600R08111
   United States
   Environmental Protection
   Agency
Results of the Lake Michigan
   Mass Balance Project:
  Atrazine Modeling Report
 RESEARCH AND DEVELOPMENT

-------
                                                     EPA/600/R-08/111
                                                       September 2008
Results  of the Lake Michigan  Mass
                Balance  Project:
                      Atrazine
                Modeling Report
                         Prepared for

             United States Environmental Protection Agency
                Great Lakes National Program Office
                   77 West Jackson Boulevard
                     Chicago, Illinois 60604
                         Prepared by

             United States Environmental Protection Agency
                Office of Research and Development
       National Health and Environmental Effects Research Laboratory
                  Mid-Continent Ecology Division
          Large Lakes and Rivers Forecasting Research Branch
                   Large Lakes Research Station
                       9311 Groh Road
                   Grosse lie, Michigan 48138
                   Kenneth R. Rygwelski, Editor
                                              Recycled/Recyclable
                                              Printed with vegetable-based ink on
                                              paper that contains a minimum of
                                              50% post-consumer fiber content
                                              processed chlorine free.

-------
                                         Notice
The information in this document has been obtained  primarily through  funding by the  United States
Environmental Protection Agency (USEPA) under the auspices of the Office of Research and Development
(ORD) and by the Great Lakes National Program Office (GLNPO).  The report has been subjected to the
Agency's peer and administrative review and it has been approved for publication as an USEPA document.
Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

-------
                                         Foreword
The  Lake Michigan  Mass  Balance Project (LMMBP) was initiated  by the United States  Environmental
Protection Agency (USEPA), Great Lakes National Program Office (GLNPO) to determine strategies for
managing and remediating toxic chemicals in the lake basin.  Within the ecosystem approach, the mass
balance framework is considered the best means of accomplishing this objective, and GLNPO requested the
assistance of the USEPA Office of Research and Development (ORD) to facilitate and produce mathematical
models that account for the sources, sinks, transport,  fate, and  food chain bioaccumulation  of certain
chemicals.  This approach has been used in the past and builds upon the modeling efforts that have occurred
in  the Assessment and  Remediation  of Contaminated  Sediments (ARCS)  Program and  the lower Fox
River/Green Bay Mass Balance Project. The feasibility of such studies and resultant alternative management
options for contaminants in  large rivers and a large embayment were demonstrated, and a logical extension
to  the entire Lake Michigan receiving water body and major tributaries was warranted. There were a large
number of cooperators in this project, and by focusing federal, state, local, private, and academic efforts and
resources on a common goal, much more was accomplished than if these entities acted independently.

The project was conducted in conjunction with the Enhanced Monitoring Program, and the approach required
that  all monitoring and field research  be coordinated and common  methodologies used.  Mathematical
modelers were consulted during planning for sample design, parameters, and temporal and spatial sampling
considerations.  This yielded a consistent and reliable database of information that was accessible by project
participants and the public. Data for  the LMMBP were collected  during 1994 and 1995 and have been
compiled according to specified quality assurance/quality control (QA/QC) requirements, and other data
assessments have been made for modeling purposes.

The  need to consider the environmental benefits and consequences  of alternative remediation choices to
protect and improve our environment continues to intensify as: 1)  environmental problems become more
complex; 2) the means to address and investigate problems become  more technical, time-consuming, and
expensive; and 3) the actual cost to implement action strategies has escalated. The integrated atrazine mass
balance modeling results are presented in this document and can aid managers in establishing priorities for
both lake-wide  and  local improvements.  Primary goals  of the modeling effort were to  determine the
persistence of atrazine and to forecast concentrations in Lake Michigan water.  The capability of forecast
modeling presented here is a salient feature of this approach directed toward providing multiple alternatives,
which then can be examined through benefit-cost analyses.

This report presents the current  status and results of the atrazine modeling effort through 2005, and it fulfills
documentation requirements as described in the Quality  Assurance Plan for Modeling: The  Lake Michigan
Mass Balance Project. Of course, a model and modeling applications are never complete, and it is expected
that further efforts will change some results, insights, and our understanding of Lake Michigan. These efforts
require an investment of resources and time, and improvements with additional  model run executions are
measured in years. In the larger picture, the need for Agency modeling technologies continues to intensify,
and the requirement for reduced uncertainty will lead to future improved generations of models.  We have

-------
placed great emphasis on following guidance provided by the USEPA and other agencies in assuring that the
scientific theory is implemented accurately and completely by model computer code and that best modeling
practices have been instituted. The fundamental principles driving the atrazine models presented in this report
have received scientific peer review using an interdisciplinary panel of scientists and experts.  The purpose
of the reviews was to ensure that decisions based on the modeling efforts  are reliable  and scientifically
credible.

This document is not intended to include all of the details and background required to understand the entire
LMMBP.  Rather, the reader should refer to the LMMBP Work Plan and other materials on the GLNPO web
site and the Lake Michigan Mass Balance Modeling Quality Assurance Plan on the ORD-Grosse lie web site
for further information.
                                               IV

-------
                                          Abstract
The Lake Michigan Mass Balance Project (LMMBP) was conducted to measure and model nutrients, atrazine,
polychlorinated biphenyls (PCBs), frans-nonachlor, and mercury to gain a better understanding of the sources,
sinks, transport,  fate, and effects of these substances within the system  and to aid  managers  in the
environmental decision-making process for the Lake Michigan basin.  The  United States Environmental
Protection Agency (USEPA) Office of Research  and Development (ORD) was requested to conduct and
facilitate modeling in cooperation with the USEPA Great Lakes National Program Office (GLNPO); the USEPA
Region V; other federal agencies; the states of Michigan, Wisconsin,  Illinois, and Indiana; the tribes; and the
public and private sectors.

This report focuses on the load sources and fate and transport modeling of atrazine only. In the Lake Michigan
basin, atrazine is used primarily as a herbicide on corn crops. With the recent increase in corn acreage in the
United States associated with biofuel (ethanol) production, increased loadings of atrazine to lakes and streams
are expected.

The atrazine modeling effort described in this report was supported by intensive sampling of the atmosphere,
major tributaries,  and water column during the 1994-1995 field years as well as by extensive quality assurance
and database development.  Using these  data and  historical data,  loadings  of atrazine  to the lake were
estimated for the tributaries and atmosphere. Multimedia, mass balance modeling frameworks were applied
to examine primary source and loss categories and make various model forecasts for a  variety of loading
scenarios.  A literature search revealed that atrazine  sorption to particles is negligible.  Hence, atrazine
transport associated with settling, resuspension,  and burial were determined to  be negligible. This  report
focuses on the modeling practices applied and results for atrazine from the MICHTOX screening-level  model
and the higher-resolution LM2-Toxic and LM3-Atrazine models.

The results of the LM2-Toxic system mass balance model show that the largest atrazine load to the lake is
from the watershed. For the year 1994, it was estimated that 5,264 kg of atrazine were discharged to the lake
via the tributaries. The second major load to the lake was from atmospheric wet deposition with a loading
estimate of 2,493 kg.  The greatest loss of atrazine from the lake was through transport to Lake Huron (2,546
kg) via the Straits of Mackinac.  Loss due to internal decay (1,662 kg) was the second largest loss mechanism.
The total inventory of atrazine in the lake was determined to be 184,310 kg in 1994. In this large,  cold northern
lake, the model suggests that in situ atrazine decay is very slow (0.009/year). This translates into an estimated
atrazine half-life  of 77 years.  Using the  model  to  forecast alternative futures,  a 35%  load reduction, if
implemented in January 1, 2005, would have been needed in order to prevent atrazine concentrations from
increasing further in the lake. If loadings and boundary conditions are assumed to be constant in the future,
the model predicts that the lake will eventually reach a steady-state concentration of 66 ng/L in the year 2194.

Our  high-resolution  model,  LM3-Atrazine,  was primarily used  to evaluate   environmental  exposure
concentrations of atrazine in 5 km x 5 km model cells receiving loadings from the major tributaries to the lake.

-------
The model segment receiving loads from the St. Joseph River, associated with the largest tributary load of
atrazine to the lake, ranged from winter concentrations of 37 ng/L to spring peaks of 100-350 ng/L.  These
predicted exposure concentrations in the lake are all below selected toxicological endpoints, including the most
sensitive, phytoplankton primary production reduction.

This synthetic lake-wide perspective is anticipated to aid lake managers in moving forward on prevention,
remedial actions, and legislative priorities associated with Lake Michigan Lake-wide Management Plans. The
models developed provide an in-depth understanding of atrazine transport and fate processes in this valuable
freshwater resource.  This abstract does not necessarily reflect USEPA policy.
                                                VI

-------
                                   Tables of Contents
Notice	     ii
Foreword	     iii
Abstract  	     v
Table of Contents	    vii
List of Figures	    xii
List of Tables	    xvi
Abbreviations	   xviii
Acknowledgments	    xx
Executive Summary 	    xxi

Part 1   Introduction 	     1

        Chapter 1 Project Overview	     1

        1.1.1  Background  	     1
        1.1.2  Description	     2
        1.1.3  Scope	     3
               1.1.3.1   Modeled Pollutants	     3
                       1.1.3.1.1  PCBs  	     3
                       1.1.3.1.2  frans-Nonachlor  	     5
                       1.1.3.1.3  Atrazine  	     5
                       1.1.3.1.4  Mercury  	     5
               1.1.3.2  Other Measured Parameters	     6
               1.1.3.3  Measured Compartments	     7
        1.1.4  Objectives	     8
        1.1.5  Design  	     8
               1.1.5.1   Organization  	     8
               1.1.5.2  Study Participants  	     8
               1.1.5.3  Workgroups	     9
               1.1.5.4  Information Management	     9
                       1.1.5.4.1  Data Reporting	     9
                       1.1.5.4.2  Great Lakes Environmental Monitoring Database	    10
                       1.1.5.4.3  Public Access to  LMMBP Data	    11
               1.1.5.5  Quality Assurance Program	    11
        1.1.6  Project Documents and Products	    13
                                               VII

-------
Chapter 2  General Information on the Herbicide Atrazine and Its Degradation
           Products	     15

1.2.1   Background  	     15
1.2.2   Physical-Chemical Properties of Atrazine	     16
1.2.3   Atrazine Degradation 	     17
       1.2.3.1  Biotic Degradation in Surface Water	     17
       1.2.3.2  Abiotic Degradation  in Surface Water	     19
               1.2.3.2.1  Hydrolysis	     19
               1.2.3.2.2  Photolysis	     19
       1.2.3.3  Atrazine Degradation in Soil	     20

Chapter 3 Atrazine Field Data Observations	     23

1.3.1   Background  	     23
1.3.2   Atmospheric Components  	     24
       1.3.2.1  Sampling and Analytical Methodology	     24
       1.3.2.2  Results	     25
               1.3.2.2.1  Atrazine in the Gas Phase Fraction	     25
               1.3.2.2.2  Atrazine in the Particulate Fraction	     25
               1.3.2.2.3  Atrazine and Degradation Products in Wet Deposition	     26
1.3.3   Atrazine in Tributaries	     29
       1.3.3.1  Sampling and Analytical Methodology	     30
       1.3.3.2  Results	     30
1.3.4   Atrazine in Lake Water	     31
       1.3.4.1  Sampling and Analytical Methodology	     31
       1.3.4.2  Results	     31
               1.3.4.2.1  Spatial Variation	     31
               1.3.4.2.2  Seasonal Variation 	     32

Appendix 1.3.1  Information Management	     35

A1.3.1.1  Overview of Information Management at the LLRS  	     35
A1.3.1.2  Summary	     37

Chapter 4 Representativeness of the Lake Michigan Mass Balance Project (LMMBP)
          Years Relative to Lake Michigan's Historic Record  	     46

1.4.1   Introduction  	     46
1.4.2   Ice Cover	     46
1.4.3   Water and Air Temperatures  	     47
1.4.4   Lake Water Levels	     50
1.4.5   Precipitation	     50
       1.4.5.1  Annual Comparisons  	     51
       1.4.5.2  Monthly Comparisons  	     51
1.4.6   Tributary Flows	     51
1.4.7   Summary  	     51
                                       VIII

-------
        Chapter 5 Atrazine Modeling Overview	     55

        1.5.1   Background  	     55
        1.5.2   LMMBP Modeling Objectives	     55
        1.5.3   Historical Modeling	     56
               1.5.3.1  Completely-Mixed Lakes-ln-Series Model  	     57
               1.5.3.2  MICHTOX  	     57
               1.5.3.3  Green Bay Mass Balance Project	     57
        1.5.4   Resolution for the LMMBP Models	     58
        1.5.5   Models Developed and Applied  	     59
               1.5.5.1  Lake Process Models  	     60
               1.5.5.2  Hydrodynamics (POM)  	     60
        1.5.6   Model Quality Assurance  	     60
        1.5.7   Model Application and Computational Aspects	     61
               1.5.7.1  Annual Simulations  	     61
               1.5.7.2  Long-Term Simulations  	     61

Part 2   Lake Michigan Mass Balance Project Atrazine Loadings to Lake Michigan	     63

        Chapter 1 Historical Atrazine Usage in the United States	     63

        2.1.1   Background  	     63
        2.1.2   Total Annual Usage Estimates	     64
        2.1.3   Future Atrazine Use Estimates	     64

        Chapter 2 Estimation of Atrazine Tributary Loadings 	     69

        2.2.1   Atrazine Tributary Load Estimates Utilizing County-Level Atrazine
               Application Data	     69
               2.2.1.1  County-Level Atrazine Application Data	     70
               2.2.1.2  The Watershed Export Percentage	     70
               2.2.1.3  Calculating the Atrazine Tributary Load	     71
        2.2.2   Estimating Atrazine Tributary Loads for Years When County-Level Atrazine
               Application Data Was Not Available	     74
        2.2.3   Atrazine Tributary Loads for MICHTOX and LM2-Atrazine	     75
        2.2.4   Atrazine Tributary Load Estimates for LM3-Atrazine  	     76
               2.2.4.1  Tributary Sampling Program  	     76
               2.2.4.2  Atrazine Load Estimation for Monitored Rivers Using the Stratified
                      Beale Ratio Estimator (SBRE) Method	     77
               2.2.4.3  Atrazine Load Estimation for Unmonitored Watersheds	     78
        2.2.5   Comments on Atrazine Tributary Loading Estimates 	     79

        Chapter 3 Estimation of Atrazine  Loads in Wet Deposition (Precipitation)	     81

        2.3.1   Atmospheric Components Considered  in Modeling Atrazine in Lake Michigan	     81
        2.3.2   Atrazine Wet Deposition Load Estimates Based on Measured Fluxes in the
               Basin	     82
        2.3.3   Atrazine Wet Deposition and Tributary  Loads for MICHTOX and LM2-Atrazine ....     83
                                               IX

-------
Part 3   Lake Michigan Mass Balance Project Level 1 Model:  MICHTOX-Atrazine	     86

        3.1    MICHTOX-Atrazine Executive Summary  	     86
        3.2    MICHTOX-Atrazine Recommendations  	     86
        3.3    Model Description	     86
              3.3.1   Model Overview 	     86
              3.3.2   MICHTOX Model Segmentation and Circulation	     87
        3.4    MICHTOX Model Application to Lake Michigan 	     88
              3.4.1   Screening Model Application	     88
              3.4.2   Enhanced Screening Model Application 	     89
                     3.4.2.1   Field Data	     89
                     3.4.2.2   Model Assumptions and Calibration Procedures  	     89
                     3.4.2.3   Tributary Loadings 	     90
                     3.4.2.4   Atmospheric Loadings  	     90
                     3.4.2.5   Model Confirmation  	     90
                     3.4.2.6   Model Application (Scenarios) 	     90
                     3.4.2.7   Discussion of Results	     91

Part 4   Lake Michigan Mass Balance Project Level 2 Model:  LM2-Atrazine	     95

        4.1    LM2-Atrazine Executive Summary	     95
        4.2    LM2-Atrazine Recommendations	     95
        4.3    Model Description	     95
              4.3.1   Model Overview 	     95
              4.3.2   LM2-Atrazine Model Segmentation and Circulation	     96
        4.4    LM2-Atrazine Model Application to Lake Michigan	     98
              4.4.1   Enhanced Screening Model Application 	     98
              4.4.2   Field Data	     98
              4.4.3   Tributary Loadings  	     99
              4.4.4   Atmospheric Loadings	     99
              4.4.5   Model Assumptions	     99
              4.4.6   Model Calibration and Application (Scenarios)  	     99
              4.4.7   Model Confirmation	    101
              4.4.8   Discussion of Results  	    101

Part 5   Lake Michigan Mass Balance Project Level 3 Model:  LM3-Atrazine	    107

        5.1    LM3-Atrazine Executive Summary	    107
        5.2    LM3-Atrazine Recommendations	    108
        5.3    LM3-Atrazine Transport and Fate Modeling	    108
              5.3.1   Purpose of High-Resolution Model 	    108
              5.3.2   Model Description and Framework	    109
                     5.3.2.1   POM Hydrodynamic Model	    109
                     5.3.2.2   Model Framework	    115
                             5.3.2.2.1   Water Quality Processes	    115
                             5.3.2.2.2   Spatial Resolution	    117
                             5.3.2.2.3   Temporal Resolution	    118
                             5.3.2.2.4   Model Assumptions	    118

-------
              5.3.3  Description of Data Used 	   118
                     5.3.3.1   Field Data	   118
                     5.3.3.2   Initial and Boundary Conditions	   118
                     5.3.3.3   Loadings	   119
                             5.3.3.3.1   Tributary  	   119
                             5.3.3.3.2   Atmospheric  	   120
              5.3.4  Description of Model Simulations and Results  	   121
                     5.3.4.1   Mass Budgets  	   124
                     5.3.4.2   Selected Model Versus Observation Statistics	   124
                     5.3.4.3   Comparison to Toxicological Endpoints 	   124
              5.3.5  Model Uncertainty	   128

Part 6   Review of Atrazine Models	   131

        6.1    LMMBP Atrazine Models  	   131
              6.1.1  Peer Reviews of LMMBP Atrazine Models  	   131
              6.1.2  Comparison of LMMBP Models	   132
        6.2    Comparison of LMMBP Models to Other Recent Atrazine Models Applied to
              Lake Michigan  	   133
              6.2.1  Schottler and  Eisenreich (1997)  	   133
              6.2.2  Tierney etal. (1999)  	   133
        6.3    Atrazine Models Applied to Lake or Deep River Systems Outside the Lake
              Michigan Basin	   134
              6.3.1  Swiss Lakes  	   134
              6.3.2  St. Lawrence  River	   135
        6.4    Atrazine Models Applied to Shallow Surface Water Systems in Agricultural
              Areas  	   135
              6.4.1  Saylorville Reservoir, Iowa  	   135
              6.4.2  Other Small Surface Water Systems	   136
        6.5    Conclusions 	   136

        Appendix 6.1  Peer Review of LMMBP Atrazine Models, September 27, 2000,
                     Romulus, Michigan	   138

        A6.1.1 Overview  	   138
        A6.1.2 Comments on Technical Issues	   139
                                              XI

-------
                                      List of Figures



1.1.1    Simplified mass balance approach	    2

1.1.2    The LMMBP sampling locations	    7

1.1.3    Flow of information in the LMMBP	   10

1.2.1    Chemical structures of atrazine and its major degradation products  	   18

1.3.1    Monthly precipitation amounts at cities in two large corn-growing regions.  Data are
        from Peoria, Illinois and Omaha, Nebraska 	   28

1.3.2    Monthly average temperatures at cites in two large corn-growing regions.  Data are
        from Peoria, Illinois and Omaha, Nebraska 	   29

1.3.3    Atrazine concentrations in Lake Michigan, 1994  	   32

1.4.1    Location of the NOAA buoys in Lake Michigan	   49

1.4.2    Monthly mean water temperatures in southern Lake Michigan	   49

1.4.3    Monthly mean water temperatures in northern Lake Michigan	   49

1.4.4    Mean June water temperatures in southern Lake Michigan	   49

1.4.5    Mean June water temperatures in northern Lake Michigan 	   49

1.4.6    Monthly mean air temperatures in southern Lake Michigan	   50

1.4.7    Monthly mean air temperatures in northern Lake Michigan 	   50

1.4.8    Mean June air temperatures in southern Lake Michigan  	   50

1.4.9    Mean June air temperatures in northern Lake Michigan  	   50

1.4.10   Record of mean monthly water levels for Lake Michigan	   51

1.4.11   Annual precipitation to Lake Michigan between 1949 and 1998	   52
                                              XII

-------
1.4.12   Comparison of 1982, 1983, 1994, and 1995 monthly mean precipitation to the mean
        for the period of 1949 through 1998	    52

1.4.13   Comparison of tributary flow for hydrodynamic model calibration (1982-1983)
        to the historic means  	    53

1.4.14   Comparison of tributary flow for the study period (1994-1995) to the historic means	    53

1.5.1    Surface water segmentation for alternative Lake Michigan mass balance
        model levels	    58

1.5.2    Model construct used for the LMMBP to model atrazine  	    59

2.1.1    Atrazine usage in the United States for 1991	    65

2.1.2    Estimates of atrazine usage in the Lake Michigan basin for 1994 and 1995  	    66

2.1.3    Historical trend of total annual usage of atrazine in the United States with acreage
        planted in corn, sorghum, and sugarcane 	    66

2.2.1    Soil textures typical for the Lake Michigan basin and part of the Lake Erie basin	    72

2.2.2    WEP-based total atrazine tributary loading estimates to Lake Michigan  	    75

2.2.3    Tributary loadings to Lake Michigan MICHTOX model segments  	    76

2.2.4    WEP-based Lake Michigan tributary loadings,  1994  	    76

2.2.5    1995 USGS SBRE atrazine loadings and median concentrations relative to median
        flow in Lake Michigan tributaries 	    78

2.3.1    Wet deposition (rain and snow)  of atrazine for  1991 for Midwestern United States 	    82

2.3.2    Gradients of atrazine in wet deposition loadings over Lake Michigan for May 1994	    83

2.3.3    Seasonality of atrazine wet deposition loadings to Lake Michigan for 1994-1995  	    83

2.3.4    Total atrazine tributary loading and wet deposition loading estimates to Lake Michigan  ....    83

2.3.5    Tributary and wet deposition loadings to MICHTOX model segments for 1994 and
        1995	    84

2.3.6    Tributary and wet deposition loadings to LM2-Atrazine model segments for 1994 and
        1995	    84

3.1      MICHTOX model segmentation	    87

3.2      Total annual estimated tributary and precipitation loadings of atrazine to Lake Michigan ....    88
                                              XIII

-------
3.3     A comparison of MICHTOX - Predicted atrazine concentrations in Lake Michigan
        to averaged Lake Michigan data for the years 1991, 1992, and 1995 are depicted	   89

3.4     Lake Michigan (open-lake) forecast scenarios: 1 - upper estimate of boundary
        condition, 2 - lower estimate of boundary condition, and 3 - estimate of average
        boundary condition 	   92

3.5     Lake Michigan (open-lake) hindcast and scenario forecasts: 4 - virtual elimination
        of all loadings and 0.0 ng/L atrazine at the Straits of Mackinac boundary, 5 - no tributary
        loads, 6 - no wet deposition, 7 - no further degradation of lake water quality	   92

4.1     Water column segmentation for LM2-Atrazine	   97

4.2     LM2-Atrazine model results for Lake Michigan and  Green Bay for the year 1994	  101

4.3     LM2-Atrazine model runs of scenarios  	  102

4.4     Historical trends of United States corn acreage planted and harvested from 1986
         to 2007	  103

4.5     Model-predicted lake-wide averaged atrazine concentrations in water related to
        increases in atrazine loadings resulting from corn crop acreage increases are depicted  ...  104

5.1     Lake Michigan hydrodynamic model 5 km x 5 km computational grid	  110

5.2     Simulated temperature (black) compared to measured temperature (gray) at two buoys
        in Lake Michigan for 1982-1983 	  111

5.3a    Time-series of simulated water temperature versus observed at 45007 for 1994-1995	  112

5.3b    Time-series of simulated surface water temperature versus observed at 45002 and
        45010 for 1994-1995	  113

5.4     Simulated mean temperature (°C) profile for 1982-1983	  114

5.5     Temporal evolution of simulated versus observed temperature profiles, Station 18M	  114

5.6     Watershed  and mid-lake sampling stations for the LMMBP study	  119

5.7     Atrazine loads for Lake Michigan tributaries, 1994-1995	  121

5.8     Comparison of field data to predicted  mid-lake surface concentrations for the 1994-1995
        model simulation and two loading conditions  	  122

5.9     Model simulation results of surface concentrations for May 29, 1995 using long-term
        WEP-based loads	  122

5.10    Comparison of near-shore surface cell model results for the 1994-1995 model simulation
        and two loading conditions  	  123
                                              XIV

-------
5.11     Mid-lake surface concentration model results for 1994-2005 model simulation and two
        loading conditions	   125

5.12     Mass budget average annual results for the 1994-1995 model simulations 	   126

5.13     Comparison of model predictions, measured data, and selected toxicological endpoints  ...   127
                                              xv

-------
                                      List of Tables



1.1.1    Characteristics of the LMMBP Modeled Pollutants	     4

1.1.2    The LMMBP Parameters	     6

1.2.1    Physical and Chemical Properties of Atrazine 	     17

1.3.1    Summary of Wet Deposition Annual Volume-Weighted Mean Deethylatrazine (DEA)
        Concentrations, Atrazine Concentrations, and Deethylatrazine/Atrazine Ratios (DAR) for
        All Stations in the Lake Michigan Basin	     27

1.3.2    Annual Mean Precipitation Amounts Measured at Chicago, Illinois; Fort Wayne,
        Indiana; South  Bend, Indiana; Muskegon, Michigan; Grand Rapids, Michigan; and
        Milwaukee, Wsconsin  	     29

1.3.3    Summary of Historical Atrazine, DEA, and DIA Concentrations in Lake Michigan 	     32

A1.3.1   List of Parameters Analyzed and Principal Investigators for the LMMBP Atrazine
        Modeling  	     36

A1.3.2   Example of Data Verification Checklist Used for the LMMBP	     38

A1.3.3   Printout of Information Stored in the LMMBP Tracking Database Related to Atrazine
        Modeling  	     42

A1.3.4   Generalized Format for the LMMBP Water Data to be Analyzed With IDL Programs  	     43

1.4.1    Summary of Lake Michigan Ice Cover Based Upon Assel (2003)  	     48

2.1.1.   U.S. Department of Agriculture Corn Crop Summaries of Atrazine Usage in the
        United States for 1991, 1994,  and 1995	     65

2.1.2    Total Annual Usage of Atrazine in the United States  	     67

2.2.1    Sources of County-Level Atrazine Application Data for the Lake Michigan Basin	     70

2.2.2    Atrazine Watershed Export Data Summarized From the Literature  	     72

2.2.3    Atrazine Watershed Export Data From Various Northern Sites 	     73
                                              XVI

-------
5.1      1982-1983 Hydrodynamic Model Evaluations for Surface Temperature at NDBC
        Buoys (45002 and 45007) and Subsurface Temperature at GLERL Current Meter
        Moorings (28 Instruments)  	   113

5.2      1994-1995 Hydrodynamic Model Evaluations for Surface Temperature at NDBC
        Buoys (45002, 45007, and 45010) and Subsurface Temperature at GLERL Current
        Meter Moorings (10 Instruments)	   113

5.3      Mass Budget Average Annual Results for 1994-1995 Model Simulations  	   125

6.1      Comparison of LM2-Atrazine Model to Other Models	   134
                                           XVII

-------
                                    Abbreviations
AOCs         Areas of Concern
AREAL        Atmospheric Research and Exposure Assessment Laboratory
CMAQ        Community Multiscale Air Quality
CMC          Criterion maximum concentration
CO2           Carbon dioxide
DAR          Deethylatrazine/atrazine ratio
DEA          Deethylatrazine
DIA           Deisopropylatrazine
DOC          Dissolved organic  carbon
DQOs         Data quality objectives
EMPs         Enhanced Monitoring Plans
ERS          Economic Research  Service
EU           European Union
FIFRA        Federal Insecticide, Fungicide, and Rodenticide Act
FQPA         Food Quality Protection Act
GBMBP       Green Bay Mass Balance Project
GIS           Geographical Information System
GLENDA      Great Lakes Environmental Monitoring Database
GLERL        Great Lakes Environmental Research Laboratory
GLNPO       Great Lakes National Program Office
GLWQA       Great Lakes Water Quality Agreement
GWP          Great Waters Program
HUC          Hydrological Unit Code
IDLs          Instrument detection limits
IJC           International Joint  Commission
IRED          Interim Reregistration Eligibility Decision
LaMP         Lake-wide Management Plan
LAPU         Load as a percentage of use
LLRFRB       Large Lakes and Rivers Forecasting Research Branch
LLRS         Large Lakes Research Station
LMMBP       Lake Michigan Mass Balance Project
MCL          Maximum Contaminant Level
MDEQ        Michigan Department of Environmental Quality
MDLs         Method detection limits
MED          Mid-Continent Ecology Division
MQOs        Measurement quality objectives
NDBC         National Data Buoy Center
NHEERL      National Health and Environmental Effects Research Laboratory
NOAA         National Oceanic and Atmospheric Administration
                                           XVIII

-------
ORD          Office of Research and Development
PCB          Polychlorinated biphenyl
PEM          Pesticide Emissions Model
Pis           Principal Investigators
POM          Princeton Ocean Model
QA           Quality assurance
QAPPs        Quality Assurance  Project Plans
QC           Quality control
RAP          Remedial Action Plan
RDMQ        Research Data Management and Quality Control System
RED          Reregistration Eligibility Decision
RMSD        Root mean square difference
RPD          Relative percent difference
SAP          Scientific Advisory  Panel
SBRE         Stratified Beale Ratio Estimator
SCFAH        Standing Committee on the Food Chain and Animal Health
SDLs          System detection limits
TiO2          Titanium dioxide
TMDL         Total Maximum Daily Load
USDA         United States Department of Agriculture
USDOI        United States Department of Interior
USEPA        United States Environmental Protection Agency
USFWS       United States Fish  and Wildlife Service
USGAO       United States General Accounting Office
USGS         United States Geological Survey
VWA          Volume-weighted averages
WEP          Watershed export percentage
                                            XIX

-------
                                  Acknowledgments
Special thanks to the United States Environmental Protection Agency, Great Lakes National Program Office
for leadership, support, and collaboration on the Lake Michigan Mass Balance Project.  The multiple efforts
by the Principal Investigators for providing data, necessary for the modeling, are greatly appreciated. Thank
you to  Ronald  Rossmann, Timothy Feist,  James Pauer, Xiaomi Zhang, and Amy Anstead for providing
valuable technical review comments. Thanks to Kay Morrison for the graphic renditions and figures and to
Debra L. Caudill for formatting and word processing. Finally, thanks to Paul Capel, Miriam Diamond, Kevin
Farley, Raymond Hoff, Robert Hudson,  and Barry Lesht for serving on the peer-review panel.
                                             xx

-------
                                  Executive Summary
The Lake Michigan Mass Balance Project (LMMBP) provided an opportunity to improve our understanding of
atrazine transport and fate in a large freshwater lake, Lake Michigan.  A rigorous, quality-assured large
supporting data set derived from samples collected in 1994-1995 was used to establish atmospheric and
tributary loads, estimate initial conditions, and perform model calibration and confirmation exercises. Historical
data collected outside of the LMMBP were also used to support the modeling effort.

Models developed at the United States Environmental Protection Agency's Large Lakes Research Station,
to assess atrazine transport and fate  in Lake  Michigan included MICHTOX, LM2-Toxic, and LM3-Atrazine.
Both LM2-Toxic and LM3-Atrazine utilized results from a hydrodynamic model to describe the lake's physics.
Results from air and tributary models were used to provide atrazine loadings to the lake.

Lake Michigan is acted upon by a number of physical parameters that impact the hydrology, chemistry, and
biology of the lake. For a lake the size of Lake Michigan, changes in these parameters can lead to significant
changes, especially when models are used  in  long-term predictions to predict the outcome of various
scenarios. The  primary driving forces are wind, air temperature, and precipitation.  These impact tributary
flows,  lake levels, waves, water circulation, water temperature, and ice cover. For the period of record, these
driving forces vary from year-to-year.  The period of 1982 to 1983 was used to calibrate the hydrodynamic
models. For this period of time, hydrodynamic conditions were not at any extreme. This is also true for the
period of 1994 and 1995 when the models were applied.

Temperature will impact contaminant modeling. Air temperature impacts how quickly the lake warms in any
one year. Water temperature impacts the volatilization of contaminants.  There appears to be a four-year cycle
of quicker warming which exists within a trend  of general warming of the lake. The trend of warming may be
part of a longer term, undocumented cycle, or may be related to climate change.

MICHTOX is a toxic chemical mass balance and food chain bioaccumulation model  developed  in the early
1990s. The model has nine water segments encompassing both Lake Michigan and Green Bay and is derived
from the general water quality model WASP4. Before the onset of the LMM BP, MICHTOX was applied to Lake
Michigan in a hindcast mode to gain an initial understanding of key atrazine processes in the lake and
controlling loads. Tributary loadings of atrazine to the lake were determined based on historical usage of the
chemical in the basin and a literature-derived Watershed Export Percentage (WEP) of 0.6%. The processes
modeled included advection, dispersion, and reaction (decay). MICHTOX was  used to provide a screening-
level analysis of the potential future trends  in atrazine concentrations  in lake water under a variety of
contaminant load scenarios. MICHTOX was run for seven scenarios to help evaluate the impacts on atrazine
trends caused by various loading sources and boundary conditions. Results using the assumption of average
boundary conditions indicate that atrazine decays at a rate of approximately 0.01/yr. This represents a half-life
of atrazine in the lake due to decay of 69.3 years.  MICHTOX modeling indicates that a total loading reduction
of approximately 37%, if implemented on January 1,2005, would be needed to keep concentrations in the lake
near steady-state.
                                              XXI

-------
LM2-Toxic is a sophisticated and state-of-the-art toxic chemical fate and transport model for Lake Michigan.
LM2-Toxic is also a revision of the USEPA-supported WASP4 water quality modeling framework.  The
processes modeled included advection, dispersion, decay, absorption, and volatilization. The transport fields
that were output from the 19-layered 5 km x 5 km gridded Princeton Ocean Model for the Great Lakes
(POMGL) were aggregated and used by LM2-Toxic.  The results of the LM2-Toxic system mass balance
model show that the largest atrazine load to the  lake is from the watershed.  For the year 1994, it was
estimated that 5,264 kg of atrazine were discharged to the lake via the tributaries.  The second major load to
the lake was from atmospheric wet deposition with a loading estimate of 2,493 kg.  The greatest loss of
atrazine from the lake was through transport to Lake Huron (2,546 kg) via the Straits of Mackinac. Loss due
to internal decay (1,662 kg) was the second largest loss mechanism. The total inventory of atrazine in the lake
was determined to be  184,310  kg in 1994. In this large, cold northern lake, the model suggests that in situ
atrazine decay is very slow (0.009/year). This translates into an estimated atrazine half-life of 77 years. Using
the model to forecast alternative futures, a 35% load reduction, if implemented in January 1, 2005, would have
been needed in order to prevent atrazine  concentrations from increasing further in the lake. If loadings and
boundary conditions are assumed to be constant in the future, the model predicts that the lake will eventually
reach a steady-state concentration of 66 ng/L in the year 2194.

LM3-Atrazine is a high-resolution (44,042  cells and 19 sigma layers) model that provides a better description
of areas such as near and offshore zones, bays, river confluences, and the thermocline. The transport fields
are provided  by output from the Princeton Ocean hydrodynamics Model.  Our high-resolution model, LM3-
Atrazine, was primarily used to evaluate environmental exposure concentrations of atrazine in 5km x 5km
model cells receiving  loadings from the  major tributaries to the  lake. The modeled processes included
advection, dispersion,  decay, absorption, and volatilization.  The atrazine decay (0.009/year) used in LM3-
Atrazine was  taken from the results  derived from the hindcast run using LM2-Toxic.

The model segment receiving loads from the St. Joseph  River, associated with the largest tributary load of
atrazine to the lake, ranged from winter concentrations of 37 ng/L to spring peaks of 100-350 ng/L.  These
predicted exposure concentrations in the lake are all below selected toxicological endpoints, including the most
sensitive, phytoplankton primary production reduction.

In comparing the results from the three LMMBP atrazine models to other models in the literature, it is apparent
that atrazine decays very slowly in large lakes that stratify in the summer months. The literature suggests that
degradation of atrazine in  small lakes and streams that are well-mixed  can be  significant. A hypothesis can
be formulated that the decay in surface water is likely to be dominated by photolytic processes, either directly
or indirectly.  In lakes that stratify in the summer, atrazine in the hypolimnion is isolated from the intense solar
radiation during the peak time of the year.  Hence, atrazine in this layer of the lake receives little degradation.

The LMMBP atrazine models differ from two other atrazine models recently applied to Lake Michigan.  The
main reason for the differences appears to be based  on how they estimated tributary  loadings - both used
higher estimates of tributary loadings. Consequently, these other models predicted much faster in situ decay.
Since tributary loadings are the major source atrazine to the lake, detailed assessments of these loads is very
important.
                                               XXII

-------
                                          PART1
                                    INTRODUCTION
Chapter 1.  Project Overview

Harry B. McCarty, Ken Miller, Robert N. Brent, and
Judy Schofield
DynCorp (a CSC Company)
601 Stevenson Avene
Alexandria, Virginia 22304
and
Ronald Rossmann and Kenneth R. Rygwelski
United States Environmental Protection Agency
Office of Research and Development
National Health and Environmental Effects
  Research Laboratory
Mid-Continent Ecology Division
Large Lakes and Rivers Forecasting  Research
  Branch
Large Lakes Research Station
9311 Groh Road
Grosse lie,  Michigan 48138

The   United   States  Environmental  Protection
Agency's (USEPA) Great Lakes National  Program
Office (GLNPO) instituted the Lake Michigan Mass
Balance Project (LMMBP) to measure and model the
concentrations of representative pollutants within
important  compartments  of the Lake Michigan
ecosystem.   For the  LMMBP,  concentrations of
polychlorinated biphenyls (PCBs), frans-nonachlor,
and  mercury were  measured  in tributaries,  lake
water, sediments, food webs, and the atmosphere
surrounding Lake Michigan. Atrazine was measured
only  in the tributaries, lake water, and atmospheric
components. This chapter provides an overview of
the entire LMMBP.  It includes a summary of the
parameters measured and identifies the participants.
Some of the data handling procedures are covered,
as well as a listing of various project reports.

1.1.1  Background

The Great Lakes, which contain 20% of the world's
freshwater, are a globally important natural resource
currently threatened by multiple stressors.  While
significant progress has been made to improve the
quality of the lakes, pollutant loads from point, non-
point, atmospheric,  and legacy sources continue to
impair ecosystem functions and limit the attainability
of designated uses  of these resources.   Fish
consumption advisories and beach closings continue
to be issued,  emphasizing  the  human  health
concerns from lake contamination.  Physical  and
biological stressors, such as invasion of non-native
species and habitat loss, also continue to threaten
the biological integrity of the Great Lakes.

The United States and Canada have recognized the
significance and importance of the Great Lakes as a
natural resource and have taken steps to restore and
protect the lakes.  In 1978,  both countries signed the
Great Lakes Water Quality Agreement (GLWQA).
This   Agreement  calls for   the  restoration  and
maintenance of the chemical, physical, and biological
integrity of the Great  Lakes by developing plans to
monitor and limit pollutant flows into the lakes.

The GLWQA,  as well as Section 118(c) of the Clean
Water Act, require the development of a Lake-wide
Management Plan (LaMP) for each Great Lake.  The
purpose of these LaMPs is to document an approach
to reduce  inputs  of critical pollutants  to the Great
Lakes  and restore  and  maintain  Great  Lakes

-------
integrity.  To assist in developing these LaMPs and
to monitor progress in pollutant reduction, federal,
state,  tribal,  and  local  entities  have  instituted
Enhanced Monitoring Plans (EMPs). Monitoring is
essential to the development of baseline conditions
for the Great Lakes and provides a sound scientific
base  of  information to  guide  future toxic load
reduction efforts.

The LMMBP is a part of the EMPs for Lake Michigan.
The LMMBP was 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
mass  balance  modeling approach provides the
predictive ability to determine the environmental
benefits of specific load reduction scenarios for toxic
substances and the time required to realize those
benefits.   This predictive ability will allow federal,
state,  tribal,  and  local agencies to  make  more
informed load reduction decisions.

1.1.2  Description

The LMMBP used a mass balance approach  to
evaluate  the  sources,  transport,  and  fate  of
contaminants in  the Lake Michigan ecosystem.  A
mass  balance approach is  based on  the law  of
conservation of mass, which states that the amount
of a pollutant entering  a  system is  equal to the
amount of that  pollutant leaving, trapped in, and
chemically changed in the system (Figure 1.1.1).  In
the Lake Michigan system, pollutant inputs  may
come from atmospheric sources, adjacent lakes,  or
tributary loads.

Pollutants may leave  the system through burial  in
bottom sediments, volatilization to the atmosphere, or
discharge into Lake Huron through the Straits  of
               Simple Mass Budget for Conservative Substances
                                    source
                 mass ;
                       in
                               water system
mass out ~ mass jn + ^sources

                                  1  source
               Mass Balance Modeling Approach
rr
air system ^
f !
ass in ^
^
! t
water system
1 i
1 1
sediment system

mass OU|
_ air sources
= mass jn + ^sources
^
± air-water exchange
± sediment-water exchange
± ^internal processes

Figure 1.1.1.  Simplified mass balance approach.

-------
Mackinac.  The relative magnitude  of these loss
mechanisms is,  in part, due  to the physical and
chemical properties of the chemicals being modeled.
Pollutants within the  system may be transformed
through  degradation  or  stored   in  ecosystem
compartments such as the water column, sediments,
or biota.

For the  LMMBP,  contaminant  concentrations in
various inputs  and ecosystem compartments over
spatial  and  temporal  scales  were  measured.
Mathematical models that track the transport and fate
of  contaminants  within   Lake   Michigan  were
developed and calibrated using these field data. The
LMMBP models will serve as a basis for future mass
budget/mass   balance  efforts  for  the   LMMBP
contaminants and other chemicals of interest.

1.1.3  Scope

1.1.3.1  Modeled Pollutants

When  the  USEPA published  the Water Quality
Guidance for  the  Great  Lakes  System (58 FR
20802), the Agency established water quality criteria
for 29  pollutants.  Those criteria were designed to
protect aquatic life, terrestrial wildlife, and  human
health.   PCBs, frans-nonachlor, and mercury are
included in the list of 29 pollutants. The water quality
criteria and values proposed in the guidance apply to
all of the ambient waters of the Great Lakes system,
regardless  of  the  sources  of pollutants in  those
waters.  The proposed criteria  provide a uniform
basis for integrating federal, state,  and tribal efforts
to protect and restore the Great Lakes ecosystem.

The number of pollutants that can  be intensively
monitored and modeled in the Great Lakes system is
limited by the resources  available to  collect and
analyze thousands of samples, assure the quality of
the  results,  manage the  data,  and develop and
calibrate the  necessary models.  Therefore, the
LMMBP  focused  on constructing  mass balance
models for a limited group of pollutants. PCBs, trans-
nonachlor, atrazine, and mercury were selected for
inclusion in the LMMBP because  these  pollutants
currently or potentially pose a risk to aquatic and
terrestrial organisms (including humans) in the Lake
Michigan ecosystem (Table 1.1.1). These  pollutants
also were selected to cover a wide range of chemical
and physical properties and represent other classes
of compounds  which  pose current  or potential
problems.   Once a  mass  budget  for selected
pollutants is established and a mass balance model
calibrated, additional contaminants can be modeled
with  limited  data and  future  resources  can  be
devoted to activities such as emission inventories
and dispersion modeling.

1.1.3.1.1 PCBs

Polychlorinated biphenyls (PCBs) are a class of man-
made,  chlorinated, organic chemicals that  include
209 congeners,  or specific PCB compounds.  The
highly   stable,   nonflammable,  non-conductive
properties of these compounds made them useful in
a variety of products including electrical transformers
and capacitors,  plastics, rubber, paints, adhesives,
and  sealants.   PCBs  were  produced for such
industrial uses in the form of complex mixtures under
the trade name  "Aroclor" and  were commercially
available from 1930 through 1977, when the USEPA
banned their production due to environmental and
public health concerns. PCBs also may be produced
by combustion processes, including incineration, and
can be found in stack emissions and ash  from
incinerators.

Because they were found by  the USEPA in the
effluents  from one or more wastewater treatment
facilities, seven Aroclor formulations were included in
the Priority Pollutant List developed by the USEPA
Office  of Water  under  the auspices of the Clean
Water  Act. Aroclors may have entered the Great
Lakes  through  other  means,  including spills  or
improperdisposal of transformerfluids, contaminated
soils washing into the watershed, or discharges from
ships.    The  PCBs   produced  by  combustion
processes may be released to the atmosphere where
they are  transported in both vapor and  particulate
phases and  enter  the lakes  through  either  dry
deposition or precipitation events (e.g., rain).

The stability and persistence of PCBs, which made
them useful in industrial applications, have also made
these compounds ubiquitous  in  the  environment.
PCBs do not readily degrade and thus accumulate in
water bodies and aquatic sediments.  PCBs also
bioaccumulate, or build  up, in living tissues. Levels
of PCBs  in some fish from Lake Michigan  exceed

-------
Table 1.1.1.  Characteristics of the LMMBP Modeled Pollutants
Pollutant
PCBs







frans-Non-
achlor3





Sources
• Waste incinerators
(unintentional
byproducts of
combustion)
• Industrial
dischargers
• Electrical power

• Application to crops
and gardens





Uses
• Electrical
transformers and
capacitors
• Carbonless copy
paper
• Plasticizers
• Hydraulic fluids

• Pesticide on corn
and citrus crops
• Pesticide on
lawns and
gardens


Toxic Effects
• Probable human
carcinogen
• Hearing and vision
impairment
• Liver function alterations
• Reproductive impairment
and deformities in fish and
wildlife
• Probable human
carcinogen
• Nervous system effects
• Blood system effects
• Liver, kidney, heart, lung,
spleen, and adrenal gland
damage
Biocon-
centration
Factor1
1,800 to
180,000






4, 000 to
40,000





USEPA
Regulatory
Standards2
MCL = 0.5 |jg/L
CCC= 14ng/L
HH = 0.17ng/L





MCL = 2 |jg/L
CMC = 2.4 |jg/L
CCC = 4.3 ng/L
HH = 2.1 ng/L



 Atrazine
 Mercury
Application to crops
Herbicide for corn
and sorghum
production
Waste disposal
Manufacturing
processes
Energy production
Ore processing
Municipal & medical
waste incinerators
Chloralkali  factories
Fuel combustion
Battery cells
Barometers
Dental fillings
Thermometers
Switches
Fluorescent lamps
Weightless               2 to 100
Cardiovascular damage
Muscle and adrenal
degeneration
Congestion of heart,
lungs, and kidneys
Toxic to aquatic plants

Possible human            63,000 to
carcinogen                100,000
Damage to brain and
kidneys
Adverse affects on the
developing fetus, sperm,
and male reproductive
organs
MCL = 3|jg/L
CMC4 = 350
                                                                                             CCC4 = 12|jg/L
MCL = 2 |jg/L
CMC = 1.4|jg/L
CCC = 0.77 |jg/L
HH = 50 ng/L
FWA5 = 2.4 |jg/L
FWC5= 12 ng/L
Wildlife6 = 1.3
ng/L
1From: U.S. Environmental Protection Agency, 1995a, National Primary Drinking Water Regulations, Contaminant Specific
 Fact Sheets, Inorganic Chemicals, Technical Version, EPA 811/F-95/002-T, USEPA, Office of Water, Washington, D.C.;
 and U.S. Environmental Protection Agency, 1995b, National Primary Drinking Water Regulations, Contaminant Specific
  Fact Sheets, Synthetic Organic Chemicals, Technical Version, EPA 811/F-95/003-T, USEPA, Office of Water,
  Washington, D.C.

2MCL = Maximum Contaminant Level for drinking water. CMC = Criterion Maximum Concentration for protection of aquatic
 life from acute toxicity. CCC = Criterion Continuous Concentration for protection of aquatic life from chronic toxicity. HH
 = water quality criteria for protection of human health from water and fish consumption. Data from:  U.S. Environmental
 Protection Agency, 1999, National Recommended Water Quality Criteria-Correction, EPA 822/Z-99/001, USEPA, Office
  of Water, Washington, D.C.

Characteristics presented are for chlordane. frans-Nonachlor is a principal component of the pesticide chlordane.
4Draft water quality criteria for protection of aquatic life.  From: U.S. Environmental Protection Agency, 2001 b, Ambient
  Aquatic Life Water Quality Criteria for Atrazine, USEPA, Office of Water, Washington, D.C.

5FWA = Freshwater acute water quality criterion. FWC = Freshwater chronic water quality criterion. From National Toxics
 Rule (58 FR 60848).

6Wildlife criterion. From the Stay of Federal Water Quality Criteria for Metals (60 FR 22208), 40 CFR 131.36 and the
 Water Quality Guidance for the Great Lakes System (40 CFR  132).

-------
the U.S. Food and Drug Administration tolerances,
prompting closure of some commercial fisheries and
issuance of fish consumption advisories. PCBs are
a probable human carcinogen,  and human health
effects of  PCB exposure include stomach, kidney,
and liver damage; liver and biliary tract cancer; and
reproductive effects,  including effects on the fetus
after exposure of the  mother.

PCB congeners exhibit a wide range of physical and
chemical   properties   (e.g.  vapor   pressures,
solubilities, boiling points), are relatively resistant to
degradation, and are ubiquitous. These properties
make them  ideal surrogates for a wide range of
organic compounds from anthropogenic sources.

1.1.3.1.2 trans-Nonachlor

frans-Nonachlor is a component  of the pesticide
chlordane.   Chlordane is a  mixture of chlorinated
hydrocarbons that was manufactured and used as a
pesticide  from 1948 to 1988.    Prior to  1983,
approximately 3.6 million pounds of chlordane were
used annually in the United States.  In 1988, the
USEPA banned all production and  use of chlordane
in the United States.

Like PCBs,  chlordane is relatively persistent and
bioaccumulative.   frans-Nonachlor is the  most
bioaccumulative of the chlordanes and is a probable
human carcinogen.  Other  human health effects
include  neurological  effects,  blood  dyscrasia,
hepatoxicity, immunotoxicity, and endocrine system
disruption.

Historically, frans-nonachlor may have entered the
Great Lakes through a variety of means related to
the application of chlordane, including improper or
indiscriminate  application, improper cleaning and
disposal  of pesticide  application equipment,  or
contaminated soil washing into the watershed.  In the
LMMBP, frans-nonachlor served as a model for the
cyclodiene pesticides.

1.1.3.1.3 Atrazine

Atrazine  is a  triazine herbicide based on  a ring
structure with  three carbon atoms alternating with
three nitrogen atoms. Atrazine is the  most widely
used herbicide in the United States for corn and
sorghum production.  Atrazine has been used as an
agricultural herbicide since 1959, and 64 to 75 million
pounds of atrazine are used annually in the United
States.  Atrazine is extensively used in the  upper
Midwest, including the Lake  Michigan  watershed,
where it is primarily associated with corn crops.

Unlike  PCBs and frans-nonachlor, atrazine  is  not
bioaccumulative.   It can  be persistent in water;
however,  it   is   moderately   susceptible   to
biodegradation in soils with a half-life of about 60-150
days. Atrazine  rarely exceeds the 3 ppb maximum
contaminant level (MCL) set by the  USEPA as a
drinking water standard, but localized  peak values
can exceed the MCL following rainfall events after
atrazine application.

On January 31,  2003, the USEPA issued an Interim
Reregistration Eligibility Decision (I RED) for atrazine.
In an October  2003 addendum to the  I RED,  the
Agency concluded that there is sufficient evidence to
formulate a hypothesis that atrazine exposure may
impact gonadal development in  amphibians,  but
there are currently insufficient data to either confirm
or refute the hypothesis.  However, in an October
2007 report to the Federal Insecticide, Fungicide, and
Rodenticide  Act  (FIFRA)  Scientific  Board,  the
Agency's  review concluded  that the  weight-of-
evidence from a literature review does not show that
atrazine  produces  consistent, reproducible effects
across the range of exposure concentrations and
amphibian species tested. Based on  available test
data, atrazine is not likely to be a human carcinogen.
The  Agency does have concern in regards to  the
potential hormonal effects observed  in  laboratory
animals  exposed  to  atrazine.    Above certain
concentration thresholds, atrazine is toxic to aquatic
plants. In the LMMBP, atrazine served as a model to
describe the  transport and fate of a water soluble
pesticide in current use.

1.1.3.1.4 Mercury

Mercury is a naturally-occurring toxic metal. Mercury
is used in battery cells, barometers, thermometers,
switches, fluorescent lamps, and as a catalyst in the
oxidation of organic compounds. Global releases of
mercury  in the  environment are both natural and
anthropogenic (caused  by human activity).   It is
estimated that about 11,000 metric tons of mercury

-------
are released annually to the air, soil, and water from
anthropogenic  sources.    These  sources  include
combustion of  various fuels such as coal;  mining,
smelting, and manufacturing activities; wastewater;
and agricultural, animal, and food wastes.

As  an  elemental  metal,  mercury  is  extremely
persistent in all  media. Mercury also bioaccumulates
with reported bioconcentration factors in fish tissues
in the range of 63,000 to 100,000.  Mercury is a
possible human carcinogen and causes the following
human  health  effects:   stomach, large  intestine,
brain, lung, and kidney damage; blood pressure and
heart rate increase;  and fetal damage.    In  the
LMMBP,   mercury   served  as   a   model  for
bioaccumulative metals.

1.1.3.2  Other Measured Parameters

In addition to  the four chemicals modeled in  the
LMMBP, many otherchemicals and parameters were
measured in the LMMBP as part  of the EMPs.  A
survey of these chemicals and parameters aids in the
understanding  of the overall ecological integrity of
Lake Michigan. These additional parameters include
various   biological   indicators;   meteorological
parameters; and organic, metal,  and conventional
chemicals in Lake Michigan. Many  of the parameters
included in this study are provided in Table 1.1.2.
Table 1.1.2.  The LMMBP Parameters
                                         Organics (Continued)
                    Organics
 acenaphthene
 acenaphthylene
 aldrin
 anthracene
 atrazine
 a-BHC
 (3-BHC
 5-BHC
 Y-BHC
 benzo[a]anthracene
 benzo[g,/?,/]perylene
 benzo[£>]fluoranthene
 benzo[/(]fluoranthene
 benzo[e]pyrene
 benzo[a]pyrene
 a-chlordane
p,p'-DDT
endosulfan sulfate
endosulfan I
endosulfan II
endrin
endrin aldehyde
endrin ketone
fluoranthene
fluorene
heptachlor
heptachlor epoxide
hexachlorobenzene (HCB)
indeno[1,2,3-cd]pyrene
mirex
frans-nonachlor
oxychlordane
                             benzo[a]pyrene
                             a-chlordane
                             y-chlordane
                             chrysene
                             coronene
                             p,p'-DDE
                             p,p'-DDD
                       frans-nonachlor
                       oxychlordane
                       PCBs congeners
                       phenanthrene
                       pyrene
                       retene
                       toxaphene
                                                 Metals
                             aluminum
                             arsenic
                             calcium
                             cadmium
                             chromium
                             cesium
                             copper
                             iron
                             mercury
                             potassium
                       magnesium
                       manganese
                       sodium
                       nickel
                       lead
                       selenium
                       thorium
                       titanium
                       vanadium
                       zinc
                                            Conventionals
                             alkalinity
                             ammonia
                             bromine
                             chloride
                             chlorine
                             sulfate
                             conductivity
                             dissolved organic
                               carbon
                             dissolved oxygen
                             dissolved phosphorus
                             dissolved reactive silica
                             dry weight fraction
                             element carbon
                             nitrate
                       particulate organic carbon
                       percent moisture
                       PH
                       phosphorus
                       silica
                       silicon
                       temperature
                       total Kjeldahl nitrogen
                       total organic carbon
                       total phosphorus
                       total suspended
                          particulates
                       orffto-phosphorus
                       total hardness
                       turbidity
                 Biologicals
fish species
fish age
fish maturity
chlorophyll a
fish lipid amount
zooplankton
fish weight
fish length
fish taxonomy
fish diet analysis
primary productivity
                Meteorological
air temperature
relative humidity
barometric pressure
weather conditions
wind direction
wind speed
visibility
wave height and direction

-------
1.1.3.3 Measured Compartments

In the LMMBP, contaminants were measured in the
following compartments:

•  Open-Lake Water Column: The water column in
  the open-lake was sampled and analyzed for the
  modeled pollutants.

•  Tributaries: Major tributaries were sampled and
  analyzed for the modeled pollutants.

•  Fish: Top predators and forage base species
  were sampled and analyzed for diet analysis and
  contaminant burden.

•  Lower Pelagic Food Chain: Phytoplankton and
  zooplankton were sampled and  analyzed  for
  species  diversity, taxonomy,  and contaminant
  burden.
•  Sediments:  Cores  were  collected  and  trap
  devices  were  used  to  collect  sediment for
  determination of contaminants and sedimentation
  rates.

•  Atmosphere: Vapor, particulate, and precipitation
  phase samples were collected  and analyzed for
  the modeled pollutants.

For the  modeled  pollutants,  more than  20,000
samples were collected at more than 300 sampling
locations and analyzed, including  more than 9,000
quality control (QC) samples (Figure 1.1.2).  Field
data collection activities were initially envisioned as
a one-year effort. However, it became evident early
into the project that a longer collection period would
be necessary to provide a full year of concurrent
                            Manistique River,
                                          uskegon River
                                          Grand River
                                                 O water survey stations
                                                 , ;j=atmospheric monitoring
                                                 D tributary monitoring
                                                   stations
Figure 1.1.2.  The LMMBP sampling locations.
                                               7

-------
information on  contaminant loads and ambient
concentrations for modeling purposes.  Therefore,
field sampling occurred from April 1994 to October
1995.

1.1.4  Objectives

The goal of the LMMBP was to develop a sound,
scientific base of information to guide future toxic
load reduction efforts at the federal, state, tribal, and
local levels.  To meet this goal,  the four following
LMMBP objectives were developed:

>• Estimate pollutant loading rates: Environmental
  sampling of major media will allow estimation of
  relative loading rates of critical pollutants to  the
  Lake Michigan basin.

>• Establish baseline: Environmental sampling and
  estimated loading rates will establish a baseline
  against which future progress  and contaminant
  reductions can be gauged.

- Predict  benefits   associated  with   load
  reductions: The completed mass balance model
  will provide a predictive tool that environmental
  decision-makers and  managers  may  use  to
  evaluate the  benefits of specific load reduction
  scenarios.

>• Understand ecosystem dynamics: Information
  from  the  extensive LMMBP  monitoring  and
  modeling  efforts  will  improve  our  scientific
  understanding of the environmental processes
  governing contaminant cycling and  availability
  within relatively closed ecosystems.

1.1.5  Design

1.1.5.1  Organization

The GLNPO proposed a mass balance approach to
provide  coherent,  ecosystem-based evaluation of
toxics in Lake Michigan.   GLNPO served as  the
program sponsor for the LMMBP. GLNPO formed
two committees to coordinate study planning,  the
Program Steering  Committee and  the  Technical
Coordinating Committee.   These committees were
comprised  of  federal,   state,   and   academic
laboratories as well as commercial  laboratories (see
Section 1.1.5.2, Study Participants). The committees
administered  a wide  variety  of tasks  including:
planning the project, locating the funding, designing
the sample collection, coordinating sample collection
activities, locating qualified laboratories, coordinating
analytical activities, assembling the  data, assuring
the quality of the data, assembling skilled modelers,
developing the models, and communicating interim
and final project results. The Mid-Continent Ecology
Division  (MED) at Duluth, in  cooperation with the
National Oceanic and  Atmospheric  Administration
(NOAA)   Great  Lakes  Environmental  Research
Laboratory (GLERL) and  the Atmospheric Sciences
Modeling   Division,   supported  the  modeling
component of the mass balance study by developing
a  suite  of integrated  mass balance models  to
simulate the transport, fate, and bioaccumulation of
the study target analytes.

1.1.5.2 Study Participants

The LMMBP was a coordinated effort among federal,
state, and  academic  scientists; and commercial
laboratories.     The  following  agencies   and
organizations  have  all played roles in ensuring the
success of the  LMMBP.   Except  for  the  three
organizations indicated with an asterisk (*), all of the
participants were members of the LMMBP Steering
Committee.

Federal and International

••  USEPA/GLNPO (Program Sponsor)
»  USEPA/Region V Water Division (WD)
»•  USEPA/Region V Air Division
-  USEPA/ORD/NHEERL/MED/LLRFRB
>•  ORD/National Exposure Research Laboratory
••  U.S.  Department of the Interior  (USDOI) U.S.
   Geological  Survey  (USGS)  Water Resources
   Division (WRD)
••  USDOI/USGS Biological Resources DivisionGreat
   Lakes Science Center  (GLSC)
-  U.S. Fish and Wildlife Service (USFWS)
>•  U.S. Department of Energy
••  U.S. Department of Commerce NOAA/GLERL
••  USEPA/Office  of Air and Radiation*
-  USEPA/Office  of Water*
>•  Environment Canada*
•*  U.S. Department of Energy Battelle NW
                                              8

-------
State

*  Illinois Department of Natural Resources
>•  Illinois Water Survey
>•  Indiana   Department   of   Environmental
   Management
>•  Michigan Department of Natural Resources
>•  Michigan Department of Environmental  Quality
   (MDEQ)
>•  Wisconsin Department of Natural Resources
>•  Wsconsin State Lab of Hygiene

Academic and Commercial

*  Indiana University
>•  Rutgers University
>•  University of Maryland
>•  University of Michigan
>•  University of Minnesota
>•  University of Wisconsin
>•  Grace Analytical

1.1.5.3 Workgroups

Eleven workgroups were formed to provide oversight
and management of specific project elements. The
workgroups facilitated planning and implementation
of the study in a coordinated and systematic fashion.
The  workgroups  communicated  regularly through
participation in monthly conference calls and annual
"all-hands"  meetings.   Workgroup  chairs were
selected and were responsible for managing tasks
under  the  purview  of  the   workgroup  and
communicating the  status of activities to other
workgroups. The workgroups and workgroup chairs
are listed below.

•  Program Steering  Committee - Paul Horvatin
   (USEPA/GLNPO)
•  Technical Coordinating Committee-Paul Horvatin
   (USEPA/GLNPO)
•  Modeling  Workgroup   -   William  Richardson
   (USEPA/ORD/NHEERL/MED/LLRFRB)
•  Air Monitoring Workgroup-Jackie Bode (USEPA/
   GLNPO)
•  Biota Workgroup  - Paul  Bertram  (USEPA/
   GLNPO) and   John Gannon (USDOI/USGS/
   GLSC)
•  Chemistry Workgroup-David Anderson (USEPA/
   GLNPO)
• Data Management Workgroup - Kenneth Klewin
  and Philip Strobel (USEPA/GLNPO)
• Lake  Monitoring Workgroup -  Glenn Warren
  (USEPA/GLNPO)
• Tributary Monitoring Workgroup - Gary Kohlhepp
  (USEPA/Region V/WD) and Robert Day (MDEQ)
• Quality Assurance Workgroup - Louis Blume and
  Michael Papp (USEPA/GLNPO)
• Sediment Monitoring Workgroup - Brian  Eadie
  (NOAA/GLERL)

1.1.5.4 Information Management

As program sponsor, GLNPO managed information
collected during the LMMBP.  Principal Investigators
(Pis) participating in the study reported field and
analytical data to GLNPO. GLNPO developed a data
standard for reporting field and analytical data and a
database  for storing  and  retrieving study data.
GLNPO was also responsible  for conducting data
verification activities and releasing verified data to the
study modelers  and  the public.   The flow  of
information is illustrated in Figure 1.1.3.

1.1.5.4.1  Data Reporting

Over  20  organizations  produced   LMMBP  data
through the collection and analysis of  more than
20,000 samples.  In the interest of standardization,
specific formats  (i.e.,  file formats  and  codes  to
represent certain data values) were established for
reporting the LMMBP data.   Each format specified
the  "rules"  by which  data were submitted, and, in
many cases, the allowable values by which they were
to be reported.   The data reporting formats were
designed to minimize the number of data elements
reported from the field crews and laboratory analysis.
Data reporting formats and the resulting Great Lakes
Environmental Monitoring Database (GLENDA, see
Section! 1.5.4.2) were designed to be applicable to
projects outside the LMMBP as well.

Principal  Investigators  (Pis) (including  sampling
crews and  the  analytical  laboratories)  supplied
sample collection and  analysis data following the
standardized reporting formats, if  possible.  The
LMMBP  data were  then  processed through an
automated  SAS-based  data verification system,
Research  Data  Management and Quality Control
System (RDMQ), for quality assurance (QA)/QC

-------
Prindpri
Cafcel and AmlyiB
I
Ripert Fltld and
AnsJyfeul Data
(ftmrdTng to LU MB Data
Slandird}



*
No
_T

*C^_ fl|l|>rOW«1 ^^S
^ _


















V















OLNPOData
Monttgienwnt


Roogtm, Stora, ind
Tnnmll bk







' ^~''*x -^~-^^__

•

QLENDA


pjlfll HtquKl
data]
1













•^^










QLNPOQA
Workgroup


Coixs uot Data
VsrMmtkn (Uarga Ffrid
and And^tfczl Data L^qg
RDMQ)
J[


^V^^
No
*
Prodint Flnl Vtrilltd
MM Flltuid Pravld* le
Approve]

Pr4d»»Fln*!Vii1hd
Dan Fll* andTnnunll
(In QLENDA-curnpelbh
Foirnat)

































LMMB Study














Input Data, to Sludy
~* MedMi






ExttrraU Pmrtn




















f























J
Figure 1.1.3.  Flow of information in the LMMBP.
checking. After verification and validation by the Pis,
the data  sets were  output in a form  specific for
upload to GLENDA.  Finally, these data sets were
uploaded to GLENDA.

1.1.5.4.2  Great Lakes Environmental Monitoring
Database

Central  to  the  data  management  effort  is  a
computerized database system to house LMMBP
and other project results.  That system, GLENDA,
was  developed to  provide data  entry,  storage,
access, and analysis capabilities to meet the needs
of mass balance modelers and other potential users
of Great Lakes data.
Development  of GLENDA began in  1993 with  a
logical model  based on the  modernized STORE!
concept  and requirements analysis. GLENDA was
developed with the following guiding principles:

•  True multi-media scope: Water, air, sediment,
  taxonomy, fish tissue, fish diet, and meteorology
  data can  all be housed in the database.

•  Data  of  documented quality:  Data quality  is
  documented by including results of quality control
  parameters.

•  Extensive contextual indicators:  Ensure data
  longevity by including enough information to allow
  future or secondary users to make use of the data.
                                             10

-------
•  Flexible and expandable:  Database is able to
   accept data from any  Great Lakes monitoring
   project.

•  National compatibility: GLENDA is compatible
   with STORE! and allows ease of transfer between
   these large databases.

In an effort to reduce the data administration burden
and ensure consistency of data in this database,
GLNPO developed  several  key tools. Features
including standard data definitions, reference tables,
standard automated data entry applications, and
analytical tools are (or will soon  be) available.

1.1.5.4.3 Public Access to LMMBP Data

All LMMBP data that have been verified (through the
QC process) and validated (accepted by the PI) are
available to the public. Currently, GLNPO requires
that written requests be made to obtain the LMMBP
data.  The data sets are available in several formats
including WK1, DBF, and SD2.  More  information
about the data sets is available on the LMMBP web
site   at:      http://www.epa.gov/glnpo/lmmb/
database.html.

The primary reason for requiring an official request
form  for the LMMBP data  is  to keep  track of
requests.  This allows GLNPO  to know how  many
requests have been made, who has requested data,
and  what use  they  intend  for the data.    This
information  assists  GLNPO   in managing  and
providing public access to Great Lakes data and
conducting  public  outreach  activities.   As  of
November 2000, 38 requests for the  LMMBP data
have been made: eight from USEPA, five from other
federal agencies, five from state agencies, five from
universities,  ten  from consultants,   three   from
international agencies, and two from non-profit or
other groups.  In the future, after all data are verified
and validated, GLNPO intends to make condensed
versions of the data sets available on the LMMBP
web site for downloading. This will allow easy public
access to the LMMBP data.

Further information on the  information management
for the LMMBP can be found in The Lake Michigan
Mass Balance Study Quality Assurance Report (U.S.
Environmental  Protection Agency, 2001 a).
1.1.5.5 Quality Assurance Program

At the outset of the LMMBP, managers recognized
that the data gathered and the models developed
from the  study  would  be used  extensively by
decision-makers   responsible   for  making
environmental,  economic,  and  policy decisions.
Environmental measurements are never true values
and  always contain some level  of uncertainty.
Decision-makers, therefore, must recognize and be
sufficiently  comfortable   with  the   uncertainty
associated with data on  which their decisions are
based.    In recognition  of this  requirement, the
LMMBP managers established a QA program goal of
ensuring that data produced under the LMMBP would
meet defined standards of quality with a specified
level of confidence.

The QA program prescribed minimum  standards to
which all organizations collecting data were required
to adhere. Data quality was defined, controlled, and
assessed  through activities  implemented  within
various parameter groups (e.g., organic, inorganic,
and biological parameters). QA  activities included
the following:

>•  QA Program:  Prior to  initiating data collection
   activities, plans were developed, discussed, and
   refined  to ensure that  study objectives were
   adequately defined and to  ensure that  all QA
   activities necessary to meet study objectives were
   considered and implemented.

••  QA  Workgroup:   USEPA established  a QA
   Workgroup whose primary function was to ensure
   that the overall QA goals of the study were met.

••  QA Project Plans (QAPPs): USEPA worked with
   Pis to  define  program  objectives,  data  quality
   objectives (DQOs), and measurement  quality
   objectives (MQOs) for use  in  preparing Quality
   Assurance Project Plans (QAPPs). Pis submitted
   QAPPs to the USEPA for review and approval.
   USEPA reviewed each  QAPP for  required QA
   elements and soundness of planned QA activities.

>•  Training:   Before beginning  data collection
   activities, Pis  conducted training  sessions  to
   ensure that individuals working on the project were
   capable of properly performing data collection
   activities for the LMMBP.
                                              11

-------
»  Monthly   Conference   Calls   and   Annual
   Meetings: USEPA, Pis, and support contractors
   participated in monthly  conference  calls and
   annual  meetings  to discuss project status and
   objectives, QA issues, data reporting issues, and
   project schedules.

>•  Standardized Data Reporting Format: Pis were
   required to submit all data in a standardized data
   reporting  format that  was designed  to ensure
   consistency  in  reporting and  facilitate  data
   verification,  data  validation,   and  database
   development.

>•  Intercomparison Studies:   USEPA conducted
   studies  to compare performance among different
   Pis analyzing similar samples. The studies were
   used to evaluate the comparability and accuracy
   of program data.

>•  Technical Systems Audits:  During the  study,
   USEPA formally audited each Pi's laboratory for
   compliance with their  QAPPs, the overall study
   objectives, and pre-determined standards of good
   laboratory practice.

>•  Data Verification: Pis and the USEPA evaluated
   project  data against pre-determined MQOs and
   DQOs to ensure that only data of acceptable
   quality  would  be  included  in  the  program
   database.

>•  Statistical  Assessments:     USEPA   made
   statistical  assessments  of the LMMBP data to
   estimate  elements  of   precision,  bias,  and
   uncertainty.

>•  Data Validation: USEPA and modelers evaluated
   the data against the model objectives.

Comparability of data among Pis participating in the
LMMBP was deemed to be important for successful
completion of the study.   Therefore, MQOs for
several data attributes were developed by the Pis
and defined in the QAPPs.  MQOs were designed to
control various phases of the measurement process
and to ensure that the total measurement uncertainty
was  within the  ranges prescribed  by the DQOs.
MQOs were defined  in terms of six attributes:
>•  Sensitivity/Detectability:  The determination of
   the low-range critical value that a method-specific
   procedure  can  reliably  discern  for  a  given
   pollutant.  Sensitivity measures included, among
   others, method detection limits (MDLs) as defined
   at 40  CFR Part 136, system  detection  limits
   (SDLs), or instrument detection limits (IDLs).

>•  Precision:  A measure of the degree to which
   data  generated  from  replicate  or  repetitive
   measurements differ from one another. Analysis
   of  duplicate samples  was  used  to assess
   precision.

>•  Bias:   The degree of agreement  between  a
   measured and actual value.  Bias was expressed
   in terms  of the recovery  of  an  appropriate
   standard reference material or spiked sample.

>•  Completeness:  The measure of the number of
   samples successfully  analyzed and  reported
   compared to the number that were scheduled to
   be collected.

>•  Comparability:  The confidence with which one
   data set can be compared to  other data sets.

>•  Representativeness: The degree to which data
   accurately and precisely represent characteristics
   of  a  population,  parameter  variations  at  a
   sampling  point,  a   process  condition,  or an
   environmental condition.

The Pi-defined MQOs  also were used as the basis
for the data verification process.  GLNPO conducted
data verification through the LMMBP QA Workgroup.
The  workgroup  was  chaired  by  GLNPO's  QA
Manager and consisted of QC Coordinators that were
responsible for conducting review of specific data
sets. Data verification was performed by comparing
all field and QC sample results produced by each PI
with their MQOs and with overall LMMBP objectives.
If a result failed to meet predefined criteria, the QC
Coordinator contacted the PI to discuss the result,
verify that it was correctly reported, and  determine if
corrective actions  were feasible.  If the result was
correctly reported  and corrective actions were not
feasible, the results were flagged to inform  data
users of the failure. These flags were not intended to
suggest that data were not useable; rather they were
intended to caution the user about an aspect of the
                                              12

-------
data that did not meet the predefined criteria. Data
that met all predefined requirements were flagged to
indicate that the results had been verified and were
determined to meet applicable MQOs.  In this way,
every data point was assigned one or more validity
flags based on the results of the QC checks.  GLNPO
also derived data  quality assessments  for each
LMMBP data set for a subset of the attributes listed
above,  specifically sensitivity, precision, and bias.
The LMMBP  modelers  and the LLRS Database
Manager also performed data quality assessments
prior to inputting data into study models.  Such
activities  included  verifying  the  readability  of
electronic files,  identifying missing data,  checking
units, and identifying outliers. A detailed description
of the QA program is included in The Lake Michigan
Mass  Balance Project Quality Assurance Report
(U.S. Environmental Protection Agency, 2001 a). A
brief summary of  quality  implementation  and
assessment is provided in each of the following parts.

1.1.6  Project Documents and Products

During   project  planning,  LMMBP   participants
developed  study tools  including work  plans, a
methods compendium, QAPPs, and data reporting
standards. Through these tools, LMMBP participants
documented many  aspects of the study including
information management and QA procedures. Many
of these documents  are  available  on GLNPO's
website at http://www.epa.gov/glnpo/lmmb.

The LMMBP Work Plan

Designers of the LMMBP have documented their
approach in a  report entitled Lake Michigan Mass
Budget/Mass  Balance  Work   Plan  (U.S.
Environmental  Protection  Agency,  1997a).  The
essential elements of a mass balance study and the
approach  used to  measure  and  model these
elements in the Lake Michigan system are described
in the work plan.  This  document was developed
based  upon the efforts of many federal and state
scientists and  staff who participated in the initial
planning workshop,  as well as Pis.

QA Program/Project Plans

The Lake Michigan Mass Balance Project: Quality
Assurance Plan for Mathematical Modeling, Version
3.0 (Richardson et a/., 2004) documents the QA
process for  the development  and application of
LMMBP models, including hydrodynamic, sediment
transport, eutrophication, transport chemicalfate, and
food chain bioaccumulation models.

The Enhanced Monitoring Program QA Program
Plan

The  Enhanced   Monitoring  Program  Quality
Assurance  Program  Plan  (U.S.  Environmental
Protection Agency, 1997b) was developed in 1993 to
ensure  that data  generated  from the  LMMBP
supported its intended use.

The LMMBP Methods Compendium

The Lake Michigan Mass Balance  Project Methods
Compendium  (U.S.   Environmental   Protection
Agency, 1997c, 1997d) describes the sampling and
analytical methods used in the LMMBP.  The entire
three volumes are available  on GLNPO's website
mentioned above.

The LMMBP Data Reporting Formats and  Data
Administration Plan

Data management for the  LMMBP was a focus from
the  planning  stage  through  data   collection,
verification, validation,  reporting, and archiving.  The
goal of consistent and  compatible data was a key to
the success of the  project. The  goal  was   met
primarily through  the development of standard
formats for reporting environmental data. The  data
management philosophy is outlined on the LMMBP
website mentioned above.

Lake Michigan LaMP

"Annex  2" of the  1972 Canadian-American Great
Lakes Water Quality Agreement (amended in 1978,
1983, and  1987) prompted development of a Lake-
wide Area Management Plan  (LaMP) for each Great
Lake. The purpose of  these LaMPs is to document
an approach to reducing input of critical pollutants to
the Great Lakes and restoring and maintaining Great
Lakes integrity. The Lake Michigan LaMP calls for
basin-wide management of toxic chemicals.
                                             13

-------
GLENDA Database

Central  to  the  data  management effort  is  a
computerized  data system to house LMMBP and
other project results.  That system, the Great Lakes
Environmental Monitoring Database (GLENDA), was
developed to provide data entry, storage, access,
and analysis capabilities to meet the needs of mass
balance modelers and other potential users of Great
Lakes data.

References

Richardson, W.L., D.D. Endicott,  R.G. Kreis, Jr., and
   K.R. Rygwelski (Eds.). 2004.  The Lake Michigan
   Mass Balance Project Quality Assurance Plan for
   Mathematical Modeling.    Prepared  by  the
   Modeling  Workgroup.    U.S.   Environmental
   Protection  Agency,  Office  of  Research  and
   Development, National Health and Environmental
   Effects  Research  Laboratory,  Mid-Continent
   Ecology Division,  Large Lakes Research Station,
   Grosse  lie, Michigan.  EPA/600/R-04/018, 233
   pp.

U.S.  Environmental  Protection  Agency.   1995a.
   National Primary Drinking Water Regulations,
   Contaminant  Specific  Fact  Sheets,  Inorganic
   Chemicals,   Technical   Version.     U.S.
   Environmental Protection Agency,  Office of
   Water, Washington, D.C. EPA/811/F-95/002-T.

U.S.  Environmental  Protection  Agency.   1995b.
   National Primary Drinking Water Regulations,
   Contaminant  Specific  Fact  Sheets,  Synthetic
   Organic Chemicals,  Technical Version.   U.S.
   Environmental Protection Agency,  Office of
   Water, Washington, D.C. EPA/811/F-95/003-T.

U.S. Environmental Protection Agency. 1997a. 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.
U.S. Environmental Protection Agency.  1997b. The
   Enhanced Monitoring Program Quality Assurance
   Program  Plan.  U.S.  Environmental Protection
   Agency, Great Lakes National Program Office,
   Chicago,  Illinois.  EPA/905/R-97/017, 61 pp.

U.S. Environmental Protection Agency. 1997c.  Lake
   Michigan  Mass Balance Study (LMMB) Methods
   Compendium,  Volume 1:  Sample Collection
   Techniques.   U.S.  Environmental  Protection
   Agency, Great Lakes National Program Office,
   Chicago,  Illinois.  EPA/905/R-97/012a, 1,440pp.

U.S. Environmental Protection Agency. 1997d.  Lake
   Michigan  Mass Balance Study (LMMB) Methods
   Compendium, Volume 2: Organic and Mercury
   Sample   Analysis   Techniques.      U.S.
   Environmental Protection Agency, Great Lakes
   National  Program  Office,  Chicago,  Illinois.
   EPA/905/R-97/012b, 532 pp.

U.S.  Environmental Protection  Agency.    1999.
   National Recommended Water Quality Criteria-
   Correction.     U.S.   Environmental  Protection
   Agency,  Office  of  Water,  Washington,  D.C.
   EPA/822/Z-99/001, 25 pp.

U.S. Environmental Protection Agency. 2001a. The
   Lake  Michigan Mass Balance Study  Quality
   Assurance   Report.     U.S.   Environmental
   Protection  Agency,   Great  Lakes   National
   Program, Chicago, Illinois. EPA/905/R-01/013.

U.S.  Environmental  Protection  Agency.   2001 b.
   Ambient Aquatic Life Water Quality for Atrazine.
   U.S. Environmental  Protection Agency, Office of
   Water, Washington, D.C.  EPA/822/D-01/002,
   230 pp.
                                              14

-------
                                          PART1
                                     INTRODUCTION
Chapter 2.  General  Information  on  the
Herbicide  Atrazine and  Its  Degradation
Products

Kenneth R. Rygwelski
United States Environmental Protection Agency
Office of Research and Development
National Health and Environmental Effects
  Research Laboratory
Mid-Continent Ecology Division
Large Lakes and Rivers Forecasting Research
  Branch
Large Lakes Research Station
9311 Groh Road
Grosse lie, Michigan 48138

1.2.1 Background

Atrazine is a triazine herbicide registered to control
broadleaf  weeds  and  some  grassy  weeds by
inhibiting photosynthesis. Its primary use in the Lake
Michigan basin  is for the control of weeds in  corn
crops.  It is estimated to be the most heavily used
herbicide in  the United States.   Usage on  corn
accounts for approximately 86%  of total  United
States domestic usage, followed by sorghum at 10%,
and sugarcane at 3% (all other uses make  up the
remaining 1%).  For corn crops, it is usually applied
in the spring  prior to, during, or after planting a  crop
or after crop  emergence. The product is formulated
as an emulsifiable concentrate, flowable concentrate,
water dispersible granular  (dry flowable), soluble
concentrate,  wettable powder, granular, and as a
ready-to-use formulation.  It may be applied to the
field  with  a  groundboom sprayer, aircraft,  or by
means  of   a   tractor-drawn  spreader   (U.S.
Environmental Protection Agency, 2003a).  In a
1990-1991 period, atrazine was the single highest-
use pesticide in the Great Lakes basin (United States
only) and represented 19.4% of all pesticides used
on  agricultural  crops  (U.S.  General Accounting
Office, 1993).

Atrazine was registered in 1958 (U.S. Environmental
Protection  Agency,  2003a),  and  is  currently
undergoing a reregistration review.  Syngenta is the
primary  atrazine registrant.   Pesticides registered
prior  to  November  1984   are  subject  to  the
reregistration  process.  On January 31, 2003,  the
U.S. Environmental Protection Agency  (USEPA)
issued an Interim Reregistration Eligibility Decision
(IRED) for atrazine (U.S. Environmental Protection
Agency, 2003b).   In October 2003, the  USEPA
issued an addendum that updates the January 31,
2003 IRED (U.S. Environmental Protection Agency,
2003c). The Agency expects the registrants to adopt
the risk management measures presented in  the
IRED.   Among other  requirements,  the  IRED
mandates  the  monitoring  of  40  representative
watersheds in the  United States  to determine if
specific atrazine levels of concern are exceeded, a
testing program  to  better evaluate  potential risk to
amphibians, and measures to mitigate exposure risk
to  applicators in both residential and agricultural
settings.  Watersheds exceeding levels of concern
criteria will  be   subject to  remedies  under  the
USEPA's  Total  Maximum   Daily  Load   (TMDL)
program  requirements.    In  the  October  2003
addendum to the IRED, the Agency concluded that
there is sufficient evidence to formulate a hypothesis
that  atrazine   exposure  may  impact  gonadal
                                              15

-------
development in amphibians, but there are currently
insufficient data to confirm or refute the hypothesis.
On October 9-12, 2007,  the Federal Insecticide,
Fungicide, and Rodenticide Act (FIFRA)  Scientific
Advisory Panel  (SAP) met with  the Agency   to
evaluate the potential for atrazine to affect  the
development of amphibian species. However, in an
October 2007 report to the FIFRA Scientific Advisory
Panel, the Agency's review of the literature indicated
that  studies do  not  show that  atrazine  produces
consistent, reproducible effects across the range of
exposure  concentrations  and  amphibian species
tested. Based on available test data, atrazine is not
likely to be a human carcinogen. The Agency does
have concern in regards to the  potential  hormonal
effects observed  in laboratory animals exposed to
atrazine. A Reregistration  Eligibility Decision (RED)
was  issued for atrazine, a triazine pesticide, in April
2006. In that RED, an evaluation was performed to
determine if the cumulative effect  from the triazine
pesticides (atrazine, simazine, propazine,  and their
chlorinated  degradates)  that  share  a  common
mechanism of toxicity are below the Food Quality
Protection Act (FQPA) regulatory level - that  the
risks associated with the pesticide residues pose a
reasonable certainty of no harm.

A  comprehensive review of atrazine  toxicity to
various freshwater trophic groups was conducted by
Solomon et al. (1996).  A total of  85 species were
tested, and the order of sensitivity from most to least
sensitive  trophic   groups  was   as   follows:
phytoplankton > aquatic  macrophytes > benthos >
zooplankton > fish. Due to limited data, amphibians
were not included in this sensitivity review. Atrazine
was  found to be more inhibitory to photosynthesis
than were its transformation products. Atrazine was
seven to 10 times more inhibitory to blue-green algae
and four to six times  more inhibitory to green algae
than  the most  potent   transformation  product,
deethylatrazine (DEA).  Young fish survival may be
at risk if the atrazine exposure concentrations  are
significant   enough   to  impact   phytoplankton
populations  and  macrophytes.   Zooplankton,  an
important food source for juvenile fish, may be
depleted if the phytoplankton are reduced, and  the
juvenile fish may become easier prey if they lose the
protective cover of macrophytes.
Atrazine  is often found in surface water and is
regulated under the  Safe  Drinking Water Act.  A
Maximum Contaminant Level (MCL) of 3 ppb was
established in 1991 by the USEPA's Office of Water
(U.S.  Environmental  Protection  Agency,  1995).
Loadings associated with run-off from farm fields are
often seasonal with the spring and early summer-
time periods being the highest.  For municipalities
dependent upon drinking water supplies from rivers,
potential  exceedences of the MCL are most likely to
occur from mid-April through  mid-July in the Lake
Michigan basin when atrazine concentrations  are
most likely to be high. Using a variety of bench-scale
water treatment processes such as  coagulation,
softening, ozonation,  chlorination,  and  powdered
activated  charcoal,   researchers   had  difficulty
adequately removing atrazine from the water and
recommended that other removal processes should
be investigated (Westerhoff et al., 2005).

Atrazine  has been banned  in the European Union
(EU) since October 4, 2003  when the herbicide was
not granted re-registration.  This decision was taken
by the Standing Committee on the Food Chain and
Animal Health (SCFAH), the EU regulatory body.

Additional background information on atrazine and
access to the documents cited in this section can be
downloaded at http://www.epa.gov/pesticides/.

1.2.2    Physical-Chemical  Properties of
Atrazine

Physical  and chemical properties of atrazine  are
given in  Table  1.2.1.   Wth  a  low  Henry's  law
constant, atrazine volatilization from the lake is low.
Also, with a moderate solubility in water, run-off from
farm fields can occur, especially  in the spring after
significant rainfall and when soil moisture content is
high.  With a low octanol-water partition  coefficient
(Kow), atrazine  is not strongly sorbed to particles in
the water, and it is not bioaccumulated to any extent.
Frank et al. (1979) analyzed suspended solids from
12 streams (45 samples) in 1974 and 1976 flowing
into the  Great  Lakes from  the Canadian  side
(Ontario) and were unable to detect atrazine in these
particulates (detection limit of 0.05 ug/g).  However,
of the 92 streams sampled in 1977, they detected
atrazine in the water approximately 80% of the time.
From that study, they concluded that atrazine was in
                                               16

-------
Table 1.2.1.  Physical and Chemical Properties of
Atrazine
Empirical Formula
Chemical Name

Chemical Family

Structural Formula:
      (CH;
C8H14CIN5
2-chloro-4-ethylamino-6-
 isopropylamino-1,3,5-triazine
Triazine
                    Cl
  \
  N
(CH3)— (CH)— (NH)— C   C— (NH)— (CH2)— (CH3)
                  \\
Molecular Weight
Melting Point
Vapor Pressure
Solubility in Water
Density
Henry's Law
  Constant

Physical state
log K,,w
                    N
215.7g/mol
173°Cto 175°C
40LiPaat20°C
33 PPM at25°C
0.35 g/ml
8.1 x 10"8 (dimensionless) at
 25°C (U.S.  Department of
 Agriculture, 2001)
White crystalline solid
2.7645
the  dissolved  phase,  rather  than  attached  to
particles. Laboratory measurements of the partition
coefficients   for   atrazine,   DEA,    and
deisopropylatrazine  (DIA) resulted in the following:
1.1, 0.4, and 0.3 (ng/g)/(ng/ml), respectively.  The
particulate substrate was Eudora Silt Loam with a
1.0% carbon content. These results indicated that
the two degradation  products are even more soluble
than the parent  compound, atrazine  (Mills  and
Thurman,  1994).    So,  models  often  omit  the
interaction of atrazine with solids (both suspended
solids  and   sediment)  and  do   not  include
bioaccumulation components.  Because atrazine is
primarily   transported   in   a   dissolved  phase,
groundwater is vulnerable to contamination as it can
receive a load associated with infiltration.
Other chemical compounds, such as cyanazine and
simazine, with the same triazine ring structure as
atrazine  have  been  used  in  the Great  Lakes
watershed.   Cyanazine  usage in the basin in the
early  1990s was about  40% that of atrazine, and
simazine was approximately 1%that of atrazine (U.S.
General Accounting Office, 1993).  Both cyanazine
and simazine were used  as herbicides.

Unless otherwise specified, the information in Table
1.2.1  was   obtained  from  USEPA's  Office   of
Pesticides (January 2003a).

1.2.3 Atrazine Degradation

Atrazine is  known to  degrade in the environment
through either biotic  or abiotic  processes.   The
specific bacteria strain and population, physical and
chemical conditions present,  and  media type all
contribute to determining the  degradation fate  of
atrazine in the  environment.

1.2.3.1 Biotic Degradation in Surface Water

Bacterial processes are known to convert atrazine to
DEA and DIA; however, this degradation is not likely
occurring in the surface water.  Abiotic processes
often  convert  atrazine  to hydroxyatrazine.   See
Figure 1.2.1 for  the chemical  structures of these
major degradation products. Biodegradation assays
of 14- to 32-days of unfiltered water from the River
Po, Italy,  spiked with  various concentrations  of
atrazine, yielded no degradation products (Brambilla
et a/., 1993).  Ingerslev  and Nyholm (2000) tested
natural water  samples  from an  unpolluted forest
stream  using  14C-labeled  atrazine.    Microbial
degradation of atrazine was evaluated by measuring
the evolution of 14C in carbon dioxide (CO2).  Testing
these samples with  a  wide  range  of  atrazine
concentrations typically found in streams showed that
the natural  population  of microbes did not degrade
the labeled atrazine. Biodegradation of atrazine was
not found in two shallow impounded small lakes in
Nebraska that  receive  agricultural inputs of atrazine
from run-off (Spalding et a/.,  1994).  Half-lives  of
atrazine in these lakes  were estimated to range from
193 to 124 days.  The biodegradation product, DEA,
was not increasing relative to atrazine in the lake,
therefore suggesting that the degradation observed
was not biotic.  They surmised that degradation was
                                               17

-------
    (CH3)2CHHN
                             NHC2H5
                                NHC2H5
               Deisopropylatrazine

                       Cl
    (CH3)2CHHN
                 Deethylatrazine
                            N
(CH3)2CHHN
                                NHC2H5
                 Hydroxyatrazine
Figure 1.2.1.  Chemical structures of atrazine and
its major degradation products.
due   to  abiotic   processes.     Evidence   of
biodegradation was not found in a study of a lake in
Nebraska (Ma and Spalding, 1997). However, these
researchers did suggest that abiotic degradation was
the likely mechanism for degradation.  A study of
atrazine degradation in an Iowa stream determined
that atrazine biodegradation was not occurring in the
river (Kolpin and Kalkhoff, 1993).  Modeling analysis
of a small Swiss lake (hydraulic detention time of 1.2
years) found that atrazine is rather stable in the lake
water with removal primarily due to export with water
flowing out of the lake (Buser, 1990; Ulrich et a/.,
1994; Mulleref a/., 1997).  Atrazine degradation via
biotic and/or abiotic processes in Lake Michigan was
found to be negligible using a mass balance model
(Rygwelski etal., 1999).  Biodegradation products of
atrazine are commonly found in surface waters, but
their  origin is likely  from  agricultural  soils  where
biodegradation is  known to occur to a significant
extent.

There are various  hypotheses why  researchers
cannot find evidence of atrazine biodegradation in
surface water.  In systems such as Lake Michigan,
this potential  biotic "food" source  (atrazine) is  very
dilute, and therefore,  it is hypothesized that bacteria
specific to atrazine  degradation do  not thrive.  If
atrazine were to substantially partition to particles in
the water, then perhaps atrazine would be in a more
concentrated form that  could  sustain  the specific
strain of atrazine-degrading bacteria.  Using granular
activated charcoal to enhance atrazine adsorption
and  the  inclusion  of  atrazine-specific  bacterial
degraders in a laboratory batch reactor, significant
reductions (45% to 86%) in atrazine concentrations
were  achieved after  a 15-day  incubation period at
10°C  (Feakin  et a/., 1994).  Also, if present in
sufficient quantities,  more readily available sources
of nitrogen other than that provided by the 1,3,5-
triazine structure may be preferentially  used  by the
atrazine-degrading bacteria.  Therefore, the atrazine
triazine structure would be left intact (Feakin et a/.,
1994). Typically, the first stage in the biodegradation
of the   1,3,5-triazines  is  deisopropylation  and
deethylation leading to the removal of nitrogen from
positions four and six  of  the 1,3,5-triazine  ring.
Feakin etal. (1994) also showed that degradation in
water without  sufficient assimilable organic carbon
did not support biodegradation.  They theorized that
the bacteria  needed a certain minimum level of
carbon for maintenance energy and growth.

While atrazine biodegradation  is not likely to occur
naturally  in surface waters, efforts have been made
to find ways  to create better  conditions for biotic
degradation in water in laboratory operations,  with
the intent of  applying the methodology to water
treatment facilities. A pilot plant operation studying
the potential to degrade atrazine in water found that
                                                18

-------
an   atrazine-specific   degrading  bacterium,
Rhodococcus rhodochrous strain SL1, was effective
in  degrading the  herbicide after the atrazine was
adsorbed  to granular activated carbon  packed in
columns (Jones et  al., 1998).  However,  periodic
reinoculation onto the columns was required to
maintain adequate numbers of SL1.  Conventional
water treatment facilities are not effective in reducing
atrazine concentrations.   Conventional activated
sludge  wastewater  treatment  plants  are  also
ineffective  at  removing  atrazine  from the waste
stream (Monteith et a/., 1995).

1.2.3.2  Abiotic Degradation in Surface Water

1.2.3.2.1 Hydrolysis

Degradation by hydrolysis is likely  in water if the
environmental conditions are favorable. Hydrolysis
was not found to occur at pH greater than 4 at 15°C
in  buffered  distilled  water or natural river water
(Comber,  1999).  Furthermore, the addition of  iron
hydroxide  and aluminum silicate  did not promote
degradation via catalysis as some researchers have
hypothesized.  The pH of Lake Michigan is relatively
high  (8.2)  and,   therefore  unlikely  to  support
hydrolysis.  However, at temperatures of 35°C,
atrazine was found to slowly degrade via hydrolysis
at a range of pHs from 3 to 8 in distilled water (Lei et
a/., 2001).  Hydrolysis rate constants were increased
(half-lives shortened) with the addition of humic acids
and nitrate ions. An evaluation of atrazine hydrolysis
in  groundwater samples at  a  pH of  7.8   and
temperatures of 4°C and 30°C showed no significant
loss (Widmer et a/., 1993).  Also, when  hydrolysis
experiments were conducted at room temperature
and a pH  of 6.5,  dissolved organic  carbon (DOC)
additions with and without nitrate did not cause  any
degradation (Hapeman et a/., 1998).  Spalding et al.
(1994) theorized that surface catalyzed  hydrolysis
was a possible mechanism for degrading atrazine in
two small lakes located in Iowa. These shallow lakes
had high turbidity with high DOC (5.1 to 8.4 mg/l).

1.2.3.2.2 Photolysis

Photolysis is enhanced when nitrate ions are present
to facilitate indirect photolysis by acting as a catalyst.
It is hypothesized that in the presence of the nitrate
ion,  hydroxy  radicals are produced  resulting in
oxidation and/or removal of the alkyl groups.  In a
small stream in Iowa, isolated from groundwater
intrusion, Kolpin and Kalkhoff (1993) found that the
atrazine half-life had a significant inverse relationship
with   sunlight,  suggesting  that  photolysis  was
responsible.  This same inverse  relationship was
noted in a reservoir in Iowa (Chung and Gu, 2003).
However,  in  both of these studies  a correlation
between atrazine half-lives and concentrations of
nitrate ions was poor. The relationship between half-
lives and nitrate concentrations may be masked in
the  natural  environment  because  of  the strong
seasonality of photodegradation with sunlight.  Using
titanium  dioxide  (TiO2)  as a  photocatalyst and
simulated  solar  light  in  a  laboratory  setting,
researchers  have  found  that atrazine  can  be
degraded very rapidly (Pelizzetti et al., 1990) with a
half-life estimated at 19 minutes (Konstantinou et al.,
2001a).    Some DOC  mimics  can  significantly
increase photodegradation of atrazine, while others
do not, leading  researchers to believe that both the
structural properties and concentration of DOC in
water are  important factors  to  consider  when
assessing   potential  photodegradation  impact
(Hapeman et al., 1998).  Using natural light sources,
some studies have found that structural properties of
some types of natural DOC present in surface water
will   actually   reduce   photodegradation  rates
(Konstantinou  et al., 2001 b).   The  degradation
products found  in  the  Konstantinou  study  using
natural  water  samples  were  the  hydroxy and
dealkylated derivatives of atrazine. It appears that
light energy at wavelengths less  than 300  nm is
necessary  to  initiate  direct  photolysis  where
photolysis occurs without the need of an intermediary
(Comber, 1999). However, natural sunlight provides
very  little   of  this   light  energy.     Direct
photodegradation produces primarily hydroxyatrazine
(Konstantinou etal., 2001 b).

Even though Lake  Michigan has  very low nitrate
(1994-1995   median  0.28   mg/L)  and  DOC
concentrations  (1994-1995 median 1.5 mg/L), it is
possible  that some degradation  is occurring via
various photolysis processes. However, it is believed
that  the impact on  the lake is small because the
depth of the lake limits light penetration through the
water column and isolates the hypolimnion during the
high solar  radiation period.   Studies of  atrazine
transport,  atmospheric deposition, and fate in Isle
                                                19

-------
Royale National Park have shown that the shallow
lakes have lower atrazine concentrations than the
deeper  lakes on this  island  in  Lake  Superior
(Thurman and Cromwell, 2000).  These island lakes
are in a pristine area and receive their atrazine input
from  the atmosphere.   If  atrazine  were  highly
persistent in water, then one would expect that the
shallow lakes would have higher concentrations than
the deeper lakes because the shallow lakes have a
higher surface  area to  depth  ratio.    However,
Thurman and Cromwell's  findings are  just  the
opposite, and a possible explanation for this is that
photolysis in the shallow lakes occurs throughout the
water column, but in the deeper lakes it may be
limited to the upper water column only.

1.2.3.3 Atrazine Degradation in Soil

The degradation of atrazine in soils is  much faster
than in water. Durand and Barcelo (1992) presented
half-life values for atrazine in soil from six studies. All
of the studies found half-lives of 125 days or less.
Nair and Schnoor (1994) found that  degradation
rates in soil  depend strongly on soil environmental
conditions.   Degradation  increased with increasing
soil water and  organic  carbon  content; however,
degradation rates decreased in low oxygenated soils.
Mirgain et al. (1993) found that bacteria degrade
atrazine in soils where the organic carbon content is
greater  than 2%.   Degradation  increased  with
increasing carbon content.   They  also noted that
repeated applications of atrazine on the same soil
sample results in  the enhancement of  degradation
with each successive application. They found that
the  reason  for this  is  that  bacteria  populations
specific to degrading atrazine increased with each
application  and  the number of bacteria strains
decreased.  Compared  to water, soil  is  better in
facilitating degradation of atrazine because the "food"
source (atrazine) is  readily available  to  support
bacterial  strains that are efficient in degrading the
herbicide.

References

Brambrilla, A., B. Rindone, S. Polesselo, S. Galassi,
    and R. Balestrini.  1993.   The Fate of Triazine
    Pesticides in River Po Water.  Sci. Total Environ.,
    132(2/3):339-348.
Buser, H.-R.  1990. Atrazine and Other s-Triazine
   Herbicides in Lakes and in Rain in Switzerland.
   Environ. Sci. Technol., 24(7): 1049-1058.

Chung, S. and R.R. Gu.  2003.  Estimating Time-
   Variable  Transformation  Rate of Atrazine in a
   Reservoir.  Adv. Environ.  Res., 7(4):933-947.

Comber, D.W.  1999. Abiotic Persistence of Atrazine
   and Simazine in Water.  Pest. Sci., 55(7):696-
   702.

Durand, G. and D. Barcelo.  1992. Environmental
   Degradation   of   Atrazine,   Linuron,   and
   Fenitrothion in Soil Samples.  Toxicol. Environ.
   Chem., 36(3/4):225-234.

Feakin, S.J., E. Blackburn, and R.G. Burns.   1994.
   Biodegradation of s-Triazine Herbicides at Low
   Concentrations in Surface Waters. Water Res.,
   28(11):2289-2296.

Frank, R., G.J.  Sirons,  R.L.  Thomas,  and  K.
   McMillan.     1979.    Triazine  Residues  in
   Suspended Solids (1974-1976) and Water (1977)
   From the Mouths of Canadian Streams Flowing
   Into the  Great Lakes.  J. Great Lakes  Res.,
   5(2):131-138.

Hapeman, C.J., S. Bilboulian, B.G.Anderson, and A.
   Torrents.  1998.  Structural Influences of Low-
   Molecular-Weight  Dissolved  Organic Carbon
   Mimics on  the Photolytic  Fate  of  Atrazine.
   Environ. Toxicol. Chem., 17(6):975-981.

Ingerslev,  F. and  N. Nyholm. 2000.  Shake-Flask
   Test for Determination of Biodegradation Rates of
   14C-Labeled Chemicals at Low Concentrations of
   Surface Water Systems.  Ecotoxicol.  Environ.
   Safety, 45(3):274-283.

Jones, L.R., S.A. Owen, P. Horrell, and R.G. Burns.
   1998.  Bacterial Inoculation of Granular Activated
   Carbon Filters for the Removal of Atrazine From
   Surface Water. Water Res., 32(8):2542-2549.

Kolpin, D.W.  and S.J.  Kalkhoff.   1993.  Atrazine
   Degradation in a Small Stream in Iowa.  Environ.
   Sci. Technol.,27(1):134-139.
                                               20

-------
Konstantinou, I.K., T.M. Sakellarides, V.A. Sakkas,
   and  T.A. Albanis.    2001 a.   Photocatalytic
   Degradation of Selected  s-Triazine  Herbicides
   and   Organophosphorus   Insecticides   Over
   Aqueous  TiO2  Suspensions.   Environ.  Sci.
   Technol., 35(2):398-405.

Konstantinou, I.K., A.K.  Zarkadis, and T.A. Albanis.
   2001 b. Photodegradation of Selected Herbicides
   in Various Natural  Waters  and  Soils  Under
   Environmental Conditions. J. Environ. Quality,
   30(1):121-130.

Lei, Z.,  C. Ye, and X.  Wang.   2001.   Hydrolysis
   Kinetics of Atrazine and Influence Factors.  J.
   Environ. Sci.,  13(1):99-103.

Ma,  L.  and  R.F.  Spalding.    1997.    Herbicide
   Persistence  and  Mobility in  Recharge  Lake
   Watershed in York, Nebraska. J. Environ. Qual.,
   26(1):115-125.

Mills, M.S. and E.M. Thurman. 1994.  Reduction of
   Nonpoint Source Contamination of Surface Water
   and  Groundwater by  Starch Encapsulation of
   Herbicides. Environ. Sci. Technol., 28(1):73-79.

Mirgain, I., G.A.  Green,  and H.  Monteil.   1993.
   Degradation   of   Atrazine   in   Laboratory
   Microcosms:  Isolation and  Identification of the
   Biodegrading Bacteria. Environ. Toxicol. Chem.,
   12(9): 1627-1634.

Monteith,  H.D., W.J.   Parker,  J.P.  Bell, and  H.
   Melcher.  1995. Modeling the Fate of Pesticides
   in Municipal  Wastewater Treatment.   Water
   Environ. Res., 67(6):964-970.

Muller,  S.R., M.   Berg,  M.M.  Ulrich,  and R.P.
   Schwarzenbach. 1997. Atrazine and Its Primary
   Metabolites in Swiss  Lakes: Input Characteristics
   and Long-Term Behavior  in the Water Column.
   Environ. Sci. Technol., 31(7):2104-2113.

Nair, D.R. and J.L Schnoor.  1994.  Effect  of Soil
   Conditions on  Model Parameters and Atrazine
   Mineralization Rates.  Water Res., 28(5):1199-
   1205.
Pelizzetti,  E., V. Maurino, C. Minero, V. Carlin, E.
   Pramauro,   and   O.  Zerbinati.     1990.
   Photocatalytic Degradation of Atrazine and Other
   s-Triazine Herbicides.  Environ.  Sci. Technol.,
   24(10):1559-1565.

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.

Solomon,  K.R.,  D.B. Baker, R.P. Richards, K.R.
   Dixon,  S.J. Klaine, T.W. LaPoint, R.J.  Kendall,
   C.P. Weisskopf, J.M. Giddings, J.P. Giesy, L.W.
   Hall, Jr., and W.M. Williams. 1996.  Ecological
   Risk Assessment of Atrazine in North American
   Surface  Waters.   Environ.  Toxicol.   Chem.,
Spalding, R.F., D.D. Snow, D.A. Cassada, and M.E.
   Burbach.  1994. Study of Pesticide Occurrence
   in Two Closely Spaced Lakes in  Northeastern
   Nebraska.  J. Environ. Qual., 23(3):571-578.

Thurman,  E.M.  and  A.E.  Cromwell.    2000.
   Atmospheric Transport, Deposition, and Fate of
   Triazine  Herbicides and Their Metabolites in
   Pristine Areas  at  Isle Royale  National Park.
   Environ. Sci. Technol., 34(15):3079-3085.

Ulrich,  M.M.,   S.R.  Muller,  H.P. Singer,  D.M.
   Imboden, and R.P. Schwarzenbach. 1994. Input
   and Dynamic Behavior of the Organic Pollutants
   Tetrachloroethylene,  Atrazine, and NTA  in a
   Lake:    A  Study  Combining  Mathematical
   Modeling and Field Measurements. Environ. Sci.
   Technol. ,28(9):1674-1685.

U.S.  Department of Agriculture.  2001.  Agriculture
   Research Service Pesticide Properties Database.
   Available from U.S. Department of Agriculture at
   http://www.ars.usda.gov.

U.S.   Environmental  Protection  Agency.    1995.
   National  Primary Drinking Water  Regulations,
   Contaminant Specific  Fact  Sheets, Synthetic
   Organic   Chemicals,   Consumer  Version.
   EPA/8 11/F-95/003-T.
                                              21

-------
U.S.  Environmental  Protection Agency.   2003a.
   Pesticides:  Topical and Chemical Fact Sheets -
   Atrazine  Background.    U.S.  Environmental
   Protection Agency, Office of Pesticides Program,
   Washington,  D.C..    Available   from  U.S.
   Environmental Protection Agency at http://www.
   epa.gov/pesticides/factsheets/atrazine_
   background.html.

U.S.  Environmental  Protection Agency.   2003b.
   Interim Reregistration Eligibility Decision (IRED)
   for  Atrazine.   U.S. Environmental Protection
   Agency,   Office   of   Pesticides   Program,
   Washington, D.C.  Case Number 0062, 285 pp.

U.S.  Environmental  Protection Agency.   2003c.
   October 31, 2003 Addendum to the January 31,
   2003  IRED.   U.S. Environmental Protection
   Agency,   Office   of  Pesticides   Program,
   Washington, D.C.  16pp.
U.S. General Accounting Office. 1993. Reporttothe
   Chairman,  Subcommittee  on  Oversight  of
   Government   Management,   Committee   on
   Governmental Affairs, U.S. Senate: Pesticides -
   Issues Concerning Pesticides Used in the Great
   Lakes Watershed.   U.S. General Accounting
   Office, Washington,  D.C.  GAO/RCED-93-128,
   39 pp.

Westerhoff,  P., Y. Yoon, S. Snyder,  and E. Wert.
   2005.     Fate   of  Endocrine-Disruption,
   Pharmaceutical,  and Personal Care Product
   Chemicals  During  Simulated Drinking Water
   Treatment Processes.  Environ.  Sci. Technol.,
   39(17):6649-6663.

Wdmer,  S.K., J.M.  Olson, and  W.C.  Koskinen.
   1993. Kinetics of Atrazine Hydrolysis in Water.
   J. Environ. Sci. Health, 28(1):19-28.
                                             22

-------
                                          PART1
                                    INTRODUCTION
Chapter   3.
Observations
Atrazine   Field   Data
Kenneth R. Rygwelski
United States Environmental Protection Agency
Office of Research and Development
National Health and Environmental Effects
  Research Laboratory
Mid-Continent Ecology Division
Large Lakes and Rivers Forecasting Research
  Branch
Large Lakes Research Station
9311 Groh Road
Grosse lie, Michigan 48138
and
Harry B. McCarty, Ken Miller, Robert N. Brent, and
  Judy Schofield
DynCorp (a CSC Company)
601 Stevenson Avenue
Alexandria, Virginia 22304

1.3.1 Background

In this chapter, a  summary of the Lake Michigan
Mass Balance Project (LMMBP) atrazine data and
historical data  are presented along  with  a brief
description of sampling and analytical methodology.
A LMMBP atrazine data report by DynCorp Science
and Engineering Group was prepared that provides
more details regarding  concentrations of atrazine and
its  degradation  products  related  to  sampling
atmospheric components, tributaries, and the open-
lake water column (Brent eta/., 2001). The DynCorp
data report also provides an in-depth discussion on
data quality implementation and assessment. Also,
see Part 1, Chapter 1,  Section 1.1.6 in this report for
references to additional  documents,  such as the
LLMBP Methods Compendium and quality assurance
plans, that provide additional details on the project.
Project  data  reside in a  Great Lakes  National
Program Office (GLNPO)-managed  Great  Lakes
Environmental Monitoring Database (GLENDA). The
data were collected for use in the mass balance
models.

For the  LMMBP, measurements  of atrazine, along
with two degradation products deisopropylatrazine
(DIA) and deethylatrazine (DEA), were attempted for
all media. However, for some media, the detection of
the  degradation products  was  difficult  because
atrazine concentrations were very low.  Whenever
possible, Principal Investigators (Pis) were requested
to report analytical results as measured, even if the
value  was lower than the method  detection limit.
This modeling report focuses on  modeling atrazine
and not the degradation products because  of the
sparsity of degradation data for some media.  Also,
triazines other than atrazine can degrade into DEA
and DIA (Thurman et a/., 1994).  So, if all  of the
parent compounds are not modeled, the degradation
products cannot be modeled.  In  a summary report
(U.S. General Accounting Office [USGAO], 1993) for
pesticide usage in the basin for 1990 and 1991, two
other  triazines  used as herbicides  in the  Lake
Michigan  basin (simazine  and  cyanazine)  can
degrade into DIA. Simazine usage in the basin was
extremely  low  compared to atrazine  usage so
degradation  products  from  simazine  would be
insignificant. Cyanazine usage, however, was about
37% of the atrazine usage estimates. Propazine can
degrade into DEA, but this chemical was not listed in
the  usage tables  of the USGAO report.   It has
                                             23

-------
been estimated that atrazine is the major source for
DEA (98%) in the Corn Belt (Thurman et a/., 1994).

Hydroxylated atrazine degradation products are also
found in the environment (Lerch et a/., 1998) and are
formed by substitution of the chlorine atom with a
hydroxyl group in the parent atrazine compound.  In
a survey of Midwestern streams, they found that
these hydroxylated atrazine degradation products
were   less   abundant   than   DEA  and   DIA.
Hydroxyatrazine is the primary hydroxylated atrazine
degradation product but was not measured  in the
LMMBP.

1.3.2  Atmospheric Components

Atrazine enters the atmosphere by volatilization from
either agricultural land (soil and plant) or water, by
wind erosion from fields where the chemical is either
sorbed onto soil particles or as  a  pure pesticide
particle from plant or soil surfaces, and by physical
drift of spray during application (Banks and Tierney,
1993). Once the chemical is airborne, a  variety of
physical  and/or  chemical processes  can  cause
degradation, and  various physical  processes can
cause deposition back to land or water. In addition to
the atrazine data report by DynCorp and the LMMBP
Methods Compendium, information on atmospheric
media sampled can be found in a master's thesis by
Sondra Miller (Miller, 1999).

1.3.2.1 Sampling and Analytical Methodology

Primary atmospheric sampling occurred at  eight
shoreline stations. Sampling locations are identified
in Figure 1.1.2. Some limited atmospheric sampling
also occurred at selected open-lake stations aboard
the research vessel,  Lake Guardian.  Also,  three
stations  (Eagle  Harbor,  Michigan; Brule  River,
Wisconsin on the southern shore  of Lake Superior;
and Bondville, Illinois) were located  outside  of the
basin and  were  established  to  characterize  air
masses from the southwest or northwest directions.
Vapor, particulate, and wet deposition were sampled
and analyzed. Atmospheric sampling occurred from
March 15, 1994 to October 20, 1995. A total of 294
vapor phase samples, 226 particulate samples, and
207  precipitation  samples  were collected.    All
samples  were analyzed for atrazine and primary
degradation products DEA and DIA  except for the
Sleeping Bear Dunes site where only atrazine was
analyzed.  From April 1994 through July 1994,
samples from the Sleeping Bear Dunes  site were
collected and analyzed at the  Illinois State Water
Survey (Clyde Sweet).  For the remaining period at
the same site, August 1994 through October 1995,
atmospheric  samples  were  both  collected and
analyzed  by  Indiana  University (Ronald  Hites).
Samples from all other atmospheric stations were
collected and analyzed by the  Illinois State Water
Survey.

Wet deposition  composite samples  were collected
over a 28-day period at the shore-based stations
using a Meteorological Instruments of Canada (MIC-
B) sampler modified  with a heater  for all-weather
sampling. Equipped with a precipitation sensor, the
samplerwas open to the atmosphere only during wet
events.   Rain and snow that was collected flowed
through  a 30 cm XAD-2 resin column that absorbed
the atrazine  and degradation  products  from  the
sample.  Glass wool  plugs, before and after the
column,  prevented   particles  from  entering  the
column.  After  the required  collection period,  the
collection funnel was rinsed with water and wiped
with clean quartz fiber filter paper  to remove any
adhering particles.  Both the filter  paper and the
rinsing became part of the sample.  Five percent of
Illinois State Water Survey wet deposition samples
were field duplicates with a system precision of 115%
for samples above the method detection limit. The
mean laboratory matrix spike recovery was reported
at 82%.  Indiana University analyzed only 14 routine
samples and 12 field  duplicates.   Their  system
precision of the duplicates was 28.1% for samples
above the method detection limit.   They achieved
laboratory matrix spike recovery of 110%.

Composite atmospheric  vapor  and  particulate
samples were collected over a period of 24 hours
every 12 days using a high-volume air sampler. Air
was passed through  a XAD-2  resin to collect the
atrazine  and  degradation products.  Air flow was
maintained at  approximately  34  m3/hour during
sampling. Resin traps were wrapped  in aluminum foil
and sealed in  tin cans and held  at -18°C until
analysis.  Particulate phase atmospheric samples
were collected on pre-fired quartz fiber filters. Filters
were wrapped in aluminum foil and sealed in tin cans
and stored at  -18°C until analysis. Multiple 24-hour
                                              24

-------
samples were often composited to yield a monthly
sample composite.  At the Sleeping Bear Dunes site,
and occasionally some other sites, 24-hour samples
were analyzed individually and then mathematically
composited to yield a monthly average.

The XAD-2 resin or filter samples were extracted by
Soxhlet extraction with 300 ml of a 1:1 hexane and
acetone mixture. The extract was concentrated by
rotary evaporation and then cleaned-up  with 3%
deactivated  silica with a  sodium  sulfate cap to
remove non-target interfering compounds.  Samples
were analyzed using gas chromatography coupled to
a mass spectrometer detector.

1.3.2.2 Results

1.3.2.2.1 Atrazine in the Gas Phase Fraction

Gas phase samples were extremely low; therefore,
quantifying over-the-lake concentrations used in the
volatilization and absorption mass balance algorithms
was  difficult.   Only  11 samples were above the
detection limits.  And of these samples, four were
flagged by the analysts as possibly contaminated due
to field or laboratory blanks, and four others were
from a station  outside the  Lake Michigan basin
located  in  Bondville,  Illinois.   This  leaves three
samples collected in the basin with measurements of
atrazine above the detection  limit-one sample from
South Haven collected July 7, 1994 through July 8,
1994 at 70  pg/m3  and the rest at Sleeping Bear
Dunes collected November 16,1994  -November 17,
1994  at  22.1  pg/m3  and  September  9,  1995-
September 14,  1995 at 31.5 pg/m3.  The sampling
stations at both the Bondville and South Haven sites
are located where local agricultural influences on the
gas phase concentrations are  likely.   Therefore,
these concentrations may not be representative of
gas phase concentrations over-the-lake.  Peck and
Hornbuckle  (2005)  measured    gas   phase
concentrations of atrazine in  the intensively farmed
state of Iowa.  They found  that gas phase atrazine
concentrations  showed a seasonal  pattern  with
highest concentrations evident during the spring and
early  summer.    In their  study,  the  average
concentration of atrazine in the air was 1,200 pg/m3.
In the  LMMBP study,  DEA and  DIA were not
detected in the gas phase.
Because gas phase measurements did not provide a
reliable  over-the-lake estimate in the LMMBP,  we
made assumptions about  this  value  based  on
detection  limits.   In  Miller (1999),  the  method
detection limit (MDL) for atrazine for the shore-based
and open-water sites gas phase concentration was
21.3 ng.  Knowing the average  flow  rate of air
through  the  sampler and  assuming  a 24-hour
collection  period,  Miller  estimated a MDL  of 9.26
pg/m3.  Modeling scenarios presented in this report
utilized this method detection limit to place an upper
expected limit on this boundary condition.

1.3.2.2.2 Atrazine in the Particulate Fraction

Atrazine in the particulate fraction in air was  low and
often difficult to detect.  This finding is also supported
by other studies, such as in rural Iowa - a state with
the highest pesticide applications in the United States
and where 94% of the  state is farmland and 60% of
that area is planted with corn (Nations and Hallberg,
1992). Only 23% of the particulate samples taken for
the LMMBP had atrazine concentrations above the
sample-specific detection limit.   Also, the chemical
was primarily  observed in the months of April, May,
June,  and July.    Only one  particulate  sample
collected  from  August through  March  contained
levels above the MDL.  Maximum monthly average
atrazine concentrations ranged from a  low of 160
pg/m3 at Sleeping  Bear  Dunes in northern  Lake
Michigan to a high of  1,400 pg/m3 at the Bondville
site. The elevated concentration at the Bondville site
is most likely related to the fact that it is in the middle
of an intensive corn-growing region. A summary of
spring/summer atrazine concentrations measured in
the particulate phase can be found in the atrazine
data report  (Brent  et a/.,  2001).  Particle size
distribution analyses  were  not conducted  on
particulates collected in the air samples.  Sweet and
Harlin (1998) estimated that approximately 1% of the
total atrazine load  associated  with wet and dry
particle  deposition  to  Lake Michigan  is  due  to
atrazine associated with particulates.

Of  the  over-water sampling  stations,  only  two
samples had  detectable atrazine in the particulate
fraction, and both of these samples were collected
close to land  in the  southern part of the lake (near
Chicago   and  Indiana   Dunes).    An  atrazine
concentration of 560 pg/m3 was measured at station
                                               25

-------
1 in May 1994, and a concentration of 280 pg/m3 was
measured at station 5. Station 1 is shown on Figure
1.1.2 as the southern-most over-water atmospheric
monitoring station,   and  station  5  is  located
immediately north and slightly west of station 1.

In the spring/summer of 1994, the LMMBP project
detected atrazine but not the degradation products at
the  Eagle  Harbor site, which is located in Michigan's
Upper Peninsula near Lake Superior.  From a period
of early  April to mid-September 1995,  atrazine
sampled   at  Eagle   Harbor   was   detected
approximately 34%  of  the time in the particulate
fraction but not in the vapor phase at this remote site
(Foreman et al., 2000).  In addition, both DIA and
DEA were detected  in the particulate phase.  This
suggests that long range transport is  possible for
both atrazine and the two degradation products via
particles.

Having a higher detection limit than the Foreman et
al. (2000) study may be one reason why atrazine was
difficult to  detect over Lake Michigan in  the LMMBP
study.  Foreman's detection limit was 6 pg/m3.  For
the  LMMBP, the detection limit ranged from 3.0 to 68
pg/m3 (average of 17 pg/m3) for particulate phase
samples analyzed at the Illinois Water Survey, and
from 26.8  to 284 pg/m3 (average of 70.7 pg/m3) for
samples analyzed at Indiana  University.   Another
possible reason for the lack of particulate atrazine
data over-the-lake  is that the type  of particulate
matter carrying atrazine may not be transported very
far from the source.   As a consequence of the low
number of detects at land-based collection sites, and
the   lack   of  evidence  of  atrazine-associated
particulate fluxes over-the-lake, these fluxes were not
estimated for modeling purposes.

1.3.2.2.3  Atrazine and Degradation Products in
Wet Deposition

Atrazine in wet deposition was primarily detected in
the  spring and summer months.  This seasonality
was also reported by Nations and Hallberg (1992)
and Goolsby et al.  (1993).  All LMMBP samples
collected in April and May had detectable levels of
atrazine.  Atrazine was not detected in samples from
November through  February.  DEA and  DIA were
also primarily detected in the spring and summer
months.   DEA was  detected in  samples collected
from March through August, and DIA was only
detected in samples collected from April through
June. DEA had a higher frequency of being detected
and also had a higher concentration on average than
DIA.   Twenty-eight  day   maximum  atrazine
concentrations  measured  over  1994  and  1995
ranged from 100 ng/L at Eagle Harbor to 2,800 ng/L
at the Indiana Dunes site. The high Indiana Dunes
value was  associated with  a low volume sample
collected over a 28-day sampling  period and may
have been influenced by emissions from  nearby
agricultural fields.   During  a rain event,  atrazine
concentrations  are often  much   higher  at the
beginning of the event compared to concentrations
measured at  the end of the event (Nations and
Hallberg, 1992; Goolsby et al., 1993). Nations and
Hallberg (1992) also found that a rain event closely
following an earlier rain event by a day or two had
much  lower   concentrations  (and  often  non-
detectable) levels of atrazine in the wet deposition
sample.  Presumably the first event scavenges the
available pesticide in the atmosphere.  Without a
detailed record of the number and duration of rain
events in the Indiana Dunes sample, it is difficult to
conclude if any of the scavenging circumstances
occurring early in a rain event(s) comprised a major
volumetric  proportion  of the sample collected.
Nations and Hallberg (1992) also found that atrazine
concentrations in wet deposition tend to be higher in
regions that have higher usage  of atrazine.   They
found consistent,  striking differences between two
stations only  11  km apart.   One  station located
adjacent to a  row-cropped field had a much higher
reported value compared to  a station located in a
forested region.   Volume-weighted mean  LMMBP
spring/summer atrazine levels  for the two-year
sampling period (1994-1995) ranged from 19 ng/L at
Eagle Harbor to 120 ng/L at Indiana Dunes.  Due to
the high variability of wet deposition concentrations
of atrazine at sites, stations around the lake were not
statistically different based on the Kruskal-Wallis test.
Sampling  at over-water stations was  limited.  A
southern central lake station  contained 7.5 ng/L on
August  20,  1994 and  a station  in Green Bay
contained 29 ng/L on April 12, 1995.

Concentrations of atrazine and DEA in wet deposition
in 1995 were much lower than observations in 1990,
1991, and  1994.  The concentrations of atrazine
collected in the Lake Michigan basin, as  reported in
                                               26

-------
Table 1.3.1 for 1990 and 1991, compare very well to
other data collected by Goolsby et al. (1997) across
the Midwestern and  Northeastern  United  States.
They found a range of 200 to 400  ng/L for 1990-
1991.  In 1994, atrazine was found  in LMMBP rain
samples collected between mid-March and mid-April,
even though  corn  planting had not yet begun in
southern Wisconsin. This suggests that atrazine was
being transported long range, originating from farm
fields in more southerly states that had been planted
earlier in  the season.   In 1995,  however, the
occurrence of atrazine in wet deposition more closely
coincided with application in the region (Sweet and
Harlin, 1998).   Further evidence that long  range
transport of atrazine to Lake Michigan was minimal in
1995 is reflected in a  low deethylatrazine/atrazine
ratio (DAR) (0.145) for 1995 (Table 1.3.1). The DAR
was calculated using the volume-weighted means for
DEA and  atrazine.  Generally, higher DAR ratios
represent higher levels of degradation of atrazine to
DEA.  Long range transport allows  more time for
degradation  of atrazine to occur in the air mass.
DAR  ratios  were  calculated for  Isle  Royale,  a
wilderness national park in Lake Superior, and the
ratio at the park was calculated to be approximately
0.4 (Thurman and  Cromwell, 2000)  for the study
period 1991-1994. So in regards to DAR and except
for  1995,  the  two areas  (Isle  Royale  and  Lake
Michigan) appear to compare very  well, suggesting
that  under  normal circumstances,   transport  of
atrazine from distant sources does occur in the wet
deposition phase.

A  possible   explanation   of  the  low  atrazine
concentrations for 1995 is that the spring of  1995
was cold and wet in major corn-growing areas south
and west of  the Lake Michigan basin  compared to
1991  and 1994.  This may have limited long range
transport to  the  Lake  Michigan basin.   Omaha,
Nebraska and Peoria, Illinois were selected as being
representative of that area south and west because
they are located in geographic areas where the
Table 1.3.1.  Summary of Wet  Deposition Annual Volume-Weighted Mean Deethylatrazine (DEA)
Concentrations, Atrazine Concentrations, and Deethylatrazine/Atrazine Ratios (DAR) for All Stations
in the Lake Michigan Basin
v Deethylatrazine Atrazine Deethylatrazme/Atrazme _ ..
Year „ „ ' .. ,„.,-.» Samp ing
ng/L ng/L Ratios (DAR) K a
19901
19911
19942
19952
Mean
Mean (Year
1995 Excluded)
101.0
233.0
32.4
4.02
92.6
122
259.0
432.0
80.6
30.0
200
257
0.402
0.540
0.422
0.145
0.377
0.455
3/27/1990
4/2/1991 -
3/15/1994
3/14/1995


Data Range
-8/14/1990
7/9/1991
-7/5/1994
-8/31/1995


1Data from Goolsby et al., 1995. All data were used in calculating the volume-weighted mean concentrations.
 Data reported with the detection limit were converted to half the detection limit. DAR represents only
 situations where both the reported DEA and atrazine concentrations were above the detection limit.

2Data from the LMMBP. All data, including zeros, were used in calculating the volume-weighted mean
 concentrations.  DAR represents only situations where both reported DEA and atrazine concentrations
 were above the detection limit.
                                              27

-------
greatest spring atrazine emissions were estimated
(Scholtz et a/.,  1997).   Figure 1.3.1  shows  the
monthly precipitation at these cities for the important
months when wet deposition fluxes are normally high
(National  Climatic Data Center, 2000).   For both
Peoria and Omaha,  1991  and 1994 were similar in
rainfall events; however, for 1995, the months of April
and  May were  wetter than the other two years.
Figure  1.3.2   shows   the   monthly  average
temperatures  for the same two cities.   For both
Peoria and Omaha,  1991  and 1994 were similar in
average temperatures; however, for 1995 the months
of April and May were colder than the two other
years.    Not  only  were the  LMMBP  atrazine
concentrations in precipitation  low for 1995, but the
total  atrazine deposition for 1995 was approximately
half of what it was in 1994.  This cannot be explained
by very low precipitation in the Lake Michigan basin
for 1995.  Table 1.3.2 displays the combined mean
precipitation amounts from Chicago, Illinois;  South
Bend, Indiana; Muskegon, Michigan; Grand Rapids,
Michigan;  Green Bay, Wisconsin; and  Milwaukee,
Wisconsin for  1994 and  1995 (34.26 and  33.73
inches, respectively) and they are close to the 30-
                     year  mean  of  all  these  sites  (34.22  inches).
                     Comparisons  to  a  50-year  mean for over-lake
                     precipitation  to Lake Michigan can  be found  in
                     Figures 1.4.11 and 1.4.12 in Part 1, Chapter 4 of this
                     report and  show similar   results.    Also, the
                     differences  between  1994 and 1995  cannot be
                     explained by  differences in  amounts  of atrazine
                     applied in the basin between the two years, because
                     these amounts are nearly the same (see the atrazine
                     loading  chapter   for more  information).    The
                     differences may be explained by the  cold and wet
                     spring in the south and  west  corn-growing regions
                     relative to the Lake Michigan  basin.  In a  cold and
                     wet spring, less atrazine emission would be expected
                     to occur because temperature is a driving force  of
                     atrazine volatilization from the soil to the air. In the
                     wet spring of 1995, among  both the  Peoria and
                     Omaha stations, there was one rain event in April
                     over one inch and seven events in May where rainfall
                     was over one inch (and as high as 2.5 inches on May
                     8 at one of the stations).  For spring 1994, there was
                     only  one rainfall   event  among the  two stations
                 12
                 10
                CO
                
-------
                              Peoria, IL

                              Omaha, NE
                         April  April  April  May  May   May  June  June June
                         1991  1994  1995  1991  1994  1995  1991  1994
                                          Month and Year

Figure 1.3.2.  Monthly average temperatures at cities in two large corn-growing regions. Data are from
Peoria, Illinois and Omaha, Nebraska.
Table 1.3.2. Annual Mean Precipitation Amounts Measured at Chicago, Illinois; Fort Wayne, Indiana;
South Bend, Indiana; Muskegon, Michigan; Grand Rapids, Michigan; and Milwaukee, Wisconsin
 Time Period

 30 Years
 1994
 1995
Total Inches (Mean at All Sites)

            34.22
            34.26
            33.73
Standard Deviation

       3.57
       6.94
       4.61
over one  inch and  that was on June 22,  1994.
Perhaps the frequent rainfall events in 1995 washed
significant quantities of atrazine from the fields into
streams,  rivers,  reservoirs, and  groundwater via
infiltration  which  allowed less of the  atrazine to
volatilize from the farm fields. Atrazine that is diluted
in reservoirs,  lakes,  and  rivers would have a lower
volatilization flux than if  it were in a concentrated
form on farm soil.  Also, heavier rainfall in the corn-
growing region could increase scavenging of the
chemical from the atmosphere, thereby leaving less
available for long-range transport.
                      1.3.3  Atrazine in Tributaries

                      Eleven tributaries to Lake Michigan were sampled
                      from April 4, 1995 through October 31,1995. A total
                      of 108 filtered samples were collected. Most tributary
                      samples  contained detectable levels of atrazine,
                      DEA, and DIA.  The tributary samples were collected
                      for purposes of estimating loadings of atrazine to the
                      lake. However, the load estimates are believed to be
                      low, and consequently, alternative tributary loadings
                      were   estimated  based on  watershed  run-off
                      algorithms using the amount of atrazine applied and
                      a watershed export factor of 0.6% for the MICHTOX
                                              29

-------
and  LM2-Atrazine  models.   For LM3, the United
States Geological Survey (USGS) provided loadings
that they calculated from flow and concentration data.
However, for a 90-day period in the spring months,
loadings were enhanced to make up for a "lost" load
(please  see section  5.3.3.3.1  in Part 5 for  more
information). Because the concentration data were
not  directly used  in  the  models,  only  a  brief
description of the data will be presented here.  For a
more complete description of these data, please refer
to Brent et a/. (2001).

1.3.3.1 Sampling and Analytical Methodology

Samples were collected as near to river mouths as
possible without being subject to  flow  reversals
where lake water moves up the river.  Composites
were collected  using  the  USGS  quarter-point
sampling procedure.  In this procedure, the river is
visually  divided  into three equal flow areas.  The
midpoint of each flow panel is sampled at 0.2 and 0.8
times the  depth.  All samples were  pumped and
composited  using a peristaltic pump through  a 0.7
urn glass fiber filter. The filtrate was  passed through
a 250 g, XAD-2 resin to trap the dissolved  atrazine.
Chilled samples were then taken to the analytical lab.
Analyses were conducted using gas chromatography
coupled to a mass spectrometer.  Full details of the
analytical  methods  have been  published in the
Methods  Compendium   (U.S.   Environmental
Protection Agency, 1997a; 1997b).

1.3.3.2 Results

Since tributary samples were only collected over a
seven-month period, full seasonal trends could not
be assessed.   For the three  tributaries  with the
highest mean concentrations of atrazine (St. Joseph
River, Kalamazoo River, and the Grand River), peaks
in atrazine concentrations occurred in mid- to late-
May.  Spring peaks were also  observed for the
degradation  products DIA and DEA.

Individual atrazine concentrations measured in the
streams ranged from a low of 0.5 ng/L in  the Pere
Marquette River to  2,700 ng/L in the St. Joseph
River.   Mean concentrations  of atrazine in the
tributaries ranged from 3.7 in the Manistique River to
350  ng/L in  the St. Joseph River. Per Brent et a/.
(2001),  these  concentrations are  comparable to
concentrations measured elsewhere  in the Great
Lakes region.   Eighty-six percent of the tributary
samples contained less than 100 ng/L of atrazine,
and all samples above  100 ng/L were in the  St.
Joseph, Kalamazoo, or Grand  Rivers.  Tributaries
with the lowest mean atrazine levels were located in
the northern portions of the  lake, where land use is
less dominated by agriculture.

Atrazine  degradation  in  the  watershed can  be
assessed  by looking  at the degradation products.
DEA and  DIA  concentrations  correlated well with
atrazine concentrations in tributary water samples
(Brent  et  a/., 2001).   As atrazine concentrations
increased, both the DEA and DIA increased. The
ratio of concentrations of [DEA]/[atrazine] or DAR is
often  used  to  assess the  extent  of  atrazine
degradation  in  a  sample.   Ratios  on  individual
measurement pairs ranged from 0.08 to 3.7, and the
median was 0.77. Mean DARs were above 1.0 at the
Pere Marquette, Sheboygan, and Milwaukee Rivers,
and were significantly higher than the mean ratios at
the Kalamazoo, Manistique, Grand, and St. Joseph
Rivers.   For all  samples,  the  mean  DAR  of 1.4
measured in October was significantly greater than
the mean ratios in April (0.75), May (0.63), and June
(0.87).  It is common to find  that the ratios increase
for a given tributary as the time since application of
atrazine increases.  Thurman et a/. (1994) also found
an increase  in  DAR from <0.1 shortly after atrazine
application to 0.4 measured  later in the year.  As the
atrazine resides in the soil,  processes (both biotic
and abiotic) are operative that degrade the chemical.
Run-off from these fields will reflect the composition
of DAR in the soil.  Furthermore, during dry spells in
the late  summer,  groundwater  can  make up  a
significant percentage of the total flow of a river.
Groundwater is often associated with high DARs.  In
July-August  1991,  Pereira  and  Hostettler  (1993)
found that the DAR for Mississippi River water was
relatively constant  at 0.2 for the entire river. This
suggests that during the travel time from Minneapolis
to New Orleans (45-65 days), the DAR showed no
evidence  of degradation.   However,  in October-
November, 1991, they found that the DAR in the river
was 0.6 in the  upper reaches of the river.  The low
DAR is believed to be associated with more run-off in
July and August.  During the  fall period, the river was
near base flow in the upper river.  During base flow,
most of the river flow is due  to groundwater.  DARs
                                               30

-------
measured in  groundwater impacted by infiltration
through an agricultural soil matrix are often high, and
exceed or are close to unity (Ma and Spalding,
1997).

1.3.4 Atrazine in Lake Water

1.3.4.1 Sampling and Analytical Methodology

Open-lake water  column samples were collected
during six cruises from April  25, 1994 to April 17,
1995.  Open-lake samples were collected from 35
sampling locations on Lake Michigan, two sampling
locations in Green Bay, and one sampling location on
Lake Huron (see  Figure 1.1.2).  The Lake Huron
samples  were collected to characterize a model
boundary  condition.   Samples were collected  at
depths ranging from 1 to 257 m. During stratification,
samples were collected at mid-epilimnion and mid-
hypolimnion, and master stations were sampled at
one meter below the surface and two meters off the
bottom.   During  non-stratification,  samples were
collected at mid-water  column depths, one meter
below the surface, and two meters off the bottom.

Water  samples were collected  using a  General
Oceanics (Model 1015) rosette sampler on board the
Lake  Guardian  research vessel.    Water  was
transferred from individual rosette canisters to amber
one-liter bottles and stored at 4°C until processing at
the testing laboratory.

Atrazine,  DEA, and DIA were isolated from filtered
water samples using  250 mg Carbopack (Supelco
Corporation) solid phase extraction (SPE) cartridges.
Analytes were eluted  from the SPE using 7 ml of a
90% dichloromethane and 10%  methanol solution
(vol:vol), followed  by 5 ml of methanol. The eluent
was then passed through clean anhydrous sodium
sulfate to remove excess water.   Extracts were
concentrated  to  <100  uL under  a nitrogen gas
stream.   Analysis of  atrazine, DEA, and DIA was
conducted using gas chromatography coupled to  a
mass spectrometer detector.  Further details of the
analytical  methods can be found  in the methods
compendium (U.S. Environmental Protection Agency,
1997a; 1997b).
1.3.4.2 Results

1.3.4.2.1  Spatial Variation

A total of 234 samples (including Green Bay and the
northern Lake  Huron boundary condition samples)
were collected  and analyzed for atrazine, DEA, and
DIA. All lake samples contained levels of atrazine
and  DEA  above the MDL.   All  but 12  samples
contained  DIA  above the MDL for that parameter.
MDLs computed were 1.25 ng/L for atrazine, 2.46
ng/L for DEA,  and 8.27 ng/L for DIA.  Skewness
characterizes  the  degree  of asymmetry  of  a
distribution.     Positive  skewness  indicates   a
distribution with an asymmetric tail extending towards
more positive  values.   In  a normal distribution,
skewness  is  approximately zero.    A  statistical
analysis of  all lake  data indicated that  atrazine
skewness  equaled 0.145.  To further evaluate the
skewness  for atrazine,  the following analysis was
performed (Tabachnick  and Fidell, 1996).

Skewness values of two  standard errors of skewness
(ses) or more (regardless of the sign) are  probably
skewed to  a significant degree. The ses for atrazine
can be estimated by:
ses =
           = 0.144
where, n = total number of open  Lake Michigan
values including duplicates and triplicates (excludes
Green Bay and the northern Lake Huron stations) =
288

2(ses) = 0.2886

Since the skewness for the atrazine lake data, 0. 1 45,
is less than 2 x ses, the distribution can be assumed
to  be normal.  The  deviation  from zero can  be
assumed to be to chance fluctuation.

Wthin Lake Michigan (excluding Green Bay  and
northern  Lake Huron stations),  lateral  and  vertical
atrazine concentrations were relatively consistent
during the LMMBP (Brent et a/., 2001). Individual
sample results ranged from 22.0 to 58.0 ng/L, and
sampling station mean atrazine concentrations only
ranged from 33.0 to 48.0  ng/L.  Similar patterns of
consistency among sampling stations were observed
                                              31

-------
for  DEA  and  DIA  concentrations.    Atrazine
concentrations  in  southern   Green  Bay  were
significantly higher than atrazine concentrations at 18
Lake Michigan sampling stations.  Due to the spatial
consistency   of  atrazine,   DEA,  and   DIA
concentrations within Lake Michigan, lake-wide mean
concentrations can be calculated to reliably represent
the lake. Schottlerand Eisenreich (1994) also  found
Lake Michigan (excluding Green Bay) to lack vertical
and lateral gradients in the 1991 and 1992 data. It is
not surprising that no vertical gradients were found,
because  most of the samples  collected  for the
LMMBP  were collected  during  times  of  non-
stratification of the lake.  Lake-wide concentrations
from the LMMBP study (April 1994-April 1995) and
previous studies are summarized in Table 1.3.3.  A
graphical representation of concentrations observed
in 1994 is depicted in Figure 1.3.3.

1.3.4.2.2 Seasonal Variation

Open-lake atrazine concentrations were measured
during  six sampling  cruises.   Brent  et a/. (2001)
concluded that statistically significant mean open-
lake  concentrations  of  atrazine,  DEA,  and DIA
increased during  the one-year LMMBP sampling
campaign (1994-1995).  Schottler  and Eisenreich
(1997) found that 1992 atrazine concentrations in the
lake  were statistically higher than  the mean lake
concentration measured  in 1991.  Based on  these
field  measurements, it appears that the  lake  is
accumulating  atrazine over time.  More information
on this accumulation will  be  discussed in the
modeling chapters.

Table  1.3.3.   Summary of Historical  Atrazine,
DEA, and DIA Concentrations in Lake Michigan
 Year
Atrazine
(ng/L)
35 (2.0)1*
37(1.8)1
38.1
DEA
(ng/L)
163
25.8
DIA (ng/L)
 1991         35 (2.0)1*    163    Not Available
 1992         37(1.8)1     243    Not Available
 4/1994-        38.1      25.8        14.9
   4/19952
1Schottler and Eisenreich, 1997
2Brentefa/.,2001
3Schottler and Eisenreich, 1994
*Values are means with the standard deviation in
parenthesis.
                                                dissolved
                                                atrazine in
                                                Lake Michigan
                                                water (ng/L)
                                                1994

                                                sample locations
                                                (0-300 meters)
                 Figure 1.3.3.  Atrazine concentrations in  Lake
                 Michigan, 1994.
References

Banks, P.A. and  D. Tierney.   1993.  Biological
   Assessment of  Atrazine  and  Metolachlor  in
   Rainfall. Ciba Plant Protection Department, Ciba-
   Geigy Corporation, Greensboro, North Carolina.
   Technical Paper  1-1993, 16 pp.

Brent,  R.N., J. Schofield, and K. Miller.   2001.
   Results of  the Lake  Michigan Mass  Balance
   Study:     Atrazine   Data   Report.     U.S.
   Environmental Protection Agency, Great Lakes
   National  Program  Office,  Chicago,  Illinois.
   EPA/905/R-01/010,92pp.
                                               32

-------
Foreman, W.T., M.S. Majewski, D.A. Goolsby, F.W.
   Wiebe, and R.H. Coupe.  2000.  Pesticides in the
   Atmosphere of the Mississippi River Valley, Part
   II -Air. Sci. Total Environ., 248(2):213-216.

Goolsby, D.A.,  E.M.  Thurman,  M.L.  Pomes, and
   W.A. Battaglin.  1993. Occurrence, Deposition,
   and  Long  Range  Transport of  Herbicides in
   Precipitation in the Midwest and Northeast United
   States.  In: D.A. Goolsby, L.L.  Boyer, and G.E.
   Mallard (Eds.), Selected Papers in Agricultural
   Chemicals  in   Water   Resources   of  the
   Midcontinental United States, pp.  75-89.  U.S.
   Geological Survey, Denver, Colorado. Open File
   Report 93-418, 89 pp.

Goolsby, D.A., E.A. Scribner,  E.M. Thurman, M.L.
   Pomes, and M.T. Meyer.   1995.   Data  on
   Selected   Herbicides  and  Two  Triazine
   Metabolites  in Precipitation of  the Midwestern
   and  Northeastern  United  States,  1990-1991.
   U.S.  Geological  Survey,  Lawrence,  Kansas.
   Open File Report 95-469, 341 pp.

Goolsby, D.A.,  E.M. Thurman, M.L.  Pomes, M.T.
   Meyer, and W.A.  Battaglin.  1997. Herbicides
   and   Their  Metabolites  in  Rainfall:  Origin,
   Transport, and Deposition Patterns Across the
   Midwestern  and  Northeastern United  States,
   1990 -1991. Environ. Sci. Technol., 31 (5): 1325-
   1333.

Lerch,  R.N.,  P.E. Blanchard,  and E.M. Thurman.
   1998.   Contribution  of  Hydroxylated Atrazine
   Degradation Products to the Total Atrazine Load
   in Midwestern Streams.  Environ. Sci.  Technol.,
   32(1):40-48.

Ma,  L  and  R.F.  Spalding.    1997.    Herbicide
   Persistence and  Mobility in  Recharge Lake
   Watershed in York, Nebraska. J. Environ. Qual.,
   26(1):115-125.

Miller, S.M. 1999. Spatial and Temporal Variability
   of  Organic  and  Nutrient   Compounds  in
   Atmospheric Media Collected During  the Lake
   Michigan  Mass Balance Study.   M.S.  Thesis,
   Department  of   Civil,    Structural,  and
   Environmental Engineering, State University of
   New York, Buffalo, New York. 181 pp.
National Climatic Data Center.  2000.  Archive of
   Climate Data.   Available from   the  National
   Oceanic and  Atmospheric  Administration  at
   http://www.ncdc.noaa.gov.
Nations, B.K. and G.R. Hallberg.
   in  Iowa  Precipitation.    J.
   21(3):486-492.
1992. Pesticides
 Environ.  Qual.,
Peck, A.M. and K.C. Hornbuckle. 2005. Gas-Phase
   Concentrations of Current-Use  Pesticides in
   Iowa. Environ. Sci. Technol., 39(9):2952-2959.

Pereira, W.E. and F.D. Hostettler.  1993.  Nonpoint
   Source Contamination of the Mississippi River
   and Its Tributaries by Herbicides.  Environ.  Sci.
   Technol., 27(8): 1542-1552.

Scholtz, M.T., A.C. McMillan, C. Slama, Y. Li, N.
   Ting, and K.  Davidson.    1997.   Pesticide
   Emissions Modeling-Development of a North
   American  Pesticide   Emissions  Inventory.
   Canadian Global Emissions Interpretation Centre,
   Ortech   Corporation,   Mississauga,   Ontario,
   Canada.  Final Report #CGEIC-1997-1, 242 pp.

Schottler,  S.P.  and  S.J.  Eisenreich.    1994.
   Herbicides in the Great Lakes.   Environ.  Sci.
   Technol., 28(13):2228-2232.

Schottler, S.P. and S.J. Eisenreich.   1997.  Mass
   Balance  Model to  Quantify Atrazine Sources,
   Transformation Rates, and  Trends in the Great
   Lakes. Environ. Sci. Technol., 31 (9):2616-2625.

Sweet, C.W. and K.S. Harlin.   1998.  Atmospheric
   Deposition of  Atrazine  to  Lake   Michigan.
   Presented at the  Air and Waste  Management
   Association's   91st   Annual   Meeting   and
   Exhibition,  June  14-18,  1998,  San  Diego,
   California.     Illinois  State  Water  Survey,
   Champaign, Illinois. Report N umber 98-TA37.02.

Tabachnick,  E.G. and L.S. Fidell.   1996.   Using
   Multivariate Statistics,  Third  Edition.   Harper
   Collin Publishers, Incorporated, New York, New
   York.
                                              33

-------
Thurman,  E.M.,  M.T.  Meyer,  M.S.  Mills,  L.R.
   Zimmerman, C.A. Perry, and  D.  A.  Goolsby.
   1994.     Formation  and   Transport   of
   Deethylatrazine   and  Deisopropylatrazine  in
   Surface  Water.     Environ.   Sci.  Technol.,
   28(13):2267-2277.

Thurman,   E.M.   and   E.  Cromwell.     2000.
   Atmospheric Transport, Deposition, and Fate of
   Triazine Herbicides  and Their Metabolites in
   Pristine Areas at Isle Royale  National  Park.
   Environ. Sci. Technol., 34(15):3079-3085.

U.S. Environmental Protection Agency. 1997a. Lake
   Michigan Mass Balance Study (LMMB) Methods
   Compendium, Volume  1:  Sample Collection
   Techniques.   U.S.  Environmental  Protection
   Agency, Great Lakes National  Program Office,
   Chicago, Illinois. EPA/905/R-97/012a, 1,440pp.
U.S. Environmental Protection Agency. 1997b. Lake
   Michigan Mass Balance Study (LMMB) Methods
   Compendium, Volume 2: Organic and Mercury
   Sample   Analysis  Techniques.     U.S.
   Environmental Protection Agency, Great Lakes
   National  Program  Office,  Chicago,  Illinois.
   EPA/905/R-97/012b, 532 pp.

U.S. General Accounting Office. 1993. Reporttothe
   Chairman,  Subcommittee  on   Oversight  of
   Government  Management,  Committee  on
   Governmental Affairs, U.S. Senate: Pesticides -
   Issues Concerning Pesticides Used in the Great
   Lakes Watershed.  U.S.  General Accounting
   Office, Washington, D.C.  GAO/RCED-93-128,
   39 pp.
                                             34

-------
                                         PART1
                                    INTRODUCTION
Appendix 1.3.1 Information Management

David A. Griesmer
Computer Sciences Corporation
Large Lakes Research Station
9311 Groh Road
Grosse lie, Michigan 48138
and
Kenneth R. Rygwelski
United States Environmental Protection Agency
Office of Research and Development
National Health and Environmental Effects
  Laboratory
Mid-Continent Ecology Division
Large Lakes and Rivers Forecasting Research
  Branch
Large Lakes Research Station
9311 Groh Road
Grosse lie, Michigan  48138

To support the modeling efforts of the Lake Michigan
Mass Balance Project  (LMMBP), samples were
collected  and  analyzed  by  the United  States
Geological Survey (USGS) and several universities
(Table A1.3.1).  The focus group acronyms in the
table provide a unique identifier of data sets. The
first two letters stand for the organization, the third
letter represents the media sampled  (air,  lake, or
tributary), and the fourth letter identifies the chemical
(atrazine) analyzed.  Project data were sent to the
United  States  Environmental  Protection  Agency
(USEPA)  Great  Lakes National  Program  Office
(GLNPO) in Chicago,  Illinois.  GLNPO staff, under
the direction of Louis  Blume, were responsible for
quality  assurance  (QA)  assessment, organization,
and consolidation of all data.  To facilitate the QA
assessment  process,   a  SAS  application,  the
Research Data Management and Quality Assurance
System (RDMQ), developed by Syd Allen, a private
contractor, was used to automate the QA process
(Sukloff et a/., 1995). RDMQ is a menu-driven SAS
program.  It has capabilities for loading data, applying
quality control (QC) checks, adding validity flags,
viewing and  editing data, producing user-defined
tables and graphs, and exporting data in ASCII files.
These tasks  are performed through a set of menu-
driven SAS programs and macros. Data which had
been  put through  the  assessment process and
approved for release  by both GLNPO and the
Principal Investigator (PI) were then sent to USEPA,
Office of Research and Development (ORD)/National
Health  and  Environmental  Effects  Research
Laboratory  (NHEERL)/Large  Lakes and  Rivers
Forecasting  Research  Branch  (LLRFRB)/Large
Lakes Research Station (LLRS) for use  by the
modeling staff.

A1.3.1.1     Overview   of   Information
Management at  the LLRS

Data received from GLNPO were usually in the form
of electronic  media.  Data were typically E-mailed,
but sometimes they were downloaded from GLNPO
databases or received  on CD-ROM.  Data  were
reformatted by GLNPO into a form facilitating  entry
into database programs at the LLRS. Upon arrival,
raw data were copied to the "Immb" folder on David
Griesmer's personal network  space ("M:\" drive).  In
addition, data were imported into one  of several
Microsoft Access databases  in the "\Access\lmmb"
folder on Mr.  Griesmer's "M:\" drive.  The "M:\" drive
was used to facilitate data security because this file
                                             35

-------
Table A1.3.1
Modeling
List  of  Parameters Analyzed and Principal  Investigators for the LMMBP Atrazine
       Parameter
            Focus
            Group
       Media
     Notes
 Principal Investigator
 Atrazine, Deethylatrazine
 (DEA),
 Deisopropylatrazine
 (DIA), Terbuthylazine
            IUAA
Atmospheric Vapor
and Particulate
Phase, Precipitation
 Atrazine, DEA, DIA,
 d5-Atrazine
            WSAA
Atmospheric Vapor
and Particulate
Phase, Precipitation
 Atrazine


 Atrazine
            RULA


            RUTA
Open-lake


Tributary
 Flow
            N/A
Tributary
Sleeping Bear
Dunes site only.
Keri
Hornbuckle, U.
of Iowa, used
these data to
calculate
loadings

All stations
except Sleeping
Bear Dunes
site.  Keri
Hornbuckle, U.
of Iowa, used
these data to
calculate
loadings
David Hall,
USGS, used
these data to
calculate
loadings
Ronald Hites, Indiana
University
Clyde Sweet, Illinois
State Water Survey
Steven Eisenreich,
Rutgers University

Steven Eisenreich,
Rutgers University
                 David Hall, USGS
space is backed up regularly and is available only to
Mr. Griesmer.  At some point in the future, the
location of these data may change; however, limited
access and backups of the data will be maintained.
Data were placed in the Microsoft Access databases
to facilitate  data  review/assessment  and  later
retrieval for the modeling team.

Prior to use, several reviews were done of the data
received to look for errors in the  data sets. At the
LLRS, this review  was broken up  into two parts.
First,  an  initial review was  made  to check  for
completeness of information; to lookfortranscription,
                                   programming, and formatting errors; and to review
                                   comments  added  by  collection   and  analysis
                                   personnel.  Second, a review was done by the data
                                   users to determine if the data made environmental
                                   sense. This type of review was conducted for the
                                   open-lake data.   Tributary atrazine loadings  and
                                   atmospheric atrazine fluxes/loadings  did  not  go
                                   through this review  process at the  LLRS, but they
                                   were  assessed by study members assigned with
                                   providing these loading values.  Tributary atrazine,
                                   deisopropylatrazine (DIA), and deethylatrazine (DEA)
                                   loading  assessments were  done  by  David Hall,
                                   (USGS).     All  atmospheric  atrazine  loading/-
                                               36

-------
concentration  data  were   assessed   by   Keri
Hornbuckle, University of Iowa.

Samples that GLNPO determined had  failed  the
RDMQ QA process were flagged with the value of
-9999 in the Grosse lie database. GLNPO preserved
all of the values in the data sets that were received
and  flagged the  analytical  remark  field for that
parameter.   Flagging  these  values  as  -9999
facilitated processing by analytical software such as
IDL.   In addition,  parameter values with analytical
remark flags of "INV" (invalid data, as determined by
the GLNPO  QA evaluation),  and "NAI"  (no  result
reported - interference) were changed  to -9999.
Samples with the analytical remark flag of "LAC" (no
results reported, laboratory accident) were removed.

Documentation  associated  with  the  data  was
reviewed.    RDMQ    data   warning  fields
(RS_NMAND,RS_WARN,   RS_UPDAT)   were
checked to  verify that there were  no   problems
flagged by RDMQ which were inadvertently included
in the database.  Every routine field sample (RFS)
and field duplicate (FD#) was  checked to verify that
a valid  station name, sampling date,  and  depth
collection  information were  included.  The  value
ranges (minimum, maximum,  average) for atrazine
and its degradation products  (DEA and  DIA) were
checked to look for any obvious errors. Data ranges
of all  data were also checked for obvious errors.
Data were checked to verify units and to confirm
whether blank, dilution, or surrogate correction were
done.   Sample QC  and station comment  fields
(RECSTAT,  RECSTATF,  and STNNOTES)  were
checked for any comments associated with a sample.
All of this information  was recorded on a Data
Verification Checklist (Table A1.3.2). If questions or
errors were  found,  they were  referred back  to
GLNPO for resolution.

Upon  completion of this initial data check, readme
files were created  to describe the data, and the raw
data set(s) and readme  files were copied to a data
archive on the LLRS Unix systems. This archive is
located at \usr\lmmbdata on the Unix servers and is
available to modeling staff at the LLRS. Each study
has its own directory (LMI0001-LMI0028) within the
Immbdata archive. The directories related to  the
atrazine modeling can be found in Table A1.3.3.
Information on  other LMMBP  directories can be
found in the LMMBP PCB report (Rossmann, 2006).

At the same time, information about data received
(metadata) was stored  in a  searchable  Microsoft
Access  database.  The database is found  on the
LLRS common drive "\\giord2\grlcommon", which is
also known as  the "L:\" drive.  This  database is
named  "Imtrack2000.mdb" and is  found  in  the
L:\Public\Access\lmmb folder.   This  database is
available to all staff. This database can be searched
by library  number  (consecutive number  assigned
when data are logged in, corresponds to LMI folder
name in Immbdata archive), PI, parameter,  PI and
parameter, or library number and parameter (Table
A1.3.3).

After initial review of a data set was completed,  data
were retrieved from the Microsoft Access databases
and exported  into files (usually Microsoft Excel) for
assessment by the modeler who would be using the
data set. Atrazine data were assessed by William
Richardson.  Initially, only routine field samples and
field duplicates were given to the data assessors. If
issues or problems were found, the person assessing
the data would then request additional QA data. If
questions/problems could not be resolved by looking
at QA data, they were referred back to GLNPO for
resolution.  GLNPO  was informed  whenever we
rejected data.

After the assessment process was completed, files
were created which could be used in IDL, which is a
software package used for visualization and analysis
of LMMBP data. A standard format was developed
for water data (Table A1.3.4).  All files were fixed
format ASCII text files. One of the principal uses of
IDL  was  to  develop  volume-weighted  averages
(VWA) estimates of parameter concentrations for
each cell in the modeling grid. These VWA estimates
could then be compared to model results.

A1.3.1.2 Summary

The  LMMBP data  received  at the  LLRS were
carefully evaluated prior to use to ensure that the
field data  being used by the  modelers were as
accurate as possible. In addition, data were archived
and cataloged to protect these valuable data sets
                                              37

-------
Table A1.3.2.  Example of Data Verification Checklist Used for the LMMBP








Data Verification Checklist



FOCUS	   Version Number	   Date Received



Description:	



1.  Read any documentation which came with data files:	



2.  Make sure I understand field names in RDMQ files:	
3.  Check fields which according to RDMQ should not be flagged/or indicate some question, with data (e.g.




   RS_NMAND, RS_WARN, RS_UPDAT).




      RS_NMAND	




      RS_WARN	




      RS UPDAT
4. Make sure every RFS and field duplicate has station, date, depth collected information.
5.  Check to make sure every sample has station name that is valid.
6.  Check number of RFS and field duplicates for every analyte. Total Samples



      Analyte	    RFS	   FDn	



      Analyte	    RFS	   FDn	



      Analyte	    RFS	   FDn	



      Analyte	    RFS	   FDn	



      Analyte	    RFS	   FDn	



      Analyte	    RFS	   FDn	
                                            38

-------
7. Analysis Results for RFS and field duplicates for every analyte.



      Analyte	Avg	Min        Max       Count



                                       Min        Max       Count



                                       Min	Max	Count



                                       Min        Max       Count



                                       Min	Max	Count



                                       Min        Max       Count



                                       Min	Max	Count



                                       Min        Max       Count



                                       Min	Max	Count



                                       Min        Max       Count



                                       Min	Max	Count



                                       Min        Max       Count



                                       Min	Max	Count



                                       Min        Max       Count



                                       Min	Max	Count



                                       Min        Max       Count



                                       Min	Max	Count



                                       Min        Max       Count



                                       Min	Max	Count



                                       Min	Max	Count



                                       Min        Max       Count
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Analyte
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
Ava
                                          39

-------
8.  Check date ranges of data to see if they are reasonable.



      Analyte	Min	Max



      Analyte	Min	Max



      Analyte	Min	Max



      Analyte	Min	Max



      Analyte	Min	Max



      Analyte	Min	Max



      Analyte	Min	Max



      Analyte	Min	Max



      Analyte	Min	Max



      Analyte	Min	Max



      Analyte	Min	Max



      Analyte	Min	Max



9.  Check to verify units information looks alright.
10. Number of significant digits for each analyte.
11.  Number of negative values for each analyte.
12. Check flags on RFS and field duplicates.
                                              40

-------
13.  Core slice range (sediment)/species, age, length, weight (fish).
14.  Check blank correction, dilution, and surrogate correction fields.
15.  Questions about QC Coordinator remarks (RECSTAT). Check flags for whole record (RECSTATF).



    Questions about Station Notes (STNNOTES), Field Remarks (FREMARK), and Sample Description




    (SAMPDESC).
16.  Additional Questions.
                                            41

-------
Table A1.3.3. Printout of Information Stored in the LMMBP Tracking Database Related to Atrazine
Modeling (L:\Public\Access\lmmb\lmtrack2000.mdb)

               LMMBP DATA ARCHIVE - QUICK REPORT.  Note: All  Data Archived on
                                superior.grl.epa.gov in /usr/lmmbdata.
 Library No.


 Description


 Library No.

 Description


 Library No.

 Description



 Library No.

 Description



 Library No.

 Description



 Library No.

 Description



 Library No.

 Description



 Library No.

 Description



 Library No.

 Description
LM10001
PI:
David Schwab
Hourly  Lake Michigan wind, wave, and atmospheric data (5 km grid) for 1982, 1983, 1994, 1995.
Original data files were converted to SEDZL and POM formerly by M. Settles.  Also, bathymetric data
for Lake Michigan.
LM10002
PI:
William Richardson
STORET conventional and general chemistry data for Lake Michigan, April 1962-August 1993. Note:
Date range varies by parameter, includes original file, reformatted spreadsheet, and MS Access file.
LM10003
PI:
David Schwab
Two-dimensional and three-dimensional GLERL hydrodynamics data forthe Lake Michigan 5 km grid.
2D  data: January  1982-September  1983;  3D:  covers January-July  1982.   Program//llrssrv2
/~model/dev/PATRIC2D/RCS is for 2D processing, no three-dimensional programming yet.
LM10004
PI:
Steven Eisenreich
Open-lake (RULA) and tributary (RUTA), atrazine, DEA, DIA data forthe LMMBP.  Open-lake 325
samples (1/17/94-4/17/95).  Tributary: 126 samples (4/4/95-5/15/96).  Revised version of data sent
2/19/98.
LM10005
PI:
Angela Bandemehr
Hourly meteorological data (airtemperature, solarradiation, relative humidity, wind speed and direction,
and precipitation) from 13 air sampling sites both in and outside of the Lake Michigan basin. 11/30/90-
12/31/96 (Dates vary by site).
LM10006
PI:
Glenn Warren
Sea bird water temperature data for seven LMMBP surveys, April 1994-October 1995.  Data collected
at 0.5 m intervals. Does not include January 1994 survey.  Note: Data received was extensively
revised from original version.
LM10007
PI:
David Hall
Tributary flow data for 11 tributaries to Lake Michigan (Fox, Grand, Indiana Harbor,  Kalamazoo,
Manistique, Menominee, Milwaukee, Muskegon, Pere  Marquette, Sheboygan, St. Joseph), 1/1/94-
12/31/95. Some data estimated.
LM10011
PI:
David Schwab
Lake Michigan final report, hourly circulation, meteorology, and wave data (5 km grid) for 1982, 1983,
1994, 1995. Includes intake, cruise, mooring, water level data.  Also, HTML files and images, model
results (XDR format), Fortran and IDL programs.
LM10020
PI:
Keri Hornbuckle
Atmospheric atrazine and nutrient (NO3, total phosphorus, TKN) wet deposition loading data for Lake
Michigan 5 km grid cells used in hydrodynamic model. Atrazine wet deposition and particulate monthly
concentration data.  Data for 10/94-10/95 (nutrient) and 5/94-10/95 (atrazine).
                                                  42

-------
 Table A1.3.3.  Printout of Information Stored in the LMMBP Tracking Database Related to Atrazine
 Modeling (L:\Public\Access\lmmb\lmtrack2000.mdb) (Continued)
 Library No.

 Description


 Library No.

 Description
 Library No.

 Description



 Library No.

 Description
LM10022
PI:
David Hall
Atrazine, DEA, DIA tributary loading data for 11 monitored tributaries and atrazine data for unmonitored
tributaries to Lake Michigan.  Data covers the time period: 1/1/94-12/31/95.
LM10026
PI:
Nathan Hawley
Current velocity, water transparency, temperature from three stations, 10/31/94-10/11/95.  In situ
sediment resuspension from sediment flume  experiments (8/12/95-9/23/98).  Also profile  data -
temperature, dissolved oxygen, conductivity, BAT, pH, fluorescence, TSM data from six stations in
Lake Michigan (1/4/95-11/29/95).
LM10027
PI:
Barry Lesht
Current velocity and direction, bottom wave orbital velocity, temperature, beam attenuation, and TSM
data collected from Tripod Station 98 (latitude 42 52.18, longitude 87 42.41), during the EEGLE project,
4/2/98-12/1/98. Data collected every 30 minutes.
LM10028
PI:
Michael Settles
NEMA and NOAA wind speed and direction, wave height and period data for six stations in Lake
Michigan, retrieved from USACOE Web Site (http://bigfoot.wes.army.mil/c300.html).  1980-1998 (not
all stations cover entire date range). NEMO-Daily data, NOAA-Hourly data.
Table A1.3.4. Generalized Format for the LMMBP Water Data to be Analyzed With IDL Programs
Beginning -
Ending Columns
1 -7
8-8
9- 14
15- 15
16-22
23-23
24-35
36 -36
37-44
45 -45
Variable Description
Cruise Name
Blank Space
Latitude (ddd.ddd)
Blank Space
Longitude (-ddd.ddd)
Blank Space
Station Name
Blank Space
Depth Sampled
Blank Space
Format (A = Alpha, F
= Floating Point No., I
= Integer, X = Skip)
A7
1X
F6.3
1X
F7.3
1X
A12
1X
F8.0
1X
Sort Order (A =
Ascending, D =
Descending,
Blank = None)
A
N/A

N/A

N/A
A
N/A
A
N/A
Missing Data
Code
Blank
N/A
Blank
N/A
Blank
N/A
Blank
N/A
Blank
N/A
                                                  43

-------
Table A1.3.4. Generalized Format for the LMMBP Water Data to be Analyzed With IDL Programs
(Continued)
Beginning -
Ending Columns
46-53
54-54
55-58
59 -59
60 -67
68 -68
69-72
73 -73
74-75
76 -76
77-79
80 -80
81 -88
89- 103
104-111
112-126




Variable Description
Sampling Start Date
(mm/dd/yy)
Blank Space
Sampling Start Time (24
hour clock)
Blank Space
Sampling End Date
(mm/dd/yy)
Blank Space
Sampling End Time (24
hour clock)
Blank Space
Filter Fraction
Blank Space
Sample Type
Blank Space
Value Parameter 1
Parameter 1 Flags
Value Parameter 2
Parameter 1 Flags
1
f
Value Parameter n
Parameter n Flags
Format (A = Alpha, F
= Floating Point No., 1
= Integer, X = Skip)
A8
1X
A4
1X
A8
1X
A4
1X
A2
1X
A3
1X
F8.0
A15
F8.0
A15


F8.0
A15
Sort Order (A =
Ascending, D =
Descending,
Blank = None)
A
N/A

N/A
A
N/A

N/A
A
N/A
D
N/A








Missing Data
Code
Blank
N/A
Blank
N/A
Blank
N/A
Blank
N/A
Blank
N/A
Blank
N/A
-9999
Blank
-9999
Blank


-9999
Blank
                                         44

-------
and make it easier for users to find the information.    Sukloff, W.B., S. Allan, and K. Ward. 1995.  RDMQ
Incorporation of this information into LLRS Microsoft       User Manual.  Environment Canada, Atmospheric
Access databases has given us flexibility in retrieving       Environment  Service,  North  York,  Ontario,
the information needed by the modeling staff at the       Canada.  91 pp.
LLRS.

References

Rossmann, R.  (Editor).  2006. Results of the Lake
   Michigan Mass Balance Project: Polychlorinated
   Biphenyl Modeling Report. 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.   EPA/600/R-
   04/167, 579 pp.
                                              45

-------
                                          PART1
                                    INTRODUCTION
Chapter 4.   Representativeness  of  the
Lake  Michigan  Mass   Balance  Project
(LMMBP)   Years   Relative   to    Lake
Michigan's Historic Record

Ronald  Rossmann,  Kenneth  R. Rygwelski,  and
Russell G. Kreis, Jr.
United States Environmental Protection Agency
Office of Research and Development
National Health and Environmental Effects
  Research Laboratory
Mid-Continent Ecology Division
Large Lakes and Rivers Forecasting Research
  Branch
Large Lakes Research Station
9311 Groh Road
Grosse lie, Michigan 48138
and
Gregory  J.  Gerstner, Xiaomi Zhang,  and Brent
Burman
Welso Federal Services, LLC
Large Lakes Research Station
9311 Groh Road
Grosse lie, Michigan 48138

1.4.1 Introduction

A major concern related to modeling contaminants in
the lake was the representativeness of the years of
sampling (1994-1995) relative to the historical record.
This  was particularly important when using the
models to predict future conditions in the lake.  The
LM2-Atrazine  and  LM3-Atrazine   models  used
hydrodynamic model output from  1994-1995 in their
construct  (see  Parts 4 and  5  of  this  report,
respectively, for  more information).   In  addition,
atrazine loading estimates for any given year can be
affected by various meteorological  conditions (see
Part 2, Chapters 2 and 3 of this report). If these data
did  not represent something  close  to  average
conditions, the resulting predictions could be biased.
Parameters  considered  most  important  to  the
performance of the  atrazine  models included  ice
cover, airtemperature, water temperature, lake water
levels, precipitation, tributary flows, wind, and waves.
Potential impacts  on  the  various  models  are
discussed below.  Each of these were investigated
for the representativeness of the 1994-1995 project
data relative to the available historical data record.

1.4.2 Ice  Cover

Ice cover impacts the volatilization,  absorption, and
physical mixing of the lake during the winter months.
In locations where there is ice cover, gas exchange
between the water and  atmosphere is prevented by
the  physical barrier.  Physical mixing includes not
only the mixing of the  water column, but also the
interaction  of  waves  with  the lake  bottom to
resuspend sediments. Winters having extensive ice
cover yield a more poorly mixed water column, and
a large region of the lake becomes depositional due
to the lack of wave resuspension of sediments. Once
ice  retreats in  the spring, sediments accumulated
during ice cover will be resuspended as a pulse.  Ice
cover can  cause significant changes  in  winter
circulation patterns in a large lake (Campbell et a/.,
1987). The years  of interest were 1982, 1983, 1994,
and 1995.  The  hydrodynamic  modeling  included
three-dimensional lake circulation,  surface flux for
atmospheric input, and wind-wave models  (Schwab
                                              46

-------
and Beletsky, 1998). These were calibrated for the
period  of  1982-1983 using temperature,  current,
water level,  and wind-wave measurements.   The
calibrated  model was  applied to 1994-1995 and
verified. There was no ice modeling component for
the version  of  the hydrodynamic model  applied.
Thus ice cover was important for understanding any
potential   weaknesses   associated  with   the
hydrodynamic results as well as the dynamics of
exchanges between the water and the atmosphere.

Ice cover  data  were available  from  the National
Oceanic and Atmospheric Administration (NOAA)/
Great Lakes Environmental Research  Laboratory
(GLERL) (Assel, 2003).  This data set is  partially
described in Assel et a/. (2002). Tabular information
presented  in Assel (2003)  were summarized in a
manner that  seemed appropriate for this discussion
(Table 1.4.1). For the period when ice was recorded
on Lake Michigan, the mean  and median daily ice
cover were 16.7% and  14.7%, respectively. An ice
year began with the first ice. For example, 1982 may
include December of 1981.   Both 1982 and  1994
were greater than the mean and median; whereas
1983 and 1995 were less than the mean and median.
None of the  four years represented an  extreme of
mean daily ice cover.  The lowest mean daily ice
cover was observed in 2002, and the highest was
observed  in  1977.    Results  for each  winter's
maximum daily ice cover were similar to mean daily
ice cover.  Ice cover is extremely variable from year-
to-year.    The  impact upon hydrodynamics  as
modeled was believed to be minimal with respect to
1983 and  1995 when  ice cover was quite low.
Though high  ice  cover occurred during the winters of
1982 and 1994,  these periods were not a part  of the
hydrodynamic   model  period.    Using   the
hydrodynamic model information for models used to
predict future conditions  could  lead  to  potential
errors. Modeled circulation patterns could be in error
and impact a high bias to modeled current velocities
during the winters of high ice cover years due  to the
lack of an ice model within the hydrodynamics model.

1.4.3  Water and Air Temperatures

Water and air temperature data were retrieved from
the National  Data Buoy Center (U.S. Department of
Commerce, 2002). Data from buoy numbers 45002
(north buoy) and 45007 (south buoy) were reviewed
(Figure  1.4.1). Water temperature  sensors  were
located  1  m  below  the  water surface, and  air
temperature sensors were located 4 m above the
surface.  Water and air temperature  data  were
available 1979 through 2002 for the north buoy and
1981 through 2002 for the south buoy.

Water temperature  is highly variable from year-to-
year.  The data had been stratified in two ways for
presentation.   First, monthly mean temperatures
were calculated and plotted for the south  (Figure
1.4.2) and north (Figure  1.4.3) buoys.  Years of
importance  to  the  hydrodynamic  model  were
highlighted. It was interesting to note that 1983 and
1995 had higher monthly  mean temperatures than
1982 or 1994.  Both 1983 and 1995  had above
normal  maximum  mean  monthly  temperature;
whereas, 1982 had atypical maximum and 1995 had
a very low  maximum.  This was reflected in the
previously  discussed  ice cover  for the  four years.
Water temperatures tended to be  higher at the
southern buoy than at the northern buoy, reflecting its
more southerly latitude.

One way to identify the relative lake warming rate
among years was to look at the mean June  water
temperature for the period of observation available
from the NOAA buoys. Mean June temperatures at
the south (Figure 1.4.4) and north (Figure  1.4.5)
showed similar patterns that were  quite  interesting.
Beginning  in  1983,  relatively  high mean  June
temperatures were observed every four years (1983,
1987,1991, 1995,1999). This cycling, as well as the
apparent increasing mean June water temperature
for  the period  of  record, should   be  further
investigated. Both of these trends can impact long-
term model  forecasts.   The years  of the  Lake
Michigan Mass Balance Project (LMMBP) (1994 and
1995) represented  a fairly average mean  June
temperature and one of the relatively high  means,
respectively.

The exchange of atrazine between the air and  water
are dependent on both water and air temperatures.
Air temperature varied from year-to-year at the south
and north buoys (Figures 1.4.6 and 1.4.7). Because
air temperature drives observed water temperatures,
it was  not surprising that patterns  observed and
                                              47

-------
Table 1.4.1. Summary of Lake Michigan Ice Cover Based Upon Assel (2003)


Year
1973
1974
1975
1976
1977
1978
1979
1980
1981
|~ 1982
!_ 1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
|~ 1994
!_ 1995
1996
1997
1998
1999
2000
2001
2002
Mean
Median
Minimum
Maximum
Mean Daily
Percent Ice Cover
During Ice Period
13.3
16.9
13.9
15.5
46.5
26.6
35.2
18.2
24.6
24.0
8.2
15.6
20.1
25.3
9.1
16.6
13.1
17.5
10.0
8.3
11.0
27.3
7.2
19.4
13.4
6.1
8.7
9.2
13.4
6.0
16.7
14.7
6.0
46.5
Days of
Observed
Ice
104
122
113
119
132
132
132
106
112
135
118
127
119
126
100
104
140
132
120
149
126
134
120
161
156
109
111
103
134
116
124
121
100
161
Maximum
Daily Percent Ice
Cover
33.0
39.4
28.1
29.5
93.1
66.6
92.3
38.6
53.8
60.2 |
23.6 !
43.3
41.3
66.8
19.3
32.7
30.9
32.4
21.5
32.8
32.2
82.7 |
21.6 !
75.0
37.8
15.1
23.0
27.2
29.5
12.4
41.2
32.8
12.4
93.1
                                          48

-------
Figure 1.4.1.  Location of the NOAA buoys  in
Lake Michigan.

        ij
Figure 1.4.2.  Monthly mean water temperatures
in southern Lake Michigan.
                                                   20-
                                                    1980
                                                            1985
                                                                    1990
                                                                            19§S
                                                                                    2000
                                                                                           2005
                                                 Figure 1.4.3.  Monthly mean water temperatures
                                                 in northern Lake Michigan.
                                                   20
                                                  o
                                                  I 12
                                                    4


                                                    0 '  •
                                                    1980
                                  south buoy
                                    June
           1985
1990
1995
2000
                                      2005
Figure 1.4.4. Mean June water temperatures in
southern Lake Michigan.
                                                  o
                                                  a>

                                                  | 12

                                                  Q.

                                                  S 8
4 '


0
1980
                                                                                   north buoy
                                                                                    June
                                                           1985
                  1990
       1995
       2000
       2005
Figure 1.4.5. Mean June water temperatures in
northern Lake Michigan.
                                              49

-------
   25
   20:

 015:
 ,i=
 <0
           south buoy
         _«_ data years for Lake Michigan
           hydrodynamic modeling
                                                    20
   -5-
   1980
   1986
 1990
 1995
 2000
 2005
Figure 1.4.6. Monthly mean air temperatures in
southern Lake Michigan.
  25:

  20:
 -15-
  north buoy
_•- data years for Lake Michigan
  hydrodynamic modeling
   1980
  1985
1990
1995
2000
2005
Figure 1.4.7. Monthly mean air temperatures in
northern Lake Michigan.
conclusions  made for water temperature  are  the
same for air temperature.  The cyclic pattern of June
mean water temperatures was also found for the air
temperatures (Figure 1.4.8 and 1.4.9). As additional
data become available, future modeling efforts will
need to address these cyclic patterns and long-term
temperature trends for water and air temperatures.

1.4.4  Lake Water Levels

Lake levels can affect model geometry.  If segment
volume deviates significantly from the volumes used
at the  time  of calibration,  model results  can be
impacted.  On a percentage basis, the impact will be
                                                    16-
                                           ,12-
                                                     8
                                                     4
                                                                            south buoy
                                                                              June
                                                     1980
                                                     1985    1990     1995     2000    2005
                                                  Figure 1.4.8.   Mean June air temperatures  in
                                                  southern Lake Michigan.
                                                      1980
                                                      1985     1990     1995     2000    2005
Figure 1.4.9.   Mean June air temperatures  in
northern Lake Michigan.
                                          most noticeable for shallow water segments and
                                          predictions  from  the hydrodynamic  model  and
                                          surface water  model could be affected.  Monthly
                                          mean lake water levels varied between 175.5 and
                                          177.5 m for the period of record (1918-1997).  Lake
                                          levels during 1994 and 1995 were near the average
                                          for the period of record (Figure 1.4.10).

                                          1.4.5 Precipitation

                                          Precipitation   influences   the  flux  of  airborne
                                          contaminants to the lake, impacts tributary loading
                                          rates, and controls water levels.  The 1982 and 1983
                                          hydrodynamic years, and the 1994 and 1995 project
                                          years were compared to the previous 50 years of
                                          data (Croley and Hunter,  1994).
                                               50

-------
               177,5   ---.-
Lake Michigan and Lake Huron water levels (1918-1997)
date; US Army Corps of Engineers, Detroit District
.............. monthly mean water levels (meters - IGLD 1985)
	all-time, record monthly high and low water levels
               175,0-
                    i I I i i I I I i f I I I ; [ I I I i j I I I i i I i i j i i I i i i I I i i I I I • i I I I i i I I i i i I I i i i I I i i i i i i i i i i : M I I ! M

                   QG3    0>O3iJ)Q5GJ05ffi0>0>O'JO5
Figure 1.4.10. Record of mean monthly water levels for Lake Michigan.
1.4.5.1 Annual Comparisons

Precipitation to Lake Michigan for 1982,1983, 1994,
and 1995 were close to the 50-year mean for the lake
(Figure 1.4.11).  1982 and 1983 were slightly above
the mean and 1994 and 1995 were slightly below the
mean. 1995 total annual precipitation was very close
to the 50-year mean for over-lake precipitation. No
visual trend was apparent in the total annual amounts
of precipitation over the 50-year period.

1.4.5.2 Monthly Comparisons

The  monthly mean  precipitation for 1982,  1983,
1994, and 1995 were compared to the 50-year mean
for the period of 1949 through 1998 (Figure 1.4.12).
For the years of interest, January, July, November,
and December of 1982; May of 1983; and October of
1995 had relatively high  amounts of precipitation,
exceeding  one  standard  deviation of the 50-year
mean.  For the four years of interest, February of
1982; June of 1983; March, May, and December of
1994; and June of 1995 had relatively low amounts
of precipitation.  This illustrates that, in any one year,
precipitation varies from month-to-month while the
precipitation for the  year can be at  or  near the
average expected.
                           1.4.6 Tributary Flows

                           Tributary flows impact the delivery of materials to the
                           lake,  including nutrients and contaminants.  During
                           high flow events triggered by spring snow melt or rain
                           events, tributary flows increase and materials can be
                           carried from the watersheds to the tributaries. Within
                           thetributary, sediments containing contaminants may
                           resuspend.  Thus the fluxes of solids, nutrients, and
                           contaminants to  the lake have the  potential  to
                           increase during  high flow events.   Tributary flows
                           were  obtained from the United States Geological
                           Survey (USGS) website (www.usgs.gov). A historical
                           average  and median daily flow were calculated for
                           each  tributary for the period of record, as well as for
                           the 1994-1995 and 1982-1983 time periods.  During
                           1982  and 1983,  tributary flows were approximately
                           20%  greater than the average flow (Figure 1.4.13).
                           The 1994-1995 time period had relatively ordinary
                           tributary flows (Figure 1.4.14).

                           1.4.7 Summary

                           Lake  Michigan is acted upon by a number of physical
                           parameters that impact the physics, chemistry, and
                           biology of the lake.   For a lake the size of Lake
                           Michigan,  changes  in these  parameters  can be
                                               51

-------
            12000
            10000-
             80QQ-
          o
          I
          I  6000 -
             4000-
             2000-
Figure 1.4.11. Annual precipitation to Lake Michigan between 1949 and 1998.
           1400 -i-
           1200
           1000
       £
       "o
       2   800 -
           600
           400 -
           200 -
  Monthly mean precipitation
-*— 1949-1998 mean
-•- - 1949-1998 mean + 1 std.dev.
-t—1949-1998 mean -1 std.dev,
 o  1982
 A  1983
 D  1994
•o  1995
                JAN    FEB    MAR   APR    MAY    JUN    JUL   AUG    SEP    OCT   NOV    DEC
Figure 1.4.12. Comparison of 1982,1983,1994, and 1995 monthly mean precipitation to the mean for
the period of 1949 through 1998.
                                                  52

-------
            5000-
            4000-
          -g 3000-
            2000-
            1000-
               Ftn
II
                    6
                    o
                    K
                            .5
                            1
                                      e-
                                      TO
                                      o
                                      •5
                                                               1
                                                                      fO
                                 a.
                                 IS
                                      O  15
          o

          ui
Figure 1.4.13. Comparison of tributary flow for hydrodynamic model calibration (1982-1983) to the
historic means.
            6000
            5000-
            4QQQ-
            3000
            2000-
             1000-
                f%-n
                              =
                              a
                                        S
                                        ef
                                        =
                            1
i  2
                                                            o
                                                                »
                                                                      1  I
Figure 1.4.14. Comparison of tributary flow for the study period (1994-1995) to the historic means.
                                               53

-------
significant, especially when models are used in long-
term predictions to predict the outcome of various
scenarios. The primary driving forces are wind, air
temperature,  and  precipitation.   These impact
tributary flows, lake levels, waves, water circulation,
water temperature, and ice cover.  For the period of
record, these driving forces varied from year-to-year.
The period of 1982 to 1983 was used to calibrate the
hydrodynamic models. Fortunately for the period of
time the models were calibrated, conditions were not
at any extreme. This was also true for the period of
1994 and 1995 when the models were applied.
However, the impact of ice cover remains a concern
and will have to be dealt with in the future.

Temperature  can  impact  atrazine  contaminant
modeling.  Air temperature impacts how quickly the
lake warms in any one year.  Water temperature
impacts the volatilization of atrazine. There appears
to be a four-year  cycle of quicker warming which
exists within a trend of general warming of the lake.
The trend of warming may be part of  a longer-term
undocumented cycle  or may  be  related to climate
change.  For  future  modeling, these cycles and
trends  will have to be considered to improve long-
term predictions.

Precipitation will impact both lake levels and tributary
flows.  In wet years, more atrazine may be delivered
to the lake (see Part 2, Chapter 2). Precipitation was
within the normal  range  for all years of modeling
interest, resulting in lake levels and tributary flows
that were within normal bounds.   Changes in lake
levels  as well as  the response of  tributaries  to
precipitation events will need to  be considered for
future  modeling   used  to   predict  changes  of
contaminants within the lake.
References

Assel, R.A., D.C. Norton, and K.C. Cronk.  2002. A
   Great Lakes Digital Ice Cover Data  Base for
   Winters  1973-2000.    National  Oceanic  and
   Atmospheric  Administration,  Great  Lakes
   Environmental Research Laboratory, Ann Arbor,
   Michigan. NOAA Technical Memorandum  ERL
   GLERL-121,46pp.

Assel, R.A.  2003. NOAA Great Lakes Ice Atlas. An
   Electronic Atlas of  Great Lake  Ice Cover.
   National   Oceanic   and   Atmospheric
   Administration,   Great   Lakes  Environmental
   Research Laboratory, Ann Arbor, Michigan.

Campbell, J.E., A.M. Clites,  and G.M. Green. 1987.
   Measurements of Ice Motion in Lake Erie Using
   Satellite-Tracked Drifter Buoys. National Oceanic
   and  Atmospheric Administration,  Great Lakes
   Environmental Research Laboratory, Ann Arbor,
   Michigan. NOAA Technical Memorandum  ERL
   GLERL-30, 22 pp.

Croley, T.E., II and T.S. Hunter.  1994.  Great Lakes
   Monthly Hydrologic Data.  National Oceanic and
   Atmospheric  Administration,  Great  Lakes
   Environmental Research Laboratory, Ann Arbor,
   Michigan. NOAA Technical Memorandum  ERL
   GLERL-83, 13 pp.

Schwab, D.J. and D. Beletsky. 1998. Lake Michigan
   Mass Balance Study: Hydrodynamic  Modeling
   Project.   National  Oceanic  and  Atmospheric
   Administration,   Great   Lakes  Environmental
   Research Laboratory,  Ann  Arbor,  Michigan.
   NOAA Technical Memorandum ERLGLERL-108,
   55 pp.

U.S.  Department of Commerce.  2002.  National
   Data Buoys. National Weather Service, National
   Oceanic and Atmospheric Administration,  Ann
   Arbor, Michigan. Available from  National Data
   Buoy Center at http://www.ndbc.noaa.gov.
                                              54

-------
                                          PART1
                                    INTRODUCTION
Chapters. Atrazine Modeling Overview

Douglas D. Endicott
Great Lakes Environmental Center
Traverse City, Michigan
and
William R. Richardson (Retired), Ronald Rossmann,
and Kenneth R. Rygwelski
United States Environmental Protection Agency
Office of Research and Development
National Health and Environmental Effects
  Laboratory
Mid-Continent Ecology Division
Large Lakes and Rivers Forecasting Research
  Branch
Large Lakes Research Station
9311 Groh Road
Grosse lie, Michigan 48138

1.5.1 Background

The atrazine mass balance project was based upon
the   Enhanced  Monitoring  Program  (EMP),   a
comprehensive,   two-year synoptic   survey  for
selected toxic  chemicals in  the Lake Michigan
ecosystem. The atrazine EMP included tributary load
and  atmospheric  deposition  monitoring;  ambient
water column;  and  additional  measurements  to
define and confirm transport  and fate  processes.
The   project  was  led  by  the  United  States
Environmental  Protection  Agency (USEPA)/Great
Lakes National Program Office (GLNPO). Modeling
support  to   the  project  was  provided  by  the
USEPA/Mid-Continent Ecology Division (MED)ADffice
of Research and Development (ORD)/Large Lakes
Research Station  (LLRS)  in cooperation with  the
Atmospheric Research and Exposure Assessment
Laboratory  (AREAL); the  National  Oceanic and
Atmospheric Administration (NOAA)/Great  Lakes
Environmental Research Laboratory (GLERL); and
other cooperators. The research developed a suite
of integrated mass balance models to simulate the
transport and fate of atrazine in Lake Michigan.

The project directly supports the development of a
Lake-wide  Management  Plan  (LaMP)  for Lake
Michigan that is mandated  under Section 118 of the
1992 Clean Water Act.  Atrazine and degradation
products are on the Lake Michigan LaMP 2008
Watch  List.  Chemicals on the Watch List include
those chemicals that have the potential to impact the
Lake Michigan ecosystem; is present in the Lake
Michigan watershed; and has the potential  for
bioaccumulation, persistence in water or sediment, or
toxicity singly or through synergistic effects.  In  a
June 1993 response to an inquiry  by U.S. Senator
Carl Levin, the United States General Accounting
Office (now called the United States Government
Accountability Office  [USGAO])  recommended that
the  USEPA assess the persistence of  pesticides,
such as atrazine, in the Great Lakes and to  report
their findings to the pesticides reregistration program
(U.S. General Accounting Office, 1993). The results
of the Lake Michigan Mass Balance Project (LMMBP)
atrazine modeling have   been  forwarded to  the
reregistration program for consideration.

1.5.2  LMMBP Modeling Objectives

Development of  effective  strategies  for  atrazine
management requires a quantitative understanding
of the relationships between sources, inventories,
                                             55

-------
concentrations,  and  effects  of  atrazine in  the
ecosystem.     This   approach   integrated   load
estimation, ambient monitoring, and research efforts
within a modeling  framework that was compatible
with   both  scientific  as  well  as   ecosystem
management  objectives.    The mass  balance
approach  estimated the magnitude of  mass fluxes
that constitute the  pathways for atrazine transport
into and out of the lake and processes that distribute
toxics within the lake water column.   Based upon
these estimates, the  mass balance was used to
determine the rate  of change in concentrations and
inventories of atrazine as inputs such as  atmospheric
and tributary loadings changes.   Thus the  mass
balance can serve as a useful tool to estimate or
predict  the   outcome  of  alternatives  under
consideration for toxics management.

In general, atrazine modeling efforts associated with
the LMMBP had the following objectives:

1. Provide a consistent framework  for integrating
   load estimates, ambient monitoring data,  and
   research   efforts,   leading  to   a   better
   understanding  of atrazine chemical  sources,
   transport, and fate in Lake Michigan.

2. Using  flow  and concentration measurements,
   estimate the loading of atrazine from all major
   tributaries to Lake Michigan for the duration of
   the study.

3. Based on county-level usage of atrazine within
   the  basin,  make  independent  estimates  of
   atrazine loading to the lake via tributaries.

4. Estimate  the atmospheric deposition  and  air-
   water exchange of atrazine, including spatial and
   temporal variability over Lake  Michigan.

5. Calibrate and confirm mass balance models for
   atrazine using project data based upon models
   for hydrodynamic and atrazine transport and fate.

6. Based upon the mass balance models, evaluate
   the magnitude  and variability of toxic chemical
   fluxes  within and between lake compartments,
   especially between the water column and  the
   atmosphere.
7.   Apply the calibrated  mass balance models to
    forecast  atrazine  concentrations  in  water
    throughout   Lake   Michigan   based   upon
    meteorological  forcing  functions and  future
    loadings based upon load reduction alternatives.

8.   Predict the water concentration of atrazine in
    lake model cells receiving loads from tributaries
    contributing a relatively high percentage of the
    total tributary load to the lake. Compare these
    predictions to water quality standards.

9.   Estimate (quantify) the uncertainty associated
    with estimates  of  tributary and  atmospheric
    loads of atrazine  and  model  predictions  of
    contaminant concentrations.

10. Identify   and  prioritize  further  monitoring,
    modeling, and research  efforts to (1) further
    reduce uncertainty and improve  accuracy of
    predictions; (2) establish additional cause-effect
    linkages, such as ecological risk endpoints and
    feedbacks; and (3) evaluate additional source
    categories, such as non-point sources in  the
    watershed.

Unlike  the   other   LMMBP-modeled  toxics
(polychlorinated  biphenyls  (PCBs),  mercury,  and
frans-nonachlor), atrazine does not sorb to solids to
any great extent, and it  does not bioaccumulate. It
is soluble in water and can migrate from farm fields
to Lake Michigan via run-off events. The herbicide is
also transported to Lake Michigan via atmospheric
pathways.

1.5.3 Historical Modeling

The modeling design  and approach for the LMMBP
reflects  a  progression of prior modeling efforts in
Lake  Michigan and throughout the Great Lakes.
These include eutrophication and toxic substance
mass balance models,  food web  bioaccumulation
models,  and predictive hydrodynamic and sediment
transport models.  Although not a comprehensive
review,  several of these prior modeling efforts  are
discussed below.
                                               56

-------
1.5.3.1 Completely-Mixed Lakes-ln-Series Model

A lakes-in-series model for conservative substances
was developed by Sonzogni et al. (1983) and applied
to  forecast chloride concentrations in each of the
Great  Lakes as  a function  of  expected future
loadings.     This  model  demonstrated  that
concentrations of non-reactive substances would
substantially "lag" the history of their input. This was
especially the  case for Lake Michigan, where
maximum chloride concentrations were not predicted
to  occur until the 22nd  Century despite declining
loads after the 1970s. Similarly strong, non-steady-
state behavior may be expected for other chemicals
which are non-reactive and weakly  associated to
particles.

1.5.3.2  MICHTOX

MICHTOX was adapted from the general model,
WASP4 (Ambrose et al., 1988), and has served well
as a screening-level model for Lake Michigan over
the past several  decades.   An  integrated mass
balance  and  bioaccumulation model  for PCBs
(modeled  as two  homologs)  and 10  other  toxic
chemicals was developed as a planning tool for the
LMMBP(Endicottefa/.,2005). The MICHTOX mass
balance model was calibrated to suspended solids
and plutonium data for the southern lake basin, while
the bioaccumulation model combined Thomann and
Connolly's  (1984)  effort  with chemical-specific
parameterization from Lake Ontario.   MICHTOX
demonstrated that  reasonable  predictions of  PCB
concentration trends in water, sediment, and  biota
could be developed although significant uncertainties
regarding sediment-water and air-water contaminant
transport remained.  These are the most significant
transport fluxes for PCBs and presumably other
hydrophobic contaminants. Major data gaps for other
priority  toxics  allowed  only  order-of-magnitude
estimates of load-concentration relationships. When
this  model  was  developed  and  run,  available
monitoring data for toxic chemical  concentration in
tributaries, air, lake water, sediment, and biota were
not adequate to define loading trends or to relate the
distribution of loadings  to contaminant  gradients
observed for sediment and  biota.  Credible model
predictions of toxic chemical transport, fate, and
bioaccumulation would depend upon developing a
comprehensive  data  set  quantifying  loadings,
sediment inventories, concentrations, and transport
fluxes on a  spatially-resolved basis  and localized
descriptions  of food web structures.

MICHTOX was also applied to model atrazine in Lake
Michigan and Green  Bay.  It was first applied prior to
the release of LMMBP data using only historical data
(Rygwelski et al., 1999),  and it was also  applied
again  after  LMMBP  data  became  available.
MICHTOX served as a low-resolution  model and the
application is discussed in this report.

1.5.3.3 Green Bay Mass  Balance Project

The  Green  Bay  Mass  Balance Project (GBMBP)
demonstrated  the   feasibility of applying  mass
balance principles to manage toxic chemicals in the
Great Lakes ecosystem.  A two-year (1989-1990)
synoptic sampling program was designed to collect
appropriate and complete data for the mass balance
study.   A suite  of  integrated mass balance and
bioaccumulation  models  were developed  which,
together, provided an ecosystem-level simulation of
sources,  transport,  fate,  and bioaccumulation  of
PCBs throughout the  Fox River and Green Bay.
These mass  balance models were also based on the
general  WASP4  model  construct.    This study
advanced  the state-of-the-art  of mass  balance
modeling, particularly the ability to construct a fairly
complete and accurate description of contaminant
mass transport.

Several  aspects  of the  Green Bay modeling effort
were noteworthy.  Particle transport and sorption
processes were  found  to  be  of  fundamental
importance   as bases for contaminant modeling.
Resuspension of contaminated sediments in the Fox
River constituted the major source of PCBs to the
river as well as the bay. In the bay, particle sorbent
dynamics were strongly affected by  phytoplankton
production and decay.  The relative significance of
hydraulic transport,  sediment  transport,  burial,
volatilization,  and open-lake boundary exchange
processes upon  the PCBs mass balance varied
considerably  with   location   in   Green   Bay.
Radionuclide tracers  were  again   essential  for
calibration of particle fluxes and confirmation of long-
term  contaminant   transport  predictions.    The
significance of contaminant accumulation at the base
of the food web  and fish  movement in relation to
                                              57

-------
exposure  gradients  were  demonstrated  in  the
bioaccumulation model. The LMMBP demonstrated
the linked submodel approach to ecosystem model
development and  application,  and the feasibility of
using such a model for assessing the effectiveness
of toxics management control alternatives.

The GBMBP models were a precursor to our LM2-
Atrazine model.  LM2-Atrazine served as our mid-
level spatial resolution  atrazine model, and  the
application is discussed in this report.

1.5.4  Resolution  for the LMMBP Models

Model resolution is the spatial and temporal scale of
predictions, as well as the definitions of model state
variables.  While  factors  such as data availability,
model  sophistication,  and  computer  resources
constrain resolution to a degree, different levels of
                model resolution are possible and  are,  in fact,
                necessary.   Three "levels" of spatial resolution,
                indicated  by the segmentation  grid  of the  lake
                surface, are illustrated in Figure 1.5.1.  Level 1 was
                resolved at the scale of lake basins (characteristic
                length, L  =  150 km) with  an associated seasonal
                temporal  resolution.   This was  a screening-level
                model resolution used in MICHTOX.   Level 2 was
                resolved at a regional scale defined by food webs (L
                = 40 km); temporal resolution was weekly-to-monthly.
                This  resolution was roughly comparable  to that
                achieved by models developed in the GBMBP.  Level
                3 was a hydrodynamic scale resolution (L = 5 km),
                with associated daily temporal resolution. Both near-
                shore and offshore regions can be distinguished with
                the Level 3 resolution.  Level 3 was scaled to resolve
                to predict  hydrodynamic transport.
  LEVEL 1 - MICHTOX
  (Screening)
  6 surface segments
  9 water segments
LEVEL2-LM-2

10 surface segments
41 water segments
LEVEL 3 - LM-3
(High resolution 5km X 5km grid)
2318 surface segments
44.042 water segments
19 sigma layers
Figure 1.5.1.  Surface water segmentation for alternative Lake Michigan mass balance model levels.
                                              58

-------
Although the LaMP and the Great Waters Program
(GWP) objectives are "lake-wide,"  both  of these
emphasize biotic impairments occurring primarily in
localized, near-shore regions.  LaMP objectives also
require  that the transport of contaminants from
tributaries and other near-shore sources to the open-
lake be resolved. Therefore, the Level 1 model was
not adequate for  the  study  objectives.  Level 2
resolution   was  adequate   for  most  modeling
objectives   but  not  for  resolution  of significant
hydrodynamic impact  or  near-shore influence  of
atrazine from tributaries.  Level 3  resolution  was
required for accurate hydrodynamic modeling and
was desirable for  predicting near-shore gradients,
especially  those formed  by  transients  such as
thermal  bars and upwelling;  as well as  more
persistent features such as tributary plumes and
thermal  stratification.   Level 3 transport resolution
also has the potential in relating toxics loading from
the 10 Areas of Concern (AOCs) adjoining Lake
Michigan which must be addressed by the Remedial
Action Plan  (RAP) process to the  LaMP via the
LMMBP.

The modeling design for the LMMBP was based on
the development of a  number of models at three
levels of resolution.  For the  atrazine contaminant
transport and fate models, MICHTOX was resolved
at Level 1;  LM2-Atrazine was resolved  at Level  2;
       and LM3-Atrazine was resolved at Level  3.  The
       Princeton Ocean Model (POM) and  atmospheric
       loading models were resolved at Level 3. Results of
       the  hydrodynamic   model  were  spatially  and
       temporally averaged prior to coupling to the Level 2
       model.   The  rationale  for  specifying  different
       resolutions was the hydrodynamic models require a
       Level 3  resolution  to offer the best capability for
       transport simulation  and  forecasting.   A  lower
       resolution was specified for LM2-Atrazine because
       this model had been demonstrated at this resolution.

       1.5.5  Models Developed and Applied

       The transport and fate atrazine models developed,
       refined, and applied by the Large Lakes and Rivers
       Forecasting  Research Branch (LLRFRB)  included
       MICHTOX, LM2-Atrazine, and LM3-Atrazine (Figure
       1.5.2).   Models  developed and  run  elsewhere
       included  a  hydrodynamics model (POM) (Schwab
       and Beletsky, 1998), an atmospheric loading model
       based  on local observations (Green et a/.,  2000;
       Miller et  a/., 2001), a tributary loading model (Hall
       and Robertson, 1998), and the Community Multiscale
       Air Quality (CMAQ) model.  CMAQ was adapted to
       simulate  the regional atmospheric fate and transport
       of atrazine (Cooter et a/., 2002; Cooter and Hutzell,
              Hydrodynamic
              and
              load models
              Contaminant
              transport
              and fate
              models
              transport
            aggregated
             to level 2
LM2-Atrazine
   level 2
   model
                          POM
                      hydrodynamic
                         model
   advective/
   dispersive
    transport
  and bottom
 shear stress
LM3-Atrazine
   level 3
   model
                                       environmental exposure
                                            concentration
Figure 1.5.2.  Model construct used for the LMMBP to model atrazine.
                                              59

-------
2002).   This atmospheric model  utilized atrazine
emissions from agricultural soils provided by the
Pesticide Emissions Model (PEM) (Scholtz  et a/.,
1999;  and Scholtz et a/., 2002).   The  CMAQ
predictions of atrazine in air and rainfall compared
well with some field observations  taken  along the
Lake Michigan shoreline in 1995.  Although the
results  from the CMAQ were not used directly in the
any of  the  LMMBP  atrazine models, the  model
demonstrated  a potential for this purpose in future
modeling  efforts.   Only  the models developed,
refined, and applied at LLRFRB will be discussed in
detail within this document.

1.5.5.1   Lake Process Models

The  mass  balance models for atrazine in  Lake
Michigan were comprised  of linked hydrodynamic
(POM)  with LM2-Atrazine and LM3-Atrazine. The
hydrodynamic  model-predicted water  movements
necessary  to  describe  the  three-dimensional
transport of dissolved  constituents  in the  water
column,  and  these  transport  parameters  were
incorporated into  the water quality models.  The
benefit of using hydrodynamic model output in this
way relieves the modeler from having to use a tracer
in the water, such  as chloride, to calibrate advective
and dispersive transport functions.  More discussion
can be found on this topic in Part 4 (LM2-Atrazine)
and Part 5 (LM3-Atrazine).

MICHTOX was not linked in any way with the POM
hydrodynamic  model.  In MICHTOX,  circulation is
specified as  advective  and  dispersive  transport
functions.  This approach suffers the disadvantages
in that calibration of the transport functions requires
extensive  tracer data (chloride), circulation  is not
predicted by meteorologic forcing functions, and the
model  loses resolution because of the difficulty in
measuring/calibrating fine-scale transport variability.
In Green Bay, chloride data was used to calibrate the
transport functions. However, in the main lake, the
chloride gradients were not evident, and  therefore,
were of no value for the purpose of calibrating the
transport functions. MICHTOX vertical and horizontal
exchange coefficients were obtained from previous
Great Lakes modeling studies. See Section 3.3.2 in
the MICHTOX chapter for more discussion on this
topic.
The models described the contaminant transport and
fate within the water column, mass transfer between
media (air and water), and atrazine degradation via
total kinetic decay  processes.   Together,  these
models formed an integrated description of atrazine
chemical cycling in the aquatic ecosystem with which
to predict the relationship  between loadings  and
concentrations of atrazine in the lake.

1.5.5.2  Hydrodynamics (POM)

The Princeton Ocean Model (POM) (Blumberg and
Mellor,  1980, 1987) was used to compute  three-
dimensional current fields in the lake.  The POM
simulated large- and medium-scale (km) circulation
patterns, vertical stratification, velocity distribution,
seiche,  and surface waves.  This model was also
used to simulate a thermal balance for the lake.  The
POM is a primitive equation, numerical hydrodynamic
circulation  model  that  predicts  three-dimensional
water column transport  in response to wind stress,
temperature, barometric pressure, and Coriolis force.
The  POM  has been demonstrated  to  accurately
simulate the  predominant physics  of large  water
bodies (Blumberg and Mellor, 1983,1985; Blumberg
and  Goodrich, 1990).  This model was  used to
develop year-long simulations on a 5 km horizontal
grid with 19 sigma-coordinate vertical layers at one-
hour  intervals for Lake  Michigan  (Schwab  and
Beletsky,   1998).    Observed   and  simulated
meteorological data were used  to  define  model
forcing  functions.   Extensive  measurements of
temperature and current  distributions collected in
Lake  Michigan during  1982-1983  were  used to
provide  the necessary  data for model calibration;
measurements of water  temperature and current
distributions were used to  confirm  hydrodynamic
simulations for 1994-1995.

1.5.6  Model Quality Assurance

A  Quality Assurance  Project  Plan  (QAPP) was
prepared and implemented for the atrazine modeling
(Richardson et a/.,  2004).  The QAPP specified
procedures   for  code  development;  testing;
modification; documentation; as well as methods and
measures applied in model calibration, confirmation,
and uncertainty analysis.
                                              60

-------
1.5.7      Model   Application   and
Computational Aspects

1.5.7.1 Annual Simulations

Annual forecast simulations were run with the LM3-
Atrazine model.  Model input reflected seasonal,
regional, and lake-wide contaminant loads.  Model
output was analyzed within the  high-resolution of
spatial and  temporal  gradients  of  contaminant
concentrations.

1.5.7.2 Long-Term Simulations

MICHTOX and LM2-Atrazine long-term simulations
were used to forecast the lake-wide impact of various
management scenarios.  Forecasts were performed
to determine time to near steady-state conditions for
both continuing and discontinued loads. Forecasts
were  also run  to evaluate  reductions in exposure
concentrations  resulting from elimination of tributary
and/or atmospheric loading.

References

Ambrose, R.B., T.A. Wool, J.P. Connolly, and R.W.
   Shanz.  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.

Blumberg, A.F.  and D.M. Goodrich. 1990. Modeling
   of Wind-Induced Destratification in Chesapeake
   Bay. Estuaries, 13(3): 1236-1249.

Blumberg, A.F. and G.L Mellor.  1980.  A Coastal
   Ocean Numerical Model. In: J. Sunderman  and
   K.P. Holtz  (Eds.),  Mathematical  Modeling of
   Estuarine Physics, pp. 203-214, Proceedings of
   the   International   Symposium,   Hamburg,
   Germany, August 1978.

Blumberg, A.F. and G.L. Mellor.  1983.  Diagnostic
   and Prognostic Numerical Circulation Studies of
   the South Atlantic  Bight.  J. Geophys. Res.,
   88(C8):4579-4592.
Blumberg, A.F. and G.L. Mellor. 1985. A Simulation
   of the Circulation in the Gulf of Mexico. Israel J.
   Earth Sci., 34:122-144.

Blumberg, A.F. and G.L. Mellor. 1987. A Description
   of  a   Three-Dimensional  Coastal   Ocean
   Circulation Model.  In: N.S. Heaps (Ed.), Three-
   Dimensional Coastal Ocean Models, Coastal and
   Estuarine  Sciences,   pp.  1-16.   American
   Geophysical Union, Washington, D.C.

Cooter, E.J. and W.T.  Hutzell. 2002.  A Regional
   Atmospheric Fate  and  Transport  Model for
   Atrazine. 1. Development and Implementation.
   Environ. Sci. Technol.,  36(19):4091-4098.

Cooter, E.J., W.T. Hutzell, W.T. Foreman, and M.S.
   Majewski.  2002. A Regional Atmospheric Fate
   and Transport Model for Atrazine. 2. Evaluation.
   Environ. Sci. Technol.,  36(21):4593-4599.

Endicott,  D.D., W.L. Richardson, and D.J. Kandt.
   2005.  1992 MICHTOX:  A Mass Balance and
   Bioaccumulation Model for Toxic Chemicals in
   Lake  Michigan.    In:    R.  Rossmann  (Ed.),
   MICHTOX:      A   Mass  Balance   and
   Bioaccumulation Model for Toxic Chemicals in
   Lake  Michigan, Part  1.   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.   EPA/600/R-
   05/158,  140 pp.

Green,  M.L, J.V. DePinto, C.W.  Sweet, and K.C.
   Hornbuckle,    2000.   Regional  Spatial and
   Temporal  Interpolation of Atmospheric PCBs:
   Interpretation of Lake  Michigan Mass Balance
   Data. Environ. Sci. Technol., 34(9):1833-1841.

Hall,  D.  and D. Robertson.   1998.   Estimation of
   Contaminant  Loading  from   Monitored  and
   Unmonitored Tributaries to Lake Michigan forthe
   USEPA  Lake Michigan Mass  Balance Study.
   Quality  Systems  and  Implementation  Plan.
   Submitted October 23,1998. U.S. Environmental
   Protection  Agency,  Great  Lakes   National
   Program Office, Chicago, Illinois.  19 pp.
                                             61

-------
Miller, S.M., M.L.  Green, J.V. DePinto, and K.C.
   Hornbuckle.    2001.  Results  from the  Lake
   Michigan Mass Balance Study: Concentrations
   and  Fluxes of  Atmospheric  Polychlorinated
   Biphenyls and  frans-Nonachlor. Environ.  Sci.
   Technol., 35(2):278-285.

Richardson, W.L, D.D. Endicott, R.G. Kreis, Jr., and
   K.R. Rygwelski(Eds.). 2004. The Lake Michigan
   Mass Balance Project Quality Assurance Plan for
   Mathematical  Modeling.   Prepared  by  the
   Modeling  Workgroup.    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.   EPA/600/R-
   047018,233pp.

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.

Scholtz, M.T., B.J. Van Heyst,  and A. Ivanhoff.
   1999.  Documentation for the Gridded Hourly
   Atrazine  Emissions  Data Set for the  Lake
   Michigan  Mass   Balance   Study.     U.S.
   Environmental   Protection  Agency, Office of
   Research and Development, National Exposure
   Research Laboratory, Research Triangle Park,
   North Carolina.  EPA/600/R-99/067, 61 pp.
Scholtz, M.T., E. Voldner, A.C.  McMillan, and B.J.
   Van Heyst. 2002. A Pesticide Emission Model
   (PEM) Part 1:  Model Development.   Atmos.
   Environ., 36(32):5005-5013.

Schwab, D. and D. Beletsky.  1998. Lake Michigan
   Mass Balance Study: Hydrodynamic Modeling
   Project.   National Oceanic  and  Atmospheric
   Administration,   Great  Lake  Environmental
   Research Laboratory, Ann  Arbor,  Michigan.
   NOAA Technical Memorandum ERL GLERL-108,
   55 pp.

Sonzogni, W.C., W. Richardson, P. Rodgers, and
   T.J.  Monteith.  1983.  Chloride Pollution of the
   Great  Lakes.   Water Pollut. Contr.  Fed. J.,
   55(5) :513-521.

Thomann, R.V. and J.P.  Connolly.  1984. An Age
   Dependent Model of PCB in a Lake Michigan
   Food Chain.   U.S.  Environmental Protection
   Agency, Office of Research  and Development,
   Environmental  Research  Laboratory-Duluth,
   Large  Lakes  Research   Station,  Grosse lie,
   Michigan. EPA/600/S3-84/026, 3 pp.

U.S. General Accounting Office.  1993. Reporttothe
   Chairman,  Subcommittee  on  Oversight of
   Government   Management,  Committee  on
   Governmental Affairs, U.S. Senate: Pesticides-
   Issues Concerning Pesticides Used in the Great
   Lakes Watershed.   U.S.  General Accounting
   Office, Washington,  D.C.  GAO/RCED-93-128,
   39 pp.
                                              62

-------
                                         PART 2

           LAKE MICHIGAN MASS BALANCE PROJECT ATRAZINE
                          LOADINGS TO LAKE MICHIGAN
                                    Kenneth R. Rygwelski
                          United States Environmental Protection Agency
                              Office of Research and Development
                  National Health and Environmental Effects Research Laboratory
                                Mid-Continent Ecology Division
                       Large Lakes and Rivers Forecasting  Research Branch
                                 Large Lakes Research Station
                                      9311 Groh Road
                                  Grosse lie, Michigan 48138
Chapter  1.  Historical  Atrazine  Usage in
the United States

2.1.1 Background

The company, J.R. GeigyA.G., of Basel Switzerland
applied for a patent with the United States Patent
Office on January 12, 1955 that described a method
for making atrazine and listed various mixtures with
the active ingredient that could be used to kill weeds
(U.S. Patent Office, 1959).  Atrazine was registered
with the United  States Department of Agriculture
(USDA)  in  1958  (U.S.  Environmental  Protection
Agency,  2003).   On June 23, 1959, the atrazine
patent (Number 2891855) was issued. By the mid-
1960s, widespread use of atrazine was observed
(Duke,  Ciba  Geigy  Patent  Office,   personal
communication,   1994;   Kells,   Michigan   State
University, personal communication, 1994).

In the Lake Michigan basin, atrazine is primarily used
on corn crops to control broadleaf and some grassy
weeds by inhibiting photosynthesis. For corn crops,
it is usually applied to the fields in the spring, prior to,
during, or after planting a crop or after emergence of
the corn seedlings.  Atrazine is usually mixed in a
water solution along with other herbicides. Estimates
by Nako and Keitt (1994)  indicate that atrazine is
relatively inexpensive compared to other herbicides.
Cost for chemically  treating one acre  in 1992 was
about three dollars (estimate does not include costs
for fuel and labor). During  1994-1995,  atrazine use
as a  percentage of total pesticide use in the basin
was  13.8% (Brody et a/., 1998).  During the same
time  frame,  corn represented 38.8% of planted
acreage.  For the period 1964 through 1993, atrazine
was the leading herbicide used in the United States
(U.S. Department of Agriculture,  1994;  Lin et a/.,
1995).  Atrazine trade  names/synonyms  include:
Aatrex, Actinite  PK,  Akticon,  Argezin, Atazinax,
Atranex,  Atrataf, Atred,  Candex,  Cekuzina-T,
Chromozin, Crisatrina, Cyazin, Fenamin,  Fenatrol,
Gesaprim,  Griffex,  Hungazin,  Inakor,  Pitezin,
Primatol,  Radazin,  Strazine,  Vectal,  Weedex A,
Wonuk, and Zeapos (U.S. Environmental Protection
Agency, 2006).
                                             63

-------
Application rates  of atrazine  to  farm  fields  have
decreased  over  time.   In  1990,  a label change
initiated by the manufacturers of  atrazine set the
maximum recommended application rate for atrazine
to three pounds active ingredient per acre. Prior to
this,  four pounds active ingredient  per acre was
recommended (Scribner et al., 2000). In 1992, the
manufacturers  again   voluntarily  reduced  the
maximum recommended application rate of atrazine
on corn and sorghum to a range of 1.6-2.5 pounds
active ingredient per acre depending on soil organic
residue and erosion  potential.  The  1992  label
changes also included atrazine mixing/loading and
application setbacks to protect various water sources
including wells,  streams,   lakes,  and  reservoirs
(Pearson  and Giles,   1993).   The lower  1992
application rate was recommended for fields with less
than 30% plant residues on the surface. The label
changes reduced recommended application rates by
nearly 50%, however, actual application rates used
by farmers decreased by only about 11% from  an
average 1.1 pounds per acre  in 1991 to 0.97-0.98
pounds per acre in 1994-1995. (U.S. Department of
Agriculture, 2006). Evidently, farmers were satisfied
with  the results from  lower  than recommended
application  rates set  by the manufacturers  of the
herbicide.  The reduced application rates in 1994-
1995 and  1998  compared to  1989-1990  were
reflected in reduced concentrations  observed in
several Midwestern streams during post-application
run-off (Scribner et a/., 2000).

2.1.2  Total Annual Usage Estimates

Usage of atrazine is predominant in the eastern half
of the United States (see Figure 2.1.1).  As depicted
in the  figure, usage  is  heavy south of the  Lake
Michigan basin in the states of Illinois and Indiana.
However, except for the northwestern part of Indiana,
most of the drainage and associated load from these
two  states  discharge  into  the Mississippi  River
watershed.  But,  the proximity of  these high-use
areas to Lake Michigan does impact the atmospheric
loading of atrazine to the southern area of the lake.

In Table 2.1.1, some statistics are presented on the
usage of atrazine on corn crops in the United States
for   crop  years   1991,  1994,  and  1995   (U.S.
Department of Agriculture, 2006). For all three years,
atrazine was the most used herbicide on corn crops.
In the survey year 1994, the ranking of the top 10
states in order of highest corn acreage to lowest was
Iowa, Illinois,  Nebraska, Minnesota, Indiana, South
Dakota, Wisconsin, Ohio, Michigan, and Montana.

Figure 2.1.2 depicts county usage of atrazine during
the  Lake Michigan Mass Balance Project (LMMBP).
Note the highest use region is in  the southwestern
part of Michigan and northern Indiana. Little atrazine
is used in the northern parts of the basin. The data
for  1994 were provided by Kirschner (International
Joint Commission,  personal communication, 1997)
and the data from 1995 were provided  by Macarus
(U.S.  Environmental  Protection  Agency, personal
communication, 1999).

Historical total annual atrazine usage estimates in the
United States are depicted in Figure 2.1.3 for years
where data were available.  The data  used in the
graphic are presented in Table 2.1.2. The atrazine
data (zero usage) for 1963 (Duke, Ciba Geigy Patent
Office, personal communication,  1994) matches
estimates made by Scribner et al. (2000).  Robert
Torla's (United  States Environmental Protection
Agency (USEPA),  personal communication, 1994)
data  (1964,   1966,  and  1971)    are   from
USDA/Economic Research Service (ERS) published
estimates (U.S. Department of Agriculture, 2003),
and the rest of the  data are from Aspelin and Nako
(USEPA, personal communication, 1997). The data
represent total annual usage (both agricultural and
non-agricultural).   However  in the 1990s, it was
estimated  that  approximately 95% was used  for
agricultural  purposes.  For some of the  years (such
as 1993 and 1995), a range of values was reported.
When this occurred, a mean of the range was used.
Also plotted on Figure 2.1.3 are  historical (1964-
2002) total United States  acreage  for corn  and
another for  the sum acreage of corn, sorghum, and
sugarcane - all crops that use atrazine  to suppress
weeds.  Notice that  the  pattern of atrazine use,
except for the earliest years, follows the pattern of
corn acreage planted in the United States.  The low
corn acreage in 1983  and 1988 were due to drought
conditions (Shapouri et al., 1995).

2.1.3 Future Atrazine Use Estimates

Atrazine  currently  holds  its large market share
because it is a pre-emergent herbicide active against
most of the serious broadleaf weeds in corn, and it is
                                              64

-------
  Kilograms per
  Square Kilometer
     I Missing or 0
      Less than 0.5
      0.5 to 2,5
      2.6 to 10.0
      10 1 to 25.0
      more than 25.0
           0        300 Miles
           I    I   II
           III  II  I
           0        500 Kilometers
Graphic by William Battaglin
  U.S. Geological Survey
Figure 2.1.1. Atrazine usage in the United States for 1991.
Table 2.1.1. U.S. Department of Agriculture Corn Crop Summaries of Atrazine Usage in the United
States for 1991, 1994, and 1995
         Number of
 Year      States
         Surveyed
 1991
 1994
 1995
17
10
15
% of Total
Corn Crop
Surveyed
90
79
90
% of Corn
Crop Treated
With
Herbicides
94
98
96
% of Corn
Crop Treated
With
Atrazine
66
68
65
Average
Application
Rate of
Atrazine
(Ibs/acre)
1.1
0.97
0.98
                        Total Amount
                         of Atrazine
                           Applied
                         (millions of
                             kg)

                            23.61
                            20.59
                            20.74
                                               65

-------
                                             0-12000
                                         •112001-24000
                                         •124001-36000
                                         tm 36001-48000
                                         • 48001-60000
                                         Kilograms of Atrazine Applied
Figure 2.1.2.  Estimates of atrazine usage in the Lake Michigan basin for 1994 and 1995.

— corn + sorghum + sugarcane
1 ... rnrn
£45-
55
o>40
o
^35-
o
« 30-
o
|2S-

0
01
s15-
•510
3
i 5
s
< 0



-A^VJ
---*'*'_."'







•* *







^
t








•x








1 total annual atrazine usage
—
-•









r







i *
> *
V







\








y
*-''







*—***








:








A








r








—\








*^













f








r120
'
-100
"en"
•80 .0
' I
-60 ~
w
2
•40 §
•
•20

-0
                _  1964 1968 1972 1976 1980 1984 1988 1992 1996 2000
                                        Year

Figure 2.1.3.  Historical trend of total annual usage of atrazine in the United States with acreage
planted in corn, sorghum, and sugarcane.
                                            66

-------
Table 2.1.2. Total Annual Usage of Atrazine in the United States (Aspelin and Nako, U.S. Environmental
Protection Agency, Personal Communication, 1997; Torla, U.S. Environmental Protection Agency,
Personal Communication, 1994)
                    Year
Millions of kg Atrazine Used in the United States
                    1963
                    1964
                    1966
                    1971
                    1974
                    1976
                    1978
                    1980
                    1982
                    1984
                    1986
                    1989
                    1991
                    1993
                    1995
                     0.0
                     6.3
                     12.0
                     25.8
                     31.8
                     36.3
                     39.9
                     38.6
                     35.4
                     39.9
                     35.8
                     34.9
                     34.0
                     34.7
                     33.8
inexpensive   (Nako  and   Keitt,  1994).    Any
replacements  must  be equally as  effective  in
controlling weeds and matching or beating costs.
Due to repeated annual usage, some weeds, such as
pigweed,  are  showing  resistance  to  triazine
herbicides.  Blending  other herbicides with atrazine
may help to eliminate some of these resistant plants.

If the  resistant plants  do not have an efficient seed
dispersal mechanism, then  these problem plants
become a local problem.  However, if the resistant
plant shows resistance to other herbicides and has
an effective seed dispersal mechanism, then usage
of atrazine may decline. Another factor to consider
in projecting  future usage  is possible  regulatory
action that could  restrict usage in some manner.
With growing ethanol demand and strong export
sales of corn, U.S. farmers planted 92.9 million acres
of corn in 2007. This exceeded the 2006 acreage by
19 percent (U.S.  Department of Agriculture, 2007).
The actual planted acreage is the highest since 1944.
           References

           Brody, T.M., B.A. Furio, and DP. Macarus.  1998.
              Agricultural Pesticide Use in the Great Lakes
              Basin:  Estimates of Major Active  Ingredients
              Applied During  1994-1995  for the Lake Erie,
              Michigan,  and   Superior  Basins.     U.S.
              Environmental Protection  Agency, Region  5,
              Chicago, Illinois.  15 pp.

           Lin, B., M. Padgitt, L Bull, H.  Delvo, D. Shank, and
              T. Harold.  1995.  Pesticide and Fertilizer Use
              and  Trends  in  the U.S.  Agriculture.   U.S.
              Department of Agriculture, Economic Research
              Service, Washington, D.C.  Document Number
              717, 56pp.

           Nako, S. and G. Keitt.  1994.  Use of Triazines and
              Other Herbicides for Broadleaf Control on Corn.
              U.S. Environmental Protection Agency, Office of
              Pesticide Programs, Biological and  Economic
              Analysis Division, Washington, D.C. 7 pp.
                                              67

-------
Pearson, D. and E. Giles.  1993.  Atrazine Label
   Changes. Resource Update One, Illinois Food
   and Agriculture Council, Urbana, Illinois.

Scribner, E.A., W.A.  Battaglin, D.A. Goolsby, and
   E.M. Thurman.   2000.  Changes in Herbicide
   Concentrations   in   Midwestern  Streams  in
   Relation to Changes in Use,  1989-1998.  Sci.
   Total Environ., 248(2/3):255-263.

Shapouri,  H., J.A.  Duffield,  and  M.S. Graboski.
   1995.   Estimating the Net Energy Balance of
   Corn Ethanol.  U.S. Department of Agriculture,
   Economic Research  Service,  Office of Energy
   and  New Uses, Washington, D.C.  Agricultural
   Economic Report  Number 721, 24 pp.

U.S.  Department of Agriculture. 1994.  Agricultural
   Resources and  Environmental Indicators.  U.S.
   Department of Agriculture, Economic Research
   Service, National  Resources and Environment
   Division, Washington, D.C. 216  pp.

U.S.  Department of Agriculture.  2003. Historical
   Track Records - National Agricultural Statistics
   Service.  Available from  U.S.  Department of
   Agriculture   at  http://usda.mannlib. Cornell.
   edu/usda/nass/96120/trackrec2003.txt

U.S.  Department of Agriculture. 2006.  Agricultural
   Chemical Usage-1991, 1994, 1995 Field Crops
   Summary.    National Agricultural Statistics
   Service, Washington, D.C. Available from U.S.
   Department   of  Agriculture  at    http://usda/
   mannlib.cornell.edu/data-sets/inputs/9x171.
U.S.  Department of Agriculture.  2007.  National
   Agricultural Statistics Service.  U.S. Department
   of Agriculture, Washington, D.C.  Available from
   U.S. Department of Agriculture at  http://www.
   nass.usda.gov.

U.S.  Environmental  Protection Agency.    2003.
   Pesticides: Topical and Chemical Fact Sheets -
   Atrazine   Background.   U.S.  Environmental
   Protection Agency, Office of Pesticide Programs,
   Washington,  D.C.   Available   from   U.S.
   Environmental Protection Agency at http://www.
   epa.gov/pesticides/factsheets/atrazine_
   background.

U.S.   Environmental  Protection  Agency.   2006.
   Consumer  Factsheet  on:   Atrazine.    U.S.
   Environmental Protection Agency, Ground Water
   and Drinking Water, Washington, D.C.  Available
   from U.S. Environmental  Protection Agency at
   http:/www.epa.gov/safewater/dwh/csoc/atrazine.

U.S.  Patent  Office.   1959.    Compositions and
   Methods for Influencing the Growth of  Plants.
   Assignors: Hans Gysin and Enrico Knusli, J.R.
   Geigy A.G., Basel, Switzerland.  Patent Number:
   2891855; Serial Number 481474.
                                              68

-------
                                         PART 2

           LAKE MICHIGAN MASS  BALANCE PROJECT ATRAZINE
                         LOADINGS TO LAKE MICHIGAN
Chapter   2.     Estimation
Tributary  Loadings
of  Atrazine
Tributary  loadings for  the  Lake Michigan  Mass
Balance Project (LMMBP) atrazine models were
estimated using an approach based on watershed
export of the applied chemical from farm fields to the
lake  and  another approach that utilized  LMMBP
measurements of atrazine concentration and flow in
the 11 monitored streams within the construct of the
Stratified Beale Ratio Estimator (SBRE) to calculate
loadings.  Watersheds that drained into a monitored
tributary were identified as a monitored watershed.
The other watersheds  in the Lake Michigan basin
were identified as unmonitored.  Both methods made
estimates of watershed loadings of atrazine to Lake
Michigan  for both the monitored and unmonitored
watersheds.   The MICHTOX  and LM2-Atrazine
models solely utilized estimates based on watershed
export.   The  LM3-Atrazine model  utilized  load
estimates based on the SBRE and on a hybrid of the
two load estimates, whereby the SBRE loads were
enhanced with  additional loadings  based on  the
annual watershed  export estimates  (see Part 5,
Chapter 3, Section 5.3.3.3.1).  Both the MICHTOX
and  LM2-Atrazine  models   utilized   annualized
loadings  only  and  are  useful  for  long-term
simulations. LM3-Atrazine loadings were calculated
on a daily basis so as to capture seasonal variations
on a finer time and spatial resolution that were not
available in either of the other two models.

Watershed loading estimates were made for all of the
Lake Michigan sub-basins. A sub-basin may have an
identifiable tributary that discharges this loading into
the lake, or  it  may  not.   However they were
calculated, both were  collectively referred  to as
tributary loadings.

2.2.1  Atrazine Tributary Load Estimates
Utilizing    County-Level   Atrazine
Application Data

Literature values for estimates of the percentage of
the amount of atrazine applied in a watershed that is
delivered to a receiving body of water were used in
the loading estimates.  This percentage is identified
as the Watershed Export Percentage (WEP), but it is
also  referred to in the literature  as Load as a
Percentage of Use (LAPU).   The calculation  of
atrazine tributary loads (mass/time) to a MICHTOX or
LM2-Atrazine  segment for a  given  year  when
application rates and corn acreage are known were
calculated as follows:

WMershetl Export Load or Tributary Load =
                where

                j  =  a county within a Hydrological  Unit Code
                      (HUC) draining into a given water segment

                n  =  total number of counties in a HUC

                k  =  the load from a given HUC in a sub-basin
                      delivered to a model segment
                                            69

-------
m =   total number of HUC loads in a sub-basin
       delivered to a model segment

A  =   atrazine application rate (mass/acre/time) for
       corn

T  =   fraction of corn acreage treated with atrazine

C  =   corn acreage in a given county

F  =   fraction of county within the HUC

L  =   (Watershed Export Percentage)/! 00

2.2.1.1 County-Level Atrazine Application Data

In this project, we received atrazine data in the form
of county-level application estimates for a given year
or the product of variables A,  T, and C in Equation
2.2.1. Sources of these data are identified in Table
2.2.1. As is evident from the table, data were only
available for six years.  Additional data beyond 2002
are likely but were not included in this analysis. The
atrazine data were reported as an active ingredient,
so no conversion was required before model loadings
were estimated.
                                  The area fraction of a given county that lies within a
                                  HUC in the Lake Michigan basin was determined by
                                  Geographical Information System (GlS)-defined HUC
                                  boundaries and county boundaries. Within the basin,
                                  there  are a number of HUCs that collectively form
                                  sub-basins.   Some of these sub-basins  defined
                                  watersheds of the 11 LMMBP major tributaries in the
                                  Lake  Michigan basin.  Other sub-basins were not
                                  readily identifiable with tributaries;  however, load
                                  estimates, identified as unmonitored tributary loads,
                                  were made for these sub-basins and the discharge
                                  into the lake was associated with a model segment.
                                  The GIS was used to calculate what fraction of a
                                  county fell into a given HUC.  Note that more than
                                  one county may fall within a given HUC.

                                  2.2.1.2 The Watershed Export Percentage

                                  The atrazine  WEP (variable L in Equation 2.2.1) is
                                  known to be  a function of soil type, population of
                                  atrazine-degrading  bacteria   in  the   soil,  field
                                  topography,  timing and  amount  of rainfall  after
                                  application,   and  other  explanatory   variables.
                                  Seventy-six  reservoir drainage  basins  in  the
                                  Midwestern United States were studied using multiple
                                  linear  regression and   logistic  regressions  to
Table 2.2.1. Sources of County-Level Atrazine Application Data for the Lake Michigan Basin
 Application
     Year
      Data Source
       Affiliation
    Date Received
     1989
     1992
     1993
     1994
     1995
     1998
W.A. Battaglin and D.A.
Goolsby (see reference)
  U.S. Geological Survey
B. Kirschner and R. Baksh    International Joint Commission
(personal communication)   (UC), Windsor, Ontario, Canada
R. Baksh (personal
communication)

B. Kirschner (personal
communication)

D. Macarus (personal
communication)

D. Macarus (personal
communication)
           UC
           UC
USEPA/Region V, Chicago,
          Illinois

     USEPA/Region V
1995 (Publication Date)
                                      1994
        1995
        1997
        1999
        2000
                                               70

-------
determine  the  significance  of  the  explanatory
variables in predicting concentrations of atrazine in
the reservoirs (Battaglin and Goolsby, 1996). Both of
the statistical tests used  in the analysis found soil
hydrologic  group values  to  be  a  significant
explanatory variable.   This same conclusion was
drawn  from  studies  performed by Blanchard and
Lerch (2000) in northern Missouri.  Small hydrologic
group values (1.75) are associated with well-drained
soil (sand and gravel), whereas larger values (>3.25)
are  associated  with  poorly-drained  soil  (clays,
wetlands,  urban).    Soil  textures  in  Michigan,
Wisconsin, Illinois, Indiana, and Ohio are shown in
Figure 2.2.1.   These data were obtained from the
State  Soil  Geographic  (STATSGO)   database
provided by   the  United States  Department  of
Agriculture  (USDA),  Soil  Conservation  Service
(SCS).  A review of  1992 and 1993 atrazine field
application data revealed that approximately 80% of
the total atrazine application in the Lake Michigan
basin is applied to crops in the sub-basins that drain
into the southeastern part of the lake.  These sub-
basins include the southwestern quarter of the lower
peninsula of Michigan including a small portion of
northern Indiana that also resides  in the Lake
Michigan basin. Soils in that part of the basin can be
identified as moderate to fine textures and have
hydrologic group values ranging from  2.51 to 3.25.

At the  start of a  rain  event, the rate of rainfall may
equal the rate of infiltration into the soil. However,
after some time, the  infiltration  rate  will start to
decrease asymptotically and reach some constant,
but lower, infiltration rate.  Run-off begins at the point
the rainfall rate exceeds the infiltration rate.

A  literature  review of atrazine watershed  export
percentages is summarized in Table 2.2.2.  The raw
data used in this summary are presented in Table
2.2.3.   Watershed export percentages for various
watersheds are grouped by soil type.  The data are
from northern watershed systems (Ontario, Canada;
northern Ohio;   northern  Iowa;  and  southern
Minnesota).  In addition, the data  reflect watershed
export percentages that were calculated for an entire
year.  Many published studies of WEP fall short of a
full year of monitoring and this causes the estimate to
be biased low for the annual estimate.  Based on this
literature review, a watershed export percentage of
0.6% was selected for the Lake Michigan and Green
Bay  watersheds to  represent the   predominant
moderate texture soil hydrologic group in this area.
Note that WEPs for clay soils (1.4) are much higher
than for sandy soils (0.2). WEP differences between
clay and sandy soils will yield  large differences on
loading estimates as Equation 2.2.1  indicates.  In a
rain event, run-off will occur sooner on non-saturated
clay soils than non-saturated sandy soils, because
sandy  soils have a higher infiltration rate.  So, it is
important to carefully assess this parameter when
estimating watershed export of atrazine.  Climatic
conditions for the 12 annual studies used in deriving
this watershed export  percentage for  loam/fine-
textured soils included a balance of five wet and five
dry years.    For  the other two years, one was
considered average in precipitation and conditions for
the other year were not reported.  Including both wet
and dry  years  should  help minimize bias in the
estimate, since atrazine-associated run-off in drought
years has been observed to be lower compared to
wet years (Richards et a/. 1996). A plot of WEPs
versus watershed size  indicated  that there was no
relationship, and this was also the conclusion by
Capel and Larsen (2001). They evaluated data from
408 observations of WEPs across numerous types of
soil textures.  Their median WEP was calculated to
be  0.66% for watersheds  less than  100,000 ha.
Although not calculated, a more rigorous derivation
of the average watershed export percentage could be
achieved if a detailed accounting of soil types was
performed for corn  croplands within the basin.  With
that detailed soil  type  information, a  weighted-
average  WEP could be calculated for each sub-
basin.

2.2.1.3 Calculating the Atrazine Tributary Load

County-level application data for a given year were
multiplied by  the fractional area of the county  in a
HUC (Equation 2.2.1). This load was further divided
if a monitored river basin occupied a portion of that
county. In that case, the atrazine load was further
divided and apportioned by area to a monitored river
load and the rest to unmonitored tributary load. This
procedure was repeated for all  counties that  had
overlap in any given HUC within the Lake Michigan
basin.  These fractional application loads (monitored
and unmonitored) were summed separately for each
HUC.   The point  of discharge  of  the monitored
tributary  into the lake was associated with a model
segment and likewise for unmonitored tributary loads.
Only those whole counties or fractional counties that
                                               71

-------
                                  300 kilometers
                               soil types
                               |    sands and gravels
                               ^^ moderate textures
                               ^| fine textures
                               ^| clays, impervious

                               (from STATSGO database)
Figure 2.2.1. Soil textures typical for the Lake Michigan basin and part of the Lake Erie basin.
Table 2.2.2.  Atrazine Watershed Export Data Summarized From the Literature. Raw Data Used to
Create This Table Can be Found in Table 2.2.3
      Soil Type

 Clay
 Loam/Fine Textured
Watershed Export   Standard   95% Confidence
   Percentage      Deviation        Level
       1.4
      0.61
0.61
0.38
0.94-1.85
0.37-0.85
 Range

0.11-2.5
0.21-1.5
Number of
 Studies

    9
    12
                                            72

-------
      Table 2.2.3.  Atrazine Watershed Export Data From Various Northern Sites
73
Water-
shed
Size (sq
km)
23.83
30.25
30.25
50.8
50.8
62
62
6790
16395
18.6
18.6
30
30
54.72
54.72
45.04
45.04
3998
17820
38585
38585


19.9
19.9
56.45
79.13
79.13
Location
Ontario
Ontario
Ontario
Ontario
Ontario
Ontario
Ontario
Ontario
Ohio
Ontario
Ontario
Ontario
Ontario
Ontario
Ontario
Ontario
Ontario
Ontario
N. Iowa
S.
Minnesota
S.
Minnesota
Ontario
Ontario
Ontario
Ontario
Ontario
Watershed
Soil Type Export Year(s)
Percentage
Clay
Clay
Clay
Clay
Clay
Clay
Clay
Clay/Loam/Sand
Clay
Loam
Loam
Loam
Loam
Loam
Loam
Loam
Loam
Sandy/Loam
Fine Textured
Fine Textured
Fine Textured


Sandy
Sandy
Sandy
Sandy
Sandy
0.11
1.47
1.35
2.51
1.40
1.28
1.31
1.49
1.50
0.32
0.54
0.21
0.26
0.50
0.35
1.09
0.80
0.65
1.50
0.33
0.62


0.15
0.20
0.17
0.29
0.18
1976
1975
1976
1975
1976
1975
1976
1981,82,83,84,85
NA
1975
1976
1975
1976
1975
1976
1975
1976
1981,82,83,84,85
84
90
90


1975
1976
1975
1975
1976
Watershed
Number R.
Twenty Mile Cr
Twenty Mile Cr
Thames R.
Thames R.
Au Sable R.
Au Sable R.
Grand R.
Maumee R.
Grand R.
Grand R.
Thames R.
Thames R.
Maitland R.
Maitland R.
Saugeen R.
Saugeen R.
Saugeen R.
Cedar R. Basin
Minn. R. Basin
Minn. R. Basin


Hillman Cr
Hillman Cr
Shelter Val. Cr
Big Creek
Big Creek
Annual
Precipita-
tion
Dry
Wet
Dry
Wet
Dry
Wet
Dry
Average
Average
Wet
Dry
Wet
Dry
Wet
Dry
Wet
Dry
Average
NA
Dry
Wet


Wet
Dry
Wet
Wet
Dry
Adjusted
% Loss to
Reference Represent
Atrazine
Only
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Logan, 1988
Richards etal., 1996
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Logan, 1988
SquillaceSThuman, 92
Schottleref a/., 1994
Schottlerefa/., 1994


Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Franks Sirons, 1979
Y
Y
Y
Y
Y
Y
Y
Y
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
N
N


Y
Y
Y
Y
Y

-------
lie within the Lake Michigan basin boundaries were
considered for tributary load estimation. When taking
fractional  areas of a  county, we assumed  that
atrazine  usage  within the  county  was  uniform.
Tributary loading estimates were  made for each of
the years in Table 2.2.1 using this method.

2.2.2 Estimating Atrazine Tributary Loads
for Years  When  County-Level  Atrazine
Application Data Was Not Available

For the six years where county-level application data
were available, tributary loads were estimated using
the approach identified in the previous section, 2.2.1;
however, to make estimates for additional years, an
approach  was selected that utilized  some  of  the
results from Section 2.2.1  and estimates of total
annual atrazine usage in the United  States.  The
approach was to  calculate a Tributary Load Ratio
(TLR) of known application rates for a given year and
divide this number by the total annual United States
usage amount of atrazine for that same year.   For
years when  application data are missing but total
annual usage is known, the ratio multiplied by  the
total annual  usage yields an estimate of tributary
load. Seventeen years of total annual United States
usage data  are displayed  in Figure 2.2.2.  This
approach was used for both MICHTOX and  LM2-
Atrazine model runs.

Tributary Load Ratio = {Tributary Load to Model
Segment) k Total Annual USA Atrazine Usage)
                                        (2.2.2)

For  any year (y), where only total annual United
States usage  is  known,  a tributary  load was
calculated utilizing a tributary load ratio:
Tributary Load = [Tributary Load Ratio) x
(Total Annual USA Usage for Any Year jy)j
                                         (2.2.3)
Due to  label changes that lowered  application
amounts and established  planting setbacks from
water bodies in 1990 and 1992, a decision was made
to use  two  TLRs  in  order to  address atrazine
application practices for pre- and post-label changes.
For the  pre-label  change  period, tributary load
estimates for years 1964 through 1986 used the TLR
calculated for 1989. We used 1989 because this was
the only  year during  that pre-label change  period
where we had both county-level  application data and
total annual United  States  usage estimates.   An
average atrazine application rate of 1.54 pounds/acre
on corn from a 1982 survey of  16 states with more
than one million acres of corn compares well to an
average application rate for Michigan and Wisconsin
of 1.5 pounds/acre for the same year (Gianessi and
Puffer, 1988). So for at least that year, the atrazine
usage  rate per acre in the Lake Michigan  basin
matches  usage rates in the rest of the major  United
States corn-growing  regions.    For comparison
purposes,  atrazine  tributary  load  estimates  to
MICHTOX segment 1 were made for the year 1984
using the TLR method based on total annual  United
States usage estimates for 1984 and also by using
available atrazine use  data (Gianessi and  Puffer,
1988) that included application rate data by state,
total corn crop acreage by state, and fraction of corn
crop that was treated with atrazine. The TLR method
yielded a total tributary load estimate of 15.4  kg/day
of atrazine to  segment one.  The tributary load
estimate  based on Equation 2.2.1 and data from
Gianessi  and Puffer yielded  a result of 17.7 kg/day.
For this latter estimate, data on the percent of corn
acreage  treated  with  atrazine was  from  1984
(Michigan),  1982   (Illinois),   and  early   1980s
(Wisconsin and  Indiana).  Also, the application rate
data are from  1984  for Michigan, Indiana,  and
Wisconsin and from 1982 for Illinois.  Data  on the
percent of  corn acreage  treated with atrazine  for
Wisconsin and Indiana were based on expert opinion
of the  U.S. Department  of Agriculture/Economics
Research Service, rather than  survey data.   Total
corn acreage in each county within the sub-basin
draining into segment 1 was based on actual  survey
data for  1984 (Kevin Pautler,  U.S. Department of
Agriculture, personal communication, 1997).  Given
the uncertainties of both methods, the two numbers
are reasonably close.

For the post-label change period, an average of the
tributary  ratios  for 1992  and  1993 was  used to
calculate tributary loadings for 1991. For the rest of
the post-label change years 1992,1993,1994,1995,
and 1998, loads were calculated based on county-
level application data  using Equation  2.2.1.  In
                                               74

-------
C/5
2 45-1
O)
i 40"
o 35-
CO
o 30-
I 25-
c
o 20-
O)
co 15-
3
TO 10-
3
c
S 5-
5 0-


i



r




r



]






















































n total annual usage
• historical tributary loa
r



























d














r












r





































!
i









1












r





1






•1-9000
-8000
-7000 J
O)
-6000 ^
-5000 "i
o
-4000 j*
ro
-3000 -5
JD
-2000 B


-1000
--0
              1964 1968 1972 1976 1980 1984 1988 1992 1996 2000
                                         year
Figure 2.2.2.  WEP-based total atrazine tributary loading estimates to Lake Michigan.
comparing tributary loading ratios for pre- and post-
label change years 1989 and 1995, the TLR for 1995
(MICHTOX segment 1), which carries most of the
atrazine tributary loading to Lake Michigan, was 26%
lower than what it was in 1989.  A similar trend was
noted for the other segments.

Yet, total annual United States usage only dropped
three percent from 1989 through 1995, and total corn
crop acreage fell just 1.16 percent (Good and Irwin,
2007).   This  indicates that usage in  the Lake
Michigan basin dropped more relative to the rest of
the United States  during  that  period.  The Lake
Michigan basin  has a number of rivers and lakes.
Perhaps the label changes requiring setbacks from
these water bodies reduced the corn acreage and
hence  usage dropped.  The state  of Wisconsin
mandated atrazine usage changes that went beyond
those related to the 1990  and 1992 label changes
and includes establishment of statewide usage rates,
application  timing,  record  keeping (Wisconsin
Register, 2004), and prohibited use areas (Wisconsin
Register, 2005). Efforts to reduce atrazine usage in
the state have  been ongoing since  1991.  Other
possible reasons for the use reduction in the Lake
Michigan  basin  include substitution  of different
herbicides, cropland taken  out of corn  production,
changes  in farming  practices,  increase of corn
acreage  outside the  Lake Michigan  basin  that
replaced   lost  acreage within  the  basin,  etc.
Regardless of the reason for the pre- and post-label
changes  reflected in  the TLRs,  use of two ratios
seems to be warranted.
2.2.3     Atrazine  Tributary
MICHTOX and LM2-Atrazine
Loads  for
Based on methods described above, total tributary
loadings to Lake Michigan are depicted in Figures
2.2.3 and  2.2.4.  Since the mid-1980s, atrazine
tributary loadings have been declining in the Lake
Michigan basin; however, total annual usage in the
United   States  has  not changed  much since
approximately   1986.    Tributary   loadings   for
MICHTOX segments are shown in Figure 2.2.3. Note
that the watershed delivering  atrazine  to  model
segment 1  delivers the highest load to the lake.  See
Figure 3.1 in Part 3 for a graphic  showing MICHTOX
segments.  This watershed drains the southwestern
part of  the state of Michigan  and a section  of
                                              75

-------
40UU'
Annn .
3500 •
3000

900
700
600
400
300
200
100











_









• 1994 loads

D1995 loads





•n

04 2



^^_^













— -I J] __,
34567
MICHTOX Segment
Figure 2.2.3. Tributary loadings to Lake Michigan
MICHTOX model segments.
     monitored
      tributary
     loads: 4305
Figure 2.2.4. WEP-based Lake Michigan tributary
loadings, 1994.
Northeastern Indiana. For the whole-lake, the total
annual tributary load estimate is the same for both
models.  However,  LM2-Atrazine had 10 receiving
surface water segments and MICHTOX had seven
(six  primary  segments  and  a  small  segment
representing the lower Fox River).  Each surface-
water segment sharing a boundary with a sub-basin
received that sub-basin load.

For years where  atrazine  application  data  or total
annual USA usage are unknown, both MICHTOX and
LM2-Atrazine calculate loads by assuming that the
loads between the two years bracketing the missed
annual loads are linear.

2.2.4  Atrazine Tributary Load Estimates
for LM3-Atrazine

The LM3-Atrazine model was not used to conduct a
hindcast simulation. This model used United States
Geological Survey (USGS) loading estimates that
were based on actual measurements of river flow
and atrazine concentration.  Because these load
estimates were low compared to the WEP-based
load estimates discussed in the previous section, the
USGS loadings  were  adjusted  upward  in  the
spring/early summer period so that the total annual
load was equal to the WEP-based annual loading.
See Section 5.3.3.3.1  in the LM3-Atrazine modeling
chapter for  more  information  on  LM3  tributary
loadings.

2.2.4.1 Tributary Sampling Program

As part of the LMMBP, the USGS calculated loads
for 11  monitored tributaries  in the Lake Michigan
basin (Hall  et a/.,  1998).   Based on these load
calculations and land  use information, estimates of
loadings  from unmonitored  areas  were  made.
Loadings  were  calculated   for  atrazine,
deethylatrazine (DEA), and deisopropylatrazine
(DIA). Tributary data used in the load estimates were
gathered  from samples collected from April 4, 1995
through October 30, 1995  (U.S.  Geological Survey
and Eisenreich, 1997). Samples were collected far
enough upstream to  minimize mixing of lake and
tributary water.  The  Grand  Calumet,  Kalamazoo,
and Pere Marquette Rivers were generally well-mixed
throughout the sampling period.  The Sheboygan,
Menominee, Manistique, Muskegon, Grand, and St.
                                             76

-------
Joseph Rivers were generally well-mixed during the
winter  months   and  stratified  with   respect  to
temperature and conductance in summer months.
The Milwaukee River, and to a lesser extent the Fox
River, were  found to be poorly mixed  at irregular
intervals  throughout  the sampling period.   The
location and identification  of the  USGS stations
sampled can be found in Hall et a/. (1998). Sampling
was conducted by the USGS in cooperation with the
Wisconsin and  Michigan Departments of Natural
Resources,  the  Wisconsin  State  Laboratory  of
Hygiene,  and the University of Wisconsin Water
Chemistry Program.  The primary  objective of the
contaminant-loading data was to provide a detailed
space and time tributary loading history for input into
the LMMBP LM3-Atrazine model.

To  reduce   errors   associated   with   the   load
calculations, sampling was deliberately biased toward
high-flow  conditions  where  more than 20%  of
samples were collected at times of discharge above
the 20% exceedance, (Dolan etal.,  1981; Hall etal.,
1998).  The assumption is that during the high-flow
periods, most of the load is transported. Sampling
for  atrazine was delayed  for one year due  to
uncertainty in selection of methods and laboratory.
As a consequence of having only seven months of
load data to quantify atrazine loadings, the USGS
believed that the atrazine load estimates based on
actual concentration and flow measurements were
not as good as estimates for the other mass balance
contaminants of interest that were based  on 19
months of measurements. Furthermore, load error
estimates  for atrazine were  especially poor,  again
due to the short sampling period.

Three to four sampling crews in three states were on
call to capture storm-induced flow events (Hall, U.S.
Geological Survey, personal  communication, 2001).
Weather   was  monitored  24  hours   per   day.
Equipment was used  to trigger pagers upon  the
onset of  rising  hydrographs.  Sampling occurred
during rising, peak, and falling hydrographs. Except
for the shallow  Pere Marquette and  Kalamazoo
Rivers, rivers were sampled at 0.2 and 0.8 of the total
depth.  These samples were taken  at the midpoints
of river panels that divided the total river flow into
three visually estimated equal flow panels that were
determined  during   discharge   calibration
measurements (Hall et a/., 1998). These six samples
were composited  into one sample.   For the Pere
Marquette  and  Kalamazoo  Rivers,  only  three
samples (one in each flow panel) were  composited
(Hall etal., 1998).  A total of 405 samples (including
quality control samples) were collected.

River discharge was measured either by stage and
discharge  techniques for the Manistique,  Pere
Marquette,  and Kalamazoo Rivers in Michigan  or
acoustic velocity meters for the Muskegon, Grand,
and St. Joseph Rivers in Michigan; Grand Calumet
River in Indiana; and Milwaukee, Sheboygan, Fox,
and  Menominee Rivers  in Wisconsin  (Hall  et a/.
1998).

2.2.4.2 Atrazine Load Estimation for Monitored
Rivers Using the Stratified Beale Ratio Estimator
(SBRE) Method

Concentration data are usually limited  due to cost
constraints; however, flow data are usually readily
available at short-time intervals.  Sampling for the
LMMBP  was  focused  on  high-flow,   high-
concentration  events.    However,   if  the   mean
concentration from  these limited samples  were
multiplied by  the total annual discharge, the load
estimate would be biased high. The reason it would
be high is that the  mean concentration observed
would be disproportionately distorted by the number
of high-flow, high-concentration samples.

The SBRE method is nearly bias-free when the data
are sufficient to give acceptable precision to the load
estimate. The SBRE method used by the USGS for
the LMMBP  can  be  found  in  Richards  (1994).
Another factor in the selection of the SBRE is that the
method is robust over a range of data distributions.
The method has been the method preferred by the
International Joint Commission (IJC) for a number of
years.  The SBRE was used for the period April 4,
1995 through October 30, 1995 when atrazine was
sampled.

For the unmonitored period, January 1,1994 through
April  3,  1995  and  October 31,  1995 through
December 31, 1995, a combination of Beale-derived
daily loads and regression loads from the monitored
period were used to adjust regression-produced daily
loads from the unmonitored period (Hall, 2004). The
Beale method does  not  provide  an algorithm  to
                                              77

-------
extend the loadings derived  from the monitored
period  to an  unmonitored period.  An adjustment
coefficient was computed by  dividing  the sum of
Beale-model daily loads from the monitored period by
the sum of the Estimator Regression Model loads for
the same period. The adjustment coefficient was
then multiplied by each daily load produced by the
selected  regression  model for each  of the  two
unmonitored  periods to  produce "corrected"  daily
loads.   For  example,  if the Beale  model  was
producing a sum of daily  loads  greater than the sum
of the regression model daily loads for the monitored
period, the adjustment coefficient would be greater
than one and the adjustment multiplication would
linearly increase each regression-daily load in each
of the two unmonitored periods.

The  1995  USGS SBRE tributary loadings  are
depicted in Figure 2.2.5.  Median river flows  and
median atrazine concentrations are also shown. The
rivers are ordered based on the highest load on the
left to the lowest load on the right. Note that although
the Grand Calumet had the lowest atrazine load, it
           did  have  the  fourth  highest  median  atrazine
           concentration.

           2.2.4.3     Atrazine  Load   Estimation  for
           Unmonitored Watersheds

           Hall (2004) presents material on the method used to
           estimate daily loading from watersheds in the Lake
           Michigan basin where no samples were taken for the
           analytes of interest.  Loading estimates derived from
           the 11 monitored tributaries were used to predict
           loadings  from  the  additional  25  unmonitored
           tributaries larger than 325 km2.  Unit area yields from
           the monitored basin were calculated as follows:
           Unit Area Yield =  !,

           where
                       (2.2.4)
           /,  =  load estimate for any given day
           A  =  area of the watershed for a monitored tributary
                     )— 90

                 1600-?8Q
               -       6
               -1200-| 601

               €1000-| 50
               o       o
               7v  800-5 40
               £ 600-£ 30-

               ™ 400-2 20-
                       "5
                  200— 10
                    0—  0
D atrazine loading (kg/yr)
B atrazine concentration (ng/L)
• flow (cfs)
                            .C  T3   X  O
                             Q.  C   O  O
                             O3  c  crash
     3  03  I-L 3 ><  'F  '-=  ,*^ ^
     ec  ^    cr O   t  tn  CD-=
     •^  ^n    t_ r^   /-^  '^    "5
                                    1995 monitored tributaries

Figure 2.2.5. 1995 USGS SBRE atrazine loadings and median concentrations relative to median flow
in Lake Michigan tributaries.
                                              78

-------
The USGS used Unit Area Yields from monitored
watersheds  that  best  matched   unmonitored
watersheds in  terms of land use and  nature  of
surficial land deposits. A GIS was used to help in the
watershed classification. Once this classification was
done, the areas of the 25 unmonitored watersheds
were  expanded  to  encompass smaller  adjacent
basins that were poorly defined in terms of land use,
discharge location, and other properties. The sum of
all monitored and unmonitored watershed loads were
designed to represent  the  total loading to  Lake
Michigan from the entire Lake Michigan watershed.

2.2.5   Comments on Atrazine Tributary
Loading Estimates

Estimates  of  atrazine  tributary  loadings to  Lake
Michigan for years  1994 and 1995 were  made
independent of  the USGS  estimates.   These
independent estimates  were  based  on  actual
application  of atrazine to the basin  and using a
literature-derived WEP of 0.6%.  The following are
the results:

       1994 USGS:         1163kg
       1995 USGS:         1426kg
       1994 WEP-Based:   5263 kg
       1995WEP-Based:   4916kg

The ratio of WEP-based  to USGS load for 1994 is
4.5, and the ratio for 1995 is 3.4.

For a discussion  on  possible reasons  for the
discrepancy between  the  two  load  estimation
techniques, see Section  5.3.3.3.1 in this report.

References

Battaglin, W.A. and  D.A. Goolsby.  1995.  Spatial
   Data in Geographic Information System Format
   on Agricultural Chemical Use,  Land Use, and
   Cropping Practices in the United  States.   U.S.
   Geological  Survey,  Atlanta,  Georgia.   Water
   Resources Investigations Report 94-4176,87 pp.
   Available  from  U.S. Geological Survey  at
   http://pubs.usgs.gov/wri/wri944176/SHDRZ.
Battaglin, W.A. and D.A. Goolsby. 1996. Using GIS
   and  Regression to Estimate Annual Herbicide
   Concentrations in Outflow From Reservoirs in the
   Midwestern USA, 1992-93.  In:  Proceedings of
   the  American  Water  Resource  Association
   Annual   Symposium  on   GIS  and   Water
   Resources,  pp.  89-98.    American   Water
   Resources Association, Middleburg, Virginia.

Blanchard, P.E. and R.N. Lerch. 2000. Watershed
   Vulnerability to Losses of Agricultural Chemicals:
   Interactions of Chemistry, Hydrology, and Land-
   Use.  Environ. Sci. Technol., 34(16):3315-3322.

Capel, P.O. and S.J. Larson. 2001. Effect of Scale
   on the Behavior of Atrazine in Surface Waters.
   Environ. Sci. Technol., 35(4):648:657.

Dolan,  D.M., A.K. Yui,  and R.D. Geist.   1981.
   Evaluation of River Load Estimation Methods for
   Total Phosphorus.  J. Great Lakes  Res., 7(3):
   207-214.

Frank, R. and G.J. Sirons. 1979. Atrazine: Its Use
   in Corn  Production  and  Its Loss  to  Stream
   Waters  in Southern Ontario, 1975-1977.  Sci.
   Total Environ., 12(3):223-239.

Frank,  R. and L. Logan.  1988.   Pesticide and
   Industrial Chemical Residues at the Mouth of the
   Grand, Saugeen and Thames Rivers, Ontario,
   Canada, 1981-85.   Arch.  Environ. Contam.
   Toxicol.,  17(6):741-754.

Gianessi, L. P.  and C.M. Puffer.  1988.  Use of
   Selected  Pesticides  for  Agricultural   Crop
   Production in the United States, 1982-1985.  U.S.
   Department of Commerce,  National  Technical
   Information   Service,  Springfield,   Virginia.
   Document Number PB89-191100, 490 pp.

Good, D. and S. Irwin.  2007. Marketing and Outlook
   Briefs-2007 U.S. Corn Production Risks:  What
   Does History Teach  Us?   U.S. Department of
   Agricultural and Consumer Economics, University
   of Illinois at Urbana Champaign.   May 2007
   lssue/MOBR01-07.
                                              79

-------
Hall. D.W., I.E. Behrendt, and P.E. Hughes.  1998.
   Temperature, pH, Conductance, and Dissolved
   Oxygen in Cross Sections of 11  Lake Michigan
   Tributaries, 1994-95.  U.S. Geological Survey,
   Middleton, Wisconsin.  Open File Report 98-567,
   85pp.

Hall,  D.W.     2004.    Quality   Systems  and
   Implementation  Plan  (QSIP)  in  the Quality
   Assurance Project Plan for the LMM BP Modeling.
   In:  W.L. Richardson, D.D. Endicott, R.G.  Kreis,
   Jr.,  and  K.R.  Rygwelski  (Eds.),  The  Lake
   Michigan  Mass   Balance   Project  Quality
   Assurance Plan  for  Mathematical  Modeling,
   Appendix G,  pp.  233.    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.   EPA/600/R-
   047018,233pp.

Richards, R.P. 1994.  Tributary Loading Estimates
   for Selected Herbicides in Lake Erie Tributaries of
   Michigan  and  Ohio.    U.S.  Environmental
   Protection  Agency,   Great  Lakes  National
   Program Office, Chicago, Illinois.

Richards, R.P., D.B. Baker, J.W. Kramer, and D.E.
   Ewing.  1996.  Annual Loads of Herbicides in
   Lake Erie Tributaries of Michigan and Ohio.  J.
   Great Lakes Res., 22(2):414-428.
Schottler,  S.P., S.J.  Eisenreich,  and P.O.  Capel.
   1994.  Atrazine, Alachlor, and Cyanazine in  a
   Large Agricultural River System.  Environ. Sci.
   Technol, 28(6):1079-1089.

Squillace, P.J. and E.M. Thurman. 1992. Herbicide
   Transport in Rivers:  Importance of Hydrology
   and  Geochemistry  in  Nonpoint  Source
   Contamination.     Environ.  Sci.   Technol.,
   26(3):538-545.

U.S.  Geological Survey and S. Eisenreich.   1997.
   USGS  Field   Operation   Plan:     Tributary
   Monitoring, Version 1.  In: L. Blume (Ed.), Lake
   Michigan Mass Balance Study (LMMB) Methods
   Compendium,  Volume  1:   Sample Collection
   Techniques, pp. 215-219.  U.S. Environmental
   Protection  Agency,  Great  Lakes  National
   Program Office, Chicago, Illinois.  EPA/905/R-
   97/012a, 403 pp.

Wisconsin  Register.   2004.   Pesticide Product
   Restrictions.   State of Wisconsin,  Madison,
   Wisconsin. Document Number 586:1244-147.

Wisconsin  Register.  2005.  Atrazine Prohibition
   Areas, Appendix A. State of Wisconsin, Madison,
   Wisconsin. Document Number 591:149-251.
                                              80

-------
                                          PART 2

           LAKE MICHIGAN MASS  BALANCE PROJECT ATRAZINE
                          LOADINGS TO LAKE MICHIGAN
Chapter 3.  Estimation of Atrazine Loads
in Wet Deposition (Precipitation)

2.3.1      Atmospheric   Components
Considered in Modeling Atrazine in Lake
Michigan

Both the MICHTOX and LM2-Atrazine models utilize
annualized wet  deposition  loadings for long-term
simulations. However, LM3-Atrazine wet deposition
loadings were calculated on a daily basis to capture
seasonal loading variations.  LM3-Atrazine was used
to make predictions in lake segments on short-time
scales  in a fine-grid framework as a function of
seasonally varying loads - both wet deposition and
tributary.

Particulate deposition was  not  considered in  the
MICHTOX, LM2-Atrazine, and LM3-Atrazine models
because studies have shown that atrazine deposition
associated with atmospheric particulates represents
a minor fraction of the total deposition of atrazine
(Nations and Hallberg 1992; Siebers etal. 1994). In
the Lake Michigan Mass Balance Project (LMMBP)
(Section 1.3.2.2.2), the detection limit for atrazine
associated  with  atmospheric  particulates  was
relatively high.  As a consequence, there was a low
number of detects at  land-based  collection sites
positioned around the lake.  Attempts to measure
atrazine-associatedparticulatesover-the-lakeyielded
only two detects,  and both of  them were in  the
southernmost part of the lake near major atrazine
sources. To make an estimate of atrazine deposition
fluxes associated with particulates, one needs both
reliable measurements of atrazine concentration on
the particles and an estimate of the deposition rate of
the particles.   To calculate a rate of deposition,
particle sizes are needed. Particle size fractionation
was not part of the LMMBP analysis.

Some  researchers   have   attempted  to  make
estimates  of  atmospheric,  particulate-associated
atrazine fluxes to  Lake Michigan  using  some
assumptions about the particle sizes.   Miller et al.
(2000) roughly estimated that the load from particles
for the high-loading  spring months (April through
June,  1994-1995), could range from  230 to 1000
kg/yr. Schottlerand Eisenreich (1997) estimated that
the atrazine-associated particulate load to the lake
for the period 1991 to 1994 was approximately 160
kg/yr.  Sweet and Harlin  (1998) estimated that the
1994-1995 atrazine-associated particulate load to the
lake using  data from  April through July to be about
220 kg/yr.   Using these estimates,  as well as wet
deposition  and tributary loadings  for  1994,  the
relative contribution of dry particulate deposition to
the  total   load  of  atrazine  to   the   lake
(wet+dry+tributary) is 2.8% to 11.4%  (Miller et al.,
2000), 2.0% (Schottler and Eisenreich, 1997), and
2.7% for Sweet and Harlin (1998).  Note that these
estimates were based on particulates  collected at
land-based stations around the lake. However, other
than the two atrazine-associated particulate detects
in the southernmost  part of the lake,  we  have no
evidence that these loadings are occurring over-the-
lake.

Vapor phase concentrations of atrazine were used in
the models as a boundary condition;  please see
                                             81

-------
Parts  4  (LM2-Atrazine) or 5  (LM3-Atrazine) for
details.

2.3.2    Atrazine  Wet  Deposition  Load
Estimates Based on Measured Fluxes in
the Basin

Over-the-lake wet deposition of atrazine for  1991
(Figure 2.3.1) was based on  data collected  from
shore-based  samples   (Goolsby   et  al.  1993).
Goolsby's study area  included  Midwestern  and
Northeastern  states  in  a  geographic  rectangle
defined by the states North Dakota, Kansas, Virginia,
and Maine.  It is interesting to note that the total
amount of wet-deposited atrazine  in this region is
calculated to be 0.6% of the amount applied in the
    0         200 Miles
          I      I
        1    T
              300 Kilometers
Estimated atrazine
deposition in micrograms
per square meter
per year -1991

  1 Less than 10
   10 to 25
   26 to 50
   51 to 100
  I more than 100
Figure 2.3.1. Wet deposition (rain and snow) of
atrazine for 1991 for Midwestern United States
(Figure by W.A. Battaglin, U.S. Geological Survey,
1997).
                     region.   This is the same percentage  used to
                     estimate the atrazine tributary load export from the
                     Lake Michigan watershed.  Higher fluxes of atrazine
                     to Lake Michigan are noted in the southern part of
                     the lake compared to the northern part. This gradient
                     is the result of higher  use of the chemical in the
                     states south and west of the lake and wind  patterns.

                     Wet deposition data for 1994 and  1995 associated
                     with the LMMBP were received from Hornbuckle
                     (University of Iowa, personal communication, 1999).
                     These over-the-lake wet deposition estimates were
                     used in all three models.   Figure 2.3.2 depicts wet
                     deposition for the month of May 1994, and again the
                     southern region depicts higher atrazine fluxes. There
                     is a strong seasonal trend of wet deposition loadings
                     to the lake (Figure 2.3.3) - high loadings in the spring
                     and early summer and  very little loading during the
                     rest of the year.  Translating Hornbuckle's loadings
                     into wet deposition fluxes  over Lake Michigan and
                     Green Bay yielded a value  of 30.8 ug/m2/yr for 1994
                     and 1995.  A similar calculation  of flux  for  1991
                     (Figure 2.3.1) yielded a value of 45 ug/m2/yr.

                     Wet deposition to the lake other than  1991, 1994,
                     and 1995 was estimated from total  annual usage
                     estimates in a  similar manner as  described  for
                     historical tributary loadings. However,  instead of a
                     "Tributary Load Ratio,"  a "Precipitation Load Ratio"
                     was defined.  Precipitation  ratios were calculated as
                     an  average for  years  1991,  1994,  and  1995  as
                     follows:

                     Precipitation Load Ratio =  I Precipitation

                     Load to a Mode! Segment \j\ Total Annual

                     USA Atrazine Usage i                  (2.3.1)
For any year (y), where  only total annual United
States usage is known,  a  segment  load was
calculated utilizing the precipitation ratio:

Precipitation Load =
! Precipitation Load Ratio \ x

I Total Annual USA Usage Year i y 11  (2.3.2)

Along with total  annual usage estimates, annual
atrazine wet deposition and tributary loadings for
Lake Michigan and Green Bay are depicted in Figure
                                              82

-------
                         wet deposition
                          atrazine load
                           (kg/month)
                           May 1994
Figure  2.3.2.    Gradients of  atrazine  in  wet
deposition loadings over Lake Michigan for May
1994.
1200-
£1000-
f. 80°"
g BOO-
a
§ 400-
200-
n-



















n _
    Mar. May  Jul.  Sep. Nov. Jan. Mar. May  Jul.  Sep.
    1994 1994  1994 1994 1994 1995 1995 1995 1995 1995

Figure  2.3.3.   Seasonality  of  atrazine  wet
deposition loadings to Lake Michigan for 1994-
1995.
                                                 2.3.4.  The wet deposition load calculated for 1995
                                                 was very low compared to 1994 (Figures 2.3.3 and
                                                 2.3.4). It is believed that a cold and wet spring in the
                                                 major corn-growing regions of the United States may
                                                 explain this low estimate (see Section 1.3.2.2.3).
n
£
5 45
o
^ 40-
"5
| 35-

1 30-
I25'

Si 20-
(0
U> 4 r-
3 1 3-
ISA annual
a on o












J_






















n total annual usage
• historical tributary loa
D wet deposition










f—i









r~i









J

























































>-
"&
ygooo e
•8000 ra
O
•7000 c
o
•6000 =
•5000 |
T>
•4000 -
s
•3000 S
•n
•2000 i
•1000 jo
••0 -Q
                                                    1964 1968 1972 1976 1980 1984 1988 1992 1996 2000    -
                                                                     year

                                                 Figure 2.3.4. Total atrazine tributary loading and
                                                 wet  deposition  loading  estimates  to  Lake
                                                 Michigan.
2.3.3    Atrazine  Wet  Deposition  and
Tributary Loads for MICHTOX and LM2-
Atrazine

Both tributary and  precipitation  loadings for  the
MICHTOX and LM2-Atrazine models' surface water
segments are shown  in Figures  2.3.5 and  2.3.6,
respectively.  In MICHTOX, the southern third of the
lake is identified as segment 1,  the central lake
region is segment 2, and the northernmost part of the
lake is segment 3 (see  Part 3, Figure 3.1).  Note that
total loadings are greater in the southern  region of
the lake compared to the northern region.  In LM2-
Atrazine, the southern third of the lake is represented
by segments 1 and 2; central lake, 3 and 4; and the
northern  lake, 5 and 6. Segments 2, 4, and 6  are
located on the eastern  side of the  lake.  The rest of
the segments are located in Green Bay.  See Figure
4.1 in Part 4 for a graphic identifying segments for
LM2-Atrazine. The highest load to LM2-Atrazine is in
segment 2.   Both  MICHTOX and  LM2-Atrazine
perform a linear interpolation  to  estimate missing
loads between dates that have known loads. For the
whole lake, the total  annual load estimates were the
                                              83

-------
4500

4000 -

3500 -
3000 -

2500 -
                                                             • 1994tributary loads
                                                             D1995 tributary loads
                                                             • 1994wet deposition
                                                             n 1995 wet deposition
                               234567
                                         MICHTOX Segment
Figure 2.3.5.  Tributary and wet deposition loadings to MICHTOX model segments for 1994 and 1995.
t DU U



3000
_._._
-^ ZDUU
1 cnn

1 UUU
•^nn -
n -









[Til


-





= 	












• 199 4 tributary loads
D1995 tributary loads
D1995 wet deposition





, ^
11 , DTl , HTI , •-• i •-•, — — , PTL
                                       4567
                                        LM2-Atraiine Segment
                                                              10
Figure 2.3.6.  Tributary and wet deposition loadings to LM2-Atrazine model segments for 1994 and
1995.
                                              84

-------
same for MICHTOX  and LM2-Atrazine; however,
MICHTOX  had  seven  receiving  surface water
segments and LM2-Atrazine had 10. See Part 5 for
information  on tributary loads and wet  deposition
estimates used in LM3-Atrazine.

References

Goolsby, D.A.,  E.M.  Thurman,  M.L.  Pomes,  M.
   Meyer, and W.A. Battaglin. 1993. Occurrence,
   Deposition,  and  Long  Range Transport  of
   Herbicides in Precipitation in the Midwestern and
   Northeastern United States. In: D.A. Goolsby,
   LL.  Boyer, and G.E. Mallard  (Eds.), Selected
   Papers on Agricultural Chemicals in  the Water
   Resources of the Midcontinental United States,
   pp. 75-89.   U.S.  Geological Survey,  Denver,
   Colorado. Document Number 93-418, 89 pp.

Miller, S.M.,  C.W. Sweet, J.V. DePinto, and K,C.
   Hornbuckle.  2000.  Atrazine and Nutrients in
   Precipitation:  Results From the Lake Michigan
   Mass Balance Study.   Environ. Sci. Technol.,
   34(1):55-61.
Nations, B.K. and G.R. Hallberg. 1992. Pesticides
   in  Iowa Precipitation.    J.  Environ.  Qual.,
   21(3):486-492.

Siebers, J., D. Gottschild,  and H.G. Nolting.  1994.
   Pesticides in Precipitation in Northern Germany.
   Chemosphere, 28(8):1559-1570.

Schottler, S.P.  and S.J. Eisenreich.  1997.  Mass
   Balance Model to  Quantify Atrazine  Sources,
   Transformation Rates, and Trends in the Great
   Lakes. Environ. Sci. Technol., 31 (9):2616-2625.

Sweet, C.W. and K.S.  Harlin.  1998.  Atmospheric
   Deposition   of  Atrazine  to  Lake  Michigan.
   Presented at the Air and Waste  Management
   Association's   91st   Annual  Meeting  and
   Exhibition,  June   14-18,  1998,  San  Diego,
   California.     Illinois  State  Water   Survey,
   Champaign, Illinois. ReportNumber98-TA37.02.
                                              85

-------
                                        PARTS

                 LAKE MICHIGAN MASS BALANCE PROJECT
                    LEVEL 1 MODEL:  MICHTOX-ATRAZINE
                                   Kenneth R. Rygwelski
                         United States Environmental Protection Agency
                             Office of Research and Development
                  National Health and Environmental Effects Research Laboratory
                               Mid-Continent Ecology Division
                      Large Lakes and Rivers Forecasting Research Branch
                                Large Lakes Research Station
                                     9311 Groh Road
                                 Grosse lie, Michigan 48138
3.1   MICHTOX-Atrazine  Executive
Summary

Our coarse-segmented model, MICHTOX, was run in
a hindcast and forecast mode under various load
modification scenarios. A calibration run based on
average boundary conditions using historical loadings
of atrazine  to Lake  Michigan   suggests  that
approximately 1% of the atrazine in the lake decays
each year.  In the forecasts of alternate futures,
eliminating all  loadings to the lake resulted in the
largest decline in model predictions. A total loading
reduction of approximately 37%, if implemented on
January 1, 2005, would have been needed in order
to prevent atrazine concentrations from increasing
further than  above those that were observed in the
lake on January 1, 2005.

3.2 MICHTOX-Atrazine  Recommendations

For long-term  forecasts, future  modeling  efforts
should utilize LM2-Atrazine as a model because it is
more  highly  resolved  and  has  advective  and
dispersive components that were derived from a
hydrodynamic  model  (see  Part  4).    The
hydrodynamic model components can be considered
to best represent "average" lake conditions because
the  various hydrodynamic forcing functions were
considered to be average (see Part 1, Chapter 4).

3.3 Model  Description

3.3.1  Model Overview

For calibration  purposes,  the  MICHTOX  mass
balance model (Endicott et a/., 2005) was used in a
hindcast mode to simulate atrazine concentrations in
Lake Michigan and Green Bay in response to mass
loadings  to  these  systems from  the  time  of
introduction in  the early 1960s to  1995.  The
calibrated model was then used in a forecast mode
to predict lake-wide concentrations in Lake Michigan
as a function of various loading scenarios.

MICHTOX was adapted from the  general water
quality model WASP4 (Ambrose et a/., 1988). The
model solves mass balance  equations based on a
finite volume spatial discretization (Thomann and
                                           86

-------
Mueller, 1987)  and Euler time  integration.   The
MICHTOX  model  framework   is  capable   of
incorporating  a full range of transport and fate
processes such as advection, dispersion,  particle
settling, sediment resuspension,  sediment burial,
transport in  sediment pore water,  partitioning  to
particles,  chemical  reaction,  volatilization,  and
absorption.

3.3.2  MICHTOX Model Segmentation and
Circulation

The segmentation schematic for Lake Michigan and
Green Bay is depicted in Figure 3.1.  Lake Michigan
and Green Bay have nine water segments.  Surface
segments 1 (southern lake), 2 (central lake),  and 3
(northern lake) cover the entire main lake. Segments
5 (southern bay),  6 (central bay), and 7 (northern
bay) coverGreen Bay. Hypolimnetic water segments
in the main lake are numbered 8 (southern lake), 9
(central lake), and 10 (northern lake).  Segment 4 is
a very small segment located in the lower Fox  River.
During  a period of approximately 100 days  in the
summer, flow and exchange across  the Straits  of
Mackinac occurs in two discrete layers between the
surface water (segment 3) and  Lake  Huron and
between the deep hypolimnetic water (segment 10)
and  water that  primarily originates from  Lake
Superior (Quinn, 1977) mixed with water from Lake
Huron.   During this period of stratification, surface
layer flow (segment 3) is from Lake Michigan to Lake
Huron,  and a deeper return flow to Lake Michigan is
observed.  It has been observed that Lake Superior
water discharging from the St. Marys River travels in
a persistent  westerly direction during stratification
and constitutes a significant component of the  return
flow to  Lake Michigan (Ayers et a/., 1956; Schelske
et a/. 1976; Saylor, J., National Oceanographic and
Atmospheric   Administration,   personal
communication, 1998).    During the  unstratified
period, all of the flow is from Lake Michigan to Lake
Huron.

Two-layered flow has been observed at the mouth of
Green Bay during thermal stratification (Martin et a/.,
1995); however, this structured flow process was not
incorporated  in the MICHTOX model framework.
MICHTOX incorporates the flows between Green
Bay and Lake Michigan as net flows.
fir
|m

__I t~>. • J
L— I wSter column
Figure 3.1. MICHTOX model segmentation.
The MICHTOX model has two water column layers
for Lake Michigan to simulate the effects of summer
stratification of the lake. Also, the model consists of
just three horizontal compartments in the main lake.
This low spatial resolution was considered adequate
to address open-lake concentrations. Water column
concentration profiles  of  atrazine  at  10  stations,
representing fourto 10 depths per station, showed no
vertical gradients during lake stratification for  the
years 1991 and 1992 (Schottler and  Eisenreich,
1997)  and   1994-1995  (Brent et  a/.,  2001).
Furthermore, they reported that analysis of data from
their 10 lake stations that covered a central north-
south axis and an  east-west  axis  showed  no
horizontal gradients in atrazine concentrations in the
lake.
                                              87

-------
MICHTOX exchange coefficients were taken from the
literature.   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).  Horizontal
exchange coefficients in Green Bay were calibrated
to reproduce chloride gradients.  In the main lake,
however, horizontal  exchange  coefficients  were
taken from work by Thomann et a/. (1979) on Lake
Ontario.

Flows in the lake were based on the whole-lake water
balance by Quinn (1977), which provided monthly
average changes in storage, tributary flow, outflow,
diversion,  precipitation,  and  evaporation.   The
hydraulic residence  time  (volume/outflow)  for the
main  lake was estimated to be 62 years (Quinn,
1992).

3.4  MICHTOX Model Application  to Lake
Michigan

3.4.1  Screening Model Application

A screening-level model of MICHTOX was applied
before Lake  Michigan  Mass   Balance  Project
(LMMBP) loadings were available (Rygwelski et a/.,
1999). This early MICHTOX application assumed
that volatilization was negligible due to a very small
Henry's law constant of 8.1 x 10"8  (U.S.  Department
of Agriculture, 2001) and that the chemical could be
modeled as a conservative substance.

For  this  screening  model, tributary  loads  were
estimated based on atrazine applications to the basin
in 1992 and 1993  using  algorithms  identified in
Equation 2.2.1. The watershed export percentage
(WEP) used was 0.6% (see Table 2.2.2). Inorderto
predict loadings for years when application data were
not available, the loads estimated for 1992 and 1993
were  divided by estimates  of total annual United
States usage of atrazine using Equation 2.2.2 (no
annual United States usage estimate was available
for 1992, so an estimate for that year was calculated
as a mean of United  States usage reported for 1991
and 1993). A mean of these two ratios was assumed
to be constant over the entire historical  record of
atrazine usage in the basin.  For years where only
total annual usage  was available,  an  estimate of
loadings  could  be determined by multiplying the
mean load ratio by total annual usage.

Loadings of wet deposition to the lake were obtained
for  1991  (Goolsby et  a/.,  1993).    These  wet
deposition  loads   were   based   on  actual
measurements of atrazine in rain and snow.  In a
similar manner as was calculated for tributary loads,
the load from Goolsby was divided by a mean of the
total annual United States usage of atrazine for the
years 1992 and 1993. Usage in the United States
between  1989 through 1995 was relatively constant
so errors in substituting a mean of 1992  and  1993
usage for 1991 were believed to be small.  In a
manner similar to the mean Tributary Load Ratio, a
mean atmospheric load ratio was used to estimate
historical  wet deposition to the lake.  See Figure 3.2
for both tributary and precipitation atrazine loads.
   7000
   6000-

 "35000 -
 g>4000

 § 3000
   2000-

   1000-

     0
                               Precipitation
      1964  1968 1972  1976 1980  1984 1988  1992 1996
                        year

Figure 3.2. Total annual estimated tributary and
precipitation  loadings  of  atrazine   to   Lake
Michigan.
Using the load history and assuming that atrazine
decay is zero with negligible volatilization, a model
hindcast run starting in 1964 yielded a good fit with
lake data (see Figure 3.3). The results shown in the
figure are from the  main lake only and does not
include Green Bay. The initial conditions in the lake
model were set to an atrazine concentration of zero.
No  calibration of the  model  was needed.   Also
depicted are the effects of using the upper and lower
95% confidence intervals on the 0.6% WEP reported
in the literature for moderate textured soils (see Part
2, Chapter 2). As a sensitivity test, a hypothetical
0.05  per  year  overall  decay   constant  was
incorporated  into  the model.  The model is very
                                              88

-------
       ^— Mean watershed export percentage 0.60
           decay constant 0.0/year

       	Upper 95% confidence interval
           watershed export percentage 0.84
           decay constant 0.0/year

       — — Lower 95% confidence interval
           watershed export percentage 0.36
           decay constant 0.0/year
           Mean watershed export percentage 0.60
           decay constant 0.05/year
           Field Data ±1 standard deviation
    1964  1968 1972 1976  1980  1984  1988
                       year
                                      1992  1996
Figure  3.3.    A  comparison  of  MICHTOX -
Predicted  atrazine  concentrations  in  Lake
Michigan to averaged Lake Michigan data for the
years 1991, 1992, and 1995 are depicted.  Field
data for 1991 and 1992  were obtained from the
literature (Schottler  and Eisenreich, 1997) and
data for 1995 are LMMBP data.
sensitive to this  decay as  shown  in Figure 3.3.
"Decay" as used in this paper is internal decay likely
due to the combined effects of abiotic and biotic
transformation of  atrazine to degradation products.
Considering that the model required no calibration
and relied mostly on data  from the  literature,  it
performed remarkably well.

3.4.2    Enhanced Screening  Model
Application

As additional county-level atrazine application data in
the basin and total  United States usage estimates
became  available,  MICHTOX  modeling in Lake
Michigan  continued to  develop.   In  the  earlier
screening  model  application,  only  two  years  of
county-level atrazine application data were available.
For the  enhanced  screening-level  model, seven
years of application data were available and used.
Also, due to label changes that  lowered application
amounts and  established  planting  setbacks from
water bodies in 1990 and 1992, a decision was made
to use two Tributary Load Ratios in order to address
atrazine application practices for pre- and post-label
changes.  New data from the LMMBP also became
available to modelers.  With these additional data,
loading  ratios for both  the tributaries  and  wet
deposition were updated (see Part 2, Chapters 2 and
3).  The model was calibrated by determining a total
decay that would yield a best fit of the  model to
observations in the lake.  Also, several forecasting
scenarios were run with the model. The efforts of
this additional modeling are described in the following
sections of this part.

3.4.2.1  Field Data

See Part 1, Chapter 3 for atrazine data obtained from
lake,  tributaries,   and  atmospheric  components
samples.

3.4.2.2   Model Assumptions  and  Calibration
Procedures

Due to atrazine's physical and chemical properties
(Part  1,  Chapters  1 and 3), processes modeled
included only advection,  dispersion,  and reaction
(decay).

Model  processes  involving   sediments   and
particulates in the water column were not included in
the MICHTOX model  runs  because  atrazine is
primarily in the dissolved state in  surface waters;
therefore, any processes that involve  sediment or
suspended  particle  interactions  are  of   minor
significance (Section 1.2.2).

A literature review of atrazine degradation processes
in surface freshwater presented in Part 1, Chapter 2
suggests that degradation is hindered in freshwater
such as in Lake  Michigan where the water is cold,
has low solids concentrations with low dissolved
organic  carbon,  has  a  high pH,  and  has  low
concentrations of  nitrate ions.    Degradation of
atrazine is  known to occur through either biotic or
abiotic   processes   in  some  environmental
compartments. Given the lack of any Lake Michigan-
specific   kinetic  information on  any  of  these
processes, the approach  taken in MICHTOX was to
estimate the loading history  of atrazine to the lake
and find an overall first-order loss rate constant to fit
the model  to observations  of atrazine in the lake
                                               89

-------
water.   Loadings were not part of the calibration
procedure.  Considerable effort was expended to
ensure that loadings were fairly represented in the
model (Part 2, Chapters 2 and 3).

Due to  a  very   small  Henry's  law  constant,
volatilization and absorption were not simulated.

3.4.2.3  Tributary Loadings

It was assumed that a WEP of 0.6% derived from the
literature for fine/moderate textured soils adequately
described the overall WEP of  the Lake Michigan
watershed.  This WEP, along with historical annual
atrazine usage  in the  United States was  used to
calculate atrazine loadings from the tributaries to the
lake. For a complete discussion on the WEP method
used to estimate MICHTOX loadings, please see
Part 2,  Chapter 2.

Utilizing flow and concentration data, the Stratified
Beale Ratio Estimator (SBRE) method was used to
estimate tributary  loads  in the  11   monitored
tributaries during the LMMBP.  Also, estimates of
loads from the unmonitored watersheds were made.
However,  loads  were  apparently  missed  and
therefore MICHTOX tributary loads were only based
on  WEP, county-level application data, and  total
United  States annual usage records. See Section
5.3.3.3.1 for a discussion of this topic.

3.4.2.4  Atmospheric Loadings

Loading  estimates of wet deposition  to Lake
Michigan and Green Bay were made for each of the
top surface water segments.  These loadings were
estimated for MICHTOX per discussion in  Part 2,
Chapter 3.

3.4.2.5  Model Confirmation

In  2005, atrazine water samples were collected in
Lake Michigan  for the  purpose of confirming the
model  predictions.   However,  as of this  printing,
these analyses were not available.

3.4.2.6  Model Application (Scenarios)

The calibration of the model was undertaken using
three scenarios (1, 2, and 3) that included lower
boundary conditions, upper boundary conditions, and
a most likely or average boundary condition scenario,
respectively.  All of these model runs started on
January 1, 1963.  The  model was  calibrated by
finding an appropriate internal decay until the model
output  best   matched  the  observed  atrazine
concentration in the lake for samples taken in 1991,
1992, 1994, and 1995.

The scenarios 4 through 7 are referred to as load
reduction  scenarios.   These  are  not necessarily
management scenarios  but can  give  managers
insight as to which loads are important in the model
and  environment for the purpose  of  predicting
concentrations of atrazine in the lake. It is believed
that they bound the entire range of potential loads
and provide some specific load scenarios within the
range. Scenario 3 was used to simulate conditions
from January 1,1996 through December 31,2004 for
scenarios  4 through  7 described below.   When
December 31, 2004 is reached, each of the load
reduction scenarios 4 through 7 began on January 1,
2005 and were run for a period of 50 years.

The Lake Superior boundary condition was assumed
linear during  the period modeled (0 ng/L at the
beginning of year 1963 and 3.5 ng/L at 1994) and
likewise for the Lake  Huron boundary condition (0
ng/L at the beginning of year 1963 and 23 ng/L at
1992). The boundary conditions were assumed to be
zero in 1963  because this was the year when  the
herbicide was first introduced to the basin.  Lake
Superior and Lake Huron atrazine concentrations for
the years 1993 and 1992, respectively, were based
on  measurements  of atrazine  in  these  lakes
(Schottler and Eisenreich, 1994).   While the Lake
Superior flow component of the return flow to Lake
Michigan is primarily characteristic of concentrations
of atrazine in Lake Superior, the actual concentration
is probably somewhere between Lake Superior and
Lake  Michigan  due to some  mixing (see section
3.3.2).

1. Calibration Based  on  Upper  Estimate  of
   Boundary Conditions -  The summer inflow
   concentration at the Straits of Mackinac was
   assumed to be 100% Lake Huron water.  Lake
   Huron water started at 0 ng/L and  was assumed
   to linearly rise to  23 ng/L as observed in 1995
   (Station 54) and then held constant at that level
   for the remainder of  the simulation.  Tributary
   loading  projections  were  set  equal  to an
                                              90

-------
   averageof loadings for 1995 and 1998, but prior
   to that time, the historical loading estimates were
   used. Wet deposition projections were set equal
   to an average of loads for 1978 through 1998.
   Wet deposition loadings before that were based
   on  historical load estimates.  The model was
   calibrated by adjusting the overall internal decay
   to  best  match  whole-lake  volume-weighted
   average concentration.

2.  Calibration  Based  on  Lower  Estimate  of
   Boundary Conditions -  The summer inflow
   concentration  at the Straits of  Mackinac was
   assumed to be 100% Lake Superior water. Lake
   Superior was assumed to begin with an atrazine
   concentration of 0  ng/L, was then assumed to
   linearly  rise to 3.5  ng/L as observed in 1994
   (Schottler and Eisenreich,  1997), and was then
   held constant at that level  for the remainder of
   the simulation. Tributary loading projections were
   set equal to an average of loadings for 1995 and
   1998, but prior to that time, the loading  history
   estimates were used. Wet deposition projections
   were set equal to an average of loads for 1978
   through 1998. Wet deposition loadings before
   that time were based on historical load estimates.
   The model was calibrated by adjusting the overall
   internal decay to best match whole-lake volume-
   weighted average concentration.

3.  Calibration  Based on "Average"  Boundary
   Conditions  - The  inflow  concentration at the
   Straits of Mackinac was assumed to be 50%
   Lake Superior and 50% Lake Huron water. This
   mix of water  was  assumed to  begin with  an
   atrazine concentration of 0 ng/L, was assumed to
   linearly rise to 13.25or[(3.5+23)/2] ng/L in 1995,
   and was then  held  constant at that level for the
   remainder of the simulation.  Tributary  loading
   projections were set equal to an average of
   loadings for 1995 and 1998, but prior to that time,
   loading  history  estimates were  used.    Wet
   deposition  projections  were set  equal  to  an
   average of loads for 1978 through 1998.  Wet
   deposition loadings before that were based  on
   historical  load estimates.   The  model was
   calibrated by adjusting the overall internal decay
   to  best  match  whole-lake  volume-weighted
   average atrazine concentration.
4.  Virtual Elimination (Lower Bound  on Model
   Prediction) - This scenario simulated a 100%
   reduction of tributary and atmospheric loads. For
   the  projections,   the   Lake  Huron/Superior
   boundary conditions were set equal to zero. This
   scenario was  run using scenario 3 for predictions
   leading up to the date when the virtual elimination
   scenario was to take place.

5.  No Tributary Loads - This scenario simulated a
   100%  reduction  of  tributary loadings.   Wet
   deposition loads were set equal to an average of
   loads for 1978 through 1998. This scenario was
   run using scenario 3 for predictions leading up to
   the date when the 100% tributary load reduction
   scenario began.

6.  No Wet Atmospheric Deposition Loadings -
   Tributary loads were set equal to an average of
   loadings for 1995 and  1998.  Atmospheric wet
   deposition  loadings were decreased by 100%.
   This scenario was run using scenario 3 for
   predictions leading up to the date when the 100%
   atmospheric load  reduction scenario began.

7.  No Further Degradation of Lake Water Quality
   - A total load (tributary  and wet deposition) was
   determined such that  no further increase in lake-
   wide   volume-weighted   concentration  was
   observed starting in January 1,2005. Up through
   December 31, 2004, scenario 3 was used.

3.4.2.7 Discussion of Results

Total internal degradation of atrazine in the water (kj
determined by model calibration  was  low in  all
scenarios where evaluated (see Figures 3.4and 3.5).
These rates of decay for scenarios 1, 2, and 3 were
0.0125/yr, 0.008/yr, and  0.01/yr, respectively.  For
the  calibration  based   on  average   boundary
conditions, MICHTOXpredictsthat approximately 1%
of the atrazine in the  lake decays each year due to
some combination of abiotic and biotic decay in the
lake.

Decay can be related to the half-life of the chemical
in  the lake by the following:
Half- Life =  f1/2 =  (In 2)/kd
(3.1)
                                              91

-------
                                                        Field Data +/-1 standard deviation


                                                        Scenario 3; kinetic decay 1%/yr


                                                        Scenario 2: kinetic decay 0.8%/yr


                                                        Scenario 1; kinetic decay 1.25%/yr
                           Sept. 9
                            1976
May19
 1990
Jan. 26
 2004
Oct. 4
2017
June 13
 2031
Feb. 19
 2045
Oct. 29
 2058
                                                    date
Figure 3.4. Lake Michigan (open-lake) forecast scenarios:  1 -upper estimate of boundary condition,
2 - lower estimate of boundary condition, and 3 - estimate of average boundary condition.
                                          Field Data +/-1 standard deviation

                                          Scenario 4: virtual elimination

                                          Scenario 5; no tributary loads

                                          Scenario 6; no wet deposition

                                          Scenario 7: 37% total load reduction
                                              Jan. 26     Oct. 4
                                               2004      2017
                                                    date
                            June 13
                             2031
                            Feb. 19
                             2045
                            Oct. 29
                            2058
Figure 3.5. Lake Michigan (open-lake) hindcast and scenario forecasts: 4-virtual elimination of all
loadings and 0.0 ng/L atrazine at the Straits of Mackinac boundary, 5 - no tributary loads, 6 - no wet
deposition, 7 - no further degradation of lake water quality.
                                                  92

-------
Assuming that scenario 3 captures typical conditions,
then the 1% internal  decay associated  with  this
scenario represents a half-life of the chemical in the
lake of 69.3 years. In scenario 3,  the water at the
Straits of Mackinac was assumed to be half Lake
Superior water and half Lake Huron water and is
believed to be a fair assessment of the conditions
during summer stratification.  It is  intuitive that the
decay rate  associated with scenario 1 that has the
highestboundary condition concentrations of atrazine
(assumed to be all Lake Huron water) is the one with
the highest decay rate because higher boundary
concentrations will mean  that  more  atrazine is
transported into the lake at the Straits of Mackinac.
This higher loading will result in  a  higher  decay
needed in the modeling  calibration exercise in order
for model  output to  match  observations.   The
opposite argument is true for the scenario where the
boundary condition at  the Straits of Mackinac is
based solely on the lower concentrations of atrazine
from Lake Superior.

In the forecasts of alternate futures (Figure 3.5),
eliminating  all  loadings to the lake resulted in the
largest atrazine decline in model predictions. A total
loading  reduction  of  approximately  37%,   if
implemented on January 1, 2005, would be needed
in order to  prevent atrazine  concentrations from
increasing higher than  what was estimated in the
lake on January 1, 2005.  If only the atmospheric
loadings ceased  (scenario 6), concentrations would
continue to increase.  However, if only the tributary
loadings ceased  (scenario 5), concentrations in the
lake would decline relative to scenario 3 predictions.

References

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.
Ayers, J.C., D.V. Anderson, D.C. Chandler, and G.H.
   Lauff.  1956.  Currents  and Water Masses of
   Lake Huron  (1954 Synoptic Surveys).   The
   University of Michigan, Great Lakes Research
   Institute, Ann Arbor, Michigan. Technical Paper
   Number 1, 101 pp.

Brent,  R.N., J.  Schofield, and K.  Miller.   2001.
   Results of the Lake  Michigan  Mass  Balance
   Study:    Atrazine   Data  Report.     U.S.
   Environmental Protection Agency, Great Lakes
   National Program Office, Chicago, Illinois.  EPA/
   905/R-01/010, 92pp.

Endicott D.D.,  W.L.  Richardson, and D.J. Kandt.
   2005.  1992 MICHTOX: A Mass Balance and
   Bioaccumulation Model for Toxic Chemicals in
   Lake Michigan, Part 1. In:  R. Rossmann (Ed.),
   MICHTOX: A Mass Balance and Bioaccumulation
   Model for Toxic Chemicals in Lake Michigan.
   U.S. Environmental Protection Agency,  Office of
   Research and Development, National Health and
   Environmental Effects Research Laboratory, Mid-
   Continent  Ecology   Division,   Large  Lakes
   Research  Station,   Grosse  lie,  Michigan.
   EPA/600/R-05/158, 140 pp.

Goolsby, D.A.,  E.M. Thurman, M.L Pomes, M.
   Meyer, and W.A. Battaglin. 1993. Occurrence,
   Deposition,  and Long  Range  Transport  of
   Herbicides in Precipitation in the Midwestern and
   Northeastern United States.  In:   D.A. Goolsby,
   LL. Boyer,  and G.E.  Mallard (Eds.), Selected
   Papers on Agricultural Chemicals in the Water
   Resources of the Midcontinental  United States,
   pp.  75-89.    U.S. Geological Survey,  Denver,
   Colorado. Document Number: 93-418,  89 pp.

Martin, S.C., S.C. Hinz, P.W. Rodgers, V.J. Bierman,
   Jr.,  J.V. DePinto,  and T.C. Young.    1995.
   Calibration of a  Hydraulic Transport Model for
   Green Bay, Lake Michigan. J. Great Lakes Res.,
   21(4):599-609.

Quinn,  F.H.  1977.  Annual  and Seasonal  Flow
   Variations Through the  Straits  of  Mackinac.
   Water Resources Res., 13(1):137-144.

Quinn, F.H.  1992.  Hydraulic Residence Times for
   the  Laurentian Great Lakes. J.  Great Lakes
   Res., 18(1):22-28.
                                              93

-------
Rodgers, P.W. and O.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.

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.

Schelske, C.L., E.F. Stoermer, J.E. Gannon, and
   M.S.Simmons.  1976. Biological, Chemical, and
   Physical Relationships in the Straits of Mackinac.
   U.S. Environmental Protection Agency, Office of
   Research  and   Development,   Environmental
   Research  Laboratory-Duluth,   Large  Lakes
   Research  Station,  Grosse  lie,   Michigan.
   EPA/600/3-76/095, 267 pp.

Schottler, S.P. and S.J. Eisenreich.  1997.  Mass
   Balance  Model  to  Quantify Atrazine  Sources,
   Transformation Rates, and Trends in the Great
   Lakes. Environ. Sci. Technol., 31(9):2616-2625.
Schottler,  S.P.  and  S.J.  Eisenreich.     1994.
   Herbicides in the Great Lakes.   Environ. Sci.
   Technol., 28(12):2228-2232.

Thomann, R.V. and J.A. Mueller. 1987. Principlesof
   Surface Water Quality Modeling and Control.
   HarperCollins Publishers, Inc., New York, New
   York.

Thomann, R.V., R.P.  Wnfield.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, 136pp.

U.S.  Department of Agriculture. 2001. Agriculture
   Research Service Pesticide Properties. Available
   from   U.S.   Department  of  Agriculture  at
   http://www.ars.usda.gov.
                                              94

-------
                                         PART 4

                  LAKE MICHIGAN MASS BALANCE  PROJECT
                        LEVEL 2 MODEL:  LM2-ATRAZINE
                                    Kenneth R. Rygwelski
                         United States Environmental Protection Agency
                             Office of Research and Development
                  National Health and Environmental Effects Research Laboratory
                                Mid-Continent Ecology Division
                      Large Lakes and Rivers Forecasting  Research Branch
                                           and
                                       Xiaomi Zhang
                             Z-Tech, an ICF International Company
                                Large Lakes Research Station
                                      9311 Groh Road
                                 Grosse lie, Michigan 48138
4.1 LM2-Atrazine Executive Summary

LM-2 Atrazine was run in a hindcast and forecast
mode  under various  load and modified  boundary
condition  scenarios.  A calibration run, based on
average boundary conditions using historical loadings
of atrazine to Lake Michigan, suggests that only 0.9%
of the atrazine in the lake decays each year. Net
volatilization of atrazine is  negligible.  Tributaries,
transporting the atrazine run-off load from farm fields,
contribute most of  the  atrazine load  to Lake
Michigan.   In the  forecasts of alternate futures,
eliminating all loadings to the lake resulted in the
largest decline in model predictions.  A total load
reduction  of approximately 35%, if implemented on
January 1, 2005, would have been needed in order
to prevent atrazine concentrations from increasing
above those that were estimated in the lake on
January 1, 2005.
4.2 LM2-Atrazine Recommendations

Due to its fast run-time speed, LM2-Atrazine can be
used to perform long-term model forecasts of lake
concentrations. As additional loading data become
available, the updated loading history can easily be
added to existing model input files.  For additional
model confirmation purposes, it is recommended that
the model predictions be compared to data from lake
samples that were collected in 2005 when these data
become available.

4.3 Model Description

4.3.1 Model Overview

As one of the models in the Lake Michigan Mass
Balance  Project   (LMMBP),   LM2-Toxic  was
specifically developed to simulate the transport and
fate  of  hydrophobic  toxic  chemicals,  such as
polychlorinated biphenyl (PCB) congeners, in both
                                            95

-------
the water and sediment of Lake Michigan (Zhang,
2006).  LM2-Toxic is a descendant of the WASP4
water quality modeling framework (Ambrose et al.,
1988).  The model solves mass balance equations
based  on a  finite  volume  spatial  discretization
(Thomann and  Mueller,  1987)  and  Euler  time
integration. Compared to WASP4, LM2-Toxichasan
updated air-water exchange formulation that includes
a Henry's law temperature-corrected coefficient as
described by  Bamford et al. (1999); water phase
mass transfer coefficient per Wanninkhoff  et al.
(1991); and the air phase mass transfer coefficient by
Schwarzenbach et al. (1993).  The LM2-Toxic model
is capable of incorporating a full range of transport
and fate processes such as advection, dispersion,
particle settling, sediment resuspension,  sediment
burial, transport in sediment pore water, partitioning
to  particles in  the  water column and  sediment,
reaction, volatilization, and gas absorption.

The  LM2-Atrazine model is  identical to LM2-Toxic
except for differences in the volatilization algorithms.
In  LM2-Atrazine, the algorithm for calculating the
temperature-dependent  Henry's   law  coefficient
follows that of Scholtzef al. (1999) and Miller (1999).
The dimensionless value for the Henry's law constant
was set to 8.1 x10"8(U.S. Department of Agriculture,
2001).   The water  and air  phase mass transfer
coefficients were that of the  O'Connor  "long form"
and O'Connor, respectively (O'Connor, 1983). The
volatilization  algorithm differences between LM2-
Toxic and  LM2-Atrazine would not be expected to
have  a  significant  impact  on  atrazine model
predictions because of the low value of the Henry's
law constant for atrazine.  As a  non-hydrophobic
chemical,  atrazine   was  not  associated  with
particulates in LM2-Atrazine.  Therefore, processes
such as resuspension, settling, burial in sediment,
transport in sediment pore water, and partitioning to
solids in the water column and sediment were not
operative. Processes such as advection, dispersion,
reaction,  volatilization,  and  gas  absorption were
active.  For information on the physical and chemical
properties of atrazine, see Part 1, Chapter 2.

In a manner similar to MICHTOX (see Part 3), LM2-
Atrazine was used in a hindcast mode  to simulate
atrazine concentrations in Lake Michigan and Green
Bay in response to mass loadings to those systems
from the time of introduction in 1964 up to 1995. The
calibrated model was then used in a forecast mode
to predict lake-wide atrazine concentrations in Lake
Michigan as a function of various loading scenarios.

4.3.2  LM2-Atrazine  Model Segmentation
and Circulation

Compared  to  MICHTOX  (Level  1  contaminant
transport and fate  model developed for Lake
Michigan)  segmentation  (Figure 1.5.1), the LM2-
Atrazine model has a finer resolution  (Figure 4.1).
Most water column segments  in the LM2-Atrazine
model segmentation schematic share  the same or
portions of the segment boundaries  used in  the
MICHTOX atrazine model. The spatial segmentation
for the LM2-Atrazine model was developed from
digitized bathymetric (5 km x 5 km grid) and shoreline
data for  Lake Michigan provided  by Dr. David
Schwab,   National   Oceanic   and   Atmospheric
Administration (NOAA) (Schwab and Beletsky, 1998).
The lake, including Green Bay, was  divided into 10
horizontal columns, five  water column layers, and
one surficial sediment layer. A detailed spatial and
cross sectional display of the  water segments for
LM2-Atrazine is illustrated in Figure 4.1.  There are
41 segments in total. Segments 1-10 are surface
water  segments   with  an   interface  with   the
atmosphere.  The rest of the  segments lie below
these surface segments.

Water balance is one of the major components in a
traditional water quality modeling framework. Water
movement directly controls the transport of solids and
chemicals in dissolved and particulate phases in a
water system.  In terms  of  LM2-Atrazine model
inputs, the data in the  transport fields  such  as
advective flows and dispersive exchanges, or mixing,
were used  to  describe  the water balance in  the
model. The components and their sources used in
LM2-Atrazine model transport fields are listed below:

1. Bi-direction horizontal advective flows (provided
   by David Schwab, NOAA;  originally based  on
   Schwab and Beletsky (1998).

2. Net vertical advective flows (provided by David
   Schwab, NOAA; originally based on Schwab and
   Beletsky (1998).

3. Tributary flows and bi-directional flows across the
   Straits of Mackinac (Endicott et al., 2005; Quinn,
   1977).
                                              96

-------
                                      21
ID
18
48
l\
/v
CO
>eg
turn


rater
umn
merit
bers

1
11
20
29
36
10
10
10
20
100
average
segment
thickness
in meters
Figure 4.1. Water column segmentation for LM2-
Atrazine.
4.  Water balancing flows.

5.  Vertical dispersion coefficients.

Components such as  precipitation, evaporation, and
groundwater infiltration were not considered in the
water transport fields used in  the LM2-Atrazine
model.
Correct water circulation is essential for the accuracy
of outputs  from the  LM2-Atrazine  model.   The
Princeton   Ocean  Model  (POM)   has  been
demonstrated   to   accurately   simulate   water
movement for a given large water body (Schwab and
Beletsky, 1997; Blumberg and Mellor, 1987).  Using
an extensively  tested version of POM for the Great
Lakes (POMGL), transport fields were generated for
Lake  Michigan at  different  spatial  and temporal
resolutions  for use in  a series of mass balance
models adapted for LMMBP (Schwab and Beletsky,
1998).  The hydrodynamic model for Lake Michigan
had 20 vertical layers and a uniform horizontal grid
size of 5 km x  5 km (Schwab and Beletsky, 1998).
Because the LM2-Atrazine model segmentation was
constructed based on the 5 x 5 km2 grid used in the
POMGL for Lake Michigan, the hydrodynamic model
results were relatively  easily  aggregated to  the
resolution   used  in LM2-Atrazine  (Schwab and
Beletsky, 1998).   The  aggregated  horizontal bi-
direction flows at  each interface  provided a good
approximation of horizontal advective and dispersive
transport  components  at  the  interface.   The
advantage of using bi-directional flows at an interface
was  that it  bypassed  the tedious and necessary
horizontal   dispersion   coefficient  calibration
procedure required when only net flow is available at
the interface.

The vertical transport field was calculated in the form
of net vertical flow [provided  by David Schwab,
NOAA and originally based on Schwab and Beletsky
(1998)].   Therefore, vertical  exchange coefficients
were calculated and calibrated to define the vertical
mixing  process   between   vertically  adjacent
segments.  A summer  period of strong stratification
and a non-stratified period of intense vertical mixing
are important  limnological  features  of the  Great
Lakes (Chapra and Reckhow, 1983;  Thomann and
Mueller, 1987). Therefore, determining the dynamics
of vertical mixing was considered an important model
development task for the LMMBP.

A  thermal   balance  model  was constructed  to
calibrate the vertical exchange coefficients at the
interfaces  (Zhang  et a/.,  1998,   2000).   The
coefficients  were  calibrated using 250  observed
vertical temperature profiles collected at 40 stations
in Lake Michigan  during the 1994-1995 LMMBP
period (Zhang,  2006).
                                              97

-------
Water  balancing  flow  was   another  advective
component added into the water transport field for
LM2-Atrazine.   The  aggregated advective flows
provided by NOAA were not balanced in individual
segments  over  the  two-year  LMMBP   period.
However,  the total  water mass  was  perfectly
balanced on a whole - lake basis. Over the two-year
LMMBP period, some segments  lost or gained a
certain amount of water. This problem could be very
significant for  long-term simulations for the LM2-
Atrazine model because the model simulation stops
once the volume of a segment reaches zero.  To
counter the amount lost or gained  in each segment,
a water balancing flow was introduced to keep the
volume of water unchanged in each segment at any
time during the simulation. The balancing flows were
generated based  on the aggregated advective flows
[provided by David Schwab, NOAA, and originally
based  on  Schwab and  Beletsky (1998)], original
volume of each  segment, and the general water
circulation patterns during the LMMBP period.

Tributary flows and flows through the Straits  of
Mackinac were based on MICHTOX model inputs
(Endicott et a/., 2005), the literature (Quinn, 1977),
and water circulation patterns during  the  LMMBP
period  [provided  by  David Schwab,  NOAA, and
originally based on Schwab and  Beletsky  (1998)].
During a period of approximately 100 days in the
summer, flow and  exchange across the Straits of
Mackinac occurs in two discrete layers formed by the
surface water and  deep, cold, hypolimnetic water.
During this period of stratification, surface layer flow
is from Lake Michigan to Lake Huron, and a deeper
return flow to Lake Michigan is observed. It has been
observed that Lake Superior water discharging from
the St. Marys River travels in a persistent westerly
direction  during  stratification  and constitutes a
significant component of the return flow  to Lake
Michigan (Ayersef a/., 1956; Schelskeefa/., 1976; J.
Saylor, NOAA, personal communication, 1998). The
remainder of this return flow to Lake Michigan is Lake
Huron water.

Hydraulic residence times  (volume/outflow) for the
main lake has been estimated to be 62 years (Quinn,
1992).

After vertical exchange coefficients were calibrated,
a conservative constituent, chloride, was simulated
using the LM2 model configuration to verify that the
water transport components described above were a
good representation of the overall water transport
field for atrazine.  The chloride model was run just
once   without   adjusting   any   parameters  or
coefficients. The model results agreed very well with
the observations during the LMMBP period (Zhang,
2006).

Water column concentration profiles of atrazine at 10
open-lake stations representing four to 10 depths per
station showed no vertical  gradients during lake
stratification for the years 1991-1992 (Schottler and
Eisenreich, 1997) and 1994-1995 (Brent et a/., 2001).
Furthermore, Schottler and Eisenreich  reported that
analysis  of data from their 10 lake  stations that
covered a central north-south axis and  an east-west
axis showed  no  horizontal gradients of  atrazine
concentrations in the lake.

4.4   LM2-Atrazine  Model Application  to
Lake Michigan
4.4.1      Enhanced   Screening
Application
Model
For the LM2-Atrazine model runs, seven years of
atrazine application data were available and used.
Also, due to label changes that lowered application
amounts and established  planting  setbacks  from
water bodies in 1990 and 1992, a decision was made
to use  two tributary load ratios in order to address
atrazine application practices for pre- and post-label
changes. New data from the LMMBP also became
available to modelers.   Wth these additional data,
loading ratios for  both the  tributaries  and  wet
deposition were updated (see Part 2,  Chapters 2 and
3). The model was calibrated by determining a total
decay  that  would yield  a best fit of the  model to
observations in the lake. Also, several forecasting
scenarios were  run with the model.  The efforts of
this additional modeling are described in the following
sections of this part.

4.4.2  Field Data

See Part 1, Chapter 3 for atrazine field data from the
lake, tributaries, and atmospheric components.
                                              98

-------
4.4.3  Tributary Loadings

It was assumed that a Watershed Export Percentage
(WEP)  of 0.6%  derived from  the literature for
fine/moderate textured soils adequately described
the overall WEP of the Lake Michigan watershed.
This WEP, along with historical annual  atrazine
usage in  the United States, was used  to calculate
atrazine loadings from the tributaries to the lake. For
a complete discussion on the WEP method used to
estimate  LM2-Atrazine loadings, please see Part 2,
Chapter 2.

The Stratified Beale Ratio Estimator (SBRE) method
was used to estimate tributary  loads in the 11
monitored tributaries during the  LMMBP utilizing
tributary  flow  and  concentration  data.    Also,
estimates of loads from the unmonitored watersheds
were  made.   However,  loads  were apparently
missed, and therefore,  LM2-Atrazine tributary loads
were based only on WEP,  county-level application
data and  total United States annual  usage records.
See Section 5.3.3.3.1 for a discussion of this topic.

4.4.4 Atmospheric Loadings

Loading  estimates  of wet  deposition  to  Lake
Michigan  and Green Bay were  made for each of the
top surface water segments.  These loadings were
estimated for LM2-Atrazine per the discussion in Part
2, Chapter 3.

4.4.5  Model Assumptions

Model   processes   involving  sediments   and
particulates in the water column were not included in
the LM2-Atrazine model runs  because atrazine is
primarily  in the  dissolved state in surface waters;
therefore, any processes that  involve sediment or
suspended particle interactions  were concluded to be
of minor significance (Section 1.2.2).

A literature review of atrazine degradation processes
in surface freshwater presented in Part 1, Chapter 2
suggests  that degradation is hindered in freshwaters
such as in Lake  Michigan where the water is cold,
has low solids concentrations, low dissolved organic
carbon, a high pH, and low concentration of nitrate
ions.  Degradation of  atrazine is known to  occur
through either biotic or abiotic processes  in  some
environmental compartments. Given the lack of any
Lake Michigan-specific kinetic information on any of
these processes, the approach taken in LM2-Atrazine
was to estimate the loading history of atrazine to the
lake  and then find an overall first-order loss rate
constant to fit the model to observations of atrazine
in the lake water.

Therefore, due to  atrazine's physical and chemical
properties (Part 1, Chapters 2 and 3), processes
modeled  included  only  advection,  dispersion,
volatilization,  absorption,  and reaction  (atrazine
decay).

4.4.6  Model Calibration and Application
(Scenarios)

The calibration of  the model was undertaken using
three  scenarios (1, 2, and 3) that included lower
boundary condition, upper boundary condition, and a
most likely or average boundary condition scenario,
respectively.  All  of these model  runs started on
January 1, 1963 with a zero load.  The model was
calibrated  by finding an appropriate internal decay
until the model  output best matched the observed
atrazine concentration in the lake for samples taken
in  1991, 1992, 1994, and 1995.

The scenarios 4 through 8 are referred to as load
reduction  scenarios.   These  are  not  necessarily
management scenarios, but they can give managers
insight as to which loads are important in the model
and  environment  for  the purpose  of predicting
concentrations of atrazine in the lake.  It is believed
that they  provide  bounds  on the  entire range of
potential loads.  Scenario  3 was used to simulate
conditions from January 1,  1996 through December
31, 2004.   Then  on  January 1,  2005,  the load
reduction scenarios 4 through 8 began and ran for a
period of 50 years.

Scenario  1  - Calibration Based on an Upper
Estimate  of  Boundary  Conditions:    In this
scenario, the initial vapor phase concentration was 0
ng/m3 and increased  linearly until December 31,
1977. Starting on  January 1, 1978,  the vapor phase
concentration was held  constant  at  the  atrazine
detection  limit  of 0.00926  ng/m3 (Miller,  1999)
throughout the remainder of the simulation period.
The summer inflow concentration  at the Straits of
Mackinac  was assumed to be 100% Lake Huron
water.  Lake Huron water was initially set at 0 ng/L
                                              99

-------
and was assumed to rise linearly to 23 ng/L observed
in 1995 (Station 54), and then remained constant for
the remainder of the simulation.  Tributary loading
projections were set equal to an average of loadings
for  1995  and 1998,  but prior  to  that time,  the
historical  loading  estimates  were  used.    Wet
deposition loads beyond  1998 were  set equal to an
average of loads 1978 through 1998.  Wet deposition
loadings before that were based on historical load
estimates.   Volatilization,  absorption,  and  other
processes were  active in the  model.  An internal
decay was then selected for the  model run that
yielded a  best fit to whole-lake volume-weighted
average concentrations.

Scenario  2 - Calibration Based on a  Lower
Estimate  of  Boundary Conditions:    In this
scenario, the vapor phase concentration was initially
set at 0 ng/m3 and remained at that concentration for
the entire  simulation period.  The  summer  inflow
concentration  at  the  Straits of  Mackinac  was
assumed to be 100%  Lake Superior water.  Lake
Superior water was initially set at 0 ng/L, and was
assumed  to rise linearly to 3.5  ng/L  observed in
1994, and then held constant at that  level for the
remainder  of  the  simulation.   Tributary  loading
projections were set equal to an average of loadings
for  1995  and 1998,  but prior  to  that time,  the
historical  loading  estimates  were  used.    Wet
deposition projections beyond 1998 were set equal to
an average of loads for  1978  through 1998.  Wet
deposition  loadings before  that were  based on
historical load estimates. Volatilization, absorption,
and other  processes were active in  the model.  An
internal decay was then selected for the  model run
that yielded a best fit to whole-lake volume-weighted
average concentrations.

Scenario  3 - Calibration Based  on "Average"
Boundary Conditions:  In this scenario, the vapor
phase concentration was initially  set at 0 ng/m3 and
then increased linearly up to 0.00463  ng/m3 (one-half
detection limit) until December 31,1977. Starting on
January 1,1978, this vapor phase concentration was
held constant at 0.00463  ng/m3  throughout  the
remainder of  the simulation period.   The  inflow
concentration  at  the  Straits of  Mackinac  was
assumed to be 50% Lake Superior  and  50% Lake
Huron water.  This mix of water started out at 0 ng/L
and was assumed to linearly rise to 13.25 ng/L or (1/4
x (3.5+23))  ng/L as observed  in  1995 in Lake
Superior and Lake Huron, respectively, and then held
constant at that level  for  the  remainder of  the
simulation.  Tributary loading projections were set
equal to an average of loadings for 1995 and 1998,
but prior to that time the variable loading estimates
were used.  Wet deposition projections beyond 1998
were  set equal to  an average of loads for 1978
through 1998.  Wet deposition  loading before that
were   based  on  historical   load   estimates.
Volatilization, absorption, and other processes were
active  in the model.  An internal decay was then
selected for the model run that yielded a best fit to
whole-lakevolume-weighted average concentrations.

Scenario 4-Virtual Elimination (Lower Bound on
Model Predictions):  In this scenario, tributary and
atmospheric loads were reduced by 100%. For the
projections, vapor  phase concentrations  and  the
Lake Huron/Superior boundary conditions were set to
zero.   All modeling processes  were active.  This
scenario was run using  scenario 3 for predictions
leading up to the date when the virtual elimination
scenario began (January 1, 2005).

Scenario 5 - No Tributary Loads: In this scenario,
the tributary loadings were reduced by 100%. Wet
deposition loads  were set equal to an average of
loads for 1978 through 1998. This scenario was run
using scenario 3 for predictions leading up to the
date when the 100% tributary load reduction scenario
was began (January 1, 2005).  All other modeling
processes were active.

Scenario 6 -  No  Wet Atmospheric Deposition
Loadings:  Tributary loads were  set equal to  an
average of loadings for 1995 and 1998. Atmospheric
wet deposition loadings were decreased by 100%.
This  scenario was  run using  scenario  3  for
predictions  leading up to the date when the 100%
atmospheric load reduction scenario began (January
1, 2005). All other modeling  processes were active.

Scenario 7 - Zero Vapor Phase Concentration:
Tributary loads were set equal  to an  average of
loadings for 1995 and 1998. Wet deposition loads
were  set equal to  an average of loads for 1978
through 1998.  Vapor phase  concentration were set
equal to zero. This scenario  was run using scenario
3 for predictions  leading up to  the date when the
zero vapor phase concentration  scenario began
                                              100

-------
(January 1,  2005).
were active.
All other modeling processes
Scenario 8 - No Further Degradation: A total load
(tributary and wet deposition) was determined using
the model such that no further increase in lake-wide
volume-weighted concentration would be observed
after January 1, 2005.  Up through December 31,
2004, scenario 3 was used.

4.4.7  Model Confirmation

In 2005, atrazine water samples were collected  in
Lake Michigan for  the purposes of confirming the
model  predictions.   However,  as of this  printing,
these analyses were not available.

4.4.8  Discussion of Results

In terms of mass flow rates, LM2-Atrazine results
from scenario 3 are depicted in Figure 4.2 for 1994.
As shown, the highest load to the lake is from the
tributaries followed by the load from the atmosphere
in the form  of wet deposition.  The greatest loss of
atrazine from the system is via export through the
Straits of Mackinac. Loss due to internal decay is the
second  highest  loss  mechanism  in the  lake.
Volatilization and gas absorption are minor processes
in terms of mass flow gain and loss.

Total internal degradation of atrazine in the water (kd)
determined  by  model  calibration was low  in  all
scenarios  evaluated.   These rates of decay  for
scenarios 1, 2, and 3 were 0.012/yr, 0.004/yr, and
0.009/yr, respectively.  For the calibration based on
average boundary conditions (scenario 3),  LM2-
Atrazine predicts that approximately 0.9% of the
atrazine in the lake decays each year due to  some
combination of abiotic and biotic decay in the lake.
Decay can be related to the half-life  of the chemical
in the lake by the following:
                              Half-Life = t1/2 = (In 2)/k,
                                          (4.1)
            absorption
              231 kg/yr
       volatilization
       51 kg/yr

          export via
           Chicago
          Diversion
          145 kg/yr
                           atmospheric
                           wet deposition
                           2493 kg/yr
                                                              input from
                                                             Lake Huron
                                                              472 kg/yr
                                 loss to decay: 1648 kg/yr
                                                          watershed
                                                            loading
                                                          5264 kg/yr
                                                             export to
                                                          Lake Huron
                                                           2531 kg/yr
                                                    Atrazine Inventory
                                                    182,779 kg
                                                 Dry deposition, settling, sediment resuspension
                                                                   and net burial are negligible
Figure 4.2.  LM2-Atrazine model results for Lake Michigan and Green Bay for the year 1994.
                                              101

-------
Assuming that scenario 3 captures typical conditions,
then the 0.9%  internal decay associated with this
scenario represents a half-life of the chemical in the
lake of  77  years.  In scenario 3, the water at the
Straits of Mackinac is assumed to be  half Lake
Superior water and  half  Lake  Huron water.  It is
intuitive that the decay rate associated with scenario
1 that has  the highest boundary concentrations of
atrazine (assumed to be all Lake Huron water) is the
one with the highest decay rate  because  higher
boundary  concentrations  will mean   that more
atrazine is transported into the lake at the Straits of
Mackinac.  Furthermore, the increased vapor phase
concentration in scenario 1 will also contribute  slightly
more to gas absorption  than  the other scenarios.
This cumulative higher mass flow will  result  in a
higher decay needed  in the  modeling calibration
exercise in  order for model output to match  lake
concentration observations. The opposite argument
is true for scenario 2 where the boundary condition at
the Straits of Mackinac is based solely on the lower
concentrations of atrazine from Lake Superior, and
the  vapor  phase  concentration   of  atrazine  is
assumed to be equal to zero throughout the entire
simulation.

In  the forecasts of alternate futures (Figure 4.3),
constant conditions  scenario  3  results in  lake
                                  concentrations   increasing   until  a   value  of
                                  approximately 66 ng/L is attained.  Scenario 3 is
                                  based on  average  boundary conditions, and the
                                  forecasts using this scenario are based on constant
                                  loadings that were observed in the mid to late 1990's.
                                  Eliminating all loadings to the lake (scenario 4)
                                  resulted in the largest decline in model predictions.
                                  A total  loading  reduction of approximately  35%
                                  (scenario 8),  if implemented on  January 1,  2005,
                                  would be  needed  in  order  to  prevent atrazine
                                  concentrations from increasing further than what was
                                  estimated in the lake on January 1, 2005. If only the
                                  atmospheric loadings ceased (scenario 6),  then
                                  concentrations in the lake would not be expected to
                                  change much after January 1, 2005, and the model-
                                  predicted  concentrations  in  the  lake  would  be
                                  expected to be only slightly higher than that predicted
                                  by scenario 8. However, if only the tributary loadings
                                  ceased (scenario 5), then atrazine concentrations in
                                  the  lake  would   decline  relative  to  scenario  3
                                  predictions.     Maintaining   the  vapor   phase
                                  concentration at  0 ng/l (scenario 7) has very little
                                  effect compared to the constant condition scenario 3.
                                  This  is intuitive because scenario 3 vapor phase
                                  concentrations are set to one-half the detection limit
                                  of atrazine in the  vapor phase.
              Ol
              c
              c
              o
              o
              c
              o
              O
              'N
              
                   1963   1993
                                                     2203   2233   2263
Figure 4.3.  LM2-Atrazine model runs of scenarios.
                                                102

-------
Related to the production of ethanol  for  motor
vehicles  in  this country,  the demand  for corn
increased the United States corn acreage planted in
2007 to 93.6  million acres, exceeding the acreage
planted in 2006 by 19.5 %  (U.S.  Department of
Agriculture, 2007). This also represents an increase
of 24.5% of the average acreage planted during the
project period, 1994-1995 (see Figure 4.4 for corn
acreage in the United States from  1986  to 2007).
This was the  largest amount of corn planted in the
United  States since 1944 when farmers planted 95.5
million acres.  It can be assumed that this increase in
corn acreage  has resulted in an increase in the use
of atrazine in the Lake  Michigan watershed.   To
estimate the potential impact on this increased usage
of  atrazine  in  the  Lake Michigan  basin,  both
atmospheric and tributary loadings were increased by
15% and 30% starting in 2007 in scenario 3 (see
Figure  4.5).  For these increases, the lake reaches
steady-state at  approximately 75.2 ng/L and 84.2
ng/L, respectively. At the time of this printing, data on
the actual usage amounts of atrazine applied to the
Lake Michigan basin were not available.  Thus the
range of percent increases for the basin is probably
the best current estimate of the potential impact of
increased loadings to the lake.

In conclusion, the  net volatilization  of atrazine is
negligible in Lake Michigan.  Furthermore,  model
calibration over a hindcast suggests that very little of
the atrazine inventory in the lake decays each year.
The chemical almost behaves as a conservative
substance in the cold, deep waters of Lake Michigan.
If loadings stay that same or increase over what was
observed in the 1990s, then the lake concentration of
atrazine is expected to increase.
            Million Acres

              100
                                    U.S. Corn Acres
                 1986    1989    1992    1995    1998    2001
                                         Planted
   Harvested
2004    2007

    USDA-NASS
      10-12-07
Figure 4.4. Historical trends of United States corn acreage planted and harvested from 1986 to 2007
(U.S. Department of Agriculture, 2007).
                                              103

-------
          90

          80-
ra  70 H


I  6°H
2  50-1
o   40 -
o
O   30

-------
Endicott,  D.D., W.L Richardson,  and D.J.  Kandt.
   2005.  1992 MICHTOX: A Mass Balance and
   Bioaccumulation Model for Toxic Chemicals in
   Lake  Michigan.    In:    R.  Rossmann  (Ed.),
   MICHTOX:     A   Mass   Balance   and
   Bioaccumulation Model for Toxic Chemicals in
   Lake  Michigan,  Part  1.  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.
   EPA/600/R-05/158, 140 pp.

Miller, S.M.  1999.  Spatial and Temporal Variability
   of  Organic  and   Nutrient   Compounds  in
   Atmospheric Media Collected  During the Lake
   Michigan Mass Balance Study.   M.S.  Thesis,
   Department  of  Civil,   Structural,   and
   Environmental  Engineering, State University of
   New York, Buffalo, New York.  181 pp.

O'Connor, D.J.  1983.  Wind Effects on Gas-Liquid
   Transfer  Coefficients.    J.  Environ.  Engin.,
   109(3):731-752.

Quinn, F.H.   1977.  Annual and Seasonal Flow
   Variations  Through  the Straits   of Mackinac.
   Water Resources Res., 13(1):137-144.

Quinn, F.H.  1992.  Hydraulic Residence Times for
   the Laurentian Great  Lakes.  J. Great Lakes
   Res., 18(1):22-28.

Schelske, C.L., E.F. Stoermer,  J.E.  Gannon, and
   M.S.Simmons.  1976.  Biological, Chemical, and
   Physical Relationships in the Straits of Mackinac.
   U.S. Environmental Protection Agency, Office of
   Research  and  Development,  Environmental
   Research  Laboratory-Duluth,   Large   Lakes
   Research  Station,   Grosse   lie,   Michigan.
   EPA/600/3-76/095,  267 pp.

Scholtz, M.T.,  B.J.  Van  Heyst,  and A. Ivanhoff.
   1999.  Documentation for the Gridded  Hourly
   Atrazine  Emissions Data Set  for  the  Lake
   Michigan   Mass  Balance   Study.      U.S.
   Environmental   Protection  Agency,  Office  of
   Research and Development, National Exposure
   Research Laboratory,  Research  Triangle Park,
   North Carolina.  EPA/600/R-99/067, 61 pp.
Schottler, S.P. and S.J. Eisenreich.  1997.  Mass
   Balance Model to  Quantify Atrazine Sources,
   Transformation Rates, and Trends in the Great
   Lakes. Environ. Sci. Technol., 31 (9):2616-2625.

Schwab,  D.J. and D.  Beletsky.  1997.  Modeling
   Thermal  Structure  and  Circulation  in  Lake
   Michigan.  In:  Estuarine and Coastal Modeling,
   pp. 511-522. Proceedings of the 5th International
   Conference  of the American  Society of  Civil
   Engineers, Alexandria, Virginia.  October 22-24,
   1997.

Schwab, D.J. and D. Beletsky.  1998. Lake Michigan
   Mass Balance Study:  Hydrodynamic Modeling
   Project.   National  Oceanic and Atmospheric
   Administration,  Great  Lakes   Environmental
   Research  Laboratory,  Ann Arbor,  Michigan.
   NOAATechnical Memorandum ERLGLERL-108,
   55 pp.

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.

Thomann, R.V. and J.A. Mueller. 1987. Principles of
   Surface Water Quality Modeling and Control.
   HarperCollins Publishers, Inc.,  New York, New
   York.

U.S.  Department of Agriculture. 2001.  Agriculture
   Research Service Pesticide Properties Database.
   Available from U.S.  Department of Agriculture at
   http://www.ars.usda.gov.

U.S.  Department  of Agriculture.   2007.  National
   Agricultural Statistics Service. U.S. Department
   of Agriculture, Washington, D.C.  Available from
   U.S.   Department   of  Agriculture    at
   http://www.nass.usda.gov.

Wanninkhoff, R., J.R. Ledwell,  and J. Crusius. 1991.
   Gas Transfer Velocities on Lakes Measured with
   Sulfur Hexafluoride. In: S.C. Wlhelm and J.S.
   Culliver(Eds.), Air-Water Mass Transfer, pp. 441-
   458.  American Society of Civil Engineers, New
   York, New York.
                                             105

-------
Zhang, X., D. Endicott, and W. Richardson.  1998.
   Transport Calibration Model With Level 2 Model
   Segmentation  Scheme.   First  Lake Michigan
   Mass Balance Project Science Panel Review,
   Southgate, Michigan.  June 23, 1998. 12pp.

Zhang, X., W. Richardson, and K. Rygwelski. 2000.
   Preparation and Verification Transport Field for
   LMMBP  Level 2 Contaminant: Transport  and
   Fate  Models.   Second  Lake Michigan Mass
   Balance   Project  Science   Panel  Review,
   Southgate, Michigan.  September 27, 2000. 15
   pp.
Zhang, X. 2006.  LM-2 Toxic.  In:  R.  Rossmann
   (Ed.),  Results  of  the Lake  Michigan Mass
   Balance  Project:    Polychlorinated  Biphenyls
   Modeling  Report,  pp.  216-452.     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.
   EPA/600/R-04/167, 579 pp.
                                             106

-------
                                        PARTS

                  LAKE MICHIGAN MASS BALANCE PROJECT
                        LEVEL 3 MODEL:  LM3-ATRAZINE
                                      Timothy J. Feist
                                       Xiaomi Zhang
                             Z-Tech, an ICF International Company
                                Large Lakes Research Station
                                           and
                                    Kenneth R. Rygwelski
                                William L. Richardson (Retired)
                                     Russell G. Kreis, Jr.
                         United States Environmental Protection Agency
                             Office of Research and Development
                  National Health and Environmental Effects Research Laboratory
                                Mid-Continent Ecology Division
                      Large Lakes and Rivers Forecasting Research Branch
                                Large Lakes Research Station
                                      9311 Groh Road
                                 Grosse lie, Michigan 48138
5.1 LM3-Atrazine Executive Summary

Most previous Great Lakes water quality models
were developed using large spatial scales.  These
models were capable of predictions on a lake-wide or
regional  scale but were not suitable for evaluating
differences on smaller  spatial  scales,  such as
between off-shore and near-shore concentrations.
The LM3-Atrazine model is a high-resolution model
that is suitable for evaluating fine-scale spatial and
temporal changes in water quality.

The LM3-Atrazine model was based upon the same
framework as the United States Environmental
Protection Agency's (USEPA)  other LM3 models.
The hydrodynamic transport was provided by the
National  Oceanic and Atmospheric Administration's
(NOAA) Great Lakes Princeton Ocean Model (POM).
The water quality framework was the same as used
by the LM3  chloride and  eutrophication  models.
Water quality components  for the  atrazine model
were  developed  at the  USEPA  Large Lakes
Research Station (LLRS) and included a small first-
order decay rate and volatilization. The model spatial
resolution consisted of a 5 km x 5 km horizontal grid
with 19 vertical layers, for a total of 44,042 model
cells. The model was run using a time step of three
hours.

Tributary loads, atmospheric loads, and boundary
conditions for the model were estimated as part of
the Lake Michigan Mass Balance Project (LMMBP).
Model simulations were conducted using  tributary
loads for the 1994-1995 LMMBP period estimated by
                                           107

-------
the United States Geological Survey (USGS) and
alternative loads with a spring seasonal adjustment
based upon long-term historical loading trends.

The LM3-Atrazine model could not be fully calibrated
because only one year of in-lake data and one year
of tributary loading data were collected during the
LMMBP.  However, confidence in model results was
provided  by the favorable  comparison of model
results to available data without adjustment of kinetic
parameters,   by   successful  calibration  of  the
hydrodynamic model, by successful calibration of the
high-resolution model  transport  (in  the form  of a
chloride model),  and  by the long-term  hindcast
calibrations of coarser segmented atrazine models
using the same water quality kinetics.

The   high-resolution   model   was   useful   in
demonstrating the  effects of tributary loadings on
near-shore waterquality. Predicted mid-lake atrazine
concentrations varied  annually less than 5  ng/L
during the two-year simulations.  In comparison, the
model  segment  receiving loads from the largest
tributary, the St. Joseph River, ranged from winter
concentrations of 37 ng/L to spring peaks of 100-350
ng/L depending upon whether 1994-1995 or long-
term tributary loads were used in the simulation.

The  USEPA collected atrazine samples from Lake
Michigan during the 2005 field season; however, the
results were not  available at the time the atrazine
modeling was  conducted.   To estimate  potential
expected concentrations, the LM3-Atrazine model
was  run for the  period 1994-2005.   Loads were
estimated by repeating the 1994-1995 loading time-
series over the 12-year period. The model was run
using both the USGS-estimated loads from 1994-
1995 and loads based upon long-term trends.  Mid-
lake  concentrations were predicted to increase from
the 1994 concentration of 37 ng/L to between 38 ng/L
and 46 ng/L in 2005.

Inflows and  outflows  of atrazine  from the Lake
Michigan system were tracked during the 1994-1995
model simulations.  Outflow through the Straits of
Mackinac  and decay  losses were  approximately
equal and were the largest loss terms.   Tributary
inputs  and atmospheric  wet deposition were  the
largest sources of atrazine. Atmospheric exchange
was  minimal.
Model results and measured data were compared to
toxicological  endpoints  to  examine  possible
ecological effects of atrazine concentrations in Lake
Michigan.     Most  model  forecast  and  data
concentrations  were   below   the   toxicological
endpoints of concern at the spatial scales used in
these modeling analyses.

5.2  LM3-Atrazine Recommendations

Because of its high-resolution (5 km x 5 km model
cells), LM3-Atrazine is useful to determine seasonal
effects of loadings to various cells.  Of  particular
interest may be the effects of high run-off in  the
spring after application to cells at the mouths of major
tributaries.  Within these cells, dramatic changes in
atrazine concentrations may be noted over relatively
short periods  of time.    Some of  the highest
concentrations in the lake would most likely be found
at these sites.    The  lower-resolution models,
MICHTOX  and   LM2-Toxic,  have   coarse
segmentation and would not respond like the high-
resolution model to these spring/early summer high
loading events. In the coarse segmented model, the
load  is instantaneously dispersed uniformly into the
much larger model segment volume receiving  the
river load.  Hence, a concentration spike  would be
low compared to a high-resolution segment receiving
the equivalent load.

5.3    LM3-Atrazine Transport  and  Fate
Modeling

5.3.1  Purpose of High-Resolution Model

Historically,  waterquality models for the Great Lakes
have been developed using large spatial scales. The
first eutrophication model for Lake Ontario (Thomann
and  Di  Toro, 1975; Thomann et a/.,  1979) was
configured   with  only  two  vertical  segments
(epilimnion and hypolimnion).  Similar scale models
were also developed for  Lake Erie (Di Toro and
Connolly, 1980), Lake Huron (Di Toro and Matystik,
1980),  and  Lake Michigan (Rodgers and Salisbury,
1981).  Even a more recent model of Green Bay was
developed on a relatively coarse-grid scale (DePinto
et a/.,  1993).   These models  were  capable  of
adequately  simulating  average water quality over
large  spatial  segments   and  projecting  future
concentrations.  However, they were not capable of
                                              108

-------
simulating spatial concentration gradients very well,
if at all. Also, there have been questions on whether
limnological   processes   could   adequately  be
represented on such a large spatial scale, particularly
sediment transport. During the design phase of the
LMMBP,  modelers were determined to  construct a
higher-resolution  model  to  overcome  these
deficiencies.

The LM3  level models include linked high-resolution
hydrodynamic and water quality components.  The
hydrodynamic  component  of the  models  was
developed by modelers at the NOAA's Great Lakes
Environmental Research Laboratory (GLERL).  The
transport framework was based on the U.S. Army
Corps of Engineers' CE-QUAL-ICM model  (Cerco
and Cole, 1994). The water quality components were
developed at the USEPA's LLRS. Completed water
quality components included a eutrophication model
(Pauer et a/., 2006), the atrazine model described
below, and,  subsequent to  the  LMMBP,  an
ecosystem model (Miller et a/., 2007).

Although  the LM3 level models have many scientific
and  technological  advantages,  there  are major
challenges. First, the LM3 models required a much
greater degree of computer resources to develop and
operate.   Second,  they required  more computer
programming support  to develop completely  new
programs.  Third, because there are over  40,000
water segments for which concentrations are being
simulated, there is much more computer  output to
manage  and evaluate.  This  has  presented  disk
storage issues and has required additional effort to
develop computer programs to analyze and  display
model output.

The following sections describe  the  LM3-Atrazine
model, the assumptions  used in  developing  the
model, the loading data and lake concentrations used
for model confirmation, and the results  of model
simulations for the 1994-1995 LMMBP period and for
forecasts.

5.3.2  Model Description  and Framework

LM3-Atrazine, as with most mass balance models,
incorporates  segment  geometry,  advective  and
dispersive transport,  boundary concentrations for
state variables, point and diffuse source loads, kinetic
parameters, constants and time functions, and initial
conditions.  These  input data, together  with  the
general  mass balance equations and the specific
chemical kinetics equations, uniquely define a special
set of water quality equations.  These equations are
numerically integrated as the simulation proceeds in
time. At user-specified print  intervals,  values of
selected state variables are saved for subsequent
evaluation  in  visualization  and  statistical  post-
processor programs.

In the Great Lakes environment,  atrazine has the
chemical  properties  of  a   mostly  conservative
substance.   The important functions of the LM3-
Atrazine model  consist of hydrodynamic transport,
external loads, atmospheric exchange, and a small
first-order decay rate.

This   section   contains   a  description   of   the
hydrodynamic model, the  kinetic processes of the
atrazine  model, and  the spatial  and  temporal
configuration of the atrazine model.

5.3.2.1   POM Hydrodynamic Model

The basis of the LM3 water quality model is water
movement and  material transport.  Hydrodynamic
simulations were conducted by Schwab and Beletsky
(1998)  who  applied the  POM.   Portions of  the
following section are  excerpted  from their report.
Subsequent to the preparation of the report, Schwab
included annual average tributary flows and average
Straits of Mackinac outflow in the final submission of
model results to USEPA for use  in mass balance
models.  In  addition, computational modifications
were made that eliminated a minor problem with
water balance [for a technical discussion of the
details, see Appendix A in Melendez et a/. (2008)].
The primary goal was to provide three-dimensional
fields  of currents, temperature,  and  wind-wave
characteristics for the study period (1994-1995) for
direct input to the LM3 water quality model.  The
model was applied to Lake Michigan using a 5 km x
5  km  grid  (Figure 5.1).    The  output  of  POM
simulations  was provided to  the water  quality
modeling team at the USEPA/LLRS  for  further
translation  for  the  water  quality  models.
                                             109

-------
Figure 5.1. Lake Michigan hydrodynamic model
5 km x 5 km computational grid.
During the development of the  POM for Lake
Michigan, the model was applied  for two  periods:
1982-1983  and 1994-1995.  The  first period was
chosen for model calibration because of an extensive
set  of  observational  data  including  surface
temperature observations at two National Data Buoy
Center (NDBC) weather buoys and current and
temperature observations  during  June 1982-July
1983 at several depths from 15 subsurface moorings.

Results were output to files  containing values for
each of the 5 km  x 5 km cells at specified time
intervals.  To compare model simulations with data,
model  results  were averaged  over  various time
periods depending on the data period. For example,
the simulated temperature time-series for the 1982-
1983 period are shown  in Figure 5.2 and for the
1994-1995   period  in  Figures  5.3a  and 5.3b.
Statistics  of  temperature   field  validation  are
presented in Table 5.1 for 1982-1983 and Table 5.2
for 1994-1995.   RMSD  is the root  mean square
difference (error) between observed and computed
temperatures.   Maximum  Error  is the maximum
temperature difference.  Average is  the arithmetic
mean.    The  correlation  coefficient provides  a
statistical indication of the  strength of the linear
relationship  between  computed  and   observed
variables.

The model  was able to reproduce all of the basic
features of  the thermal structure of Lake Michigan
during the 600 day period of study:  spring thermal
bar, full stratification, deepening of the thermocline
during the fall cooling, and the overturn in the late fall
(Figure 5.4).

Another model validation was made  by  comparing
observed temperature profiles acquired  during the
seven  Great   Lakes  National  Program  Office
(GLNPO) water quality surveys during 1994-1995 to
simulated  temperature  profiles  at  20 locations.
Figure 5.5 depicts  one of these locations, Station
18M. In addition, the USGS conducted several near-
shore transect surveys and compared simulated and
observed temperatures.

Schwab and Beletsky (1998)  provided additional
information  on model development and validation.
The basic conclusion was that, overall, the models
simulated   the  large  scale  thermal  structure,
circulation, and waves quite realistically on the 5 km
x 5  km  grid.   There were some  qualifications,
however. First, lack of an ice model will be a serious
problem if the model is applied during a year with
normal or severe ice conditions.  It will cause both
significant violations of the lake's heat balance  and
errors in calculating transfer of momentum  from air-
to-water  because  of  the  difference  in   surface
roughness  of  ice  and  water  and  momentum
absorption  by  the  ice. The  1994-1995 POM
simulation  assumed a  constant uniform  water
temperature of 2°C for  the period  January 1 to
March 31, 1994.  Because  no hydrodynamic data
                                              110

-------
s
30
g 25
ffi 20
2 15
Q)
E" 1°
0
1
30
0 25
S 20
TO 15
a. 10
h- 5
0
1
30
o 25
£ 20
3
a 15
0)
E 10
,2 5
0
1
30
0 25
2 20
2 15
0)
g- 10
K 5
0
1
urface Temperature at 45007 Surface Temperature at 45002
1 	 • 	 ' 	 ' 	 ' 	 ' 	 30
il l, o" 25
ii* y» " ^ 20
|"% f V 2 15
v1,. ....... 	 , 	 , 	 , 	 	 	 	 0
I
: P' f '{
00 200 300 400 500 600 100 200 300 400 500 600
Temperature at 1 5 m Temperature at 1 5 m
.,,,„„„,,„.,„,.„„.,„„,„,,,„„,„.,„„.,„.,„,„! jQ
^ 0 25
r - £ 20
L (. ; 1 JL , "S 1 5
r^_., 	 ,...^ 	 , 	 n
> 	 	 ' 	 l
: 4
i
'• \
00 200 300 400 500 600 100 200 300 400 500 600
Temperature at 50 m Temperature at 50 m
pm,mrm, 	 	 	 	 ^Q
f - o 25
i- - 2 20
3
r - S 15
'• CD
1 0. 10
r~*^ 0
4
•
". i
• 1
i j
• 1
00 200 300 400 500 600 100 200 300 400 500 600
Temperature at 148 m Temperature at 154 m
, ,„ „. „,-,„,-„„.„.„,,, -„,-,„ 	 ^ jQ
§ 25
£ 20
3
2 15
(D
Q. -j n
EE
_^_____,%-^ , 	 ^ H 5
br^ 	 	 	 , 	 ., 	 , 	 o
1
; 1
^^_x___ ,,^,^-
00 200 300 400 500 600 100 200 300 400 500 600
JULIAN DAYS, 1982-83 JULIAN DAYS, 1982-83
Figure 5.2. Simulated temperature (black) compared to measured temperature (gray) at two buoys in
Lake Michigan for 1982-1983 (Schwab and Beletsky, 1998).
                                          111

-------
                           Surface Temperature at 45007
                       o
                       a?
                       3
                       as
30
25
20
15
10
 5
 0
                        e
                        2
                        £0
                        I
30
25
20
15
10
 5
 0
                           100  200  300 400  500 600
                                Temperature at 12m
                           100  200  300  400  500 600
                                Temperature at 42 m
                           100  200  300 400  500 600
                       3
30
25
20
15
10
                               Temperature at 102 m
                           5
                           0
                           100  200  300 400  500 600
                                JULIAN DAYS, 1994-95
Figure 5.3a. Time-series of simulated watertemperature versus observed at 45007 for 1994-1995. Gray
line is observation; black line is model simulation (Schwab and Beletsky, 1998).
                                         112

-------
     o
     5
     CO
     ***
     


-------
                  Simulated Average Temperature of L. Michigan
          100
              100      200      300      400      500      600

                                 JULIAN DAYS,  1982-83


Figure 5.4. Simulated mean temperature (°C) profile for 1982-1983 (Schwab and Beletsky, 1998).
        Cruise 1   Day 126  Year 94
      50-
    £ 100-
    CL
    0)
    Q
     150-
     2001	
    Cruise 3  Day 234  Year 94
    Cruise 4 Day 303  Year 94
                                 SO
£100
 -
fa
a
                                 150
                                 2001	
                                                             SO
£100
0)
Q
       0   5   10  15  20  25  30     OS  10   15  20   25  30
                                                            150
                                                            20QI.... I.
              Temperature (C)
          Temperature (C)


    Cruise 6  Day 225  Year 95
                                 50



                               £
                               £ 100
                                 150
                                 200
   0   5   10  15  20  25  30
          Temperature (C)


    Cruise 7 Day 281  Year 95
                             50






                           £100
                           Q.

                           Q



                             150
                                   0   5  10  15  20  25  30      0   5   10  15   20  25   30
                                         Temperature (C)                  Temperature (C)
Figure 5.5.  Temporal evolution of simulated versus observed temperature profiles, Station 18M
(Schwab and Beletsky, 1998). Black line is model simulation; gray line is observation.
                                           114

-------
were available after December 21, 1995, the LM3-
Atrazine model used the corresponding 1994 data for
the last 10 days in 1995.

Second, the model did not perform as well in the
thermocline area as it did near the surface.  The
simulated thermocline was too diffuse. Although this
problem might be overcome by development of a
higher-resolution model, this problem is probably not
significant for the mass balance study in comparison
to other uncertainties with data and chemical and
biological processes.

Lastly, while the MlCHTOXand LM2-Atrazine models
have  bidirectional  flow  through  the Straits  of
Mackinac, the present configuration of LM3-Atrazine
only uses a net, annual average outflow. In reality,
there is a return flow to Lake Michigan at the Straits
during stratification for a period of approximately 100
days in the summer.  However, to  include this
process within POM would have required significant
additional  resources  including  the running of a
simultaneous Lake Huron hydrodynamic model. The
absence of bi-directional flow at the Straits was not
expected to have a significant impact on circulation
predictions in the main portion of the lake.

5.3.2.2  Model Framework

The LM3-Atrazine model uses the same computer
code and spatial resolution  as other LM3 models
(LM3-Eutro and LM3-Eco). Detailed documentation
of the LM3 models has been provided by Melendez
etal. (2008). The documentation provides a history
of the  models' development  and  a  complete
description of the model framework, equations, and
use.  Documentation of the LM3-Eutro application is
included in Pauerefa/. (2006). Version 3.2.15 of the
LM3 model code was used for the  LM3-Atrazine
model analyses.

The transport  model  incorporated within the  LM3
framework was based  on the ULTIMATE QUICKEST
transport scheme, originally developed by Leonard
(1991)  and subsequently augmented for use with
variable grid sizes by Chapman et a/. (1997).  The
transport algorithm was  coded  in  Fortran  and
previously applied to  the Chesapeake Bay model
(CE-QUAL-ICM) (Cerco and Cole, 1994,1995). The
transport model calculation performed numerical
integration of spatially varying concentrations using
quadratic interpolations of the concentration to infer
its value at flow faces and analytic integration over
space- and time-variables to account for changes in
the concentration at the cell wall during the course of
the time step.   Further details  of the dimensional
derivation  of the ULTIMATE  QUICKEST transport
method can be found in Melendez et a/. (2008).

Because  atrazine   is  relatively  stable  in  Lake
Michigan,  only  a subset  of the  model's  kinetic
processes were used:  hydrodynamic  transport,
atmospheric exchange, and  degradation.   These
processes, and  the spatial and temporal resolution
used in the simulations, are described below.

5.3.2.2.1 Water Quality Processes

The LM3 models are mass balance models based on
the principle of conservation of mass. They use the
same finite segment modeling approach used in the
USEPA-supported WASP4  and the CE-QUAL-ICM
modeling framework. The models  describe where
and how a mass of constituent is transported and
transformed.  The mass of a chemical or solid in
each water segment is controlled by water movement
between adjacent segments, solids and chemical
dynamics within the system,  internal and external
loads, and boundary concentrations.

For  LM3-Atrazine,  external  loads,  hydrodynamic
outflow, and chemical transformation are the most
significant   processes   affecting  atrazine
concentrations  in Lake  Michigan.    Atmospheric
exchange  (volatilization  and absorption  at  the
water/air interface) was also included  in  the model
kinetic  process,  although  the  mass involved  is
considerably smaller than that involved with outflow
or chemical transformation.   Atrazine  does  not
partition onto solids.  Thus the settling and sediment
interaction portions of the LM3 water quality model
were not utilized.

Mass balance equations representing the  above
processes were used in the model to  compute the
change of mass of atrazine in each segment at a
certain  time.   A  general  time-dependent  finite
differential equation in  a given segment  can be
written to describe the change of mass for a  state-
variable at a certain time.  The change in mass of
atrazine in the LM3-Atrazine model for a given water
column segment is described as:
                                             115

-------
where
n

R
                                        (5.3.1)
Vj   =  volume of segment] (L3)

Cj   =  concentration of water quality constituent in
        segment] (M/L3)

C,   =  concentration of water quality constituent in
        segment I (M/L3)

C,j   =  concentration of water quality constituent at
        the interface between segment I  and  j
        (M/L3)

QIJ   =  net flow  across  the interface  between
        segment I and j (defined as positive when
        entering  segment j  and negative when
        leaving segment]) (L3/T)
     =  number of adjacent segments
                  ,   bulk   dispersion/diffusion
        coefficient (L3/T)

     =  mixing  (dispersion/diffusion)  coefficient
     =  interfacial area between segment I and j
        (L2)

     =  characteristic   mixing   length   between
        segments I and j (L)

     =  external loading rate of segment] (M/T)

     =  mass  change  rate  due  to  air-water
        exchange process between segment] and
        air directly above segment] (M/T)

     =  mass change  rate due to sum  of kinetic
        transformation processes within segment]
        (M/T), positive is source, negative is sink
Note:   L = length; M = mass; T = time.
                                                 The mass change due to  kinetic transformation
                                                 processes, SkJ, is represented in the atrazine model
                                                 by  a single  first-order  decay  rate.   The decay
                                                 coefficient  was  determined  during the  long-term
                                                 hindcast simulations using the MICHTOX and LM2-
                                                 Atrazine models (Parts 3 and 4 of this report) and
                                                 was set at 0.009 year1 (2.854 x 10'10 s'1).

                                                 Gas exchange  (volatilization and absorption)  of
                                                 atrazine between the lake and the atmosphere is a
                                                 potential source or loss of atrazine to Lake Michigan.
                                                 Computing the atrazine mass transfer across the
                                                 water-air interface was necessary to satisfy the
                                                 overall atrazine inventory and mass budget in the
                                                 Lake Michigan system for the LMMBP period.  The
                                                 mass change rate term (SawJ) for atrazine due to air-
                                                 water  exchange  processes was  calculated  in
                                                 Equation 5.3.2 as a product of the overall net mass
                                                 exchange  flux and  surface area  of  the water
                                                 segment].

                                                                                          (5.3.2)
where

kol   =  the overall mass exchange rate coefficient
        (L/T)

CHail =  dissolved atrazine concentration in water
                                                  'dwj
                                                          (M/L3)
                                                  CaJ  =   atmospheric atrazine concentration  over
                                                          segment] (M/L3)

                                                  H'   =   temperature-dependent   Henry's   law
                                                          constant (dimensionless)

                                                  Aj   =   surface area of the water segment j (L2)

                                                  The overall mass exchange rate coefficient (/c0/) was
                                                  calculated  using  the  Whitman  two-film  theory
                                                  formulation (Whitman, 1923) given as:
                                                                                          (5.3.3)
                                                         kt    kg*Hl
                                              116

-------
where

k,    =  the liquid film mass transfer rate coefficient
        (L/T)

kg   =  the gas film mass transfer rate coefficient
        (L/T)

The  parameters H',  k, and kg were  calculated  at
every time  step  for  each LM3  segment.   The
Wanninkhoff  (1992)  formulation for  water mass
transfer  resistance  and  the   Schwarzenbach
(Schwarzenbach et a/.,  1993)  formulation  for gas
mass transfer resistance were used for modeling the
air-water exchange of atrazine in Lake Michigan.
The Wanninkhoff equation for k,, with  correction for
atrazine molecular diffusivity in  reference to carbon
dioxide (CO2)  molecular diffusivity  across  the air-
water interface, is given as:
                                        (5.3.4)
where
u
 10
       =  chemical molecular diffusivity  in water
       =  CO2 molecular diffusivity in water (L2/T)

       =  wind velocity measured at 10 m above
          water surface (L/T)
The Schwarzenbach formulation for kg with correction
of atrazine molecular diffusivity in reference to water
vapor  molecular  diffusivity  across  the air-water
interface is given as:
                                        (5.3.5)

where

Da     =  chemical molecular diffusivity in air (L2/T)

Dg_H20  =  water vapor molecular diffusivity in gas
          phase (L2/T)

The  atrazine  model  calculated  a  temperature-
corrected dimensionless Henry's law coefficient using
                                                  equations derived from Scholtz  et a/.  (1999) and
                                                  Miller (1999).
                                                                                 2.3Q3FI
                                                 where
                                                                                          (5.3.6)
                                                  H'    =  temperature-dependent  Henry's   law
                                                           constant (dimensionless)

                                                  HTref   =  Henry's  law  constant  at the reference
                                                           temperature

                                                  AHH   =  the enthalpy of phase change (kJ/mol)

                                                  R     =  the  ideal gas  constant, 8.315  x  10"3
                                                           kJ/(mol)(°K)

                                                  T     =  interfacial temperature  (°K)
                                                  ' ref
                                                        =  reference temperature of 298.16 K (25°
                                                           C)
The value for the dimensionless Henry's law constant
at 25°C was set to 8.1 x 10'8 (U.S. Department of
Agriculture, 2001). The enthalpy of phase change
was  set to 50 kJ/mol (Scholtz et a/., 1999; Miller,
1999).

5.3.2.2.2 Spatial Resolution

Developing the  high-resolution grid  for the LM3
models  required  compromises between  different
spatial configurations and the difficulties in translating
the 5 km x 5 km grid hydrodynamic output.  The best
approach was to develop the fine-grid model at  the
same 5 km scale as the POM (Figure 5.1). The high-
resolution  LM3  grid   consisted   of  2,318
horizontal segments with 19 vertical "sigma" layers,
resulting in a total of 44,042 water column cells.

A  linkage between POM and the LM3 model was
developed by Chapman et al. (1997). The linkage
mapped   POM  cell   numbers  with  ULTIMATE
QUICKEST flow face numbers and  the relationship
between horizontal and vertical components.  LM3-
Atrazine  inputs  included  POM output for water
temperature, horizontal and vertical  dispersion, and
horizontal and vertical currents for each segment in
the water column.
                                              117

-------
5.3.2.2.3 Temporal Resolution

The LM3-Atrazine model simulated the period from
January 1, 1994 through December 31, 1995 for the
LM M BP study period. To forecast the possible range
of atrazine concentrations expected in Lake Michigan
during the 2005 sampling surveys, the model was
also run for the 12-year period January 1,  1994-
December31, 2005.

The LM3-Atrazine model was run using a variable
time  step  based upon model stability.  Over the
course of  the  1994-1995 simulation,  the  average
value of the time step was approximately three hours.
Output from the POM hydrodynamic model was
averaged over three-hour intervals for input  to the
LM3-Atrazine model.  LM3-Atrazine model results
were output at a daily interval for two-year  model
runs.  Atrazine almost behaves  as a  conservative
constituent in Lake Michigan (has an extremely slow
chemical  transformation),   and  daily  behavior
provided sufficient resolution for interpretation of
simulation results. Results from some of the 12-year
forecast model  runs  were  output at a six-day
frequency to maintain reasonable output file sizes for
long-term output animations.

5.3.2.2.4 Model Assumptions

The conceptualization of processes  in the LM3-
Atrazine model was based upon literature review
(Part 1, Chapter 2) and  previous LLRS  atrazine
modeling efforts (Part 3 and  Part 4).  Atrazine
essentially behaves as a conservative  substance in
Lake   Michigan.     Previous  LLRS   modeling
demonstrated that external loading and outflow from
the Straits  of Mackinac were the most important
processes effecting  atrazine concentrations  in the
lake  (Part  3;  Part  4; Rygwelski et a/.,  1999).
Although  it occurs  slowly,  degradation of  the
chemical is also an important process because of the
slow rate of export.  Exchange between the water
surface  and atmosphere was modeled, although it
only had a small effect on lake concentrations.

Atmospheric loads were assumed to  be primarily
through wet deposition.   Dry deposition was not
found to be significant based upon LMMBP sampling
reports (Brent et a/., 2001; Miller, 1999).  Sections
1.3.2.2.2 and 1.3.2.2.3 of this report summarize the
atmospheric deposition sampling. Later papers have
suggested that dry  deposition may  be significant
(Miller  et a/., 2000)  but only provided a range of
possible loads and no spatial or temporal resolution
consistent with the  LM3 models.  The  range of
possible  dry  deposition loads  was  taken  into
consideration when  estimating  loads for long-term
forecasts.

Atrazine is primarily  found in the dissolved state in
Lake  Michigan,  and  sediment  interactions with
atrazine are  minor (Part 1, Chapter 2; Rygwelski et
a/., 1999).  Sediment processes were assumed to be
negligible and were not included in the LM3-Atrazine
model kinetics.

5.3.3  Description of Data Used

The data used for the LM3-Atrazine  modeling was
collected during the 1994-1995 LMMBP studies. The
data were  reviewed in  Brent et al.  (2001) and
summarized in Part  1, Chapter 3 and Part 2  of this
report.

5.3.3.1 Field Data

Model  simulation  results were compared to data
collected during the LMMBP field surveys.  Lake
water samples were collected  for atrazine analysis
during six cruises from April 1994 through April 1995.
Data  from   mid-lake stations  were  selected for
comparison purposes because these stations were
sampled during most cruises. While data were also
collected from near-shore stations,  these stations
were not routinely sampled.

5.3.3.2 Initial and Boundary Conditions

The   Lake   Michigan  atrazine  model   initial
concentrations were estimated  based upon  the
LMMBP field survey data. A uniform concentration of
37 ng/L was set for all main lake and northern Green
Bay model cell initial  concentrations based upon the
average concentrations measured during the  spring
1994 sampling cruise.  Southern Green Bay cells
nearest to the Fox  River were assigned an initial
concentration of 50 ng/L based  upon the limited
Green  Bay sampling from the  fall 1994 and  spring
1995 cruises.

While the MICHTOX and LM2-Atrazine models have
bidirectional flow through the  Straits of Mackinac, the
                                              118

-------
present configuration of LM3-Atrazine only uses a net
outflow. For this assumed configuration, Lake Huron
boundary  conditions  are  not  necessary because
there is no flow to Lake Michigan at the Straits of
Mackinac.

Atmospheric atrazine vapor samples were collected
from   March   1994   through   October  1995.
Atmospheric sampling did not detect vapor phase
atrazine concentrations in 86% of the samples (Brent
et  a/.,  2001).    For  modeling  purposes,   the
atmospheric concentration for all locations and times
was set to a single value  equal  to one-half  the
average  method  detection  limit  (MDL)  of  the
samples, 4.63 pg/m3 (Miller 1999).

5.3.3.3 Loadings

5.3.3.3.1 Tributary

Watershed atrazine loadings to Lake Michigan were
estimated by Hall and Robertson (1998). Loads were
calculated for 11 tributaries  that were sampled as
part of the LMMBP  field  program  and for  18
unmonitored watersheds  (Figure 5.6).    For  the
monitored tributaries, event and base flow samples
were collected  from  April 1995 through  October
1995.  The Stratified Beale Ratio Estimator (SBRE)
was used to calculate  loads for 1995 with these
sample  data and the USGS flow data.   Because
tributary samples were not collected in 1994,  loads
for  1994 were estimated  using  USGS regression
methods and the 1995 data (Hall  and Robertson,
1998).  Loads for the unmonitored watersheds were
estimated  using load to watershed area ratios from
monitored watersheds with  similar soils and land
uses. Part 2, Chapter 2 discusses the tributary loads
in more depth.

The USGS estimated  loads for 1994-1995  were
substantially smaller than what  would  have  been
expected based upon long-term loading patterns.
Rygwelski et a/. (1999) reviewed previous studies
and found that, for soils similar to those in the corn-
producing watersheds of Lake Michigan, 0.6% of the
atrazine applied  to  the watershed reached  Lake
Michigan.  This amount is also  referred to as  the
Watershed Export Percentage (WEP). Rygwelski et
a/. (1999) also conducted long-term hindcast atrazine
modeling that confirmed the  appropriateness of the
0.6% WEP (see Parts  3 and 4).  The 1994-1995
   Lake Michigan
    watersheds
       monitored
       tributary
       basins
       unmonitored
       tributary
       basins
       sampling
       locations
Figure 5.6.  Watershed and mid-lake sampling
stations for the LMMBP study.
USGS estimated tributary loads were only 30% of the
load estimated using the long-term WEP and  1994-
1995 atrazine application data for the Lake Michigan
watershed.   Using the USGS-estimated  loads for
1995, a WEP of 0.12% was calculated.  Using this
WEP derived from the USGS load and no atrazine
decay in a MICHTOX hindcast, the model-predicted
less than one-half of the measured concentration in
the lake as observed in the mid-1990s.

The 1994-1995 loads may have been substantially
lower because of a number of possible factors:  loads
were lower than  normal due to the WEP possibly
decreasing   over  time   because   of   improved
agricultural management practices, significant peaks
in tributary loads may have been missed because the
weekly storm event sampling was discontinued too
early, or atmospheric dry deposition may be higher
than expected.  It is  known that dry years can
depress atrazine  watershed  loadings.  However,
                                             119

-------
precipitation to the lake was near long-term averages
(see Sections 1.3.2.2.3, 1.4.5.1, 1.4.5.2, and 2.2.1.2
for  information  on rainfall and impact on WEPs).
Other potential meteorological forcing functions were
also near average conditions during 1994-1995 (see
Part 1, Chapter 4).  Therefore, weather conditions
are an  unlikely  cause of  the low  USGS  export
estimates.

The USGS loadings were based on an average of 14
atrazine  samples per  year per tributary  (range:
seven to  20).  When  compared  to  other  similar
atrazine load estimation  studies (Schottler  et  al.,
1994;   Richards et a/.,  1996; Capel and Larson,
2001) this represents a very low number of samples
collected  and  thus could  have  contributed   to
underestimation  of  loads.  A study by Leu et al.
(2004) found that a single run-off event that occurred
on day 23 after application of atrazine exported 70%
of the total cumulative load measured  during  a one-
to 67-day period after application.  On a fine-loamy
field in Ohio, a rainstorm occurred just two days after
atrazine application to a no-till  field.  That rain event
accounted for only 3% of the yearly rainfall and 6% of
the yearly run-off; yet it produced 78% of the yearly
atrazine loss (Shipitalo and Owens,  2003).  The
Shipitalo and  Owens' study also concluded that  the
timing of  rainfall  and run-off relative to atrazine
application can have a much greater effect on yearly
losses than agronomic management  practices  (till
versus no-till).  So, a lack of adequate sampling
during an event shortly  after atrazine application
could cause significant underestimation of the total
annual loading from  a watershed. Also, Schottler et
al. (1994) and Williams etal. (1995) have noted that
the spring atrazine concentration often peaks in
streams  just  before  the maximum  flows   are
achieved.   One possible theory suggests  that a
fraction of atrazine on the soil  immediately following
application is readily available for transport by  run-off
during a precipitation event.  However, later in  the
season,  the peak concentration may actually lag  the
peak flow suggesting that export from the fields is
associated with water that has infiltrated the soil and
carried  via  shallow saturated  zones or surface
drainage tile networks to receiving tributaries. High
frequency sampling  just before, during, and after a
flow event are  important in order to fully capture
atrazine loading events.
To evaluate the possible range of loads occurring in
the Lake Michigan system, additional model runs
were conducted with annual loads set equal to those
expected based upon the long-term WEP of 0.6%.
The USGS loads were adjusted by multiplying loads
from each tributary during a 90-day period from April
15 to July 13 by a factor that resulted in the loads for
that tributary being equal to  the expected WEP-
based load.  Only the spring period was multiplied
since this is the period when the majority of atrazine
loads enter the lake and this is the period when
tributary and atmospheric loads have the  largest
uncertainty.    Load   multiplication  factors  were
calculated as the multiplier for the  specified time
period loads that set the total 1994 and 1995 USGS-
calculated loads  for each  tributary equal  to the
combined WEP loads for both years for that tributary.
Computer code in the LM3-Atrazine model conducted
the multiplication  during the  model  simulation by
reading inputs for  the scaling time periods for each
year and the multiplication factors for each tributary.
WEP-based loads and USGS-estimated loads are
listed in Section 2.2.5.  Figure 5.7 displays the USGS
and WEP-based loading time-series for the three
largest tributary  loadings  of  atrazine  to  Lake
Michigan.

A loading series for  the  12-year model runs was
developed by repeating the loads for the 1994-1995
period six times, using USGS-calculated  or WEP-
adjusted loads as appropriate.  The LM3-Atrazine
model did this automatically by looping over the two-
year loading input deck  and  applying  the load
multiplier factors as needed.

In Figures 5.7-5.10 and for the remainder of this part,
the "long-term WEP  loads" in the legends refer to
USGS loads that were adjusted as described in the
preceding paragraphs and "USGS estimated loads"
refer to the loads as received from the USGS.

5.3.3.3.2 Atmospheric

Atmospheric deposition samples were collected from
March 1994 through October 1995. Wet deposition
was the dominant  atmospheric source of atrazine to
Lake Michigan. The monthly average 1994 and 1995
wet  deposition  loading  time-series  data  were
provided by Hornbuckle (University of Iowa, personal
communication, 2002;  Miller  et al.,  2000).  Dry
deposition was not included in the atmospheric loads.
                                              120

-------
O
0

-------

40-
I 30-
«



0 -
Station 27M
A JL A




01 ^
30-


	 USGS estimated loads
A field data (0-1 Oml




Station 47M

Z
A *














Jan July Jan July Jan Jan-1994 JuIy-1994 Jan July Jan
1 994 1 994 1 995 1 995 1 996 1 995 1 995 1 996
40-

1» 30-
g
c
g 20-

Station 18M







40-

0130-
i
5 20-

A Stations 40M/41

A











Jan July Jan July Jan Jan July Jan July Jan
1994 1994 1995 1995 1996 1994 1994 1995 1995 1996
Figure 5.8. Comparison of field data to predicted mid-lake surface concentrations for the 1994-1995
model simulation and two loading conditions. Station locations are shown in Figure 5.6.
May 29. 1995
600 •
1 Manistique ^_V ~~-\
42

41

40
39



w r
Menominee ^HP? /s \
^Br/ f^f^m
Oconto^BW/ f \j/
Fox* /
Manitowac ;, ji /
ff M^Pere Marquette
Sheboygan«f /
/ K^Muskegon
Milwaukee *« W Grand
Root <• m '
/f Kalamazoo
                      38
                                 Calumet*
                                                   St. Joseph
                      37
                     Atrazine
                      (ng/L)
Figure 5.9. Model simulation results of surface concentrations for May 29,1995 using long-term WEP-
based loads.  Selected tributary input locations are labeled.
                                             122

-------



-250-
c
liso-
"100-
50-
0-


• long term WEP loads
""""""i USGS estimated loads

.1
segment181 near J||
St. Joseph River mouth |lj
SCr 1 ft 	
o* Mk





F




                                                     Fox'  i
                                                     Rirer /
              Jan
              1994
July
1994
Jan
1895
July
1995
Jan
1998
                                                             SI. Joseph River
Figure 5.10. Comparison of near-shore surface cell model results for the 1994-1995 model simulation
and two loading conditions.
While the model is useful for demonstrating near-
shore impacts, it does not simulate concentrations in
a river plume  entering the lake or in the tributary
itself. During model simulations, at each time step,
any tributary load is immediately mixed throughout
the 5 km by 5 km model cell near the tributary mouth,
and thus predicted near-shore concentrations are a
function  of the size of the model cells and  not
representative of concentrations in river plumes in
the lake.  Furthermore, the LM3-Atrazine model was
not  designed  to  make  predictions  of atrazine
concentrations in any of the tributaries.

The  model results reasonably fit the available data,
and  no adjustments were made to the initial model
kinetic parameters.    A better model  fit to data
probably could have been obtained by using different
initial concentrations in different regions of the main
lake   rather  than  a  uniform,  lake-wide  initial
concentration. However,  since the January 1994
initial concentrations were estimated  from spring
1994 data, it was believed that the sampling data
were not sufficient to justify that change.

A longer-term data  set would be required to fully
calibrate  the   LM3-Atrazine   model.    However,
confidence in model predictions was provided in two
                             ways.  First,  atrazine in Lake  Michigan acts as a
                             mostly conservative chemical,  and the  model was
                             previously calibrated to the conservative chemical
                             chloride in Lake Michigan (Richardson et a/., 2001).
                             Thus  there  is  confidence that  the transport  of
                             substances,  one of the primary loss processes of
                             atrazine, was being correctly simulated.  The high-
                             resolution transport is the primary difference between
                             LM3-Atrazine and the MICHTOX and LM2-Atrazine
                             models.   Second,  an  acceptable simulation  of
                             atrazine concentrations was obtained using model
                             parameters  derived  from  literature and previous
                             modeling studies, providing confidence in the kinetic
                             formulations  and  kinetic  parameterization  of  the
                             model (Rygwelski et a/., 1999; Part 3; Part 4).

                             As part of the 2005 Lake Michigan sampling effort,
                             the USEPA  collected  atrazine  samples at multiple
                             stations during multiple cruises.  These data were not
                             available  at  the time of  this report,  but,  when
                             available, they will provide  a comparison of atrazine
                             concentrations to those measured in 1994-1995 and
                             an estimate  of the change in atrazine mass in the
                             system over that time  period. To estimate potential
                             changes in Lake Michigan atrazine concentrations
                             during  the 1994-2005 time period, additional LM3-
                             Atrazine model  simulations were conducted.  Two
                                               123

-------
model runs were conducted: one with the 1994-1995
USGS-estimated loads and one with the WEP-based
loads. The 12-year loading time-series for these runs
was developed  by repeating the appropriate 1994-
1995  loading time-series  and hydrodynamics  six
times.    Assuming  that  atrazine usage  in the
watershed  did  not change significantly from that
during   1994-1995,   results  from  these   model
simulations will likely bracket the concentrations from
the 2005 sampling  period.   Predicted mid-lake
concentrations for 2005 ranged from 38 ng/L for the
USGS tributary loading time-series to 46  ng/L for the
WEP-based tributary  loading  time-series  (Figure
5.11). Tributary loads were not sampled during the
2005 surveys, but by calculating the change in the in-
lake atrazine inventory and comparing it to the load
scenarios used for model runs, the actual magnitude
of present watershed  loads  will be  able to  be
estimated.

5.3.4.1  Mass Budgets

Inflows and outflows of atrazine to the Lake Michigan
system were tabulated during the  1994-1995 model
simulation runs (Table  5.3, Figure 5.12).   For the
model run using the USGS-estimated tributary loads
for 1994-1995, the largest source of atrazine was wet
deposition from  the atmosphere. The percentage of
loads from tributary sources was  only slightly less.
For the  model run using the long-term WEP-based
1994-1995 tributary loads, tributary loads dominated
and  were  almost three times  higher than wet
deposition.   Absorption  from  the  atmosphere
(volatilization in) was minimal for both cases.

Losses from the Lake Michigan system were similar
for both loading scenarios. The largest losses of
atrazine were from decay  and outflow through the
Straits of Mackinac, though the mass lost  through
these processes is relatively small compared to the
total atrazine inventory in the lake.  Outflow  through
the Chicago Ship and Sanitary Canal was  a  small
percentage of total mass lost, and volatilization from
water to the air was negligible.

The annual net gain of atrazine to the system for the
model run using the USGS-estimated loads was 380
kg/year, or 11 %  of the measured 1994-1995  loads to
the system. For the model run using long-term WEP-
based load estimates the net gain increased to 3,842
kg/year, equal to 55% of incoming  sources.
5.3.4.2   Selected  Model Versus Observation
Statistics

The variability in the field data made any comparison
with  model results difficult.  There was as much
variation between atrazine field duplicate samples as
there was seasonal variation predicted by the model.
Fifty-seven field duplicate and  two  field  triplicate
samples were  collected  as part of the  LMMBP
atrazine sampling.  The median absolute difference
between field duplicate samples was 1.8 ng/L,  with
the average relative percent difference (RPD) equal
to 6%.  Maximum seasonal variation in model results
from representative mid-lake stations was 1.5  ng/L
for the model run using USGS-estimated loads and
6.6 ng/L for the model run using WEP-based loads.

There were also no significant spatial  or temporal
trends in the Lake Michigan data (Brent et a/., 2001)
that  would  have  assisted  in  evaluating model
prediction capabilities. This may have  been due to
an actual lack of trends or because there  was no
near-shore sampling during the late spring and early
summer period when the lake concentrations were
predicted  to  be  most  affected  by  seasonal
atmospheric and tributary loadings.

5.3.4.3  Comparison to Toxicological Endpoints

Model simulation and  forecast results were plotted
with measured data against toxicological endpoints to
examine potential  ecological effects of  predicted
atrazine concentrations  in Lake Michigan  (Figure
5.13).  Most forecast and data concentrations were
below the selected toxicological endpoints of concern
at the spatial scale used in these modeling analyses.

The toxicological endpoints selected for Figure  5.13
were developed as part of a review of toxicity studies
used for determining  the eligibility of  atrazine for
reregistration as an herbicide (U.S. Environmental
Protection Agency, 2003a).  Endpoints for important
ecological components of the Lake Michigan system
included fish, zooplankton, other invertebrates, and
phytoplankton.  Mortality endpoints  correspond to
acute,  or short-term,  toxicity studies.   Growth or
population  reduction   endpoints correspond to
chronic, or long-term, toxicity studies.
                                              124

-------

^r 40-
1

a 20
10-
Station 27 M
,_^_

V '




Jan Jan Jan
1994 1997 2000
SO -
j40-
« 30-
JE
l»-

Station ISM
— ^OJ^,.J^.-*"' • ' '

I
fl
1
1
7i
Jan Jan
2003 2006
long term WEP loads
— USGS esfcmated loads

- '




Jan Jan Jan
1994 1997 2000
i
Jan Jan
2003 2006

•40-
)
' 30 -
20-


Station 47M
_„ «^—-~— ~~




Jan Jan Jan Jan Ja
1994 1997 2000 2003 2
-------
                            wet
                          deposition
                            1784
          volatilization
                  out
                   13

          volatilization
              in
              58
   Chicago
    River
   outflow
     110
                                                                                  Mackinac
                                                                                   outflow
                                                                                    1302
                                                USGS loads
                                           Atrazine Inventory

                                   water column = 179,459 kg
                                   mass change = + 380 kg/yr
                                               1578
                                     monitored and unmonitored
                                          tributary loading
                                     (Lake Michigan watershed)
                            wet
                         deposition
                            1784
          volatilization
                  out
                  14

          volatilization
              in
              58
decay
 1647
                                                                                 Mackinac
                                                                                  outflow
                                                                                   1313
                                                                     Long-term WEP loads
                                                                        Atrazine Inventory
                                                               water column  = 182,979 kg
                                                               mass change =+ 3842 kg/yr
   Chicago
    River
   outflow
     112
                 5086
        monitoted and unmonitored
            tributary loading
        {Lake Michigan watershed)
Figure 5.12.  Mass budget average annual results for the 1994-1995 model simulations. All mass
flow rates are in kg/yr.
                                               126

-------
               10,000=1
                1.000 -=
           o>    100 —
                  10 •=
                   1 •=
                 0,1 -=
                0.01
fish mortality (5300)


draft acute toxicity criteria CMC (1500)

invertebrate mortality (720)
invertebrate population reduction (62)
fish population reduction (62)
phytoplankton acute toxicity (32)

zooplankton population reduction (10)

human drinking water MCL (3)
                                                                o measured data
                                                                  model results
                                                                  endpoint/criteria
                                maximum measured 1995 tributary concentration
                                      in St. Joseph River (2.7)
phytoplankton primary production reduction (2.3)
highest Lake Michigan predicted concentration,
      USGS 1994-1995 loads (0.10) -St. Joseph River mouth

Lake Michigan 2263 MICHTOX forecast concentration (0.066)

Lake Michigan 2005 LM3-Atrazine forecast range (0.038 - 0.046)
                                Lake Michigan 1994 average concentration (0.037)
Figure 5.13.  Comparison of model predictions, measured data, and selected toxicological
endpoints.
Regulatory endpoints were also included in Figure
5.13.  These endpoints included proposed criteria for
environmental protection and established limits for
human health protection.  Water quality criteria for
the protection  of  aquatic ecosystems  have  been
proposed foratrazine (U.S. Environmental Protection
Agency, 2003b) but were not finalized at the time of
this  report.   The draft  acute  toxicity  Criterion
Maximum  Concentration  (CMC) was  included in
Figure 5.13.  While a  draft chronic criteria was also
published, it was not included in the figure because
it was not based upon a single concentration. The
draft  chronic criteria  were  based  upon  modeling
ecological  community changes  in aquatic  plants
using both exposure concentration and duration. The
human drinking water Maximum Contaminant Limit
(MCL) is also included in the graph.

Measured atrazine data collected during the LMMBP
were below endpoints  of toxicological concern except
for one tributary sample from the St. Joseph River in
May 1995. This measurement, 2.7 ug/L, exceeded
                    the endpoint of 2.3  ug/L at which  reductions in
                    primary production of phytoplankton were estimated
                    to  occur.  The  St. Joseph River sample was also
                    close to the human drinking  water MCL.  Detailed
                    information on determining compliance with the MCL
                    for atrazine can be found in 40 CFR 141.24(h). The
                    second highest measured tributary concentration,
                    0.55 ug/L, was a sample from the Grand River in May
                    1996 and  was below  all  selected  toxicological
                    endpoints. The 1994 Lake Michigan annual average
                    atrazine concentration of 0.037 ug/L was well below
                    the selected toxicological endpoints.

                    Model forecasts were below all selected endpoints.
                    The  MICHTOX long-term   steady-state  forecast
                    concentration   of  0.066 ug/L was  well   below
                    toxicological endpoints.  The LM3-Atrazine 12-year
                    (2005) forecast lake-wide concentration range was
                    lower than  the MICHTOX  steady-state forecast
                    concentration.  The highest simulated single model
                    cell  concentration  from  the high-resolution  LM3-
                    Atrazine model was also below selected endpoints.
                                                127

-------
The  highest  simulated  concentration,  using  the
USGS loading time-series, was 0.10 ug/L near the
mouth of the  St.  Joseph  River.    It must be
remembered that this concentration represents an
average prediction from a volume representing a 5
km by 5 km area of  the  lake which provides
significant dilution to tributary  event  loads.  The
WEP-based loading time-series was not used in this
analysis because the distribution of the long-term
annual loads among seasons and short-term loading
events was somewhat subjective.  Thus, presenting
a concentration  prediction based upon this loading
time-series at a single  location and  point in  time
would have a  large amount of uncertainty.

5.3.5 Model Uncertainty

While the LM3-Atrazine could not be fully calibrated
because of insufficient data, the basis upon which the
model was developed   provided confidence  that
model results  were reasonable. The hydrodynamic
model was successfully compared to two separate
datasets  (Schwab and Beletsky, 1998) and model
transport of a conservative substance, chloride, was
also calibrated (Richardson et a/., 2001). The only
additions to the chloride model for the LM3-Atrazine
model were volatilization and kinetic  decay terms.
Volatilization  was a minor  effect on  the fate of
atrazine in the lake, and the decay term was based
upon  long-term hindcast calibrations with   the
MICHTOX and LM2-Atrazine models.  Furthermore,
the model provided reasonable fits to data without
changing model kinetic parameters from the initial
values based  upon literature  studies  and previous
atrazine model calibration studies. There may be
some uncertainty about the decay term because the
LM2-Atrazine  model used to  calibrate the  term
incorporated  bi-directional  flow at the Straits  of
Mackinac outflow while the LM3-Atrazine model used
a  net outflow from the Straits  to  Lake  Huron.
However, this  would only  have had a minor effect on
the atrazine mass in the lake for the  time periods
modeled with the LM3-Atrazine model.

There was probably more uncertainty from  the
loading data used in the model and the field  data
than   from  the  model   kinetic processes.    The
estimated  1994-1995   tributary   loads   were
significantly less than those expected based upon
previous long-term modeling studies, and it was not
known if 1995 was a year of low atrazine loading, if
storm events  were missed  during  the tributary
sampling,  or  if there  were  additional  significant
sources of loads such as dry deposition that were not
measured.

References

Brent,  R.N., J. Schofield,  and  K.  Miller.  2001.
   Results of the Lake Michigan Mass Balance
   Study: Atrazine Data Report. U.S. Environmental
   Protection  Agency,  Great   Lakes  National
   Program   Office,   Chicago,   Illinois.
   EPA/905/R-01/010,92pp.

Capel, P.O. andS.J. Larson. 2001. Effect of Scale
   on the Behavior of  Atrazine in Surface Waters.
   Environ. Sci. Technol., 35(4):648:657.

Cerco,  C.  and T. Cole. 1994. Three-Dimensional
   Eutrophication Model of Chesapeake Bay.  U.S.
   Army Corps of Engineers, U.S. Army Engineer
   Waterways  Experiment  Station,  Vicksburg,
   Mississippi.  Technical Report EL-94-4, 658 pp.

Cerco, C. and T. Cole.   1995.  User's Guide to the
   CE-QUAL-ICM    Three-Dimensional
   Eutrophication  Model.   U.S. Army Corps  of
   Engineers,   U.S. Army  Engineer  Waterways
   Experiment  Station,  Vicksburg,  Mississippi.
   Technical Report EL-95-15, 2,420 pp.

Chapman, R.S., T.M. Cole,  and T.K. Gerald. 1997.
   Development  of Hydrodynamic/Water Quality
   (POM-IPXMT) Linkage  for the Lake Michigan
   Mass Balance Project.   Final Report.    U.S.
   Environmental  Protection  Agency,  Office  of
   Research   and  Development,  Environmental
   Research   Laboratory-Duluth,   Large  Lakes
   Research Station, Grosse lie, Michigan.  63 pp.

DePinto, J.V.,  R.  Raghunathan, P. Sierzenga,  X.
   Zhang, V.J. Bierman, Jr., P.W. Rodgers, andT.C.
   Young.  1993.  Recalibration of  GBTOX:   An
   Integrated Exposure Model for Toxic Chemicals
   in Green Bay, Lake  Michigan.  Final Report.  U.S.
   Environmental  Protection  Agency,  Office  of
   Research   and  Development,  Environmental
   Research   Laboratory-Duluth,   Large  Lakes
   Research Station, Grosse lie,  Michigan. 132 pp.
                                              128

-------
Di  Toro,   D.M.   and  J.P.   Connolly.    1980.
   Mathematical  Models of Water Quality in Large
   Lakes, Part 2: Lake Erie.  U.S. Environmental
   Protection Agency,  Office  of  Research  and
   Development,   Environmental   Research
   Laboratory-Duluth,   Large  Lakes   Research
   Station,    Grosse   Me,    Michigan.
   EPA/600/3-80/065, 97 pp.

Di Toro,  D.M.  and  W.F. Matystik,  Jr.   1980.
   Mathematical  Models of Water Quality in Large
   Lakes, Part 1: Lake  Huron and Saginaw Bay.
   U.S. Environmental Protection Agency, Office of
   Research  and  Development,  Environmental
   Research   Laboratory-Duluth,  Large  Lakes
   Research   Station,   Grosse   lie,   Michigan.
   EPA/600/3-80/056, 180 pp.

Hall, D. and D.  Robertson.   1998.  Estimation of
   Contaminant  Loading  from  Monitored  and
   Unmonitored Tributaries to Lake Michigan forthe
   USEPA Lake Michigan Mass Balance  Study.
   Quality  Systems  and Implementation Plan.
   SubmittedOctober23,1998.  U.S. Environmental
   Protection  Agency,   Great Lakes  National
   Program Office, Chicago, Illinois. 19 pp.

Leonard, B.  1991. The ULTIMATE  Conservative
   Difference Scheme Applied to Unsteady One-
   Dimensional Advection. Comp. Methods Appl.
   Mechan. Engin., 88(1):17-74.

Leu, C., H. Singer, C. Stamm,  S.R. Muller, and R.P.
   Schwarzenbach.     2004.    Simultaneous
   Assessment of Sources, Processes, and Factors
   Influencing Herbicide Losses to Surface Waters
   in a Small Agricultural Catchment.  Environ. Sci.
   Technol., 38(14):3827-3834.

Melendez, W., M. Settles, J. Pauer. 2008. LM3: A
   High-Resolution Lake  Michigan Mass Balance
   Water  Quality  Model.   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. 285 pp.
Miller, D.H., R.G. Kreis, Jr., W. Huang, and X. Xia.
   2007.  Application  of an Ecosystem Modeling
   Approach for Investigating Population Dynamics
   of the Invasive Species Bythotrepheslongimanus
   in Lake Michigan.  International Association for
   Great Lakes Research  Annual Meeting,  State
   College, Pennsylvania, May28-June 1, 2007.

Miller, S.M. 1999. Spatial and Temporal Variability
   of  Organic  and   Nutrient   Compounds   in
   Atmospheric Media Collected  During the Lake
   Michigan Mass Balance Study.   M.S.  Thesis,
   Department   of   Civil,   Structural,   and
   Environmental Engineering, State University of
   New York, Buffalo,  New York.  181 pp.

Miller, S.M.,  C.W. Sweet, J.V. DePinto, and K.C.
   Hornbuckle.  2000. Atrazine  and Nutrients in
   Precipitation:  Results from the Lake Michigan
   Mass Balance Study.   Environ.  Sci. Technol.,
   34(1):55-61.

Pauer, J.J., K.W. Taunt, and W. Melendez.  2006.
   LM3-Eutro. In:  R.  Rossmann  (Ed.), Results of
   the  Lake  Michigan Mass  Balance  Project:
   Polychlorinated Biphenyls Modeling Report,  pp.
   120-182.  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.  EPA/600R-04/167, 579 pp.

Richards, R. P., D.B. Baker, J.W. Kramer, and D.E.
   Ewing.  1996.  Annual  Loads  of Herbicides in
   Lake Erie Tributaries of Michigan and Ohio.  J.
   Great Lakes Res., 22(2):414-428.

Richardson, W.L., K.R. Rygwelski, X. Zhang, J.J.
   Pauer, and W.  Melendez.   2001.  Models  of
   Chloride and Atrazine in Lake  Michigan at Two
   Spatial Resolutions.   Presented at the 44th
   Conference  on   Great   Lakes   Research,
   International  Association  for  Great   Lakes
   Research, University of  Wisconsin, Green Bay,
   Wsconsin, June 10-14, 2001.
                                             129

-------
Rodgers, P.W. and D.  Salisbury.   1981.   Water
   Quality   Modeling   of  Lake   Michigan  and
   Consideration of the Anomalous Ice Cover of
   1976-1977. J. Great Lakes Res., 7(4):467-480.

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.

Scholtz, M.T., B.J. Van Heyst, and A. Ivanoff. 1999.
   Documentation for the Gridded Hourly Atrazine
   Emissions Data Set for the Lake Michigan Mass
   Balance Study.  U.S. Environmental Protection
   Agency, Office of Research and Development,
   National   Exposure  Research  Laboratory,
   Research Triangle Park, North Carolina. EPA/
   600/R-99/067, 61 pp.

Schottler, S.P., S.J. Eisenreich, and  P.O.  Capel.
   1994.  Atrazine, Alachlor, and Cyanazine  in a
   Large Agricultural River System.  Environ.  Sci.
   Technol.,  28(6):1079-1089.

Schwab, D.J. and D. Beletsky. 1998. Lake Michigan
   Mass Balance Study:  Hydrodynamic Modeling
   Project.   National  Oceanic and Atmospheric
   Administration,  Great  Lakes   Environmental
   Research  Laboratory,  Ann Arbor,  Michigan.
   NOAA  Technical Memorandum  ERL GLERL-
   108, 55 pp.

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.

Shipitalo, M.J. and LB. Owens.  2003.  Atrazine,
   Deethylatrazine,  and   Deisopropylatrazine in
   Surface   Runoff  from  Conservation  Tilled
   Watersheds. Environ. Sci. Technol., 37(5):944-
   950.

Thomann,   R.V.   and  D.M.  Di   Toro.    1975.
   Mathematical Modeling of Phytoplankton in Lake
   Ontario,  Part  1 -  Model  Development  and
   Verification.   U.S.  Environmental Protection
   Agency, Office of Research and Development,
   ERL-Corvallis,  Large Lakes Research Station,
   Grosse lie, Michigan.  EPA/600/3-75/005,  178
   pp.
Thomann, R.V., R.P. Wnfield, and 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,   Minnesota.
   EPA/600/3-79-094, 136 pp.

U.S.  Department of Agriculture.  2001.  Agriculture
   Research Service Pesticide Properties Database.
   Available from U.S. Department of Agriculture at
   http://www.ars.usda.gov.

U.S.  Environmental Protection  Agency.   2003a.
   Interim Reregistration Eligibility Decision (IRED)
   for  Atrazine.   U.S.  Environmental Protection
   Agency,   Office  of  Pesticides  Program,
   Washington, D.C.  Case Number 0062, 285 pp.

U.S.  Environmental Protection  Agency.   2003b.
   Ambient Aquatic Life Water Quality for Atrazine -
   Revised  Draft.  U.S. Environmental Protection
   Agency,  Office of  Water,  Washington,  D.C.
   EPA/822\R-03\023,171 pp.

Wanninkhoff, R.J. 1992. Relationship Between Gas
   Exchange and Wnd  Speed Over the Ocean. J.
   Geophys. Res.,  97:7373-7381.

Whitman, W.G. 1923.  A Preliminary Experimental
   Confirmation  of the Two-Film  Theory  of Gas
   Absorption. Chem. Metall. Eng., 29:146-148.

Williams, R.J., D.N. Brooke, P. Matthiessen, M. Mills,
   A. Turnbull, and R.M. Harrison.  1995. Pesticide
   Transport  to  Surface   Waters  Within   an
   Agricultural Catchment. J. Inst. Water Environ.
   Manag., 9(1):72-81.
                                             130

-------
                                         PART 6
                          REVIEW OF ATRAZINE MODELS
                                    Kenneth R. Rygwelski
                          United States Environmental Protection Agency
                              Office of Research and Development
                  National Health and Environmental Effects Research Laboratory
                                Mid-Continent Ecology Division
                       Large Lakes and Rivers Forecasting  Research Branch
                                            and
                               Timothy J. Feist and Xiaomi Zhang
                             Z-Tech, an  ICF International Company
                                 Large Lakes Research Station
                                      9311 Groh Road
                                  Grosse lie, Michigan 48138
6.1  LMMBP Atrazine Models

6.1.1  Peer Reviews of LMMBP Atrazine
Models

Two modeling science peer reviews were conducted
on the  Lake  Michigan Mass  Balance  Project
(LMMBP)  atrazine modeling  products.    These
reviews were conducted near the beginning and final
phases of the atrazine  modeling work.  The first
review was general in nature and was conducted on
June  23-25, 1998 in  Southgate,  Michigan  and
covered all components of the LMMBP modeling
effort including  project  design and organization;
project management, including an  evaluation  of
resources;  model linkages;  sediment transport;
loadings; hydrodynamics; model construct; atrazine;
polychlorinated biphenyls  (PCBs);  eutrophication;
and mercury. The second review was conducted on
September27,2000 in Romulus, Michigan and solely
focused on atrazine modeling.
In the first review, panel members recommended that
atrazine modeling advance to a level 2 type model
(LM2-Atrazine) with more resolution than MICHTOX.
Also, they recommended thatmanagementscenarios
for the prediction of alternative futures include model
sensitivity  runs  that include both zero atrazine
concentrations in the vapor phase and non-zero
concentrations, because measurements of the vapor
phase concentrations in the  basin were difficult to
detect  (see  Part  4 for LM2-Atrazine modeling
results).  Reviewers  encouraged  the development
and application of the high-resolution model, LM3-
Atrazine (see Part 5 for results of LM3-Atrazine, a 5
km x 5 km gridded model). The reviewers included
United  States Environmental  Protection  Agency
(USEPA),  Great  Lakes National  Program Office
(GLNPO);  Dr.  Paul Capel, United States Geological
Survey (USGS); Dr.  Miriam Diamond, University of
Toronto; Dr. Kevin Farley, Manhattan College;  Dr.
Raymond  Hoff, Environment Canada;  Dr. Robert
Hudson, University of Illinois - Urbana Champaign;
and Dr. Barry Lesht, Argonne National Laboratory.
                                            131

-------
Review  comments from the second peer  review
appear in Appendix 6.1  of this Part.   In general,
comments received on atrazine modeling for the
LMMBP were very favorable.  Reviewers included
USEPA/GLNPO;  Dr.  Paul Capel, USGS; and Dr.
Robert Hudson, University of Illinois.

The  reviewers acknowledged  that although the
modified  Stratified  Beale  Ratio Estimator (SBRE)
method  and the  USGS ESTIMATOR  used in the
LMMBP  are  standard  and  reliable methods  to
estimate loadings, the length of the data record (one
year) for the LMMBP was perhaps too short, and the
number  of samples taken from the tributaries  to
estimate  loads was limited.  Typically,  multi-year
records are used. There was follow-up discussion
and evaluation of another load estimation procedure
by Dr. Robert Hudson after the formal peer review
comments were submitted. All of the necessary files
were  provided to  Dr.  Hudson  to  make  loading
assessments using rating curves similar to what is
used in ESTIMATOR, but also to look at all the sites
together  rather than individually.   He  consulted
USGS, who performed the LMMBP load  estimates,
before performing his analysis. The new attempt was
not successful.  The  reviewers  concluded that the
load estimates made for the LMMBP  using the
watershed export percentage (WEP) approach were
most  likely the best estimates  available  for the
project.

It was recommended that further literature research
be conducted to determine what type of degradation
mechanisms may be operative in  Lake Michigan.
This was done and the results were reported in Part
1, Chapter 2.

The reviewers concluded that there are a number of
combinations  of watershed  export percentages
(WEPs) and in situ decay rates that could achieve a
model "fit" to the data. It is true that if the WEP were
increased, the decay rate would have to increase.  A
concern was raised  by  the  reviewers that these
variables are somewhat unconstrained.  However,
the WEP was constrained by focusing only on
northern freshwater drainage basins with soil texture
similar to that of the Lake Michigan basin.  Also,
since rainfall can have an effect on measured WEPs,
a balance of both  wet and dry years were included in
our long-term model  runs.  Using the mean WEP
from these studies reported in the literature was the
best estimate of the WEP's central tendency in the
Lake Michigan basin.  Indeed, one of the reviewers,
Paul Capel, looked at WEPs from 408 observations
across numerous types of soil textures after the peer
review and calculated a mean WEP of 0.66%, which
was close to our mean of 0.6% (Capel and Larsen,
2001).

Other comments included a recommendation for a
follow-up atrazine sampling of Lake Michigan water
to help confirm short-term model predictions. This
sampling was done in 2005; however, the results
were  not yet available at  the time of this printing.
Also,  the  reviewers  suggested  that  a  model
sensitivity analysis   be  conducted.    Sensitivity
analyses were performed using both the MICHTOX
model and  the  LM2-Atrazine models  and  are
reported in Parts 3 and 4 of this report.

The reviewers also were very pleased  with the
progress made with the LM3-Atrazine application and
suggested that this  high-resolution  model would be
very  useful  for  making   local   environmental
management decisions. The modelers agree  with
this  assessment and  have demonstrated  local
applications  in the vicinity of the St. Joseph River,
Fox River, Grand River, and the Kalamazoo River
mouths.  Some  of the details  of  the  St. Joseph
application are discussed in Part 5.

6.1.2 Comparison of LMMBP Models

The LMMBP models are  those discussed in this
report: MICHTOX, LM2-Atrazine, and LM3-Atrazine
(Part  3,  4, and 5, respectively).  The differences in
the  model construct among these models has been
discussed.   Total annual  atrazine  loadings for all
three  models were the same and were based on an
estimate of the  0.6% WEP.  Both MICHTOX and
LM2-Atrazine were calibrated using historical loading
estimates and comparing model output to  available
lake data.   Calibration consisted  of selecting an
appropriate in situ total decay so that model output
matched lake data.   For Scenario  3, based on
average  conditions and the most  likely scenario,
MICHTOX yielded a half-life of atrazine in the lake of
69.3 years (kinetic decay of 0.01/year). LM2-Toxic
predicted a similar half-life of 77 years (kinetic decay
of  0.009/year).   LM3-Atrazine  model used  the
0.009/yr decay derived from calibration of decay in
the  LM2-Atrazine model.
                                             132

-------
6.2   Comparison  of  LMMBP  Models to
Other Recent Atrazine Models Applied to
Lake Michigan

Within the last decade, three Lake Michigan atrazine
modeling papers have been published.  All  three
models  were  based  on the  principles of  mass
balance.  However, the  three models yielded very
different  estimates of  in situ atrazine  decay.
Tributary loads carry the most atrazine to the lake
compared  to  other  sources.    Therefore,  any
significant differences in the amount of atrazine
delivered among the models will result in a range of
internal   decay  estimates.    There  are   many
differences among these models, but the analysis
here will specifically focus on the main reasons why
these models differ.

6.2.1 Schottler and Eisenreich (1997)

Schottlerand Eisenreich (1997) predicted an internal,
overall, 14-year half-life for atrazine in Lake Michigan
using a  mass balance model called Stella.   They
used an atrazine WEP of 1% obtained from studies
on  basins  outside  of  the  Lake Michigan  basin.
However, their selection  of WEP's did not appear to
be based on soil textures that match those of the
Lake Michigan  basin.  Also, it was not clear if the
WEP they  used reflected  wet or dry  years  (or a
combination of both).  These  considerations  could
have an impact on selecting a representative WEP
for the Lake Michigan basin (See Part 2, Chapter 2).
The watersheds  were  from   both  northern  and
southern regions. The higher WEP used by Schottler
and  Eisenreich will yield higher atrazine tributary
loads to be delivered to the lake (approximately 67%
more mass loading from tributaries than the LMMBP
models delivered) and therefore more internal decay
was required in the lake to achieve a model fit to the
lake data.   Their  model  predicted that atrazine
concentrations  in the  lake were at a steady-state
concentration  of 34 ng/L in 1994, but the  model
predicted that the lake concentration was close  to
this value since the  late 1980s. The LMMBP models
suggest that the lake, under constant 1995 loadings
into the future, will reach a steady-state concentration
of 66 ng/L in the year approximately 2194.
6.2.2  Tierney et a/. (1999)

Tierney et a/. (1999) predicted that the half-life of
atrazine in Lake Michigan is about two years.  The
authors used atrazine run-off concentration data
derived from the Lake  Erie basin (Richards and
Baker, 1993), and from  Bodo (1991), who studied
watersheds  in  Southwestern Ontario  to  make
estimates of atrazine loading in the Lake Michigan
basin.  The soils in the Lake Erie basin have much
more clay (Richards and  Baker, 1993) than the soils
in  the  Lake  Michigan basin and  run-off (WEP) of
atrazine in the Lake Erie basin would likely approach
percentages over 1% (see Table  2.2.2 in Part 2 of
this report). The Lake Michigan basin has moderate
textured soils, and the run-off WEP would be closer
to  0.6%.  Using atrazine concentration data from
Lake Erie tributaries with high WEPs and applying
them to characterize tributaries in the Lake Michigan
basin would result in more atrazine loadings to Lake
Michigan than what is likely, and therefore, in situ
decay will need to be high in their  model in order for
the model to match observed lake data. High decay
is associated with the short half-life that they report.

Run-off loads of atrazine also is a strong function of
the amount of  atrazine applied to  corn  in  the
watershed.   Predicted run-off concentrations in the
Lake Michigan basin by Tierney etal. (1999) did not
appear to be based on relating corn crop acreage in
Lake  Erie basin  and Lake Michigan basin.  They
related flow-weighted concentrations in tributaries to
% total agricultural land use and then applied them to
the Lake Michigan basin. Total agricultural land use
would   be   a   poor  predictor  of  atrazine
usage/discharge if  corn crop acreage per acre
agricultural land varies within or between the Lake
Erie and the Lake Michigan basins. The reason is
that atrazine is used almost exclusively on corn crops
in the Great  Lakes  basin. There  is no indication in
the paper that an  analysis of corn crop acreage
variation within agricultural lands was performed. To
further complicate this issue, the amount of atrazine
applied to corn acreage can vary from state-to-state.

Loadings in their model (both from watershed run-off
and  precipitation)  appear to  be fixed  to levels
observed in  the  early 1990's and  applied for  the
entire historical usage period of the chemical. This
would  have  overestimated  loads from  the period
leading  up  to   approximately   1978.     This
                                              133

-------
overestimation of loads in those early years would
require  that they  include a  significant non-zero
atrazine decay term in their mass balance.

The  Tierney model predicted that  Lake Michigan
reached steady-state atrazine concentrations in the
mid- to late-1970s with a concentration of 33 ng/L.

In  contrast,  Richardson and Endicott  (1994) and
Rygwelski et al. (1999) and the modeling work in this
paper organized WEPs from the  literature and used
a WEP based  on moderate textured soils typical of
the Lake Michigan basin of 0.6%.  Furthermore,
Rygwelski et al. (1999) and  this  paper selected
WEPs  only  from northern  watersheds  only and
included a mix of both wet and dry years (see Part 2,
Chapter 2).  Also, only corn crops grown in the Lake
Michigan basin were included in  this  analysis to
determine atrazine  loadings on a county-by-county
basis.

The  results of the three atrazine models applied to
Lake Michigan  are displayed in Table 6.1. A WEP of
approximately  5.6% was calculated for the Tierney
model, based on their estimates of loads to the lake
and amount of atrazine applied to the Lake Michigan
                                  watershed.  It is clear from the table, that higher
                                  WEPs are associated with shorter atrazine half-lives.

                                  6.3  Atrazine  Models  Applied to  Lake or
                                  Deep River Systems  Outside  the  Lake
                                  Michigan Basin

                                  Other atrazine models have been applied to large
                                  freshwater lakes and  rivers.  Consistent with the
                                  results of the LMMBP models, these models have
                                  shown that little to no atrazine decays in these lakes
                                  and that  loss via outflow from the lakes or rivers is
                                  the primary atrazine removal mechanism.

                                  6.3.1  Swiss Lakes

                                  Ulrich et al. (1994) modeled atrazine in an eutrophic
                                  lake, Greifensee, in Switzerland.   The lake has a
                                  maximum depth  of  32 m with a  mean of 17.8 m.
                                  They found that, except for a short time in July and
                                  August, atrazine  showed a somewhat conservative
                                  behavior.  Within the overall mass balance, in situ
                                  decay accounted for only 5% of total annual loss of
                                  atrazine from the lake.  Ninety-five percent of the loss
                                  from the lake was attributed to outflow. The authors
Table 6.1  Comparison of LM2-Atrazine Model to Other Models
    Model
     WEP
Half-Life
Watershed Load
 Methods Used
Estimated Year to
 Reach Steady-
     State
 Atrazine Steady-
State Concentration
       ng/L
Rygwelski and
Zhang, 2007
(Part 4 of this
report)
Schottler and
Eisenreich,
1997
0.6%



1.0%


77 yrs. County
Application and
WEP

14 yrs. County
Application and
WEP
2194



1994 But
Approached Near
Steady-State
66



34


                                                            Concentration in
                                                            the Late-1980s.
 Tierney et al.,
 1999
Not Used Directly    2 yrs.
(Approx. 5.6%)
          Run-off Flow and
          0.23 |jg/L
          Forested; 1.6 |jg/L
          Agricultural (Flow-
          Weighted)
                 Mid-to Late-1970s
                          33
                                              134

-------
noted that decay in the epilimnion layer of 0.003 per
day was needed only in July and August to get the
model  to  fit observations.  They  also  noted that
during that time, nitrate levels in the lake increased.
High nitrate concentrations and high  solar energy
have  been associated  with  indirect   photolytic
degradation of atrazine in water (see Section 1.2.3.2
of this report).  Since the lake stratified in the warm
months of the year, water in the hypolimnion would
be somewhat more  isolated  from  photolytic decay
than the epilimnion.  During the rest of the year,
atrazine was modeled without  decay.   Modeled
processes such as volatilization and sedimentation
were negligible.

Buser  (1990)  modeled  atrazine  in Lake  Zurich,
Switzerland. The maximum depth of the  lake is 136
m with an average depth of 50 m.  His results also
showed atrazine to be rather stable and  its removal
primarily via outflowing waters compared to other
loss processes such  as sedimentation, degradation,
and volatilization. This lake also stratified during the
warm months of the year.

Muller et a/. (1997) modeled atrazine in three Swiss
lakes: Greifensee, Murtensee, and Sempachersee.
The maximum/mean depths for the Murtensee and
Sempachersee are 45.5 m / 23.3 m and 87 m / 44 m,
respectively. Except for the July and August period
when they used an in situ decay of 0.003 per day in
the  epilimnion,   atrazine  was   modeled  as  a
conservative  substance.   Good  agreement was
achieved  between  model output and  measured
concentrations of atrazine in the lakes.

6.3.2  St. Lawrence River

Over an 18 month period in 1995 and 1996, Pham et
a/. (2000)  measured  the  inputs  and  outputs  of
loadings of atrazine to a reach of the St. Lawrence
River.  The atrazine load was measured  in both the
upper  part of the  river  near Cornwall,  Ontario,
Canada and  at the outflow to  the estuary, near
Quebec City, Quebec, Canada.  Taking into account
loadings from the watershed, their measurements
indicated that atrazine does not degrade during the
three day transit in the river.  This  large  river has a
mean discharge of  approximately 12,000 m3/s  at
Quebec City.  At Cornwall, the depth is about 8.2 m
and at Quebec City the depth is approximately 11m.
6.4  Atrazine Models Applied to Shallow
Surface  Water Systems in  Agricultural
Areas

Atrazine  degradation  seems to  be  occurring  in
shallow surface water systems in agricultural areas.
A hypothesis is that in these shallow systems, light
energy penetrates a greater percentage of the water
column than in lakes that show thermal stratification
in the summer. Compared to these lakes,  shallow
rivers have fast mixing due to turbulence. This brings
a fresh supply of atrazine close to the surface where
photolysis can more easily degrade it.  Rivers also
generally  have higher solids concentrations that
could act as catalysts for hydrolysis. In deep lakes,
summer stratification isolates water from photolysis
in the hypolimnion and solids concentrations tend to
be lower than that found in rivers.  See  Part  1,
Chapter 2 for more discussion on this topic.

6.4.1 Saylorville Reservoir, Iowa

The Saylorville Reservoir is located on the upper Des
Moines River basin in Northern Iowa near the city of
Des  Moines.  Seventy-nine percent of the basin is
cropland, mostly corn and soybeans. The reservoir
is shallow, with a mean depth of only 4.3 m.  Chung
and Gu (2003) modeled atrazine transport and fate in
1997. During the study period, the reservoir showed
very weak thermal stratification  in  the summer
months, which allowed them to assume well-mixed
conditions.  The  authors found a strong  inverse
relationship between  half-life and  daily hours  of
sunlight. This supports the notion that photolysis was
probably operative as a loss mechanism.  In  this
system,  approximately 60% of the atrazine that
entered   the   reservoir  was  released  through
discharge.  Approximately 40%  of atrazine in the
reservoir  was  transformed  via  kinetic  loss
mechanism(s)  such as photolysis,  hydrolysis,  etc.
The half-life of atrazine in the reservoir varied from
two to 58 days. Their analysis found that the half-life
of  atrazine  did  not  correlate  well  with  nitrate
concentrations, suggesting that photolysis was not
nitrate-mediated  indirect photolysis.  Rather, they
indicated that  direct photolysis, aided  by the high
concentrations of dissolved organic carbon  (DOC),
was probably operative.
                                              135

-------
6.4.2  Other Small Surface Water Systems

Other modeling studies in small lakes and a shallow
creek in agricultural regions  have shown similar,
relatively short half-lives of atrazine.

Spalding et a/. (1994) estimated the atrazine half-
lives  in  two very  small lakes  in  Northeastern
Nebraska ranged from 124 to 193 days.  Spalding
suggested   that  hydrolysis   may  have  been
responsible  for degradation of atrazine  in  these
lakes. However, these lakes had relatively high pH's
averaging 8.1 for one lake and 8.2 for the other, and
some researchers have found that hydrolysis above
pH 4 was difficult to achieve in the laboratory. The
authors did  not rule out  photolytic decay.  These
small  lakes were very turbid where average Secchi
readings were less  than  1 m.  None of the other
atrazine  modeling  papers  reviewed  suggested
hydrolysis as a  possible explanation of atrazine
decay.

In a small creek in Iowa, Kolpin and Kalkhoff (1993)
found that atrazine half-lives had a significant inverse
relationship  with sunlight, therefore  suggesting
photolysis was responsible. To rule out temperature
as a confounding variable, they found that comparing
atrazine half-lives to  water temperature did not yield
a significant correlation.  The degradation in Roberts
Creek was very rapid.  Half-lives ranged from 168
hours (November 1990) to 35.6 hours (July 1990).
The stream discharge ranged from 0.264 to  0.086
m3/s during the sampling period, April 11,  1990 to
December 2, 1990.

 6.5  Conclusions

A likely reason for the discrepancies (see Table 6.1)
between the three Lake Michigan atrazine modeling
efforts has to do with the wide range of estimates
made for tributary loadings of atrazine to  the lake.
Since tributary loadings  are the  major source  of
atrazine to the lake, rigorous detailed efforts need to
be taken to ensure that these loads are as accurate
as possible.

Atrazine decay in large surface water systems (lakes
and rivers) appears to be much lower than decay
found in shallow water systems. In larger systems,
decay is very slow with half-lives estimated in years.
In shallow, small systems with a high degree  of
mixing, atrazine decay can be rapid with  half-lives
estimated in days or even hours.

Research suggests that decay in surface water may
be  linked to photolysis,  either direct of indirect.
Modeling studies in deeper lakes suggest that this
happens  in the summertime when  solar energy is
high.  Photolysis is limited in lakes that are stratified
or deep rivers, because the exposure of light energy
to the inventory of atrazine  in these systems is
limited. Systems that are well-mixed further facilitate
photodegradation, because a fresh supply of atrazine
is constantly being brought to the water surface
where light energy would be the greatest. Atrazine in
a hypolimnion layer would  be less  available for
photolysis because it is somewhat isolated from the
mixed epilimnion layer due to the thermocline.

In regards to Lake Michigan, can other degradation
processes besides  photodegradation explain the in
situ decay?  Per Part 1, Chapter 2,  Section 1.2.3.1,
biodegradation  in  surface waters is  not likely.
Hydrolysis in Lake Michigan is not likely because of
the high pH of 8.4, low solids, and low DOC (see Part
1, Chapter 2, Section 1.2.3.2.1).

References

Bodo,  B.A.   1991.   Trend Analysis and  Mass-
   Discharge Estimation of Atrazine  in Southwestern
   Ontario Great Lakes  Tributaries:   1981-1989.
   Environ. Toxicol. Chem., 10(9):1105-1121.

Buser, H.-R.  1990. Atrazine and Other s-Triazine
   Herbicides in Lakes and in Rain in Switzerland.
   Environ. Sci. Technol., 24(7): 1049-1058.

Capel, P.O. and S.J. Larson. 2001.  Effect of Scale
   on the Behavior of Atrazine in Surface Waters.
   Environ. Sci. Technol., 35(4):648:657.

Chung, S. and R.R. Gu.  2003.  Estimating Time-
   Variable Transformation  Rate of Atrazine  in a
   Reservoir. Adv.  Environ. Res., 7(4):933-947.

Kolpin, D.W. and S.J.  Kalkhoff.   1993.   Atrazine
   Degradation in a Small Stream in Iowa. Environ.
   Sci. Technol.,27(1):134-139.
                                               136

-------
Muller,  S.R.,  M.  Berg,  M.M.  Ulrich,  and  R.P.
   Schwarzenbach.  1997. Atrazine and Its Primary
   Metabolites in Swiss Lakes: Input Characteristics
   and  Long-Term Behavior in the Water Column.
   Environ. Sci. Technol., 31(7):2104-2113.

Pham, T.-T., B. Rondeau, H. Sabik, S. Prouix, and D.
   Cossa.  2000.  Lake Ontario: The Predominant
   Source of Triazine Herbicides in the St. Lawrence
   River.  Can. J. Fisher. Aquat. Sci., 57(Suppl.
   1):78-85.

Richards, R.P. and D.B.  Baker.  1993.  Pesticide
   Concentration  Patterns in Agricultural  Drainage
   Networks in the  Lake Erie Basin.  Environ.
   Toxicol. Chem., 12(1):13-26.

Richardson,  W.L  and  D.D.  Endicott.   1994.  A
   Screening Model for Establishing Load-Response
   Relationships  for  Toxic  Chemicals  in   Lake
   Michigan.  Presented at the Fifteenth Annual
   Meeting   of  the   Society  of  Environmental
   Toxicology and Chemistry  (SETAC),  Denver,
   Colorado, October 30 - November 3, 1994.

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.
Schottler, S.P. and S.J.  Eisenreich.  1997.  Mass
   Balance  Model to Quantify Atrazine Sources,
   Transformation Rates, and Trends in the Great
   Lakes. Environ. Sci. Technol., 31 (9):2616-2625.

Spalding, R.F., D.D. Snow, D.A. Cassada, and M.E.
   Burbach.  1994.  Study of Pesticide Occurrence
   in Two Closely Spaced  Lakes  in Northeastern
   Nebraska.  J. Environ. Qual., 23(3):571-578.

Tierney, D.P.,  P.A. Nelson,  B.R. Christensen, and
   S.M. Kloibery Watson. 1999. Predicted Atrazine
   Concentrations in the Great Lakes: Implications
   for Biological  Effects.   J. Great  Lakes Res.,
   25(3):455-467.

Ulrich, M.M.,  S.R.  Muller,  H.P.  Singer,  D.M.
   Imboden, and R.P. Schwarzenbach. 1994. Input
   and Dynamic Behavior of the Organic Pollutants
   Tetrachloroethylene,  Atrazine,  and  NTA  in  a
   Lake:    A  Study   Combining  Mathematical
   Modeling and Field Measurements. Environ. Sci.
   Technol.,28(9):1674-1685.
                                              137

-------
                                         PART 6
                          REVIEW OF ATRAZINE MODELS
Appendix  6.1   Peer  Review  of LMMBP
Atrazine  Models,  September  27,  2000,
Romulus, Michigan

Report of  the  Second  Review  Panel Meeting
Submitted to:
Dr. Glenn Warren
United States Environmental Protection Agency
Great Lakes National Program Office
77 W. Jackson Boulevard
Chicago, Illinois 60606-3590
by
Paul Capel
United States Geological Survey
Water Science Center of Minnesota
Mounds View, Minnesota 55112
and
Robert Hudson
University of Illinois
S-518 Turner Hall, MC-047
1102 South Goodwin Avenue
Urbana, Illinois 61801

A.6.1.1  Overview

The second review meeting was focused solely on
the work that the Lake  Michigan  Mass Balance
Program  (LMMBP)  has  completed on  atrazine.
Presentations were made on the following subjects:
data  quality  assurance;  summary  statistics
measurements in air, rain, tributaries, and lake water;
tributary load calculations;  modeling atmospheric
transport  and deposition;  atmospheric deposition
calculation results; hydrodynamic transport in the 41
segment model; hind/forecasting using MICHTOX;
and, simulation results from 41-segment and high-
resolution models.

The review team feels that the LMMBP has generally
met its goals for modeling atrazine loading to and
fate  and transport within  Lake Michigan.   The
following aspects of the work were notably strong:

A. The data management system and data quality
   assurance program were excellent. A great deal
   of work was expended to develop the platforms
   and communication that was needed to make
   such a large data set useful.   This work had
   recently undergone an independent review.

B. The atmospheric modeling (from volatilization to
   deposition) is an important contribution, both to
   the LMMBP and to the scientific community. This
   is the first attempt at a regional model for a semi-
   volatile chemical.  Although the work is  still
   underway, the  planned attempts to compare
   model predictions with the field measurements is
   commended.

C. The hydrodynamic  components  of the  41-
   segment model appear to be complete and well-
   calibrated, based on the results for temperature
   and chloride.  These components of the model
   will be  further tested when the focus  shifts from
   atrazine,  which  is largely  dissolved,  to  the
   particle-associated  chemicals   (mercury  and
   PCBs).

D. The simulations of atrazine fate within the lake
   based on the MICHTOX, 41-segment, and high-
   resolution models agreed well  with each other
                                            138

-------
   and with the measured data.  The fact that the
   measured atrazine concentrations were relatively
   homogeneous throughout the lake  (22  to  58
   |jg/L) made the comparison of measured and
   modeled results  a  "relatively" straightforward,
   albeit necessary test of the model.

E. TheuseofMICHTOX(Rygwelskiefa/., 1999) to
   simulate the evolution of atrazine levels  in the
   lake since atrazine use began (hindcasting) and
   to forecast future levels was an excellent way to
   tackle  the issue of  the rate of atrazine  decay
   within the lake.  It also plays an important  role in
   testing the consistency of the loading and  decay
   estimates.

F. The high-resolution modeling has significance far
   beyond the potential improvements in scientific
   understanding of atrazine fate it may bring. The
   model should be of great use in making local
   environmental  management   decisions.    In
   addition, the animations produced from the daily
   simulations  should  serve  as  an  excellent
   communications tool for environmental managers
   to reach  the public with.  This work should
   continue to be strongly encouraged.

A.6.1.2.  Comments on Technical Issues

A. Tributary Loads of Atrazine - The LMMBP
   work to date indicates that  about 2/3  of the
   atrazine load to  Lake Michigan is borne by rivers.
   The tributary loads were estimated using various
   statistical approaches, such  as the modified
   Beale  method  and the  USGS ESTIMATOR
   software, to derive loads from a limited number of
   dissolved atrazine measurements in water from
   the rivers in the Lake Michigan basin.  Although
   determining "true"  loads is impossible,  these
   estimation methods  have proved reliable and are
   considered  standard  where  non-point source
   loads  need to  be  quantified.   In  this  case,
   however, the length of the data record for each
   tributary (one-year) is short with a limited number
   of measurements in comparison to multi-year
   records that are typically used. Therefore, the
   reviewers suggest  exploring  other  statistical
   approaches that can be used on the existing data
   set.  Appendix 1 (in preparation) presents  a brief
   description of one such approach that could be
   considered.  [Note to readers: Appendix 1 was
   not completed by the review panel. However, Dr.
   Hudson did make  some loading estimates using
   rating  curves   similar  to  what  is  used  in
   ESTIMATOR,  but looked at all of the  sites
   together, rather than individually.  He consulted
   USGS,  who  performed  the  LMMBP  load
   estimates, before  performing his analysis. The
   new attempt was not successful.]

   The tributary loads were also estimated using the
   "watershed export percentage" (WEP) approach
   and the estimated annual use of atrazine in each
   watershed of rivers flowing  into the lake.  This
   approximate method serves as a good check on
   the tributary load calculations and has the benefit
   of allowing  the tributary  loads  to  easily  be
   estimated each year for the hind/forecasting.

B. Atmospheric Deposition - The magnitude of
   atmospheric deposition was estimated through
   field measurements (for rain) and simple models
   (for dry deposition).  It appears that inputs via
   rain are dominant. A single, typical value for the
   particle depositional velocity was chosen and all
   of the estimates based on this single value. The
   reviewers suggest that the  model sensitivity to
   this  approximation  should  be   examined  by
   choosing  an  appropriate  range  of   particle
   depositional velocities.  Large particles, coming
   from Chicago,  have been shown  to have much
   higher depositional velocities than the "typical"
   value used, although it is unknown how much
   atrazine is  on  these  larger  particles.   [Note to
   readers:  Only wet deposition was estimated for
   the Lake Michigan atrazine models, because dry
   deposition was negligible.  See Part 1, Chapters
   for more information.]

C. Atrazine Decay  Processes -  Atrazine was
   initially selected for study in the  LMMBP as a
   model of a reactive,  biodegradable compound
   (see Section 1.1 of Statistical Assessment of QA
   Data documents).  A half-life of  14 years was
   estimated by Schottler and Eisenreich (1997)
   based on the assumption that atrazine should be
   approximately  at  steady-state within  the lake.
   Rygwelski  et al.  (1999)  showed that current
   atrazine levels within the lake could be predicted
   from  plausible  historical  loading  estimates
                                              139

-------
   assuming no decay of atrazine within the lake.
   This approach leads to predictions of very large,
   continued  increases  in  lake  atrazine  levels.
   Although atrazine levels are not likely to exceed
   current drinking water standards, this scenario is
   obviously of  greater concern than the steady-
   state assumption.

   Further  literature  review of  mechanisms  of
   atrazine  decomposition is warranted  to  help
   determine which mechanisms are most likely to
   be significant in the lake. Given the current state
   of knowledge, it may be difficult  to resolve this
   issue.  However, the full range of processes -
   biodegradation,  photochemical decomposition,
   and chemical hydrolysis should be considered.
   The possibility of more significant decay within
   the lake needs to be kept open.

D. Summary - Both the data and modeling results
   suggest  that atrazine may not be as  reactive
   within  the  lake as originally anticipated.   This
   question is probably best resolved by continued
   monitoring of atrazine levels  in lake water.  The
   model  results  can  be used  to ensure  that
   sampling locations are  not unduly affected by
   tributary  inputs.  Further modeling work in this
   area should combine the historical approach with
   parameter sensitivity analyses.   The  results
   presented in the review meeting showed that the
   measured  data  can be correctly modeled by
   different combinations of  WEP  and  atrazine's
   degradation rate in Lake Michigan.  At this time,
   neither  parameter  is well-constrained.   It is
   suggested  by  the  panel  that the LMMBP
   investigate the  relationship  between values of
   WEP and degradation rate  that yield accurate
   estimates of current atrazine levels from historical
   loading   rates.     Presumably,   an  inverse
   relationship  between the two will result, with an
   acceptable range for each.

   The above discussion concerns an example of
   variables  in  the models  that are constrained at
   this time only to a range of values, rather than a
   single correct value.  The LMMBP might wish to
   consider other model variables to evaluate  the
   model's sensitivity to the appropriate ranges of
   these values and to the relationships between
   parameters.

References

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.

Schottler, S.P.  and  S.J. Eisenreich.  1997.   Mass
   Balance Model to Quantify Atrazine Sources,
   Transformation Rates, and Trends  in the Great
   Lakes. Environ. Sci. Technol., 31(9):2616-2625.
                                               140

-------