
-.. 0 MICHTOX
Q> 0.8
C>
~
-
c
0 0.6
-.;::;
C'O
....
+-'
C
Q)
u 0.4
c::
0
U
ca
() 0.2
£l.
0
plankton Mysis Diporeia sculpin/alewife
0.6
-..
Q> 0.5
C>
~
-
c 0.4
0
:;
m
....
- 0.3
c
Q)
u
c
0 0.2
u
t-
0
Q 0.1
0
plankton Mysis Diporeia sculpin/alewife
0.6
.-..
C>
--
C> 0.5
~
-
c
0 0.4
+:;
en
....
+-'
C
Q) 0.3
u
c
0
u
Q) 0.2
c
Q)
..c
a. 0.1
C'O
x
B
0
plankton Mysis Diporeia sculpin/alewife
Figure 1.37. Validation of bioaccumulation predictions for lower trophic levels in Lake Michigan.
39
-------
9
° MICHTOX
00 0
8- 0 LO PCB congeners
0
. LO other HOCs 0 f °
o..~ go
0
7 to
u.
-------
3
2
oClJqg~
ot 1 0 Do aD 0
a::: 0
en 0 0
CD 0
"5 0 OTCDF TCDD
o. 0
e Cb . . .
....
en B(a)P
.Q 0 0
-1 . .
.
. .
-2
o MICHTOX
o LO PCBs/other HOCs
. LO PCDDs/PCDFs
.
-3
4
5
6
7
8
9
109 Kow
Figure 1.39. MICHTOX predicted trout biota-to-
sediment ratio and data from Lake Ontario.
towards understanding differences in the mass
balance results between chemicals, which can be
quite large. The third application was sensitivity
analysis, which displays the effect of varying
individual model parameters upon predictions.
Finally, the steady-state model was used as the
principal vehicle for uncertainty analysis. Uncertainty
analysis provides quantitative estimates of the
uncertainty of model predictions.
Steady-state spreadsheet model output for each toxic
chemical is attached as an appendix to this report.
The spreadsheets document model input
parameters, intermediate computations, results in
terms of chemical concentrations in water and
sediment in each model segment, and biota
concentrations in southern Lake Michigan.
1.6.1 Steady-State
Predictions
Load-Response
Steady-state is the condition where concentration
change (dc/dt) in response to a constant loading
becomes negligibly small. Predicted steady-state
concentrations of each toxic chemical in water,
sediment, and lake trout are tabulated in Table 1.9.
These are chemical concentrations predicted for
southern Lake Michigan for a total unit loading to that
segment of 1 kg/d. At steady-state, the model
predicts a linear relationship between total loading
and concentration. Thus, the results for a loading of
1 kg/d may be proportioned to any other load. This
can be represented as a load-concentration diagram;
the relationship for PCBs in water, for example, is
plotted in Figure 1.40.
1.6.2 Mass Fate and Transport
The mass balance model works by computing the
flux of chemical lost from the system (fate) and the
fluxes transported between model segments. In the
steady-state model, the total chemical load must be
balanced by the losses and net transport. Therefore,
the magnitude of all mass fluxes in a segment or
segments of the model may be expressed as
fractions of total load. This allows convenient
comparison of the magnitude of chemical fluxes for
different processes and for different chemicals. Such
a comparison of steady-state mass fluxes for toxic
chemicals in southern Lake Michigan is presented in
Table 1.10. These fluxes were taken from model
simulations made with expected air concentrations
(Table 1.3) and, for PCBs, 1990 tributary loadings as
well. The largest mass fluxes are internal cycling
associated with particles (settling and resuspension)
and particle burial, volatilization (absorption was
included in the load), and, for BaP, photolysis. The
other transport fluxes - advective transport,
dispersive exchange, and sediment-water diffusion-
are, in comparison, small to negligible. The mass
balance processes identified as having the largest
mass fluxes for a chemical also generally control the
model predictions.
The magnitudes of fate and transport fluxes vary
between chemicals, as shown in Table 1.10. They
also vary significantly between different model
segments. In Table 1.11, fate and transport fluxes
for PCBs are presented for southern Lake Michigan
and central Green Bay. The particle-associated
fluxes of settling and resuspension, relative to
segment loading, are much greater in central Green
Bay. Transport, exchange, and volatilization fluxes
are all larger in central Green Bay, whereas the burial
flux is greater in southern Lake Michigan. In general,
more fate and transport fluxes "participate" in the
mass balance for the shallower Green Bay
segments. This can be seen in Figure 1.41, which
41
-------
Table 1.9. Predicted Steady-State Concentrations of Critical Pollutants In Lake Michigan for Unit
Loading
Chemical Water (pglL) Sediment (nglg) Trou' . nglg)
BaP 56 14 ' '
\ .J
Chlordane 140 31 680
Dieldrin 440 51 190
DDT 150 32 700
HCB 60 9.6 89
Heptachlor epoxide 200 20 47
PCB4 84 16 240
PCB5 86 26 1200
PCBs 85 21 770
TCDD 87 44 41
TCDF 250 43 67
Toxaphene 940 39 15
10000
::; 1000
0,
S:
c:
.2
~
C 100
II>
u
c:
0
u
B
ftI 10
~
1
.01 1 10 100
load 10 SlM (kg/d)
100000
:e; 10000
CI
..s
c:
.2
~
C 1000
II>
u
c:
0
u
:;
g 100
10
.01 .1 1 10 100
load 10 SlM (kg/d)
Figure 1.40. Load-co-
. i"ation relationship for PCBs in southern Lake Michigan.
42
-------
Lake Michigan Mass Fluxes Expressed as Fractions of Segment Load
Chemical Volatilization Photolysis Transport Exchange Settling Resuspension Diffusion Burial
BaP 1.ge-3 0.72 2.7e-3 4.8e-4 0.52 0.22 0.017 0.28
Chlordane 0.39 0 7.Oe-3 6.0e-3 1.1 0.49 0.037 0.60
DOT 0.20 0.18 7.2e-3 6.1 e-3 1.2 0.50 0.038 0.62
Dieldrin 0.038 3.5e-3 0.022 0.037 1.8 0.79 0.064 0.97
Heptachlor 0.61 0 9.8e-3 5.8e-3 0.72 0.31 0.026 0.38
epoxide
HCB 0.81 2.2e-3 2.ge-3 5.ge-4 0.35 0.15 0.012 0.19
PCB4 0.68 0 4.1 e-3 1.ge-3 0.58 0.25 0.019 0.31
PCB5 0.50 0 4.2e-3 1 .1 e-4 0.92 0.40 0.029 0.49
TCDD 0.042 0.11 4.2e-3 5.5e-3 1.6 0.68 0.049 0.85
TCDF 0.17 0 0.012 0.016 1.6 0.67 0.052 0.83
Toxaphene 0.24 0 0.046 0.037 1.4 0.60 0.063 0.75
Table 1.11. Comparison of PCBs Mass Fate and Transport for Critical Pollutants in Southern Lake
Michigan and Central Green Bay
Segment Volatilization Transport Exchange Settling Resuspension Diffusion Burial
Southern Lake Michigan 0.59 4.2e-3 1.0e-3 0.75 0.32 0.024 0.40
Central Green Bay
1.9 0.31 0.76 4.3 4.1 0.056 0.14
portrays the relative magnitude of fate and transport
fluxes for PCBs in southern Lake Michigan and
central Green Bay. Settling and resuspension, which
are internal cycling fluxes, have been removed from
this figure.
A factor which can lead to confusion in interpreting
model results (whether steady-state or dynamic) is
the treatment of absorption. The confusion arises
because absorption can be either treated as a
component of net volatilization (a flux) or as a part of
the atmospheric load. The motivation for the latter
approach is that the three air-to-water chemical
fluxes (wet and dry deposition and absorption) are all
modeled as proportional to air concentration, which
must be specified externally to this model. If the
three are lumped together, then the resulting "total
atmospheric load" is proportional to air concentration,
and all other fluxes are proportional to water
concentrations computed "inside" the model. If
absorption is excluded from the atmospheric load,
then atmospheric load will have an apparent greater
effect upon predicted concentrations than other load
southern Lake Michigan
central Green Bay
Figure 1.41. Relative magnitude of PCBs fate and
transport fluxes.
43
-------
components. Whether or not absorption is properly SLM
treated as a load or a boundary flux is academic; 200
however, it is important to understand the distinction 160
between atmospheric deposition and total :J
atmospheric load. The significance of this distinction ~
c:
is displayed in Figure 1.42 for PCBs in southern Lake .2 120
~
Michigan. Absorption is seen to be the largest air-to- 1:
'"
u
c:
water flux of PCBs, although the net volatilization flux 8 80
is comparatively small. S!
CII
~
40
1.0
0
xl10 xiS xl2 base 2x 5x 10x
case
0.5 CGB
:tJ 3000
CiI
~
)( net
~ i
I;: volatilization
10 0.0
U c: 2000
Cl. .Q
.... g
! c:
IU '"
! u
c:
0
u
'n; -0.5 Cii 1000
i;j
~
-1.0
Figure 1.42. PCBs air-water fluxes at steady-state
in southern Lake Michigan.
1.6.3 Sensitivity Analysis
Sensitivity analysis is a general method for model
calibration; here it is used to demonstrate how
MICHTOX is sensitive to individual chemical- and
system-specific model parameters using PCBs as an
example. The model was run repeatedly with a
range of values for the parameter of interest. The
change observed in model predictions provides an
indication of model sensitivity to that particular
parameter. Results of sensitivity analysis for PCBs
are presented graphically in Figures 1.43 through
1.58. In most cases, parameter values were varied
from one-tenth to ten times the estimated value,
which generated the expected ("base case")
prediction. Parameter values were varied
simultaneously in all model segments; results are
shown for southern Lake Michigan and central Green
Bay segments.
o
xl10
xiS
xl2
2x
5x
10x
base
case
Figure 1.43. Sensitivity of water concentrations
to Kow.
10000
8000
Oi
"0
.s
c 6000
Q
~
c
8
c 4000
0
u
'S
~
2000
0
xl10 xl5 x/2 base 2x 5x 10x
case
Figure 1.44. Sensitivity of trout concentrations to
Kow'
44
-------
ClM SlM
500 300
~ 400 ::;-
~ 200
.a co
c: .2
.12 300 1i!
i! "
c :s
2! co
0
c: 200 '-'
8 $ 100
f '"
~
100
0
0 xl10 xl5 xl2 base 2x 5x 10x
xl10 xl5 xl2 base 2x 5x 10x case
case
CGB
CGB 1200
5000
1000
::;-
4000 C,
::; ..e,
:g co 600
.2
c: 1i!
.12 3000 " 600
i! :s
c co
.. 8
u
c: ~ 400
8 2000 '"
Ii; ~
iii 200
~
1000
0
xl10 xiS xl2 base 2x 5x 10x
o case
xl10 xl5 xl2 base 2x 5x 10x
case Figure 1.47. Sensitivity to dry deposition
Figure 1.45. Sensitivity to Henry's constant. velocity.
SlM
SLM 200
200
160 ;;[ 150
i ~
c:
g .2
120 Jg
.!: ~ 100
ij 8
8 60
S '"
~ ~ 50
40
0 0
x/5 x/2 base 2x 5. xl10 x/5 xl2 base 2x 5x 10x
case case
CGB CGB
3000 1000
i ;;[ BOO
2000 ~
j c:
g g 600
~ .J1!
~
8 8
.'!! 1000 400
~ '"
1ii
~
200
0
xl5 xl2 base 2x 5x 0
case xl10 x/5 xl2 base 2x 5x 10x
case
Figure 1.46. Sensitivity to volatilization rate. Figure 1.48. Sensitivity to sediment-water
diffusion coefficient.
45
-------
Ci
Q
S.
c
o
~
C
Q)
u
c
o
u
:;
g
3000
2000
1000
o
0.2
base
case
0.8
0.4
Figure 1.49. Sensitivity of trout concentration to
chemical assimilation efficiency.
Ci
Q
S.
c
.2
e
c
~
c
o
u
:;
g
2000
1500
1000
500
o
0.0
0.2
0.4
base
case
0.8
Figure 1.50. Sensitivity of trout concentration to
pelagic diet fraction of forage fish.
-
SLM
200
i 160
e:
.51 120
i!
C
..
g
8 80
.!i
..
~
40
0
xl10 xl5 xl2 b... 2x 5x 1ax
ce..
CGB
1000
;;! 800
~
B 600
I!!
C
f!
e:
0 400
u
.!i
~
200
0
xl10 xl5 xl2 b... 2x 5x 1ax
ce..
Figure 1.51. Sensitivity to dispersive exchange
coefficient.
SLM
400
i 300
~
~
~ 200
8
.!i
..
~ 100
0
xl5 xl2 ba.. 2x 5x
ca..
CGB
4000
~ 3000
~
~ 2000
8
e:
8
;;;
-; 1000
0
xl5 xl2 bliSS 2x 5x
case
Figure 1.52 Sensitivity to suspended particle
concentration.
46
-------
SLM SLM
300 200
i ~ 150
~
200 c
c .2
0 ~
g 100
c:; g
2!
c:; 8
8 S
S 100 co
; 3: 50
0
0 0.055 0.100 0.150 base 0.2 0.25
xl10 xl5 xl2 base 2x 5x lOx case
case
CGB
CGB 1200
1000
1000
800 ~
i ~
c 800
.2
c:; I
.2 600
i! 600
! c
8
c:; ~ 400
8 400
!i 3:
~ 200
200
0
0.055 0.100 0.150 base 0.2 0.25
0 case
xl10 x/5 xl2 base 2x 5x 10x
case Figure 1.55. Sensitivity of water concentrations
Figure 1.53. Sensitivity to particle burial velocity. to suspended particle foc.
SLM
50
SLM
200
c;; 40
~
~ 150 c
8 .2
}g 30
c c
.2 8
i! c:;
C 100 8
~ C 20
0 "
... E
;; '6
n; 51 10
3: 50
0
0.055 0.100 0.150 base 02 025
o case
xl10 xl5 xl2 base 2x 5x lOx
case
CGB
CGB 300
1000
~
i 800 .::.
c: 200
.2
i!
c:; C
.2 600 "
i! ...
c:
C 0
8 ...
c
c: " 100
8 400 E
J!! '6
; 51
200
0
0 0.055 0.100 0150 base 0.2 0.25
case
xl10 xiS xl2 base 2x 5x lOx
case Sensitivity of sediment
Figure 1.56.
Figure 1.54. Sensitivity to particle settling concentrations to suspended particle foc.
velocity.
47
-------
2000
"3
0,
-S
1500
1 .
~ "....,i(~
1000 11,0,:;' ,..'
1',\,,:', ;T--
-,;'::
'..;,~ .~
"
Q
'-'~-"
''='
>~~.:.
" ~<
~
8
"
8
:;
~
'7
t -.. . ';:~
I;
. .- "~~,
',~'"
'.
--,~
:~!"t; ~
500
~~.:.~.'.
.,,.:;~-
.t-~l
o
0055
02
025
0150
0100
base
C8se
Figure 1.57. Sensitivity of trout concentrations to
suspended particle foe.
10COCO
I
PCB tnbulary load distributions
. accord,ng to Marti & Armstrong
. proportioned by flow
o proportioned by surface area
o ali Fo> R"er
~, 1C00V
'"
c
Q
~
c
t 1000
c
o
u
r.
-------
Sensitivity of model predictions to circulation
transport was evaluated by varying the dispersive
exchange coefficient. Figure 1.51 displays the
sensitivity to this parameter. In southern Lake
Michigan, the coefficients must be increased ten-fold
to observe any response, reflecting both the relative
insignificance of horizontal exchange here and the
lack of a chemical gradient between the main lake
segments. In central Green Bay, the model is more
sensitive to horizontal exchange, the concentration
gradients are greater, and the circulation is a more
significant mass balance process. Concentrations in
central Green Bay decrease for exchange
coefficients both lesser and greater than the
expected values, a sensitivity more complex than
observed for other model parameters. At low
exchange values, the flux of PCBs to central Green
Bay from the Fox River and inner Green Bay is
retarded, which lowers central Green Bay
concentrations. At high values of exchange, PCBs in
Green Bay are substantially diluted by main lake
water, again lowering central Green Bay
concentrations.
Model sensitivity to several particle-related
parameters was also tested. Sensitivity to
suspended particle concentration (M) is plotted in
Figure 1.52. Water concentrations increase with M
because of the shift in particle-sorbed chemical from
the sediment to the water column. The dissolved
chemical fraction is also lowered and, hence,
volatilization. Sensitivity of trout concentrations to M
(not plotted) is virtually negligible. The increase in
water concentration with M is offset by the decline in
the dissolved chemical fraction, so that the chemical
exposure to biota is relatively unaffected. Figure
1.53 displays the model sensitivity to particle burial
velocity; increasing the burial velocity reduces the
water concentration. Because burial is a
predominant loss process for PCBs in southern Lake
Michigan, sensitivity to burial velocity is observed
across the range of parameters. In central Green
Bay, however, sensitivity is only observed for high
burial velocities. Sensitivity to particle settling
velocity is plotted in Figure 1.54. As settling velocity
increases, the solids balance requires resuspension
to increase as well. Thus, the intensity of particle
mixing between water and sediment grows with
increasing settling velocity. Because particulate
chemical concentrations (r) are nearly equal in the
water and sediment at steady-state, the model's
sensitivity to this parameter is minimal.
The sensitivity of model predictions to suspended
particle organic carbon was also evaluated because
it is the organic carbon fraction of particles which
actively sorbs HOCs. The model sensitivity to foe is
somewhat different than that to M because organic
carbon is not constrained by a mass balance in this
model. The development of a carbon-based particle
balance was a principal goal of the modeling effort
for the GBMBP. Sensitivity of water concentrations
to foe is displayed in Figure 1.55, indicating different
responses in southern Lake Michigan and central
Green Bay. In southern Lake Michigan water,
concentrations decline slightly with increasing foe, as
the flux of chemical settling to the sediment
increases. In central Green Bay, there is a much
greater increase in concentration with foe, which
lowers dissolved chemical fractions and volatilization
loss. In the sediment (Figure 1.56), however, both
segments show an increasing concentration with foe.
Figure 1.57 sensitivity of trout concentrations to foe;
as was the case for M, biota concentrations are fairly
insensitive to this parameter.
One additional factor, the spatial distribution of
tributary loading, was evaluated for model sensitivity.
Loading distribution may be expected to affect the
distribution of chemical concentrations throughoutthe
system. Thus sensitivity in each model segment was
evaluated. Several possible loading distributions
were considered, although the total PCBs tributary
loading to the lake, 1 kg/d, was not varied. These
included the distribution based upon Marti and
Armstrong's (1990) tributary sampling, distributions
based upon tributary flow to each segment and
segment surface area, and distributions based on
allocating all loading to the Fox River. The results
appear in Figure 1.58. Although water column
concentrations are plotted, sediment concentration
distributions were similar. Concentrations in the main
lake segments (1, 2, and 3) are relatively insensitive
to load distribution and show little spatial gradient.
Because of their large surface areas, these
segments receive much of their PCBs load from the
atmosphere. Thus their insensitivity to tributary load
distribution is not unexpected. In Green Bay
segments (5, 6, and 7), both sensitivities to tributary
load distribution and the spatial concentration
gradients are much more pronounced.
49
-------
Concentrations are sensitive to the distribution of
tributary (or other spatially variable) loading in these
segments primarily because tributary load makes up
most of the total chemical loading to Green Bay and
because horizontal transport and exchange are more
important processes in these relatively isolated
segments. The "all Fox River" case demonstrates
the predicted effect of PCBs loading from that
tributary to the entire lake. Much of PCBs mass from
that source has been lost (principally by volatilization)
before it reaches the main lake. Of the 1 kg/d
loading to the Fox River, only 31 g/d are predicted to
reach the main lake.
A number of general observations about model
behavior may be drawn from the sensitivity analysis.
The first is that sensitivity to parameters do not vary
uniformly throughout the model's segments and state
variables. Rather, the magnitude and even the
direction of changes in concentration vary according
to segment and state variable. The observed
sensitivities are strongly related to the relative
magnitude of the various chemical fluxes in a
particular model segment. Parameter sensitivity is
otten not uniform across a range of parameter
values. Finally, biota concentrations are largely
sensitive to bioaccumulation model parameters but
are less sensitive to parameters of the mass balance
model.
1.6.4 Uncertainty Analysis
Model predictions may be erroneous for a number of
reasons, including parameterization errors,
conceptual and descriptive errors, and algorithm
errors (bugs). Calibration and verification procedures
are usually relied upon to detect and correct such
errors in water quality models. Because extensive
calibration/verification of this model was not possible,
the possibility of undetected errors makes MICHTOX
predictions uncertain. Uncertainty analysis was used
to address and quantify uncertainty in model
predictions, particularly relating to parameterization
errors. Conceptual and descriptive errors in the
model were neglected because these factors relate
to possibilities which would change model results to
an unforeseeable extent. Uncertainty due to these
errors can only be identified by more comprehensive
model calibration and verification. Modeling quality
assurance, hopefully, prevented major mishaps due
to bugs.
For the most part, uncertainty analysis was
performed on the steady-state model. This was a
choice constrained by the 8000 model runs
necessary to perform the analysis. A more limited
analysis of the dynamic model was performed; this is
described in a later section of the report.
1.6.4.1 Analysis 01 Model Uncertainty
Uncertainty analysis was performed by the Monte
Carlo method. This method allows direct analysis of
the consequence of model parameter uncertainty
because the model is used to compute changes in
concentration resulting from changes in parameter
values. This is achieved by performing repeated
simulations of the model with randomly selected
values from defined probability distributions. For
each simulation, parameter values are defined as
random variables whose distribution is a measure of
uncertainty in the real but unknown value of the
parameter. In this application, parameter variability
was assumed to be uncorrelated, and parameter
values were chosen at random from specified (exact)
frequency distributions. This is known as the Latin
Hypercube method (McKay et al., 1979). The
process is repeated for a number of iterations
sufficient to converge upon an estimate of the
frequency distribution of the output variables. Monte
Carlo analysis allows a probabilistic statement 01
uncertainty to be made because a distribution of
model predictions are produced (Gardner and O'Neill,
1983). Details of the implementation of uncertainty
analysis for toxic chemical models is provided in
Endicott et al. (1990, 1991). Parameter distributions
were formed from data available in the literature and
from experience gained in calibrating other models.
Probability distributions for model parameters treated
as uncertain are tabulated in Table 1.12. Because
many of these parameters had expected values that
varied between model segments or trophic levels,
scale factors were used to simultaneously vary the
parameter values. For example, flows between
water column segments were varied by multiplying
the expected value 01 each flow by the scale factor,
the probability distribution of which is found in the
table. In this way, parameter variability in each
segment and trophic level was generated, and
random selection of every parameter value was
avoided. This reduced the number of parameter
selections per iteration from 80 to 31.
50
-------
Table 1.12. Probability Distribution for Steady-State Model Uncertainty Analysis
Group Parameter Type" Distribution Mean CV 95%CI
Lake Flow SF LN 0.10
circulation Dispersive exchange SF LN 0.10
Pore water diffusive exchange PV LN 3.0e-4 0.64 1.0e-4 1.0e-3
Particle Settling velocity SF N 0.11
transport Burial velocity SF LN 0.18
Suspended particle concentrations SF LN 0.35
Organic Suspended particle foe SF LN 0.25
carbon Water column colloidal binding efficiency PV LU 0.50 1.0
Sediment particle foe PV N 0.24 0.0017 0.047
Pore water dissolved organic carbon PV N 0.050 0.25 0.34 1.0
binding efficiency
Log Koe error RE N 0.67 0.30
Atmospheric Aerosol volumetric fraction PV LN 0.046 6.3e-12 6.3e-11
Particle scavenging ratio PV N 2.0e-11 0.37 55,000 350,000
Dry deposition velocity PV LN 200,000 0.43 86 430
190
Bioaccumu- Dissolved oxygen concentration PV N 0.10 8.0 12.0
lation Alewife pelagic diet fraction PV N 10 0.11 0.50 0.80
Lipid content SF N 0.062 0.11
Benthic log chemical assimilation RE N 0.26
efficiency error
Lake trout chemical assimilation efficiency PV N 0.085 0.50 0.70
Food assimilation efficiency SF N 0.06 0.064
Growth rate SF LN 0.18
Respiration rate SF LN 0.18
Chemical metabolism: BaP PV LN 0.64 0.0072 0.072
TCDOrrCDF PV N 0.023 0.36 0.0010 0.0060
Chemical uptake efficiency error RE N 0.0035 0.051
.SF: scale factor; PV: parameter value; RE: regression error; N: normal distribution; LN: lognormal
distribution; LU: log uniform distribution.
1.6.4.2 Results
Three different tests of predictive uncertainty were
performed on the steady-state model. The first test
performed for each toxic chemical evaluated
predictive uncertainty due solely to uncertain
parameters with a fixed air concentration of 1 ng/m3.
The second test treated air concentration as
uncertain, and was again performed for each
chemical. This represented the more realistic
condition where loadings as well as model
parameters are uncertain. The third test, performed
on PCBs, included a constant tributary load to
simulations in which air concentration and model
parameters were both treated as uncertain. This test
was repeated over a range of tributary loads; it
displays the predicted response (with uncertainty) to
partial control of loadings.
The output probability distributions for 200 iterations
were used to assure convergence in the uncertainty
analysis. The model output distributions were
approximately lognormal for all chemicals and all
tests. The logarithmic means of the Monte Carlo
output distributions agreed with predictions of the
steady-state model made with expected (mean)
values of all parameters. Thus, the results of the
uncertainty analysis were consistent with the central
limit theorem.
Results for the first test are summarized in Table
1.13, including logarithmic mean, logarithmic
51
-------
Table 1.13. Summary of Results for First Test of Model Uncertainty.
Concentrations for Fixed Air Concentrations of 1 ng/m3 for Each Chemical
Predicted Steady-State
Water Concentration Lake Trout
(pglL) 95%CI Concentration (nglg) 95%CI
Chemical Log Mean InCV LCL* UCL ** Log Mean InCV LCL UCL
BaP 9940 0.46 410 2500 26 1.9 2.2 270
Chlordane 1200 0.40 560 2500 4900 1.8 330 3.5e+04
DDT 1900 0.43 870 4400 8000 1.3 910 4.ge+04
Dieldrin 7200 0.82 1900 31000 2700 2.8 140 4.8e+04
HCB 32 0.90 6.7 140 42 1.7 4.3 410
Heptachlor epoxide 1500 0.74 370 5000 340 2.5 20 5000
PCBs 450 0.66 130 1350 3200 1.0 550 1.6+04
TCDD 1600 0.44 730 3700 730 1.2 150 5600
TCDF 3900 0.55 1600 12000 1000 1.2 130 5000
Toxaphene 11 000 4100 4100 31000 190 1.1 41 1300
.Lower confidence limits
..Upper confidence limits
coefficient of variation, and 95% confidence intervals
of the distribution of model predictions for water and
trout in southern Lake Michigan. The results of this
test show that the uncertainty of water and,
particularly, trout predictions are very large for each
channel. The widths of the 95% confidence intervals
for predicted water concentrations generally span a
factor of 10; HCB, dieldrin, and heptachlor epoxide
had greater uncertainty. For trout, confidence
intervals span widths varying from factors of 30
(PCBs) to 300 (dieldrin). Trout concentration
predictions are expected to be more uncertain than
those for water because uncertainty in the
bioaccumulation model amplifies the uncertainty
generated in the mass balance model.
Although these uncertainties are large, they are fairly
comparable to the variability reported for toxic
chemical concentrations in aquatic ecosystems. For
example, the coefficient of variance (CV) for
predicted PCBs trout concentrations is 1.0; if these
predictions are converted to bioaccumulation factors,
then the variability is reduced to a CV of 0.76. In
comparison, the CV calculated from 1971 Lake
Michigan fish data (normalized for lipid) is about 0.75
(Connolly, 1992). Across a range of ecosystems and
HOCs, significant variability in BAF is observed. For
chemicals in the range 6 < 10gKow < 7, Connolly
reports a bioaccumulation factor CV of 1.6. In this
context, the uncertainty of model predictions appears
more reasonable, and it reflects the magnitude of
variability in the data.
The results in Table 1 .13 may be used to define
confidence limits for the steady-state load-
concentration relationships because air
concentrations may be converted to total chemical
load. Such a result is displayed for PCBs in southern
Lake Michigan in Figure 1.59. The results also
predict the relative potential of each toxic chemical,
for a given air concentration, to accumulate in Lake
Michigan. Toxac'"'ene has the greatest accumulation
potential in wat"- -iCS has the least. Fortrout, DOT,
chlordane, PCf:. and dieldrin have the greatest
accumulation pOtential; SaP has the least.
In the second test, air concentrations were treated as
uncertain. Thus, greater predictive uncertainty was
expected. Results of this test, presented in -able
1.14, show that uncertainties have grown from the
previous test, particularly for those chemicals with
largely uncertain air concentrations. This is most
apparent for water concentration predictions with
much larger CVs for chlordane, DOT, TCOD, TCOF,
and, particularly, toxaphene. Uncertainty in trout
concentration predictions were generally less
affected by air concentration uncertainty, except for
dieldrin and toxaphene. Apparently, the impact of
uncertainty in air concentrations upon predictive
52
-------
10000
:::;
0,
.s
c;
.2
~
"E
'"
o
c;
o
o
1000
oS!
co
~
100000
10000
c;;
0,
.s
c; 1000
o
~
C
'"
0
c; 100
o
o
:;
g
10
1
01
.'
"
100
10
"""", .
- expected
- - - - - - 95% confidence limits
.1
.01
.1
1
load to SLM (kg/d)
10
100
".
,
".
-
,
.1
1
load to SLM (kg/d)
10
100
Figure 1.59. Load-concentration relationship for PCBs in southern Lake Michigan including
confidence limits.
Table 1.14. Summary of Results for Second Test of Model Uncertainty. Predicted Steady-State
Concentrations for Expected Air Concentrations of Each Chemical
Water Concentration Lake Trout
(pg/L) 95%CI Concentration (nglg) 95%CI
Chemical Log Mean InCV LCL UCL Log Mean InCV LCL UCL
BaP 510 0.57 1.7 1400 0.13 2.0 9ge-03 1.4
Chlordane 48 0.70 14 170 190 1.9 19 2300
DDT 57 0.74 16 210 240 1.4 26 1500
Dieldrin 240 1.00 48 1300 88 3.6 3.5 2000
HCB 1.4 1.1 0.22 7.1 1.9 1.7 0.19 17
Heptachlor 25 0.92 5.4 120 5.6 2.7 0.32 96
epoxide
PCBs 110 0.64 33 330 780 1.0 150 4000
TCDD 0.049 0.77 0.012 0.18 0.023 1.3 3.7e-03 0.20
TCDF 1.3 0.83 190 5.1 0.34 1.4 0.039 2.5
Toxaphene 2000 1.7 190 1 .8e+04 35 2.4 2.2 5010
53
-------
uncertainty varies among chemicals, and the impact
may be different in water and trout.
These results may also be considered to be
predicted steady-state concentrations for the toxic
chemicals, if the predominant continuing load is
atmospheric. Conversely, the results may be used to
test whether the estimated air concentrations for
each chemical are consistent with observed
concentrations in Lake Michigan. If observed
chemical concentrations fall outside the 95%
confidence limits of the predictions, then the
estimated air concentrations used in the model are
probably incorrect. Data used to evaluate the
predictions were obtained from a variety of sources
(Environment Canada, 1991 ; Michigan Department of
Natural Resources, 1990; U.S. Environmental
Protection Agency, 1989b; DeVault et al., 1986).
Results of this test suggest that BaP, TCDD, and
TCDF air concentrations are probably lower than
their expected values, and air concentrations of HCB
are probably higher. Observed concentrations of the
other chemicals fall within the confidence limits of the
predictions.
The third uncertainty analysis test was a modification
of the second, with a constant tributary loading in
addition to uncertain model parameters and air
concentrations. This test, performed for PCBs only,
was repeated over a range of tributary loadings from
0.02 to 2 kg/d. The spatial distribution of tributary
loads was the same described for PCBs verification.
Results of this test appear in Table 1.15. As tributary
loading increases, the uncertainty in predicted
concentrations declines. This is because the
influence of air concentration uncertainty upon
predictions diminishes as atmospheric deposition
becomes a relatively smaller load component in
comparison to the growing tributary load. In other
words, increasing the tributary load reduces total load
uncertainty, which was reflected in the uncertainty of
predictions. The second insight offered by this test
was that predicted concentrations do not respond
proportionately with tributary loading, independently
of air concentrations. This can be seen in Figures
1.60 (water) and 1.61 (trout) in which the results of
this test appear as load-concentration plots. As
tributary loading declines, so do water and trout
concentrations - but only so far. The factor which
limits the effectiveness of tributary loading reduction
is the air concentration. Large reductions in tributary
loading cannot be expected to produce similarly large
improvements in water quality if atmospheric loading
is not reduced as well. Because the estimated
tributary loading of PCBs is currently about 1 kg/d,
Lake Michigan is already in the situation where water
quality improvement for this chemical must come
largely from reductions in atmospheric loading.
1.6.4.3 Critical Parameterization Uncertainty
Given that uncertainties in model predictions are
undesirably large, how can the situation be
improved? The large predictive uncertainties may be
"reigned in" by further calibration and verification of
the model. This would require collection of additional
monitoring data for toxic chemicals, emphasizing
determination of chemical loads. Aside from
estimating the confidence intervals for model
predictions, uncertainty analysis can identify the
"contribution" of model parameters to uncertainty in
Table 1.15. Summary of Results for Third Test of Model Uncertainty. Predicted Steady-State PCBs
Concentrations for Varying Tributary Loads and Expected Air Concentration
Water Concentration Lake Trout Concentration
(pglL) 95%CI (nglg) 95%CI
Load (kgld)1 Log Mean InCV LCL UCL Log Mean InCV LCL UCL
0.02 110 0.63 34 330 780 0.99 160 4000
0.2 120 0.60 38 340 840 0.96 180 4200
2 190 0.46 81 450 1400 0.91 290 6000
54
-------
10000
prediction with constant air concentration.
- expected
1000 - 95% confidence limits
:J' ,
c. '
8 "
c "
,g 100 "
e . "
c ' . ' "
, . ,
8 ' . ,
" . .'
c " . ,
8 10 ' . ' "
, . ,
~ . ,
S . ' "
III . ' "
~ . ,
"
" no air concentration:
,
...... expected
- - - - - - 95% confidence limits
.1
,001 .01 .1 10 100
tributary load to SLM (kg/d)
Figure 1.60. Load-concentration relationship for
PCBs in southern Lake Michigan water: Effect of
constant air concentration.
100000
prediction with constant air concentration:
- expected
10000 - 95% confidence limits
,
~
.:.
c 1000
B
!!
C
8
c 100
o
u
"5
g
10
.
.
,
, .
, .
, .
, .
zZ;/'
, .' "
,'.# "
" ,
" ,
" # ..'
. ,
, . ,
, . ,
, . ,
, . ,
. ,
. ,
. ,
. ,
. .
. "
, .' no air concentration:
.. . ... expected
- - - - - - 95% confidence limits
1
.001
.01
10
100
.1
tributary load to SLM (kg/d)
Figure 1.61. Load-concentration relationship for
PCBs in southern Lake Michigan trout: Effect of
constant air concentration.
predictions. Therefore, an alternative approach to
reducing uncertainty would be to prioritize research
efforts intended to improve the accuracy of model
predictions by accurately measuring "critical"
parameters identified as major contributors to
predictive uncertainty. The optimum strategy for
reducing predictive uncertainty would be a
combination of the two approaches.
The degree of correlation observed between
parameter and prediction values in Monte Carlo
testing indicates the relative importance of an
uncertain parameter in contributing to prediction
uncertainty. The square of the correlation coefficient,
r2, is an estimate of the fraction of prediction
uncertainty attributable to each uncertain parameter.
Correlation analysis was performed on the results of
the second uncertainty analysis test (uncertain model
parameters/uncertain air concentrations) for five of
the toxic chemicals: SaP, chlordane, dieldrin, PCBS,
and TCDD. Results are presented graphically in
Figure 1.62 for water concentration predictions and
in Figure 1.63 for trout predictions. For both water
and trout, a relatively few critical parameters were
identified which contribute most of the uncertainty in
model predictions. The relative contribution of these
parameters to predictive uncertainty was, however,
fairly chemical-specific and related to the
hydrophobicity and reactivity of each chemical. As
expected, air concentrations contribute large
uncertainties to water concentration predictions.
Parameters related to air-water chemical flux -
scavenging ratio, Henry's constant, and aerosol
volume fraction - were also found to be significant
contributors to predictive uncertainty. Other critical
parameters for water prediction uncertainty include:
suspended particle concentration, particle burial rate,
and for BaP, photolysis rate. Significantly, several
parameters were not identified as critical: advective
and dispersive transport, air-water transfer
coefficients, and organic carbon fractions. The
1.0
08
c
"
'0
'"'
"
8 0.6
c
o
11\
!1
5
" 0.4
""0
e
..
"
r:r
II>
0.2
o air concentration
o sClV8nging ratio
~ photolysis rale
~ Henry's constsnl
OKOC
i!jJ particle bunal IIOlocily
. suspended particle cone
~ aerosol VOlume fradlOn
. NSOM b.ndlng efticiency
0.0
dieldrin chlordane B (a) P
PCBS
TCDD
Figure 1.62. Contribution of critical parameters
to steady-state model uncertainty: Water
concentration.
55
-------
10
08
o metabolism me
0KDW
Ii dry deposlbon velocrty
~ partide burial rate
o respinl\lOn rate
D ... concentration
~ Qlem. assm eft (benthos)
. Henry's constant
8KOC
. bIOta "pod fraalon
C
II
U
e
~ 0.6
c
.2
ii
!i
8 04
" .
II
iO
"
a
..
02
00
dleldnn chlordane B (a) P
PCBS
TCDD
Figure 1.63. Contribution of critical parameters
to steady-state model uncertainty: Trout
concentration.
organic carbon partition coefficient, while critical for
the less hydrophobic chemicals (dieldrin and
chlordane), was not identified as critical for the more
hydrophobic chemicals (BaP, PCBS, and TCDD).
Generally, the critical parameters are related to
chemical sources (in this case, atmospheric input)
and loss mechanisms (burial, photolysis).
For trout predictions, air concentrations are a lesser
but still significant source of uncertainty, except for
PCBS. Critical parameters for trout include Kow
(primarily by determining excretion rate) and other
bioaccumulation parameters: metabolization (for BaP
and TCDD), respiration, benthos chemical
assimilation efficiency, and lipid fraction; burial rate,
dry deposition velocity, Henry's constant, and I
-------
transport parameters, are expected to be
substantially cross-correlated. Neglecting this
covariance inflates the uncertainty estimates.
Parameter covariance may be estimated during
model calibration. Application of these methods was,
however, considered beyond the scope of this effort.
1.7 Dynamic Model Applications
The dynamic MICHTOX model was used to predict
temporal changes in toxic chemical concentrations
under transient conditions. The most important of
these is the prediction of "lag time," the period of time
for concentrations to respond to a change in loading.
Dynamic simulations were also made for several
PCBs management scenarios and for a hypothetical
severe storm "event." Finally, uncertainty in the
dynamic simulations will be examined, including
consideration of factors not included in the model
framework.
1.7.1 Toxic Chemical Lag Time
The most important dynamic prediction is Jag time.
It is the time required for a specific load reduction to
achieve a desired water quality objective, such as
attaining a water quality standard or lifting an
advisory on fish consumption. The lag time depends
upon the magnitude of the loading reduction, the
difference between present and desired
concentrations, and the intrinsic responsiveness of
the system which is predicted by the model. Greater
loading reductions will result in shorter lag times,
although the difference may be small, and lag times
will be longer to achieve lower concentration
objectives.
where:
cj = initial concentration before load reduction
cf = final steady-state concentration at new load
for which the rate of change dx/dt is independent of
the magnitude of load reduction, and is determined
solely by the model parameterization for each toxic
chemical. Thus, results from one dynamic simulation
may be used to predict dx/dt for load reductions of
any magnitude for each toxic chemical. Since dx/dt
equals dc/dt if loading is completely eliminated, this
"cutoff" simulation was used to determine rates of
concentration change for each toxic chemical: a
constant loading was eliminated after steady-state
had been reached, and the subsequent decline in
concentrations simulated for the next 30 years.
Results of this simulation for each toxic chemical are
presented in Table 1.16; the results for PCBs are
also plotted in Figure 1.64. Simulated concentrations
initially decline rapidly, particularly in water.
Thereafter, the concentration decline approaches a
steady, first-order loss rate in each media.
Chemicals are ordered in the table according to this
chemical loss rate for water and sediment, with HCB,
PCBs, SaP, and heptachlor epoxide concentrations
declining most rapidly.
Table 1.16. Predicted Long-Term Chemical Loss
Rates for Lake Michigan Critical Pollutants
Following Loading Reduction. First-Order Loss
Rates in Units of 1Near in Southern Lake
Michigan
The rate of concentration change (dc/dt ) in response Chemical Water Sediment Trout
to a reduction in loading is determined by the HCB 0.16 0.16 0.16
magnitude of the load reduction as well as the model PCBs 0.15 0.15 0.12
parameterization. Accordingly, dc/dt will be different BaP 0.14 0.14 0.14
for every load reduction, which severely limits the Heptachlor epoxide 0.12 0.13 0.12
Chlordane 0.091 0.093 0.092
generality of model predictions. However, a relative DDT 0.085 0.086 0.086
concentration x may be defined: Toxaphene 0.063 0.063 0.064
c - cf TCDF 0.063 0.063 0.063
x = TCDD 0.061 0.060 0.061
c; - cf Lead 0.046 0.047
Dieldrin 0.040 0.041 0.041
57
-------
10
.......... water
\
\
\
: \
Qj' 0.8 \ '.
u ,\
I : \
o ~ \
g 0.6 \ '.
~ \
---- sediment
- trout
Qj'
U
,
~ 0.4
II
x
0.2
0.0
o
5
10
15
20
25
30
years after loading reduction
Figure 1.64. Predicted PCBs time response to
loading reduction in southern Lake Michigan.
The practical significance of lag time may be
demonstrated in terms of the load-concentration
relationship. At steady-state, concentrations are
proportional to total loading. Thus the load-
concentration relationship is linear: a 50% reduction
in concentration is achieved by cutting the total
chemical load in half. However, this outcome will
only be achieved after a relatively long time. The
maximum rate of concentration decline is controlled
by the lake itself, and the effectiveness of loading
reduction is constrained by this rate. This constraint
causes the load-concentration relationship to diverge
from linearity, as shown in Figure 1.65. For a given
time after reducing total lake loading (0, 5, and 10
years), PCBs in trout will not decline below a limiting
concentration, even if total load is reduced to zero.
As time after reducing total load increases, the
limiting trout concentration will decline, allowing
smaller loads to become effective. Every five years,
the limiting trout PCBs concentration decreases by
about 50%. Only after a very long time will the load-
concentration curves converge upon the steady-state
relationship.
The loss rate predictions for concentrations in trout
may be applied to tentatively evaluate available
monitoring data for priority toxics in Lake Michigan.
By comparing the load "cutoff' predictions to the
trend in concentration data, the hypothesis that
priority toxicsloading has been essentially eliminated
may be tested. If concentrations appear to decline
-
10000
~.
~.
~.
~
----:;
---------- ~
....~.
.......................... .~
.~
.~
.~
. .#'
...-.- ~..
~.
~.
~.
:e;
Cj
.s
c:
.2
e
"E
Q)
u
c:
o
u
:;
g
1000
100
time after
reducing loading:
- Oyrs
---.Syrs
.......... 10 yrs
-- --. 20 yrs
_'_1' S5
10
.1
.
1
10
100
tributary load to SLM (kg/d)
Figure 1.65. Response of trout PCBs
concentrations at various times after reducing
PCBs load (ss = steady-state).
as rapidly as the load "cutoff' predictions, then it may
be concluded that the rate of decline is controlled by
the long-term loss rates of the system. Figure 1.66
presents such a comparison for chlordane (data
sources are indicated on figures). Although no clear
trend of declining concentration is apparent in
chlordane data, the slow rate of predicted
concentration decline following load cutoff suggests
that such a trend could be obscured by the variability
of these data. Comparison of data to the load cutoff
prediction for DDT is displayed in Figure 1.67.
Because DDT concentrations are declining, at least
as rapidly as the load cutoff prediction, DDT loading
in recent years appears to be negligible. Essentially
the same conclusion is drawn for dieldrin (Figure
1.68), PCBs (Figure 1.69), and TCDD (Figure 1.70).
For TCDF, the comparison leads to an ambiguous
result (Figure 1.71) particularly because the
variability associated with this data is unknown.
Generally, however, it appears that recent trends in
trout concentration for these chemicals are
consistent with model simulations of load "cutoff." It
cannot be concluded from this that loads are zero;
rather, it suggests that present loads are small in
comparison to past loadings which have resulted in
extensive sediment contamination. Further, this
sediment reservoir maintains water column and biota
chemical concentrations at their present levels. At
58
-------
1.2
1.0
c;; 0.8
0,
2.
c:
0
'=" 0.6
~
c
.,
u
c:
8 0.4
.,
c:
..
"E
0
:2
u 0.2
0.0
M
co
0'1
.....
. age 6 trout data
o age 7 trout data
.. .. .. .... age 7 prediction
..r
CO
0'1
....
Il')
CO
0'1
....
to
CO
0'1
....
,....
CO
0'1
....
CO
CO
0'1
....
0'1
CO
0'1
....
year
Figure 1.66. Simulation of chlordane in Lake
Michigan trout (Michigan Department of Natural
Resources, 1990).
25
20
~
.!!>
CI
'"
~ 15
o
:;:>
!!!
"E
B
c:
8 10
I-
o
o
'.
5
o
o - N
... ... ...
2! ~ ~
o EPAlFWS data
III age 6 trout data
. age 7 trout data
- - - - - age 7 prediction
'.
.
'. ..!
'-
"'''",,,,,,,
.....---.........
CO') "II;f' lot) (C ,....
..... ..... ..... ..... ,....
~ 2? 2? ~ 2?
ex> 0) 0 ..-
..... to-. co co
2? 2? 2? 2?
N M ...
ex> ex> ex>
2? 2? 2?
year
Figure 1.67. Simulation of DDT in Lake Michigan
trout (DeVault et al., 1986; Michigan Department
of Natural Resources, 1990).
~
C>
2-
.~ 0.6
~
"E
OJ
to
c:
a
to 04
c:
~
a;
'i5
1.0
0.8
. age 6 trout data
II! age 7 trout data
o EPAlFWS data
- - - - - age 7 prediction
.......- ..
..- ........-
-........
. -
... -."'.......
"""''''''
0.2
0.0
N M "II;f' I,() CD ..... CD
..... ..... ...... ...... ,.... ...... .....
2? 2? 2? 2? ~ 2? 2?
0> 0 -
... ex> co
2? 2? 2?
N M v an
CD co co CX)
2? 2? 2? 2?
to ,.... ex) 0)
co co co co
2? 2? 2? 2?
year
Figure 1.68. Simulation of dieldrin in Lake
Michigan trout (DeVault et al., 1986; Michigan
Department of Natural Resources, 1990).
30
III EPAlFWS data
o MDNR
- - - - - age 7 prediction
--..--..- - .
'
.
"
'
.
. -
"""....
".......
n n I'n'-'-n-
25
~ 20
2-
c:
.9
Iii
~ 15
Q)
to
c:
a
to
~ 10
11.
5
o
o ..... N M v
..... ..... ...... ..... .....
~ ~ ~ 2? ~
~ ~ ~ ~ ~ g ~ ~ ~ ~ ~ ~ ~ ~ m ~
~ ~ 2? ~ ~ ~ ~ ~ 2? ~ 2? 2? 2? 2? 2? 2?
year
Figure 1.69. Load cutoff simulation of PCBs in
Lake Michigan trout.
59
-------
~
'"
2-
c
o
i
~ 4 OOe-6
"
c
8
o
o
u
....
O.oOe+O
8 oOe-6
Ii] reDO data
... .. .. .... age 7 prediction
6 00e-6
2 OOe-6
f'-CDChO....
~ ~ ~ ! !
N M ~ \t) CD
CD co co CD co
~ ~ ~ ~ ~
..
-------
1.7.3.1 PCBs Control Scenarios
Polychlorinated biphenyls have been identified as a
priority toxic in Lake Michigan because of persistent
elevated concentrations in Lake Michigan biota,
especially lake trout. Reduction of present
atmospheric and tributary loadings may be
contemplated by the LaMP as control actions to
lower PCBs concentrations. Aside from issues of
feasibility or expense, the effectiveness of PCBs load
reduction upon reducing concentrations in lake trout
should be considered before taking such action. To
this end, MICHTOX was used to simulate the
expected change in future southern Lake Michigan
lake trout PCBs concentrations in response to load
reduction scenarios initiated in 1990. A No-Action
scenario was based upon extending the duration of
the verification simulation and holding PCBs loadings
constant after 1990. The second scenario was the
elimination of tributary loading at 1990. The third
scenario for controlling PCBs was to eliminate all
loading (total), atmospheric as well as tributary.
Resulting trout concentration predictions for the three
scenarios are plotted in Figure 1.72. Trout PCBs
concentrC}tions are predicted to decline 55% in 10
years in the No-Action scenario. The additional
reduction in predicted trout concentrations due to
eliminating tributary PCBs loading is barely
detectable, suggesting that controlling tributary
loading alone will be ineffective for PCBs in southern
Lake Michigan. The third, "Zero-Load" scenario is
predicted to have a more substantial effect upon trout
PCBs concentrations, resulting in a 74% reduction in
10 years. Viewed another way, these simulations
suggest that a 2 ~g/g (ppm) PCBs lake trout
standard would be achieved five years sooner for
the Zero-Load scenario than for No-Action.
However, even for the Zero-Load scenario, 0.4 ~g/g
of PCBs would remain in trout after 20 years.
According to these predictions, virtual elimination of
external sources of PCBs will have, at best, only
long-term effectiveness in reducing concentrations
in Lake Michigan.
It should be noted that these predictions were
extrapolated beyond the end of the verification
simulation, which appeared to overpredict PCBs trout
concentrations after 1980 by a factor of two. While
this discrepancy was acceptable from the standpoint
of preliminary verification, it leads to some ambiguity
in predicting future PCBs concentrations. For
10
- no action (constant loads after 1990)
.......... tributary loads eliminated at 1990
8
. - - - all PCB loading eliminated at 1990
Qi
CD
2-
c
.9
~
"E
Q)
"
c
o
"
en
c..>
c..
1990
6
4
~
,
,
, .
" "'.
',~<:::::.:....................................
....
....
-..------
:;
~
2
o
1985
1990
1995
2000
2005
2010
year
Figure 1.72. Predicted effectiveness of PCBs
load reductions.
example, would the prediction be more accurate if
the trout concentrations were lowered to match the
data? This could be achieved by changing the
loading time-series, the model calibration, or some
combination of the two. Given the available
information, this would be a somewhat arbitrary
decision that would lead to different future
predictions. It would be preferable to base predictions
of future PCBs concentrations on a simulation which
better matches present conditions. These conditions
include sediment, trout, and water concentrations
and atmospheric and non-atmospheric loads. The
collection of data for the purpose of defining present
conditions for PCBs or other priority toxics in Lake
Michigan should be prioritized because this
information will be essential for making reliable
predictions of future toxic chemical concentrations.
1.7.3.2 Severe Storm Event
Analysis of chemical distribution in sediment cores
indicates that normal sedimentation rates in the
Great Lakes are periodically disrupted; these
disruptions have been related chronologically to
major storms (Robbins et al., 1978). Lick (1993) has
suggested that such events are of considerable
significance in determining the distribution, transport,
and fate of particle-associated contaminants such as
PCBs. To pursue this suggestion, the impact of such
61
-------
a severe storm on PCBs concentrations in Lake
Michigan was simulated. A storm occurring in winter
of 1990 was simulated; during the two-day event, the
entire surficial sediment layer was eroded from the
southern lake basin (Segment 11) and resuspended
through the water column. Afterwards, particle fluxes
were returned to their normal time-series values and
the solids balance was allowed to recover. Predicted
suspended particle and PCBs water concentrations
in southern Lake Michigan are plotted in Figure 1.73.
20
/',
.r./ \
,/ \
.-" .
............. .
.s
c
,2
1! 120
C
..
<.)
c
8
III
~ 80
L...."""'''''''''''''''''''''''''''''''''''''''''''''''''''''''''''-::::'''''''"''''
c
.1
'a
:II 40
o
1989
1994
1992
1993
1991
1995
1990
year
Figure 1.74. Effect of storm event on PCBs In
southern Lake Michigan sediment.
Because most of the resuspended PCrjS are
partitioned into non-dissolved fractions, ttley are
largely unavailable for accumulation by biota.
Results of this simulation suggest that the effect of
severe storms upon PCBs concentrations in the main
lake are short-lived and do not lead to significant
additional accumulation in biota.
1.7.4 Uncertainty in Dynamic Simulations
Dynamic model simulations are uncertain due to
factors in addition to those considered in the steady.
state model. These factors include the additional
significant model parameter, the surficial sediment
thickness, and uncertainty of initial conditions and the
time-series of loadings. The surficial sediment
thickness defines the residence time of particles and
chemicals in the mixed layer, which controls the long-
term rate of concentration change in the model.
Sediment thickness was parameterized in MICHTOX
according to values suggested by vertical
concentration profiles of lead-21 0 and cesium-137 in
sediment cores. In addition, this parameterization
was verified for water column plutonium
concentrations. However, the modeling assumption
that the mixe:j-Iayer residence' ...,e will be the same
for all toxic~nemicals has no; Jeen validated. In
particular, the thickness of the mixed layer may relate
to the loading history of a particular chemical. If, for
example, the loading of a chemical were to increase
62
-------
faster than the rate at which it could be incorporated
and mixed throughout the surficial sediment layer,
then the mixed layer thickness would effectively
decrease. Intensity and depth of sediment mixing
also depend upon the abundance and type of
benthos; Diporeia, for instance, mix only the upper 1
to 2 cm of sediment (Eisenreich et al., 1989). A
thinner surficial sediment thickness results in
predictions of a more rapid change in concentrations.
This sensitivity is demonstrated for the simulation of
PCBs concentrations in trout in Figure 1.75. The
"thin sediment" simulation, using a surficial sediment
thickness of 1.1 cm (one-third the base
parameterization), does predict PCBs concentration
change in better agreement with the data. This
simulation would also change the prediction of future
PCBs concentrations, indicating less significant
decline in concentrations over the next 10 years.
The parameterization of surficial sediment thickness
is, therefore, one potentially significant cause of
uncertainty in dynamic model predictions.
30
o EPAlFWS
. MDNR
- verification simulation
I : ----.tth
g 15 """""
o ~
C.)
IX!
~ 10
"5
g
........ thin sediment layer simulation
5
! Ii>..
!
i
o
1970
.
1975
1980
1985
1990
year
Figure 1.75. Sensitivity of PCBs concentrations
in trout to thin (1.1 cm) surficial sediment layer
thickness.
Another source of dynamic prediction uncertainty is
the determination of initial concentration conditions
for model simulations. If model simulations begin
with a "clean" system, as was the case for MICHTOX
verification, then this is not an issue. However, such
simulations may result in excessively long model
runs, and may not be possible for toxic chemicals
whose past loadings are unknown. If, for instance,
one wished to simulate trout PCBs concentrations
after their maximum in 1974, the initial PCBs
concentrations would have to be input. For age
seven trout, this could range from 15 to 27 jJg/g,
according to the data. Depending upon the selection,
quite different model predictions could result.
However, even more critical is the selection of initial
chemical concentrations in the surficial sediment,
because under reduced loading conditions sediment
concentrations "drive" the model simulation.
Computer resources were insufficient to perform full
uncertainty analysis on the dynamic MICHTOX
model. However, a limited test was run for PCBs to
evaluate the significance of the factors discussed
above, in relationship to those already considered for
the steady-state model, upon uncertainty in dynamic
simulations. Ten Monte Carlo simulations of the
PCBs Zero-Load scenario were run; model
parameters, including surficial sediment thickness,
were varied by the Latin Hypercube method to
simulate uncertainty. The results, for water and trout
concentration predictions, are plotted in Figures 1.76
and 1.77. Although 10 runs are not sufficient to
resolve the model output distributions, they do
provide a qualitative indication of uncertainty in
dynamic predictions. The lag time for a 90%
1.25
c:
.2
~
~ 1.00
o
c:
o
o
~ 0.75
:i=
ca
U
a..
00.50
u
Q.
CD
.~ 0.25
10
~
0.00
o
25
35
40
30
15
20
5
10
year
Figure 1.76. Predicted water PCBs
concentrations for ten realizations 01 dynamic
model.
63
-------
c:
.Q
1ii
~ 1.00
c:
Q)
CJ
c:
o
CJ
In 0.75
u
a.
'S
e
:::.. 0.50
o
~
~
.~ 0.25
1ii
~
1.25
0.00
o
5
10
15
20
25
30
35
40
year
Figure 1.77. Predicted trout PCBs concentrations
for ten realizations of dynamic model.
reduction in PCBs water concentration varied from
six to 20 years, with a mean of 12 years. In trout, the
90% lag time varied from 10 to 24 years, with a mean
of 16 years: The surficial sediment thickness was
found to contribute more than 85% of the uncertainty
in water concentration lag time predictions. For trout
lag time predictions, the most significant source of
uncertainty was the plankton BCF (30% of lag time
variability). Twenty-four percent of the trout lag time
variability could be attributed to the lag time in water
concentrations. As was the case for the steady-state
model, analysis reveals that the dynamic model
predictions are highly uncertain. Consequently, as
was the case for steady-state, quantitative results
from the dynamic model simulations should not be
considered reliable. Reducing this uncertainty would
require additional calibration and/or verification data
and measurement of critically uncertain parameters.
It should also be considered that the loading time-
series are themselves somewhat speculative and
uncertain. Historical loadings must usually be
inferred from sedimentary records (plutonium being
the notable exception) from which the loading time-
series may not be fully deconvoluted (Christensen
and Goetz, 1987). As a result, it may be difficult
during calibration and verification to distinguish
model error from errors in the loading time-series.
This is demonstrated by comparing the verification
and load cutoff simulations (Figures 1.31 and 1.69).
If one compares the fit of these two simulations to the
trout PCBs data, the cutoff of PCBs loads at 1974
would appear to be the better loading time-series.
Because relatively little data are available to
independently confirm the PCBs loading time-series
developed for Lake Michigan, the loading history
itself must be considered as a source of error to the
simulation. Accurately determining loads is critical
for detecting and correcting model errors and,
ultimately, to reducing predictive uncertainty.
One additional aspect of MICHTOX which may lead
to erroneous long-term predictions is the lack of
structural variability in the bioaccumulation model. In
particular, the MICHTOX trophic structure is static;
the model neither predicts nor does it respond to
factors such as changing forage composition, trophic
status in response to nutrients, exotic species
invasion, or fisheries management. Yet such factors
do affect the trophic structure in the Great Lakes and
may be expected to affect bioaccumulation at the top
of the food chain. Long-term bioaccumulation
simulations well-parameterized for present conditions
are likely to diverge from future reality as the lake
trophic structure varies. In some cases, the
prediction divergence may be small, as was the
predicted change in trout PCBs bioaccumulation due
to benthic coupling. However, this may not generally
be the case. Uncertainty in future bioaccumulation
predictions due to the dynamics of trophic structure
in the Great Lakes is, as far as existing
bioaccumulation models are concerned, in the realm
of unforeseeable possibilities.
1.8 References
Alberts, J.J. and M.A. Wahlgren. 1981.
Concentrations of Pu, Cs, and Sr in the Waters
of the Laurentian Great Lakes: Comparison of
1973 and 1976 Values. Environ. Sci. Technol.,
15(1 ):94-98.
Ambrose, R. B., T. A. Wool, J. P- Connolly, and
R.W. Schanz. 1988. WASP4, A Hydrodynamic
and Water Quality Model- Model Theory, User's
Manual and Programmer's Guide. U.S.
Environmental Protection Agency, Office of
Research and Development, Environmental
Research Laboratory, Athens, Georgia.
EPA/600/3-87/039, 297 pp.
64
-------
Assel, R.A., F.H. Quinn, G.A Leshkevich, and S.J.
Bolsenga. 1983. Great Lakes Ice Atlas.
National Oceanic and Atmospheric
Administration, Great Lakes Environmental
Research Laboratory, Ann Arbor, Michigan. 115
pp.
Auer, M.T. 1989. 1982 Green Bay Data. Michigan
Technological University, Houghton, Michigan. 1
pp. and Disk.
Ayers, J.C., D.C. Chandler, G.H. Lauff, C.F. Powers,
and EB. Henson. 1958. Currents and Water
Masses of Lake Michigan. The University of
Michigan, Great Lakes Research Division, Ann
Arbor, Michigan. Publication Number 5,220 pp.
Baker, J.E, P.D. Capel, and S.J. Eisenreich. 1986.
Influence of Colloids in Sediment-Water Partition
Coefficients of Polychlorobiphenyl Congeners in
Natural Waters. Environ. Sci. Techno!.,
20(12):1136-1143.
Baker, J.E and S.J. Eisenreich. 1990.
Concentrations and Fluxes of Polycyclic Aromatic
Hydrocarbons and Polychlorinated Biphenyls
across the Air-Water Interface of Lake Superior.
Environ. Sci. Techno!., 24(3):342-352.
Burkhard, L.P. and D.W. Kueh!. 1986. N-
Octanol/W ater Partition Coefficients by Reverse
Phase Liquid Chromatography/Mass
Spectrometry for Eight Tetrachlorinated Planar
Molecules. Chemosphere, 15(2): 163-167.
Cahill, R.A 1981. Geochemistry of Recent Lake
Michigan Sediments. Illinois Geological Survey,
Champaign, Illinois. Circular 517,94 pp.
Capel, P.D. and S.J. Eisenreich. 1990. Relationship
Between Chlorinated Hydrocarbons and Organic
Carbon in Sediment and Porewater. J. Great
Lakes Res., 16(2):245-257.
Carey, AE., N.S. Shifrin, and A.C. Roche. 1990.
Lake Ontario TCDD Bioaccumulation Study.
Final Report. U.S. Environmental Protection
Agency, Region II, New York, New York. 671 pp.
Christensen, E.R. and R.H. Goetz. 1987. Historical
Fluxes of Particle-Bound Pollutants from
Deconvolved Sedimentary Records. Environ.
Sci. Techno!., 21 (11 ):1 088-1 096.
Clark, T.P., R.J. Norstrom, G.A. Fox, and H.T. Won.
1987. Dynamics of Organochlorine Compounds
in Herring Gulls (Larus argentatus): II. A Two-
Compartment Model and Data for Ten
Compounds. Environ. Toxico!. Chem., 6(7):547-
559.
Connolly, J.P. and R.V. Thomann. 1985. WASTOX,
A Framework for Modeling the Fate of Toxic
Chemicals in Aquatic Environments. Project
Report. U.S. Environmental Protection Agency,
Office of Research and Development,
Environmental Research Laboratory-Duluth,
Large Lakes Research Station, Grosse lie,
Michigan. 52 pp.
Connolly, J.P. 1991. Documentation for Food Chain
Model, Version 4.0. Manhattan College,
Department of Environmental Engineering and
Sciences, Riverdale, New York.
Connolly, J.P. 1992. Bioaccumulation of
Hydrophobic Organic Chemicals by Aquatic
Organisms. Presented at the Workshop on
Bioaccumulation of Hydrophobic Organic
Chemicals by Aquatic Organisms, National
Institute of Environmental Health Sciences,
Leesburg, Virginia. June 28 - July 1, 1992.
de Wolf, W., J.H.M. de Bruijn, W. Seinen, and
J.L.M. Hermens. 1992. Influence of
Biotransformation on the Relationships Between
Bioconcentration Factors and Octanol-Water
Partition Coefficients. Environ. Sci. Techno!.,
26(6):1197-1201.
DePinto, J.V. 1990. Lake Ontario Nutrient
Cycle/Foodweb Modeling. Workshop Report.
State University of New York at Buffalo, Great
Lakes Program, Buffalo, New York.
Dermott, R. and K. Corning. 1988. Seasonal
Ingestion Rates of Pantoporeia hoyi (Amphipoda)
in Lake Ontario. Canadian J. Fish. Aquat. Sci.,
45: 1886-1895.
65
-------
DeVault, D.S., W.S. Willford, R.J. Hesselberg, D.A.
Nortrupt, E.G.S. Rundberg, A.K. Alwan, and C.
Bautista. 1986. Contaminant Trends in Lake
Trout (Salvelinus namaycush) from the Upper
Great Lakes. Arch. Environ. Contam. Toxicol.,
15:349-356.
DeVault, D., W. Dunn, P Bergqvist, K. Wiberg, and
C. Rappe. 1989. Polychlorinated Dibenzofurans
and Polychlorinated Dibenzo-p-dioxins in Great
Lakes Fish: A Baseline and Interlake
Comparison. Environ. Toxicol. Chem.,
8(11):1013-1022.
Di Toro, D.M. 1985. A Particle Interaction Model of
Reversible Organic Chemical Sorption.
Chemosphere, 14(10): 1503-1538.
Di Toro, D.M. 1987. Modeling the Fate of Toxic
Chemicals in Surface Waters. Presented at the
8th Annual Meeting of the Society for
Environmental Toxicology and Chemistry,
Pensacola, Florida. November 9-12, 1987.
Eadie, B.J., J.A. Robbins, P.F. Landrum, C.P. Rice,
M.J. Simmons, M.J. McCormick, S.J. Eisenreich,
G.L. Bell, RL. Pickett, K. Johansen, R.
Rossmann, N. Hawley, and T. Voice. 1983. The
Cycling of Toxic Organics in the Great Lakes: A
3- Year Status Report. National Oceanic and
Atmospheric Administration, Great Lakes
Environmental Research Laboratory, Ann Arbor,
Michigan. NOAA Technical Memorandum ERL
GLERL-45, 163 pp.
Eadie, B.J., RL. Chambers, W.S. Gardner, and G.L.
Bell. 1984. Sediment Trap Studies in Lake
Michigan: Resuspension and Chemical Fluxes in
the Southern Basin. J. Great Lakes Res.,
10(3):307-321.
Eadie, B.J., N.R Moorehead, and P.F. Landrum.
1990. Three-Phase Partitioning of Hydrophobic
Organic Compounds in Great Lakes Waters.
Chemosphere, 20( 1-2): 161-178.
Eadie, B.J., G.L. Bell, and N. Hawley. 1991.
Sediment Trap Studies in the Green Bay Mass
Balance Program: Mass and Organic Carbon
Fluxes, Resuspension, and Particle Settling
Velocities. National Oceanic and Atmospheric
Administration, Great Lakes Environmental
Research Laboratory, Ann Arbor, Michigan.
NOAA Technical Memorandum ERL GLERL-75,
29 pp.
Eck, G.W. and E.H. Brown. 1990. Status of Forage
Fish Stocks in Lake Michigan, 1990.
Unpublished Report. U.S. Department of the
Interior, U.S. Fish and Wildlife Services, National
Fisheries Research Center-Great Lakes, Ann
Arbor, Michigan.
Edgington, D.N. and J.A. Robbins. 1976. Records
of Lead Deposition in Lake Michigan Sediments
Since 1800. Environ. Sci. Technol., 10(3):266-
274.
Edgington, D.N. 1991. Sediment Core Data for the
Green Bay Mass Balance Study. University 01
Wisconsin, Center for Great Lakes Studies,
Milwaukee, Wisconsin.
Eisenreich, S.J., B.B. Looney, and J.D. Thornton.
1981. Airborne Organic Contaminants in the
Great Lakes Ecosystem. Environ. Sci. Techno!.,
15(1 ):30-38.
Eisenreich, S.J., P.D. Capel, J.A. Robbins, and R.
Bourbonniere. 1989. Accumulation and
Diagenesis of Chlorinated Hydrocarbons in
Lacustrine Sediments. Environ. Sci. Techno!.,
23(9):1116-1126.
Endicott, D.D., W.L. Richardson, T.F. Parkerton, and
D.M. Di Toro. 1990. A Steady-State Mass
Balance Bioaccumulation Model for Toxic
Chemicals in Lake Ontario. U.S. Environmental
Protection Agency, Office of Research and
Development, Environmental Research
Laboratory-Duluth, Large Lakes Research
Station, Grosse lie, Michigan. 121 pp.
66
-------
Endicott, D.O., W.L. Richardson, and D.M. Di Toro.
1991. Modeling the Partitioning and
Bioaccumulation of TCDD and Other
Hydrophobic Organic Chemicals in Lake Ontario.
Presented at the 11th International Symposium
on Chlorinated Dioxins and Related Compounds,
Research Triangle Park, North Carolina.
September 23-27, 1991.
Environment Canada. 1991. Toxic Chemicals in the
Great Lakes and Associated Effects.
Environment Canada, Communications
Directorate, Ontario Region, Toronto, Ontario,
Canada. 488 pp.
Evans, M.S. and P.F. Landrum. 1989.
Toxicokinetics of DOE, Benzo(a)pyrene, and
2,4,5,2',4,4',5'-Hexachlorobiphenyl in
Pontoporeia hoyi and Mysis relicata. J. Great
Lakes Res., 15(4):589-600.
Evans, M.S., G.E Noguchi, and C.P. Rice. 1991.
The Biomagnification of Polychlorinated
Biphenyls, Toxaphene, and DOT Compounds in
a Lake Michigan Offshore Food Web. Arch.
Environ. Contam. Toxicol., 20(1):87-93.
Flint, R. W. 1986. Hypothesized Carbon Flow
through the Deepwater Lake Ontario Food Web.
J. Great Lakes Res., 12(4):344-354.
Frank, R., F.L. Thomas, M. Holdrinet, A.L.W. Kemp,
and H.E. Braun. 1979. Organochlorine
Insecticides and PCB in Surficial Sediments
(1968) and Sediment Cores (1976) from Lake
Ontario. J. Great lakes Res., 5(1):18-27.
Franz, T.P. and S.J. Eisenreich. 1991. Wet
Deposition of Polychlorinated Biphenyls to Green
Bay, Lake Michigan. Draft Report. U.S.
Environmental Protection Agency, Great Lakes
National Program Office, Chicago, Illinois. 32 pp.
Gardner, W.S. and R.V. O'Neill. 1983. Parameter
Uncertainty and Model Predictions: A Review of
Monte Carlo Results. In: M.B. Beck and G. van
Straten (Eds.), Uncertainty and Forecasting of
Water Quality, pp. 245-257. Springer-Verlag,
New York, New York.
Gottleib, ES., J.H. Saylor, and G.S. Miller. 1990.
Currents and Water Temperatures Observed in
Green Bay, Lake Michigan. National Oceanic
and Atmospheric Administration, Great Lakes
Environmental Research laboratory, Ann Arbor,
Michigan. NOAA Technical Memorandum ERL
GlERL-73, 90 pp.
Hallam, T.G., A.A. lassiter, J. Li, and W. McKinney.
1990. Toxicant-Induced Mortality in Models of
Daphnia Populations. Environ. Toxico!. Chern.,
9(5):597-621.
Hermanson, M.H. and ER. Christensen. 1991.
Recent Sedimentation in Lake Michigan. J.
Great Lakes Res., 17(1 ):33-50.
Hermanson, M.H., ER. Christensen, D.J. Buser, and
L. Chen. 1991. Polychlorinated Biphenyls in
Dated Sediment Cores from Green Bay and Lake
Michigan. J. Great lakes Res., 17(1):94-108.
Hesselberg, A.J., J.P. Hickey, D.A. Nortrupt, and
W.A. Willford. 1990. Contaminant Residues in
the Bloater (Coregonus hOYI) of Lake Michigan,
1969-1986. J. Great lakes Res., 16(1):121-129.
Hoff, A.M., D.C.G. Muir, and N.P. Grift. 1992.
Annual Cycle of Polychlorinated Biphenyls and
Organohalogen Pesticides in Air in Southern
Ontario. I. Air Concentration Data. Environ. Sci.
Techno!.,26(2):266-275.
Hornberger, G.M. and R.C. Spear. 1981. An
Approach to the Preliminary Analysis of
Environmental Systems. J. Environ. Mgt., 12:7-
18.
Hudson, A.J.M., S.A. Gherini, and R.K. Munson.
1991. The MTL Mercury Model: A Description of
the Model, Discussion of Scientific Issues and
Presentation of Preliminary Results. Mercury in
Temperature lakes Project. Annual Report to
Electric Power Research Institute and Wisconsin
Department of Natural Resources. EPRI Report
2020-10.
67
-------
Jobes, F.W. 1949. The Age, Growth, and
Bathymetric Distribution of the Bloater,
Leucicthys hoyi (Gill) in Lake Michigan. Papers
of the Michigan Academy of Sciences, Arts and
Letters, 33:135-172.
Landrum, P.F. and J.A. Robbins. 1990.
Bioavailability of Sediment Associated
Contaminants to Benthic Invertebrates. In: R.
Baudo, J.P. Giesy, and H. Muntau (Eds.),
Sediments: Chemistry and Toxicity of In-Place
Pollutants, pp. 237-263. Lewis Publishers,
Incorporated, Ann Arbor, Michigan.
Lick, W. 1993. The Importance of Large Events. In:
Reducing Uncertainty in Mass Balance Models of
Toxics in the Great Lakes - Lake Ontario Case
Study, pp. 286-307. Donald W. Rennie Memorial
Monograph Series, Great Lakes Monograph
Number 4, State University of New York, Buffalo,
New York.
Liss, P .S. 1973. Process of Gas Exchange Across
an Air-Water Interface. Deep Sea Res., 20:221-
228.
Lyman, W.J., W.F. Reehl, and D.H. Rosenblatt.
1982. Handbook of Chemical Property
Estimation Methods - Environmental Behavior of
Organic Compounds. McGraw-Hili Book
Company, New York, New York. 938 pp.
Mabey, W.R. and J.H. Smith. 1982. Aquatic Fate
Process Data for Organic Priority Pollutants.
U.S. Environmental Protection Agency, Office of
Water Regulations and Standards, Washington,
D.C. EPA/440/4-81/014, 407 pp.
Mackay, D. and S. Paterson. 1986. Model
Describing the Rates of Transfer Processes of
Organic Chemicals Between Atmosphere and
Water. Environ. ScL Technol., 20(8):810-816.
Mackay, D. 1989. Modeling the Long-Term
Behavior of an Organic Contaminant in a Large
Lake: Application to PCBs in Lake Ontario. J.
Great Lakes Res., 15(2):283-297.
Marti, EA. and D.E Armstrong. 1990.
Polychlorinated Biphenyls in Lake Michigan
Tributaries. J. Great Lakes Res., 16(3):396-405.
McCarthy, J.F. and B.D. Jiminez. 1985. Reduction
in Bioavailability to Bluegills of Polycyclic
Aromatic Hydrocarbons Bound to Dissolved
Humic Material. Environ. Toxico!. Chern.,
4(4):511-521.
McKay, M.D., R.J. Beckman, and W.J. Conover.
1979. A Comparison of Three Methods for
Selecting Values of Input Variables in the
Analysis of Output From a Computer Code.
Technometrics, 21 (2):239-245.
Michigan Department of Natural Resources. 1990.
Fish Contaminant Monitoring Program, 1990
Annual Report. Michigan Department of Natural
Resources, Surface Water Quality Division,
Lansing, Michigan. Report Number 90/077.
Miller, M.A. and M.E Holey. 1991. Diets of Lake
Trout Inhabiting Nearshore and Offshore Lake
Michigan Environments. J. Great Lakes Res.,
18(1 ):51-60.
Mortimer, C.H. 1971. Large-Scale Oscillatory
Motions and Seasonal Temperature Changes in
Lake Michigan and Lake Ontario. University of
Wisconsin, Center for Great Lakes Studies,
Milwaukee, Wisconsin. Special Report Number
12, 111 pp.
Mudroch, A. and D. Williams. 1989. Suspended
Sediments and the Distribution of Bottom
Sediments in the Niagara River. J. Great Lakes
Res., 15(3):427-436.
Murphy, T.J. and C.P. Rzesezutko. 1977.
Precipitation Inputs of PCBs to Lake Michigan. J.
Great Lakes Res., 3(3/4):305-312.
Neidermeyer, W.J. and J.J. Hickey. 1976.
Chronology of Organochlorine Compounds in
Lake Michigan Fish, 1929-1966. Pest. Monit. J.,
10(3):92-94.
Niimi, A.J. and B.G. Oliver. 1983. Biological Half-
Lives of Polychlorinated Biphenyl (PCB)
Congeners in Whole Fish and Muscle of Rainbow
Trout (Sa/mo gairdnerl). Canadian J. Fish.
Aquat. ScL, 40(9):1388-1394.
68
-------
O'Connor, D.J. 1983. Wind Effects on Gas-Liquid
Transfer Coefficients. J. Environ. Engin.,
109(3):731-752.
Oliver, B.G. 1987. Partitioning Relationships for
Chlorinated Organics Between Water and
Particulates in the St. Clair, Detroit and Niagara
Rivers. In: K.L.E. Kaiser (Ed.), QSAR in
Environmental Toxicology II, D. Reidel Publishing
Company, Dordrecht, Holland.
Oliver, B.G. and A.J. Niimi. 1988. Tropodynamic
Analysis of Polychlorinated Biphenyl Congeners
and Other Chlorinated Hydrocarbons in the Lake
Ontario Ecosystem. Environ. Sci. Technol.,
22(4):388-397.
Oliver, B.G., M.N. Charlton, and R. W. Durham.
1989. Distribution, Redistribution and
Geochronology of Polychlorinated Biphenyl
Congeners and Other Chlorinated Hydrocarbons
in Lake Ontario Sediments. Environ. Sci.
Techno!., 23(2):200-208.
Oppenhuizen, A. and D.T.H.M. Sijm. 1990.
Bioaccumulation and Biotransformation of
Polychlorinated Dibenzo-p-dioxins and
Dibenzofurans in Fish. Environ. Toxicol. Chern.,
9(2):175-186.
Prospero, J.M. 1978. The Tropospheric Transport
of Pollutants and Other Substances to the
Oceans. National Academy of Science,
Washington, D.C. 243 pp.
Quinn, F.H. 1977. Annual and Seasonal Flow
Variations through the Straits of Mackinac.
Water Resources Res., 13(1):137-144.
Rice, C.P., P.J. Samson, and G.E. Noguchi. 1986.
Atmospheric Transport of Toxaphene to Lake
Michigan. Environ. Sci. Techno!., 20(11 ):11 09-
1116.
Robbins, J.A and D.N. Edgington. 1975. Stable
lead Geochronology of Fine-Grained Sediments
in Southern lake Michigan. Argonne National
laboratory, Radiological and Environmental
Research Division, Argonne, Illinois. Annual
Report AN-75-3, Part III, pp. 32-39.
Robbins, J.A, D.N. Edgington, and AL. Kemp.
1978. Comparative Pb, Cs, and Pollen
Geochronologies of Sediments From lakes
Ontario and Erie. Quatern. Res., 10:256-278.
Robbins, J.A 1985. The Coupled lakes Model for
Estimating the Long-Term Response of the Great
Lakes to Time-Dependent loadings of Particle-
Associated Contaminant. National Oceanic and
Atmospheric Administration, Great Lakes
Environmental Research laboratory, Ann Arbor,
Michigan. NOAA Technical Memorandum ERL
GLERL-57, 41 pp.
Robbins, J.A and B.J. Eadie. 1991. Seasonal
Cycling of Trace Elements, Cs, Be and Pu in
lake Michigan. J. Geophys. Res., 96(C9):17081-
17104.
Rodgers, P.W. and D.K. Salisbury. 1981. Water
Quality Modeling of Lake Michigan and
Consideration of the Anomolous Ice Cover of
1976-1977. J. Great Lakes Res., 7(4):467-480.
Rodgers, P.W. and W.R. Swain. 1983. Analysis of
Polychlorinated Biphenyl (PCB) Loading Trends
in lake Michigan. J. Great lakes Res., 9(4):548-
558.
Rordorf, B.F. 1989. Prediction of Vapor Pressures,
Boiling Points and Enthalpies of Fusion for
Twenty-Nine Halogenated Dibenzo-p-dioxins and
Fifty-Five Dibenzofurans by a Vapor Pressure
Correlation Method. Chemosphere, 18(1/6):783-
788.
Rossmann, R. and J. Barres. 1988. Trace Element
Concentrations in Near-Surface Waters of the
Great Lakes and Methods of Collection, Storage,
and Analysis. J. Great Lakes Res., 14(2):188-
204.
Rudstam, L.G., F.P. Binkowski, and M.A. Miller.
1992. A Bioenergetics Model for Analysis of
Food Consumption Patterns by Bloater in Lake
Michigan. University of Wisconsin, Center for
limnology, Madison, Wisconsin.
69
-------
Schwab, D.J. and D.L. Sellers. 1980. Computerized
Bathymetry and Shorelines of the Great Lakes.
National Oceanic and Atmospheric
Administration, Great Lakes Environmental
Research Laboratory, Ann Arbor, Michigan.
NOAA Technical Data ERL GLERL-16, 13 pp.
Schwab, D.J. 1983. Numerical Simulation of Low-
Frequency Fluctuations in Lake Michigan. J.
Phys. Oceanogr., 13:2213-2224.
Skoglund, R.S. and D.L. Swackhamer. 1991.
Spatial and Seasonal Variations in the
Bioaccumulation of PCBs by Phytoplankton in
Green Bay, Lake Michigan. Presented at the
34th Conference on Great Lakes Research,
International Association for Great Lakes
Research, University of New York, Buffalo, New
York. June 2-6, 1991.
Smith, M.S., P.W. O'Keefe, K.M. Aldous, H. Valente,
S.P. Connor, and R.J. Donnelly. 1990.
Chlorinated Dibenzofurans and Dioxins in
Atmospheric Samples From Cities in New York.
Environ. Sci. Techno!., 24(10):1502-1506.
Sonzogni, W.C. and M.S. Simmons. 1981. Notes on
Great Lakes Trace Metal and Toxic Organic
Contaminants. Great Lakes Basin Commission,
Great Lakes Environmental Planning Study, Ann
Arbor, Michigan. Contribution Number 54.
Stevens, R.J.J. and M.A. Neilson. 1989. Inter- and
Intralake Distributions of Trace Organic
Contaminants in Surface Waters of the Great
Lakes. J. Great Lakes Res., 15(3):377-393.
Strachan, W.M. and S.J. Eisenreich. 1988. Mass
Balancing of Toxic Chemicals in the Great Lakes:
The Role of Atmospheric Deposition. In:
Workshop on the Estimation of Atmospheric
Loading of Toxic Chemicals to the Great Lakes,
International Joint Commission, Windsor,
Ontario, Canada.
Sunito, L.R., W.Y. Shiu, D. Mackay, J.N. Seiber, and
D. Gotfelty. 1988. Critical Review of Henry's
Law Constants for Pesticides. Rev. Environ.
Contam. Toxico!., 103:1-59.
Swackhamer, D.L. and D.E. Armstrong. 1986.
Estimation of the Atmospheric Contributions and
Losses of Polychlorinated Biphenyls for Lake
Michigan on the Basis of Sediment Records of
Remote Lakes. Environ. Sci. Techno!.,
20(9):879-883.
Swackhamer, D.L. and D.E. Armstrong. 1987.
Distribution and Characterization of PCBs in Lake
Michigan Water. J. Great Lakes Res., 13(1):24.
36.
Swackhamer, D.L. and D.E. Armstrong. 1988.
Horizontal and Vertical Distribution of PCBs in
Southern Lake Michigan Sediments and the
Effect of Waukegan Harbor as a Point Source.
J. Great Lakes Res., 14(3):277-290.
Sweet, C.W. and T.J. Murphy. 1991. The Role of
the Atmosphere in the Mass Balance of PCBs in
Green Bay. Presented at the 34th Conference
on Great Lakes Research, International
Association for Great Lakes Research, University
of New York, Buffalo, New York. June 2-6, 1991.
Syracuse Research Corporation. 1988. ChemFate
Database. Syracuse Research Corporation,
Syracuse, New York.
Thomann, R.V., R.P. Winfield, and J.J. Segna.
1979. Verification Analysis of Lake Ontario and
Rochester Embayment Three-Dimensional
Eutrophication Models. U.S. Environmental
Protection Agency, Office of Research and
Development, Environmental Research
Laboratory-Duluth, Large Lakes Research
Station, Grosse lie, Michigan. EPA-600/3-79-
094, 136 pp.
Thomann, R.V. and M.T. Kontaxis. 1981.
Mathematical Modeling Estimate of
Environmental Exposure Due to PCB
Contaminated Harbor Sediments of Waukegan
Harbor and North Ditch. U.S. Environmental
Protection Agency, Industrial Environmental
Research Laboratory, Cincinnati, Ohio. 109 pp.
Thomann, R.V. and D.M. Di Toro. 1983.
Physicochemical Model of Toxic Substances in
the Great Lakes. J. Great Lakes Res., 9(4):474-
496.
70
-------
Thomann, R.V. and J.P. Connolly. 1984. Model of
PCB in the Lake Michigan Lake Trout Food
Chain. Environ. Sci Technol., 18(2):65-71.
Thomann, R. V. and J .A. Mueller. 1987. Principles of
Surface Water Quality Modeling and Control.
Harper and Row Publishers, Incorporated, New
York, New York. 644 pp.
Thomann, R.V. 1989. Bioaccumulation Model of
Organic Chemical Distribution in Aquatic Food
Chains. Environ. Sci Techno!., 23(6):699-707.
Thomas, R.L., A.L. Kemp, and C.F.M. Lewis. 1972.
Distribution, Composition and Characteristics of
the Surficial Sediments of Lake Ontario. J.
Sediment. Petrology, 42(1 ):66-84.
U.S. Environmental Protection Agency. 1989a.
Green Bay/Fox River Mass Balance Study - A
Strategy for Tracking Toxics in the Bay of Green
Bay, Lake Michigan. U.S. Environmental
Protection Agency, Great Lakes National
Program Office, Chicago, Illinois. EPA-905/8-89-
001 , 52 pp.
U.S. Environmental Protection Agency. 1989b.
PCDD/PCDF Concentration in Great Lakes Fish.
Unpublished Report. U.S. Environmental
Protection Agency, Office of Research and
Development, Environmental Research
Laboratory-Duluth, Duluth, Minnesota.
Van Hoof, P.L. and A.W. Andren. 1989. Partitioning
and Transport of 210-Pb in Lake Michigan. J.
Great Lakes Res., 15(3):498-509.
Veith, G.D., D.L. DeFoe, and B.V. Bergstedt. 1979.
Measuring and Estimating the Bioconcentration
Factor of Chemicals in Fish. J. Fish. Res. Bd.
Canada, 36(9): 1040-1048.
Wahlgren, M.A., D.M. Nelson, and E.T. Kucera.
1977. Seasonal Cycling of Plutonium in the
Water Column of Lake Michigan: 1975-1977.
Argonne National Laboratory, Radiological and
Environmental Research Division, Argonne,
Illinois. Annual Report ANL-77-65, Part III.
Weininger, D., D.A. Armstrong, and D.L.
Swackhamer. 1983. Application of a Sediment
Dynamics Model for Estimation of Vertical Burial
Rates of PCBs in Southern Lake Michigan. In:
D. Mackay, S. Paterson, S.J. Eisenreich, and
M.S. Simmons (Eds.), Physical Behavior of PCBs
in the Great Lakes,. Ann Arbor Science
Publishers, Incorporated, Ann Arbor, Michigan.
Whitman, W.G. 1923. A Preliminary Experimental
Confirmation of the Two-Film Theory of Gas
Absorption. Chem. Metallo Eng., 29:146-148.
Winchester, J.W. 1969. Pollution Pathways in the
Great Lakes. Limnos, 2:20-24.
Zepp, G.R. and D.M. Cline. 1977. Rates of Direct
Photolysis in Aquatic Environment. Environ. ScL
Techno!., 11 (4):359-36.
71
-------
1.9 APPENDIX
STEADY-STATE MODEL OUTPUT FOR EACH TOXIC CHEMICAL
72
-------
......
(..)
m
CD
:J
N
o
-
D)
-
"'C
'<
"'I
CD
:J
CD
en
-
CD
D)
Q,
'<
I
en
-
D)
-
CD
s:
o
Q,
9?
~r...r.....
GLI n..
div.u,;Oft 93.'
Q..,2 1036
Q..' ""
Q",,& 169
1iL5 '6~
Q...6 3"
COLl 383
..d...,.(_SI_,"
f...122110
f...2' 2110
f-3H 118'
f...05 2.6
f...56 400
F-61 600
f...21 1.,0
uu&821;
A-1 2.61[-10
A...2 2.11£-10
A-3 USE-OS
"""0 6.58E-06
1\..5 3.80[-08
"'-.6 1.42[-09
,,-, 2.83E-09
..tllI.,,'.'dll-
...~_t 1.5
...,_2 1.5
...,,_3 1.5
u_& 1.5
,,~_5 1.5
",,_61.5
...~_1 1.5
d.p..IiIl..~_I."
,eLl 3.08[-06
veL2 2.69[.06
veL3 3,&2[-01
vd.....4 1.6"£-05
vcL5 :I 48[-01
vcL6 5.15[-01
v 16> 116
o.ow lUC-IM lUC.os .!SI£...fN
.160£.01 "'.$11 ;'$/1 2$'"
"" 0...11$1 (J.~ b./"'f
lH£..Q1 lUEIo(JS ~=.01
l...~ J.()6 ~.64
2..'6 o.ssr q.1Q QI.?6
T ahl Ale-ife = q....
-------
......_......111. I..... ....-..1....................... -.....-... .-........
-
Q...t ." let-kow '00 8ef,...a:.c ,.11
....,.. "I tI( t.UO
Q..2 10)6 Vp l_.cn '''-' 11.141
Q..' 1133 KI 2.64 '4w_1 0'"
Q.... ". It"' 14"'_111."
CL516. '4w_. 11."
Q..'''' ~....- ........,.,.... '"_511.-
Q..I'" .....0 f"_' .4$1
......--I.SI.I!! ILl , 00f -0. ".-' f/.S1it
~.. lno ,,",-I UtJI
Las 2no ........ "".""'..'''11- "'_1 UtJI
L'" nit 1:" 0 'pw_' AM
La' I' \;,_1 " ,,._c 11.111
L".OO W_' " ,,.J os".
Eo. 6l 600 W_' t1 rpw_' 43f1.1
(..., 7(30 "-' 0 fp._' III'"
- V_'" h_1 ,,/t1tIII
"-116tr-tO '-'-' " ""'_I II/IINIII
1\..2211[010 W_' II ""_3 IlII/IIJII
"'-' UXoOI ,"_a 4""
",-.1.511:-06 "_"....._""a...J--"I" '''w_' iI..,ItU
o "-' 3.10[001 C_wH ,...-OS '"-' 11.11.1
"-' U'loOI '''.-' a""
~ "'-' 2.83(00' ... .._.......-I."'..'_!l1I '48-1 ""X"
o ..1.111"".'411- C_1t 00380 '4--' .."X"
~ "--' 15 '48-' UX'''
C. "...2 " .............. .---.-.. '*-' 1.1."
I» ""-' 15 """'L' 2.~-1I 1«,,-' '4X'4I
..--. I' ~n" lGOr.os '''-I ".IX"
:J "'-' 15 ,,,. 2 25(.0' 'cH_' ".IX'''
CD ..~, U dell, tJ3 "'-' II"'"
en ,,«-' IS AT '"' ,,....144#
.. ~V-."""Jw"p- Ip,-S 11..#1
CD 'eL' 3.08r'06 IL, II.~ "--' II"",
...... I» ,eL2 '-6Ir.06 --.11 II UII ",_5 4'"
1'Cl.3 ,ur-o' IL'_"" 011( "--' "..,
~ C. ..4..,' ....[.05 ,-,_1l1li. a"" Ip'-' «4#
'< ,4.,5 368[.0' K, 4.SIC-M n....' f.NC-IN
. ~6 5.15[-01 'p,,,~ ~,_2 f.N£-IN
se ,4..,' 2.16(-Or "" .6Jt."., .....- r.#llC4I
I» ..-.aJ; '..... r.Me 4tI
S' M-,' OIl ..--_................,..,..,- ...... '_41
M.,I 0" AQ..ltUN '....' "'UN
Hl..30" AIL' 11.1U .....' '_41
== M..' 28 AIL3 tJ.#.IdI w_I.l.IK-i/II
o M..5 IU AD-. .Mr" ",-' I.RC4tI
HL' 218 IdLS 1.1«4.1 "'-, LMC4I
C. r-L' 1.68 AIL' 0#111 "-' '_41
!!. M...Hd 260000 AD..' 11.### ,,_5 s....VC.flS
f-.aI:; "-' "tIC..
'.t-f o.ne tAD "'" "-' .t.tAK"'"
'o~2 0.,,,
foc..' 0'''
Ioc-,one
f.t-5 O.'"
,.~, 0. n.
foc..' 0'"
I~.,4 0.032
.....-...d DOe ........i..-
...... ..
po'" ° I
.""111--. .-.--........
'co.,
b20H 0.61
_MIS. .. ....
.,ce., -v-
/C." «.It 1£." JC." /C." JC.IS /C." JC.IS /C." .IC.~ IE-I! -IC-~ I{.I! -.lE./2 D
fC. ./C41 X-o, oK-o, lUll .IUII lUll .IUII /C -QI ./C-QI fC -QI .1'(" -QI ..-c..IJ1 oK -0' D
1'("." JC." .!HD JC.1t X.1t JC.1t x.1t .JiE-" 3£-" JC.1t JE-" 4E-" /C." -IE-II) D
.4" oK -0' X. /C41 K.Qf ./C. K. ./C. K. .IC. K. ~'I!-Df I£-Df «. D
x.., -/C-I$ JC." ...'C-If fC.1t( -;C-III fC." -,y.. «.., «.11 /C-" .1£-11 X.II .IC.1t al
«-III "C-D8 X-Df ./C-Df .C41 «41 «.qr -.I~'" «." -.lC4' 1E"""0 -IE..IJ1' /C-Df ./C-Df D
1'(".11 JC.II K.IS -1£-13 X.II JC.II Ie -QI «-
-------
....'.-<11.....i'.... i.._. ......._....,......'i. i.._. ...:11......1:11..." .a.a.....~
IJ8aiaJlaJ;
Q...1 na l0t-Ko. 6.00 Iot-koc '.M
dinr,iG" t3.1 Hie 0533
11..21036 Vp 1.0'f.0I 'dw_1 /).148
Q...3 1133 KI 2.11 'dw_2 11.-
Q...~ ,,, Ie, 33' 'dw_' 11.-
Q...5 16' 'd._~ 11.-
C 11..63" ,....., . - di"".'.''''- 'ft_5 11.-
Q...J '13 ILp 0068 fdw_6 l1.'U
n' ........,.(.!I'~)- K..,I 3.00£.0. IcIw_' 11.$&1
~12 anD Ipw_1 /)..t(J1
::r ~23 2no di.... '.adct..'d.- fpw_2 (l.2()1
0" t- 3" It8S r'" 0 Ip.._3 1I..;q1
f..CS 2.6 "'-' 0 'pw_. 11.621
... L!6 '00 "'_2 0 'p,,_5 11.~
o t-61 600 "'_3 0 fpw_6 11..1U
E: t-21 1UO "'-' 0 fp,,_, /)'2~
"C .uut82J; "'-' 0 fbw_' 11.-
"'-I 2.61£-10 "'_6 0 fbw_2 (1-
::r 1\..2 2 tlE.10 "'_I 0 fbw_3 0.-
CD 1\..3 U'f.O' fbw_. 03/1
~ 1\..& 6.58f.06 h._."..~ ....oIi.....,,,,.,.,- 'b,,_'5 0.-
'< 1\..5 3.80£.08 C_wH 2.0[-05 fbw_6 0.-
::; 1\..6 UaE.O' 'bw_' 0.1n
... "'-I 2.83E-03 ai, ...........a..i...,..,.S, fds_' 6. 'b~_6 f6.t£-IN
en M...30.83 AO_2 O$M 'b~_' f 84£-IN
- FYL.. 28 1.0_3 0.1/111 "_, .2.4#C.(Jf
CD M..5 8.CI AD_. .t.64£-41 '1'1_2 .l.6,.C.(Jf
DJ M..6 2" 1.0_5 JJi'C.os "_3 UX.fJ6
Q. M-' U8 AD_6 00..'tM "-' IJK-O<
M...,ed 2C0000 "'0_1 0./N11 ¥r_5 I..tX-4$
'< fDWALIIC; ,,_6 lJ()£.fI1
I
en '0,..1 0.118 rAO 0...., "-' ~1)SC.fJ6
- 'oL2 0.118
DJ 'oL3 0118
- 10<... 0118
CD '0<-.50118
s: '0<..6 0.118
10<;"" 0.118
0 'o<..:lcd 0 032
Q. -e..dl...1 nnr. "....-ell.'
pwdoe 10
m. POlO, 0'
....1111..1_, .......I..~-
'e 05
b2elf 061
I....... .f "M"
...ct.r.V"
1£-111 -'"c-II 6£-12 -X-I? IC.I? -£-12 1£.12 -te-12 E-I.! -I£-I.t .2£-12 ...'£-13 4£-12 .../C-I.t (I
~.fJ6 .IC.fI1 SC-tJ;" ...'£-tJ; 6<41 -!C-N i'C-N .6£41 i'C-N -,Y.{)6 X-N oK-N ..'£-tJt ...~-tJt IJ
,"c-!,! ..$C-£! '!c-I() oJiC.II X-/I -x-n ~£-II ..sc-/I 1$ 2i'4£..()S :!'U£.QS $.
-------
.....,
0>
c
i.
c:
~.
:s
o
...
CD
DJ
Q,
'<
I
o
...
DJ
...
CD
3:
o
Q,
!!.
=:""1"'."
"-' ."
....... '3 I
"-, tOM
"-' 11)3
!iLl III
CLS "'
"-' 3f'
CLf 38'
.....".--'-~,..,.
E..tI.no
!...2:J ITJO
r..:ItIltU
r...S 1.1
r...-.OO
t..." 100
r..2' 1'UO
--
0\..11.1':-10
"""1...10
1\..' ,.,(.0,
t\.' 6.58(-06
"-' 3.101:-.
"-' loA.-OI
t\.' 1.'..0'
......liI."""....
.....115
,....- 15
"-' LS
....., IS
w,-S "
"-ItS
"-' U
...ha.....'~.I.
9(..1 :1.0.-
Y4..1 t.1I!-06
94...33.'.'0'
94l.1 lNr-os
" ,...-0'
"""..0,
....' UK.oI
~
PILI 0 IS
.......oea
",-,O.U
",-, II
M...S ....
h\.' 1.&8
""-' l..'
PtLH4 ..0000
Ju&II8LIIC;
'M..' 0..,'
foc...I 0.11'
'",-1 0.118
foc....O.,18
'0<...101"
'oc...' o.n.
foc...ro.fT8
loc..,M4 0.032
.......... DiIIIIIe ........
,w4M 10
pot.. 0.'
~_..&....&.... ---...."'"
leU
W:cfr 0."
,,".-"""__...,UlI........
1....1(- SSO
tic 0 02..
Vp :J.tH.04
lo,n
k, )2'
.".1.' ... - ..,-. ,..,..,.
U 3'U.O'
11:..1 300(.04
....... ........,.".,.
t'"'O
V_' #
"'_1 iI
"'_3 iI
'W'_' II
'wI_' "
W_6 tI
"'-' tJ
"..a".I- .._"..1.-'._'"
C-wtt :I 68(.0&
.1. .._....I...ti_-'..-'_~.
c..", O~.
....._"..... ....-..t..."
".....u a 00f-"
seM 1000.os
f.. 22U:-O:J
dlep 11'3
Rf 2'"
Lv .$.~#
.'IL" 4Di1
L._I.tIt .$.MC4S
lLp-hh .$.I«-IN
ICp 113E.."
",. tUtll'4
OWv Ift#C..f11
---_.....£. a...... ...,.,.
AIU IU~'
1\0..1' I1J111
AQ..' lun
AQ..4 1.«""'"
AD..S ~ SIC 4S
AD-' IJ.IJNI
AIL' 11..-
tAD IN
............ -_-.a....-
let-be I."
'''_1 6#.#1
'*-' -,NI
'*-' 6.141
,..-. ,"MI
Ift_' 6.1U
'''-' II-II
'''-' ..,.,
~w_1 «161#
'pw_2 IUIIU
,,._3 D.1tJU
'''-. 6.1"
,,._5 .'If
'pw-' 6.""
Ipw_' II./UI
fltw_' t1t'U~
'"-' 11.111»
'bw_' -.-.u
'h_.II.NI
fIN_5 #.111
1Iw_' 11.111'
fttw_' 1/..'"
'--' UK.,
hlt_' U/C-M
'--' U/r..,
141_. U/C..,
'41_' U/C..,
'._6 U!r.()l
14._' UI'E-M
Ip._' 1/.."
'P'-' I/..NI
Ipc- 3 6""
'p,_.I1..,
Ip'_' ",.
Ip$_' 1/..#1
Ip,-' 11."
'N_' ,.NC-IN
'b._, r.NL..,
fIN_3 r.NL'"
fb._. ,,,,,../111
"'s...5 f.NL-N
"'s..' r.1«-IN
Ibr...' !"CoIN
"'_, 1.11€4I
",_. 6.IM-IJtI
"_3 1.I.tC-IJtI
"_4 I."'"
,,-5 1."..41
1'c' 1.1«41
v,-' ~1»tC..
, -..-
...ct.. .V-
..........
«.. .JC." «.Il oK.1l X.1l ....fIC-" X.1l .JC'1l X-" .x./I Xofl .JC." 1£./1 -.Ie.I' "
'UII ./C4I /C" .tr"," Jr"," .Jr"," X4' *"'" X"'" .JC4' X4' .JC4' /C"" ./C" "
Jr.1l ./C." Jr." .JC." X.. oK." X-" .JC.. X.. .JC." 1£.11/1 ../£.11/1 Jr." ~C'1I/1 "
/C.". .IF"," /C4I oN4I Ie"" -Ie"" X.". *.". /C4I ./c4I /C4I ./c4I /C4I ./C4I "
X.1l .1£." X." oN." X-l» ,x-l» X." ,x." X." ,x.1I) X." ,x." x.1I) .JC." 11.1
X4l -Ie"," ... -le4f 1e41 ..41 Ie.". .Jr4f Ie" -Ie.". Ie" -Ie"" ..". *.. "
x.1$ .E.U X.II .#C.II X.I$ .JC.II IF-- -Ie -- «.11) -Ie.1I) /C.II) .#C." X.II .JC.II "
/C-l» .w-" /C-- .Jr-l» SC-l» -Ie-M X-N ./c-l» X4f oK" Jr4' .1C41 'C-- "C4I "
1C.1l -/C.II /C.II) ".Il Ie." 4E." X4I ,x- x- ..41 IF-l» .IF-l» 1e.1I) -Ie.1I) Uti,
IC"," "41 X.Qt! .."'" 1C4' -1C41 X-iN ..-iN X4I -«-iN .41 *41 IF"" .IF.. "
IC." .Ie." IC." ./C" X.1l ..." X-l» *-l» /C- ./C4I /C4I ./C4I x.1I) ".11) "
Jr4' ,x"," SC" ,x4f 6" -.Ie" K4I -.le41 X-iN ,x-iN X-iN -X-iN .K4I ,x41 "
Jr.1l ,x.1l «.11) ,x-II) X.II) -x.1I) /C-l» ./C-l» /C-l» ./C-l» /C-l» ./C-l» .'C-l» ~'C-M IJ.I1S
/C" -le41 /C4I -Ie" SC" .SC"" X4I ..41 «41 -44S 'C4S ..C4S JC41 "41 "
e'.8k81 e.-c......I..
.""1.. ..c...'''''' C. I."" . ..AI.,
~wl U/C-N 1 lU ».,
-r.-,t /1.$ 1 ., n..
;-.W.. ,"IICoN . cn 3"
... l# . - ..,
C w' ,. fie -IN , »I '0
c" ,U , UIoC 11.1
;-:: ~~: . ,» ".M1
. $C' ".M1
C_w5 I.SI'C.(/II ' fit .!2.t
C 1:5 ISI , ,~ 0'
C_" 1.1N..tN . - '.11
C .6 1fI.4 . N. ..,
C_w' .$.1«.(/11 , .""-' .$l4
C_sJ 1.11/1 , 0» ....
,.
.......,. ""
."'tfCMC(eIJI"I)" lME..
,.,'.MtfCellC(.,n). '-"IE"
n......' ~e C!tfl,). II~
cIiM8hcd ",..1...,). '10.0
pel..et lrartMMHlewff.. = O.UG
l!ipi4'IKttol:
ear (l'kt.11'
8CfCUtl,).
(OIlUIIU"'. (ItI,) ..
P.I...c
e..,wc
-. ru...... AI...,. In AIe.If. '.' L".T....
0.0100 0."" G."" 0.0800 0.""
11.- 0."' 11.- 11.- .....
11.- 11.- 11.- 11.- 11.-
0.'" O.ono 0.'" 0-'" .....
tJ.», UJfJ "~K'(» Ute-l» ..-...
U«-DS 2.001:-03 8.'0[-0. 8.'..eM ,._-
0."" ..- 0."" 0.01" 5.00E-cIS
0." 0... G."
,.,IC'" I.I/C43 .tIC'" If«4$ 'MIC-l»
,..., ... .11 CII IIJ.,
11.- 11.- '--... '-_-l» 'NC-l»
,."'" NIl I..I~"" .1..3tC4I Co."'"
U. IU 11.- 11.- Co"
ISlC.., ~."-M '.f."" ltte"
'.N Uol ,.U It'!
01.11 $1.1 IU II' M-'
IpW ,..clio..
dIea. ........ tIf .
"''''''.'''.''
,..111..... .fflchllCf"
("'''''''.flltltld) =
...~ rat. (VIII)..
ru,irw...".ltI"4) ..
........ rate IVd)
,u,.rlI,02ltld)=
..,.........(ll'ltlIII).
"(Vol).
BCI'(IrII,)..
'8M tMillnY.""'", .
8/011"'0'.
..,(11M) .
C8llC.1ltr1l8oll(.'):
c. (p,n) .., ..,...,
. (I",) ... uput
=dI3ii ~~
III :
. .
w"a eNe_....i.. (u..... I; ..,,)
I~~I
T .... Al8wi1. .
-"'3
1 .
. .
'OS
.oo
50s
-------
~......~....r... i.p.. ......,;..I-~...",,;.. ..... .....1...1....01 p.....'.'4o
--
8...1 tl6 Io~Kow 560 Ooo-<~ ,~
dI...rll'l08 8~1.I Hlc l03
1.\..2 1036 Yp 0.0506 fdw_' 0603
8...31133 ICI 2.ft fdw_2 0603
1.\..6 168 Kt 36' Idw_3 1J.8tI.!
1.\..' 163 fdw_6 OIU
Q..6"S .1..., . - "1"..'.'.'- 'dw_' 0»1
Q..T 383 ",-p 0 fd,,_6 0-
..""..f.("'~'c" "-' 3.ooE'-04 fdw_' 11.,...",
(..122no 'pw_I 4.11411
(..23 2no .i....I......I..'." 'Pw_2 11M"
(..," IU" r'" 0 rpw_' Of»"
(..45 2.6 ~~'-' I) f,.._4 0116
(..'6400 "'_2 0 'pw_5 1)4t11
!...6T 600 "'_3 0 'p._6 11..'JfJSJ
f...2r '430 "'-' 0 'pw_' o.U(J()
.au.oI82J; "'-' 0 Ibw_' Q.tN6Q
:t o\..1261E.10 "'_6 0 'bw_2 0f)/6()
0\..2211£.10 "'-' 0 'bw_3 0f)/6()
CD 0\..3 U'E.O' 'bw_4 O.!i"4
"C 0\..4 6.'8E.06 i............ .......i_i...I..,n. 'bw_' Q.,N,
- 0\..5 3.80E.08 C_wH I nE.O& 'bw_6 0 Id?$
AJ 0\..6 1.42[.0' Ibw_' ""ttJ()
n 0\..' 2.83E.0' .b ........I..II.._I_..,.~, fd,_, ,UEoN
':S' c...II_.I.'~'- C_'ilt 0.0160 Id;_2 lUC-lN
0" "JII_I U Id1_3 U4E.()j
u_215 ._...e...i..I.. .........._..... Id,_. lUE.tU
~ v'_3 15 aero,oLr 200f-1I Id1_5 IUE..()I
m ",_4 U In" 2.00[.05 rd,_6 lUE-N
",_515 ni" 225E-03 'd._' IUC..(N
"C ,,~_6 15 ddcp "3 fp,_1 un
o "'-1' 1.5 RT 2353 'p1-2 un
>< ..~....I_I..'.I"'- "'-' M.n
is: "c:L1308E.(I6 "'-" IJ.U,1 'p,_4 Mn
"4..2 265[.06 Iftu_k 0"" 'p'_5 0'"
....... CD "eL' 3.42[.0' k..,,_lalle (USt 'p'_6 0."
....... veL& 1.6&[,05 k..p_blle 000 Ip~_J 0."
en "cL5 :U8[-0' K.()I
CD l'Y\..,ud 240000 "'0_' Q.Ql5f n_' I.1X.Qf
s: ~ wI_6 lJOe.Q1
fo<-IO.118 rAD 0= no' ~OJCoOIS
o 10<..2 o.ne
Q. 10<..3 0118
'oL4 0.118
~ 10(..5 o till
10<..6 On8
10<...1 On8
10c...1~d 0.032
c.di...t nnr.: ,.....i..'
pwdoc '0
..",;~i:::..: :..._....n.
Ie 0,'
b2df 0.6J
i..cnt .r -M-
.ee'.1 .V.
..'C./Q ~." 1£./1 .sc.u !JC-12 .....'£.12 1£.12 .1£.12 1£.12 -I£-I.! ..'C.!..? ....'£-u 6£'12 -6£.D 0
SUM -JCoOIS ..'C.(J! -1C-tJf 'C.Q6 ",C.Q6 X .Q6 -3£.Q6 «.Q6 ",C.Q6 SC.Q6 .SC.Q6 ..,£.qr -.1C4Jt 0
1£.11 .K.I.? ..'C-M -IE-II) 4£.11 -4£." $£-/1 .x.1I X-II ..$C-II 1£-11 -1£-11 1£.111 .1£'10 0
x-tJr -/E.qr SC.QII -1£.Q1 $C..Qr ..x.gr I~.qr .,Y.Qt lC..Qr .g..Qt /£oOIS ./£... XoOIS -3£oOIS 0
.1£.12 .1£.12 J£-/I .JC.1/ ,Y-IO .;C-M ,''£.I.! .,Y.U Be-I.? ...J£.12 1£." .1£-1/ X." ..$£./1 01
iF .Q6 -3£.Q6 I£.QII .;C.,,,- _'C.Q1 ",C.Q1 X..q,. .../C,o! ..,£.0," .....~.(!,.. .JC.qt ..sc.qt 6C.fJf -IC.qr 0
/£.,. .g.11 ..'£.1$ ./C.IS X." -3£.,. 6C.Q6 .6£.Q6 K.1t/ -,2£.111 X." ...x.1I I£-I.? .1£,12 0
1£-/1 ....'£-11 K.III -$C-KJ ..'£.1() .1£.fI() X.()I .1£4S 1£... .1£... lE.or ./C.qr iF-I» .iF-I» 0
;'£-1$ -3£.1$ l£.fI .;c.u .2C.U .....'£-l.? 1£.Q6 .1£.Q6 .'C.Q6 ~'C.Q6 6-1» ~'C-II> /£.111 -1E.1() 0-
6C-II> ",C-I» ..'£4t -8C.Q6 X.". ~'C.Q6 .JC.()I .I£.()I ,1C.()I ~'C.()I _'C.Q1 JC.Q1 /£.QII ./£oOIS 0
IE.A.? .re.n X.II ./E./I 1£.12 .$£.12 .1£-11> ~'C-I» .1£-1» -3£-11> 'C-II> ",C-I» ..'Col(J .....'C.KJ 0
.1£.Q6 .1£.". SC41l ...$C.qr .1£.Q6 -JC.". SC"'" .1£"'" sc"'" .SC.Q1 6C"", -I£.()I 'CoOlS "'CoOlS 0
l£.J2 -3£.12 1£.111 -;"E-ll _¥-II ....,£./1 6.111 ....'£.1() 6.111 .....0£-10 ./£.", -3£.111 /£-1» .1£-1» OIAI
..'£.0, -6C.Q6 X""" ~'C'" SE..()! .sc.o, 1£""" .SC.QII SC... .SC.QII 6C.QII -8C -00 .'C"", .6£"", 0
(....lc.1 c..ceO"''tl..
..I.ti.. ..ct.r. " C. " . ,
C_wt .t61E..(J1 1 226 2,"$
C_:1' Q.6» 1 1111 .."
C_w2 .1.1i'C.Q$ 2 ."6 S."l
C,2 0,." 2 12' .2..1l
c...3 4ti'C-IJ$ 3 '" OS
C_," tit 3 .'()I 4.6l
c...' oZ..1i'C0C/6 , o.~~ 0""'"
C_,4 o.qU,p ' 2~t 0""'"
C__5 4.1/C.QS , ,,S JS<
C_,5 0- , 3t~ J$S
C_w6 4.$/C-IJ$ . 2.1.6 $.Jr
C_~ t1.846 6 ",. S$2
C_wT .$J6C.IJ1 ' 26.., ssr
C_:11 11102 ! t» ss<
I. .t
log(Kow):r 14"
w~ttf COliC (ug/l):r .t..!K.QJ
pOle "~el COliC (ug/l) c lltJC-N
,.drlll.llt cone (1I9/g):r 2.6<1
di:110md olyg~. (lI'Iglf): 100
pcl~gi( ditt 11~(tloll-~lewlh: 0620
lipidfnctioll:
(MID ~:1:1i,.jl~tlon.1I :
cllcln.tl:ut1'tlcll_",
lood~"iln elficrcllty"'
(oIl1umpt.'t.(g'g/d):
glowth f~te (Ud) :
rUpilaholi It (glg/d):
IIIttaboIr1rn'~tt(1fd)
rt~ ,. (9 02/g'd):
vpt~hentt(l/k9/d):
< (Ud) ,
BCF(lf1Ig):
'ood chill tnll:1 r~tio.':
B"'f(1/1I9):
109(8"'f'):
cOllctltntIOIl(1I9/9) :
Cd (p,lI) ., n,..,.'
r (-",) ., SIt,.nt
~~iIbu
Iihu
.i... c..u.tnti.. (n:,...t t; ..,,)
[il
-I
10
2 I
3'
, 3
, .
lipid fraction:
Ber(l/kg.l)t
BCf (1/kg) ~
cOIICCflu~trOIl("9"9) :
M sis
00400
06l1tJ
06l1tJ
0400
112$t
J ~O[-O~
00&00
P.I.,lc
AI.wife t
0.0800
0.-
0-
0800
.1~~.q.s
8&OE.04
0011&
0.00
61lC.q.s .Pi'4£.os
6U .?1.1
ft.b4l6r OOf()..1
i'S66 '$i'£..tM
'IS OU6
.UJE~ ~4..'E..o.I
4U 461
.ri'$ 10'
Tot~1 AI.wif. :
Di .rlt-i.
00300
1000
0-
0.0120
0100
2,00[.03
0.0340
.161£.tJS
1'69
OOMJ
91i',$
2..~
USC"",
'6()
0-
Bnt.ie
AI.wif. 2
00800
0-
0-
0800
J.61£.tJS
8&Of-0&
0011&
000
21~£4S
.f?"
001f»
1$,Y..ttI
0»0
ISO
211
h..Tr..'
OtJ4
0600
0-
0.800
66fC4S
& '8[-0&
500[-03
000
1..~4S
6()1
Ui'E.().J
$21£..tM
nil
.!,~£.Q1
'"
6_'"
-------
_._.._.-".~"I.. I..... ......-........."..."&,,,,,,,- ....................--...--
-
CLI ar. ~".w,n ""'
-------
...I........i,i.. ..._, ""._.".'.cp...i'i.. i.._1
JIa.-J;
~14'4 'CHj...Kow 5." Iog....ICoc .."
.""11'10.83.1 HIe 6.25
GL2 1036 Vp 2.&0[.03 fdw_1 o.m
Q...31133 K12.8r fct._2 o.m
Q...4 168 K9 364 Idw_' 0.'31
Q....5 168 fdw_4 IJ.11411
Q..' 3" ,....1 , - .;".. ,.,... f._5 tJ.I.2r
Q..1383 ",-.0 Idw_6 0.1»
.........,.'_Sl~'- K....f 3.00[-0" fdw_' f/.""~
t...12 ano Ip._' 0.111
l.23 2110 .1..... I...d..'... Ipw_2 01"
t...3H un t... 0.'00 fpw_3 Q.nr
t..4,2.6 \\1_10.111 'pw_" IJ.6/.1
t...!6 ..00 "'_2 o.OUtIJ 'pw_5 IJ.$11
t...6J 600 ""'_3 .s.u£.()$ Ipw_6 0.=
~ t..2J '430 "'-' 0..?S.1 fpw_' 0.210
uuIaZJ; "'-' 0. Ibw_1 0.0IIJ
.... "'-1 2.6IE-10 ""_6 1I.fJS2F' Ibw_2 0.0IIJ
... A..22.11E-10 "'_, l26£41 Ibw_3 0.0IIJ
D» A...3 148[-08 'bw_' 0.-
n "'-4 6.58[-06 h...":... ........"...'..'1 Ibw_5 0...'U
::T A....5 :) 80[-08 C_wH 316[.0. Ibw_6 0./6"
0' "'-6 1.&2[-08 Ibw_' 0.<."
"'-' 2.83[00' .i.. ............,...(_,'.3 'd,_1 ,....r.1C.Q!
... ....Ii..'.''',. C_,I0I20 fd,_2 ".i'2C.Q1
o "_'1.5 fd,_3 ;";1'-~.QJ
C" 's_2 '5 "'..'1:.....1.. "".!111..'..'I: 'd:_' i'.r.1C..fJJ
is' ,,_31.5 IIItfo,oLf 2.00[.n fd,_5 U2C.Q!
"-" 1.5 ~c.,. 2.00E-05 fd,_6 i'.1'-¥.()1
::T ",_5 " nlll 225E.03 fd,_J U_¥.QS
CD ,,_6 1.5 ddep H3 fp~_1 0.""
~ u_I15 AT 2353 'p'_2 0..0»
....a.I,I..'.'4'- fp~_3 0.""
...... "S. .,eL 1 3.08[.06 ....' Oi'.?S Ip$_' 0.,,"
CD .,eL2 26U.06 111111....81 0."" 'p~_5 0.""
en .,eL3 3.42[.01 1L,_I~kt 04.1r Ip~_6 0.""
CD .,eL" UU.05 h....p_I~t!e 0.00 fp'_' 0..0»
veL5 3.48[.0' Kp .t.kJt£oI» Ib,_' i'8IE.(J.I
D» 'eL6 5 15[.01 'p' OtMFI Ib,_2 '''''$lE..tN
C. ,eLI2.16l.01 'wi.. .N&2 IbL3 ,.64£..Q.I
'< - Ib,_' ;:N£..tN
. M.,.1088 ....a"l'...." I.." '..'41 fb~_S ,~&lE..tN
en M...2 0" AD_1 011'1 Ibc6 ;:64E..tN
- M...30.8' AD_2 0.4601 Ib,_1 ,""8I£../N
D» M..e 28 AD_3 0...1 .r_' Jl46C-4tI
- M...5842 "D_4 .1.14C~ ",_2 28;e-4tl
CD M...62.8 AD-5 0.0/# .,_3 1..2~#C..tJt6
M...1148 AD_6 0."''' .'-' ,.,UN
s: M.,nd 2.0000 AD_1 OQ.M .,_5 $.!X.()1
o hK1iaa.JlI:; ,,_6 tJ«.oJ
lOLIOH8 tAD 1$0 ".' ~OX.o#
C. 10(,..2 O. n8
~ 10(,..3 0118
fo(,.." 0"18
fo(,..5 one
10<..6 0,n8
fo(,..1 0.118
fo(,..ud 0032
e."I.... nn~ ..a..ei...
pwdoc 10
PO'O' 08
,..'.'Ia.i.. "".!III..I......
fc 05
b2df 061
i...I..lIeo' 8M-
gectol."""
8£.11 -4£.11 .1£.1.1 .1£.1.1 _11£./.1 .1£./.1 ~." -$C-" ./£-14 -4£-U 3£." ~." Ie." .fe." I
'C.o# ~.o# 3£.()6 -JC.()6 1£ -1M .i"E-lM .'C-IM ./C -1M .'C-IM ~'C-IM 'c -1M ..£-IM ~.()6 -$C.()6 0.
...#C.J.! .I£.~ /C./tJ -K.II Ie-a -«-f.? ..¥-I.? ......'C.f.? .1£-1.! ......,£.1.2 4£-1,! -4E.1.1 .1£.11 .JC.II 0.
I£-ot -JC.()6 JC.o# ./C.Q.J 4£-0" -4L -0; 1e.()6 .;e.Q6 3£.()6 -JC-N N:.fIr ......11£-0; X.o# ./C.o# 0.
~." .1£-1$ 1£.11 .fe-a ~-/tJ -$C./tJ x." ......'£-IS X-IS ......¥-I$ JC." .1£-/.1 ./£.12 -4£.1.1 U
1£.()6 .fe -1M JE-o; -X-IJi' .'C.Q.J ..C.Q.J JC -1M -JC -1M fE.()6 ./C.()6 X.()6 ..2£.()6 .1C.or ......#c.or 0.
~... .1£.16 1£." .fe-" x... ~-.. 1e.f16 .fC.f16 /C./tJ -1£.10 ~-t.? .,y-t.r ..'£.1$ ......11£.1.1 0.
/C.a .fe." 1£.11 -X-II g.1.? -.lC-1.! ~.Q.I ./C.fI$ ;e-o; -;"£-0; '£.f16 -$C-N JC./tJ -JC./tJ 0.
X-" ~-11 Ie." "£-1$ Ie." .fe." Ie -1M ~.fI$ X.fl$ -JC.fI$ JC./tJ .$C-KJ 1£-11 .1£.11 0.0."
~-/tJ .I£.KJ fE.()6 .fe.fl$ x./tJ -JC-/tJ I£.Q.I ./C.Q.I /C.Q.I ~'C.Q.I 1£.0# -1£.0# ...'C-or ../£.0;" 0.
IC-" -$C." .$C./,! .1£.&1 ..'£-1$ .....'£-'" 1C./tJ .fe-ill i'£-KJ .;"/:-M X -1M ~'C.fI$ ./E.II -4£.11 0.
.'C.fI$ -JC./tJ IC-N ..£-N ,Y-IM .fC.fI$ .'C.Q.J ~'C.oJ .'C.Q.J ~'C.oJ JC.oJ -JC.Q.J fE.o# .fE.fII 0.
,Y.I$ -$C." $£.11 ..2£." ~.a .....'£-12 X-II .....11£.11 X.II ...sc-II ,Y-II -6£.11 1£.111 .1£.10 0.111
~-N .I£-N fE.o# .FC-4'" E.o;" .1£.0,. 1£.fII .1£.0# 1£4)# -I£.()f ~4)# -$C.o# .'C.oJ .fC.Q.J 0.
c'e.lc.1 c..ceat..U..
,.I.ti.. ...ct.,'../I) ell'..11 , ..1.'
C_w1 IOIC.QS " " 120.
C_,1.U" 1 2" n.
C_w2 $4i'C.QS 2 "".1 /tJ~
C_~2 21.$ . ... /tJ.$
C_w3 rSJC.Q.J ' 1" 10
C sJ $.1$ , 2;".1 1M
~.... 01ltt6$ . - .>6(}
C_~& 11.1 , 66i'6 .>6(}
C_w5 2.S6£.()$ . m ".
C_'S .u I . ..~"YJ ".
C_" $tnC.Q.i . '"
-------
....._......IIa.. ~ ..""_1......_....111.. ...... ............... ----.....-
~
IliLI Cl. .....t.. .28 Iot-~ ..~.
dI¥".- '31 Hie '38
0.2103 Vp .6K.04 '4w_1 tUN
1liL3 1133 lei 210 '__I .."
Q..'.' 1e,356 '''_3 tUM
Q..5 ". '''-' tJ.INII
CL' :n. ......, . - ""0. ,aloC,. '"_5 4.1$J
ILl'., UO '"-' 4.$U
.11......~'_Sl.t IlJ :. oor-o, f"_1 .-
loG ano "'-' UI'I
1o13 1110 .a....... --.........,.,. ,,._2 1.31'1
lo:III ...1 t'" 0'" ,.,._3 4.~l'I
r...." 2' "'_I 4.111 "w_' .Uf
LS6.oo "'_I .,,1» 'pw-' .IM
r...6' 600 ""_3 $'4£-.1» l,w_6 D4#
L21 1&30 "'-' iI..,m rpw_' 11.-
..- "'-' D 181._1 6. UN
" "'-.1 261(.10 "'-' DII»1 'k_2 tJ US6
"'-.1 2tE.1O "'-' l~W -.I» "",_3 6 till
(1) .\..3 "'..01 "w_' .-
;:, "'-.. 6 51(.06 ".-.8111'. .....&.1......,,,. rbw_' 6.'-'
- .\..' 310[.01 C_wH 3 "f.eM Ih_' 6..!D
&» A....6 1.&2(.0' '''w_' lUl"l
n "-I 2.'3(.01 ..I. .................J.:S' Idt_1 I.I1M41
:r ....u..-'.,.,. c.... 012'0 'd~2 "'1£41
0" "s..ltS 'd,_3 I.ISC"
"«...2 tS ..._."..1.. --.-...... 'd,_", 4.11M"
... "-' 1$ ."..ol..f 2 oor." '*_$ ~IIK"",
o ""-' 1$ un 2oof.05 fch_' I.IIE'''
g: ....S U '.111 2.2S!.03 1._' ~.iJSC"
v,-6 1$ clckp "3 ',,-' tJ~
'C ,,-I 1$ AT 23" ",-2 "-""
:r ".~_.I"'_"..J.'. tps..3 1/.'"
(1) .tl.1 3.01E.06 '-, il..AN ""-"' UU
;:, 'eLl 1 '..06 .a..t "-1.1i1 ",_, 6-»1
(X) ~ "cL3 :UK.OI ,-,,_kilt D"~ "'L6 1/.'"
o ",-a UU-os '-P-".' "-"'" ",,-' iJ..-,
en wd-$ "...01 ., l.#X./tI .....' fi11C""
- ,""'6 S.15I:.0' Ip. tJ..Ii/II r'-L2 181£""
(1) ~I 216f.0' 'wi" 1m" '"L' r_....
&» -- '"L' r_....
Q, WL' on ...--....1. .......... f..'.'- ...... 1ISC""
M...2 Oil AD..' 4- 'b~. I"ISC -i/III
'< M....3 0.81 AD..2 iJ.NI "U 1.,.-fN
.
en hL' 18 AD..3 0.2$./ ~_t .t4K-4tI
- OIL' ua A[Lt ,,-,~£.{N ,,_2 1.#fC'"
&» OIL'I... "11..1.- ,,-' I~.
- "4..1148 AD..' 6_1 ....' l$NoM
(1) t.l,nd 240000 Alt.' 11./11 "-' S.IX..
i: ~ w_' 11«41
'~1 0111 tAD ,. w_' UX41
o '0(..10,178
a. '0(..3 0111
!!. '0c...4 0.111
'0e-5 011.
'oe...6 on.
'0,-1 Dna
loe-M 0.032
........1 nnr: ..-......
..... ..
"'0111 0.$
_....I.I~ .~.......
fe OS
We" 061
....... ., ~M-
Met.. -yo
N-n -«.n X-" -/r.II X.I$ -N." tr.1$ ->C-II tr-I$ -te.J.S X." -N.I$ X.I$ ->C.I$ 1
1e41 ->C4I tr-#l' 4C4I X""" .tr""" i'C# -Ie# i'C# .,"C"- tr""" .tr""" #C""" -Ie""" 6
N-" -/r.1I tr.1/1 -le." tr.n .1E.1f 1£." -4£.13 1£." -/E.U X.U 4C.U n." -«.n 1
NC C'" ICC." X.J' ->e.lt ,,~
N." -«.., X""" -X""" ~C# ->C'" X"" -X"" X"" ->e.... X-i11 -X41 1£41'" .1£41 ,
tr." -Ie-II 1£-" ->e-" 1£-" .tr.1S /C'" ./C... /C'" ./C... X'" -X... N-/I -«-II II
fC# -X'" Xe..,. ,
tr.U .IE.U ~r.11 -x.n g.U -4E.U /'C-lt -Ie.lt /'C.n -re-" /C.., ./E.III /'C.., -Ie." "U
Ie-N ->e-N X4I ./C..,. X
-------
..,c..--.........;';" ..,.. ......_;...I...........;.i.. ;..,.. ...1...1.1." ....-.....c
-
Q...t n
~ "-.6 U2[.09 tbw_1 0.211
o ILl 2..83£.09 .i. ".."."',,.'i..(.1P'.~' fd,_' LMc.tH
9: ..""..,.,,-1' C_" 3.16E-05 Ids:_2 lIJC.tJ.S
'5_' 1.5 'd,_3 lUC-t11
tT u_21.$ .'..'I:....rI~ ,.........e- 'ds:_' l/.!J£-tIS
CD ,,_3 1.5 iIItro'oLf 2.00[-11 fd,_5 IJJC.tJ.S
~ ,,_I 1.S il'U' 2.00E.05 'd,_6 l UiE-tIS
N ,,~51.5 ,abl 2.25[-03 Id,_l 'JJC.tJ.S
o "-' 1.5 ddtp t13 ",_, ,,~
I Y'-' 1.5 AT 2353 'p~2 ,,~
'C ".p.cil...'.'.)- Ips:....3 tJ.~
I ycl.' 308E.06 ,,-. tI.lJ6l6 'p'-' ,,~
a. .eL2 269E.06 mar...h 'USt) 'ps:_5 ,,~
(XI ()" .eL3 3.42[-01 1L,_lillh. 0.0618 Ip,_6 ,,~
.eL& 1.6.E.05 ILp_lillht ,,/U 'p~_l O~
-" ~. ,cLS 3,48E.or '. 4.S6C"1I Ib,_' 'nc.QI
.,eL6 5.15E-01 '.' "6# fbs:_2 l6SE../N
~ veL' 2.16[-01 "', ,-"" fbs_3 ,~m.()l
o ~ 'bs:..' ,"'.4'J£.{J4
en M..I 0.89 :a.._..~.....lt I..... ...'.1- 'b,_5 1'.11C-N
M..2 0.89 "0_' ~_-IN 'bs:..6 '_-IN
- 1'Y\..30.89 "'0_2 s..HC.()I Ib~_l f.6SC 44
CD M...' 20 "0_3 U/JC-tU w_' .r.46C.1J6
S» M..S tU2 AlU t8J1£.Q1' .,,_2 2.8,.OC.()6
a. M...6 2.'8 },0_5 1.«"'" ,,_3 J.a.¥:-b6
'< M...l ue "'0_6 ~U£.Q$ .,,-. '_-IN
I fvt.~td 2'0000 AO_' $.SIC.JJS ,,_5 12.$£.Q1
en 1oWaa..III:; ,,_6 lRJE .Q1
- '0<,..1 0.118 I:AD lllE.Q.S "'-' .J.0SE..Q6
S»
- fOL2 0.118
CD '0c..3 0,118
loe..." 0.tT8
s: '0(..50.118
0 '0(..6 0118
'oLl 0118
a. loc...sed 0032
~ c..i..." nnc ,......i..'
p",do< 10
po,o, 03
",:Iord.ie.i.. "2o..s..I..c'
'e 05
b2df 061
8£-/1 ../£-11 ..11£-12 -1E.~ J£-1$ ../£-13 ;"£-1$ -1£.1$ ;C-I$ -lE-1S «.1$ -$£.1$ 2E-1.? ...'£-12 q
1C.tJ.S -1C.tJ.S SC..fJ1' -1£4,.- $C.1J6 .$C.1J6 «.1J6 .lC.(J8 JC.1J6 -$£.1J6 Ie"'''' -JC.1J6 ..,£-tJ,. ....¥:./)r q
SC-I.? -1£-12 I£-It) -1£." ...11£-11 .....'JI£-fI X-II ..$£.1/ $E.fI ..x.1I 4£.11 -4E-1f 6£./1 «-" q
.sc.o; ./C.Q; 1£.tJ.S -.2C.tJ.S X.QtS ~'£.QtS ~£.QtS ..£.QtS ~£.QtS ..£.QtS K.QtS .K.QtS 1C.tJ.S -1C.tJ.S 0
fE.1S oX.1J" X.II ...#£-11 «.It) -;e.1(1 ~-I.? «.12 X.I.? .x-I.? 1£-11 -IE-II ,EC-1I ...'C." qq
JC.()6 "£.()6 ~£.QtS ~'£.QtS I£-()I .I£-()I 1£.QtS .1£.QtS 1C.QtS .1C.QtS 1C.QtS -/C.QtS X.QtS oX.QtS 0
K./S ...'£.6 ..'£-1$ ./£-1$ J£-U -.lC-1I 6£.()6 -I£.()6 X.1t) ..$£.It) $E-fl -1£-11 Je-I,! .J:E-I:! q
X-If ./C-If 1£-1» -I£.1t) ..'£-/111 -.2C-1t) ~£-()I ./C.fM 2£.QtS ~'£.QtS J'C.qr ~..q; 2£.()6 ~'£-N q
X.IJ" -.2C.1J" /C." -4C-l? X.I.? ..sc-t..! K-N ~,£.()6 2£.1J.8 ~'£.1J.8 ~£-IM ..$£-1» .!C.1t! ....'C-1tJ q"""
6£-1» oX-I» JIC.ol' .1E4J," $C.()6 .$C.()6 .£ -IN ..$£-()I ,£-()I "£-()I i"£.tJ.S -I£.tJ.S .£.tJ.tS ..£.116 q
K.J$ oX.1J" .!C.II -1£-1/ "£.I.t 4£./.1 $C -1M "£-1» KoM -$CoM 6£-IM -1£-1» X.1t) ..$£.N) 0
X.()6 ./c-$9I
tJ(u~1
Q.OIbf81
I. .,
log (KowJ" ,"..N
¥Ahr COliC (lIglI) == 13i"£.Q$
portw::r.lt'Collc(ugll)== 6.UC..tJ8
'tdilfltftICOIIC(ftg'g)== 0.(/
di"o!.,td oyygen (mgll)" 10.0
pclillglc dlt' frilietloll-iIIl~w.'e.. 0.620
lipid h::r.ctloll "
chcn. iII~~imil::r.lioatff "
ehcn. tnll~Itlt" "
1004 u~im effie/tile, c
eo",umpl 'I. (g/9/4) II'
9,owth"tt(1/d)c
lupir::to'iofl rt.. (9/9/4) c
I8tt::r.boli~1fI '::tole (Vd)
rup. II. (9 02/9/4) =
upt::tokt,::r.lt(11Jr9/d)=
k (lid) I:
OC'("'9) .
food c!\::to'" t,.a,. 'MIO, f II'
BAf(lJkg)==
109 (BAr) "
eOflctfttntion(flg'g)=
r (_,I,) ., se,.ut
c. (p,II) ., ..,.ut
..,,~
.-n
w
I.It
.
I I I . , I ,
"-
w
."
1.11
Ut
.
, I . . I I ,
.iota c..cntrati.. (..,...t t; .,1,)
:~~
I , . ,
1 00020&983
2 001882(5
3 00232.36
. 00195508
5 0.0201nr
fipldfrl)clioft:l
Ocrl"'..I).
BCf(lJkg)-
eOl'lCtfttH,tioft (lIg/g) =:
PI..'t..
0.0200
J.6l'E.()6
Ul'E-.(J1
IkJE.f1J"
M sis
0.0'00
1J.2JJ
0..24$
0."00
tI.2$r
1.30[-03
0.0'00
P.lagic eut.ic
Di o..ia AI..il. 1 AI..if. 2
0.0300 0.0800 00$00
00301 0.24J 024J
O"4J 024$ 0.241
° 0120 0800 0.800
0100 '4!J£..o.r 161£.()$
200[-03 6."0[-0' 8.40E-0'
0.03"0 0.011" 0.0114
3.50[.03 3.50[.03
6.li"£4S 2N£.f).t 2N£..()$
_%JI 6"-.2 6r2
6.S6C.()4 6.OX-tlS 6.OX-tlS
r6b£..(J.I 1.1...¥..(N 11.2C"{u
1.14 tJ.13f b..$,!$
4.M£-tH r~..(J1
168 J9(I
0.02» 1)8"" "44C-M
Total AI..ih = 0.01.16
"-61E.()$
.!J"6
1.6SC.()4
.$.0bC.(N
i'.#
I4K.()6
61t
IJ.OIM
...ct., -y-
131.4316602
h'eTro.t
on.
0,600
0;,..'
0.800
66lC..o.r
'.98E.O.
5.00(.03
350E.03
IN£.()$
»1
164£.(J1
=
IOS
1$JC.(JfJ
6.'fJ
4tA.:IfM
-------
""...._.....M&" ...... ....-.&.-.1.""..."'1. &.a.. .-.1......... .__.1....
~
""1C'C ......It- 580 ......ItM .."
4,.,.111.- ,,. I'k ales
8....1 to)6 Vp , C..os '''_1 ~,..,
...' 11:)3 lei Ifl """.' #183
fiLC ., Ie, '" '''_3 ~,.,
8....5.' ,...c """
8....' '" ""_AI . - ."°. ,.....,- '''-' ~IU
8..' 381 '-. 0 '''_6 6""
..."~'_~.'. IlL' ,.00(.0& '''...' OM4
f-. 1"0 Ipw...' 11.111:1
L,." 1110 ....d """""""00- ,,....1 11111:1
r....3H 11" ['" 0 ,,.-' .II~
r....&S.1 "_16 ,,..& -.
~ f-"COO "'_I - ".-, .~"
r.....rr.oo "'-' - r,._1 tI."'"
CD r....ar "'0 "'-' - '''-' UIM
- ....c..aJ; ioU - ,"_I II.tIntI
...
I» A..121':.1O "'-' - 'h_' 11..,..
n A..a alR.tO 'W.' 11 ""'-' .".
::r A..' UK.O' Ib-.C 11301
A..& 6.,.1:.06 ~ . '--"-"-'---")- Ibw.' 11M"
0" A..' :UOI:.oe c.... 0 'b._' ..,.
... A..I '...01 ~,.""
o A..' 2,"I:.ot ... ._.......1..'.,,'.5' '4t... I..w..
e: ..........,.t ~.. "6«.0. '*-.. 1..w41
,,_I U ,...' ..1«41
g .,....1 U .1_"".......1.. ~..._....... '.... I..w./ll
n_' 15 Hr.soL' a.CIOf.tt 14....' 1../tIC./II
:J .,...., " .c.. 1000.OS '''''6 4.w..
.,...., u ,.,." 2.22-0' '4....' .Me.
N ....., 'S .., n3 'p,-' """,
S. "Lr is AT 2353 ,,....2 II.IN
C ..r-III_'....'. 'PI...' ....,
'Hl.' 3.Get.06 ~,OAUtJ ,,....'.4N
()) ... ,4...2 2 ".06 ~. tI.", ',s...,tI.~
I»
I\) :J ,4...33UI:.0' ~,-.... .tUN ""-, #.1»
'Hl.' t6&r-os k..,...,I... UtI "'s...' II.IN
rn ~s 3."'E.or .. 1.1'N." filii...' rME4I
en 'l4..6I.R-O' Ip. #.111 ...... 'NUN
- ~'2"K..oI "'. I.IIFHIS ...... f.ME4I
CD - ....., r.ME4I
I» M...' 0.88 ...-",,~...I.. I.'" ,..1.'. ~._s tNC-tN
C. ht.10.0$ AQ."t I.IN£ 41 '.,..., 'MC...,
ht.' 0.'. AQ..I ""le41 ....., ;N~-IN
'< M.,6 I' AIL' 'BE-IU w_' L"«-I/If
. ht.SU. Alt.' ,Ite-- ~2 Llir-l/lf
en r.t..12.'" Alt.S 1'.3IIC..tIS "_3 1.,IItE..
-
I» "l.' t... AlLI I.NC- ,,_c 1._-
- M..H4 2.0000 AQ..' U4C-N 'If-' 1..RC4I
CD fudi8a.JII:; "_I '1«41
3: 10(..,10.'" [AD '"'''' ",_I ,.MC..
'0(..,2 O.n.
0 'oc...' 0.'"
C. 'oc...6 o.n.
'0<..,$ u.n.
!!. f~' o.n.
'oc...' o.n'
'0(....4 0.032
..~ ~ -_.~
'-'10
,.,.. 0.'
,.....;.1aaI... ,---......
'co.$
bo2d. 0&1
....... ., ,...
..c... .,.,-
}C.., ./E.II N." -IE.II IE.II .!E." «.11 .1C.U «.11 ~.II X-II -IE.II iC.n ./E.II -
iC~ .iC" IF.", -84' N4f ..N4' X4' ...1C4,. X41 -IE"" "C41 -84' 1E4' «4' -
N." -IE-II X." ..Jr." X," -IE." iC-" .1£-11 iC.., .I£./tl N." ..Jr." X." ..Jr.., -
«41 -841 iC4I .iC4I «.. -8" IE" -IE" IE 41 «41 IF4I «41 iC~ .iC41 -
«.11 -8-11 iC.. .fF-R iC. -iC-IU IE.II «.n IE." «-" IF." -/r-II iC." .I£./tl --
«4' ..N 4' Ie" -lE4I 1E41 ~~ X" ..N" X.. -IE" X" -IE" IE" .IE" "
IE-" -IE-" X-II -Ie." X-II -IE-II IF.. «.. X-" -IE-" re." .fF-II IE-il .~.II -
X." .E-/tl ~C-IU ..IE. iC -IU .iC-IU X-iN .iC-IU .'C 41 ..IE" ~C41 -«4' X" -IE" -
«.iI ..N.II 4U' -IE." N." ..Jr-" N" ..N" .'C" ..Jr" IE"'"' -IE-IU .,c.1t1 ~C.I# --
IE" -IE" X4' ~E4' X4f -lE4' X-iN -IE-iN X4I 4C-iN IF41 .IF~ IE 41 -IE" -
IE.II .JC-£! X." .IE.II X." -IE." IF"'"' -« -IU reoM -IF-IU XoM .,y-t» IE." «.., "
~'C-41 «.. x.. -iC" X41 ~.", N-iN ..Jr-iN .'i!-iN oK -iN .'C-iN -IE-iN N41 ..Jr41 -
N." -te-II N-III .E.. x." .IF." X-" -!C.III X.., «.111 x.., -IE." iCoM .iCoM _tJtJ
1E4' -lE4' X" -1E4I X" -IE" X41 -lE41 X41 -IE~ X.(J1 -lE41 Ie~ -lE41 -
c~~ C8M:....Mi-
.8ht. ..ce.. .
c..t 1~1C'"
:II'tl.M.U
c....2 t'E -
.. --
c..w3 .l.Me-
" .",
c...wC /..MC-61
,...Me-
C_w5 ~U£-1Jt6
C_,$ "tJ1IJ,
C..,w6 .1t1E-tJI
,6 filii'
c....' ~.1t1Co//II
:liT IIi¥#t1
c.
1 --
, ~.U
2 l...'"
. UI
:) .!.tU
. ."
. #1111
. _tM.Y
, --
S -.u
, '#3
, ~tJI
, UI
. ,»
_J8
UIII
tI.-
_",M
,,~r.!
uu
.24dC -IU
.! &JC.;M
_.11»
0--
IN_PI
_~11
-...,
--
.. .
lot(ICow)' SIII1
.IICfC08lC(..lI)a "~41
,.......erc08lC(.,n), ~#K"
HIIIMAt COM (.",), tl.1I"
.,..... ..". (8fII) a to.O
pdlltK diet "Kt_-.wif., 0.120
IIipMIfrKtiM.
CM_. ."iIIIhti.. .If. ,
.--If""',,"'.'
,..4,,- efficieH,'
.--.pt, ft. (""4) .
9"0wtll,II.(II4) .
"',.IIIO.,t.(""d).
Mtlltoi::8'M. (114)
"51'. ,t. I' 011"41'
ept,.,,".(IIII"4J'
k(V4l-
lief ....,-
,... d. tAu. ''''',' .
1M IV',)-
lot (811') -
cnc..."tt8ri (III",)..
"'IC~
lf4
--
I._oN
Ut
lllC..
~III
-'"
c. ....... ., .......
-,...,1" ..-
HII13ft
~~iH
, J:
Wet. C88C..i. ." .. 6."''''~-''''''~' I. .....
~~-
I.IH.H
..-...
._"
.....
..-..
..-.. -
t , ..
, 258.01
20-
, 0.'806
6 0.""
5"""
llpidlrlelio..
ec' P",'" -
ecr (II.,) "
coocnt,8tiOll (.,t,) I
..Ie,lc
... Alawif. .
"0300 0.0100
(\UI 11.II/11II
'tJ.1I/1III 11.II/11II
0.0J20 0800
11.- ,.<14£.
2 00[.03 8.'Of .0'
0.030&0 O.O"C
..101["",
1.11C-R 1.;11£4$
IN I"
__, ~_-IU
tlXoM UNoN
I.. 11.1'<1
«.-#11 1m..
.,." 1.1111
tJ.., t1.161
T.... AI.wiI. .
......-
00100
I.~.
~'IC~
tI...-
e...We
......
0.""
tI.-
--
.-
S.IIC.
8._.0'
0.011&
'.101:-41,
1.;11£#
I"
LNC-IU
I.-oN
,~
--
--
""aT....
o.n.
0-
tI.-
0.800
'.~4J'
'....0.
$.001:""
'.101:"'"
UC-IU
4tl.1
UIF4I
UNoN
1/.-
3.'/~"
1.SI
--
-------
CX)
to)
-I
o
><
I»
'tJ
:J'
CD
~
CD
en
..
CD
I»
Co
'<
I
en
..
I»
..
CD
s:
o
Co
!!.
.."I._~._..rIi'I... I....
--
Q...t .,"
diYufio8 $31
Q.,2t036
Q...31133
G...& 16'
Q...S '"
Q...63U
Q...1 383
.11.."....'-~'~11'
Lt2 2no
[..23 2nD
[..3" "83
f-4S 2.6
t.,.56 .00
f....61 600
t..21 1.30
.uuIaZJ;
"'-, 2.61£-10
A....22.11[-1O
A...3 1.4'[.0'
1\..' 6.58[006
A....S 3.80[.08
1\..6 U2E.O'
A....' 283f.O'
"."Ii.,~_,.).
'f_' 15
.,,_2 15
",_315
1$_415
'1:::_5 U
'1:::_6 U
11'.::_, t5
d.,8.....li...'.'.1I.
yeL' 3.08[-06
'.eL2 2.6'!:'06
veL3 3 '2£.01
veL' 1.64[-05
vcL5 3.48E-01
veL' 5.15[-07
'4..' 2.'6[-01
.IaIiddasIQ;
M-1 0.8'
M...2 0.89
M...3 0.8'
Mo." 28
M.J 8 C2
M...6 2.&8
M...T ue
M-nd 20&0000
JudUta.JU:;
'0(..1 0 ITa
'0<-2 0 He
10c:..3 0.n8
10(,.,' o.ne
'0c...5 one
'0(..6 0.118
'0<-' 0 Ifa
'o<-ud 0.032
.....i.... nnc ,......el'.'
pwdoc 10
pOfO~ 0'
."",tlti..I_".."".IIo..'.''''
'e 05
b2t" 06'
.r.._I...,",.-~.....i'i.. '.p.'
109-IC.. '.'2
Hk 0.0152
Yp 3..2u-oa
.....
1e,323
....., .. - ""'. ..,..-
......0
KJ 3.00£-04
.Ii...... I........,..,.
1;'" 0
"'_11/
"'_2 1/
"'_3 1/
"'_4 fJ
"'_5 IJ
""_6 (J
"'-' 1/
..-".81:. r.."I.'..'.."'.
C_wH 0
.1. ....r...I."".i..f..'.3,
C_,t 0.181
""...c....I.. _30130.....C.
'IoCfOfOLf 2 OOE.tI
,cn 2.ool.05
"1ft 2.25[.03
ddep tr3
AT 2353
fL' fJ.(J1fJ1
ftI,,,-k fJ.1JtJ
IL'",-"he fJ.(JkJS
ILp_l,ke fJ."'"
ICp l8SC#'/If1
'p' lJ.ar"
'vi, r.#1C#'/U
...."1,...1.. I..... ~.,'''II.
AD_1.1.1i"
1.0_2 llt
AD_3 IIUI
1.0_4 1../1£41
1.0_5 IJ.#$IS
AO_6 fJ.11I
AO_' IJ.ISf
E""o./.n
.......1.... ..a.""..''''''
IOIJ-Koe s.~,
IcI,,_t o.~2
fd._2 1J..N.t
Idw_3 tJ.N2
'd._' O.OUZ
'"_5 1J.6SI
Id.._6 tJ6S./
'dw_' 1/408
'pw_1 fJ.1JSN
'p.._2 (J.(J$8I
'p"'_3 fJ.DSN
fpw_& Q~
'pw_5 Q-~~
'pw_6 OMi"O
'pw_' Q.Ct611
IbwU I/(JW
'bw_2 QQW
Ibll'_' Q(JIM
Ibw_4 Q.!U
fbw_5 fJ.1.?.?
'bw_6 o.tM61
'bw_1 (JDSOG
'ds_1 ./ IKo(U
'd,_2 ~ 11£.(N
,d,_' ~JSC.(N
,d,_& ~ISCo(U
'd,_5 ~ IJ£.(N
'd:L6 ~11£44
'4,_, ~1S£44
'p'_1 o.~
'p'_2 0.""
'p,_3 0."""
Ip,_& "".",
fp,_5 fJ.~
Ip,_6 0."",
fp,_' o.~
'b,_1 ,"'84£44
Ib,_2 r:.6.lE.(N
'b,_3 fUC-N
fb,_' ,.$i£..(N
fb,_5 f.6./£..&W
fb,_6 ,"'N£44
fb,_, r:..uE.Q.I
"'-' 2~6'£-b6
"_2 ~~.(16
'1"-' J.22£.1J6
'1',_4 l$K44
YI_5 I..!JIC-t1S
",,_6 IJQ£-t1S
YI_1 .t.1»C-tJt6
i..erst 0' "M"
.tct.r "'fl.
4£-w) .../£.1tJ ..'C.1f) -1£-/1 l£-/1 06£." E-w) ..!/C.II E-W) .1£.1(1 E-w) .1£.1(1 E-W) -E-w) 3
4£.... -E.tJ1 1£.... -6£ -fir lE.IJ,. -6£.lJi' 1£.... ...K.lJt 1£"" -tC"" E"" -E"" E"" -1£"" fJ
..,£.11/ ..6'£.11 4£-w) .1£.'" ./£.1(1 4C.'" I£-w) .1£.'" I£-w) -6£-w) I£-w) -6£-W) 4£-w) -6£-w) 3
2<:"" -4£ -fir 4£"" -E.tJ1 IC"" olE.... 1£.... -rUM 1£.... -6£.... 1£.... -6£.... 4£.... -4£.... fJ
E-W) .$C.II Ie-w) -JC.W) Ko/» ~>Co/» X.W) -JC-W) ~£.1tJ 4£-10 ~C.KJ 4C.KJ Ie-W) .re.W) 06
1£.... .I£.IJ,. Ie"" -JC4f 2<:.tJ1 -IC.tJ1 IC4f -JC"" 'C"" -IC.... .NJ6 -IC"" Ie"" .Ie"" 0
,"£.1$ 4£.1$ ./£-1.1 ....OC-J.! .!!£.,p .1£.,2 lC.tJ6 -6£.tJ6 ./£.10 4C-N) /C.JO 'IE-If) 6£./1 .1£./1 0
Xo/» -Eo/» /C.tJ6 -,y~ 1£0/» -6£0/» .'C.fJI -/C-D$ /C"" ./C-tJ6 ~C",H 4C..(J; .C.tJ6 -IC.tJ6 fJ
Ie-n ..tC-II K-w) ./E.KJ /C-JO -1E-1(J .>C.tJ6 ~>C.tJ6 X.tJ6 -JC.tJ6 fC -1M -6£0/» ,"C.- .,y.1O fJ=
~C.(Jl ...."E.lJt .>C-tJ6 ~-flr lC.IJ,. .,,,£.or .'C.fJI ~>C.fJI .>C.fJI -K.fJI 1e.tJ1 -Ie -0> Ie"" -le-tJ6 4
6C-n 4C.1I ./C.", ....~.'" K-w) -K-w) Xo/» -4£ -M tC.tJ6 ~o/» /C.tJ6 -1e.tJ6 Eo/» -/Co/» 0
i'E",1f .sc .(11. X-tJ6 ~>C4f 2<:.... -/C-tJ6 X.tJ1 -4£.tJ1 X.tJ1 ~.tJ1 X.tJ1 ./C -tN X-tJ6 ~-tJ6 fJ
-'C.fI() -FE.II 4£-w) ~E.1t1 X-W) -JC.W) Ko/» -/Co/» No/» ~'Co/» .'Co/» ~>Co/» Ko/» -Ko/» (1.2.1
K4f -,y.ql toe.... -IC4f X.... -JC-tJ6 /C.tJ1 -/C.tJ1 .>C.tJ1 ~>C -0> .>C -0> ~>C-O> _>C.tJ1 -6£.tJ1 fJ
c..aln,l c..cc..lnti..
sol.tie. .ut.. . ,.
C_w1 .!.NC-#$
C_:::t ..'fl..!
C_w2 .t.,si'E..().1
C_,2 2$.6
C_w3 .t.Me-IM
C-::::3 ./!6.$
C_",& 1'.1J1C.(JJ
C-,. Q_'41
C_w5 .?$6E.()$
C_d 16~
C_w6 .t.66E.fM
C_,6 .N.6
C_w1 2.6E.o.t
C_,1 .t1.,.
Cd I.
I 3..'0
I ~IM
2 21Sf
2 ll)l'£...Q.I
3 2~./1
3 l31C-tU
& ~~.I
. al
5 U$I
5 ;"U
6 2.ti'J
6 l«..,
, ~.ou
1 l.Ii'Eo.IN
. . ,
6f"
614
11M
N.
la
IW)
IW)
'JO
661
66.3
IIN
ItM
106
W)"
I. .t
log (Kow) '" ./.8.1
w::IIhr cone (119/1) '= lMCoQ;S
porew'tCfconc(lIg/l)c fJ.1JtIJ4.!
'cdbtleJ'llconc(flg/g): N
dl"olvcd 011)191:11 (mg/l) I: to 0
pcl~ic diet "'
-------
PART 2
2002 LAKE MICHIGAN MASS BALANCE PROJECT: MODELING TOTAL
POLYCHLORINATED BIPHENYLS USING THE MICHTOX MODEL
Douglas D. Endicott
Great Lakes Environmental Center
Traverse City, Michigan
2.1 Executive Summary
The MICHTOX model was used to perform a
preliminary mass balance modeling assessment for
polychlorinated biphenyls (PCBs) in Lake Michigan.
Comparison of model predictions to data generated
by the Lake Michigan Mass Balance Project
(LMMBP) provided a unique opportunity to confirm
the MICHTOX model using data not available at the
time of its development. Principal results of the
MICHTOX assessment include:
1. Total PCBs forcing function estimates from the
LMMBP are compatible with the original
MICHTOX model estimates for atmospheric
deposition and tributary loading. However,
atmospheric total PCBs vapor concentrations
from the LMMBP are significantly higher than
estimated in the original model. Consequently,
all total PCBs forcing functions were recalculated
using the LMMBP estimates. '
2. Changes to the model formulation and
parameterization, recommended by the LMMBP
Atmospheric Workgroup, enhance the volatility of
total PCBs and increase the volatilization mass
transfer rates. As a result of these changes,
PCBs equilibrium was shifted significantly
towards the atmospheric vapor phase, and this
shift occurred more rapidly than previously
predicted. The enhanced volatility of total PCBs
was found to completely offset the greater
absorption resulting from higher vapor
concentrations. Mass balance diagnostics also
demonstrate that air-water fluxes now clearly
predominate the transport pathways for PCBs in
Lake Michigan.
3. Model simulations were conducted using different
assumptions regarding long-term total PCBs
forcing functions. The most reasonable
predictions were obtained by assuming that total
PCBs forcing functions peaked in 1961-1963.
The simulation made using this assumption
offered a better prediction of the total PCBs
concentrations observed in water, sediment, and
fish than the original MICHTOX model simulation.
4. The model was applied to forecast total PCBs
concentrations in lake trout for a number of
scenarios in which the future PCBs forcing
functions were changed from their 1994-1995
estimated values. These changes were intended
to represent alternative strategies for managing
PCBs in Lake Michigan, and the model was used
to forecast the effectiveness of these alternatives
in terms of reducing lake trout total PCBs
concentrations. The results of the toxic chemical
management forecasts demonstrated that
properly evaluating the effectiveness of control
action depends upon understanding which forcing
functions are controllable and what the future
trend in forcing functions (especially atmospheric
84
-------
vapor concentrations) will be in the absence of
control actions.
5. The uncertainty of MICHTOX predictions arising
from errors in model parameterization and forcing
functions was evaluated using Bayesian Monte
Carlo (BMC) analyses. Based on the uncertainty
in model predictions arising from these factors,
observed average total PCBs concentrations
should be well within a factor of two of predicted
values.
6. The MICHTOX model does not predict
equilibrium between air and water total PCBs
concentrations.
7. Total PCBs bioaccumulation predictions were not
sensitive to initial concentration conditions in fish,
after an initial simulation period. Fish total PCBs
concentration predictions were demonstrated to
be quite sensitive to bioaccumulation model
parameterization.
8. Bioaccumulation factors were predicted to vary
continuously throughout the model simulations.
9. The MICHTOX model was found to provide a
reasonably accurate simulation of total PCBs in
the Lake Michigan ecosystem. However, the
preliminary MICHTOX assessment will be
superceded in the next several years by mass
balance models with significantly better resolution
and better process and state variable
representations. Furthermore, these newer
models will be capable of making full use of the
LMMBP data; for example, they can be used to
model PCBs as individual congeners.
2.2 Recommendations
After revising and successfully running MICHTOX as
a screening-level model, a number of issues became
apparent which require further consideration. These
are as follows:
1. For the forcing function assumption used, the
model does not predict equilibrium between the
air and water total PCBs concentrations. Other
forcing functions scenarios should be done to
determine whether this is a general result or a
result specific to the chosen function.
2. Chemical assimilation efficiency, diet
composition, growth rates, lipid contents,
chemical excretion rates, and phytoplankton
bioconcentration should be examined in any
recalibration of the model.
3. Bioaccumulation factors should be used with
caution in Lake Michigan because they are
expected to vary with time.
2.3 Introduction
The LMMBP is a coordinated effort among federal,
state, and academic scientists to monitor tributary
and atmospheric pollutant loads, develop source
inventories of toxic substances, and evaluate the fate
and effects of these pollutants in Lake Michigan. A
key objective of the LMMBP is to construct mass
budgets and mass balance models for a limited
group of contaminants that are present in Lake
Michigan at concentrations that pose a risk to aquatic
and terrestrial organisms within the ecosystem. The
mass balance modeling is being conducted to
support both regulatory and research agendas, as
described in the LMMBP Study Plan (U.S.
Environmental Protection Agency, 1997). Elements
of the LMMBP which provide information for the
modeling objective include:
1. Monitoring of atmospheric and tributary toxic
chemical sources and estimation of loadings and
other forcing functions at spatial and temporal
scales necessary for mass balance modeling.
2. Measurement of toxic chemical concentrations in
lake water at 41 stations and six cruises (Le,
sampling events) over the course of the two-year
mass balance study period.
3. Concurrent measurement of suspended solids,
organic carbon, and nutrients in all water column
samples.
4. Measurement of toxic chemical and associated
state variables in surficial sediments throughout
Lake Michigan.
5. Measurement of toxic chemicals in representative
biota for pelagic and benthic food chains for two
top predator fishes, lake trout and coho salmon,
for multiple locations and seasons.
85
-------
The sampling design and analytical parameters of
the LMMBP were specified to support the
development of hydrodynamic, sediment transport,
eutrophication, contaminant transport/fate, and food
web bioaccumulation models (U.S. Environmental
Protection Agency, 1997). The various models and
submodels that comprise this system are under
development by the United States Environmental
Protection Agency (USEPA) at the Large Lakes
Research Station (LLRS), Grosse lie, Michigan.
This report presents a preliminary mass balance
modeling assessment for PCBs in Lake Michigan
using the MICHTOX model (described in Part 1).
MICHTOX was developed as a planning tool for the
LMMBP using information available at the time to
construct a screening-level model for the transport,
fate, and bioaccumulation of total PCBs in Lake
Michigan. Because of its availability, MICHTOX is
now being applied as a tool for the rapid, preliminary
assessment of the LMMBP data for PCBs.
MICHTOX was previously applied in a similar manner
to assess atrazine (Rygwelski et al., 1999), another
toxic chemical prioritized by the LMMBP. The
MICHTOX assessment is, in part, being conducted to
generate preliminary modeling results for inclusion in
the 2002 Lake Michigan Lake-wide Management
Plan (LaMP) Report. This assessment includes
comparisons of MICHTOX simulation results to total
PCBs concentration data generated by the LMMBP.
Such a comparison provides a unique opportunity to
confirm the MICHTOX model using data not available
at the time of its development. The assessment also
compares total PCBs forcing function estimates from
the LMMBP to those estimated previously for
MICHTOX. In addition, the model was applied to
forecast the effectiveness of several toxics
management alternatives for PCBs in Lake Michigan.
It should be recognized, however, that the
preliminary MICHTOX assessment will be
superceded in the next several years by mass
balance models with significantly better resolution,
better process, and better state variable
representations. Furthermore, these newer models
will be capable of making full use of the LMMBP
data; for example, they can be used to model PCBs
as individual congeners.
2.4 Description of Model, Data, and
Simulations
2.4.1 MICHTOX Model
The MICHTOX model was developed to simulate the
transport, fate, and bioaccumulation of PCBs and
other toxic chemicals in Lake Michigan. The
development and original application of MICHTOX
are documented in Part 1 of this report, which
includes a thorough description of the model. A
schematic diagram of the MICHTOX contaminant
transport and fate model is presented in Figure 2.1.
Figure 2.2 displays the spatial segmentation of the
MICHTOX model, which includes 17 water column
segments (divided into epilimnion and hypolimnion
segments in the main lake), and seven surficial
sediment segments. MICHTOX was implemented
using the USEPA WASP4 modeling framework
(Ambrose et al., 1988) and the Manhattan College
food chain model (Version 3.20; Connolly, 1991).
The model was originally developed and run on a
MicroVax minicomputer and was later modified by
the USEPA to run on Compaq (Digital) Alpha OSF1
workstations.
In October 2001, the Great Lakes Environmental
Center (GLEC) began resurrection of the MICHTOX
source code and input data using Secure Remote
Access to connect to the USEPA workstation IIrssrv2
over the Internet. All necessary files were identified
and reorganized in a user directory, -dde/MICHTOX/
GLEC. The original MICHTOX PCBs simulations
were rerun to confirm that the model reproduced the
earlier results. Because Secure Remote Access was
disrupted in December 2001, it was necessary to
move the model to GLEC personal computers for
final PCBs simulations. The MICHTOX source code
was C preprocessed at LLRS, e-mailed to GLEC,
and recompiled using Compaq Visual Fortran 6.1.
Results of personal computer simulations were
compared to the same simulations run on Alpha
workstations, again to confirm that model results
were independent of the computer platform.
The goal of this Work Assignment was to rapidly
evaluate the LMMBP data for PCBs; therefore,
MIC~TOX was not significantly revised or
recahbrated. However, a number of modifications
86
-------
Watershed
Tributary
Loading
Epilimnion
Hypolimnion
SLirficial
Sediment
Subsurficial
Sediment
Layers
Deep
Sediment
Atmosphere
r.: - b~u~d- - - - - dissolved - - - ~orbed chemical ~
: ~_<;h~mt~L l1li(". chemical J --sfc--- ---P-C,-C-- I
I~_D~~_- ""------ /_--- ____I
Transport and
Exchange
Exchange
Resuspension
Settling
----------- ---1
~I - bound - -. sorbed chemical I Exchange
chemical l1li( > dlssolyed
I U - -O
-------
'\
Straits of
Mackinac
'...,
~
><
o
u..
~ seasonally-stratified
W water column
r-;:;--ln completely-mixed
~ water column
Q!l] surficial sediment
r-. grey scale intensity
L...;;...J Indicates lake depth
Lake
Michigan
'~
Chicago
Diversion
Figure 2.2. Spatial segmentation for the 17 segment MICHTOX model.
88
-------
were made to the MICHTOX model for this work.
These included:
1. Correction of water balance to maintain continuity
- The specification of advective flow between
water column segments in the model input
contained small errors, so that the volume of
some segments increased during model
simulations, while the volume decreased in other
segments. The problem was most severe in
Segment 3 (northern lake epilimnion), which lost
1.7% of its volume each year. Consequently,
very long (>100 year) MICHTOX simulations
aborted, because Segment 3 volumes eventually
reached zero. To correct the flows so as to
maintain continuity on an annual time scale, small
adjustments were made to June surface layer
flows, entrainment flows, and southern lake
tributary inflows.
2. Treatment of the boundary condition at the Straits
of Mackinac - The boundary condition at the
Straits of Mackinac was modeled as a
seasonally, bidirectional flow. To specify
concentrations of model state variables for the
reversing flow components, available data,
including the LMMBP data, were examined to
estimate the concentration gradients across the
Straits of Mackinac. The LMMBP total PCBs
data for both surface and deepwater samples at
Stations MB72M (northern Lake Michigan) and
LH54M (northern Lake Huron) suggested that
Lake Huron boundary concentrations were about
5% lower than concentrations in northern Lake
Michigan. Therefore, boundary concentrations
were specified to be 95% of the value simulated
for northern Lake Michigan in the previous model
time-step. This treatment is consistent with other
analyses, such as Robbins (1985) and Thomann
and Di Toro (1983), which suggest that similar
concentration trends are expected in northern
Lake Michigan and Lakes Huron and Superior.
3. SDecification of atmosDheric vapor
concentrations - The model program was
modified for the input of atmospheric vapor
concentrations as segment-specific forcing
functions. Originally, a single forcing function
was used for atmospheric vapor concentrations
in all MICHTOX surface water segments.
4. UDdating of chemical volatilization rate
formulations - Revision of the mass transfer
formulations used to calculate volatilization rates
was recommended by the LMMBP Atmospheric
Workgroup. Specifically, the Workgroup
recommended the Wanninkhoff (1992)
formulation for water mass transfer resistance
and the Schwarzenback et al. (1993) formulation
for gas mass transfer resistence as being the
most appropriate for modeling the air-water
exchange of PCBs in Lake Michigan. The
volatilization rates input to MICHTOX were
recalculated on a monthly basis using these
formulations. In addition, the subroutine used to
calculate chemical volatilization in the LMMBP
Level 2 and 3 mass balance models were also
modified to include these formulations.
5. UDdatina Darameterization for Henry's constants
- Another recommendation of the Atmospheric
Workgroup was to make use of recently
developed data for the Henry's Law constant for
PCBs congeners published by Bamford et al.
(2000). These data included measurements of
Henry's constants at different temperatures. A
model which extends these data to the
predictions of Henry's constant for all PCBs
congeners was obtained (Bamford et al., 2002)
from the authors. The Bamford model was
applied to predict values of Henry's constants for
PCBs in MICHTOX, using average monthly
surface water temperatures. The Bamford model
was also implemented as a spreadsheet and
provided to the USEPA for inclusion in the Level
2 and 3 LMMBP mass balance models.
6. DeveloDment of an alternative formulation for
disDersive water column transDort - An
alternative formulation for dispersive water
column transport was developed for MICHTOX in
1997 (M. Settles, personal communication). This
was implemented to replace the original WASP4
mass transport term, which was based on a
simple concentration gradient formulation. Tests
conducted with both versions of the model
indicated that predicted PCBs concentrations
were about 15-20% higher in hypolimnetic water
column segments in model simulations using the
alternative formulation; differences were smaller
in other model segments. This modification,
intended to address a potential instability in the
89
-------
model, was apparently never fully tested.
Therefore, the original formulation for dispersive
water column transport was restored in the
version of MICHTOX used for this work.
2.4.2 Mass Balance Data for PCBs
MICHTOX resolves total PCBs (Le., the sum of
congener concentrations) as the sum of two
homologs, tetrachlorobiphenyl (PCB4) and
pentachlorobiphenyl (PCBS). Although this is
technically incorrect (Le., there are ten homologs),
this representation of total PCBs as two homologs
was considered a reasonable compromise between
pre-LMMBP loading and concentration data, mostly
quantified as total PCBs and/or Aroclors, and
congener-specific estimates available for
physicochemical model parameters.
PCBs concentrations were reported in the LMMBP
for some 90 congener peaks; 36 were selected for
mass balance modeling based upon their
detectability in various media. Forcing functions
(atmospheric vapor concentrations, deposition
fluxes, an,d tributary loadings) were also estimated for
total PCBs and the selected congeners.
Unfortunately, similar estimates were not developed
for PCBs homologs. Thus it was necessary to use
the total PCBs forcing functions in MICHTOX. Total
PCBs forcing functions were evenly split between
PCB4 and PCBS homologs, the same assumption
that was made in the original MICHTOX application.
It was not possible to confirm this assumption within
the time constraints of this project. The total PCBs
forcing functions and concentration data are
summarized below:
1. Atmoscheric vacor concentrations- The LMMBP
atmospheric data were used to develop regional
spatial and temporal interpolations of PCBs vapor
concentrations (Green et al., 2000). The 5 km
interpolated total PCBs vapor concentrations
were averaged on a monthly basis for each
MICHTOX surface water segment. The resulting
time functions are plotted for the three main lake
surface water segments in Figure 2.3. Patterns
of both seasonal and spatial variability were
evident in the vapor concentration estimates.
Because the model predictions of surface water
concentrations were found to be sensitive to the
monthly variability of PCBs vapor concentrations,
this and the other forcing functions were specified
on a monthly basis for all MICHTOX simulations.
2. Atmoscheric decosition fluxes (wet and dry) -
The LMMBP atmospheric data were also used to
develop regional spatial and temporal
interpolations of PCBs wet and dry deposition
fluxes. The deposition fluxes were averaged on
a monthly basis and converted into loadings
using the surface area of each MICHTOX surface
water segment. The resulting time functions for
wet and dry deposition are plotted for the
southern main lake surface water segment
(Segment 1) in Figure 2.4. Estimated deposition
fluxes for total PCBs were much higher in 1994,
the first year of the LMMBP, due to the enhanced
wet deposition flux in the spring of that year.
3. Tributarv 10adinQs - Ten Lake Michigan
tributaries were monitored during the LMMBP,
and the data were used to estimate loadings of
toxic chemicals and other constituents to the
lake. Tributary loading estimates for total PCBs
were summed for each surface water segment,
and input to the model as monthly average
loadings. The tributary loading time functions for
model Segments 1 and 5 are plotted in Figure
2.5; tributary loadings to other segments were
very small. As was the case for the other PCBs
forcing functions, there was a definite pattern to
the seasonal trend of tributary loadings, with
highest values estimated in the spring of each
year.
4. Water column concentrations - Total PCBs
concentrations were averaged for each cruise
and water column segment using a volume-
weighted averaging (VW A) procedure. Dissolved
(filtered) and particulate total PCBs
concentrations were averaged separately.
Dissolved fraction of total PCBs concentrations
are presented in Table 2.1, and particulate
concentrations are presented in Table 2.2.
5. Surficial sediment concentrations - Total PCBs
concentrations were measured in 133 surficial
sediment samples; 50 were collected from the
top 1 cm increment of box cores, and 65 were
Ponar samples. All total PCBs data were
interpolated onto a uniform grid using a natural-
neighbor algorithm and then averaged for each
90
-------
<0
~
3::!!
Ota
::sc
- ..
::rCD
-<::!:
(1)3
CD I»
CQ-
3 m
CD 0
a-
0-
...A 2,
'I»
~-
"'tJ
o
m
o
I»
-
3
o
o
"
::r
CD
..
n'
D1
"
o
...
()
o
~
()
CD
:J
-
...
I»
-
0'
:J
o
"
...
o
()
CD
o
o
CD
Q.
I»
o
1.6
1.4
1.2
-
C")
E
-
C)
c
-
c
o
:;:::
I.!
-
C
ell
U
C
o
u
...
o
Do
ra
>
m
o
D.
-
0.8
0.6
0.4
0.2
--. -. seg-1
- seg-2
..-.
. ,
-- seg-3
..
. .
,
.
------.
.
o
o
10 12 14
month (0 = January 1994)
6
8
16
18
2
4
20
22
24
-------
3!!
oco
::::s C
~Ci1
-< -.
cn3 1.2
CD!.
co CD -
3 0 'C
Q
CDS, ~
::::s-
-0 m
c
...1.- ;;
-I»
0- IV
O-g ..2
CO u 0.8
~OJ 'i:
CD
CD 0 .z:::
CD .. I» a.
I\) ::::s ... ..
.3 0
I» 0 5 0.6
~0 IV
CD'tJ m
3:~ (,)
-. CD Q.
n" - 0.4
2;n
co~
I» CD
::::S...
-
. I» 0.2
::::s
Q.
Q.
-<
Q. 0
CD 0 2 4 6 8 10 12 14 16 18 20 22 24
'tJ
0 month (0 . January 1994)
0
::;:
o'
::::s
'tJ
a
n
11
i
I»
en
-------
c.o
VJ
i:::!!
ocg
%~
-tCD
01\)
>
-------
Table 2.1. Cruise- and Segment-Specific Dissolved Fraction of Total PCBs Concentrations (ng/l)
Date Seg-1 Seg-2 Seg-3 Seg-5 Seg-6 Seg-7 Seg-8 Seg-9 Seg-10
May-94 0.53 0.51 0.35 0.36 0.36 0.36 0.54 0.50 0.34
Jun-94 0.59 0.64 0.61 0.59 0.60 0.58
Aug-94 0.85 0.85 0.84 0.50 0.49 0.77 0.69 0.75 0.75
Oct-94 0.67 0.81 0.83 0.40 0.40 0.61 0.68 0.81 0.84
Jan-95 0.54 0.54 0.54
Apr-95 0.66 0.68 0.66 0.36 0.37 0.49 0.66 0.67 0.66
Aug-95 0.82 0.86 0.88 0.74 0.74 0.74 0.71 0.79 0.81
Sep-95 0.87 0.88 0.87 0.31 0.31 0.53 0.83 0.87 0.87
Table 2.2. Cruise- and Segment-Specific Average Particulate Total PCBs Concentrations (ng/L)
Date Seg-1 Seg-2 Seg-3 Seg-5 Seg-6 Seg-7 Seg-8 Seg-9 Seg-10
May-94 0.147 0.137 0.132 1.653 1.659 0.272 0.114 0.120 0.123
Jun-94 0.088 0.065 0.071 0.090 0.070 0.072
Aug-94 0.031 0.030 0.024 0.574 0.577 0.103 0.080, 0.055 0.045
Oct-94 0.102 0.042 0.030 0.608 0.598 0.121 0.099 0.045 0.030
Jan-95 0.138 0.136 0.138
Apr-95 0.099 0.084 0.075 0.805 0.789 0.245 0.089 0.073 0.071
Aug-95 0.046 0.026 0.025 0.048 0.050 0.064 0.111 0.058 0.050
Sep-95 0.036 0.023 0.020 0.572 0.571 0.126 0.052 0.026 0.025
MICHTOX surficial sediment segment. Average
total PCBs concentrations were also calculated
for the box core samples in each main
lakesediment segment (Segments 11-13).
Relatively few sediment samples were collected
in Green Bay. Thus surficial sedimenttotal PCBs
concentrations measured in box cores collected
during the 1989-1990 Green Bay Mass Balance
Project (GBMBP) were used to calculate average
concentrations in Green Bay sediment
(Segments 15-17). Segment-specific average
total PCBs sediment concentrations are
presented in Table 2.3.
According to the LMMBP Quality Assurance
Project Plan (QAPP), samples from deeper
intervals in selected sediment cores were to be
analyzed for PCBs in addition to the 0-1 cm
interval. However, these data were not available
at the time of this report.
6. Biota concentrations - Fish and lower food chain
organisms were sampled in three biota zones.
Two of the zones, Saugatuck and Sheboygan
Reef, fall within the southern lake basin
(Segments 1/8). A third zone, Sturgeon Bay, is
located in the northern lake basin (Segments
2/9). It should be noted that there was some
confusion regarding the identification of biota
samples from Saugatuck and Sturgeon Bay in the
database from which these data were retrieved.
Although GLEC believes that the data have now
been associated correctly with biota zones, this
has not been confirmed by the USEPA. Although
biota sampling was conducted on a seasonal
basis, all samples were averaged for use with
MICHTOX. Total PCBs concentrations measured
in the three biota zones, based on age or size
classes, are presented in Table 2.4(a-c).
94
-------
Table 2.3. Segment-Specific Average Surficial Sediment Total PCBs Concentrations (ng/g)
Segment
Surficial Samples
From LMMBP Box
Cores
All LMMBP Surficial
Sediment Samples
56.2
35.2
4.99
17.1
127
52.9
11
12
13
15
16
17
102
63.4
27.9
Surficial Samples
From GBMBP Box
Cores
695
643
97.3
Table 2.4a. Average Total PCBs Concentrations in Fish in the Saugatuck Biota Zone
Species
Alewife < 120 mm
Alewife> 120 mm
Bloater < 160 mm
Bloater> 160 mm
Deepwater Sculpin
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Smelt (Adult)
Slimy Sculpin
Average PCBs
Concentrations (ng/g)
304
592
586
875
340
175
904
883
1287
2068
3185
3609
4511
5728
8209
7477
8116
6666
6799
4014
294
390
Age (Years)
1-2
3-7
1-3
4-7
4+
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1-7
1-6
PCBs Standard
Deviation (ng/g)
167
140
201
250
101
171
288
241
532
1126
809
921
1645
4101
2515
2997
872
794
3268
69
149
95
-------
Table 2.4b. Average Total PCBs Concentrations in Fish in the Sheboygan Reef Biota Zone
Species
Alewife < 120 mm
Alewife> 120 mm
Bloater < 160 mm
Bloater> 160 mm
Deepwater Sculpin
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Smelt (Adult)
Age (Years)
1-2
3-7
1-3
4-7
4+
3
4
5
6
7
8
9
11
12
13
14
1-7
Average PCBs
Concentrations (nglg)
347
540
753
876
427
547
706
1202
1395
1974
2668
3102
5322
4692
4466
3483
305
PCBs Standard
Deviation (nglg)
211
106
132
148
90
184
217
204
192
320
1001
1022
1215
1234
217
133
Table 2.4c. Average Total PCBs Concentrations in Fish in the Sturgeon Bay Biota Zone
Species
Alewife < 120 mm
Alewife> 120 mm
Bloater < 160 mm
Bloater> 160 mm
Deepwater Sculpin
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Lake Trout
Age (Years)
1-2
3-7
1-3
4-7
4+
1
2
3
4
5
6
7
8
9
10
11
12
Average PCBs
Concentrations (nglg)
170
589
604
739
325
350
395
889
1268
1707
2487
2656
3360
4211
5283
5939
4420
PCBs Standard
Deviation (nglg)
71
171
155
189
59
163
107
159
270
309
577
509
559
757
1168
1543
1185
96
-------
2.4.3 Revised MICHTOX and LMMBP
Forcing Functions
New total PCBs forcing functions were developed for
the MICHTOX simulations presented in this report.
As mentioned previously, these forcing functions
included: atmospheric vapor concentrations
atmospheric (wet and dry) deposition loadings, and
tributary loadings. The new total PCBs forcing
functions were based upon the LMMBP estimates
presented previously in conjunction with other
information regarding long-term trends in PCBs
usage, loadings, and concentrations in Lake
Michigan and the Great Lakes. This latter
information was used to develop the continuous total
PCBs forcing functions necessary to run the
MICHTOX simulation from an uncontaminated initial
condition. It is important to consider trends in long-
term loads and other forcing functions. These relate
directly and indirectly to chemical transport and fate
over comparable time scales. Other simulations
were conducted with MICHTOX to predict the
effectiveness of several toxic chemical management
scenarios.
The forcing functions developed for PCBs by the
LMMBP are believed to be accurate estimates for the
1994-1995 period, based upon the data quality
objectives and well-developed estimation
procedures. A number of sources of information
were available to characterize trends of the usage
and release of PCBs in the Great Lakes during the
20th Century, and these were used to extrapolate the
LMMBP forcing functions both backwards and
forwards in time. Although somewhat speculative a
similar procedure had been demonstrated for PCBs
in Lake Ontario (Mackay, 1989; Gobas et al., 1995).
The procedure requires the following information:
1. The date when contamination begins.
2. The rate of increase in the magnitude of the
forcing function.
3. The date and duration of the loading/forcing
function peak.
4. The rate of decline in the magnitude of the
forcing function.
Rates of change in vapor phase PCBs
concentrations for Lake Michigan and the Great
Lakes region have been published by a number of
researchers (Hillery et al., 1997,1998; Baker and
Eisenreich, 1990; Green et al., 2000; Schneider et
al., 2001). Although there is some disagreement as
to whether atmospheric measurements support the
notion that vapor PCBs concentrations are declining
over Lake Michigan, Schneider et al. (2001) indicated
that PCBs concentration profiles in highly-resolved
sediment cores from Grand Traverse Bay support the
view that vapor phase PCBs concentrations have
been declining at a rate of about 0.115/year, which
corresponds to a six-year half-life over the past 25
years.
Similarly, rates of change in PCBs tributary loadings
can be determined from loading estimates based
upon measurements from the Fox River (in 1989-
1990 by Velleux and Endicott, 1994 and in 1994-
1995 by the LMMBP) and major tributaries
throughout the Lake Michigan basin (in 1982 by Marti
and Armstrong, 1990 and in 1994-1995 by the
LMMBP). This information yields estimates for the
rate of decline in tributary loadings of 0.053 to
0.054/year, corresponding to a 12- to 13-year half-
life.
To complete the long-term total PCBs forcing
functions, a number of other assumptions were
made:
1. PCBs contamination
commenced in 1940.
of
Michigan
Lake
2. The rate of increase in vapor concentrations and
tributary loadings was the same as the rate of
decline.
3. Atmospheric deposition loadings followed the
same long-term trends as vapor concentrations.
4. Monthly variability in the magnitude of forcing
functions followed the 24-month pattern
established by the LMMBP estimates.
The date and duration of the peak in the PCBs
forcing functions was not so easily defined.
Schneider et al. (2001) suggested that forcing
functions peaked in 1970, and declined with the
decline in chemical production after 1972. On the
97
-------
other hand, Gobas et al. (1995) estimated that PCBs
loading to Lake Ontario peaked much earlier, in
1961. Why there would be such a difference
between Lakes Michigan and Ontario is not clear,
and perhaps it reflects the subjectivity of these
estimates. Ultimately, three different estimates for
long-term total PCBs forcing functions were
developed:
1. Scenario A - Total PCBs forcing functions peak
in 1970 and decline after 1972.
2. Scenario B - Total PCBs forcing functions peak
in 1961 and decline after 1963.
3. Scenario C - Total PCBs forcing functions peak
in 1961 and decline after 1972.
Plots of the three forcing function scenarios are
provided on a whole-lake basis for vapor
concentrations (Figure 2.6), atmospheric deposition
(Figure 2.7), and tributary loading (Figure 2.8). In
each plot, the forcing function estimate used in the
original MICHTOX application are also presented.
Forecast simulations using LMMBP data to define
initial conditions for PCBs and alternative forcing
functions (representing general toxic chemical
management alternatives) using those measured by
the LMMBP as a baseline included:
1. No-Chanae - Total PCBs vapor concentrations,
deposition fluxes, and tributary loadings continue
in the future at the levels estimated by the
LMMBP for 1994-1995.
2. No-Action - Total PCBs deposition fluxes and
tributary loadings continue in the future at levels
estimated by the LMMBP; however, atmospheric
vapor concentrations decline in the future at a
rate equal to the observed rate of decline over
the past 25 years. This assumes that PCBs
vapor concentrations are declining due to the
slow depletion of regional-scale inventories (e.g.,
transformers, contaminated soil, landfills, sewage
sludge, etc.) from which PCBs evaporate and act
as sources to the atmosphere. It is not clear to
what extent these declines may be related to
toxic chemical management.
3. Fifty-Percent Function Reduction - Total PCBs
vapor concentrations, deposition fluxes, and
tributary loadings are reduced from LMMBP
1994-1995 levels by 50% at the start of 2002.
4. Fifty-Percent Reduction - Deposition fluxes and
tributary loadings are reduced from LMMBP
1994-1995 levels by 50% at the start of 2002.
Total PCBs vapor concentrations continue in the
future at the levels estimated by the LMMBP for
1994-1995. I n other words, this simulation
assumes that vapor concentrations are not
controllable.
5. Elimination - Total PCBs vapor concentrations,
deposition fluxes, and tributary loadings are
eliminated at the start of 2002.
Twenty-year forecasts, simulating the period 1994
through 2014, were made with MICHTOX for each of
these toxics management alternatives.
2.5 Results and Discussion
2.5.1 Confirmation of MICHTOX PCBs
Bioaccumulation Predictions
Initially, predictions of total PCBs bioaccumulation
were made by running the MICHTOX food web
model to steady-state, using the average dissolved
and particulate total PCBs concentrations observed
during the LMMBP in southern Lake Michigan. This
was done to separate the bioaccumulation
predictions from the transport/fate predictions of the
model, which were used in other model runs to define
chemical exposure. The bioaccumulation predictions
were then compared to average total PCBs
concentrations in fish from the Saugatuck and
Sheboygan Reef biota zones. These comparisons
are plotted in Figures 2.9 and 2.10. MICHTOX
consistently underpredicts PCBs bioaccumulation in
all fish from Saugatuck, except bloater (Figure 2.9).
Model predictions are much more consistent with
data for PCBs concentrations in fish from Sheboygan
Reef (Figure 2.10), although the model again tends
to underpredict PCBs bioaccumulation for older age
classes of lake trout. Differences in PCBs
concentrations measured in fish at these two
locations most likely reflect different local PCBs
exposures; sediment PCBs concentrations, for
98
-------
""
cs.
e
CiJ
N
~
r-
o
::J
cc
,.!..
CD
...
3
CD
UJ
:!:
3
D)
...
CD
UJ
o
-
r-
D)
~
3:
n'
::J"
cs'
<0 D)
::J
<0 -
o
-
D)
."
o
OJ
UJ
<
D)
"C
o
...
n
o
::J
n
CD
::J
...
...
D)
-
c)"
::J
~
16
14
12
M
.E 10
CI
.s
c:
:8
~
-
c:
GI
U
c:
o
u
..
o
Co
C'CI
>
8
6
4
2
o
1940
r-.
. \
I .
. \
r .
. \
I .
. \
I .
. \
I .
. \
I .
. \
I .
. \
I .
. \
I \
i \
.i \.
I
i
i
i
i
./.
./
.~
- original MICHTOX time function
- - - - - - scenario A estimate
-.-scenario B estimate
- scenario C estimate
1950
1960
1970
1980
2010
1990
2000
year
-------
"'T1
CS'
c
...
CD
N
~
r-
0
:J
CO
I
- 10000
CD
...
3
CD 9000
UI
~
3
m
- 8000
CD
UI
So >;
r- CfI 7000
m ~
~
CD OJ
C
s: ~ 6000
n" .2
::T
cO' c
0
~ m :e 5000
o :J II)
0 - 0
Co
0 CD
- '0
!!!. .~ 4000
." ...
CD
0 .s::.
m Co
I/)
UI ~ 3000
m -
- I'll
3
0 2000
UI
"
::T
CD
... 1000
c:r
Q.
CD
"
0 0
UI 1940
;::;
S'
:J
0
m
Q.
:i
cp
ro'\
I .
. \
I .
. \
I .
. \
I .
. \
I .
. \
I .
. \
I .
. \
I .
,
i
,
I
I
j
.
- original MICHTOX time function
on. -- scenario A estimate
-.- scenario B estimate
- scenario C estimate
..., --.. ....., ............".
1950
1960
1970
1980
1990
2000
2010
year
-------
"'"
ca"
c
;
N
90
r-
0
::J
(Q 2000
I
-
C1I
...
3
C1I 1800
UI
:::!:
3
I» 1600
-
C1I
UI
0
- 1400
r- -
I» ~
-
~ C)
~
s: - 1200
cr C)
~ c
cS" "C 1000
C'IS
...... 1» ..2
0 ::J
...... - ~
0 800
-
I» .s
~ :J
0 .c
".: 600
m -
UI
-
:!"
C" 400
c
-
I»
...
'< 200
0"
I»
Co
:J 0
(Q
1940
r--'\
I "
" \
I "
" \
I "
I
I
I
I
I
..
--.. --.-..... -.. .--..........
1950
1960
1970
1980
year
- original MICHTOX time function
-- -- -- scenario A estimate
-. - scenario B estimate
- scenario C estimate
1990
2000
2010
-------
~
o
I\)
- ."
CD -.
...cg
a c:
... ...
o-CD
I»~
"'co
rn.
Q.o
CDO
S 3
--0
CD I»
- ...
:;. in.
!!to
I» ~
~o
Q.-
-=:
~-
:;.0
Q.%
,a-f
cO
1»)(
::I.rn
=Ci
CD I»
.!.Q.
eo<
.
rn
-
I»
-
CD
-
o
-
!!.
"tJ
o
m
rn
n
o
~
n
CD
~
-
AI
-
o.
~
(I)
-
o
en
I»
c
cg
a
c:
n
~
-
in.
~
Q.
;
10000
9000
8000
7000
c;
-
C)
.E. 6000
c:
o
:;:::
f! 5000
-
c:
CD
u
~ 4000
u
CD
o
2= 3000
2000
1000
o
small large small large lake
alewife alewife bloater bloater trout
1
lake
trout
2
lake
trout
3
lake
trout
4
lake
trout
5
o average PCB
. model
lake
trout
6
lake
trout
7
lake
trout
8
lake
trout
9
lake
trout
10
lake
trout
11
lake
trout
12
-------
a. "TI
I» -.
_co
I» C
Ci'i
"I\)
...
0-"
"0
0".
1»(')
.. 0
Ut 3 7000
0."
C1» I»
:2 ..
0 -.
-Ut
C1» 0 0 average PCB
~:2 6000
.. 0 . model
Ut-
-=:
1»-
::J(')
o.:t 5000
;-.
-.0
0.>< -.
0>
,gUt -
c- 0>
I» m c:
'-' 4000
::10. c:
='< 0
C1» I
Ut!!1. +::
co
"':-"1» ....
...... - .....
0 C1» c: 3000
- Q)
c.> 0 (.)
- c:
e!. 0
~ (.)
(') CO
aJ U 2000
Ut a..
n .....
0
:2
n
C1» 1000
:2
-
..
I»
-
o.
:2
Ut
-
0
en
~
CD
0"
0
'<
CD
I»
:2
:JJ
C1»
C1»
-
~
Ut
=r
o
small large small large
alewife alewife bloater bloater
lake
tro ut
1
lake
trout
2
lake
trout
3
lake
trout
9
lake
trout
4
lake
trout
5
lake
trout
6
lake
trout
7
lake
trout
8
-------
example, are substantially higher at Saugatuck than
at Sheboygan Reef. Because both biota zones fall
within the same southern Lake Michigan model
segment, differences in PCBs exposure
concentrations are not reflected in the
bioaccumulation model results. Differences in PCBs
bioaccumulation may also result from regional
variation in trophic dynamics or bioenergetics;
LMMBP data characterizing these factors were not
examined for this project. The sensitivity of
bioaccumulation model predictions are specifically
addressed later in Section 2.3.10 of this report.
2.5.2 Comparison of Original MICHTOX
PCBs Simulations to the LMMBP Data
The next step taken in the evaluation of the
MICHTOX model was to compare PCBs forcing
functions (Le., atmospheric vapor concentrations,
atm~spheric deposition fluxes, and tributary
loadings) used in the original model to estimates
generated by the LMMBP. The original and the
LMMBP forcing functions for total PCBs are
compared on a whole-lake basis for years 1994-1995
in Table 2.5. The LMMBP estimates for atmospheric
deposition and tributary loading are reasonably close
to. ~he original MICHTOX forcing functions; the
onglnal forcing functions are both about 30% lower
than the LMMBP estimates. However the
discrepancy is much greater for ~apor
concentrations; the average total PCBs vapor
concentration estimated by the LMMBP is 4.5 times
higher than the original MICHTOX value. This
discrepancy is much greater than expected, and
results from the lack of adequate data available to
define lake-wide average PCBs vapor concentrations
in the original MICHTOX model. Given the model's
sensitivity to vapor concentrations, it was determined
that all PCBs forcing functions should be updated
based upon the LMMBP estimates, and that these
forcing functions should be used to rerun all
MICHTOX simulations.
Prior to rerunning the model with the revised forcing
functions, the original MICHTOX model predictions of
PCBs concentrations in water, sediment, and fish
were compared to concentrations measured in the
LMMBP. These comparisons are presented in
Figures 2.11-2.14. MICHTOX predictions of total
PCBs concentrations in main Lake Michigan
segments are compared to available water column
measurements (including those from the LMMBP) in
Figure 2.11. Predictions made with the original
model (although referred to as the "original
MICHTOX model" in this section, the simulations are
based on a version of the model which incorporated
the continuity balance and Straits of Mackinac
boundary condition modifications discussed
previously) appear to be quite consistent with the
data over the past 25 years. Predictions of surficial
sediment total PCBs concentrations made using the
original MICHTOX model are compared to data from
three Lake Michigan sediment core profiles in Figure
2.12. Data from these sediment cores, collected by
the USEPA Great Lakes National Program Office
(GLNPO) in 1991-1992, were used because no
sediment cores were analyzed for PCBs in the
LMMBP. The agreement between predictions and
data is not very satisfactory. In part this is because
the transformation of sediment core data from depth
intervals to dates was done based upon
sedimentation rates, without consideration of the
effects of sediment mixing in the surface layers.
Table 2.5. ~omparison of Origin~1 MICHTOX Total PCBs Forcing Functions for 1994-1995 to the
LMMBP Estimates (Whole-Lake Average)
Forcing Function
Atmospheric Vapor Concentration (ng/m3)
Atmospheric Deposition (kg/y)
Tributary Loading (kg/y)
Original MICHTOX
0.092
151
261
LMMBP Estimate
0.41
232
346
104
-------
"TI
cO'
c
CiJ
~
.....
.....
.
0
...
ca'
::J 5.0
!..
3:
0 4.5
:I:
d ,/ "
/ \
>< 4.0
"C \
... \ - Segment 1
(1)
Q.
n' - 3.5 ...... Segment 2
- -.J
0' - - Segment 3
C>
::J C
tn - 3.0 - - - Segment 8
I» C - Segment 9
::J 0
Q. ~ - - Segment 10
...... Q. co 2.5
'-
0 1» "" water data
01 - c: .
I» Q)
- 0 2.0
0 c:
... 0
3 0
I»
::J 1.5
Dr
,...
(1) 1.0
-
0
-
I»
." 0.5
0
OJ
en 0.0
n
0 1940 1950 1960 1970 1980 1990 2000 2010
::J
n
(1) year
::J
-
...
I»
-
o'
::J
tn
-------
~
o
m
-."
fit -.
CDca
a.c:
-. ~
3 CD
CDN
::J'
..~
n~
~o
CD ~.
fltca
n -.
o ::J
-I»
Ci'3:
n -
c;to
a.:J:
-. -t
~o
~><
CD"
CD~
~CD
~a.
CD n'
CD::!;
~o
. ::J
fit
I»
::J
a.
a.
I»
..
I»
-
o
~
350
3
!.
::J
ii
'1l:
CD
fit
CD
a.
3
CD
::J
-
..
o
..
!!.
."
o
OJ
fit
n
o
::J
n
CD
::J
-
~
I»
..
o'
~
fit
-Segment 11
300 - - - Segment 12
-Segment 13
. site 47s
250 . t;. site 18
0 site 68k
c;
--
C) 200 0
c
.......
c
0
..
IU
...
- 150
c
B 0 t;.
c t;.
0
0 .
100
0 t;.
0 .
50
t;.
t;.
t;. 0
0
1900 1920 1940 1960 1980 2000 2020
year
-------
en "II
:::r -.
(Dca
O"c
o i
'<
"
...
(I)
Q,
n'
-
O'
:J
UI
o
-
-
..... 0
-
o I»
.......
"Q
o
aJ
UI
(')
o
:J
(')
(I)
:J
-
...
I»
-
0'
:J
UI
:J
-
0'
J
I»
:J
Q,
(')
o
3
"
I»
...
0'
o
:J
-
o
9000
8000
7000
6000
en
C> . LMMB data
.s
c: 5000 0 model
o
~
...
"E 4000
Q)
u
c:
0
u
CD 3000
U
2::
2000
1000
o
small large small large
alewife alewife bloater bloater
lake
trout
3
lake
trout
4
lake
trout
5
lake
trout
6
lake
trout
9
lake
trout
12
lake
trout
11
lake
trout
7
lake
trout
8
-------
D
D 0
C CO
0)
D ...
D
"-
!II
Q)
>-
o
,....
0)
...
c
c: c: c: 0
o 0 0 :;:;
:u to :u :u .~
.- ... .- .- 't:J
't:J to 't:J 't:J G)
~~~~o.
c.:; 0. 0.0
...... to CO 0) .....
G)~G)G)G)
g>Og>g>g>
: D ~ I I
,t' /
:/
// D
/1
: / D
..,/ / D
...' I
.,' I
..., /
... /
/
/
D
D
D
...- ./
...- /"
..- /
/
............:///
... D I
: I
...~..... \,
....
"" .....
'. "-
....... .......
". "-
"" "-
".
'.
c
D
o
(")
10
N
o
N
10
...
o
.....
1.0
(6/6n) UO!t8JtUa:)uro 8~d.
o
...
o
N
o
o
o
N
o
0)
0)
....
o
CO
0)
....
o
1.0
0)
...
o
~
0)
0...
Figure 2.14. Original MICHTOX predictions of total PCBs concentrations in lake trout and comparison
to DeVault et al. (1986) data.
108
-------
MICHTOX predictions of total PCBs concentrations
in fish are compared to concentration data for fish
collected at Sheboygan Reef in Figure 2.13. The
observed trend of increasing total PCBs
concentrations with fish size and age is captured by
the model predictions, although the original model
consistently tends to overpredict bioaccumulation in
fish at Sheboygan Reef. This tendency for the
original model to overpredict bioaccumulation is also
evident in Figure 2.14 which compares predictions of
total PCBs concentrations in lake trout to a long-term
dataset provided by DeVault et al. (1986 and
personal communication). Predictions for several
lake trout age classes are plotted, corresponding to
the range of average fish age inferred from weight
data for the fish composited in these samples. The
original model overpredicts total PCBs concentrations
in lake trout for all years except for the early 1970s
when the highest PCBs concentrations were
observed.
The results of the original MICHTOX model
simulations were also interrogated in terms of total
PCBs mass transport fluxes and inventories for the
1994-1995 period. These are summarized in Table
2.6, with results for Green Bay separated from the
main lake. According to the original model, both air-
water and water-sediment fluxes dominate the
transport of PCBs in Lake Michigan. The negative
PCBs flux associated with the Straits of Mackinac
export arises from the boundary condition treatment
in which higher PCBs concentrations are associated
with the reverse-flow component from deeper layers
of the water column at this boundary.
2.5.3 Updating Parameterization for
Henry's Constants
Henry's constants for PCBs congeners estimated
using the Bamford et al. (2000) model are
significantly higher than those used in the initial
MICHTOX parameterization. This is illustrated in
Table 2.7 where Henry's constants for the modeled
homologs are compared on a monthly basis. These
values were calculated from the average Henry's
constants for all congeners in each homolog using
Table 2.6. Mass Balance Diagnostics for Total PCBs in the Original MICHTOX Simulation (Year 1994-
1995)
Flux (kg/d)
Mass Transport Pathway
Green Bay Export
Straits 01 Mackinac Export
Chicago River Export
Tributary Loading
Atmospheric Deposition
Net Volatilization
Volatilization (Gross)
Gas Absorption
Settling
Resuspension
Burial
Net (Mass In - Mass Out)
Main Lake
Green Bay
91
-4
1
128
137
1256
2170
914
2141
2632
815
-1712
-91
133
14
391
420
29
2574
2802
54
-390
Inventory (kg)
Total PCBs Inventory
Water Column
Surficial Sediment
Main Lake
Green Bay
1630
16,500
104
7530
109
-------
Table 2.7. Comparison of Original and Revised Henry's Constant (atm m~"mol) Parameterization
Month Revised PCB4 Original PCB4 Revised PCBS Original PCBS
January 1.3E-04 2.5E-05 1.6E-04 2.1 E-05
February 1.2E-04 2.5E-05 1.6E-04 2.1 E-05
March 1.2E-04 2.5E-05 1.5E-04 2.1 E-05
April 1.2E-04 2.5E-05 1.6E-04 2.1 E-05
May 1.3E-04 3.7E-05 1.7E-04 3.2E-05
June 1.7E-04 6.2E-05 2.1 E-04 5.3E-05
July 2.2E-04 9.9E-05 2.8E-04 8.8E-05
August 2.6E-04 1.4E-04 3.2E-04 1.2E-04
September 2.4E-04 1.2E-04 3.0E-04 1.1E-04
October 1.9E-04 7.4E-05 2.4E-04 6.4E-05
November 1.6E-04 6.2E-05 2.0E-04 5.3E-05
December 1.4E-04 3.7E-05 1.8E-04 3.2E-05
Annual Average 1.7E-04 6.1 E-05 2.1 E-4 5.2E-05
the mean monthly temperature. The Bamford
Henry's constants are two to seven times higher than
the original values which were based on data from
Burkhard (1984). Interestingly, the model based
upon Bamford's measurements predicted that the
higher molecular weight congeners had generally
higher Henry's constants. This was a consequence
of the number of ortho-chlorine substitutions, which
the model correlated to volatility, increasing with
molecular weight (Le., homolog number). Previously,
higher chlorinated PCBs had generally been
assumed to be less volatile (Brunner et al., 1990;
Dunnivant et al., 1992).
2.5.4 Updating the Chemical Volatilization
Rate Formulations
The total PCBs volatilization rates calculated using
the Wanninkhoff (1992) and Schwarzenbach et al.
(1993) formulations were substantially higher than
the rates computed in the original MICHTOX model.
Monthly rates of PCBs homolog volatilization,
calculated using the original and revised MICHTOX
formulations, are compared in Table 2.8. The revised
volatilization rates are two to three times higher than
those computed in the original model. The revised
volatilization mass transfer rate calculations were
confirmed by reproducing the congener-specific rates
presented in Totten et al. (2001). The original model
computed volatilization rates using the formulations
of O'Connor (1983) and Liss (1973).
In terms of PCBs transport and fate, the implications
of higher Henry's constants and volatilization rates is
that PCBs equilibrium will be shifted significantly
towards the atmospheric vapor phase, and this shift
will occur more rapidly than previously predicted.
2.5.5 Long- Term
Simulations
Hindcast/Forecast
The first simulations conducted with the revised
MICHTOX model were long-term simulations from a
zero (Le., "clean") initial condition in 1940. These
were conducted for each of the long-term forcing
function scenarios: A, B, and C. The results of these
simulations were compared to LMMBP and long-term
PCBs concentration data, with the goals of
confirming model predictions and determining which
long-term loading scenario best simulated the data.
Predictions for Scenario A, in which the PCBs forcing
functions were assumed to peak in 1970-1972, are
plotted and compared to data in Figures 2.15 to 2.19.
Predictions in the main Lake Michigan segments are
compared to available water column measurements
(including those from the LMMBP) in Figure 2.15.
Clearly, the model predictions of total PCBs
concentration in the water column are low for this
scenario. For reference, the Scenario A predictions
in the southern Lake Michigan segments (Segments
1 and 8) are plotted together with the original model
long-term predictions in Figure 2.16. Predicted
surficial sediment total PCBs concentrations are also
very low in comparison to the data as shown in
110
-------
-
Table 2.8. Monthly PCBs Volatilization Rates (mId) Calculated by Original and Revised MICHTOX
Formulations
Month Original PCB4 Rate Revised PCB4 Rate Original PCBS Rate Revised PCBS Rate
January 0.44 1.72 0.37 1.83
February 0.42 1.63 0.36 1.74
March 0.41 1.59 0.35 1.71
April 0.39 1.52 0.33 1.63
May 0.45 1.26 0.39 1.34
June 0.37 0.75 0.33 0.77
July 0.37 0.55 0.34 0.56
August 0.61 0.74 0.55 0.74
September 1.05 1.27 0.93 1.29
October 0.98 1.72 0.85 1.78
November 0.98 0.95 0.85 2.02
December 0.60 1.72 0.52 1.81
Annual Average 0.59 1.29 0.51 1.43
20
1.8
1.6
1.4
::; 1.2
0,
.s
c: 1.0
.12
~
i:
11 0.8
c:
8
0.6
0.4
0.2
0.0
1940 1950
o
- Segment 1
- - - Segment 2
- Segment 3
- Segment 8
- - Segment 9
- Segment 10
o water data
1960
1970
1980
1990
2000
2010
year
Figure 2.15. Long-term Scenario A predictions 01 main lake total PCBs concentrations.
111
-------
cn"TI
n -,
(1)> co
:J C
G)CiJ
~
-, N
0,
~~
- '
o
03:
~-
-0
ce.:I:
:J-4
!.o
3><
00
0.0
(1)> C
--
~:T
~ (1)>
(1)> ~
Q,:J
-, r-
~G)
-, ~
0(1)>
:J
!"3:
C:;'
:T
cO'
G)
:J
.....
.....
I\)
-
o
-
!.
"U
o
m
o
~
~
(1)>
0.
2
0'
:J
!"
o
o
3
~
I»
...
ii'
o
:J
o
-
is'
:J
co
cr
i
:::J
--
C)
.s
c:
o
:p
('II
....
-
c:
Q)
(J
c:
o
(J
cc
()
2:
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
,..' ,/
......;.",
....,,:.:..~
..."s...-
0.0
1940
....- /
..." /'
,," /'
.,' .I'
1950
.
.
.
.
.
. /
. /
"
'.
,
.
.
.
.
.
- - segment 1 - original model
...... segment 8 - original model
- segment 1 - scenario A
- segment 8 - scenario A
:
.
.
.
.
.
,/-
/
I
"
,
.
.
.
/
/
/
\
,
\
\
.
.
.
.
.
/
.
.
.
.
.
.
.
",
"
...................
1960
1970
1980
2010
1990
2000
year
-------
......
......
U)
n"TI
g cs.
n C
CD CD..
~
::;~
I»~
:::!:......
O'
~O
-0
fA 3
&'0
-.1»
3 :!.
CD fA
~ 0
-~
n 0
~-
CD 0-
fA
n ~
oCQ
=,!..
CD CD
n ..
(i3
Q.cn
-. n
~ CD
~~
CD I»
CD ::::!.
7'0
co»
!R'C
.v ..
~CD
Q.
(i"
-
0"
~
fA
-
o
3
!.
~
Di'
~
fA
CD
Q.
3
CD
~
-
-
o
-
!.
""0
o
m
fA
300
250
200
-
C>
....
C>
c::
-
c::
o
:;::;
ro
....
-
c::
CD
u
c:
o
u
150
100
50
.
o
1900
o
1920
.
o
o
.
1940
o
.
o
o
. .
.
o
o
.
o 0 0
.
o
.
- Segment 11
- Segment 12
- -- --. Segment 13
. site 18
. site 47s
o site 68k
.
.
.
.
.
.
.
,1.\
.
1960
1980
2000
2020
year
-------
. .
.
o
....
o
N
o
o
o
N
o
0)
0)
....
o
CO
0)
....
0 ...
D m
o 0)
o >-
o
0 0
""
0)
....
Ih
Ih II) II) C
C ceo
o 00..
.- .-.- (J
:2S:2:2:c
't) III 'C 't) 0)
~~~~c.
o.'S 0.0.0
""111(00)....
0) >0) 0) 0) 0)
C) '" C) C) C)
III ..... III m III
I 0 : I
It)
N
o
N
It)
....
o
~
(6/6n) Uo!teJtu~uo::l 8::>dt
o
c.o
0)
....
o
It)
0)
....
It)
o
'It
0)
0""
Figure 2.18. Comparison of long-term Scenario A predictions to DeVaultet al. (1986) lake trout data.
114
-------
"'71
c.
c
...
CD
N
:...
!D
0
0
3
" 7000
I»
...
iir
0
:J
0
- 6000
~
0
:J:
-I
0 5000
><
en
n
CD 0)
:J en
I» .s 4000
...
C)" c::
0
....... » ~
....... -
01 0 C 0 average PCB
- Q)
!!. 0 3000 . model
c::
"'U 0
0 0
m
OJ U
tn ~
n 2000
0
:J
n
CD
:J
-
...
I» 1000
-
o.
:J
tn
-
0
en
J
CD
a
0
'<
CC
I»
:J
:D
CD
CD
-
C.
I»
-
~
o
small large small large
alewife alewife bloater bloater
lake
trout
3
lake
trout
4
lake
trout
5
lake
trout
6
lake
trout
7
lake
trout
8
lake
trout
9
lake
trout
11
lake
trout
12
-------
Figure 2.17. The same tendency of Scenario A to
underpredict total PCBs concentrations occurs with
the fish data, both DeVault's long-term data (Figure
2.18) and the size- and age-class specific data from
the LMMBP (Figure 2.19).
Predictions for Scenario B, in which the PCBs forcing
functions were assumed to peak in 1961-1963, are
plotted and compared to data in Figures 2.20-2.29.
Predictions in the main Lake Michigan segments are
compared to available water column measurements
(including those from the LMMBP) in Figure 2.20.
Although still somewhat low, Scenario B predictions
are definitely an improvement over Scenario A.
Scenario B water column predictions are also
compared to the original MICHTOX model
predictions in Figure 2.21. Figure 2.22 compares the
deepwater (hypolimnetic) predicted dissolved total
PCBs concentrations to segment-average data for
the eight LMMBP cruises. The model accurately
predicts both the trend of decreasing dissolved total
PCBs concentrations moving from south to north in
the lake, as well as the build-up of deepwater
dissolved total PCBs concentrations during the
stratified period of each year. This agreement is
encouraging because dissolved-phase deepwater
concentrations were used as PCBs exposure
concentrations for the bioaccumulation model.
Predicted surficial sediment total PCBs
concentrations for Scenario B are compared to
sediment core data in Figure 2.23. The comparison
of model predictions to average surficial sediment
concentrations based on LMMBP sediment core
samples (Figure 2.24) is somewhat more informative.
Surficial sediment total PCBs concentrations in
southern and central Lake Michigan and mid-Green
Bay are about 60% low in comparison to the data,
while predicted sediment concentrations are much
closer to average concentrations measured in other
sediment segments.
Scenario B predictions of total PCBs concentrations
in fish are compared to DeVault's (DeVault, 1986;
DeVault, personal communication) long-term data
(Figure 2.25) and the size- and age-class specific
data from the LMMBP for Sheboygan Reef (Figure
2.26). In general, this simulation compares well with
both the long-term and LMMBP data, although
aspects of both simulations deserve comment. The
Scenario B lake trout predictions agree well with
long-term data after the mid-1970s; however, they
locate the peak total PCBs concentrations about five
years earlier than observed by DeVault. The lake
trout total PCBs predictions are also 20-30% lower
than the data for most of the Sheboygan Reef age
class data. In general, however, the Scenario B
simulations offer a better prediction of the total PCBs
data than the original MICHTOX model simulation.
The results of the Scenario B MICHTOX model
simulations were also interrogated in terms of total
PCBs mass transport fluxes and inventories for the
1994-1995 period. These are presented in Table 2.9.
Comparison of these results to the mass balance
diagnostics from the original model (Table 2.6)
demonstrate how the transport and fate predictions
for PCBs have changed as a result of updating the
Henry's constant parameterization, revising the
volatilization rate formulations, and making use of the
LMMBP estimates to develop long-term PCBs forcing
functions. Most notably, the total PCBs inventories
are depleted by 40-70% in the Scenario B simulation
(in comparison to the original long-term model
simulation) due to the increased chemical volatility.
Enhanced volatility of total PCBs completely offsets
the enhanced absorption resulting from higher vapor
concentrations. Scenario B mass balance diagnostics
also demonstrate that air-water fluxes now clearly
predominate the transport pathways for PCBs in
Lake Michigan.
Predictions for Scenario C, in which the PCBs forcing
functions were assumed to peak over a longer
duration, from 1961-1972, are plotted and compared
to data in Figures 2.27-2.31. Although Scenario C
forcing functions are substantially different, the
predictions are qualitatively similar to those for
Scenario B, at least for the period during which data
are available.
In general, the Scenario B long-term simulations tend
to agree most favorably with the available PCBs
data. The model predictions for this scenario are
probably at least as accurate as the forcing functions
themselves; this was judged to be an adequate level
of model confirmation for this assessment. Further
refinement of forcing functions and model parameters
could improve the agreement between data and
~redictions; however this was not possible given the
time and resource constraints of this project.
116
-------
'TI
CS'
c
CiJ
N
i\)
!::t
I
0
::s
(Q
..!. 4.0
CD
~ r"
3
en i \ - Segment 1
() 3.5
CD 1'-, \ .--u. Segment 2
::s
S» . I .
~ J' \ \ -- Segment 3
o' , . \.
OJ 3.0 - - - Segment 8
'C - Segment 9
~
CD -.. Segment 10
a.
n :::J 2.5 c water data
o' --
C>
::s c:
tn -
c:
~ 0 0 2.0
- +-
~ 3 III
........ ... c
-
S» c:
S' Q)
u
Dr c: 1.5
0
~ u []
- []
0
- 1.0
!!!.
"'tJ
0
OJ
tn 0.5
()
0 . " .. C "CD
::s
() ""0_.
CD
::s 0.0
-
~ 1940 1950 1960 1970 1980 1990 2000 2010
S»
-
S' year
::s
~
-------
en","
n -.
('Dca
~ c
wCi1
~
-. N
o.
m~
-.
0
oS:::
~-
-.0 4.0
C!:!.:c
~~
!.O
3>< ......
0 tn 3.5
0.0 \
('D c
-- \
'OJ \
~ ('D
('D ~ 3.0
o.~
-.....
2,w - - - segment 1 - original model
-. ~ -
0 ('D ....I - - segment 8 -original model
~s::: -- 2.5
en
c -' segment 1 -scenario B
. -. -
n
J c
0 - segment 8 - scenario B
ca. ~
w 2.0
~ -
...... c
...... - Q)
CD g, 0
c
!. 0
0 1.5
"tJ m '
0 U '
m D..
tn -
'0 1.0
a; ,
0. ,
n.
-
o. 0.5
~
!"
0
0 0.0
3
'0 1940 1950 1960 1970 1980 1990 2000 2010
w
... year
i"
0
~
0
-
0'
::s
ca
S
~
3
-------
-"",
0 -.
-CO
!!.c
"tJCi1
01\)
m.
fnl\)
,,~
0 0
::I
" 0
CD 3 0.3
::1'0
::1»
1»...
- -. X
-'fn
0 0
::I ::I - Segment B
!'» 0 0.25 Segment 9
- - - -
0' - Segment 10
::I x Segment B data 0
CO
. - Segment 9 data A
- 0
CD ...J
-- 0.2
... en .A. Segment 10 data
3 .s
en c:
" .Q X
CD 16
::I ...
-
I» c: .A.
... (J) 0.15
o' (,)
...... c: .A. .A. A
o
...... OJ (,)
O' (5
~ II)
fn II)
:0
- A
0
- 0.05
J
CD
I'
3:
3:
m
"tJ
Q.
CD
CD
'a
~
I»
-
CD
...
Q.
iir
fn
0
<'
CD
Q.
o
o
100
200
300 400 500
day (0 = January 1, 1994)
600
700
800
-------
- 'TI
tn -.
CDca
o.c
-. ~
3 CD
CDN
:J'
_N
n~
~o
CD 0
tn 3 350
n"C
2.11
-~
CD -.
n tn \
-0 I \
CD:J I
0.0 300
-. - , \
:Jcn I I
-4n I q
CO CD I
co:J
-411 I
I ~ - 250
-4 -. 0 0
coo ~
:Sm OJ
c:
":""'''C - _Segment 11
CiJ c:
0 200
0. :; - Segment 12
~ (U
...
- - - - Segment 13
o. c:
...... :J ~ Site 47s
N tn c 0 0
o - 8 150 Site 1 B
o 0 D A
3 m l:J. l:J.
U 0 Site 6Bk
II 2:
5.
Di' 100
i;" 0 A
tn
CD
e: 50 0 D
3
CD
:J A
- A
..
~ A 0
0
"'0 1900 1920 1940 1960 1980 2000 2020
o
m year
tn
n
0
:J
n
CD
:J
..
~
a
6"
~
..
-------
n"TI
0 -.
;:,cc
n C
CD ...
;:, CD
-",
DlN
:::!:~
o.
;:,
00 1000
-0
13
3:"tJ
3:m
m~.
"'tJ0
I» 0
;:,:s
c.o
C')-
m-
s:g
mea
"'tJ.!. ,...,
C) 100
CD .....
C"... C)
o 3 :]
>< -
(')0 c
0
0 (') ~
... CD II!
CD :s ....
0 m -
c
...... m ~. Q)
I\) 3 0 0
c
...... '2,m 8
CDog co
.!!!.... u
. CD a..
c. - 10
n'
-
o'
:s
0
-
0
I»
<
CD
...
I»
ea
CD
-
0
-
I»
"'tJ
0
m
0
0
CD
c.
3
CD
:s
-
. model prediction
o box core averages
1
Segment 11
Segment 12
Segment 13
Segment 15
Segment 16
Segment 17
-------
"
cS'
c
CiJ
~
N
~
n
0
3 25
'0
I»
...
iii' C
o C
~
0
-
0' 20
~
CO C
..!.
CD
... - age 7 prediction
3
en c; C DeVault data
n - 15
CD C) ...--- age 8 prediction
~ .2.
I» c - age 9 prediction
... ~
0' ~ -- age 10 prediction
...... m -
I\) c
I\) '0 8
... c
CD 0 10
Q. 0 ,
n' co
- U '
0' 2: '
~ '
(I) ,
- '
0 '
C 5 '
CD
<
I»
C
::;
CD
-
Q»
:--
-
...a.
CD
CD
en
-
Dr
71:'
CD
-
...
0
C
-
a.
J
o
1940
J
......
~ ~....-::;;.....
1950
1960
1970
1980
1990
2000
2010
year
-------
a."TI
I» -.
_co
!» c
CiJ
N
N
0)
0
0
3
'0 7000
I»
~
iir
0
~
0 6000
-
0'
~
CO
I 0 average PCB
-
CD 5000
~ . model
3
en -
.!;!!
(") C>
CD .s
~ c: 4000
I»
~ .2
0' 1\1
....
...... ED 1::
I\) - Q)
(.J
U> 0 c: 3000
- 8
I»
'1J ro
U
0 2:
ED
UI 2000
(")
0
::I
(")
CD
~
.. 1000
~
I»
...
0'
~
UI
..
0
en
~
CD
C'
0
'<
co
I»
~
:D
CD
CD
-
-
0'
::r
o
small large small large
alewife alewife bloater bloater
lake
trout
3
lake
trout
4
lake
trout
5
lake
trout
6
lake
trout
7
lake
trout
8
lake
trout
9
lake
trout
11
lake
trout
12
-------
."
cs.
c
...
CD
N
N
~
r-
0 2.0
~
co
..!.
CD 1.8
...
3
en
n
CD 1.6
~
I» - Segment 1
...
c)" 1.4 ....- Segment 2
0
'0 \. 0 - Segment 3
CiJ .\
:J 1.2 -.. Segment 8
a. -
n. C) - - - Segment 9
- oS
c)" c: --- Segment 10
~ 0 1.0
fn ..
0 !!! 0 water data
.- -
- c:
I\) 3 ~ 0.8
~
I» c:
:so 0
u
Ai' 0.6
~
CD
-
0
- 0.4
!.
'1J
0
OJ 0.2
fn
n
0
~ 0.0
n
CD 1940 1950 1960 1970 1980 1990 2000 2010
~
-
... year
I»
-
o.
~
sn
-------
cnon
C') -.
C1Ice
::::J C
I» CiJ
~
-. I\)
O.
O~
- .
o
05:
~-
-.0
ce.::t
::::J-I
!.O
3><
o rn
0.0
C1I C
--
,,::r
~ C1I
C1I ~
0.::::J
-',
2,1»
-. ~
o C1I
::::J
!II~
C')
::r
ce'
I»
::::J
.......
I\) -
o
0'1 -
e?.
"'tI
o
I:D
rn
"
~
C1I
0.
c:r
-
o'
::J
!II
o
o
3
"
I»
~
fir
o
::J
o
-
o
::J
ce
I
-
C1I
~
3
.........
-I
g> 2.5
-
c
o
:;:;
~ 2.0
c
Q)
o
c
o
~ 1.5
()
a..
4.0
3.5
3.0
1.0
0.5
-
/
I
I
\
\
\
\
- - - segment 1 - original model
- - segment 8 - original model
- segment 1 - scenario C
- segment 8 - scenario C
I
I
I
I
I
I
I
I
I
\
\
\
, - ,
,
I
,
,
"-
\
,
\
\
I
I
,
I
I
,
I
,
\
,
\
\
\
\
\
\
/
/
/
I
,
\
\
\
\
\
\
,
/
/ '
/ "
/ "
,
,
"
"
"
"
/ ,
;"
;" "
...... '
...... '
."...--:: ' , ,
0.0
1940
1950
1960
2010
1970
1980
1990
2000
year
-------
- ."
U) -.
&cg
-. ..
3 CD
CDN
:J'
_N
()!D
~o
CD 0
U) 3 300
()'tJ
2.1»
- ..
CD -.
() U)
-0
CD:J
0.0 250
-. -
:Jcn
....()
CO CD
co:J
""1»
I ..
.... -.
coo 200
~O -
~'tJ OJ
--
; OJ
c:
Q, -
r)' c:
0
- .. 150
o' cu
...... :J '"
-
I\) U) c:
0> - ~
0 c:
3 0
0
I» 100
5'
ii
~
CD
en
CD 50
Q,
3
CD
:J
-
- .
o
- 0
!.
"'tJ 1900
o
CD
U)
()
0
:J
()
CD
:J
-
a
6'
::s
U)
o
1920
.
o
.
.
1940
o
o
o
.
o
o
o
.
.
/' -- - ,8
/
/
o
.
1960
year
1980
. .
o 0
.
o
.
o
.
- Segment 11
- Segment 12
- - Segment 13
. site 18
. site 47s
o site 6Bk
.
.
8
\
"-
"-
"-
.....
---
2000
2020
-------
Table 2.9. Mass Balance Diagnostics for Total PCBs in MICHTOX Scenario B Simulation (Year 1994-
1995)
Flux (kg/d)
Mass Transport Pathway Main Lake Green Bay
Green Bay Export 38 -38
Straits of Mackinac Export 0
Chicago River Export 0
Tributary Loading 126 220
Atmospheric Deposition 216 15
Net Volatilization 758 432
Volatilization (Gross) 3000 502
Gas Absorption 2243 70
Settling 948 1641
Resuspension 1152 1811
Burial 349 28
Net (Mass In - Mass Out) -726 -262
Inventory (kg)
Total PCBs Inventory Main Lake Green Bay
Water Column 690 57
Surficial Sediment 7070 4370
25
o
o
20
o 0
;;
! 15
.~
!
c:
~
8 10
In
o
'=
o
- age 7 prediction
o DeVault data
nnn age 8 prediction
- age 9 prediction
- - age 10 predIction
o
1940
1950
1960
1970
1980
1990
2000
2010
year
Figure 2.30. Comparison of long-term Scenario C predictions to DeVault et al. (1986) lake trout
data.
127
-------
0.."
m -,
_co
!» c
ft1
N
W
....
,
0
0
3
'0
m
...
0' 7000
0
~
0
-
s:
(; 6000
%
d
)( 0 average PCB
en 5000 . model
n
CD
~
m -
... C)
o' ~
.... 0 ..s 4000
I\) - c::
00 g .Q
~
!!. c:
'1J ~ 3000
0 c::
aJ 0
0
fI) cc
n u
0 2=
~
n 2000
CD"
~
-
...
m
-
0' 1000
~
fI)
-
0
tn
~
CD
c:T
0
~
m
~
2J
CD
!!.
-
C;;'
~
o
small large small large
alewife alewife bloater bloater
lake
trout
3
lake
trout
4
lake
trout
5
lake
trout
6
lake
trout
7
lake
trout
8
lake
trout
9
lake
trout
11
lake
trout
12
-------
Although it is tempting to conclude from the model
confirmation, that long-term forcing functions such as
Scenario B are accurate estimates of past PCBs
inputs to Lake Michigan, it must be cautioned that
these estimates are highly speculative. For example,
according to the Scenario B estimates, the
cumulative tributary loading of PCBs to Lake
Michigan through 2002 is 41 ,000 kilograms. It is not
clear how this estimate could be reconciled with
Swackhamer and Armstrong's (1988) estimate that
500,000 kg of PCBs were released to Lake Michigan
from Waukegan Harbor alone. Clearly, the timing
and magnitude of a PCBs input, such as Waukegan
Harbor, would have a profound impact on model
simulations. It is very difficult to make reliable
estimates of past forcing functions for PCBs that are
both consistent with, and constrained by, the
available data.
2.5.6 Toxic
Forecasts
Chemical
Management
The revised MICHTOX model was next applied to
forecast total PCBs concentrations in lake trout for a
number of scenarios in which the future PCBs forcing
functions were changed from their 1994-1995
estimated values. These changes were intended to
represent alternative strategies for managing PCBs
in Lake Michigan, and MICHTOX was used to
forecast the effectiveness of these alternatives in
terms of reducing lake trout total PCBs
concentrations. The forecast simulations
commenced in 1994 using LMMBP average total
PCBs concentrations to define initial conditions in
water, sediment, and fish. Forcing functions were
changed starting in year 2002 of the simulations,
which was treated as the effective date of the
scenario-specific control actions. The predictions for
total PCBs concentrations in age seven lake trout in
the southern lake segment are displayed in Figure
2.32. With the exception of the No-Action scenario,
all forecast predictions are the same until simulation
year 2002, at which point predictions diverge due to
changes in the forcing functions. None of the
forecasts reach steady-state (Le., constant total
PCBs concentrations) within the 12 years simulated
following the control actions. These predictions are
also presented in Table 2.10 in terms of total PCBs
concentrations five and ten years after the change in
forcing functions.
The No-Change and Elimination scenarios bracket
the range of possibilities, in terms of total PCBs
concentrations in lake trout, that can be expected as
a result of managing loadings and forcing functions.
The No-Action forecast predicts that total PCBs
concentrations in lake trout will decline by 50% in 10
years, if loadings and vapor concentrations continue
over this period at their 1994-1995 levels. This
decline occurs because the inventory of total PCBs
in the lake sediments is being slowly depleted by
volatilization and burial processes, and this depletion
results in declining exposure concentrations in both
the water column and the surficial sediment. The
Elimination forecast predicts that total PCBs
concentrations in lake trout will decline by 75% in 10
years if loadings and vapor concentrations are
eliminated in 2002. According to these predictions,
the maximum achievable reduction in future lake
trout total PCBs concentrations over this 10-year
period is from 0.82 to 0.38 ~g/g, or 55%. This
reduction would require the cessation of all PCBs
inputs to Lake Michigan.
~
The Fifty-Percent Forcing Function Reduction and
the Fifty-Percent Loading Reduction scenarios
explore the effectiveness of incremental toxic
chemical management alternatives. As expected,
the Fifty-Percent Forcing Function Reduction
scenario is forecast to be half as effective as
Elimination in terms of predicted total PCBs
concentrations in lake trout after 10 years. The
forecast total PCBs concentration is 0.60 ~g/g after
10 years, a 60% reduction from the predicted
concentration in 2002. The Fifty-Percent Loading
Reduction scenario demonstrates how the
effectiveness of toxic chemical management is
diminished if atmospheric vapor concentrations are
not brought under control. For this scenario, tributary
and atmospheric deposition loadings are reduced
50%, but vapor concentrations are held at 1994-1995
values. In this case, the forecast total PCBs
concentration is 0.77 ~g/g after 10 years, only a 50%
reduction from the predicted concentration in 2002,
and only six percent lower than the 10-year
concentration forecast for the No-Change scenario.
The No-Action scenario differs from the others. It
assumes that atmospheric vapor concentrations will
continue to decline in the future according to the rate
of decline observed over the past 25 years. As noted
previously, there is no scientific consensus as to
129
-------
......
Co)
o
en."
n» -.
-------
Table 2.10. MICHTOX Predictions of Total PCBs Concentrations (JIg/g) in Lake Trout for Toxic
Chemical Management Alternatives
Years After Action
Simulated Control Action
No-Change
No-Action
Fifty-Percent Forcing Function Reduction
Fifty-Percent Loading Reduction
Forcing Function Elimination
o
1.57
1.46
1.56
1.57
1.56
5
1.09
0.88
0.91
1.04
0.74
10
0.82
0.55
0.60
0.77
0.38
whether this is a realistic expectation. For this
scenario, the forecast total PCBs concentration is
0.55 j.Jg/g after 10 years, a 60% reduction from the
predicted concentration in .2002. In other words, the
reduction in total PCBs concentration for the No-
Action scenario is about the same as that forecast for
the Fifty-Percent Forcing Function scenario.
Comparison of the No-Action scenario to the No-
Change and the other toxic chemical management
forecasts demonstrates that the effectiveness of
control action depends upon understanding which
forcing functions are controllable and what the future
trend in forcing functions will be in the absence of
control actions. It could be dangerous to assume
that atmospheric vapor concentrations will continue
to decline in the future according to the rate of
decline observed in the past. Clearly, a better
understanding of long-term trends in total PCBs
vapor concentrations would lead to more accurate
forecasts of toxic chemical concentrations expected
from control actions. Ultimately, this will be essential
if models are to inform decisions regarding the
control of PCBs in Lake Michigan and the other Great
Lakes.
2.5.7 Uncertainty of MICHTOX Model
Predictions
Model predictions are uncertain for a number of
reasons, including: conceptual errors and/or
omissions, errors in parameterization,
uncharacterized system variability, and systematic
errors in forcing functions and calibration data. The
uncertainty of MICHTOX predictions, arising from
errors in model parameterization and forcing
functions, has been evaluated using conventional
and BMC analyses. Uncertainty analyses have been
conducted on a steady-state version of MICHTOX
due to the computational requirements of the
methods involved. The results of BMC analysis of the
original steady-state model are presented in Table
2.11. The "prior" (a p~jOf/) results are comparable to
conventional Monte Carlo analysis, in which the
uncertainty of parameters and forcing functions is
defined as uncorrelated probability distributions. The
parameters and forcing functions used to generate
the posterior results have been "informed" by
application of the likelihood function (Dilks et al.,
1992) based upon the residuals between model
predictions and confirmation data. In other words, the
reduction in uncertainty evident in the posterior
results (indicated by the 95% confidence intervals in
Table 2.11) reflects the utility of the confirmation
data. Although the BMC analysis shown here was
conducted prior to the availability of the LMMBP data,
the results are expected to be representative of the
uncertainty in MICHTOX total PCBs predictions: total
PCBs concentrations should be well within a factor of
two of the model predictions. It is also possible that
repeating the BMC analysis using LMMBP data and
forcing functions would result in even smaller
confidence intervals for predictions and, hence, less
uncertainty .
2.5.8 Are Lake Michigan Total PCBs
Concentrations in Equilibrium With
Atmospheric Vapor Concentrations?
For the past decade, scientists have debated
whether PCBs concentrations in the Great Lakes
131
-------
Table 2.11. Results of BMC Uncertainty Analysis for Original Steady-State MICHTOX Model
Prior Total PCBs Posterior Total PCBs
Concentration Predictions Concentration Predictions
Model State 95% Confidence 95% Confidence
Variable Units Mean Interval Mean Interval
Water Column ngiL 0.280 (0.14 - 0.57) 0.297 (0.20. ~4)
Surficial Sediments nglg 62.6 (30-130) 71.4 (48 - 105)
Lake Trout nglg 2354 (720 - 7700) 2620 (1700 - 4100)
have reached a state of equilibrium with atmospheric
vapor concentrations (International Joint
Commission, 1996; Stowe et al., 1995; Smith, 1995,
2000; U.S. Environmental Protection Agency, 1993).
This interpretation is an alternative to the hypothesis
that PCBs concentrations largely reflect the
resuspension of PCBs from the lake sediments or
other ongoing sources such as tributaries. Obviously,
the outcome of this debate has significant
implications regarding how best to manage PCBs
and possibly other toxic chemicals in Lake Michigan
and the other Great Lakes (as was illustrated by the
forecasts of toxic chemical management alternatives)
and was a primary motivation for conducting the
LMMBP.
Figure 2.33 plots the observed trends of total PCBs
concentrations in air (vapor concentrations for the
Great Lakes region reported by Schneider et al.,
2001) and Lake Michigan water. The similarity
between these trends suggests that PCBs
concentrations may be at or near equilibrium. To
explore this further, predictions from the No-Action
forecast scenario were used to calculate the water-
to-vapor ratio of total PCBs concentrations (dissolved
total PCBs concentration in water/atmospheric vapor
total PCBs concentration). The result, plotted in
Figure 2.34, indicates that for this scenario, the
water-to-vapor concentration ratio increases with
time. Therefore, the model does not predict
equilibrium between air and water total PCBs
concentrations. This analysis should be repeated for
other forcing function scenarios to determine whether
this result is general or specific to the forcing function
assumption.
2.5.9 Sensitivity of Bioaccumulation
Predictions to Initial Total PCBs
Concentrations in Fish
Models such as MICHTOX are generally quite
sensitive to initial concentration conditions. Total
PCBs concentration predictions from the initial years
of the forecast scenc!rios (Figure 2.32) display a
transient increase, which is attributable to initial
conditions in the model. Tests of the
bioaccumulation model sensitivity to initial conditions
indicated that the model predictions were not
sensitive to initial conditions atter a number of years
equal to the age class of fish being examined. This
is illustrated in Figure 2.35, where substantially
different initial total PCBs concentrations were used
for the two simulations. Comparison of predictions
for age seven lake trout show that sensitivity to ir 11
conditions disappears within the first six to se .-,n
years of simulation. Therefore, the five- and ten-year
predictions for toxic chemical management forecasts
(Figure 2.32 and Table 2.1 0) do not depend upon the
specification of initial total PCBs concentrations in
fish.
2.5.10 Sensitivity of Bioaccumulation
Predictions to Food Chain Model
Parameterization
Confirmation of the total PCBs bioaccumulation
predictions for both steady-state exposure (Figures
2.10 and 2.11) and Scenario B simulations (Figures
2.25 and 2.26) suggest that predictions are biased
low b~ 20~30% for most fish age and size classes.
Recahbratlon of the MICHTOX food chain model may
132
-------
"'"
US"
c
CiJ
N
W
~
~ 2
o
...
!!!.
"'a 1.8
o
IJJ
In
C') 1.6
o
::;,
C')
CD
::;, 1.4
...
..,
C»
...
0" c 1.2
::;, .Q
In 10
~
::;, 'E
Q)
r- u 1
C» c
0
; u
co
...... 3: c..> 0.8
c.:>
c.:> n" ~
~
ca'
C» 0.6
::;,
C»
::;"
C» 0.4
::;,
Q.
:e 0.2
C»
...
CD
~
0
1970
o
. vapor concentration (ng/m3)
o water concentration (ng/L)
.
o
o
.
.
.
o
.
0 0 .
.
0 D~
~
,
1975 1980 1985 1990 1995 2000
year
-------
l:!!
_cc
-.c
0 ~
;:, CD
-N
O.
~~
l:u
~a 40000
Cr
g,
-
0 35000 -+- surface water/vapor
-
!!. --0- deep water/vapor
."
0
m c:: 30000
0 0
n ~
0 ~
;:, .....
c::
n CD 25000
CD u
;:, c::
- 0
... u
C» 0
-
o. Q.
(1j 20000
;:, >
0 c
... CT .Q
(J.)
~ CD i6
i ...
C 15000
m CD
u
;:, c::
0
Co u
iir "0
CD 10000
0 >
0 (5
~ II)
.!a
Co "0
~ 5000
II)
-
CD
~
II)
;:,
Co
Dl
'1:J
0
~
-
i:
n
:J:
d
)(
z
9
o
1990
1995
2000
"2005
2010
year
2015
2020
-------
-"'n
0 -"
CjJca
n C
m (;
!1N
"w
~
en
CD
::J
en
~ 2.5
<"
~
'<
0
-
-
0
0
0.
n
::r
m
:i" .-..
3 OJ
--
0 OJ
0. :J
~ -; 1.5
... 0
0 +::
ro
:j" ...
...... -
c.:I ~ c:
Q)
01 Dr u
c:
n 8 1
o
::J CD
0. ()
~ Q.
c)" -
::J
en
;g 0.5
CD
en
CD
<
CD
::J
i»
'"
CD
...
..
0
C
~
Z
0
.
0
::r
m
::J
co
CD
-+- initial condition from steady-state model
- initial condition from Sheboygan Reef average concentrations
o
1995
2000
2005
2010
2015
2020
year
-------
be appropriate, given that the parameterization of this
model has never been optimized using field data.
Although recalibration was beyond the scope of this
project, sensitivity analysis was used to demonstrate
that an optimal fit of the LMMBP total PCBs data
could be readily achieved by adjusting a single food
chain model parameter. The parameter of interest,
the chemical assimilation efficiency, was increased to
0.8. This was the maximum value suggested by the
literature and prior model applications. Comparisons
of the model predictions, made with this parameter
adjustment, to the initial model predictions and the
data for total PCBs concentrations in fish at
Saugatuck, are presented in Figure 2.36.
Adjustment of chemical assimilation efficiency alone
increases the total PCBs concentrations predicted by
the food chain model by over 50%. Other important
bioaccumulation parameters, including diet
composition, growth rates, lipid contents, chemical
excretion rates, and phytoplankton bioconcentration
factors, should also be examined in any recalibration
effort. However, sensitivity analysis demonstrates
that it should be possible to optimize the
bioaccumulation predictions to data collected in either
Saugatuck or Sheboygan Reef biota zones or some
aggregation of these data.
2.5.11 Are Total PCBs Bioaccumulation
Factors Constant for Lake Trout in Lake
Michigan?
Bioaccumulation factors (BAFs) are defined as the
ratio of chemical concentration in fish (normalized by
fish lipid content) to dissolved concentrations in
water. BAFs are often used as simple substitutes for
food chain model predictions, for example to estimate
chemical concentrations in fish from measured
concentrations in water (U.S. Environmental
Protection Agency, 2000). However, BAFs are not
necessarily constant in the ecosystem. This can be
demonstrated by calculating BAFs from the food
chain model results. BAFs calculated from the No-
Change forecast scenario, plotted in Figure 2.37,
show that BAFs vary continuously throughout the
simulation. The BAFs initially increase rapidly, reach
a maximum of over 7 x 107, and then decline at a
constant rate for the duration of the simulation.
Thus, BAFs should be used with caution in Lake
Michigan, as they are expected to vary with time.
2.6 References
Ambrose, A.B., T.A. Wool, J.P. Connolly, and R.W.
Schanz. 1988. WASP4, A Hydrodynamic and
Water Quality Model - Model Theory, User's
Manual and Programmer's Guide. U.S.
Environmental Protection Agency, Office of
Research and Development, Environmental
Research Laboratory-Athens, Georgia.
EPA/600/3-87/039, 297 pp.
Baker, J.E. and S.J. Eisenreich. 1990.
Concentrations and Fluxes of Polycyclic Aromatic
Hydrocarbons and Polychlorinated Biphenyls
Across the Air-Water Interface of Lake Superior.
Environ. Sci. Techno!., 24(3):342-352.
Bamford, H.A., D.L. Poster, and J.E. Baker. 2000.
Henry's Law Constants fo Polychlorinated
Biphenyl Congeners and Their Variation With
Temperature. J. Chern. Engin. Data, 45:1069-
1074.
Bamford, H.A., D.L. Poster, A.E. Huie, and J.E.
Baker. 2002. Using Extrathermodynamic
Relationships to Model the Temperature
Dependence of Henry's Law Constants of 209
PCB Congeners. Environ. Sci. Techno!.,
36(20):4395-4402.
Brunner, S., E. Hornung, H. Santi, E. Wolff, O.G.
Piringer, J. Altschuh, and R. Bruggemann. 1990.
Henry's Law Constants for Polychlorinated
Biphenyls: Experimental Determination and
Structure-Property Relationships. Environ. Sci.
Techno!., 24(11 ):1751-1754.
Burkhard, L.P. 1984. Physical-Chemical Properties
of the Polychlorinated Biphenyls: Measurement,
Estimation, and Application to. Environmental
Systems. Ph.D. Dissertation, Water Chemistry
Department, University of Wisconsin, Madison,
Wisconsin. 275 pp.
Connolly, J.P. 1991. Documentation for Food Chain
Model, Version 4.0. Manhattan College,
Department of Environmental Engineering and
Sciences, Riverdale, New York.
136
-------
......
c..>
......
CD"TI
=-.
_.cc
n c
-. ...
CD CD
::;,
nl\)
'
-------
-"'"
0 -.
~co
CD c
() ~
I» CD
!!l.N
0(.)
-.......
3.
5.~
I» ~ 8,E+07
~CD
00.
::::I -.
":,,,,2.
CD
0. 7.E+07
tT
o'
I»
()
() 6.E+07
c ij)
3 ~
c ~
ii u.
- ~
o' a:I 5.E+07
::::I Q)
- UJ
I» m
() .l:
- a.
0 1J
~ Q)
0 >
...... - (5
Co) 0 UJ
CD ~ .!!!
- 1J
0 ,j
-
I» Q)
.~
~ iii
0 E 2.E+07
ED ....
0
0 c:
I
1J
::::I 'a'
I» 1.E+07
CO
CD
0
CD
<
CD O.E+OO
::::I
ii 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015
r year
-
~
0
c
-
-
Z
0
.
0
:T
I»
::::I
co
CD
-------
DeVault, D.S., W.A Willford, R.J. Hesselberg, D.A
Nortrupt, EG.S. Rundberg, AK. Alwan, and C.
Bautista. 1986. Contaminant Trends in Lake
Trout (Salvelinus namaycush) from the Upper
Great Lakes. Arch. Environ. Contam. Toxico!.,
15:349-356.
Dilks, D.W., R.P. Canale, and P.G. Meier. 1992.
Development of Bayesian Monte Carlo
Techniques for Water Quality Model Uncertainty.
Eco!. Mode!., 62:149-162.
Dunnivant, F.M., A.W. Eizerman, P.C. Jurs, and M.N.
Hasan. 1992. Quantitative Structure-Property
Relationships for Aqueous Solubilities and
Henry's Law Constants of Polychlorinated
Biphenyls. Environ. Sci. Techno!., 26(8):1567-
1573.
Gobas, F.A.P.C., M.N.Z. Graggen, and X. Zhang.
1995. Time Response of the Lake Ontario
Ecosystem to the Virtual Elimination of PCBs.
Environ. Sci. Techno!., 29(8):2038-2046.
Green, M.L., J.V. DePinto, C. Sweet, and K.C.
Hornbuckle. 2000. Regional Spatial and
Temporal Interpolation of Atmospheric PCBs:
Interpretation of Lake Michigan Mass Balance
Data. Environ. Sci. Techno!., 34(9):1833-1841.
Hillery, B.L., I. Basu, C.W. Sweet, and R.A Hites.
1997. Temporal and Spatial Trends in a Long-
Term Study of Gas-Phase PCB Concentrations
near the Great Lakes. Environ. Sci. Techno!.
31(6):1811-1816.
Hillery, B.L., M.F. Simcik, I. Basu, R.M. Hoff, W.M.J.
Strachan, D. Burniston, C.H. Chan, K.A. Brice,
C.W. Sweet, and R.A. Hites. 1998. Atmospheric
Deposition of Toxic Pollutants to the Great Lakes
as Measured by the Integrated Atmospheric
Deposition Network. Environ. Sci. Techno!.,
32(15):2216-2221.
International Joint Commission. 1996. Workshop of
the International Joint Commission (IJC) Great
Lakes Science Advisory Board's Workgroup on
Parties Implementation: PCBs, The New
Equilibrium? International Joint Commission,
Windsor, Ontario, Canada.
Liss, P.S. 1973. Processes of Gas Exchange
Across an Air-Water Interface. Deep Sea Res.,
20:221-228.
Mackay, D. 1989. Modeling the Long-Term
Behavior of an Organic Contaminant in a Large
Lake: Application to PCBs in Lake Ontario. J.
Great Lakes Res., 15(2):283-297.
Marti, EA. and D.E Armstrong. 1990.
Polychlorinated Biphenyls in Lake Michigan
Tributaries. J. Great Lakes Res., 16(3):396-405.
O'Connor, D.J. 1983. Wind Effects on Gas-Liquid
Transfer Coefficients. J. Environ. Engin.,
109(3):731-752.
Robbins, J.A. 1985. The Coupled Lakes Model for
Estimating the Long-Term Response of the Great
Lakes to Time-Dependent Loadings of Particle-
Associated Contaminants. National Oceanic and
Atmospheric Administration, Great Lakes
Environmental Research Laboratory, Ann Arbor,
Michigan. NOAA Technical Memorandum Report
ERL GLERL-57, 41 pp.
Rygwelski, K.R., W.L. Richardson, and D.O.
Endicott. 1999. A Screening Model Evaluation of
Atrazine in the Lake Michigan Basin. J. Great
Lakes Res., 25(1):94-106.
Schneider, A.R., H.M. Stapleton, J. Cornwell, and
J.E. Baker. 2001. Recent Declines in PAH,
PCB, and Toxaphene Levels in the Northern
Great Lakes as Determined from High Resolution
Sediment Cores. Environ. Sci. Techno!.,
35(19):3809-3815.
SChwarzenbach, R.P., P.M. Gschwend, and D.M.
Imboden. 1993. Environmental Organic
Chemistry. John Wiley and Sons, Incorporated,
New York, New York. 681 pp.
Smith, D.W. 1995. Are PCBs in the Great Lakes
Approaching a "New Equilibrium"? Environ. Sci.
Techno!., 29(1 ):42a-46a.
Smith,D.W. 2000. Analysis of Rates of Decline of
PCBs in Different Lake Superior Media. J. Great
Lakes Res., 26(2):152-163.
139
-------
Stowe, C., S.R. Carpenter, L.A. Eby, J.F. Amrhein,
and R.J. Hesselberg. 1995. Evidence That
PCBs are Approaching Stable Concentrations in
Lake Michigan Fishes. Ecol. Appl., 5(1 ):248-260.
Swackhamer, D.L. and D.E. Armstrong. 1988.
Horizontal and Vertical Distribution of PCBs in
Southern Lake Michigan Sediments and the
Effect of Waukegan Harbor as a Point Source. J.
Great Lakes Res., 14(3):277-290.
Thomann, R.V. and D.M. Di Toro. 1983. Physico-
Chemical Model of Toxic Substances in the Great
Lakes. J. Great Lakes Res., 9(4):474-496.
Totten, L.A., P.A. Brunciak, C.L. Gigliotti, J. Dachs,
T.R. Glenn, E.D. Nelson, and S.J. Eisenreich.
2001. Dynamic Air-Water Exchange of
Polychlorinated Biphenyls in the New York-New
Jersey Harbor Estuary. Environ. Sci. Technol.,
35(19):3834-3840.
U.S. Environmental Protection Agency. 1993.
Proposed Great Lakes Water Quality Guidance.
Federal ~egister, 58:20806-20809.
U.S. Environmental Protection Agency. 1997. Lake
Michigan Mass Balance Modeling Work Plan.
U.S. Environmental Protection Agency, Great
Lakes National Program Office, Chicago, Illinois.
36 pp.
U.S. Environmental Protection Agency. 2000. Lake
Michigan Lakewide Management Plan (LaMP
2000) - Main Report. U.S. Environmental
Protection Agency, Great Lakes National
Program Office, Chicago, Illinois. 254 pp.
Velleux, M. and D. Endicott. 1994. Development of
a Mass Balance Model for Estimating PCB Export
from the Lower Fox River to Green Bay. J. Great
Lakes Res., 20(2):416-434.
Wanninkhoff, R.J. 1992. Relationship between Gas
Exchange and Wind Speed over the Ocean. J.
Geophys. Res., 97:7373-7381.
140
-------
vvEPA
United States
Environmental Protection
Agency
Please make all necessary changes on the label
below, detach or copy, and return to address in upper
right-hand comer.
If you do not wish to receive these reports, CHECK
HERE D, detach or copy this cover, and return to
address in upper right-hand comer
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT No. G-35
Research and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
EPA/600/R-05/158
December 2005
www.epa.gov
Recycled/Recyclable
Printed with vegetable-based ink
on paper that contains a minimum
of 50% post-consumer fiber
content processed chlorine-free
------- |