Preliminary Interpretation of the Ecological Significance of Atrazine
Stream-Water Concentrations Using a Statistically-Designed Monitoring
Program 

In Support of an

Interim Reregistration Eligibility Decision

on Atrazine

Submitted to the FIFRA Scientific Advisory Panel

for Review and Comment

December 4 – 7, 2007

Prepared by the

Office of Prevention, Pesticides, and Toxic Substances

Office of Pesticide Programs

Environmental Fate and Effects Division

Washington, D. C.

Office of Research and Development

National Health and Environmental Effects Research Laboratory (NHEERL)

Mid-Continent Ecology Division

Duluth, MN

and

Western Ecology Division,

Corvalis, OR

November 15, 2007

Contents

  TOC \o "1-5" \h \z \u    HYPERLINK \l "_Toc182871016"  I. Introduction
  PAGEREF _Toc182871016 \h  8  

  HYPERLINK \l "_Toc182871017"  A.	Purpose of the Scientific Advisory
Panel Meeting	  PAGEREF _Toc182871017 \h  8  

  HYPERLINK \l "_Toc182871018"  B.	Background	  PAGEREF _Toc182871018 \h
 8  

  HYPERLINK \l "_Toc182871019"  C.	Monitoring Study Objectives	  PAGEREF
_Toc182871019 \h  9  

  HYPERLINK \l "_Toc182871020"  D.	Ecological Endpoint and Level of
Concern for Atrazine	  PAGEREF _Toc182871020 \h  10  

  HYPERLINK \l "_Toc182871021"  1.	Freshwater Microcosm/Mesocosm Studies
on Atrazine	  PAGEREF _Toc182871021 \h  10  

  HYPERLINK \l "_Toc182871022"  2.	Determining a Level of Concern Based
on Microcosm/Mesocosm Studies	  PAGEREF _Toc182871022 \h  11  

  HYPERLINK \l "_Toc182871023"  E.	Atrazine Ecological Monitoring
Program	  PAGEREF _Toc182871023 \h  13  

  HYPERLINK \l "_Toc182871024"  F.	Key Study Documents	  PAGEREF
_Toc182871024 \h  14  

  HYPERLINK \l "_Toc182871025"  1.	CASM Model Development	  PAGEREF
_Toc182871025 \h  14  

  HYPERLINK \l "_Toc182871026"  2.	Watershed Selection Process	  PAGEREF
_Toc182871026 \h  15  

  HYPERLINK \l "_Toc182871027"  3.	Overview of Atrazine Monitoring Study
Results	  PAGEREF _Toc182871027 \h  15  

  HYPERLINK \l "_Toc182871028"  4.	Sampling Frequency Analyses	  PAGEREF
_Toc182871028 \h  15  

  HYPERLINK \l "_Toc182871029"  5.	GIS Approaches to Assessing Watershed
Vulnerability	  PAGEREF _Toc182871029 \h  16  

  HYPERLINK \l "_Toc182871030"  G.	Issues To Be Addressed in Future SAP
Consultations	  PAGEREF _Toc182871030 \h  16  

  HYPERLINK \l "_Toc182871031"  II. Use of a Community Simulation Model
for Extrapolation of Atrazine Levels of Concern for Exposure Time-Series
  PAGEREF _Toc182871031 \h  18  

  HYPERLINK \l "_Toc182871032"  A.	Problem Definition and General
Strategy	  PAGEREF _Toc182871032 \h  18  

  HYPERLINK \l "_Toc182871033"  1.	Nature of Assessment Problem	 
PAGEREF _Toc182871033 \h  18  

  HYPERLINK \l "_Toc182871034"  2.	General Strategy for Assessment
Methodology	  PAGEREF _Toc182871034 \h  20  

  HYPERLINK \l "_Toc182871035"  B.	Model Formulation and
Parameterization	  PAGEREF _Toc182871035 \h  22  

  HYPERLINK \l "_Toc182871036"  1.	General Model Formulation and
Parameterization for Reference Conditions	  PAGEREF _Toc182871036 \h  22
 

  HYPERLINK \l "_Toc182871037"  2.	Incorporation of Toxicity Information
into the Model	  PAGEREF _Toc182871037 \h  23  

  HYPERLINK \l "_Toc182871038"  C.	Model Implementation and Example
Application	  PAGEREF _Toc182871038 \h  27  

  HYPERLINK \l "_Toc182871039"  1.	Model Effects Index Selection	 
PAGEREF _Toc182871039 \h  27  

  HYPERLINK \l "_Toc182871040"  2.	LOC Determination for Base Model
Configuration	  PAGEREF _Toc182871040 \h  28  

  HYPERLINK \l "_Toc182871041"  3.	Example Applications	  PAGEREF
_Toc182871041 \h  30  

  HYPERLINK \l "_Toc182871042"  D.	Sensitivity Analysis	  PAGEREF
_Toc182871042 \h  31  

  HYPERLINK \l "_Toc182871043"  1.	Overview	  PAGEREF _Toc182871043 \h 
31  

  HYPERLINK \l "_Toc182871044"  2.	Model Effects Index	  PAGEREF
_Toc182871044 \h  32  

  HYPERLINK \l "_Toc182871045"  3.	Exposure Start Date	  PAGEREF
_Toc182871045 \h  33  

  HYPERLINK \l "_Toc182871046"  4.	Environmental Parameters	  PAGEREF
_Toc182871046 \h  34  

  HYPERLINK \l "_Toc182871047"  5.	Plant Toxicity EC50s	  PAGEREF
_Toc182871047 \h  34  

  HYPERLINK \l "_Toc182871048"  6.	Chronic, Low-Level Exposures	 
PAGEREF _Toc182871048 \h  37  

  HYPERLINK \l "_Toc182871049"  7.	Slope of the Toxicity Curve	  PAGEREF
_Toc182871049 \h  37  

  HYPERLINK \l "_Toc182871050"  E.	Summary and Next Steps	  PAGEREF
_Toc182871050 \h  37  

  HYPERLINK \l "_Toc182871051"  F.	SAP Charge Questions on the Use of
the CASM Model	  PAGEREF _Toc182871051 \h  37  

  HYPERLINK \l "_Toc182871052"  III. Atrazine Monitoring Program For
Ecological Effects: Determining The Extent Of Waters Exceeding
Effects-Based Thresholds For Atrazine	  PAGEREF _Toc182871052 \h  39  

  HYPERLINK \l "_Toc182871053"  A.	Monitoring Study Objectives	  PAGEREF
_Toc182871053 \h  39  

  HYPERLINK \l "_Toc182871054"  B.	Study Design	  PAGEREF _Toc182871054
\h  40  

  HYPERLINK \l "_Toc182871055"  1.	Site Selection	  PAGEREF
_Toc182871055 \h  40  

  HYPERLINK \l "_Toc182871056"  a)	Watershed Vulnerability Assessment	 
PAGEREF _Toc182871056 \h  40  

  HYPERLINK \l "_Toc182871057"  b)	Representative Sampling of Vulnerable
Watersheds	  PAGEREF _Toc182871057 \h  45  

  HYPERLINK \l "_Toc182871058"  (1)	Target Population	  PAGEREF
_Toc182871058 \h  45  

  HYPERLINK \l "_Toc182871059"  (2)	Survey Design	  PAGEREF
_Toc182871059 \h  46  

  HYPERLINK \l "_Toc182871060"  (3)	Sample Size And Analysis Weights	 
PAGEREF _Toc182871060 \h  47  

  HYPERLINK \l "_Toc182871061"  (4)	Sampled Population	  PAGEREF
_Toc182871061 \h  48  

  HYPERLINK \l "_Toc182871062"  (5)	Monitoring Question	  PAGEREF
_Toc182871062 \h  50  

  HYPERLINK \l "_Toc182871063"  c)	Selection Of Monitoring Site Within
The Watershed	  PAGEREF _Toc182871063 \h  50  

  HYPERLINK \l "_Toc182871064"  2.	Sampling/Monitoring Design	  PAGEREF
_Toc182871064 \h  51  

  HYPERLINK \l "_Toc182871065"  C.	Monitoring Results	  PAGEREF
_Toc182871065 \h  52  

  HYPERLINK \l "_Toc182871066"  1.	Atrazine Chemographs	  PAGEREF
_Toc182871066 \h  52  

  HYPERLINK \l "_Toc182871067"  2.	Analyzing Atrazine Chemographs: LOC
Exceedances	  PAGEREF _Toc182871067 \h  54  

  HYPERLINK \l "_Toc182871068"  D.	Percentage Of Watersheds Exceeding
The LOC Threshold	  PAGEREF _Toc182871068 \h  57  

  HYPERLINK \l "_Toc182871069"  1.	Statistical Analysis Based On GRTS	 
PAGEREF _Toc182871069 \h  57  

  HYPERLINK \l "_Toc182871070"  a)	Methods	  PAGEREF _Toc182871070 \h 
57  

  HYPERLINK \l "_Toc182871071"  b)	Assumptions	  PAGEREF _Toc182871071
\h  58  

  HYPERLINK \l "_Toc182871072"  c)	Results	  PAGEREF _Toc182871072 \h 
59  

  HYPERLINK \l "_Toc182871073"  2.	Population Estimates	  PAGEREF
_Toc182871073 \h  59  

  HYPERLINK \l "_Toc182871074"  a)	Excluded Sites	  PAGEREF
_Toc182871074 \h  59  

  HYPERLINK \l "_Toc182871075"  b)	Sites That Did Not Exceet the LOC	 
PAGEREF _Toc182871075 \h  60  

  HYPERLINK \l "_Toc182871076"  c)	Sites That Exceeded the LOC in At
Least 2 Years	  PAGEREF _Toc182871076 \h  60  

  HYPERLINK \l "_Toc182871077"  d)	Sites That Exceeded the LOC in 1 Year
  PAGEREF _Toc182871077 \h  60  

  HYPERLINK \l "_Toc182871078"  e)	Uncertain Sites	  PAGEREF
_Toc182871078 \h  60  

  HYPERLINK \l "_Toc182871079"  f)	Reliability of WARP estimates for
HUCs and Sub-watersheds	  PAGEREF _Toc182871079 \h  62  

  HYPERLINK \l "_Toc182871080"  g)	Comparison of Alternative WARP
estimates for HUCs	  PAGEREF _Toc182871080 \h  65  

  HYPERLINK \l "_Toc182871081"  E.	Sensitivity / Uncertainty Analyses	 
PAGEREF _Toc182871081 \h  70  

  HYPERLINK \l "_Toc182871082"  1.	Weather	  PAGEREF _Toc182871082 \h 
71  

  HYPERLINK \l "_Toc182871083"  2.	Flow rates / Low flow sites	  PAGEREF
_Toc182871083 \h  72  

  HYPERLINK \l "_Toc182871084"  a)	Low Flow Sites	  PAGEREF
_Toc182871084 \h  72  

  HYPERLINK \l "_Toc182871085"  b)	Flow Measured at the Monitoring Sites
  PAGEREF _Toc182871085 \h  73  

  HYPERLINK \l "_Toc182871086"  3.	Sampling Frequency / Auto-samples	 
PAGEREF _Toc182871086 \h  77  

  HYPERLINK \l "_Toc182871087"  a)	Evaluation of Sampling Frequency
Using AEMP Data	  PAGEREF _Toc182871087 \h  78  

  HYPERLINK \l "_Toc182871088"  b)	Auto Sample Results	  PAGEREF
_Toc182871088 \h  80  

  HYPERLINK \l "_Toc182871089"  c)	Syngenta PRZM Augmentation	  PAGEREF
_Toc182871089 \h  82  

  HYPERLINK \l "_Toc182871090"  d)	Evaluation of Alternate Sampling
Frequency Strategies	  PAGEREF _Toc182871090 \h  86  

  HYPERLINK \l "_Toc182871091"  e)	Conclusions	  PAGEREF _Toc182871091
\h  91  

  HYPERLINK \l "_Toc182871092"  F.	SAP Charge Questions on the Atrazine
Monitoring Results	  PAGEREF _Toc182871092 \h  91  

  HYPERLINK \l "_Toc182871093"  IV. Approaches To Address The Question
“Where Are The Waters That Are Exceeding Effects-Based Atrazine
Thresholds?”	  PAGEREF _Toc182871093 \h  93  

  HYPERLINK \l "_Toc182871094"  A.	From Watersheds To Waterbodies	 
PAGEREF _Toc182871094 \h  93  

  HYPERLINK \l "_Toc182871095"  B.	From The 40 Sampled Watersheds To The
Larger Population Of Vulnerable Watersheds	  PAGEREF _Toc182871095 \h 
95  

  HYPERLINK \l "_Toc182871096"  1.	Evaluation of WARP Parameters	 
PAGEREF _Toc182871096 \h  95  

  HYPERLINK \l "_Toc182871097"  2.	Evaluation Of Other Soil- And
Hydrology-Related Parameters	  PAGEREF _Toc182871097 \h  97  

  HYPERLINK \l "_Toc182871098"  C.	SAP Charge Questions Relating to
Identifying Where Atrazine Exceedances Are Likely to Occur	  PAGEREF
_Toc182871098 \h  102  

  HYPERLINK \l "_Toc182871099"  V. References	  PAGEREF _Toc182871099 \h
 103  

  HYPERLINK \l "_Toc182871100"  VI. Appendix 1: Microcosm and Mesocosm
Studies Used in CASM_Atrazine	  PAGEREF _Toc182871100 \h  114  

  HYPERLINK \l "_Toc182871101"  VII. Appendix 2: Summary of Atrazine
Monitoring Data	  PAGEREF _Toc182871101 \h  126  

  HYPERLINK \l "_Toc182871102"  VIII. Appendix 3: Comparisons of
precipitation during monitoring years to historical averages	  PAGEREF
_Toc182871102 \h  135  

  HYPERLINK \l "_Toc182871103"  1.	IA-01, 2004-05	  PAGEREF
_Toc182871103 \h  136  

  HYPERLINK \l "_Toc182871104"  2.	IA-02, 2004-05	  PAGEREF
_Toc182871104 \h  138  

  HYPERLINK \l "_Toc182871105"  3.	IL-01, 2004-05	  PAGEREF
_Toc182871105 \h  140  

  HYPERLINK \l "_Toc182871106"  4.	IL-02, 2004-05	  PAGEREF
_Toc182871106 \h  142  

  HYPERLINK \l "_Toc182871107"  5.	IL-03, 2005-06	  PAGEREF
_Toc182871107 \h  144  

  HYPERLINK \l "_Toc182871108"  6.	IL-04, 2005-06	  PAGEREF
_Toc182871108 \h  146  

  HYPERLINK \l "_Toc182871109"  7.	IL-05, 2004-05	  PAGEREF
_Toc182871109 \h  148  

  HYPERLINK \l "_Toc182871110"  8.	IL-06, 2004-05	  PAGEREF
_Toc182871110 \h  150  

  HYPERLINK \l "_Toc182871111"  9.	IL-07, 2004-05	  PAGEREF
_Toc182871111 \h  152  

  HYPERLINK \l "_Toc182871112"  10.	IL-08, 2005-06	  PAGEREF
_Toc182871112 \h  154  

  HYPERLINK \l "_Toc182871113"  11.	IL-09, 2004-05	  PAGEREF
_Toc182871113 \h  156  

  HYPERLINK \l "_Toc182871114"  12.	IN-01, 2004-05	  PAGEREF
_Toc182871114 \h  158  

  HYPERLINK \l "_Toc182871115"  13.	IN-02, 2004-05	  PAGEREF
_Toc182871115 \h  160  

  HYPERLINK \l "_Toc182871116"  14.	IN-03, 2005-06	  PAGEREF
_Toc182871116 \h  162  

  HYPERLINK \l "_Toc182871117"  15.	IN-04, 2004-06	  PAGEREF
_Toc182871117 \h  164  

  HYPERLINK \l "_Toc182871118"  16.	IN-05, 2004-06	  PAGEREF
_Toc182871118 \h  167  

  HYPERLINK \l "_Toc182871119"  17.	IN-06, 2005-06	  PAGEREF
_Toc182871119 \h  170  

  HYPERLINK \l "_Toc182871120"  18.	IN-07 2005-06	  PAGEREF
_Toc182871120 \h  172  

  HYPERLINK \l "_Toc182871121"  19.	IN-08 2005-06	  PAGEREF
_Toc182871121 \h  174  

  HYPERLINK \l "_Toc182871122"  20.	IN-09 2005-06	  PAGEREF
_Toc182871122 \h  176  

  HYPERLINK \l "_Toc182871123"  21.	IN-10 2005-06	  PAGEREF
_Toc182871123 \h  178  

  HYPERLINK \l "_Toc182871124"  22.	IN-11 2005-06	  PAGEREF
_Toc182871124 \h  180  

  HYPERLINK \l "_Toc182871125"  23.	KY-01 2005-06	  PAGEREF
_Toc182871125 \h  182  

  HYPERLINK \l "_Toc182871126"  24.	KY-02 2005-06	  PAGEREF
_Toc182871126 \h  184  

  HYPERLINK \l "_Toc182871127"  25.	MN-01 2005-06	  PAGEREF
_Toc182871127 \h  186  

  HYPERLINK \l "_Toc182871128"  26.	MO-01, 2004-06	  PAGEREF
_Toc182871128 \h  188  

  HYPERLINK \l "_Toc182871129"  27.	MO-02, 2004-06	  PAGEREF
_Toc182871129 \h  191  

  HYPERLINK \l "_Toc182871130"  28.	MO-03, 2004-06	  PAGEREF
_Toc182871130 \h  194  

  HYPERLINK \l "_Toc182871131"  29.	NE-01, 2004-2005	  PAGEREF
_Toc182871131 \h  197  

  HYPERLINK \l "_Toc182871132"  30.	NE-02 2005-06	  PAGEREF
_Toc182871132 \h  199  

  HYPERLINK \l "_Toc182871133"  31.	NE-03, 2004-2005	  PAGEREF
_Toc182871133 \h  201  

  HYPERLINK \l "_Toc182871134"  32.	NE-04 2005-06	  PAGEREF
_Toc182871134 \h  203  

  HYPERLINK \l "_Toc182871135"  33.	NE-05 2005-06	  PAGEREF
_Toc182871135 \h  205  

  HYPERLINK \l "_Toc182871136"  34.	NE-06 2004-06	  PAGEREF
_Toc182871136 \h  207  

  HYPERLINK \l "_Toc182871137"  35.	NE-07 2005-06	  PAGEREF
_Toc182871137 \h  210  

  HYPERLINK \l "_Toc182871138"  36.	OH-01 2004-05	  PAGEREF
_Toc182871138 \h  212  

  HYPERLINK \l "_Toc182871139"  37.	OH-02 2005-06	  PAGEREF
_Toc182871139 \h  214  

  HYPERLINK \l "_Toc182871140"  38.	OH-03 2004-05	  PAGEREF
_Toc182871140 \h  216  

  HYPERLINK \l "_Toc182871141"  39.	OH-04 2005-06	  PAGEREF
_Toc182871141 \h  218  

  HYPERLINK \l "_Toc182871142"  40.	TN-01 2005-06	  PAGEREF
_Toc182871142 \h  220  

 

List of Tables

  TOC \h \z \c "Table"    HYPERLINK \l "_Toc182871143"  Table II-1
Atrazine plant toxicity tests used in assessment methodology.	  PAGEREF
_Toc182871143 \h  24  

  HYPERLINK \l "_Toc182871144"  Table II-2 Ten sets of randomly selected
EC50s (ug/L atrazine) for sensitivity analysis.	  PAGEREF _Toc182871144
\h  36  

  HYPERLINK \l "_Toc182871145"  Table III-1 Evaluation of selected
watershed vulnerability approaches for atrazine monitoring, based on
Williams et al, 2004a	  PAGEREF _Toc182871145 \h  43  

  HYPERLINK \l "_Toc182871146"  Table III-2 Target Population and Survey
Design Stratification	  PAGEREF _Toc182871146 \h  46  

  HYPERLINK \l "_Toc182871147"  Table III-3 Watersheds selected for
monitoring using the GRTS approach.	  PAGEREF _Toc182871147 \h  49  

  HYPERLINK \l "_Toc182871148"  Table III-4 Summary of chemograph
shapes, numbers and magnitudes of atrazine peaks for the 40 monitoring
sites	  PAGEREF _Toc182871148 \h  52  

  HYPERLINK \l "_Toc182871149"  Table III-5 Summary of monitoring
results and LOC exceedances for the 40 watersheds.	  PAGEREF
_Toc182871149 \h  55  

  HYPERLINK \l "_Toc182871150"  Table III-6 Population Estimates for the
1,172 Vulnerable Watersheds Based on  the 40 Monitoring Sites	  PAGEREF
_Toc182871150 \h  61  

  HYPERLINK \l "_Toc182871151"  Table III-7 Population Estimates
Assuming Non-Monitored HUCs Missing at Random	  PAGEREF _Toc182871151 \h
 61  

  HYPERLINK \l "_Toc182871152"  Table III-8 Atrazine Use Data and Model
Inputs Used for Alternative WARP Estimates	  PAGEREF _Toc182871152 \h 
66  

  HYPERLINK \l "_Toc182871153"  Table III-9 Correlations of Alternative
WARP Predictions for 1172 vulnerable HUCs	  PAGEREF _Toc182871153 \h  66
 

  HYPERLINK \l "_Toc182871154"  Table III-10 Correlations of Alternative
WARP Predictions for 5860 High Atrazine Use HUCs	  PAGEREF _Toc182871154
\h  68  

  HYPERLINK \l "_Toc182871155"  Table III-11 Comparison of HUCs in Top
20% WARP Predictions	  PAGEREF _Toc182871155 \h  68  

  HYPERLINK \l "_Toc182871156"  Table III-12 Summary of annual average
flow rate for the 40 monitoring sites by year.	  PAGEREF _Toc182871156
\h  73  

  HYPERLINK \l "_Toc182871157"  Table III-13 Comparison of estimated
atrazine concentrations between 4-day and 12-day grab samples, IN-04,
2004	  PAGEREF _Toc182871157 \h  80  

  HYPERLINK \l "_Toc182871158"  Table III-14 4-day Grab Samples
Augmented with Auto Samples - Rolling Averages (ppb)	  PAGEREF
_Toc182871158 \h  81  

  HYPERLINK \l "_Toc182871159"  Table III-15 Percent Difference Between
Grab Averages and Auto Sample Adjusted Averages using CASM SSI%	 
PAGEREF _Toc182871159 \h  82  

  HYPERLINK \l "_Toc182871160"  Table III-16 Summary of Percent
Differences Between Original Rolling Average Concentrations and SSI% and
revised Rolling Average Concentrations (ppb) and SSI% for PRZM Augmented
Chemographs from Snyder, et al (2007)	  PAGEREF _Toc182871160 \h  84  

  HYPERLINK \l "_Toc182871161"  Table III-17 Percentile of Ranked
Distribution of Percent Differences from Table III-16	  PAGEREF
_Toc182871161 \h  86  

  HYPERLINK \l "_Toc182871162"  Table III-18 Sampling frequency
strategies used by Crawford (2004)	  PAGEREF _Toc182871162 \h  87  

  HYPERLINK \l "_Toc182871163"  Table III-19 Error Distribution for
Various Sample Strategies of Crawford (2004) for Different Concentration
Profiles	  PAGEREF _Toc182871163 \h  87  

  HYPERLINK \l "_Toc182871164"  Table III-20 Variability of estimated
concentrations based on different sample frequencies (Crawford, 2004)	 
PAGEREF _Toc182871164 \h  89  

  HYPERLINK \l "_Toc182871165"  Table VI-1 Micro- and mesocosm studies
table with Brock scores and estimated average % change in community
structure (Steinhaus similarity) of primary producers	  PAGEREF
_Toc182871165 \h  114  

  HYPERLINK \l "_Toc182871166"  Table VII-1 Summary of Detection
Frequencies by Monitoring Site and Year	  PAGEREF _Toc182871166 \h  126 


  HYPERLINK \l "_Toc182871167"  Table VII-2 Summary of Rolling Averages
From Ecological Watershed Monitoring Data for Comparison with CASM
Thresholds	  PAGEREF _Toc182871167 \h  128  

  HYPERLINK \l "_Toc182871168"  Table VII-3 Sites Ranked from Lowest to
Highest Based on Annual Average Flow	  PAGEREF _Toc182871168 \h  130  

  HYPERLINK \l "_Toc182871169"  Table VII-4 Summary of Rolling Average
Concentrations (ppb) and SSI% for PRZM Augmented Chemographs from
Snyder, et al (2007)	  PAGEREF _Toc182871169 \h  133  

 List of Figures

  TOC \h \z \c "Figure"    HYPERLINK \l "_Toc182871170"  Figure I-1
Species sensitivity distributions of various atrazine-induced toxicity
endpoints.	  PAGEREF _Toc182871170 \h  10  

  HYPERLINK \l "_Toc182871171"  Figure II-1 Examples of atrazine
exposure time series (chemographs) in natural systems.	  PAGEREF
_Toc182871171 \h  19  

  HYPERLINK \l "_Toc182871172"  Figure II-2 Score for the severity of
effects in microcosm and mesocosm atrazine exposures versus exposure
duration and average concentration.	  PAGEREF _Toc182871172 \h  20  

  HYPERLINK \l "_Toc182871173"  Figure II-3 LOC=4.0 based on correlation
of micro/mesocosm effect scores with model effect index values (upper
panel), and comparison of model LOC to micro/ mesocosm effect scores at
different exposure concentration/ durations (lower panel)	  PAGEREF
_Toc182871173 \h  29  

  HYPERLINK \l "_Toc182871174"  Figure II-4 Model effect index values
and multiplication factors for sixteen sample chemographs, using base
model configuration	  PAGEREF _Toc182871174 \h  31  

  HYPERLINK \l "_Toc182871175"  Figure II-5 Comparison of multiplication
factors for the base model configuration to alternatives with different
model effect indices	  PAGEREF _Toc182871175 \h  33  

  HYPERLINK \l "_Toc182871176"  Figure II-6 Comparison of multiplication
factors for base model configuration to alternative configurations with
different start dates	  PAGEREF _Toc182871176 \h  34  

  HYPERLINK \l "_Toc182871177"  Figure II-7 Comparison of multiplication
factors for the base model configuration to alternative configurations
with lower and higher nutrients.	  PAGEREF _Toc182871177 \h  34  

  HYPERLINK \l "_Toc182871178"  Figure II-8 Comparison of multiplication
factors for the base model configuration to alternative configurations
with lower and higher temperature	  PAGEREF _Toc182871178 \h  35  

  HYPERLINK \l "_Toc182871179"  Figure II-9 Comparison of multiplication
factors for the base model configuration to alternative configurations
with lower and higher light.	  PAGEREF _Toc182871179 \h  35  

  HYPERLINK \l "_Toc182871180"  Figure II-10 Comparison of
multiplication factors for the base model configuration (bold solid
line) to the mean (bold dashed line) and mean +/- one standard deviation
(narrow dashed lines) of the multiplication factors for ten alternative
plant EC50 sets.	  PAGEREF _Toc182871180 \h  36  

  HYPERLINK \l "_Toc182871181"  Figure III-1 Ranking of watersheds in
quintiles using the USGS WAtershed Regression on Pesticides model	 
PAGEREF _Toc182871181 \h  42  

  HYPERLINK \l "_Toc182871182"  Figure III-2 Location of atrazine use
area (gray), areas of use > 0.25 lb/ac (green), most vulnerable 20th
percentile of watersheds (yellow), and sampled watersheds (red).	 
PAGEREF _Toc182871182 \h  44  

  HYPERLINK \l "_Toc182871183"  Figure III-3 Location of the 40
watersheds (red) sampled for the atrazine eco-monitoring study.	 
PAGEREF _Toc182871183 \h  48  

  HYPERLINK \l "_Toc182871184"  Figure III-4 Estimated Cumulative
Distribution for CASM Score.	  PAGEREF _Toc182871184 \h  62  

  HYPERLINK \l "_Toc182871185"  Figure III-5 Comparing WARP scores for
HUC-based watersheds vs. sub-watersheds	  PAGEREF _Toc182871185 \h  63  

  HYPERLINK \l "_Toc182871186"  Figure III-6 Comparison of WARP scores
for sub-watershed and HUCs	  PAGEREF _Toc182871186 \h  63  

  HYPERLINK \l "_Toc182871187"  Figure III-7 Illustrating WARP’s
capabilities to accurately predict moderately high vs. high
vulnerability to atrazine by relating it to resulting CASM scores.	 
PAGEREF _Toc182871187 \h  64  

  HYPERLINK \l "_Toc182871188"  Figure III-8 Comparing how well WARP
predicts maximum 14 day rolling averages	  PAGEREF _Toc182871188 \h  65 


  HYPERLINK \l "_Toc182871189"  Figure III-9 Scatterplots of Alternative
WARP Predictions for 1172 vulnerable HUCs	  PAGEREF _Toc182871189 \h  67
 

  HYPERLINK \l "_Toc182871190"  Figure III-10 Comparison of the spatial
distribution of the original vulnerable watersheds used to select the
monitoring sites with watersheds identified with the full WARP equation
using Dunne Overland Flow.	  PAGEREF _Toc182871190 \h  69  

  HYPERLINK \l "_Toc182871191"  Figure III-11 Comparison of original
vulnerable watersheds identified using 1998-2002 use data with
vulnerable watersheds identified using 2001-03 use data.	  PAGEREF
_Toc182871191 \h  70  

  HYPERLINK \l "_Toc182871192"  Figure III-12 Atrazine concentrations
and measured stream flow, NE-04, 2005	  PAGEREF _Toc182871192 \h  75  

  HYPERLINK \l "_Toc182871193"  Figure III-13 Atrazine concentrations
and measured stream flow, NE-05, 2005	  PAGEREF _Toc182871193 \h  76  

  HYPERLINK \l "_Toc182871194"  Figure III-14 Atrazine concentrations
and measured stream flow, NE-07, 2005	  PAGEREF _Toc182871194 \h  76  

  HYPERLINK \l "_Toc182871195"  Figure III-15 Atrazine concentrations
and measured stream flow, MO-01, 2005	  PAGEREF _Toc182871195 \h  77  

  HYPERLINK \l "_Toc182871196"  Figure III-16 Graphical Representation
of PRZM Augmentation of a 4-Day Grab Sample Time series (taken from
Syngenta Presentation Titled “Atrazine Ecological Monitoring Program
review” dated December 14, 2006)	  PAGEREF _Toc182871196 \h  83  

  HYPERLINK \l "_Toc182871197"  Figure IV-1 Extent of stream segments in
the MO-02 HUC that met the sample selection criteria (from Harbourt et
al, 2004).	  PAGEREF _Toc182871197 \h  94  

  HYPERLINK \l "_Toc182871198"  Figure IV-2 Comparison of WARP values to
%SSI deviation for the 1,172 vulnerable watersheds (sites exceeding the
4% LOC trigger are labeled)	  PAGEREF _Toc182871198 \h  96  

  HYPERLINK \l "_Toc182871199"  Figure IV-3 Comparison of original WARP
values based on 1998-2002 use data with new WARP values based on
2001-2003 use data	  PAGEREF _Toc182871199 \h  97  

  HYPERLINK \l "_Toc182871200"  Figure IV-4 Depth to restrictive soil
layers in MO-01 and MO-02, located in the Central Claypan MLRA.	 
PAGEREF _Toc182871200 \h  99  

  HYPERLINK \l "_Toc182871201"  Figure IV-5 Distribution of Hydrologic
Group C and D Soils in MO-01 and MO-02, located in the Central Claypan
MLRA.	  PAGEREF _Toc182871201 \h  100  

  HYPERLINK \l "_Toc182871202"  Figure IV-6 Distribution of Hydrologic
Group C and D Soils in the Atrazine Monitoring Sites in Nebraska,
Missouri, Iowa, and western Illinois	  PAGEREF _Toc182871202 \h  101  

 

Preliminary Interpretation of the Ecological Significance of Atrazine
Stream-Water Concentrations Using a Statistically-Designed Monitoring
Program

 Introduction

Purpose of the Scientific Advisory Panel Meeting

In January, 2003, the US Environmental Protection Agency (US EPA) issued
an ecological risk assessment as part of the Interim Registration
Eligibility Decision (IRED) for atrazine (US EPA, 2003a).  As a
condition of re-registration, the atrazine registrants were required to
develop a monitoring program to determine whether atrazine
concentrations in streams associated with corn and sorghum production
were exceeding a designated effects-based threshold. This threshold was
based on aquatic plant community effects. If this threshold is exceeded,
then a watershed-based mitigation program would be required.

This document summarizes the Agency’s preliminary review and
interpretation of the results of a three-year atrazine monitoring
program conducted in flowing water bodies associated with corn and
sorghum production. The Agency is consulting with this Scientific
Advisory Panel (SAP) on scientific issues related to the interpretation
of the results of the monitoring program and the extent to which the
methods used by the Agency could be used or adapted in any future
atrazine aquatic assessments or monitoring efforts to determine the
extent to which water bodies exceed atrazine thresholds of concern for
aquatic community effects.

In its review and sensitivity analysis of the aquatic community model
used to relate atrazine monitoring concentrations to effects found in
microcosm and mesocosm studies, the Agency identified some issues
related to the model’s response to low concentrations and to slope
responses. The Agency has posed questions for the SAP on these issues
and, once they are resolved, will return to the SAP with an update.

Background

Atrazine, a triazine herbicide currently registered for use against
broadleaf and some grassy weeds, was first registered for use in 1958,
and is estimated to be one of the most widely used herbicides in the
United States.  Atrazine inhibits primary production by reversibly
blocking photosynthesis. It is both mobile and persistent in the
environment.  

The atrazine ecological risk assessment for the January 2003 IRED
identified ecological risk concerns from the use of atrazine based on
the potential for community- and population-level risks to aquatic
ecosystems at prolonged concentrations of atrazine ranging from 10 to 20
µg/L. The Agency required the atrazine registrants, in consultation
with the US EPA, to develop a program under which the registrants
monitor for atrazine concentrations and mitigate environmental exposures
if the US EPA determines that mitigation is necessary. This program was
to focus on stream systems within a watershed context (US EPA, 2003a).

The scope of the monitoring program included identification of an
appropriate ecological level of concern (LOC) based on the IRED and
development of a protocol for monitoring that specifies the frequency,
location, and timing of sampling. The program also identified atrazine
exposures that would trigger mitigation measures and described resultant
mitigation measures. This monitoring and mitigation program was
designed, conducted and implemented on a tiered watershed level basis
consistent with existing state and federal water quality programs (US
EPA, 2003a).

On October 31, 2003, the US EPA issued an addendum to update the January
2003 IRED. This addendum specified the key questions the US EPA wanted
the monitoring study to address, outlined the methodology for
determining the LOC trigger, and briefly described the monitoring study
design and proposed protocol submitted by Syngenta Crop Protection,
Inc., the principal atrazine registrant. The LOC trigger was based on
results of microcosm and mesocosm studies. An existing, previously
published aquatic community model was used to compare atrazine atrazine
monitoring data to exposure profiles observed in the microcosm and
mesocosm studies to ascertain if the level of concern derived from these
studies was exceeded. The monitoring protocol described a
watershed-based approach that identified 40 monitoring sites
representing watersheds associated with corn and sorghum production that
are the most vulnerable watersheds for atrazine contamination and
specified a minimum of two years of monitoring to determine if LOC
triggers were exceeded. The 40 watersheds are statistically
representative of 1,172 potentially vulnerable watersheds. Results of
the monitoring study will be used to determine if further monitoring or
remedial efforts are needed (US EPA, 2003b).

Mitigation actions in the watershed were triggered if atrazine
concentrations from the monitoring site exceeded the LOC trigger for two
years. If the site exceeded the LOC trigger in any one year, then a
third year of monitoring would be required before mitigation actions
would be undertaken (US EPA, 2003b). 

Monitoring Study Objectives

Based on the 2003 ecological risk assessment for atrazine, the Agency
identified several key questions to be addressed in the monitoring study
(US EPA, 2003b). These questions formed the study objectives required to
meet the US EPA’s needs under the Federal Insecticide, Fungicide, and
Rodenticide Act (FIFRA) and support the development of an atrazine
aquatic life criterion under the Clean Water Act.

Identify aquatic community level thresholds of concern based on
available microcosm and mesocosm studies and develop a method that
relates these aquatic community responses to atrazine exposure profiles
in a reasonable and transparent manner. 

Design a monitoring program to estimate the extent of watersheds in corn
and sorghum producing areas that have flowing waters which exceed
atrazine LOC triggers for aquatic community effects.

Based on results of the monitoring study, identify watershed attributes
that can be used to identify where these higher atrazine exposure areas
are likely to occur.

Syngenta designed the study to address these objectives. This document
focuses on the first two objectives. While initial approaches concerning
the third objective are presented, a subsequent document and SAP
consultation will focus on the final objective.

Ecological Endpoint and Level of Concern for Atrazine

Freshwater Microcosm/Mesocosm Studies on Atrazine

Figure   STYLEREF 1 \s  I -  SEQ Figure \*
ARABI⁃獜ㄠᐠᔱ匠数楣獥猠湥楳楴楶祴搠獩牴扩瑵潩獮
漠⁦慶楲畯⁳瑡慲楺敮椭摮捵摥琠硯捩瑩⁹湥灤楯瑮⹳

EC50s for aquatic plant growth (●), LC50s/EC50s for aquatic animal
acute lethality/immobiliation (□) and various chronic/ sublethal
animal toxicity endpoints (■), from Tables 5.1 and 5.2 of Giddings

Numerous toxicity tests with a variety of plant and animal species have
shown aquatic plant growth to generally be much more sensitive to
atrazine than are various effects on aquatic animals.  Figure I-1 shows
species sensitivity distributions of EC50s for plant growth, EC50s for
acute lethality/immobilization to aquatic animals, and various
chronic/sublethal endpoints for aquatic animals, based on the Table 5.1
and 5.2 summaries of Giddings et al. (2000).  At comparable percentages
in these cumulative distributions, plant growth is circa 3-10 times more
sensitive than chronic endpoints for animals and circa 20-100 times more
sensitive than acute toxicity to animals.  Within each distribution, the
data used for Figure I-1 vary with respect to the duration of exposure
and nature of effects, so this figure is intended to only illustrate
general differences in taxa sensitivity, not rigorously define these
distributions.  Because of the plant sensitivity illustrated in Figure
I-1, the problem definition here focused on effects to the aquatic plant
community.

Applying individual toxicity tests such as those summarized in Figure
I-1 to plant community risk assessments is made difficult by issues
related to the rapid, reversible nature of small to moderate effects of
atrazine on plants, the influences of competition and compensation
within the community, recovery rates of community perturbations, and the
seasonality of community sensitivity.  In contrast, microcosm and
mesocosm studies with atrazine address the aggregate responses of
multiple species in aquatic plant communities. Mesocosm and microcosm
studies also allow observation of population and community recovery from
atrazine effects and of indirect effects on higher trophic levels.  

Atrazine has been the subject of many mesocosm and microcosm studies. 
The durations of these studies have ranged from a few weeks to several
years at exposure concentrations ranging from 0.1 to 10,000 µg/L.  Most
of the studies have focused on atrazine effects on phytoplankton,
periphyton, and macrophytes; however, some also included measurements on
animals.

Based on these studies, as described in the 2003 IRED for atrazine (U.S.
EPA, 2003a), potential adverse effects on sensitive aquatic plants and
non-target aquatic organisms, including populations and communities, are
likely to be greatest when atrazine concentrations in water equal or
exceed approximately 10 to 20 µg/L on a recurrent basis or over a
prolonged period of time.  Appendix 1 summarizes the freshwater aquatic
microcosm, mesocosm, and field studies that were reviewed as part of the
2003 IRED.  An open literature search for studies not included in the
2003 IRED was completed in May 2007.  It found that the available
studies all showed effects levels to freshwater fish, invertebrates, and
aquatic plants at concentrations greater than 10 µg/L. 

 

Determining a Level of Concern Based on Microcosm/Mesocosm Studies

Based on the results of available micro- and mesocosm studies for
atrazine, the US EPA identified changes in the aquatic plant community
structure as the endpoint of concern.  This appeared to be the most
sensitive endpoint affecting aquatic plants. Further, the effect of
atrazine on aquatic plants, whether direct or indirect, appeared to be
more sensitive than effects on other organisms in the aquatic ecosystem,
e.g., aquatic invertebrates or fish. Thus, by focusing on aquatic plant
community structural changes, the Agency would, in effect, protect
against adverse effects on the rest of the aquatic community.  

The degradates of atrazine were not included in the determination of the
endpoint of concern since their toxicity to aquatic plants, freshwater
fish, and aquatic invertebrates is much lower than that of the parent
molecule (US EPA, 2007, 2003a). Given the lesser toxicity of the
degradates compared to the parent, and the relatively small proportion
of the degradates expected to be in the environment and available for
exposure relative to atrazine, the focus of this assessment is parent
atrazine.  

The ecological level of concern (LOC) was based on 77 results from 25
micro- and mesocosm studies on atrazine (Appendix 1). The analysis
evaluated the change in aquatic community structure and function of
primary producers. The studies measured growth (rate) and biomass in the
laboratory and reduction in primary production and changes in structure
of primary producer communities in the field.

Establishing the LOC required quantifying the results of the microcosm
and mesocosm studies using a comparable measure of effects on aquatic
plant productivity and community structure. An analysis of the reported
effect(s) and the atrazine exposure profile (i.e., the magnitude and
duration of atrazine concentrations) revealed a wide range of study
designs and quality and also indicated that a wide range of atrazine
exposure profiles could result in significant change in aquatic
community productivity and structure. A method was developed to separate
the reported results on plant community productivity and structure
observed in these studies into those that were significant versus those
with slight to no effects.  

First, the severity of effects of atrazine on the aquatic plant
community were quantified. Because there was not a single, consistent,
quantitative effects measure that could be compared among these studies,
Brock et al. (2000) analyzed many of these microcosm/mesocosm studies
using an effects score summarized below: 

Effect Scores:

1   = 	no effect

2   = 	slight effect

3   = 	significant effect followed by return to control levels within 56
d

4   = 	significant effect without return to control levels during an
observation period of

less than 56 d

5   = 	significant effect without return to control levels for more than
56 d

The US EPA subjected those studies not included in the Brock analysis to
the same scoring system.  For the 77 results from the microcosm/mesocosm
studies, 15 had a Brock score of 1, 12 had a score of 2, 12 had a score
of 3, 23 had a score of 4, and 15 had a score of 5. 

Since atrazine exposure profiles in natural systems, in this case
streams, are typically complex, a method was needed to analyze
monitoring data to determine when monitored exposure profiles are
functionally equivalent to those profiles observed in mesocosm and
microcosm studies that showed significant changes or, conversely, that
showed no significant effects.  A model that predicts changes in aquatic
communities in streams, the Comprehensive Aquatic Systems Model (CASM)
(DeAngelis et al. 1989; Bartell et al. 1999, 2000; Bartell 2003), was
used to assess whether an atrazine exposure profile from a monitoring
study would likely be associated with a significant effect on aquatic
communities.  CASM is an ecological food web model that can indicate
changes in the aquatic community structure and function of primary
producers due to the addition of a chemical to the system.  CASM can
include a large number of species in the model structure and readily
accepts complicated chemical exposure profile inputs.

Syngenta commissioned the CASM model developer to provide the Agency
with a version of the model specifically tailored to atrazine,
containing information on how atrazine affects different species.  This
model, the Copmprehensive Aquatic Systems Model for Atrazine
(CASM_Atrazine), was developed to simulate complex ecological production
dynamics of a 2nd or 3rd order Midwestern stream. (Volz, et al., 2007)
This Agency is using this model version to interpret the data generated
by the ecological monitoring program.  This current version has a shell
which will allow users to input their data. Once current development and
testing are concluded, it will be made publicly available  

One output from the CASM_Atrazine model, the Steinhaus Similarity Index
(SSI) deviation, represents the relative magnitude of change in aquatic
community structure and function of primary producers between results of
the simulated atrazine concentration profile and results of a control
simulation, i.e., the same community over the same time period with no
atrazine input.  SSI deviation is expressed as a percentage, with higher
percentages indicating greater deviation of the atrazine-exposed aquatic
community from the control.  

With model simulations of the exposure profiles in the microcosm and
mesocosm studies, a value for the SSI can be selected to segregate those
studies that exhibited significant effects (Brock scores of 3-5) from
those that did not (Brock scores of 1-2).   An initial assessment of
such an LOC was provided in the IRED addendum (US EPA 2003b) and is
updated in Section II of this document. 

CASM simulations can be used to test whether the LOC was exceeded for
any exposure time-series (e.g., a chemograph from a monitoring study).
Section II also provides examples of this approach and describes a
sensitivity analysis of the CASM model that addresses possible sources
of uncertainty in the approach.

Atrazine Ecological Monitoring Program

The next step in the assessment involved designing a monitoring program
that generated chemographs sufficient to assess the magnitude, duration,
and frequency of atrazine exposures, given the potentially flashy nature
of atrazine exposure in stream systems. Because the LOC trigger reflects
both magnitude and duration of exposure, the chemograph must reflect
sampling at a sufficient frequency to characterize atrazine exposure in
flowing waters. The monitoring program needed to determine the extent to
which waters in vulnerable watersheds exceed the effects-based
thresholds for atrazine. In addition, Syngenta was asked to collect
watershed and sub-watershed data that could be used to identify other
areas where waters exceeding the LOC thresholds are located. 

Syngenta monitored 40 sites for at least 2 years based on a
probability-based survey design that sampled 1,172 vulnerable
watersheds. The study used a watershed-based approach to focus on those
water bodies that would likely be most vulnerable to atrazine loadings.
Surveying vulnerable watersheds increases the probability of finding
waters with atrazine exposures above the levels of concern. Conversely,
if no problems were found in these watersheds, the probability is low
that problems would exist in less vulnerable watersheds. HUC-10/11 scale
watersheds (typically 40,000 to 250,000 acres in size) provided a
workable scale for evaluating the vulnerability of watersheds and
establishing the watershed population of interest. A tiered approach
based first on atrazine use intensity and then on the USGS’ Watershed
Regression for Pesticides (WARP) model (Larson et al, 2004) defined the
most potentially vulnerable tier of watersheds based on a ranking of
WARP values. From this tier, a spatially-balanced survey design was used
to select a representative sample of 40 watersheds. Williams et al.
(2004a) describe the watershed vulnerability approach and site selection
process. Section III of this document provides the US EPA review of the
approach and interpretation in relation to the monitoring study
objectives. 

Section III also presents the Agency’s review and interpretation of
Syngenta’s monitoring study for the 40 watersheds (Hampton et al.,
2007a). Both Syngenta (Hampton et al., 2007b) and the US EPA are
continuing to evaluate the results to address the third study objective
relating to identifying where the watersheds that exceed the LOC might
occur. With regard to this latter issue, Section IV of this document
describes some approaches the US EPA is considering and requests comment
from the SAP. A future SAP consultation will focus on interpretations of
the study results with a particular focus on applying the results to
identify watershed attributes and watersheds with corn and sorghum
production most likely to have streams with the highest atrazine
concentrations.

Key Study Documents

Syngenta has submitted a number of documents to the US EPA related to
the development of a method of determining the threshold trigger for
atrazine impacts on aquatic communities and to monitoring for atrazine
in watersheds representative of the most vulnerable watersheds. The
Agency is reviewing these studies as a part of its assessment of the
monitoring program and is presenting its preliminary assessment to the
SAP.  The following documents sufficiently capture the details and
results of the study and have been provided to the SAP for additional
background.

CASM Model Development

Volz, D.C., S.M. Bartell, S.K. Nair, and P. Hendley. 2007. Modeling the
Potential for Atrazine-Induced Changes in Midwestern Stream Ecosystems
using the Comprehensive Aquatic Systems Model (CASM). Final Report. MRID
47174103.

This report contains background on the CASM model, documentation of the
development of the model specifically for atrazine (CASM_Atrazine),
development of rolling average triggers, and an uncertainty analysis
conducted by the model developer. The report concludes that the model
provides a flexible tool for predicting community-level impacts from
atrazine exposure in low-order Midwestern streams using monitoring data.

As a result of the US EPA evaluation, some model parameters have been
revised, resulting in a revised LOC trigger. The US EPA evaluation and
sensitivity analyses are reported in Section II. 

Watershed Selection Process

Williams, W. M., C.M. Harbourt, M.K. Matella, M.H. Ball, and J.R. Trask.
2004a. Atrazine Ecological Exposure Flowing Water Chemical Monitoring
Study in Vulnerable Watersheds Interim Report: Watershed Selection
Process.  Prepared by Waterborne Environmental, Inc., Leesburg, VA for
Syngenta Crop Protection, Inc., Greensboro, NC. 

Williams, W. M., C.M. Harbourt, M.H. Ball, M.K. Matella, J.R. Trask, and
N.J. Snyder. 2004b.  Atrazine Ecological Exposure Monitoring Program
Interim Report: Supporting Spatial Data.  Prepared by Waterborne
Environmental, Inc., Leesburg, VA for Syngenta Crop Protection, Inc.,
Greensboro, NC.  

Williams et al (2004a) describes the steps Syngenta used to identify the
most vulnerable watersheds based on atrazine use intensity and factors
affecting the potential for atrazine runoff using the USGS Watershed
Regression on Pesticides (WARP) model. It also describes the criteria
used for selecting a monitoring site within the watersheds. Williams et
al (2004b) provides detailed documentation on the spatial and monitoring
data used in the watershed vulnerability assessment and site selection
process.

Section III describes US EPA’s evaluation of the watershed
vulnerability and selection approach.

Overview of Atrazine Monitoring Study Results

Hampton, M., Burnett, G., Carver, L.S., Harbourt, C.M., Hendley, P.,
Johnston, E.A., Perez, S., Snyder, N.J., and Trask, J.R., 2007a.  2007
Interim Report - 2004 - 2006 Data Overview - Atrazine Ecological
Exposure Flowing Water Chemical Monitoring Study in Vulnerable
Watersheds Interim Report.  Prepared by Waterborne Environmental, Inc.,
Leesburg, VA for Syngenta Crop Protection, Inc., Greensboro, NC.  MRID
47174102.

This report includes a description of the sampling instrumentation and
methods and provides the monitoring results for the 40 sites. 

The US EPA used the data in this report for its analysis of the
monitoring results in Section III. In addition, the US EPA has
interpreted the results based on a revised version of the CASM-Atrazine
model described in Section II. The results, compared against a revised
LOC, are discussed in Section III.

Sampling Frequency Analyses

Snyder, N.J., Harbourt, C.M., Miller, P.S., Trask, J.R., Prenger, J.J.,
Hendley, P., and Johnston, E.A., 2007.  Atrazine Ecological Exposure
Flowing Water Chemical Monitoring Study in Vulnerable Watersheds:
Analysis of Chemograph Behavior between Grab Samples - Measurement and
Hybrid PRZM Approaches.  Prepared by Waterborne Environmental, Inc.,
Leesburg, VA for Syngenta Crop Protection, Inc., Greensboro, NC.  

Snyder et al (2007) compared the results of the 4-day grab samples to
both autosampler data triggered by flow events at selected sites and to
data where the Pesticide Root Zone Model (PRZM) version 3.12.2 (Carsel
et al, 1998) was used to estimate atrazine concentrations between
sampling events. While the CASM SSI deviation values generally increased
with both the autosample data and the PRZM-estimated concentrations,
neither resulted in chemographs that triggered the threshold LOC.

The USEPA also evaluated sampling frequency uncertainties and various
approaches for describing this uncertainty in Section III. 

GIS Approaches to Assessing Watershed Vulnerability

Hampton, M. Prenger, J.J., Harbourt, C.M., Hendley, P., and Miller,
P.S., 2007b.  Atrazine Ecological Exposure Flowing Water Chemical
Monitoring Study in Vulnerable Watersheds: Approaches to Assessing
Potential Watershed Scale Vulnerability for Atrazine Runoff.  Prepared
by Waterborne Environmental, Inc., Leesburg, VA for Syngenta Crop
Protection, Inc., Greensboro, NC.  MRID 47174101.

This report provides a brief summary of the original approach for
identifying watersheds vulnerable to atrazine runoff (see Williams et
al, 2004a) but focuses primarily on additional analyses with spatial
data that was not as readily available at the time of the original study
design.

The USEPA has completed preliminary analyses of this study, as well as
its own assessments. Those analyses are discussed briefly in Section IV
in terms of future directions, but will be the subject of a future SAP.

Issues To Be Addressed in Future SAP Consultations

As noted in the purpose section, the Agency is consulting with this SAP
on the use of a community simulation model as a means of extrapolating
the results of microcosm/mesocosm studies to stream monitoring exposure
data and on the preliminary interpretation of the results of a
watershed-based monitoring study in the corn and sorghum growing region.
Several issues noted in this document will be brought to a future SAP.
These are highlighted below:

Once issues related to the way in which the CASM model appears to both
overestimate the effects of extended-duration, low-level exposures and
underestimate the effects of short-term, high-level exposures are
resolved (see Section II.D), the Agency will return to the SAP with a
revised interpretation of the monitoring study.

Monitoring continues in 2007 for a number of sites. The results of this
monitoring may change the current interpretation of some of the
monitoring sites (see Section III.D.2). These updates will be folded
into the revised interpretation.

A separate pilot monitoring study is underway for water bodies in
sugarcane areas. Additionally, the Agency is evaluating existing
monitoring of reservoirs to determine the potential impact of atrazine
on aquatic communities in static water bodies. The interpretation of
these results may require an aquatic community model specific for the
sugarcane areas. To the extent that such a model departs from the
current CASM approach, the Agency may determine that a consultation with
the SAP is warranted.

Based on input from this SAP, the Agency will continue to work on
approaches to (a) identify streams that exceed the LOC and (b) identify
additional watersheds and/or streams beyond the 40 monitoring sites that
exceed the LOC. The Agency plans a future SAP consultation on the
results of this effort.

 Use of a Community Simulation Model for Extrapolation of Atrazine
Levels of Concern for Exposure Time-Series

This section addresses procedures for assessing the ecological level of
concern (LOC) for atrazine exposures in freshwater systems.  It first
defines the problem and presents a general strategy for the assessment
methodology. Second, it describes the formulation and parameterization
of a community simulation model to be used in the methodology. After a
discussion of the implementation of the model with example applications,
a sensitivity analysis is presented to address the uncertainties
introduced into assessments due to various choices for model formulation
and parameterization.

Problem Definition and General Strategy

Nature of Assessment Problem

The problem being addressed here is a common one in aquatic risk
assessments, namely that exposure time-series in natural ecosystems for
which risks need to be assessed are markedly different from exposure
time-series in experimental systems for which effects have been
determined. Experimental system exposures usually involve a relatively
constant concentration over a pre-determined duration, sometimes
followed by assessment of recovery after exposure termination.  Field
exposures tend to be more variable and open-ended, without a fixed
duration.  This is particularly true for atrazine, which enters aquatic
systems largely in rainfall-driven runoff, resulting in highly variable
and episodic exposures that depend on rainfall distribution, atrazine
application patterns, topography, and soil properties.  The assessment
methodology therefore needs to address effects under exposure scenarios
such as those in   REF _Ref180376001 \h  Figure II-1 , which includes a
variety of exposure time-series derived from atrazine monitoring
efforts.

As summarized in Section   REF _Ref182025133 \w \h  I.D , LOCs are based
on plant community effects observed in micro/mesocosm studies,
discriminating between no or slight effects (Brock Scores 1-2) and
significant effects (Brock Scores 3-5).  A major assumption here is that
these micro/mesocosm data collectively describe a relationship of
effects to exposure that is applicable to the field sites of interest. 
Although concentrations during the exposure periods in these studies are
not absolutely constant, exposures can be approximately characterized by
a concentration and duration, which are related to the effect scores in 
 REF _Ref181755648 \h  Figure II-2 .  The two groups of scores to be
discriminated in setting an LOC (1-2 versus 3-5) are largely separated (
 REF _Ref181755648 \h  Figure II-2 ); however, some overlap exists,
indicative of variable effects among the different types of systems
and/or uncertainty in evaluation methods, the causes of which are
uncertain.  The assessment methodology must therefore address how LOCs
should be defined relative to the presence of false negatives
(micro/mesocosm scores of 3-5 which lie below the LOC) and false
positives (scores of 1-2 which lie above the LOC).   

Figure   STYLEREF 1 \s  II -  SEQ Figure \* ARABIC \s 1  1  Examples of
atrazine exposure time series (chemographs) in natural systems.

Chemograph numbers 7 and 9 are not shown because they are identical to
numbers 6 and 8, respectively, except shifted to two weeks earlier.

  

Figur⁥–呓䱙剅䙅ㄠ尠⁳䤔ᕉጞ匠充䘠杩牵⁥⩜䄠䅒䥂
⁃獜ㄠᐠᔲ匠潣敲映牯琠敨猠癥牥瑩⁹景攠晦捥獴椠⁮
業牣捯獯⁭湡⁤敭潳潣浳愠牴穡湩⁥硥潰畳敲⁳敶獲獵
攠灸獯牵⁥畤慲楴湯愠摮愠敶慲敧挠湯散瑮慲楴湯‮

 denotes score 2, and ∆ denotes score 1.

General Strategy for Assessment Methodology

Although the use of the micro/mesocosm data better addresses integrated
and cumulative aquatic community effects than would application of
toxicity test results for individual plant species, it still presents
the challenge of extrapolating to highly variable field exposures (  REF
_Ref180376001 \h  Figure II-1 ) from the more regular exposures in the
experimental systems.  Thus a method is needed to provide a consistent,
quantitative index of effect for the entire range of exposures in both
the experimental systems and the field systems of interest.  This
methodology does not need to provide any absolute predictions of effects
in any system, but rather must be able to provide a measure of the
relative severity of effects resulting from different exposure patterns.
 Conceptually, this would allow LOCs defined based on micro/mesocosm
data to be extrapolated to different exposure time-series in field
systems.  It should be emphasized that the purpose of this methodology
is only to address extrapolations across exposure time-series, and not
to address other factors which might make atrazine effects vary between
systems (such as might be partly responsible for the overlap of the
score on Figure II-2) and which are not predictable based on current
knowledge. 

Just as it was decided to base LOCs on plant community effects observed
in experimental ecosystems rather than individual species toxicity
tests, it was decided to use a simulation model that addresses an entire
aquatic community for the desired extrapolations among exposure
time-series.  By incorporating representations of major aquatic
community processes and the effects of atrazine on these processes, this
model is expected to provide useful assessments of the relative impacts
of different exposures, and thereby support extrapolations among
exposure time-series for different experimental and field ecosystems. 
For example, if 100 ug/L atrazine was the LOC for a 60 day constant
exposure, what would be the LOC for a 30-day constant exposure?  Or what
would be the LOC based on the highest peak (or some average
concentration) in a chemograph with a complex shape, such as any of
those in   REF _Ref180376001 \h  Figure II-1 ?

The general steps in the assessment methodology are as follows:  

(1)  A single version of an aquatic community model is formulated and
parameterized. This model formulation should include a sufficient number
of plant species to accommodate assigning a range of atrazine
sensitivities reflective of available single-species toxicity tests ( 
REF _Ref181755648 \h  Figure II-2 ).  Because relative effects of
different exposures are of concern, rather than absolute effects on a
specific system, this formulation and parameterization will not address
any specific system of interest, but still should have some general
relevance to the types of experimental and natural systems of interest. 
Just as it was assumed that the micro/mesocosm results have generic
relevance to field assessments, it is assumed that exposure
extrapolations based on this single model version would have generic
applicability to the experimental and field ecosystems.  The sensitivity
analysis in Section   REF _Ref180376659 \w \h  II.D  will evaluate this
assumption and some uncertainties associated with it.

(2)  A model simulation is conducted for each of the atrazine
concentration/duration combinations in the micro/mesocosm data set, as
well as for the absence of atrazine exposure (control).  The duration of
the model simulations is not the exposure duration, but rather a fixed
length long enough to assess effects over an entire growing season (with
the exposure period being a subset of the total simulation). The model
simulation also should be timed to start at a typical expected onset
date of significant atrazine exposure in natural systems.  A measure of
the change of the model output variables between control and exposed
conditions that provides a consistent measure of effects independent of
exposure duration is designated as a model effects index.

(3)  The observed effects scores (1-5) for each micro/mesocosm treatment
are then correlated to the model effects index values computed by the
model for the exposure concentrations/durations of each treatment.  An
LOC for the model effects index is selected that provides the best
discrimination between the effects score groups (i.e., 1-2 vs 3-5).  The
model output is thus "calibrated" to the micro/mesocosm data so that the
LOC for the model effects index is related to effects of concern
identified in the micro/mesocosm experiments.  It is important to
emphasize that the model effects index value selected as the LOC does
not define the level of ecological protection being provided.  Rather,
the significance of the model LOC is in (a) its correlation to the
effects in the micro/mesocosm experiments that do define the desired
level of ecological protection and (b) its utility as a reference value
for assessing the relative impact of different exposure time-series.    

 

(4)  A model simulation is then conducted for each of the field exposure
time-series of interest, and the model effects indices for these
exposures are examined to determine if the LOC is exceeded.  In
addition, the model could iteratively estimate the factor
(multiplication factor) by which the time-series would need to be
increased or decreased for the model effects index to exactly equal the
LOC, thereby providing further information useful for risk assessment
and management. 

Model Formulation and Parameterization

General Model Formulation and Parameterization for Reference Conditions

The aquatic community simulation model selected to provide the desired
extrapolation tool was the "Comprehensive Aquatic Systems Model" (CASM).
 This selection was based on the extensive history of this model (and it
predecessors and derivatives) in aquatic risk assessments (O'Neill et
al. 1981, 1983; DeAngelis et al. 1989; Bartell et al. 1992, 1999, 2000;
Hanratty and Stay 1994; Hanratty and Liber 1996; Naito et al. 2002,
2003; Bartell 2003) and its ready adaptability for use in the atrazine
assessment methodology.  Other conceptually similar models (e.g.
AQUATOX) could have been used and would likely have provided similar
extrapolations if comparably formulated and parameterized.  General
information regarding the CASM model in the aforementioned references
and its specific adaptation for atrazine is described in Volz et al.
(2007).  Only a summary of the general nature of the model and its basic
parameterization for reference conditions will be given here.   

The state variables for CASM are (a) the biomasses for various species
defining a simplified aquatic community and (b) concentrations for
dissolved oxygen, dissolved and particular organic matter, and certain
nutrients.  The state equation for each biological species is a
bioenergetics equation that includes terms, as appropriate to each
species, which define gains/losses of biomass from photosynthesis,
respiration, food consumption, export/import, mortality, etc.  For
example, for phytoplankton species i, the basic bioenergetics equation
is:

 	(1)

where Bi is the biomass of the species, t is time, pi is the gross
photosynthesis rate, ri is the respiration rate, si is a sinking rate,
mi is a natural mortality rate, and gij is the grazing rate by a
consumer species j.  Each of these rates requires specification of (a) a
functional form that relates the rate to various state and input
variables and (b) one to several parameters required for function; for
example, pi is a function of light, temperature, and nutrients which
includes parameters for maximum photosynthesis rate, light saturation,
optimal temperature, and nutrient half-saturation concentrations
(DeAngelis et al. 1989; Bartell et al. 1999, 2000; Bartell 2003).  
Input variables to CASM include light, temperature, nutrient imports,
water flow, and depth.  

        

For atrazine assessment needs, a version of CASM was formulated and
parameterized to represent a second- to third-order midwestern U.S.
stream that would be a typical recipient of runoff from atrazine corn
applications (Volz et al. 2007).  This model aquatic community includes
ten species of phytoplankton, ten species of periphyton, six species of
macrophytes, two species of zooplankton, five species of benthic
invertebrates, seven species of fish, and bacteria in both the water
column and sediment.  Physicochemical environmental input variables were
based on data for an Ohio stream, and parameters for the bioenergetics
equations were obtained from peer-reviewed technical literature (Volz et
al. 2007).  The simulation is for a calendar year and, in the absence of
atrazine exposures, the model shows seasonal trends for the populations
of different taxa that made the model appropriate for assessing stressor
effects on community dynamics (Volz et al. 2007).

Incorporation of Toxicity Information into the Model

One challenge in applying a bioenergetics-based community model such as
CASM to toxicity assessments is that toxicity tests do not directly
address the effects of the chemical on the specific parameters within
the bioenergetics state equation (e.g., Equation 1).  For example, a
plant growth test will provide information on the net growth rate, but
not on how specific parameters within the photosynthesis and respiration
functions are being affected.  There is just not enough information in
the toxicity tests to independently adjust all the model parameters.

	

To address this problem, CASM uses the bioenergetics equations to
simulate the concentration/effects curve of toxicity tests, and
estimates a "toxic effects factor" (TEF) for each concentration that,
when used to adjust several model parameters, will reproduce the effect
level at that concentration (DeAngelis et al. 1989; Bartell et al. 1999,
2000; Bartell 2003; Volz et al. 2007).  For plants, the version of CASM
developed for this atrazine assessment offers two alternatives regarding
what model parameters are altered by the TEF.  For the "General Stress
Syndrome" (GSS), the TEF is used to (a) reduce photosynthesis by
decreasing the maximum photosynthesis and increasing light and nutrient
saturation parameters and (b) increase the various loss terms, such as
respiration, in the bioenergetics equation.  For the "Photosynthesis
Stress Syndrome" (PSS), only the maximum photosynthesis parameter is
modified.  

To model the response of each plant species, CASM therefore requires the
specification of a function for the fraction reduction in plant growth
versus atrazine concentration over a specified exposure duration.  The
assumed functional shape of percent growth inhibition versus
log10(concentration) is the cumulative Gaussian distribution, so that
the probit-transform of the percent growth reduction versus
log10(concentration) is linear.  This function can be specified by
inputting either (1) the slope and intercept of this linear transform or
(2) an exposure concentration showint 50% effect (EC50) and No Observed
Adverse Effect (NOEC) which defines the function under an assumption
that the NOEC is five probit units below the EC50.  For the exposure
concentration on each day of a simulation, CASM estimates a TEF that
causes the bioenergetics equation to produce the growth reduction
specified by the toxicity function over the specified test duration
under optimal nutrient, light, and temperature conditions. 

To provide the needed plant toxicity information, individual plant
species toxicity tests collated by Giddings et al. (2000) and identified
from other sources were reviewed. A subset of these were identified for
use in model parameterization based on a suitably-defined growth EC50
over a specified duration (  REF _Ref180376742 \h  Table II-1 ).  Next,
the geometric mean EC50s for each species were computed, and assigned to
the various plant species within CASM based on bioenergetics similarity
of the test species to the modeled species (  REF _Ref180376742 \h 
Table II-1 ).  The probit versus log10(concentration) slope for each
species was set to 3.3 based on the mean of the slopes reported in the
U.S.EPA Office of Pesticide Programs one-liner database (Montague 1998).
 This slope and the mean EC50s for each species were used to calculate
the probit line intercepts for input into CASM. 

The assignment of EC50s to the modelled plant species is a possible
major source of uncertainty given the variability of results within
species (  REF _Ref180376742 \h  Table II-1 ).  The EC50s in Table II-1
were subject to an analysis of variance to compare between-species and
within-species variability.  For log10-transformed EC50s, this analysis
indicated an overall mean of 2.01 (corresponding to an untransformed
EC50 of 103), a within species mean squared error of 0.481 (df=45), a
between species mean squared error of 0.486 (df=36), and a
non-significant F-statistic (1.02). This lack of a demonstrated 
difference between the within-species and between-species variability
indicates a source of variability which makes any single EC50 for a
species subject to considerable uncertainty.  Additionally, it may also
make mean EC50s based on a large number of tests subject to some amount
of uncertainty in any particular situation.  This is not to say that
sensitivity to atrazine does not vary among species, but rather that
various sources of variability make the selection of specific EC50s
uncertain.  Alternative sets of plant EC50s should therefore be examined
(see Section   REF _Ref180376659 \w \h  II.D ).        

Toxicity values were also assigned to the animal species (Volz et al.
2007).  Because the lowest EC50 assigned to an animal species was much
greater than the EC50s for most of the plant species, substantial
effects on the plant community would occur at concentrations below any
toxic effects on animals in these simulations.  Thus, assessment results
are not sensitive to the animal toxicity value selection, which will not
be further discussed or evaluated here.        

Table   STYLEREF 1 \s  II -  SEQ Table \* ARABIC \s 1  1  Atrazine plant
toxicity tests used in assessment methodology.

Species	EC50 (ppb)	Dura-tion (d)	Reference	Species Geometric Mean EC50
CASM

Population 

Assignment

GREEN ALGAE

Ankistrodesmus braunii	60	11	Burrell et al. 1985	61	#5 – Phytoplankton

#15 - Periphyton

	61	24	Larsen et al. 1986



Chlamydomonas nostigana	330	3	Kallqvist and Romstad 1994	330

	Chlamydomonas reinhardi	10	10	Schafer et al. 1993	59	#4 - Phytoplankton

	19	1	Larsen et al. 1986



	176	4	Fairchild et al. 1998



	350	3	Schafer et al. 1994



Chlorella fusca	15	1	Faust et al. 1993	15

	Chlorella pyrenoidosa	125	1	Stratton and Giles 1990	280



175	5	Gramlich and Frans 1964



	282	5	Montague 1998



	1000	14	Stratton 1984



Chlorella vulgaris	94	4	Fairchild et al. 1998	88	#6 – Phytoplankton

#20 - Periphyton

	25	11	Burrell et al. 1985



	293	1	Larsen et al. 1986



Scenedesmus obliquus	38	1	Larsen et al. 1986	38	#3 – Phytoplankton

#14 - Periphyton

Scenedesmus quadricauda	169	4	Fairchild et al. 1998	169

	Scenedesmus subspicatus	21	4	Kirby and Sheahan 1994	76



72	3	Schafer et al. 1994



	110	4	Geyer et al. 1985



	200	3	Zagorc-Koncan 1996



Selenastrum capricornutum	4	4	University of Mississippi 1991	84	#16 -
Periphyton

	26	4	Caux et al. 1996



	50	4	Versteeg 1990



	53	5	Montague 1998



	55	5	Hoberg 1993c



	70	1	Turbak et al. 1986



	95	5	Roberts et al. 1990



	117	4	Fairchild et al. 1998



	118	3	Radetski et al. 1995



	128	4	Gala and Giesy 1990



	130	4	Hoberg 1993b



	158	3	Mayer et al. 1998



	200	3	Kallqvist and Romstad 1994



	214	7	Abou-Waly et al. 1991



	235	4	Fairchild et al. 1997, 1999



	359	3	van der Heever and Grobbelaar 1996



	42	1	Larsen et al. 1986



CRYPTOMONADS

Cryptomonas pyrinoidifera	500	6	Kallqvist and Romstad 1994	500	#17 -
Periphyton

BLUEGREEN ALGAE

Anabaena cylindrica	178	1	Stratton 1984	178	#19 - Periphyton

Anabaena flos-aquae	58	3	Abou-Waly et al. 1991	342	#10 - Phytoplankton

	230	5	Hughes et al. 1988



	3000	4	Fairchild et al. 1998



Anabaena inaequalis	100	14	Stratton 1984	100	#8 - Phytoplankton

Microcystis aeruginosa	129	5	Parrish 1978; Montague 1998	285



630	6	Kallqvist and Romstad 1994



Microcystis sp.	90	4	Fairchild et al. 1998	90	#7 – Phytoplankton

#18 - Periphyton

Synechococcus leopoliensis	130	3	Kallqvist and Romstad 1994	130	#9 -
Phytoplankton

DIATOMS

Cyclotella sp.	430	6	Kallqvist and Romstad 1994	430	#13 - Periphyton

Navicula pelliculosa	60	5	Hughes et al. 1988	60	#1 – Phytoplankton

#11 - Periphyton

Skeletonema costatum	265	2	Walsh 1983	69	#2 – Phytoplanton

#12 - Periphyton

	260	2	Mayer 1987



	50	2	Walsh et al. 1988



	24	5	Parrish 1978; Montague 1998



	55	5	Hoberg 1993c



	24	4	Montague 1998



MACROPHYTES

Ceratophyllum demersum	22	14	Fairchild et al. 1998	22	#23 - Macrophyte

Elodea canadensis	1200	10	University of Mississippi 1991	159	#24 -
Macrophyte

	21	14	Fairchild et al. 1998



Hydrilla verticillata	110	14	Hinman 1989	110

	Lemna gibba	22	14	Hoberg 1993a	159

(Genus Mean)	#25 – Macrophyte

         

	180	7	Hoberg 1991



	170	5	Montague 1998



Lemna minor	8700	14	University of Mississippi 1991



	92	4	Fairchild et al. 1998



	56	10	Kirby and Sheehan 1994



	153	4	Fairchild et al. 1997



Myriophyllum heterophyllum	132	14	Fairchild et al. 1998	132	#22 -
Macrophyte

Najas sp.	24	14	Fairchild et al. 1998	24

	Thallassia testudinum	320	1.7	University of Mississippi 1991	320

	Potamogeton perfoliatus	474	21	Forney and Davis 1981	170	#26 -
Macrophyte

	130	28	Kemp et al. 1985



	80	0.1	Jones et al. 1986



Vallisneria americana	163	42	Forney and Davis 1981	163	#21 - Macrophyte



Model Implementation and Example Application

	

Model Effects Index Selection

The basic output of CASM is a time series of daily biomass values for
each modeled species over the simulation year.  Various possibilities
exist for synthesizing this information into a model effects index for
the primary producer community.  The index might address how just total
plant biomass or production is perturbed in an exposed system compared
to an unexposed system, or might also reflect perturbations in the
relative biomasses of each species.  The index might indicate the
maximum perturbation across the entire simulation, the average
perturbation over the entire simulation, an average perturbation during
a fixed time window during the simulation, or a maximum running average
of the perturbations.  

	

The model effects index selected is based on the average, over the
entire simulation, of the daily Steinhaus Similarity Indices (SSI).  The
SSI quantifies the similarity between two communities (in this case,
exposed and control model-calculated communities).  For each day of the
simulation, the SSI is calculated as:

 

where BR,i is the daily biomass of the ith species in the reference
(control) simulation and BE,i is the daily biomass of the ith species in
the atrazine-exposed simulation.  SSI values thus reflect changes in
both absolute and relative population sizes, ranging from 1.0 for
identical community structures to 0.0 for completely disjoint
communities.  The model effects index will be expressed as the percent
reduction of this average SSI from the maximum value of 1.0 (i.e., a
model effects index of 10 represents an average SSI of 0.90).  

This type of measure was selected in order to include changes in
relative species abundance as part of the risk assessment, rather than
just some measure of total primary productivity.  The entire simulation
period was chosen as the averaging period to provide a consistent
measure appropriate to a variety of exposure time series shapes. For
example, two exposure time series might have the same initial atrazine
peak that produces the same maximum perturbation of the SSI, but one of
the time series might have additional peaks that cause more effects on
average than the other time series.  Although a year-long average for
the model effects index will be reduced by including pre-exposure
periods in which both the control and exposed communities are identical
(daily SSI=1.0), this will not alter risk assessments significantly
because both the LOC and model effects indices for systems of interest
will be reduced to similar degrees.

Although the rationale for choosing the average SSI over the entire
simulation as the model effects index has been described above, it is
not certain a priori how well this choice will support discrimination of
micro/mesocosm effects scores or their extrapolation to field exposures.
 Therefore, the sensitivity analysis below will address what level of
uncertainty in risk assessments might exist because of other possible
choices for the model effects index.

LOC Determination for Base Model Configuration

A "base model configuration" was selected for conducting assessments
that was (a) the Midwestern U.S. stream community structure discussed
above and described in Volz et al. 2007, (b) the Ohio stream
physicochemical parameters also discussed above and in Volz et al. 2007,
(c) the use of the GSS for toxicity calculations, and (d) the toxicity
relationships summarized in   REF _Ref180376742 \h  Table II-1 .  In
addition, for model simulations of the exposures used in the
micro/mesocosm data base, exposures were started on Julian day 105,
selected as a typical average starting date for atrazine applications in
the corn belt.

Figure   STYLEREF 1 \s  II -  SEQ Figure \* ARABIC \s 1  3  LOC=4.0
based on correlation of micro/mesocosm effect scores with model effect
index values (upper panel), and comparison of model LOC to micro/
mesocosm effect scores at different exposure concentration/ durations
(lower panel)

Simulations were conducted without atrazine exposure (control) and with
each atrazine exposure concentration/duration from the micro/mesocosm
database, and used to calculate a year-long average SSI for each
atrazine exposure.  For each micro/mesocosm exposure, the upper panel of
  REF _Ref180377373 \h  \* MERGEFORMAT  Figure II-3  compares the
micro/mesocosm effects score to the model effect index (percent
reduction in the average modeled SSI from its maximum value of 1.0). 
The LOC for the model effects index for this consultation was selected
simply by determining the midpoint of the range of values for which
false negatives (i.e., micro/mesocosm scores of 3-5 for exposures with a
model effect index below the LOC) equaled false positives (i.e.,
micro/mesocosm scores of 1-2 for exposures with a model effect index
above the LOC). In a similar fashion, the LOC could also be identified
in order to minimize the number of false negatives.   REF _Ref180377373
\h  \* MERGEFORMAT  Figure II-3  illustrates the LOC determination of
4.0, with seven false positives and seven false negatives, all of which
are within a factor of 3 of the LOC.

The lower panel of   REF _Ref180377373 \h  \* MERGEFORMAT  Figure II-3 
illustrates the discrimination of the micro/mesocosm 1-2 and 3-5 score
groups with the model LOC=4.0. This figure shows the
concentration/duration dependence of micro/mesocosm scores and includes
a model-calculated line showing the concentration/duration dependence of
the LOC. It also shows how close the false negatives and false positives
are to the LOC on an exposure concentration scale, differing again by no
more than a factor of 3.  Better performance than this would not be
expected given the overlap between the score groups.  The model does
show more time dependence than is evident in the micro/mesocosm data,
because the model effects index reflects effects over an entire season,
and not just during the exposure period.  This results in some tendency
for the false positives to be at longer durations than the false
negatives.  

The base model described here differs somewhat from that described in
the October 2003 IRED addendum (US EPA, 2003b) or by Volz et al (2007),
most notably that the LOC is 4.0 rather than 5. Several changes are
responsible for this change in the LOC.  Most importantly, the model
effects index previously had only been averaged from Julian Day 105 to
the end of the year, but is now averaged over the entire year to include
any effects before this date.  Because the early parts of the year
typically have low enough exposures to have no effects, this longer
averaging period will include significant stretches of time when the SSI
deviation is zero, thus reducing the average deviation (by a factor of
1.4).  This has little impact on actual assessments because the model
effects index for the field chemographs of interest will be reduced by
the same factor as the LOC (i.e., this change will result in a greater
likelihood of exceeding the LOC only for those field chemographs with
early exposures).  To a lesser degree, the LOC has also been affected by
other changes from the earlier version, including improvements in how
toxicity test information is translated into bioenergetics equations,
applying the results of just one simulation using the best estimate for
toxicity factors rather than the median results of multiple simulations
based on random selections from a distribution of toxicity factors, and
direct simulations of each exposure in the micro/mesocosm dataset rather
than interpolations from a matrix of different exposure/duration
combinations.  Again, however, higher or lower values for the LOC are
largely compensated for by higher or lower effects in the simulations
for the field chemographs of interest.  The previous CASM implementation
also differed in having average screening concentrations that needed to
be exceeded before CASM was even run; the purpose of this was to avoid
unnecessary use of the substantial computing resources this earlier CASM
implementation required. In the current version, this is no longer an
issue due to faster computing resources and the elimination of multiple
simulations . 

    

Example Applications

Once an LOC is determined based on comparing micro/mesocosm effect
scores with the model effects indices, application of the model simply
involves conducting a simulation for an exposure time series of interest
and determining whether the model effects index from this simulation
exceeds the LOC.  The upper panel of   REF _Ref181757713 \h  \*
MERGEFORMAT  Figure II-4  shows the model effects indices for sixteen
such simulations, using the example chemographs from   REF _Ref180376001
\h  Figure II-1 .  Only two of the chemographs exceeded the LOC, but a
few others have model effects index values close to the LOC.

Figure   STYLEREF 1 \s  II -  SEQ Figure \* ARABIC \s 1  4  Model effect
index values and multiplication factors for sixteen sample chemographs,
using base model configuration

In addition to simply determining whether the model effect index exceeds
the LOC, the model can also evaluate the multiplication factor by which
the exposure time-series must be changed so that the model effects index
equals the LOC. If the LOC is exceeded, this factor indicates the
reduction in exposure necessary to be below the LOC. If the LOC is not
exceeded, this factor indicates the amount exposure could be increased
and still not exceed the LOC.  Such information can benefit remediation
efforts, evaluations of "margins of safety", and assessments of possible
risks from future activities and conditions.  The lower panel of   REF
_Ref181757713 \h  \* MERGEFORMAT  Figure II-4  provides the
multiplication factors for the sixteen example chemographs.  The
multiplication factors for the two chemographs that exceeded the LOC
trigger are less than 1, indicating the amount of reduction needed to
reduce effects to the LOC.  The other multiplication factors range from
barely over 1 to about 5, indicating the margin of safety present in
these exposure time series.

Sensitivity Analysis

	

Overview

The proposed methodology depends on whether the predicted relative
effects in different exposure time-series from a single model
configuration are useful approximations for extrapolations among a range
of experimental and natural ecosystems.  Although suitable data from
experimental and natural ecosystems are not available to directly
validate the utility of such extrapolations, this methodology can be
validated to some degree by evaluating how sensitive results compare to
alternative model configurations.  Such a sensitivity analysis has three
potential benefits.  First, if relative effects are sufficiently similar
across a range of possible model configurations, the need to
independently justify specific options for model configuration is
reduced.  Second, a lack of sensitivity of results to a range of
possibilities in a simulated system increases confidence that
extrapolations among natural systems also would not be highly sensitive
to system properties.  Third, a sensitivity analysis will provide some
quantitative information on certain sources of uncertainty, which can
inform risk management decisions for assessments with exposures near the
LOC.  

A variety of decisions went into formulating the base model
configuration used in Section   REF _Ref180377931 \w \h  II.C  and
warrant consideration in this sensitivity analysis.  The community
structure and bioenergetics equations of the base model configuration
were intended to describe a second- to third-order Midwestern stream. 
Alternative community structures currently under development will be
part of an expanded sensitivity analysis in the future and thus are not
considered here.  Areas of method development that are part of the
sensitivity analysis include (a) the selection of the model effects
index, (b) the start date for model simulations of the micro/mesocosm
exposures, (c) the environmental driving variables (nutrients,
temperature, and light), and (d) the EC50 selection.  A comparison was
also made regarding use of the GSS versus the PSS, but no significant
differences in results were found.

Because the LOC is dependent on the model configuration and
parameterization, the LOC for each model configuration addressed in the
sensitivity analysis must first be determined.  Each model configuration
requires new simulations of an unexposed system and of each exposure
concentration/duration in the micro/mesocosm data set, after which a new
LOC is selected for discriminating the micro/mesocosm scores with this
model configuration.   Simulations are then done again for the 16
example time series in   REF _Ref180376001 \h  Figure II-1 , determining
the model effects index to compare to the LOC and also the
multiplication factors that specify how much the exposure must be
changed to equal the LOC.  Because each model configuration has its own
LOC, the absolute model effects index values cannot be compared across
model configurations to assess their similarity.  However, the
multiplication factors for each configuration can be compared, because
they each address changing the exposure to equal the respective LOCs,
and thus address how similar the risk assessments are for each
configuration.  Therefore, the basic criterion for assessing similarity
in the subsequent sections will be to compare the multiplication factors
for the alternative model configurations to the base model
configuration.  

Model Effects Index

As already noted, alternatives to the selected model effects index (the
percent reduction in the annual average of the daily SSI values from the
maximum value of 1.0) could include (a) a measure of total plant
community biomass/production, not considering relative changes in
different plant species, and (b) a different averaging period.  For this
sensitivity analysis, two alternatives were examined.  One alternative
(AVP) used the difference between the total plant biomass in the control
and exposure simulations, but still averaging over the entire
simulation.  The other alternative (MXS) still used the daily SSI
values, but used the maximum percentage deviation from 1.0 anytime
during the simulation as the effects index, to contrast the smallest
averaging period (one day) with the average over the entire simulation.

Figure   STYLEREF 1 \s  II -  SEQ Figure \* ARABIC \s 1  5  Comparison
of multiplication factors for the base model configuration to
alternatives with different model effect indices

 

The upper panel of   REF _Ref180378699 \h  \* MERGEFORMAT  Figure II-5 
contrasts the multiplication factors for these two alternative effects
indices with the effects index for the base model configuration (solid
line).  To better show the magnitude of deviations, the lower panel of  
REF _Ref180378699 \h  \* MERGEFORMAT  Figure II-5  show the relative
multiplication factors (the multiplication for the alternatives divided
by that for the base model configuration) on a finer scale, with
horizontal lines denoting deviations of a factor of 1.2, 1.5, and 2.0.

Although the AVP endpoint had a markedly lower LOC (2.70) and the MXS
endpoint a markedly higher LOC (7.45), than the base case (4.00), the
multiplication factors for these different model configurations are all
approximately the same.  This is because a change in the LOC arises from
an alternative model configuration producing generally higher or lower
effects than the base case for all exposures, so that a higher LOC is
being compared to a similarly higher effects index for any simulation,
and a lower LOC is being compared to lower model effects indices. 
Therefore, there is little change in the multiplication factors.  In  
REF _Ref180378699 \h  \* MERGEFORMAT  Figure II-5 , alternative model
effects indices indicate exposures of concern within 20% of that of the
base model effects index for all the example exposures, with average
deviations of less than 10%.  This close agreement indicates that the
choice of model effects index does not result in any substantial changes
to risk assessments using this methodology, so that the index selected
for the base model has general applicability within reasonable
uncertainty.  

Exposure Start Date

The simulations of the exposures of the micro/mesocosm data set to
determine an LOC for the model effects index assume a start date for the
model exposure on Julian day 105 (April 15) for the base model
configuration.  In natural systems, the first significant atrazine
exposure will vary depending on application practices and rainfall, and
so different options for this start date are of obvious interest.  Start
dates 15 days before and after the base choice were therefore tested. 
This shift only affects simulations of the constant exposure/fixed
duration combinations used in the LOC determination, and does not affect
the example field time series to which the model is then applied.  Thus
any change in LOC will lower or raise the multiplication factor by a
fairly uniform amount.

Figure   STYLEREF 1 \s  II -  SEQ Figure \* ARABIC \s 1  6  Comparison
of multiplication factors for base model configuration to alternative
configurations with different start dates

  REF _Ref180380098 \h  \* MERGEFORMAT  Figure II-6  shows the results
for these alternative start dates.  An earlier start slightly increases
model effect index values, leading to a higher LOC (4.25).  Because
model simulations for the example time series are not affected by the
changed start date, multiplication factors are slightly elevated, but by
less than 10%.  In contrast, a later start slightly decreases model
effect index values, lead to a lower LOC (3.60) and slightly lower
multiplication factors (range from 10-18% lower (average 14%).  As was
true for the choice of the model effects index, these limited effects
indicate little sensitivity of results to the choice of start date and
good applicability of the start date choice for the base model
configuration. 

Environmental Parameters

Figure   STYLEREF 1 \s  II -  SEQ Figure \* ARABIC \s 1  7  Comparison
of multiplication factors for the base model configuration to
alternative configurations with lower and higher nutrients.

The base model configuration employs environmental data (nutrients,
light, temperature) based on an Ohio stream (Volz et al. 2007).    REF
_Ref180380641 \h  \* MERGEFORMAT  Figure II-7  shows the effect of
altering nutrients (nitrogen, phosphorus, silicon) up or down, together,
by a factor of 2.    REF _Ref180380907 \h  Figure II-8  shows effects of
changing temperature up or down by 5o C.    REF _Ref180381121 \h  Figure
II-9  shows the effects of halving or doubling light intensity.  

For all these variables, the multiplication factors were never more than
20% different from those for the base model configuration, and the mean
deviation was less than 10%.  As for the model effects index and the
starting date, risk assessments using this methodology are insensitive
to the choices for these environmental variables.

Plant Toxicity EC50s

As discussed in Section   REF _Ref180833304 \w \h  II.B.2 , there is
considerable uncertainty in the assignment of EC50s to the modeled plant
species, especially because of the large variation of values within a
species.  Based on the analysis of variance of the plant toxicity data
discussed in Section   REF _Ref180833318 \r \h  B.2 , ten alternative
sets of plant EC50s were randomly selected from a log-normal
distribution with median 100 ug/L and log standard deviation of 0.4 ( 
REF _Ref180381356 \h  Table II-2 ).  The methodology was applied using
each of these toxicity data sets, and the mean and standard deviation of
the log multiplication factors was computed for each sample chemograph. 


Figure   STYLEREF 1 \s  II -  SEQ Figure \* ARABIC \s 1  8  Comparison
of multiplication factors for the base model configuration to
alternative configurations with lower and higher temperature

Figure   STYLEREF 1 \s  II -  SEQ Figure \* ARABIC \s 1  9  Comparison
of multiplication factors for the base model configuration to
alternative configurations with lower and higher light.

  REF _Ref180381585 \h  Figure II-10  compares the mean and standard
deviation of the multiplication factors for the alternative plant EC50
sets to the multiplication factors for the base model configuration. 
Compared to the other factors in this sensitivity analysis, this
variation of EC50s resulted in greater deviations of multiplication
factors from that of the base case, and thus represents the greatest
source of uncertainty.  However, the base model configuration is still
within one standard deviation of the mean of the alternative toxicity
data sets and within 20% of the mean except for chemograph #1. 
Furthermore, even if the range of the alternative toxicity values were
extended to two standard deviations, this differs by no more than a
factor of two, except for the upper tail of chemograph #1, for which the
base configuration is more conservative.  This justifies the use of the
base model configuration as being within this uncertainty of toxicity
values, and documents the magnitude of an important uncertainty.       

Table   STYLEREF 1 \s  II -  SEQ Table \* ARABIC \s 1  2  Ten sets of
randomly selected EC50s (ug/L atrazine) for sensitivity analysis.

Mode Species	Set 1	Set 2	Set 3	Set 4	Set 5	Set 6	Set 7	Set 8	Set 9	Set
10

Phytoplankton 1	48	241	34	200	225	179	109	102	1196	292

Phytoplankton 2	125	96	89	65	84	112	11	42	150	160

Phytoplankton 3	54	42	65	55	168	167	641	63	162	89

Phytoplankton 4	149	185	438	41	122	148	31	164	351	788

Phytoplankton 5	198	53	241	56	13	283	67	22	133	324

Phytoplankton 6	66	23	44	121	325	384	208	205	90	12

Phytoplankton 7	1077	282	411	116	107	16	148	78	31	167

Phytoplankton 8	37	39	141	410	154	109	131	220	76	836

Phytoplankton 9	142	16	137	252	29	203	176	108	528	58

Phytoplankton 10	365	64	102	58	188	234	248	62	94	130

Periphyton 1	62	561	57	14	164	111	74	25	94	63

Periphyton 2	24	336	112	113	66	82	231	106	26	23

Periphyton 3	204	204	102	125	819	107	140	112	388	51

Periphyton 4	73	121	74	80	329	158	112	245	124	128

Periphyton 5	41	341	31	23	48	47	97	187	63	241

Periphyton 6	289	151	59	143	94	161	207	326	30	50

Periphyton 7	163	58	29	232	420	83	55	33	160	97

Periphyton 8	47	156	300	87	220	128	67	308	69	84

Periphyton 9	187	142	68	140	118	16	96	152	128	107

Periphyton 10	317	54	20	84	59	99	200	852	53	823

Macrophyte 1	269	62	285	407	86	105	85	136	44	90

Macrophyte 2	70	246	101	92	59	19	51	30	86	334

Macrophyte 3	288	50	61	51	214	116	76	123	27	677

Macrophyte 4	88	6	46	477	192	51	50	55	130	59

Macrophyte 5	55	55	30	38	136	58	485	230	248	16

Macrophyte 6	47	419	708	81	47	49	129	416	176	116



Figure   STYLEREF 1 \s  II -  SEQ Figure \* ARABIC \s 1  10  Comparison
of multiplication factors for the base model configuration (bold solid
line) to the mean (bold dashed line) and mean +/- one standard deviation
(narrow dashed lines) of the multiplication factors for ten alternative
plant EC50 sets.

Chronic, Low-Level Exposures

Although this sensitivity analysis was oriented to choices for model
configuration and inputs, it required testing of a broad set of
representative chemographs (Figure II-1) to define a universe of
exposures to which conclusions about sensitivity apply.  These
chemographs all lacked any significant atrazine exposure during the
first few months of the year, but other recent model simulations with
extended exposures of a few ug/L during this period indicated
significant model responses which could have significance to risk
assessments.  Further evaluations of these responses have identified
possible problems in CASM regarding (a) translation of the lower tails
of the toxicity dose/response curves into the bioenergetics equations
and (b) the sensitivity of some components of the modeled community to
small, prolonged effects during the first few months of the year.  These
issues are currently being addressed, and will entail the development of
alternative model configurations which more appropriately address low
concentration and early-year exposures.  Once these changes are made,
they will support revisions to the analyses reported in this paper.

Slope of the Toxicity Curve 

The current CASM_Atrazine configuration used a default probit slope of
3.31. This sensitivity analysis did not assess the performance of the
model with varying the slope of the toxicity curve or the EC50 within
the bounds of uncertainty. The slope of the toxicity curve might impact
whether certain types of chemographs – in particular, short-term, high
concentration exposures characteristic of Midwestern streams – exceed
the LOC trigger. While the Agency plans to include this in future
sensitivity analyses, it is also seeking SAP input on this issue. 

Summary and Next Steps

The methodology described here addresses the need to assess the likely
ecological effects of diverse, highly time-variable atrazine exposures. 
This was accomplished by using a community simulation model to estimate
the relative effects of different exposure time series, with the
absolute level of concern for model-predicted effects based on
correlations with effects observed in microcosm and mesocosm studies. 
The feasibility of this methodology has been established for
well-defined exposure time-series and the model-defined level of concern
has been demonstrated to discriminate well between acceptable and
unacceptable effects in the microcosm and mesocosm studies.  A
sensitivity analysis demonstrated that results are similar across a wide
range of options for model formulation, parameterization, and inputs. 
Additional work in progress will expand the sensitivity analysis,
further refine estimates for uncertainty associated with the base model
configuration, address further changes to CASM algorithms for
incorporating toxicity information, and quality assure the software
being used. 

SAP Charge Questions on the Use of the CASM Model

(1) 	Please comment on the use of a community simulation model for
assessing the relative effects of different exposure time series. Please
provide any recommendations for a community response model other than,
or along with, CASM that could be used for assessing the effects of
atrazine. What are the strengths and weaknesses associated with the
other model(s). Please comment on approaches that do not require an
aquatic community response model and discuss the advantages and
disadvantages of any alternative non-modeling approaches for
extrapolating the effects seen in micro/mesocosm data to the effects
resulting from field exposure.

(2) 	The general methodology employed in this analysis consists of (a)
correlating model outputs to micro/mesocosm data to determine a model
LOC and (b) applying the model to chemographs of interest to determine
whether the LOC is exceeded.  Please comment on the scientific strengths
and limitations of this approach. 

(3) 	Please comment on the reasonableness of the general CASM_Atrazine
model formulation and parameterization, and the various options selected
for the base model configuration.

(4) 	Please comment on whether the described sensitivity analyses are
suitable for characterizing uncertainties associated with the choice of
options for configuration of the base model and the input variables. 
What additional sources of uncertainty alternatives should be examined
in this analysis?  Please comment on whether the sensitivity of results
to the slope of the toxicity curve, as well as the EC50, should be
examined to address possible effects on responses to short pulses.

(5)	During its review of the CASM_Atrazine model, the Agency found that
the model appears to overestimate the effects of low, chronic
concentrations possibly due to the way the model simulates population
levels and decline of macrophytes early in the year.

The Agency sees two approaches for addressing this issue: (1) exclude
early season atrazine exposures from the chemograph inputs, or (2)
modify the model to better account for the impacts of early-season
exposures. Please comment on the strengths and weaknesses of the
Agency’s approaches and provide recommendations for any alternatives.

Given that the Agency identified this issue during the exposure
evaluation, please provide recommendations on additional steps the
Agency could take for quality assurance for the model and methodology.

 Atrazine Monitoring Program For Ecological Effects: Determining The
Extent Of Waters Exceeding Effects-Based Thresholds For Atrazine

Over the years, atrazine surface water monitoring data have been
collected by numerous programs and sources, including the U.S.
Geological Survey (Goolsby and Battaglin, 2003; USGS, 2006b; USGS,
2007), registrants, states, and universities (see Williams et al, 2004b
for a listing and description of atrazine monitoring studies). However,
these monitoring studies were conducted for different purposes and were
not intended to evaluate the extent to which atrazine concentrations in
water bodies would exceed the LOC or to identify the areas/conditions
under which the LOCs might be exceeded.

This section describes the objectives of the monitoring study, the
watershed selection process, and the design and conduct of the
monitoring study. The subsequent analysis of the monitoring results
includes an evaluation of whether exposure exceeds the LOC, sensitivity
analyses of the impacts of sampling and environmental factors, and
extrapolation of the results for the monitoring sites to the larger
population of vulnerable watersheds. The Agency is soliciting feedback
from the SAP related to the interpretation of the results and
sampling/study design issues critical to the application of this
approach for assessing potential atrazine impacts to other water bodies.
 

Monitoring Study Objectives

The purpose of the monitoring program in flowing waters is to measure
the magnitude and extent of atrazine concentrations in water bodies with
the greatest potential vulnerability to atrazine exposure from use in
corn and sorghum production (based on atrazine use and runoff potential)
and to estimate percentage that exceed the LOC. The study focused on
water bodies in the most vulnerable watersheds because this is where
elevated concentrations of atrazine were most likely to occur,
increasing the likelihood of identifying water bodies that might exceed
the ecological thresholds of concern. If no exceedances occurred in
these watersheds, the likelihood of finding atrazine exposures of
concern would be lower in less vulnerable watersheds. 

The 2003 addendum to the Atrazine IRED specified that the monitoring
program initially focus on flowing waters associated with corn and
sorghum production (US EPA, 2003b). A separate pilot monitoring study is
underway for water bodies in sugarcane areas. Additionally, the Agency
is evaluating existing monitoring of reservoirs to determine the
potential impact of atrazine on aquatic communities in static water
bodies. The sugarcane and reservoir efforts are not the subject of this
SAP. 

The atrazine monitoring program was intended to identify watersheds in
corn and sorghum areas that contained water bodies exceeding the
effects-based LOC trigger described in Section II. The monitoring
program also gathered additional information on atrazine use and
practices, watershed and water body characteristics, and other factors
to facilitate identifying water bodies beyond the initial sampling pool
that have the potential for atrazine loadings that exceed effects-based
thresholds. 

The atrazine ecological monitoring program was designed to:

Determine the extent to which waters in vulnerable watersheds exceed
effects-based thresholds for atrazine, and

Collect information that will help identify where the waters that exceed
the effects-based atrazine thresholds occur.

This section focuses on the first objective. Section   REF _Ref180988777
\r \h  IV  describes the approach the USEPA proposes to take for
addressing the second objective.

Study Design

The study design relied on the wealth of information already known about
the fate and transport properties of atrazine, its use, and its history
of occurrence in water. The following sections briefly describe the site
selection process, from identifying vulnerable watersheds to selecting
monitoring sites, and the sampling/monitoring methods. Williams et al
(2004a) provide details of the study design.

Site Selection

The site selection process used existing knowledge about atrazine
occurrence in water to maximize the potential for finding waters that
exceed the LOC. Key factors were:

Use: amount and intensity of atrazine use on corn and sorghum

Vulnerability based on use/runoff: integrating both atrazine use and
surface runoff vulnerability

Representative sampling: 40 sites selected from most vulnerable
watersheds, with probability of selection proportional to use

This section summarizes the steps used to identify those watersheds that
are expected to be most potentially vulnerable to atrazine loading. From
this upper tier of vulnerable watersheds, a spatially-balanced survey
design was used to select a representative sample of 40 watersheds.
Sub-watershed land use and flow data were used to identify candidate
stream segments for monitoring within those watersheds.  Details and
documentation are available in Williams et al (2004a). 

Watershed Vulnerability Assessment 

The monitoring program used a watershed-based approach to target areas
for monitoring. Because of the widespread use of atrazine on corn and
sorghum, the initial approach focused on nationally-available data. 
Watershed, or hydrologic unit, boundaries define the extent of surface
water drainage to a stream point and provide a framework for organizing
and processing spatial data pertinent to the vulnerability of watersheds
and receiving water bodies to atrazine loading. 

Hydrologic unit maps defined by the USGS (Seaber et al, 1987; USGS,
2006a) provide a hierarchical system of mapping watersheds. This system
includes six levels, with national maps available for the first four
levels at the time the study was being designed. The assessment used the
fifth, HUC-10 or HUC-11, level, representing watersheds that are
typically 40,000 to 250,000 acres in size. Because mapping at this level
was not available for all states, Syngenta collected the best available
coverages for the 37-state area that encompassed the extent of
significant atrazine use on corn and sorghum (Williams et al, 2004a,
2004b).

HUC-10/11 watershed boundaries were used to analyze a number of spatial
data layers that might serve as indicators of the potential for
pesticide (i.e., atrazine) movement from the fields of application to
water bodies. Williams et al. (2004a, 2004b) provide details of those
watershed vulnerability analyses and the spatial data sources used for
the analysis, as well as documentation of additional data manipulation.

Syngenta evaluated a number of potential factors affecting pesticide
runoff and watershed vulnerability, including atrazine use intensity,
land use types, flow/drainage under row crops, precipitation, and soil
characteristics (Williams et al., 2004a). At the same time, the US EPA
looked at the potential for using the Watershed Regression for
Pesticides (WARP) model, developed by the USGS for atrazine (Larson et
al, 2004), as a tool for identifying the watersheds potentially most
vulnerable to atrazine loading in flowing water bodies.

The WARP model integrates use intensity, watershed area, soil
susceptibility to runoff and rainfall intensity with available water
monitoring data (Larson et al., 2004). WARP estimates various
percentiles of the annual distribution of atrazine concentrations using
separate regression equations based on the following parameters: 

Atrazine use intensity: amount of atrazine applied per watershed area

Area: Total watershed area

Rainfall intensity: R factor from USDA/NRCS

Soil erodibility: K factor from USDA/NRCS

Dunne Overland Flow: fraction of runoff from saturated soils

The atrazine watershed vulnerability assessment is based on
th⁥敲牧獥楳湯攠畱瑡潩⁮桴瑡攠瑳浩瑡摥琠敨㤠琵⁨
数捲湥楴敬映潲⁭湡愠湮慵⁬楤瑳楲畢楴湯漠⁦瑡慲楺
敮挠湯散瑮慲楴湯㩳ഠ

WARP = 10^ (− 4.60 + (0.67U0.25 )+ (1.12LogR)+ (3.59K)+ (0.0006A0.5
)−(0.11*D))

where, 

U = watershed cropped area use intensity in kg/km2

R = the USLE rainfall erosivity factor

K = the USLE soil erodibility factor

A = watershed area (km2)

D = Dunne overland flow 

The WARP model calculations required collecting and re-aggregating
values for atrazine use intensity (market research data on atrazine use
data by county, averaged across the years 1998-2002), R factor (USDA,
2001, Natural Resource Inventory), K factor (average for the STATSGO
mapping units, USDA, 1994), and Dunne overland flow (Wolock, 2003) at
the HUC-10/11 watershed-scale. Details are provided in Williams et al
(2004a and 2004b).   REF _Ref180382416 \h  Figure III-1  shows the
results of the WARP assessment for those watersheds that intersected
counties with atrazine use intensities of 0.25 lb ai/ac or greater
(5,860 watersheds). The watersheds are ranked in five quintiles based on
the estimated 95th percentile WARP concentrations. 

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  1  Ranking of
watersheds in quintiles using the USGS WAtershed Regression on
Pesticides model

Williams et al. (2004a) also considered watershed vulnerability
approaches based on flow accumulation under row crops (derived from the
National Elevation Dataset and the 1992 National Land Cover Dataset) and
a soil surface runoff potential rating based on intrinsic soil
properties only (Pierce and Anderson, 1992). 

Syngenta evaluated the various watershed vulnerability approaches using
atrazine monitoring data from a number of sources, including USGS, their
own monitoring programs, and various state and university programs
(Williams et al, 2004a, 2004b). Williams et al (2004a) grouped
monitoring stations with no reported atrazine concentration greater than
0.1 ppb into a “low” detection group and monitoring stations with at
least one detection of atrazine greater than 3.0 ppb into a “high”
group. 

The low detection group included 526 monitoring stations that
represented the lower 21st percentile of the atrazine monitoring data.
The high detection group included 487 monitoring stations, representing
the upper 18th percentile of atrazine monitoring data. Syngenta used GIS
tools to evaluate how well the various watershed vulnerability
approaches separated sites with detections >3 ppb from those with
detections <0.1 ppb. The highest 20% of vulnerable watersheds identified
using WARP provided the most effective distinction between high and low
detection sites (  REF _Ref181161804 \h  Table III-1 ).

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  1  Evaluation of
selected watershed vulnerability approaches for atrazine monitoring,
based on Williams et al, 2004a

Watershed Vulnerability Approaches1	Total HUCs included	HUCs with sample
stations2	Stations with atrazine detects >3 ppb	Stations with atrazine
detects <0.1 ppb



	No.	Pct	No.	Pct

Initial Pool of 3,440 HUCs in 11 states

WARP upper 90th %ile	339	51	45	88%	0	0%

WARP upper 80th %ile	632	102	85	83%	0	0%

Flow Acc. Upper 90th %ile 	287	32	15	47%	7	22%

Sev. Soil Runoff upper 90th %ile	304	43	17	40%	10	23%

Combined upper 90th%ile WARP + Flow Acc	626	83	60	72%	7	8%

Combined upper 90th%ile WARP + Flow Acc + Sev. Soil Runoff	930	126	77
61%	17	13%

Combined upper 95th%ile WARP + Flow Acc + Sev. Soil Runoff	493	59	39	66%
8	14%

Final Pool of 5,860 HUCs in 37 states

WARP upper 80th %ile	1172	195	156	80%	2	1%

Flow Acc. Upper 90th %ile not included in WARP upper 80th %ile	263	26	5
19%	16	62%

Combined WARP upper 80th and Flow Acc. Upper 90th %ile	1435	221	161	73%
18	8%

1 Watershed vulnerability approaches:

WARP = Watershed Regression for Pesticides 95th percentile annual
residue prediction algorithm

Flow Acc = Flow Accumulation under Row Crops (from NED, NLCD)

Sev. Soil Runoff = Soils rated as having a severe potential for surface
runoff of pesticides

2 Pool of sampled HUC’s for the 11 state group was 574 (295 in the
“upper” group and 87 in the “lower” group). For the 37 state
group was 797 (333 in the “upper” group and 198 in the “lower”
group).

The upper 10th and upper 20th percentiles of WARP effectively grouped
the highest percentile of monitoring stations with atrazine detections
>3 ppb (80-88% of the stations included in these tiers of vulnerable
watersheds had at least one reported detection of atrazine >3 ppb) with
the fewest numbers of stations that never had detects >0.1 ppb (  REF
_Ref181161804 \h  Table III-1 ). Vulnerability approaches based on flow
accumulation under row crops and percentage of soils rated as having a
severe potential for surface runoff included more monitoring stations
with low or no atrazine detections. Most likely these approaches did not
account for atrazine use and, thus, could include areas that, while
vulnerable to pesticide runoff, did not have any atrazine use in the
watershed.

Using the watershed vulnerability approach, the atrazine use area in the
corn and sorghum-growing regions of the U.S. was divided into three
tiers (  REF _Ref180382390 \h  Figure III-2 ):

Approximately 10,000 HUC-10/11 watersheds in which atrazine was used on
corn and sorghum. 

5,860 HUC-10/11 watersheds that intersected counties with use
intensities of 0.25 lb ai/county acre or higher, based on atrazine use
between 1998 and 2002. 

High vulnerability tier of 1,172 HUC-10/11 watersheds that represent the
upper 20th percentile of the second tier of watersheds, based on USGS’
WARP model. 

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  2  Location of
atrazine use area (gray), areas of use > 0.25 lb/ac (green), most
vulnerable 20th percentile of watersheds (yellow), and sampled
watersheds (red).

Several assumptions are implicit in the watershed vulnerability
approach:

The WARP estimates reflect watershed vulnerability over time. In
particular, the approach assumes that year-to-year variability in
atrazine use intensity, one of the major driving factors in the
regression model estimates, will not result in drastic changes in the
relative ranking of the watersheds. This assumption may not be
unreasonable since atrazine use has remained relatively steady over
time. However, if shifts in relative usage of atrazine on corn and
sorghum or if shifts in corn-sorghum cropping patterns occur, this
assumption may not hold.

The WARP estimate of the 95th percentile atrazine concentration in a
year captures the major factors affecting the relative vulnerability of
watersheds to atrazine loading in water bodies. Since WARP was based on
nationally-available datasets, it does not necessarily capture locally
important characteristics that may influence atrazine movement to water
bodies.

The distributional approach to atrazine concentrations (i.e., WARP
estimates of the 95th percentile of atrazine concentrations in a year)
are a reasonable surrogate for the magnitude-duration chemographs that
impact aquatic communities.

A watershed-based vulnerability is reflective of the relative
vulnerability of the water bodies within that watershed.

Representative Sampling of Vulnerable Watersheds 

The goal of the atrazine monitoring program was to employ a
probability-based, spatial survey design to select sampling locations so
that the extent to which waters are exceeding effect-based atrazine
thresholds could be defensibly estimated via the use of the appropriate
statistical methods. For clarity, the components of the population and
sample are first summarized.

Target Population 

The critical question for the monitoring program was “To what extent
are waters exceeding effects-based thresholds for atrazine?”  One
interpretation of the target population is then defined as the
collection of all watersheds within the United States (the 48
conterminous states in this instance).  Since watersheds are defined for
all locations on streams and rivers, the monitoring question of interest
can be stated as “How many kilometers of all streams and rivers within
the United States exceed the established level of concern (LOC)?” In
actuality, to focus the field monitoring on areas considered to be most
vulnerable and to address limitations on where field monitoring could be
implemented, a restricted set of watersheds, based on HUCs, was defined
for the target population (  REF _Ref181271383 \h  Table III-2 ).  It is
important to understand this restriction and how it impacts the survey
design and its interpretation. 

  REF _Ref181271383 \h  Table III-2  distinguishes two alternative
target populations: (1) Stream and river based and (2) HUC-based.  The
HUC-based target population is a subset of the stream and river based
target population since HUC pour points are a subset of all possible
locations on the stream and river network.  The HUC-based target
population was used because of uncertainties associated with
county-level atrazine use estimates and because, at the time, the
framework was not available to compute WARP scores for all stream and
river locations within the corn/sorghum use areas (see Section   REF
_Ref181761340 \w \h  IV.A  for some potential future approaches). 
Consequently, rather than defining the target population as watersheds
defined by all locations on stream and river network, the target
population is defined in terms of HUCs where it is feasible to compute
WARP scores for the 5,860 HUCs that intersect counties with atrazine use
>0.25 lb ai/county acre.  This is Stratum B for the HUC target
population in   REF _Ref181271383 \h  Table III-2 .  This stratum was
then split into two additional strata based on the WARP scores.  Stratum
B WARP > 80 Percentile includes the set of 1,172 HUCs in areas where
corn and sorghum are grown that were predicted by the USGS WARP model to
be the 20% most vulnerable watersheds to atrazine out of the 5,860
watersheds.  This stratum is further split into the those HUCs predicted
by USGS WARP model to be between the 80th and 95th percentile (874 HUCs)
and those above the 95th percentile (298 HUCs).

Because the intent of the monitoring study was to focus on the most
vulnerable watersheds, Stratum A and Stratum B WARP 0-80 Percentile are
explicitly excluded from the survey design.  HUCs in these two strata
are not included in the atrazine monitoring program.

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  2  Target
Population and Survey Design Stratification

Survey Design Strata	Stream and River Target Population	Hydrologic Unit
(HUC-10) Target Population

Stratum A	All streams and rivers in HUC watersheds that do not intersect
counties with atrazine use >0.25 lb atrazine/county acre	All HUC
watersheds that do not intersect counties with atrazine use >0.25 lb
atrazine/county acre

Stratum B	All streams and rivers in HUC watersheds that intersect
counties with atrazine use >0.25 lb atrazine/county acre	All HUC
watersheds that intersect counties with atrazine use >0.25 lb
atrazine/county acre (5,860 HUCs)

Stratum B WARP 0-80 Percentile	All streams and rivers in Stratum B with
WARP scores less than or equal to 80th percentile of all WARP scores in
Stratum B	All HUC watersheds in Stratum B with WARP scores less than or
equal to 80th percentile of all WARP Scores in Stratum B (4,688 HUCs)

Stratum B WARP > 80 Percentile	All streams and rivers in Stratum B with
WARP scores greater than 80th percentile of all WARP scores in Stratum B
All HUC watersheds in Stratum B with WARP scores greater than 80th
percentile of all WARP Scores in Stratum B (1,172 HUCs)

Stratum B WARP (2,4]

All HUC watersheds in Stratum B WARP >80 with WARP scores between 2 and
4 (874 HUCs)

Stratum B WARP (4,14]

All HUC watersheds in Stratum B WARP >80 with WARP scores > 4 (298 HUCs)

  

Survey Design

The probability-based survey design is stratified based on the
categories described for the HUC-10 target population.  Stratum A and
stratum “B WARP 0-80 Percentile” were not sampled.  Only Stratum
“B WARP (2,4]” and stratum “B WARP (4,14]” were sampled.  Given
the available resources and the desire to focus the sampling on HUCs in
areas where corn and sorghum are grown that are the most vulnerable to
exposures to atrazine, samples were selected only from HUCs within the
top 20% of USGS WARP scores.  That is the sampling units (i.e. the units
actually sampled) were the HUCs, with the sampling frame (i.e. the list
of sampling units) being a GIS coverage of the centroids of each of the
HUCs.  

Within each of the two strata sampled, the HUCs were selected with
probability proportional to the value of atrazine use estimated for the
HUC from the county-level use data.  Consequently, HUCs with higher
estimated atrazine use were more likely to be selected than those with
lower estimated use within each stratum. The rationale was to focus
sampling in HUCs more likely to result in an exceedance of the LOC. 

Sample Size And Analysis Weights

The total sample size was 40 HUCs with 20 allocated to each of the two
strata WARP (2,4] and WARP (4,14].  In addition, 5 additional HUCs were
selected in each stratum to be used as backup replacement HUCs in the
event that one or more of the original 20 HUCs could not be monitored
due to field operation limitations.

One stratum, where N1=874, represents “moderately high” WARP scores
of (2,4].  The other stratum represents watersheds with the overall
highest WARP scores, which fall in the range of (4, 14] where N2=298. 
Since the survey design includes stratification and unequal probability
of selection proportional to atrazine use (kg/km2), the statistical
analyses must incorporate this information through the use of weights
associated with each HUC.  The weights are inversely proportional to the
probability of inclusion; therefore, each of the selected watersheds in
the WARP (2,4] stratum (N1=874) represent 20 to 93 watersheds, where a
HUC with higher atrazine use would represent fewer HUCs and a HUC with
lower atrazine use would represent more HUCs (since more HUCs have lower
atrazine use than high atrazine use within the stratum.  Similarly, in
the WARP (4,14] stratum (N2=298) the selected watersheds represent
approximately 9 to 38 watersheds.

The survey design selection also incorporated a process that ensures
that every potential sample selection would be spatially-balanced.  That
is, the algorithm guarantees that the HUCs will always be spatially
representative over the extent of the population of 1,172 watersheds. 
The method is termed a generalized random tessellation stratified (GRTS)
design (Stevens and Olsen, 2004; Stevens and Olsen, 1999).  The GRTS
method was developed as part of the USEPA Office of Research and
Development (ORD) Environmental Mapping and Assessment Program (EMAP)
(McDonald et al, 2002) and has been used extensively by EPA and many
state programs.  Actual selection of HUCs was completed using the
spsurvey library (ARM 2007) developed by EMAP for use with the R
Statistical Software program (R Development Core Team, 2006).

  REF _Ref181163437 \h  Figure III-3  shows the location of the 1,172
HUC-10s and the subset of 40 HUC-10s that were actually selected and
monitored.

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  3  Location of
the 40 watersheds (red) sampled for the atrazine eco-monitoring study.

Sampled Population

The survey design selected 40 HUCs to be sampled (  REF _Ref182036577 \h
 Table III-3 ). Each of the selected HUCs was evaluated to determine if
it was feasible to locate and maintain a stream monitoring site within
the HUC.  In seven of the initial 40 HUCs a suitable stream sampling
site could not be determined (4 from WARP (2,4] stratum and 3 from WARP
(4,14] stratum).  Following procedures for replacing HUCs for GRTS
sampling these were replaced by other HUCs within each stratum. 
Although only 40 HUC-10s were monitored, the actual sample size is 47
with missing data for the 7 non-sampleable HUCs.  It is critical that
the seven non-monitored HUCs be included since they represent the
collection of HUCs within the 1,172 HUCs where it would not have been
possible to locate and maintain a monitoring site.  Unless additional
assumptions are made it is not possible to know whether the 40 monitored
HUCs are representative of this collection of non-monitored HUCs.  One
potential assumption would be that this collection is “missing at
random” and hence 40 monitored HUCs would represent all 1,172 HUCs. 
Note that the number of HUCs in this collection can, and will, be
estimated.

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  3  Watersheds
selected for monitoring using the GRTS approach.

Site	Watershed_Name	HUC11 Code	Stratum WARP	Years Monitored	Auto-samples

IA-01	Wolf Creek	07080205090	(2,4]	2004-05

	IA-02	Nishnabotana River	10240002080	(2,4]	2004-05	2004-05

IL-01	Pecatonica River	07090003140	(2,4]	2004-05

	IL-02	Pine Creek	07090005050	(2,4]	2004-05

	IL-03	Spring Creek	07090007050	(4,14]	2005-07 1

	IL-04	Iroquois River	07120002160	(2,4]	2005-07 1

	IL-05	Panther Creek	07130004040	(2,4]	2004-05

	IL-06	Sugar Creek West Fork	07130009060	(2,4]	2004-05	2004-05

IL-07	Grindstone Creek	07130010060	(2,4]	2004-05

	IL-08	Horse Creek	07130007060	(4,14]	2005-07 1	2005-06

IL-09	Muddy Creek, Illinois	05120112070	(4,14]	2004-05

	IL-NS

07140202090	(2,4]	Not sampled

	IN-01	Mill Creek	05120106070	(2,4]	2004-05

	IN-02	Eel River	05120104040	(2,4]	2004-05

	IN-03	Eightmile Creek	05120101110	(4,14]	2005-06

	IN-04	Rock Creek	05120105020	(4,14]	2004-06

	IN-05	Limber Lost Creek	05120101050	(4,14]	2004-06

	IN-06	Vermilion River, North Fork	05120109090	(4,14]	2005-07 1	2005-07

IN-07	White River	05120201070	(4,14]	2005-06

	IN-08	Whitewater, Nolans Fork	05080003030	(4,14]	2005-06

	IN-09	Raccoon Creek	05120108160	(4,14]	2005-06	2005-06

IN-10	Brandywine Creek	05120204040	(4,14]	2005-06

	IN-11	Little Pigeon Creek	00514020114	(2,4]	2005-07 1	2005-07

IN-NS1

05120111170	(2,4]	Not sampled

	IN-NS2

05120111090	(4,14]	Not sampled

	KY-01	Brashears Creek	05140102090	(4,14]	2005-06

	KY-02	Twomile Creek	05110005130	(4,14]	2005-06

	KY-NS

05110004110	(2,4]	Not sampled

	LA_NS

01114020891	(4,14]	Not sampled

	MN-01	Whitewater North Fork	07040003110	(2,4]	2005-06	2005-06

MO-01	South Fabius River	07110003040	(2,4]	2004-07 1	2004-07

MO-02	Youngs Creek	07110006030	(4,14]	2004-07 1

	MO-03	Little Sni-a-Bar Creek	10300101130	(2,4]	2004-06	2006

MO-NS

10240011060	(2,4]	Not sampled

	NE-01	Wahoo Creek	10200203090	(2,4]	2004-05

	NE-02	Middle Loup River	10210003080	(2,4]	2005-06

	NE-03	Platte River	10200101040	(2,4]	2004-05

	NE-04	Big Blue River, Upper Gage	10270202060	(4,14]	2005-06	2006

NE-05	Muddy Creek, NE	10240008081	(4,14]	2005-06	2006

NE-06	Crooked Creek	10250016081	(2,4]	2004-06	2004-06

NE-07	Big Blue River, Lower Gage	10270205011	(4,14]	2005-06

	NE-NS

10270203060	(4,14]	Not sampled

	OH-01	Kokosing River	05040003020	(2,4]	2004-05

	OH-02	Licking River, North Fork	05040006010	(4,14]	2005-07 1

	OH-03	Mad River	05080001160	(2,4]	2004-05	2004-05

OH-04	Deer Creek	05060002020	(4,14]	2005-06

	TN-01	Obion Middle Fork	00801020303	(4,14]	2005-06

	1 – Monitoring data for 2007 are still undergoing QA/QC and are not
included in this assessment. 



Monitoring Question

The monitoring question initially posed was “To what extent are waters
exceeding effects-based thresholds for atrazine? The extent of
exceedances will be quantified in terms of X% of watersheds having
flowing water bodies that exceed the trigger with Y% confidence.”  The
actual monitoring question that can be answered by the survey design is
“How many (%) HUCs in corn and sorghum growing regions in the United
States where predicted exposure to atrazine is greatest are estimated to
have atrazine concentrations that exceed the LOC in at least one
sub-watershed?”  The latter phase “in at least one sub-watershed”
is necessary since only one potential monitoring site on the stream and
river network within the HUC was monitored.  The next section describes
the process for selecting monitoring sites within a HUC.

Selection Of Monitoring Site Within The Watershed

Williams et al (2004a) details the criteria used to select monitoring
sites within the targeted watersheds (described in Section   REF
_Ref181170489 \w \h  III.B.1.b) . The criteria for selecting stream
segments for monitoring focused on stream segments relevant to study
goals, e.g., sub-watersheds with higher row-crop densities (and, thus,
higher likelihood of atrazine use) and sub-watersheds with minimal urban
influences. The selection process also avoided sub-watersheds that may
be subject to major annual variation in atrazine load as a result of
crop rotation or highly “flashy” hydrology which may minimize longer
atrazine exposures.

Syngenta identified stream segments with maximum drainage areas that
would 

exclude major river stems running through the interbasin HUC10/11
watersheds, 

avoid larger streams/rivers (5th and 6th order) within larger HUCs, and 

allow for a sufficient watershed size to minimize the likelihood of
monitoring data distortion due to annual crop rotation changes.

The following criteria were used to identify eligible stream segments
using GIS data and tools:

Minimum drainage area of 9 square miles

Maximum drainage area of no more than one-half of the HUC11 watershed,
unless total watershed area is less than 50 square miles. In such
instances, tributaries of larger streams will be manually identified 

Percent flow accumulation under urban land use is less than 10% 

Percent flow accumulation under cropland is in the upper 50th percentile

Syngenta randomly numbered all stream segments that met the criteria for
field evaluation. Beginning with the most downstream point of the first
randomly-ordered segment and working upstream, the field crew looked for
a suitable sampling site. The field crews verified that corn and/or
sorghum agriculture was present and that the streams were accessible for
monitoring. Conditions that would exclude a stream segment included
areas with few acres in corn and sorghum, a prevalence of
herbicide-tolerant corn or use of herbicides other than atrazine, point
sources such as pesticide distributors, or other anomalies. If no
suitable location was found for the first segment, the process was
repeated for the next randomly selected segment until a suitable site
was located. 

Sampling/Monitoring Design

Williams et al (2004a) and Hampton et al (2007a) provide details of the
monitoring design and conduct. Half (20) of the monitoring sties were
sampled in 2004-05 and the remaining 20 sites were sampled in 2005-06 ( 
REF _Ref181271383 \h  Table III-2 ). Each site was sampled for at least
two years. Monitoring extended into a third year for some sites in which
atrazine concentrations exceeded the LOC in one or more years, elevated
levels of atrazine below the LOC occurred in lower-than-normal rainfall
years, or low flow conditions affected sample collection.

Sampling began in early April, before planting, and continued for
approximately 120 days after an estimated 50% of the local corn acres
were planted (Hampton et al, 2007a). The sampling period typically ran
from April through August at most sites. 

At every site, water samples were collected every 4 days during sampling
period. In addition, flow-triggered automatic samplers were installed at
10 of the sites to provide a comparison between the 4-day grab samples
and autosamples. The automatic samplers were triggered to start and stop
at specified changes in stream flow, determined specifically for each
site (Hampton et al, 2007a). In addition, all sites were instrumented to
measure stream flow and to collect weather data (e.g., meteorological
conditions, rainfall, soil moisture). 

All water samples were analyzed for atrazine using immuno-assay (limit
of detection of 0.05 ug/L), with detections confirmed using gas
chromatography/mass spectrometry (GC/MS). After May 2, 2005, all samples
were analyzed using liquid chromatography/ mass spectrometry/mass
spectrometry (LC/MS/MS) (Hampton et al, 2007a).

 

Monitoring Results

Hampton et al (2007a) and Volz et al (2007) provide results of the
monitoring study. This section provides the US EPA’s analysis of the
data.

Atrazine Chemographs

μg/L from the IN-11 site in 2005.  The mean annual concentrations
ranged from a maximum of 9.5 μg/Ll from the MO-01 site in 2004 to a low
of 0.1 μg/L for the NE-06 site in 2006, while the median values ranged
from 4.2 μg/L for the MO-02 site in 2004 to 0.08 μg/L for the OH-03
site in 2005.  For all 40 sites, peak concentrations ranged from 0.13 to
208.76 μg/L, 14-day average concentrations from 0.1 to 80 μg/L, 30-day
average concentrations from 0.1 to 45 μg/L, 60-day average
concentrations from 0.1 to 26 μg/L, and 90-day average concentrations
from 0.1 to 18 μg/L.  

μg/L. Most of the sites had multiple atrazine peaks (defined as
detections greater than 1 μg/L), but these were generally of short
duration, usually no more than 1 to 3 sample points (  REF _Ref181177077
\h  Table III-4 , Appendix 3). The exceptions are MO-01 and MO-02, both
of which had extended periods of elevated atrazine concentrations. Forty
of the site-years had atrazine peaks greater than10 μg/L; 25 had peaks
greater than 20 μg/L; and 10 had peaks greater than 50 μg/L, with 4 of
those peaks associated with the same monitoring site (MO-01). 

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  4  Summary of
chemograph shapes, numbers and magnitudes of atrazine peaks for the 40
monitoring sites

Site	Year	Chemograph Shape	# Peaks	Max. Peak (μg/L)	Range in
consecutive  days >1 μg/L2



Peaks >1 μg/L	Pattern 1	10-20 μg/L	20-50 μg/L	>50 μg/L



IA-01	2004	2	M-L	1

	10.0	8-12

IA-01	2005	2	L-L



1.2	4

IA-02	2004	5	L-L-L-L-L



1.8	4-16

IA-02	2005	2	L-L



5.5	4-12

IL-01	2004	4	L-M-L-L	1

	13.2	4-16

IL-01	2005	0



	0.0	na

IL-02	2004	3	L-L-L 



4.9	4-12

IL-02	2005	1	L



2.9	12

IL-03	2005	1	L



5.6	4

IL-03	2006	1	L



2.5	4

IL-04	2005	1	L



2.8	12

IL-04	2006	4	M-L-L-L	1

	11.5	4-12

IL-05	2004	3	L-L-H

1

22.1	4

IL-05	2005	1	L



1.8	4

IL-06	2004	3	L-L-L



2.2	4-8

IL-06	2005	0



	0.0	na

IL-07	2004	3	L-H-L

1

21.8	4-12

IL-07	2005	2	L-L



2.3	1-8

IL-08	2005	3	L-L-L



5.6	4-20

IL-08	2006	2	H-M	1	1

33.1	8

IL-09	2004	4	M-M-M-L	3

	13.3	4-20

IL-09	2005	2	M-L	1

	16.0	8-12

IN-01	2004	3	L-L-L



8.6	8-24

IN-01	2005	2	L-L



4.4	4-12

IN-02	2004	5	L-L-L-L-L



9.3	12-20

IN-02	2005	3	L-H-L

1

20.3	4-12

IN-03	2005	3	L-L-L



7.6	8-12

IN-03	2006	6	M-L-L-L-L-L	1

	16.9	4-12

IN-04	2004	3	VH-L-L

	1	78.1	4-8

IN-04	2005	3	L-L-L



8.7	4-8

IN-04	2006	3	M-L-L	1

	10.2	4-8

IN-05	2004	4	L-H-H-L

2

28.9	4-12

IN-05	2005	3+	M-L-M	2

	17.3	8-36

IN-05	2006	4	L-H-H-L

2

41.0	4-12

IN-06	2005	2	L-L



7.2	4-8

IN-06	2006	5	L-L-L-L-L



9.4	8-20

IN-07	2005	3	H-M-L	1

	22.6	4-11

IN-07	2006	3	M-L-L	1

	10.5	4-16

IN-08	2005	4	L-H-L-L

1

21.1	4-8

IN-08	2006	5	L-H-H-L-L

2

20.7	4-12

IN-09	2005	4	L-L-L-L



9.4	4-12

IN-09	2006	2	L-L



8.3	4-8

IN-10	2005	3	M-L-L	1

	12.4	4-20

IN-10	2006	3	M-M-L	2

	16.4	4-12

IN-11	2005	1	VH 

	1	208.0	24

IN-11	2006	3	L-L-L



9.8	4-12

KY-01	2005	2	L-L



2.2	12-20

KY-01	2006	3	L-H-L

1

22.4	4

KY-02	2005	5	L-M-M-L-L	2

	19.3	8-12

KY-02	2006	2	M-L	1

	14.3	8-12

MN-01	2005	1	L



5.9	16

MN-01	2006	0



	0.0	na

MO-01	2004	4	VH-VH-M-L	1

2	66.0	12-24

MO-01	2005	3	H-VH-H

2	1	182.0	24-32

MO-01	2006	3	VH-L-L

	1	82.8	8-32

MO-02	2004	5	H-VH-H-L-L

2	1	54.0	20-24

MO-02	2005	5	M-M-H-L-L	2	1

28.1	12-76

MO-02	2006	3	H-L-L

1

43.2	8-60

MO-03	2004	2	L-VH

	1	59.0	13-79

MO-03	2005	4	L-M-L-L	1

	12.3	20-56

MO-03	2006	1	L



3.9	52

NE-01	2004	2	L-M	1

	19.3	16-36

NE-01	2005	3	L-M-M 	2

	16.7	8-16

NE-02	2005	5	M-H-M-M-L	3	1

20.7	8-12

NE-02	2006	2	VH-L

	1	82.0	4-16

NE-03	2004	3	L-L-L



2.3	4-8

NE-03	2005	2	L-M	1

	11.3	4

NE-04	2005	3	H-L-L

1

49.9	4-60 3

NE-04	2006	1	L



4.1	28

NE-05	2005	3	L-H-H

2

49.9	8-44 3

NE-05	2006	1	L



6.8	52

NE-06	2004	5	L-L-L-L-L



7.8	4-8

NE-06	2005	2	M-H 	1	1

33.1	8-12

NE-06	2006	0



	0.0	na

NE-07	2005	3	L-M-VH	1

1	112.2	8-24 3

NE-07	2006	1	L



1.9	37

OH-01	2004	3	M-M-L	2

	18.3	8-12

OH-01	2005	3	L-L-L



3.0	4

OH-02	2005	4	M-L-L-L	1

	18.2	8-24

OH-02	2006	2	L-M	1

	14.0	8-20

OH-03	2004	4	M-M-H-L	2	1

21.5	8-20

OH-03	2005	2	L-L



8.2	4-39

OH-04	2005	4	H-L-L-L

1

20.2	4-16

OH-04	2006	2	L-L



6.3	4

TN-01	2005	5	L-L-L-L-L



7.6	8-20

TN-01	2006	4	L-L-L-M	1

	10.7	4-8

1 – Pattern refers to the relative size of the peak in sequence. L =
1-10 ug/l; M = 10-20 ug/l; H = 20-50 ug/l; VH = >50 ug/l

2 – the range in consecutive days >1 ug/l is based on 4-day sample
intervals, except where autosample data were available. For the 4-day
samples, the measured concentration applied to a 4-day period.

3 – the span of consecutive days with concentrations >1 ug/L for
NE-04, NE-05, and NE-07 in 2005 include sample gaps (see text for
discussion). 

Analyzing Atrazine Chemographs: LOC Exceedances

Ultimately, the site chemographs must be interpreted against the LOC,
i.e., relating the magnitude and duration of atrazine concentrations to
results of the microcosm/mesocosm studies using CASM.   REF
_Ref181177887 \h  Table III-5  summarizes results of the CASM analysis
for the site-year chemographs. Each site/year of data was expanded to a
365 day time series. Preliminary data interpolation used a stair-step
method where the grab sample concentration was extended (imputed) to the
subsequent 3 days between sampling events. Samples prior to the first
sample date were given the same concentration as the first sample date
from that year, and a similar approach was taken for the dates after the
last sampling event. Sample results from each date that were reported as
a non-detection were conservatively assigned an assumed value of the
detection limit. Because of the low concentrations, this assumption did
not impact the LOC determination.  Finally, dates where no sample was
collected or analyzed were assumed to be equal to the nearest previous
sample with a result.  This final assumption resulted in significant
uncertainty for three sites in Nebraska, where dry or low-flow
conditions resulted in fewer samples being collected. It is not clear
how much the lack of samples are due to dry stream/no flow conditions
and how much are due to low flow with stream bed morphologies that
precluded the collection of samples. The uncertainties arising from the
approach the Agency took in extrapolating concentrations across periods
when no samples are taken are addressed in Section   REF _Ref182290615
\w \h  III.E.2 .

μg/L)	%SSI Dev.	Mult. Factor	LOC Status



Peak	14-day	30-day	60-day



	IA-01	2004	10.0	3.7	2.1	1.2	0.3	12.1	Below LOC 2 or more years

	2005	1.2	0.5	0.3	0.3	0.0	13.2

	IA-02	2004	1.8	1.1	0.8	0.6	0.4	12.2	Below LOC 2 or more years

	2005	5.5	2.1	1.4	0.8	0.3	14.1

	IL-01	2004	13.2	6.6	4.1	2.5	0.8	5.5	Below LOC 2 or more years

	2005	0.6	0.3	0.3	0.3	0.0	7.2

	IL-02	2004	4.9	2.9	2.3	1.5	0.5	8.6	Below LOC 2 or more years

	2005	2.9	1.8	1.1	0.6	0.2	8.5

	IL-03	2005	5.6	1.8	1.0	0.6	0.1	23.1	Below LOC 2 or more years

	2006	2.5	0.9	0.5	0.3	0.1	18.4

	IL-04	2005	2.8	1.4	0.8	0.5	0.1	16.9	Below LOC 2 or more years

	2006	11.5	3.4	1.8	1.8	0.6	7.5

	IL-05	2004	22.1	7.2	3.6	2.0	0.7	7.9	Below LOC 2 or more years

	2005	1.8	0.7	0.4	0.3	0.1	19.1

	IL-06	2004	2.2	1.1	0.7	0.5	0.3	15.3	Below LOC 2 or more years

	2005	0.2	0.2	0.2	0.2	0.0	13.3

	IL-07	2004	21.8	7.0	4.2	2.4	0.9	6.6	Below LOC 2 or more years

	2005	2.3	0.9	0.6	0.5	0.1	14.3

	IL-08	2005	5.6	4.4	2.8	1.8	1.1	3.6	Below LOC 2 or more years

	2006	33.1	11.0	8.1	4.4	1.9	2.9

	IL-09	2004	13.2	8.1	6.3	4.6	1.4	2.9	Below LOC 2 or more years

	2005	16.0	6.2	3.4	2.3	1.0	3.2

	IN-01	2004	8.6	4.0	3.5	2.4	0.7	5.9	Below LOC 2 or more years

	2005	4.4	1.4	1.0	0.7	0.3	5.5

	IN-02	2004	9.3	6.3	4.5	2.8	0.8	4.6	Below LOC 2 or more years

	2005	20.3	6.3	4.3	3.0	1.0	4.5

	IN-03	2005	7.6	4.3	3.3	2.3	0.7	5.1	Below LOC 2 or more years

	2006	16.9	10.6	6.2	3.9	1.3	3.7

	IN-04	2004	78.1	23.8	12.0	6.4	1.7	2.7	Below LOC 2 or more years

	2005	8.8	3.6	2.1	1.4	0.4	7.2



2006	10.2	5.6	3.7	2.2	0.6	6.0

	IN-05	2004	28.9	14.9	11.9	7.0	2.2	2.2	Below LOC 2 or more years

	2005	17.3	7.8	4.5	4.1	1.5	2.8



2006	41.3	17.9	13.1	7.4	2.4	2.2

	IN-06	2005	7.2	2.9	1.8	1.0	0.4	11.5	Below LOC 2 or more years

	2006	9.4	4.0	2.7	1.9	0.9	5.4

	IN-07	2005	22.6	9.6	6.4	3.9	0.6	7.6	Below LOC 2 or more years

	2006	10.5	5.4	3.6	2.0	0.6	7.4

	IN-08	2005	21.1	6.9	4.9	2.8	1.0	5.5	Below LOC 2 or more years

	2006	20.7	8.9	7.7	4.4	1.5	3.8

	IN-09	2005	9.4	3.7	2.4	1.7	1.2	4.2	Below LOC 2 or more years

	2006	8.3	3.0	1.8	1.2	0.6	7.1

	IN-10	2005	12.4	6.1	4.0	2.8	0.9	4.8	Below LOC 2 or more years

	2006	16.4	7.5	6.3	3.6	1.2	4.4

	IN-11	2005	208.8	65.1	31.5	16.2	5.6	0.7	Exceeds LOC in 1 year

	2006	9.8	5.9	3.3	1.9	1.3	3.5

	KY-01	2005	2.2	1.5	1.2	0.9	0.3	14.2	Below LOC 2 or more years

	2006	22.4	6.9	3.6	1.9	0.7	8.5

	KY-02	2005	19.3	8.7	7.1	4.5	1.7	2.0	Exposures within 2x of LOC in 1
year

	2006	14.3	4.7	3.9	2.3	0.6	3.9

	MN-01	2005	5.8	2.6	1.4	0.8	1.2	3.8	Below LOC 2 or more years

	2006	0.2	0.2	0.1	0.1	0.0	20.4

	MO-01	2004	65.9	39.6	28.6	19.4	4.4	0.9	Exceeds LOC in 2 or more years

	2005	182.8	78.1	42.5	25.7	6.8	0.5



2006	82.8	48.2	31.6	17.5	4.4	0.9

	MO-02	2004	53.8	33.0	25.9	16.8	4.7	0.8	Exceeds LOC in 2 or more years

	2005	28.1	18.7	14.6	11.5	5.2	0.8



2006	43.2	34.7	27.4	15.4	4.5	0.8

	MO-03	2004	59.0	23.3	13.1	8.1	3.0	1.3	Exposures within 2x of LOC in 2
years

	2005	12.3	8.7	6.9	5.5	1.9	2.0



2006	3.9	2.3	1.9	1.5	1.0	4.8

	NE-01	2004	19.2	13.0	7.5	4.3	1.3	3.8	Below LOC 2 or more years

	2005	16.7	6.6	5.6	3.6	1.2	4.4

	NE-02	2005	20.7	11.4	10.7	6.3	2.0	2.4	Below LOC 2 or more years

	2006	82.0	28.6	14.1	7.3	1.8	2.3

	NE-03	2004	2.3	1.1	1.0	0.6	0.1	19.0	Below LOC 2 or more years

	2005	11.9	3.7	2.1	1.2	0.3	10.0

	NE-04	2005	36.0	36.0	27.3	17.4	4.5	0.8	Low flow/ exceeds 1 year

	2006	4.1	4.1	3.1	1.7	2.1	2.1

	NE-05	2005	49.9	23.8	19.9	16.5	4.6	0.8	Low flow/ exceeds 1 year

	2006	6.8	6.8	6.8	5.1	0.8	4.5

	NE-06	2004	7.7	2.8	2.1	1.7	0.5	7.3	Below LOC 2 or more years

	2005	33.1	20.6	11.4	6.0	1.9	2.8



2006	0.1	0.1	0.1	0.1	0.0	20.5

	NE-07	2005	112.2	80.0	45.2	22.7	5.2	0.7	Low flow/ exceeds 1 year

	2006	na	na	na	na	0.6	8.5

	OH-01	2004	18.3	8.8	5.7	3.2	1.0	5.3	Below LOC 2 or more years

	2005	3.0	1.0	0.9	0.6	0.2	12.0

	OH-02	2005	18.1	7.1	4.0	2.9	1.0	5.1	Below LOC 2 or more years

	2006	14.0	5.9	5.2	2.9	0.8	5.3

	OH-03	2004	21.5	9.1	7.3	4.5	1.4	3.9	Below LOC 2 or more years

	2005	8.2	3.0	1.6	0.9	0.5	11.5

	OH-04	2005	20.2	8.0	4.7	2.7	1.0	5.5	Below LOC 2 or more years

	2006	6.3	2.6	1.7	1.0	0.3	12.9

	TN-01	2005	7.6	5.6	4.0	2.9	1.0	3.9	Below LOC 2 or more years

	2006	10.7	3.6	2.4	2.0	0.7	3.5

	

The level of concern is triggered when the Steinhaus Similarity Index
(SSI) exceeds 4 percent. Only two sites – MO-01 and MO-02 exceeded the
4% SSI trigger in multiple years. Those sites exceeded in all three
years of monitoring. IN-11 exceeded the 4% SSI in 2005 but not in 2006.
Monitoring continued at this site in 2007, but the results are not
available at this time. Three sites in Nebraska – NE-04, NE-05, and
NE-07 – exceeded the 4% SSI in one year (2005), but not in subsequent
years when there was sufficient flow to allow for sampling on a 4-day
interval. The chemographs associated with the LOC exceedances in these
sites include periods in which no samples were taken because of low
stream flow (Hampton et al, 2007a). Two sites – KY-02 and MO-03 did
not exceed the 4% SSI trigger but had exposures that were within a
factor of 2 of exceeding the trigger based on the multiplication factor.
Because of uncertainties identified in the way the model may
underestimate the effects of short-duration high exposures and in the
extent to which the sample frequencies may capture peak concentrations,
these sites are flagged as uncertain. These conditions are grouped
separately in the statistical assessment below.

Percentage Of Watersheds Exceeding The LOC Threshold

Statistical Analysis Based On GRTS

Methods

This monitoring study was specifically designed to answer the question,
“How many watersheds in corn and sorghum growing regions in the United
States where predicted exposure to atrazine is greatest are estimated to
have atrazine concentrations that exceed the LOC in at least one
subwatershed?” This section explains how the statistical analysis
performed directly corresponds with the survey design that was used, in
order to accurately address the question.

The primary objective of the analysis was to determine how many
watersheds out of the 1,172 are estimated to fall within each of three
basic categories: 1) considered an excluded site 2) are not likely to
contain a water body with atrazine concentrations that exceed the LOC,
and 3) are likely to contain a water body with atrazine concentrations
that exceed the LOC.   The analysis provides the estimated proportion
(expressed as a percent) of watersheds as well as the number of
watersheds, along with 95% confidence intervals for each.  

This monitoring study required 47 watersheds be visited in order to
successfully attain an overall sample size of n=40 (i.e. 7 sites had to
be excluded because of logistical constraints to sampling).  Although it
may seem intuitive to completely disregard these 7 excluded sites and
only focus on the 40 in which monitoring was actually performed, doing
so would result in the loss of key information describing the
characteristics of the watersheds within the population and potentially
result in a bias in the estimates.  Therefore, in order to accurately
represent the status of the population of 1,172 watersheds, the
estimated number of excluded sites in the population is calculated,
where n=24 for stratum (2,4], and n=23 for stratum (4,14], totaling an
overall n=47.  

Proportion (expressed as a percent) and unit estimates are calculated
using the Horvitz-Thompson ratio estimator (Horvitz and Thompson, 1952).
 For each stratum, the numerator of the ratio is calculated by summing
the design weights for each sample response within a category and then
dividing by the sum of the design weights over the sample size for the
stratum.  The number of watersheds for a response category is then
estimated simply by multiplying the estimated proportion of watersheds
by the number of watersheds within that stratum.  Estimates and standard
errors are then calculated for all strata combined by weighting each
stratum by its respective population size, with N1=874 and N2=298
representing approximately 75 and 25 percent, respectively, of the
population of 1,172.    

      

Variance estimates are calculated using the local mean variance (Stevens
and Olsen, 2003), which utilizes the x- and y-coordinates for each
watershed, and confidence intervals are calculated using a Normal
distribution multiplier value of 1.96, multiplied by the standard error.
 Simulation studies reported by Stevens and Olsen (2003) have shown that
this variance estimator performs better than the Horvitz-Thompson
variance estimator. 

Assumptions

The following assumptions drive the accuracy of the estimates presented
in the results.

The 1172 HUCs accurately represent the 20% most vulnerable watersheds
within the true population.  

All other HUCs within the United States are assumed not to exceed the
LOC.

If the monitored site within each HUC shows an exceedance, then any
other similar site within that HUC would have exceeded the LOC if it had
been selected to be monitored. 

The atrazine chemograph reported for each site is independent of whether
that site was sampled during the period of 2004-2005 or from 2005-2006. 
This validates the data being grouped together and generalized as a
single 2004-2006 sampling period.

The true atrazine chemographs that occurred throughout the sampling
period were in fact captured by the data. 

Data responses have been correctly categorized.

Results

The use of the phrase “based upon this data” at the beginning of
each statement refers to the fact that the estimates and their
confidence intervals are completely dependent upon the sample size of
data used, the circumstances under which the data was collected, and the
way that the data was manipulated and/or used prior to the calculation
of the estimates themselves. A level of concern (LOC) exceedance is
based on CASM SSI % deviation calculation for each sub-watershed.  

Population Estimates 

Population estimates have been generated for 47 HUCs representing the
1172 watersheds.  These estimates have been broken into four categories:

(1) 	Excluded sites - 7 HUCs

(2)	Sites that did not exceed the LOC in either year - 32 HUCs

(3)	Sites that exceed the LOC in multiple years - 2 HUCs

(4)	Sites that exceed the LOC in one year with additional monitoring
pending – 1 HUC

(5)	Sites where the interpretation is uncertain - 5 HUCs

This last category includes one site where the LOC was exceeded 1 year
and a third year of data is pending, three sites in Nebraska where
low/no flow conditions limited sample collection, and two sites where
application of an uncertainty bound based on model sensitivity analysis
(see Section II) suggests that exceedances could occur.  The following
sections summarize the population estimates for each category separately
however the Agency is only certain about exceedances of the LOC for the
two sites with more than one year of exceedances.  All other categories
are subject to interpretation based on the uncertainty analysis
discussed in Section II (model sensitivity) and Section III (monitoring
data uncertainty).  

Excluded Sites

Based upon the data, an estimated 255 (22%) HUCs out of the population
of 1172 HUCs during 2004-2006 would have had to be “excluded,”
whereby monitoring activities could not have been successfully performed
(  REF _Ref181764879 \h  Table III-6 ). The 95% confidence interval for
the estimate is from 119 to 390 HUCs (i.e. 10%-33%).  No inference is
possible whether these HUC-10s would have had one or more subwatersheds
with LOC exceedances, without making additional assumptions.  The fact
that a monitoring site could not be located within these HUCs may result
in having different characteristics with respect to atrazine chemographs
than HUCs where a monitoring site could be located.

Sites That Did Not Exceet the LOC

Out of the 47 HUC-10s in the sample, 32 of the 40 monitored
sub-watersheds within the HUC-10s reported atrazine concentrations
resulting in an assessment that they were below the level of concern. 
Based upon the data, it is estimated that 669 HUC-10s (57% of 1172)
would have atrazine concentrations resulting in an assessment that they
were below the established level of concern.  The 95% confidence
intervals for the estimates are 47-67% in terms of percents and 551-787
in terms of HUC-10s.

     

Sites That Exceeded the LOC in At Least 2 Years

Two watersheds (MO 01 and MO 02) have experienced exposures that exceed
the LOC in two or more years (LOC Exceeds 2YR).  The LOC exceeds 2YR
category includes an estimated 101 (9% of 1172) HUCs that would contain
at least one sub-watershed having atrazine concentrations resulting in
an exceedance of the designated level of concern in at least 2 years,
with 95% confidence limits of 0 and 218 HUCs, or 0% to 19% of 1172 HUCs.
These are sites that triggered follow-up mitigation actions as a result
of 2 years of LOC exceedances.

Sites That Exceeded the LOC in 1 Year

One watershed (IN-11) exceeded the LOC in one year of monitoring but not
in a second year. The LOC exceeds 1YR category includes an estimated 46
(4% of 1172) HUCs that would contain at least one sub-watershed having
atrazine concentrations resulting in an exceedance of the designated
level of concern in 1 of 2 monitoring years, with 95% confidence limits
of 0 and 125 HUCs, or 0% to 11% of 1172 HUCs. Monitoring has been
extended for a third year at this site.

Uncertain Sites

Two categories of watersheds are considered uncertain in the
interpretation of the LOC exceedance for the years that were monitored:
1) Sites in which samples were not taken during low flow but exceeded
the LOC in one year based on extrapolated monitoring (Low Flow Exceeds),
and 2) Sites with exposures below the 4% SSI but within 2 times of
exceeding (LOC within 2X).  These estimates are shown in Table III-6. 
These categories are uncertain and final determination of whether they
should be considered as exceeding the LOC will depend on the outcome of
this SAP meeting.  The population estimates are simply provided to give
context to the estimates.

The Low Flow Exceeds category includes an estimated 34 (3% of 1172) HUCs
that would contain at least one sub-watershed having atrazine
concentrations resulting in an exceedance of the designated level of
concern due to low flow conditions, with 95% confidence limits of 15 and
54 HUCs, or 1% to 5% of 1172 HUCs.

The LOC within 2X category includes an estimated 67 (6% of 1172) HUCs
that would contain at least one sub-watershed having atrazine
concentrations that are within 2X of an exceedance of the designated
level of concern, with 95% confidence limits of 0 and 154 HUCs, or 0% to
13% of 1172 HUCs.

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  6  Population
Estimates for the 1,172 Vulnerable Watersheds Based on  the 40
Monitoring Sites

Category	Fre-quency	Estimated Population Percent 	95% Lower Confidence
Bound	95% Upper Confidence Bound	Estimated Number of Watersheds	95%
Lower Confidence Bound	95% Upper Confidence Bound

Excluded	7	22%	10%	33%	255	120	390

LOC Below	32	57%	47%	67%	669	551	787

LOC Exceeds 2YR	2	9%	0%	19%	101	0	218

LOC Exceeds 1YR	1	4%	0%	11%	46	0	125

Low Flow Exceeds	3	3%	1%	5%	34	15	54

LOC within 2X	2	6%	0%	13%	67	0	154

Total	47	100%	 	 	1172	 	 



Previously, the question was raised about what assumptions could be made
about the seven HUCs initially selected that could not be monitored. 
One assumption is that they are missing at random, i.e., that the HUCs
that could not be monitored have the same atrazine chemograph
characteristics as HUCs that could be monitored.    REF _Ref181765022 \h
 Table III-7  summarizes population estimates for HUCs exceeding the LOC
when this assumption is made.  Typically, the estimates under this
assumption are 1-2% greater for LOC exceedances while the estimated
number of HUCs that are below the LOC exceedance changes from 57% to 73%
(669 to 855 HUCs).  

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  7  Population
Estimates Assuming Non-Monitored HUCs Missing at Random

Category	Fre-quency	Estimated Population Percent 	95% Lower Confidence
Bound	95% Upper Confidence Bound	Estimated Number of Watersheds	95%
Lower Confidence Bound	95% Upper Confidence Bound

LOC Below	32	73%	59%	87%	855	695	1016

LOC Exceeds 2YR	2	11%	0%	24%	130	0	283

LOC Exceeds 1YR	1	5%	0%	13%	61	0	150

Low Flow Exceeds	3	4%	2%	5%	40	19	61

LOC within 2X	2	7%	0%	17%	86	0	199

Total	40	100%

	1172





Another way to summarize the population estimates is to estimate the
cumulative distribution for CASM SSI deviation (%) scores.    REF
_Ref181272450 \h  Figure III-4  shows the estimated cumulative
distribution and 95% confidence limits.  The flat portion of the plot
between approximately 2 and 4 is potentially indicative that two
populations of HUC-10s may exist with respect to CASM scores.

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  4  Estimated
Cumulative Distribution for CASM Score.

  

 Reliability of WARP estimates for HUCs and Sub-watersheds

The WARP predictions for the sampled HUCs closely correlated (r= 0.874)
with those predicted for the monitored sub-watersheds within the
selected HUCs (  REF _Ref181765305 \h  Figure III-5 ). Sub-watershed
WARP scores tend to be greater than those predicted for HUCs (  REF
_Ref181272652 \h  Figure III-6 ).  Given that the data was collected at
the sub-watershed level, WARP predictions for sub-watershed, rather than
HUCs will be used in subsequent plot comparisons.   

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  5  Comparing
WARP scores for HUC-based watersheds vs. sub-watersheds

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  6  Comparison
of WARP scores for sub-watershed and HUCs

The monitoring results indicate that WARP is not able to successfully
distinguish “moderately high” vulnerability from “high to very
high” vulnerability predictions.  If WARP were in fact able to
distinguish between the two, we would expect to see a positive
relationship between CASM SSI deviations and WARP scores (  REF
_Ref181272792 \h  Figure III-7 ).  It might also be expected that all
WARP predictions between (2,4] would have values falling in Quadrant 4,
indicating that they also received a low CASM score.  Likewise, it would
be expected that all WARP predictions with values greater than 4 would
appear in Quadrant 2, having CASM scores greater than 4%.  It is notable
for these for sub-watersheds that no CASM scores occur between
approximately 2 and 4.  It is yet to be determined whether or not the 6
points with CASM SSI scores >4% share a common factor that could
potentially be added to the WARP model in order to improve prediction
capabilities within the most vulnerable sites.  For CASM scores less
than or equal to 2%, however, an overall direct linear relationship can
be seen in WARP model, although some discrepancies appear to be present
around 2%.  A similar plot is shown in   REF _Ref181765560 \h  Figure
III-8 , plotting WARP scores against the maximum 14-day rolling average
across year per site.  

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  7 
Illustrating WARP’s capabilities to accurately predict moderately high
vs. high vulnerability to atrazine by relating it to resulting CASM
scores.

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  8  Comparing
how well WARP predicts maximum 14 day rolling averages

The WARP predictions in this analysis were used as a surrogate for the
%SSI to identify vulnerable watersheds that were likely to exceed LOC
when the study was designed.  This analysis asked the question "What
evidence does the monitoring results give to indicates that WARP was a
reasonable surrogate for SSI and consequently that the study focused on
the appropriate set of watersheds?"  Unfortunately the only data
available for comparison is the 40 monitored watersheds which are at the
upper end of WARP predictions.  At this upper end, the Agency sees no
evidence of a relationship.  Potentially this suggests that WARP may be
a "weak" surrogate for %SSI. The analysis is limited because %SSI is not
available for any watersheds below the 80th percentile.  A relationship
may exist but is not evident at the upper end of the plot.

 Comparison of Alternative WARP estimates for HUCs

WARP model predictions depend on the values of the input variables. The
initial WARP estimates used to define vulnerable watersheds used
atrazine use data averaged for the years 1998 to 2002 and did not
include Dunne overland flow as a variable (WARPDOFZ). Subsequently,
Syngenta calculated two updated WARP predictions (WARP and WARP_NewUse)
where WARP used same atrazine use data as WARPDOFZ and WARP_NewUse used
atrazine use data updated for 2001 to 2003 (Hampton et al, 2007b).  The
US EPA used atrazine data from 1999 to 2001 in its original watershed
vulnerability assessment (WARP_EFED_NoPrecip) and considered a version
of the WARP model that included May-June Precipitation as one of the
predictive parameters (WARP_EFED_Precip). The question is whether these
alternatives would identify the same set of vulnerable watersheds. The
various WARP estimates are listed in   REF _Ref182294339 \h  Table III-8
. 

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  8  Atrazine Use
Data and Model Inputs Used for Alternative WARP Estimates

WARP Estimate	Atrazine Use Data	WARP Model Parameters (1)

WARPDOFZ	1998-2002	U, R, K, A

WARP	1998-2002	U, R, K, A, D

WARP_NewUse	2001-2003	U, R, K, A, D

WARP_EFED_NoPrecip	1999-2001	U, R, K, A, D

WARP_EFED_Precip	1999-2001	U, R, K, A, D, P

1 U = Use Intensity; R = R Factor; K = K Factor; A = Watershed Area; D =
Dunne Overland Flow; P = May-June Precipitation

  REF _Ref181766115 \h  Table III-9  gives the correlations for the five
WARP predictions while   REF _Ref181766154 \h  Figure III-9  shows the
relationship for the 1172 vulnerable HUCs identified by WARPDOFZ.  In
general correlations are higher when the same atrazine use data is used.

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  9  Correlations
of Alternative WARP Predictions for 1172 vulnerable HUCs

	WARPDOFZ  	WARP	WARP_NewUs	NOPRECIP	PRECIP95

WARPDOFZ  	1	0.895	0.616	0.620	0.583

WARP	0.895	1	0.563	0.597	0.564

WARP_NewU 	0.616	0.563	1	0.777	0.704

NOPRECIP	0.620	0.597	0.777	1	0.965

PRECIP95	0.583	0.564	0.704	0.965	1



Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  9 
Scatterplots of Alternative WARP Predictions for 1172 vulnerable HUCs

The above comparison is based only on the 1172 vulnerable HUCs.  To
determine whether the five alternatives would identify the same top 20%
vulnerable HUCs requires the comparison to be made on the 5860 HUCs with
high atrazine use.  This information is available for all alternatives,
although only the actual 1172 WARP predictions are available for
WARPDOFZ.  First, the correlations for the five WARP predictions based
on 5860 HUCs are stronger, as would be expected, than when only the 1172
HUCs are considered (  REF _Ref181766255 \h  Table III-10 ). Note the
correlations with WARPDOFZ are the same since only the 1172 HUCs are
available for it.

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  10  Correlations
of Alternative WARP Predictions for 5860 High Atrazine Use HUCs

	WARPDOFZ  	WARP	WARP_NewUse	NOPRECIP	PRECIP95

WARPDOFZ  	1	0.895	0.616	0.620	0.583

WARP	0.895	1	0.824	0.818	0.811

WARP_NewUse 	0.616	0.824	1	0.777	0.704

NOPRECIP	0.620	0.818	0.919	1	0.986

PRECIP95	0.583	0.811	0.900	0.986	1



The top 20% of the WARP predictions were identified for each alternative
assuming 5860 HUCs.  Then all HUCs that were common across all
alternatives were identified (Top20_All) and all HUCs that were in the
top 20% for at least one alternative were identified (Top20_Any). Across
the alternatives, 819 HUCs are identified by all alternatives and 1585
are in the top 20% for at least one alternative.    REF _Ref181766547 \h
 Table III-11  compares how the alternatives classify vulnerability of
HUCs compared to the initial classification based on WARPDOFZ.  Each
small table would have no HUCs in the off diagonals of the table similar
to the WARPDOFZ first table. Each of the four new alternatives
classifies from 87 to 218 HUCs as being in the top 20th percentile
compared to the original WARPDOFZ classification.  Note that since
NOPRECIP and PRECIP WARP predictions were available for all 9510 HUCs
where atrazine was used, 34 and 41 of the top 20th percentile HUCs
occurred outside the high atrazine use study region.

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  11  Comparison
of HUCs in Top 20% WARP Predictions

Alternative WARP Prediction Vulnerable HUCs	WARPDOFZ Vulnerable HUCs
Total

	<80%	80-95%	>95%	Outside 5860 HUCs

	WARPDOFZ

	<80%	4688	0	0

4688

	80-95%	0	874	0

874

	>95%	0	0	298

298

	Total	4688	874	298

5860

WARP

	<80%	4601	87	2

4690

	80-95%	71	774	28

873

	>95%	16	13	268

297

	Total	4688	874	298

5860

WARP_NewUse

	<80%	4518	172	1

4691

	80-95%	166	572	134

872

	>95%	4	130	163

297

	Total	4688	874	298

5860

NOPRECIP

	<80%	4501	225	14	3599	8339

	80-95%	178	537	115	43	873

	>95%	9	112	169	8	298

	Total	4688	874	298	3650	9510

PRECIP

	<80%	4470	235	17	3616	8338

	80-95%	210	520	119	27	876

	>95%	8	119	162	7	296

	Total	4688	874	298	3650	9510



The Agency then looked at the spatial distribution of the vulnerable
watersheds identified by the different WARP estimates.   REF
_Ref181767332 \h  Figure III-10  compares the locations of the 1,172
vulnerable watersheds identified using WARPDOFZ (WARP estimates with
Dunne Overland Flow set to 0) and the watersheds identified using WARP
with Dunne Overland Flow included. The green areas show where the HUCs
overlap; watersheds in yellow are in the original 1172 but not in the
revised WARP set; and watersheds in blue are in the revised group only. 

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  10  Comparison
of the spatial distribution of the original vulnerable watersheds used
to select the monitoring sites with watersheds identified with the full
WARP equation using Dunne Overland Flow.

The extensive blue area is misleading because it includes larger HUC-8
watersheds where no HUC-10 watersheds were available for TX, OK, or AR.
All 40 of the watersheds selected for monitoring are included in both
sets of vulnerable watersheds. That does not preclude the possibilty
that the GRTS process would possibly have selected a different suite of
watersheds based on the revised WARP group.

  REF _Ref181767664 \h  Figure III-11  contrasts the locations of the
original 1172 HUCs selected with WARPDOFZ (yellow) with the upper 20th
%ile set of HUCs that would have been identified using the newer
atrazine use (WARP_NewUse), shown in red. The orange represents the
areas of overlap. In this case, 3 of the monitoring sites would not have
been in the most vulnerable tier with the new use data (1 in MN, 1 in
KY, 1 in NE).

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  11  Comparison
of original vulnerable watersheds identified using 1998-2002 use data
with vulnerable watersheds identified using 2001-03 use data.

Sensitivity / Uncertainty Analyses

This section focuses on factors that might impact how the results could
be interpreted or how future monitoring studies and data could be used
to assess the impacts of atrazine on aquatic communities. 

Weather

One of the assumptions of the population estimates (Section   REF
_Ref181719026 \w \h  III.D ) is that the chemographs are independent of
the year of sampling. Sources of variability in monitoring data
collected over multiple years include:

variability in weather patterns, particularly extreme events

timing of rainfall in  relation to application period

amount of flow in the water body prior to runoff events 

stark differences in rainfall amounts, timing, and intensity that make
comparisons among sites or among years at the same site difficult.

In the 2003 IRED addendum (US EPA, 2003b), the Agency specified that
monitoring might be extended for a third year at a site if either of the
following occurred:

atrazine concentrations in water triggered an LOC exceedance in any of
the sampling years, or

unusual meteorological conditions (e.g., high or low rainfall) occurred
during the monitoring period. 

The US EPA analyzed the rainfall data two ways (see Appendix 3):

Rainfall totals by month were compared against historical totals, with
special emphasis placed on rainfall amounts from April to June, which is
considered the principal atrazine use period and coincides with the bulk
of the analytical sampling;  

Daily precipitation data was plotted against the monitoring data and
overlaid the planting season for a more detailed evaluation. 

In the first analysis, the site-specific precipitation data were summed
by month and compared with historical monthly totals. A site year with
precipitation less than the 25th percentile for one or more of the
critical months is considered below normal precipitation and the results
from that year would be considered not representative of a normal, or
above normal climatic conditions.  Site years where data was between the
25 percentile and the 75th percentile were considered normal, and above
the 75th percentile were considered above normal.

  

In general, more sites had months with lower-than-normal than
higher-than-normal precipitation. In particular, 2005 was a low rainfall
year (based on the monthly rainfall comparisons for the Apr-Aug sampling
period) for a number of sampling locations, including nine sites in
Illinois, three in Indiana (IN-02, -06, -11), two in Ohio (OH-02, -03),
and one each in Kentucky (KY-02) and Minnesota (MN-01). Such rainfall
discrepancies may make it difficult to draw definitive conclusions
regarding the relative vulnerabilities of the water bodies to atrazine
runoff.  

However, lower-than-normal precipitation during the likely application
months does not necessarily mean that atrazine concentrations will be
correspondingly low. The timing of the precipitation is critical. If
precipitation is low in the time leading up to planting, then low-flow
conditions may exist in the streams and a rainfall event after
application could result in high concentrations because the relative
amount of dilution (runoff water into stream water) would be lower than
in a “normal” rainfall year with “normal” stream flow. The high
concentration (208 ppb) at IN-11 in 2005 is an example of this. In 2006,
when monthly rainfall totals at IN-11 were higher than normal, measured
atrazine concentrations were much lower than in 2005. 

A comparison of total precipitation against yearly or monthly averages
may not be a sufficient assessment of the potential for high atrazine
concentrations to occur in water body without an additional evaluation
of the timing of rainfall events and of stream flow at the time of the
runoff event. The US EPA has requested that Syngenta continue monitoring
for a third year in IL-03, IL-04, IL-08, IN-06, IN-11, MO-01, MO-02, and
OH-02, collecting data on stream flow as well as rainfall during the
monitoring period. 

Flow rates / Low flow sites

Low Flow Sites 

Sampling frequency is a critical issue for understanding uncertainty in
monitoring data and by extension the interpretation of the SSI scores
from CASM.  Although robust in both number of samples and geographic
diversity the AEMP data does include years where samples could not be
collected at some sites.  Of particular interest were sites monitored in
Nebraska.  Three of the seven sites in Nebraska experienced a higher
frequency of missed samples compared to other sites both within Nebraska
and across the entire AEMP geographic range.  According to the
registrant (Hampton, et al, 2007) three sites in Nebraska – NE-04,
NE-05, and NE-07 – experienced low to no-flow conditions that resulted
in significant number of missed grab samples in 2005.  

Because of the missing samples, Syngenta did not use CASM to evaluate
the three NE sites (Hampton, 2007a), suggesting that a dry stream would
experience “significant stressors” other than atrazine runoff. They
implied that these low- to no-flow events would make exposure to
atrazine irrelevant because the stressors on aquatic communities due to
low flow conditions are significant and would overwhelm these
communities (Hampton et al, 2007a, Hampton et al, 2007b).  

The issue raises several important questions.  First, if sites exhibit
low- or no-flow conditions such as those which may have been encountered
in the three Nebraska sites, does this preclude the analysis of these
data for evaluating aquatic community impacts?  Meyer et al (2007) have
demonstrated that headwater streams, including both low flow and
intermittent streams, can have significant biodiversity.  They note that
these types of streams are critical to the integrity of the entire river
network (Meyer, et al 2007).  Given this, the Agency believes the
question which should be asked is whether the type of streams
represented by the three Nebraska sites with missing samples due to
low/no flow conditions can be evaluated in the context of the microcosm
and mesocosm studies that are the foundation of the LOC (4% SSI response
from CASM).  

The second question is, if the microcosm/mesocosm studies are
representative of low/no flow sites (intermittent streams), what is the
impact of missing samples on interpreting CASM output?  If sites have a
“significant” number of missing samples, is it appropriate to use
these data in CASM?  A logical follow up question would be that if it is
inappropriate to use sites with missing data, then how many samples must
be missing to exclude the data from use?  

Finally, the Agency evaluated a series of landscape and watershed
metrics for any obvious explanations for the uniqueness of the three
Nebraska sites. These three sites with low flow conditions are in
southeastern Nebraska while other sites in western Nebraska did not
experience similar conditions.  Metrics evaluated included flow
information (both from AEMP and USGS Gages), stream order,
precipitation, base flow conditions, recharge potential, soils, geology,
irrigation use, and presence of tile drains.  Generally, there does not
appear to be an obvious explanation from any of these metrics accounting
for why these three sites should have experienced conditions unique from
the other 37 sites and, in particular, the other four sites in Nebraska
and nearby areas of Iowa and Missouri.  Interestingly, it did become
apparent during this evaluation that sites exceeding the LOC tended to
be lower flow sites. 

Given this uncertainty the Agency decided to conduct stair step
interpolation for the three sites and run each chemograph through CASM. 
The sites exceeded the CASM LOC, principally due to the fact that
several of the sample concentrations were high and interpolation using
the stair step approach extended these exposures well beyond the
standard 4 day gap, possibly overestimating the importance of individual
sample results.  For now, the three Nebraska sites are grouped
separately in the population estimates (see Section   REF _Ref181719026
\w \h  \* MERGEFORMAT  III.D ) in order to evaluate what these sites
would represent if they were considered to be exceeding.  Given the
uncertainty in the appropriateness of the use of these data, these
population estimates could be useful for determining how many sites
might be expected to experience similar conditions whether they be
interpreted as simply low/no flow or exceeding the LOC.

Flow Measured at the Monitoring Sites

Stream flow was the most obvious factor to compare between the
monitoring sites.Flow data for each year was summarized for each site by
calculating the annual average flow rate for each site by year (  REF
_Ref181697430 \h  Table III-12 ). 

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  12  Summary of
annual average flow rate for the 40 monitoring sites by year.

Site	2004 flow (cfs)	2005 flow (cfs)	2006 flow (cfs)

IA 01	50.20	69.95	37.70

IA 02	20.81	21.09	97.78

IL 01	102.25	15.05

	IL 02	71.71	56.08

	IL 03

46.34	37.70

IL 04

21.45	97.78

IL 05	102.89	83.12

	IL 06	30.02	20.49

	IL 07	29.71	20.82

	IL 08

146.39	125.51

IL 09	32.95	27.57

	IN 01	63.86	145.85

	IN 02	30.57	14.22

	IN 03

19.49	36.36

IN 04	42.96	14.77	12.55

IN 05	8.28	7.22	9.26

IN 06

132.89	213.52

IN 07

199.77	202.98

IN 08

27.60	42.37

IN 09

56.13	79.52

IN 10

15.23	29.35

IN 11

11.43	36.79

KY 01

6.55	14.30

KY 02

46.46	63.70

MN 01

28.33	16.46

MO 01	2.73	3.13	2.30

MO 02	28.30	25.86	29.69

MO 03	85.10	69.73	24.78

NE 01	42.85	57.50

	NE 02

84.83	37.90

NE 03	2.73	4.82

	NE 04

9.50	6.59

NE 05

35.22	22.35

NE 06	3.62	1.04	4.40

NE 07

15.37	8.56

OH 01	34.63	26.97

	OH 02

27.41	36.24

OH 03	21.14	21.42

	OH 04

192.55	208.70

TN 01

17.22	11.97



In addition, the 86 years of flow data were ranked relative to each
other (  REF _Ref182217040 \h  Table VII-3  in Appendix 2  REF
_Ref181697520 \h  ).  The three NE sites did not have the lowest average
flow: NE-04 in 2005 had the 15th lowest flow out of 86 site years, NE-07
in 2005 was 24th, and NE-05 was 51st.  Interestingly, three of the five
lowest flow site years were the three sample years forMO-01, which
exceeded the LOC each year.  

  REF _Ref180382723 \h  Figure III-12  through   REF _Ref180382745 \h 
Figure III-15  plot atrazine concentrations against measured stream flow
for NE-04, NE-05, NE-07, and MO-01 in 2005 to illustrate the
relationship between atrazine concentrations and flow conditions. In
these figures the missing samples are shown as 0. While   REF
_Ref180382723 \h  Figure III-12  and   REF _Ref180382745 \h  Figure
III-15  suggest a relationship between atrazine concentrations and flow
events, even it is a delayed response, the relationship is not as
evident in   REF _Ref181772715 \h  Figure III-13  and   REF
_Ref181076461 \h  Figure III-14 . One possible explanation is that
atrazine concentrations may be higher in some low-flow streams because
there is less receiving waters to dilute the atrazine load in runoff
waters. It is also possible that some unexplained feature of streambed
morphology could influence the ability to collect samples.

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  12  Atrazine
concentrations and measured stream flow, NE-04, 2005

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  13  Atrazine
concentrations and measured stream flow, NE-05, 2005

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  14  Atrazine
concentrations and measured stream flow, NE-07, 2005

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  15  Atrazine
concentrations and measured stream flow, MO-01, 2005

Sampling Frequency / Auto-samples

The AEMP data set represents a robust and targeted monitoring data set
available for atrazine with 4-day grab samples augmented with more
frequent flow triggered auto-samples.  The AEMP site averaged 36 samples
per year for 37 of the 40 sites with an average of 17 samples per year
for the three sites (NE-04, NE-05, and NE-07) that experienced more
missing samples than the others.  On average, this equates to roughly 7
samples per month during the sample season (generally from April to
September) while the sites with missing samples averaged 3 samples per
month during the sample season.  The AEMP has been input into CASM in
order to determine whether community level effects can be expected given
the duration and magnitude of atrazine exposures in each chemograph. 
Even with the robust nature of this data, CASM requires a complete 365
day chemograph for estimating effects levels.  Thus each AEMP chemograph
is augmented by interpolation between sample events and extrapolation to
estimate atrazine concentrations prior to the first sample and after the
last sample.  

Sampling frequency is a critical component of uncertainty analysis for
this evaluation because atrazine exposure and interpretation of
potential community responses relative to the microcosm and mesocosm
studies is driven by the magnitude and duration of exposure.  In this
construct the means of “filling in” (e.g. interpolation or
imputation) missing data is critical and thus the more data that is
missing (e.g. less frequent sampling) the greater uncertainty there will
be in determining if the LOC has been exceeded based on increasing
uncertainty in the magnitude and duration of atrazine exposure.  Thus an
understanding of the inherent uncertainty in the AEMP data is critical
to evaluating the risk conclusions drawn from CASM.  Sensitivity
analysis of the CASM model informed model uncertainty, however, this
analysis did not address the potential uncertainty in the AEMP data. 
Understanding uncertainty associated with the AEMP data due to
interpolation, missing samples, and less frequent sampling is critical
and can assist in interpreting uncertainty in the model estimates and
the amount of conservativeness of overall risk conclusions.  Also, a
better understanding of how these factors might influence model output
for the AEMP data can inform how other monitoring data not collected as
part of the AEMP might, or might not be reasonably interpreted when run
in CASM 

For the AEMP, the method chosen to interpolate or impute a ‘missing’
value is the stair step approach whereby a concentration at a given
sampling event is carried forward across the four day window to the next
sampling date.  Estimating atrazine concentrations pre- and
post-sampling occurs by extrapolation of the first sample backwards to
the beginning of the year and the last sample forward to the end of the
year except where a chemograph represents a second or third year at a
site, in which case the last sample result is carried beyond the end of
the year to the first sampling event of the following year.  Thus even
though this represents a robust data set, there is uncertainty
associated with the means of estimating the chemographs to obtain a 365
day profile.

The impact of the infilling process was evaluated to determine how
important the assumptions used to infill are on CASM output (e.g. risk
conclusions for each site).  First, sampling frequency was evaluated by
simulating a less frequent sampling design using the AEMP data.  The
ramifications of this are then compared with selected sites within the
AEMP where sampling frequency was less (e.g. the three Nebraska sites
with low flow conditions and missing data) and how missing samples or
less robust data might influence uncertainty.  Second, an alternative
approach to assessing uncertainty in different sampling frequencies was
considered.  Third, alternative approaches to infilling are considered
in the context of uncertainty.

Evaluation of Sampling Frequency Using AEMP Data

As noted in the analysis of the three Nebraska sites there will clearly
be instances where less robust monitoring data can be encountered.  The
following evaluations were completed to determine how important sampling
frequency is in the use of the CASM model relative to AEMP data, to
determine if uncertainty with missing samples can be characterized, and
by extension if monitoring data that is less robust can and/or should be
evaluated using CASM.

The CASM model uses 365 day chemographs as the exposure profile input
and thus even for the robust AEMP data some level of interpolation, or
infilling of missing data points, is required.  The method chosen for
interpolation in CASM is a stair-step approach by which each sample
result is carried forward across the un-sampled days until the next
sample result is reached.  The infilling then proceeds across the
profile until the last sample is reached.  How dates prior to the first
sample event and after the final sample event in a given year are
imputed raises an issue of uncertainty.  In their analysis, Syngenta
assumed that all days prior to the first event and after the last event
were 0.  The Agency’s approach is predicated on an assumption
(supported by other data sets such as the Heidelberg College data; Volz
et al, 2007) that there will be quantifiable exposure year round
including these early and late time frames.  Thus, the Agency has
estimated the last days of the 365 day profile with the results from the
last day sampled.  For the days prior to the first sample we have
estimated using two methods.  First, if no sampling has occurred in the
previous year, the result of the first sample date was assigned to every
day extending backwards to January 1.  If a previous year’s sampling
has occurred, the last value from the previous year is carried forward
to the first date of the following year.  Comparison of rolling average
estimation and CASM risk conclusions indicate that the Agency’s
approach increases the atrazine exposure compared to that used by the
registrant and increases the overall exposure profile slightly but does
not change the ultimate conclusion for the AEMP sites (the same sites
are identified to exceed the LOC using both approaches).

The second issue associated with interpretation of the AEMP data in CASM
is interpolation between sampling events and the sampling frequency of
individual chemographs.  CASM requires a 365 day profile and will
“interpolate” between sampling events where data is lacking. 
Options in CASM for interpolation include linear and stair-step
interpolation between sampling events.  In both instances, there is a
question of whether peak concentrations are being missed by 4 day grab
samples and what impact potentially missed pulses of atrazine exposure
might be having on model output based on a magnitude and duration of
exposure.

In order to test the importance of sampling frequency the Agency
conducted an analysis where data from one of the AEMP sites was reduced
to a 12-day grab sample profile.  First, the first sampling date was
intentionally skewed to miss the peak concentration from the 4 day grab
sample data.  Then an alternative 12 day profile was created which
included the peak concentration.  14-day, 30-day, 60-day, and 90-day
rolling average concentrations for each of the three profiles were then
calculated.

Comparisons for 2004 chemograph for IN-04 show a wide range in peak and
average concentrations depending on the length and timing of sampling
events (  REF _Ref182191164 \h  Table III-13 ). Missing the peak
concentration of 78 ppb resulted in much lower longer-term average
concentrations than those found with the 4-day grab samples while
hitting the peak concentration resulted in higher longer-term average
concentrations.  This analysis suggests that sampling frequency and when
the samples are collected can have a significant impact on how the
results are interpreted.  In this case, neither of the alternative 12
day grab sample profiles exceeded the LOC although the 12 day profile
that captured the peak was 3.8% compared to the 4% LOC and was well
within the 2x multiplication factor uncertainty bound for CASM. 

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  13  Comparison
of estimated atrazine concentrations between 4-day and 12-day grab
samples, IN-04, 2004

	Atrazine concentration (ug/L) and CASM SSI% for various average periods

	Peak	14-da	21-da	30-da	60-da	90-da	Annual Avg.	SSI%

4-day grab samples w/stair step interpolation	78.08	23.81	16.33	12.05
6.35	4.37	1.20	1.7

12-day grab samples w/peak EEC missed	2.26	2.04	1.60	1.48	1.06	0.88	0.34
0.3

12-day grab samples w/peak EEC captured	78.08	67.38	45.97	33.20	17.29
11.68	3.03	3.8



Auto Sample Results

In addition to 4-day grab samples, 13 sites were instrumented with
auto-samplers.  The auto-samplers were installed to collect 8-hour
composite samples triggered by increased flow events within a stream
site.  A total of 25 site years of data was collected using the auto
samplers.  These data were used to augment the 4-day grab time series to
investigate the impact of grab samples on uncertainty in the CASM runs
using 4-day samples. Both the 4-day grab sample chemographs and the
chemograph created from 4-day grab samples augmented with autosample
results were interpolated/extrapolated to create 365 day profiles.  

In order to conduct this analysis the 4-day time series was compared
with the auto-sample results for each site-year of data and the
auto-sample results were added to the time series.  For days with more
than one auto-sample result the highest value from that day was used as
though that value was the grab sample for that day.  The auto-sample
results were compared against the 4-day grab samples and where an
auto-sample event occurred on the same day as a 4-day grab sample the
highest of the two sample types was used in the distribution.  If the
auto-sample event occurred on a day when no 4-day grab sample was
collected then the auto-sample result was placed in the time series on
the day it was collected.  Data interpolation then proceeded as with the
4-day samples with the exception being that if there were un-sampled
days after an auto-sample event these days were assumed to have the same
value as the auto-sample result.  In essence, the auto-sample results
were used to fill exposure gaps between the 4-day grabs and then used as
part of the stair-step interpolation process in the same manner as the
4-day grab sample interpolation.

In order to test the robustness of the 4 day samples an analysis was
conducted whereby various rolling average concentrations (  REF
_Ref182191200 \h  Table III-14 ) were calculated for each site by year
for the 4 day grab samples augmented with auto-samples (the 4 day grab
sample rolling averages and SSI% were summarized previously in   REF
_Ref182368984 \h  Table III-5 .  

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  14  4-day Grab
Samples Augmented with Auto Samples - Rolling Averages (ppb)

	Maximum atrazine concentration, ug/L (ppb) over averaging periods of

Site	Year	Peak	14 days	21 days	30 days	60 days	90 days	365 days 

IA 02	2004	5.37	2.41	1.83	1.84	1.26	0.92	0.17

IA 02	2005	5.53	2.14	1.53	1.39	0.85	0.63	0.58

IL 06	2004	5.26	1.07	0.92	0.74	0.53	0.42	0.19

IL 06	2005	0.23	0.15	0.14	0.13	0.12	0.11	0.09

IL 08	2005	24.67	9.22	7.11	5.82	3.53	2.61	0.55

IL 08	2006	50.70	11.74	8.06	10.06	5.54	3.78	0.84

IN 06	2005	7.23	3.85	2.93	2.31	1.22	0.88	0.24

IN 06	2006	24.30	5.73	4.79	3.87	2.39	1.79	0.47

IN 09	2005	34.49	8.81	6.08	5.40	3.45	2.46	0.40

IN 09	2006	13.80	4.27	3.13	2.85	1.93	1.37	0.31

IN 11	2005	237.50	65.13	44.41	31.51	16.18	11.32	3.80

IN 11	2006	15.89	5.90	4.36	3.34	1.92	1.37	0.83

MN 01	2005	15.03	5.10	3.62	2.61	1.41	1.04	0.33

MN 01	2006	0.22	0.16	0.14	0.13	0.12	0.11	0.10

MO 01	2004	65.94	28.83	21.92	22.08	15.43	11.10	3.20

MO 01	2005	182.75	78.06	54.19	45.66	26.98	19.18	4.98

MO 01	2006	106.00	45.11	39.06	30.20	17.01	11.69	3.09

MO 03	2006	20.31	5.00	3.82	3.12	2.10	1.68	0.86

NE 04	2006	125.00	21.08	14.10	13.98	8.00	5.47	1.45

NE 05	2006	6.76	5.29	4.49	3.94	3.18	2.22	0.63

NE 06	2004	10.99	3.08	2.75	2.20	1.70	1.56	0.45

NE 06	2005	36.13	19.29	13.01	10.85	5.79	3.95	1.05

NE 06	2006	0.13	0.11	0.11	0.10	0.10	0.10	0.10

OH 03	2004	21.50	8.14	7.60	6.97	4.25	2.86	0.74

OH 03	2005	15.84	4.06	2.78	2.43	1.29	0.90	0.28



Once the chemographs were augmented with auto-sample results and
interpolated/extrapolated to a 365 day profile using the stair-step
approach each data set was input into CASM where a SSI% and
multiplication factor (MF) was estimated.  The SSI% and MF for the
auto-sample augmented chemographs were then compared with the original
chemograph SSI% and MF to understand what influence the incorporation of
continuous auto-sample results into the profile had on model behavior ( 
REF _Ref182191236 \h  Table III-15 ).  While none of the auto-sample
augmented profiles exceeded the 4% SSI LOC (except where the original
chemograph exceeded) the overall influence on model response was an
average 64% increase in SSI% and a 18% decrease in MF across all sites
with auto-samplers.

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  15  Percent
Difference Between Grab Averages and Auto Sample Adjusted Averages using
CASM SSI%

Site Name	Auto Sampled Sites	Four Day Sites	Percent Difference

	AD365	MF	AD365	MF	AD365	MF

IA02 2004	0.35	12.22	0.14	18.75	150	-35

IA02 2005	0.33	14.06	0.28	16.13	18	-13

IL06 2004	0.27	15.25	0.14	21.00	93	-27

IL06 2005	0.00	13.32	0.00	27.16	0	-51

IL08 2005	1.13	3.63	0.61	5.70	85	-36

IL08 2006	1.91	2.87	1.62	3.72	18	-23

IN06 2005	0.37	11.47	0.28	14.45	32	-21

IN06 2006	0.87	5.44	0.68	6.50	28	-16

IN09 2005	1.19	4.16	0.56	7.72	113	-46

IN09 2006	0.64	7.09	0.43	9.80	49	-28

IN11 2005	5.64	0.71	5.09	0.78	11	-9

IN11 2006	1.32	3.50	0.53	7.81	149	-55

MN01 2006	0.00	20.40	0.00	20.40	0	0

MN01 2005	1.15	3.84	0.22	13.75	423	-72

MO01 2004	4.44	0.85	5.04	0.70	-12	22

MO01 2005	6.80	0.53	6.31	0.57	8	-7

MO01 2006	4.43	0.87	4.48	0.84	-1	3

MO03 2006	0.96	4.78	0.78	5.41	23	-12

NE04 2006	2.14	2.10	0.44	7.97	386	-74

NE05 2006	0.78	4.54	1.05	2.77	-26	64

NE06 2004	0.54	7.31	0.53	7.13	2	3

NE06 2005	1.94	2.85	2.04	2.75	-5	4

NE06 2006	0.00	20.49	0.00	20.49	0	0

NE07 2006	0.55	8.46	na	na	na	na

OH03 2004	1.41	3.91	1.46	3.66	-3	7

OH03 2005	0.54	11.47	0.36	15.38	50	-25





Average	64	-18



In general, overall rolling average concentrations increased across all
sites and the overall change in CASM SSI% was 64% across all sites.  The
analysis indicates that incorporation of continuous sampling is likely
to increase the overall exposure profile vis. the profile presented by
grab samples, and that use of less frequent [grab] samples is likely to
underestimate exposure. However, none of the revised chemographs
exceeded the LOC (4% SSI) except for sites already above the LOC (e.g.
MO 01).

Syngenta PRZM Augmentation

In Snyder et al (2007) Syngenta presented additional analysis on the
importance of data interpolation.  This analysis involved augmentation
of the 4-day grab sample profile using the Pesticide Root Zone Model
(PRZM) (Carsel et al, 1998) to predict edge of field concentrations with
site-specific climate data and application information.   Predicted peak
concentrations were matched to 4-day grab sample results to establish
dilution and dissipation factors.  The remaining PRZM concentrations not
corresponding with 4-day grab samples were then “augmented” to the
time series profile.  Syngenta completed this analysis for all 40 sites
and years monitored (80 site years in all) available at the time of
Snyder (2007).  Each PRZM augmented chemograph was then run through CASM
to generate rolling average concentrations for 14-day, 30-day, 60-day,
and 90-day durations as well as calculated SSI%.  More detail on the
methodology and assumptions employed in this evaluation are presented in
Snyder (2007).  A graphical representation of how the analysis was
performed is presented in   REF _Ref182215828 \h  \* MERGEFORMAT  Figure
III-16 .  

Figure   STYLEREF 1 \s  III -  SEQ Figure \* ARABIC \s 1  16  Graphical
Representation of PRZM Augmentation of a 4-Day Grab Sample Time series
(taken from Syngenta Presentation Titled “Atrazine Ecological
Monitoring Program review” dated December 14, 2006)

 

Additional analysis of this evaluation was conducted by the Agency and
consisted of comparing the rolling average concentrations and SSI% from
the original un-adjusted chemographs and the PRZM augmented chemographs.
 The SSI% are presented for completeness but not used for further
analysis because the SSI% were created using a different version of CASM
than currently being evaluated and because the majority of the 80
chemographs remained unchanged at 0% SSI (24 or 30% of all site years)
or increased but the original SSI% was 0 (36 or 45% of all site years)
and no percent difference could be calculated.  For all 80 chemographs
the percent change in rolling average concentrations and SSI% (where
values greater than 0 were available for both original and augmented
chemographs) were calculated.  The PRZM augmented rolling average
concentrations and SSI% is presented in   REF _Ref182215973 \h  Table
VII-4  in Appendix 2 while the percent change from the original rolling
average and SSI is summarized   REF _Ref182215941 \h  Table III-16 .  
The original rolling average and SSI% are presented in   REF
_Ref181177887 \h  Table III-5 . 

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  16  Summary of
Percent Differences Between Original Rolling Average Concentrations and
SSI% and revised Rolling Average Concentrations (ppb) and SSI% for PRZM
Augmented Chemographs from Snyder, et al (2007)

Site	Year	% Difference in the



14-day rolling average	30-day rolling average	60-day rolling average
90-day rolling average	SSI%

IA 01	2004	180	273	323	284	a

IA 01	2005	842	767	587	387	b

IA 02	2004	604	535	485	402	a

IA 02	2005	100	135	235	210	a

IL 01	2004	22	43	56	48	b

IL 01	2005	940	1180	1320	900	a

IL 02	2004	367	251	252	235	a

IL 02	2005	32	9	23	62	b

IL 03	2005	384	437	368	443	a

IL 03	2006	287	292	273	435	a

IL 04	2005	103	163	122	113	b

IL 04	2006	105	222	166	144	a

IL 05	2004	66	153	173	182	43

IL05	2005	459	430	267	300	b

IL 06	2004	524	763	652	560	a

IL 06	2005	1095	1040	710	460	a

IL 07	2004	78	120	129	122	a

IL 07	2005	653	610	302	348	a

IL 08	2005	61	94	137	129	a

IL 08	2006	58	40	50	78	140

IL 09	2004	54	75	72	70	a

IL 09	2005	16	57	61	95	600

IN 01	2004	232	158	223	217	a

IN 01	2005	226	220	281	276	a

IN 02	2004	32	60	65	55	a

IN 02	2005	-7	-8	-3	1	b

IN 03	2005	-2	-1	0	11	b

IN 03	2006	7	22	66	64	a

IN 04	2004	26	49	68	64	24

IN 04	2005	388	338	443	437	a

IN 04	2006	101	80	83	78	a

IN 05	2004	2	5	17	20	-44

IN 05	2005	2	25	-3	15	a

IN 05	2006	-31	-19	-6	-1	-68

IN 06	2005	19	18	12	27	b

IN 06	2006	24	27	15	31	b

IN 07	2005	-29	-16	-8	-4	-69

IN 07	2006	65	57	121	121	a

IN 08	2005	0	0	0	10	b

IN 08	2006	25	19	25	37	a

IN 09	2005	213	184	245	225	a

IN 09	2006	63	70	89	107	b

IN 10	2005	-1	30	43	40	b

IN 10	2006	49	42	42	49	a

IN 11	2005	-30	-29	-23	-21	-15

IN 11	2006	24	58	71	66	b

MO 01	2004	-29	-24	-14	-14	-15

MO 01	2005	-16	-13	-10	-33	-41

MO 01	2006	-8	-2	10	10	-5

MO 02	2004	-5	-1	1	2	-2

MO 02	2005	5	17	25	21	144

MO 02	2006	-13	-8	0	-2	-4

MO 03	2004	-18	-12	-6	-4	-6

MO 03	2005	20	24	21	18	b

MO 03	2006	125	69	44	80	b

NE 01	2004	3	4	3	6	b

NE 01	2005	26	30	42	40	a

NE 02	2005	-5	-9	-1	0	a

NE 02	2006	-2	13	14	14	6

NE 03	2004	557	301	227	220	a

NE 03	2005	5	12	21	19	b

NE 06	2004	92	49	67	44	a

NE 06	2005	-29	-19	-16	-16	-71

NE 06	2006	1440	1310	640	420	a

KY 01	2005	25	27	21	33	b

KY 01	2006	7	38	67	79	-71

KY 02	2005	23	17	17	16	a

KY 02	2006	60	82	87	82	b

MN 01	2005	97	172	186	185	a

MN 01	2006	1170	1460	950	770	a

OH 01	2004	-9	6	16	19	b

OH 01	2005	108	86	130	82	b

OH 02	2005	8	13	50	74	300

OH 02	2006	23	26	44	43	b

OH 03	2004	-1	5	45	41	a

OH 03	2005	42	115	102	105	a

OH 04	2005	-2	35	83	107	1600

OH 04	2006	123	171	230	233	a

TN 01	2005	-12	-6	5	5	b

TN 01	2006	1	11	-8	1	b

a – 0% SSI in both original and augmented chemographs

b – 0% SSI in original chemograph

Finally, the percent differences were ranked and percentile generated
for each duration and the SSI% separately.  In addition, a single
percentile distribution was created for all four rolling averages.  The
result of this analysis is presented in   REF _Ref182216184 \h  Table
III-17 .  

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  17  Percentile
of Ranked Distribution of Percent Differences from Table III-16

	14-day rolling average	30-day rolling average	60-day rolling average
90-day rolling average	SSI%1	All rolling averages

average	151	162	145	132	122	148

median	26	42	61	66	-4	49

99th	1224	1340	1024	796	1410	1280

95th	842	767	640	443	650	653

90th	524	437	368	402	330	436

75th	108	163	186	185	67	171

50th	26	42	61	66	-4	49

25th	1	11	15	18	-42	10

10th	-12	-8	-3	0	-69	-6

5th	-29	-16	-8	-4	-71	-15

1st	-31	-25	-17	-24	-71	-29

1 – SSI% percent difference distributions are suspect due to limited
number of suitable data.

This overall analysis suggests that augmentation of the original 4-day
grab sample chemograph using a modified PRZM approach can increase
thresholds by roughly 50% at the median and as much as 150% for the
average.  This analysis is consistent with the comparison of grab sample
only chemographs versus grab sample augmented with autosample results
chemographs that suggests that 4 day grab sample profiles by themselves
are likely to miss peak exposures and thus underestimate overall
exposure. This does not suggest that all sites will increase by this
amount using this type of approach to augment a time series but it does
provide context to the uncertainty in using a stair step interpolation
method for the 4-day grab samples.

Evaluation of Alternate Sampling Frequency Strategies

Application of the CASM model beyond the AEMP will be highly influenced
by the type of sample program implemented as well as type of streams
being assessed.  Setting aside stream type which is addressed elsewhere,
the analysis presented above clearly suggests that sample programs less
robust than the AEMP are likely to miss peak concentrations and may
result in an under-estimation of exposure.  Many monitoring programs are
being conducted for the implementation of TMDL programs and for the
evaluation of the establishment of aquatic life criteria.  It is highly
unlikely that these sample designs will include the level of monitoring
captured in the AEMP.

One possible solution to the uncertainty associated with differing
sample frequencies may be found in the work of Crawford (2004). 
Crawford (2004) used a robust data set of 10 years of atrazine data
collected by Heidelberg College Water Quality Laboratory (WQL) from four
streams in Ohio was linearly interpolated to provide a synthetic
chemograph of hourly observations.  Crawford (2004) conducted a Monte
Carlo assessment using the Heidelberg data to approximate different
sampling strategies ranging from a limited sampling consisting of one
sample per quarter to a more robust program of 10 samples per month. 
The probabilistic assessment ran 1,000 iterations and error
distributions based on percent difference between computed and actual
mean, 90th, 95th, and 99th percentile concentrations.  As expected, the
analysis shows that as sampling frequency increases the difference
between actual and sampled values decreases.  A summary of the various
monitoring programs evaluated by Crawford (2004) and the +/- differences
at the 10th and 90th percentiles are presented in   REF _Ref181071264 \h
 Table III-18  and   REF _Ref181071320 \h  Table III-19 .

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  18  Sampling
frequency strategies used by Crawford (2004)

Abbreviation1	Samples per Year	Samples per Month during Runoff Period

Q	4	1

W7	7	1

W10	10	2

M	12	1

W14	14	2

W22	22	4

M2	24	2

W462	46	10

M4	48	4

M10	120	10

		1 – Detailed descriptions of sample strategies may be found in
Crawford, 2004

		2 – Sample strategy W46 most closely matches AEMP

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  19  Error
Distribution for Various Sample Strategies of Crawford (2004) for
Different Concentration Profiles

	Mean Concentration	90th % Concentration	95th % Concentration	99th %
Concentration

Site	Method	10th %	90th %	10th %	90th %	10th %	90th %	10th %	90th %

Honey Creek	Q	-77.18	98.70	-79.54	213.79	-91.46	48.56	-97.72	-36.24

	W7	-48.70	66.14	-70.15	65.97	-54.91	215.90	-84.40	0.12

	W10	-37.47	49.94	-36.07	42.27	-49.99	40.44	-74.46	14.50

	M	-49.68	69.82	-59.79	35.26	-50.32	214.69	-84.96	3.60

	W14	-38.71	35.94	-35.26	34.61	-50.16	33.22	-74.70	10.83

	W22	-21.74	20.06	-19.90	28.96	-30.36	21.97	-48.23	69.39

	M2	-38.80	47.36	-35.60	39.11	-50.12	33.49	-74.33	29.90

	W46	-9.02	10.85	-13.16	10.43	-17.01	19.05	-32.34	7.61

	M4	-23.29	15.22	-15.77	27.96	-28.89	19.00	-52.02	58.51

	M10	-12.82	7.34	-12.66	15.33	-14.31	18.32	-41.83	6.31

Maumee River	Q	-68.50	83.64	-76.56	105.18	-86.91	22.96	-93.66	-16.57

	W7	-29.16	26.37	-59.24	48.51	-33.64	52.33	-69.42	-0.69

	W10	-14.73	26.17	-31.84	26.35	-37.52	11.80	-43.41	5.09

	M	-27.13	26.67	-56.22	15.50	-31.51	63.34	-62.37	0.41

	W14	-17.66	16.70	-32.15	15.26	-36.16	8.66	-43.57	4.59

	W22	-9.18	8.72	-16.64	25.50	-17.12	9.12	-22.01	15.55

	M2	-17.55	17.23	-29.31	19.41	-38.95	12.69	-43.20	7.37

	W46	-3.42	6.30	-7.41	5.77	-8.18	6.37	-18.27	2.75

	M4	-7.33	6.56	-14.23	29.85	-14.12	8.31	-22.66	16.60

	M10	-2.99	1.30	-8.99	4.47	-6.51	6.40	-18.72	2.75

Rock Creek	Q	-79.62	138.39	-82.18	327.73	-92.57	71.38	-97.59	-38.85

	W7	-59.24	65.53	-72.11	176.81	-61.16	198.07	-87.08	-10.84

	W10	-41.47	44.17	-48.40	66.81	-56.96	48.68	-73.12	17.40

	M	-58.29	84.30	-68.55	70.73	-63.04	249.73	-89.30	4.42

	W14	-42.95	32.08	-41.81	45.60	-59.16	41.70	-73.23	6.54

	W22	-28.33	28.20	-27.45	41.90	-31.46	39.77	-54.50	66.09

	M2	-44.46	47.89	-38.90	60.55	-58.63	54.55	-72.53	33.67

	W46	-11.35	13.34	-18.34	17.04	-18.20	20.10	-29.03	9.24

	M4	-27.61	20.80	-26.37	34.03	-34.32	42.30	-55.79	58.42

	M10	-6.20	13.20	-5.64	17.09	-16.44	30.79	-32.79	5.09

Sandusky River	Q	-73.39	104.09	-78.24	173.84	-89.14	38.23	-95.61	-27.29

	W7	-44.29	54.28	-68.72	63.49	-46.41	94.62	-80.06	3.32

	W10	-26.11	34.14	-44.31	78.33	-47.57	31.43	-52.67	12.70

	M	-40.06	44.07	-64.72	33.93	-43.37	110.12	-76.09	3.76

	W14	-27.50	25.15	-44.05	40.15	-47.26	26.84	-54.63	10.40

	W22	-15.51	18.96	-17.63	41.40	-23.50	30.03	-34.11	31.54

	M2	-28.93	26.86	-39.14	44.16	-48.33	30.17	-53.83	15.93

	W46	-6.10	10.70	-8.57	13.21	-10.93	16.72	-21.08	9.50

	M4	-15.34	15.12	-17.87	41.97	-21.32	30.19	-33.32	28.75

	M10	-6.99	6.45	-8.54	12.56	-9.10	20.99	-22.11	11.39



The data and analysis in Crawford (2004) were evaluated with two
objectives.  First, can the error estimates provide additional
information on the uncertainty associated with a sampling frequency
comparable to the AEMP relative to a continuous 365 day exposure profile
(e.g. an analysis similar to the grab versus autosample analysis above)?
 In this case, it is assumed that sampling strategy W46 from Crawford
(2004) is the closest approximation of the AEMP.  It is then assumed
that the error estimate for this strategy at the 99th percentile
concentration can provide context to the autosample analysis above.  In
essence the Heidelberg data approximates the autosample augmented
profile described above.  The error estimates for W46 at the 99th
percentile range from 18 to 32%.  This comparison is limited because the
autosample analysis above includes interpolation between events while
the Crawford (2004) data does not estimate the intermittent peak
concentration.

In the summary above the potential for underestimation of exposure in a
sampling design such as W46 can vary by as much as 30% depending on
target concentration (e.g. 99th percentile of the yearly distribution)
keeping in mind that the error estimate represents the uncertainty in
the ability of a sampling strategy to capture a certain percentile of a
365 day time series.  This analysis suggests that it may be possible to
estimate how likely a given sampling strategy will capture a given
target exposure.  Interestingly, the error estimates for W46 are similar
to the estimated uncertainty seen when comparing 4-day grab sample
chemograph with autosample augmented chemographs.

An additional value in the Crawford (2004) data is the ability to
compare uncertainty in different sampling frequencies (e.g. monitoring
other than the AEMP).  As an example of how this approach might work the
following table (  REF _Ref181698327 \h  Table III-20 ) was created to
show the difference in error estimates for the various sampling
strategies for the annual mean concentration relative to W46.  The W46
sampling strategy evaluated by Crawford (2004) best represents the AEMP
and thus could form the foundation for comparison.  It should be noted
that the error estimates provided below do not include the influence of
interpolation/extrapolation used on the AEMP.  Thus the analysis is
useful for comparing uncertainty in raw data (e.g. data that has not
been interpolated to a 365 day profile) from sampling scheme similar to
the AEMP with other less frequent sample strategies that could
potentially be input in CASM.  This table provides a clearer picture of
the trend summarized above suggesting that as sampling frequency
decreases the error differential increases.  In this example, it could
be interpreted that a sampling strategy approximated by the monthly
sampling strategy (i.e. method M below) would be expected to
under-predict exposures relative to W46 by 40%.  

It is suggested that the different error distributions surrounding the
various sampling strategies will allow for an estimation of the
uncertainty in less robust data.  For example, the percent difference
for one watershed (Sandusky River) in Crawford (2004) for the annual
time weighted mean concentration when sampling 10 times per month ranged
from -13% at the 5th percentile up to roughly +7% at the 95th percentile
while quarterly sampling yielded a range of -77% to 152%.  Additionally,
it should be possible to provide uncertainty bounds on less robust
monitoring data based on the differences seen in the Crawford data.  It
is speculated that these data can allow for an uncertainty factor to be
put in place which, when monitoring yields exposure duration profiles
within a certain bound of the LOC established by CASM, would be similar
to the MF approach based on the model sensitivity presented in Section
II.

Table   STYLEREF 1 \s  III -  SEQ Table \* ARABIC \s 1  20  Variability
of estimated concentrations based on different sample frequencies
(Crawford, 2004)

Sample Site Name	Sample Method Code

	% Difference from W46



p10	p90	p10	p90

Honey Creek	Q	-77.18	98.70	-68	88

Honey Creek	W7	-48.70	66.14	-40	55

Honey Creek	W10	-37.47	49.94	-28	39

Honey Creek	M	-49.68	69.82	-41	59

Honey Creek	W14	-38.71	35.94	-30	25

Honey Creek	W22	-21.74	20.06	-13	9

Honey Creek	M2	-38.80	47.36	-30	37

Honey Creek	W46	-9.02	10.85	0	0

Honey Creek	M4	-23.29	15.22	-14	4

Honey Creek	M10	-12.82	7.34	-4	-4

Maumee River	Q	-68.50	83.64	-65	77

Maumee River	W7	-29.16	26.37	-26	20

Maumee River	W10	-14.73	26.17	-11	20

Maumee River	M	-27.13	26.67	-24	20

Maumee River	W14	-17.66	16.70	-14	10

Maumee River	W22	-9.18	8.72	-6	2

Maumee River	M2	-17.55	17.23	-14	11

Maumee River	W46	-3.42	6.30	0	0

Maumee River	M4	-7.33	6.56	-4	0

Maumee River	M10	-2.99	1.30	0	-5

Rock Creek	Q	-79.62	138.39	-68	125

Rock Creek	W7	-59.24	65.53	-48	52

Rock Creek	W10	-41.47	44.17	-30	31

Rock Creek	M	-58.29	84.30	-47	71

Rock Creek	W14	-42.95	32.08	-32	19

Rock Creek	W22	-28.33	28.20	-17	15

Rock Creek	M2	-44.46	47.89	-33	35

Rock Creek	W46	-11.35	13.34	0	0

Rock Creek	M4	-27.61	20.80	-16	7

Rock Creek	M10	-6.20	13.20	5	0

Sandusky River	Q	-73.39	104.09	-67	93

Sandusky River	W7	-44.29	54.28	-38	44

Sandusky River	W10	-26.11	34.14	-20	23

Sandusky River	M	-40.06	44.07	-34	33

Sandusky River	W14	-27.50	25.15	-21	14

Sandusky River	W22	-15.51	18.96	-9	8

Sandusky River	M2	-28.93	26.86	-23	16

Sandusky River	W46	-6.10	10.70	0	0

Sandusky River	M4	-15.34	15.12	-9	4

Sandusky River	M10	-6.99	6.45	-1	-4



At this point, this analysis is not proposed specifically as a metric
for others implementing lower frequency sampling to use but as an
outline of a potential approach.  Additional tools for interpolation of
daily concentrations from temporal monitoring may include kriging,
conditional simulation (stochastic sequential simulation), and  time
series interpolation techniques (Deutsch and Journel, 1998; Isaaks and
Srivastava, 1989).   As Crawford (2004) suggests, use of these tools
requires an analysis of data stationarity (constant mean and variance)
and development of correlation functions using semi-variagrams and
correlograms.  None of these approaches have been evaluated as an
alternative to the stair step method of interpolation at this time.  

Conclusions

The AEMP represents one of the most robust and targeted monitoring data
sets for atrazine submitted to the Agency.  However, no large-scale
monitoring can provide continuous exposure data and the analysis above
suggests that there is still uncertainty with even a robust data set
such as the AEMP.  The comparison of sampling frequency results suggests
that sampling design is critical to incorporation of a chemograph (i.e.
monitoring data profile) into the LOC evaluation.  Monitoring programs
that are less robust than AEMP will have greater uncertainty in
capturing maximum peak and even some longer-term exposures.  Analysis of
grab versus auto sample results suggest that auto-samplers provide an
excellent choice for replacing or augmenting a grab sample program but
resources may limit this approach.  The auto sample analysis conducted
above suggests that some uncertainty is expected even with a robust data
set like the AEMP but the analysis also suggests that this type of
uncertainty can be quantified.  In addition, Syngenta has evaluated an
approach of augmenting monitoring data with PRZM estimates that also
suggests that even data as robust as the AEMP can underestimate
continuous exposure.  Finally, the analysis of sampling frequency using
the data from Crawford (2004) suggests that error bounds on a sampling
design may be appropriate for less robust monitoring data.  This could
be a key element to allowing application of uncertainty criteria to
monitoring results. 

SAP Charge Questions on the Atrazine Monitoring Results

(1)	The monitoring program used a tool (WARP) designed to assess the
vulnerability of watersheds and stream segments to (1) identify
watersheds within the corn/sorghum growing region that are likely to be
most vulnerable to atrazine exposure and, (2) select sampling sites
within the watersheds that are likely to be more susceptible to atrazine
runoff.  

Please comment on the use of WARP predictions for hydrologic units (HUC
10/11) to restrict the survey design to those HUCs in the upper 20th
percentile and then (1) to stratify by WARP predictions between 80th –
95th percentiles and above 95th percentile and (2) to select HUCs with
probability proportional to higher atrazine use rates.

Comment on the use of survey design population estimation approach for
estimating the number (and %) of HUCs that may have LOC exceedances.

(2)	Once the vulnerable HUC 10/11 watersheds were selected for
monitoring, specific monitoring sites were selected within each
watershed using criteria that were designed to maximize the potential
for selecting the streams most vulnerable to atrazine exposure. 
However, with only a single point monitored per watershed, estimates of
within-HUC variability for detections of atrazine could not be
calculated. The resulting population estimates reflect variability
across watersheds but not within the monitored watersheds. Please
comment on this approach and identify and discuss any alternative
approaches to extend the results of the monitoring sites.

(3)	Three monitoring sites in NE experienced low- or no-flow conditions
that precluded sampling. While Hampton et al. (2007a) suggest that these
sites with intermittent or low flow are already stressed by other
factors, Meyer et al. (2007) indicate that such aquatic communities are
rich in diversity. The Agency has generated statistics for these three
sites as a separate stratum, however the meaning of these separate
population estimates is uncertain.

Please comment on whether the Agency should consider the low flow sites
and/or intermittent streams as a part of the population estimates or
treat them separately. 

Please comment on whether the aquatic systems and exposure conditions of
the existing microcosm and mesocosm studies adequately represent these
low flow and/or intermittent stream communities.  If not, how could EPA
determine an LOC for low flow conditions?

(4)	The monitoring study sampled for atrazine concentrations at 4-day
intervals to characterize the atrazine chemograph in these low-order
Midwestern streams. The CASM_Atrazine model used these chemographs with
a stair-step interpolation between samples dates to relate atrazine
exposures in the streams to microcosm/mesocosm studies in order to
determine whether the exposures triggered LOC thresholds.

What other approaches for interpolation should be considered?  Given the
concentration-duration endpoint, how frequently must sampling occur to
appropriately capture the magnitude and durations of exposure associated
with atrazine?

Sensitivity analysis of CASM_Atrazine model inputs suggests that some
uncertainty bound on model results is appropriate.  The Agency used a 2x
multiplication factor from the model sensitivity analysis  to estimate
uncertainty in model output.  The sample frequency analysis indicates
that there is uncertainty associated with monitoring data that may not
be accounted for by the model uncertainty factor of 2x.  Given the
importance of sample frequency and interpolation, please comment on
whether consideration should be given to placing additional uncertainty
bounds on monitoring data to account for uncertainty in the ability of
the sampling strategy to capture the magnitude and duration of atrazine
exposures. Please provide any suggestions for how to proceed with this
approach.

 Approaches To Address The Question “Where Are The Waters That Are
Exceeding Effects-Based Atrazine Thresholds?”

From Watersheds To Waterbodies

The monitoring study design was based on a watershed vulnerability
assessment. As noted earlier, it addresses the question regarding how
many vulnerable watersheds in corn and sorghum growing regions in the
United States exceed the LOC for atrazine in at least one of its
subwatersheds.  However, ultimately, the value in the results of the
study is in identifying water bodies (stream miles) where atrazine
concentrations exceed the LOC.  While the initial results of the study
indicate that, for example, 9% (0 to 19%, based on a 95th percent
confidence bound) of the HUC-10 watersheds in this vulnerable tier had
streams that triggered the LOC for primary producers in multiple years,
that does not necessarily imply that all of the flowing waters in those
watersheds exceed the LOC. The results imply that conditions exist
within those HUC-10 watersheds that trigger an LOC in at least some of
the waters. 

Since watershed-based approaches make sense for addressing nonpoint
sources of impact, that is where the follow-up actions should be
focused. Part of the follow-up action specified in the atrazine IRED
addendum was to identify other areas in the larger HUC-10 watershed that
might have similar conditions and concerns. This work is ongoing and
includes more detailed watershed study, discussions with local soil and
water quality experts, and additional monitoring. 

The US EPA is exploring several options for addressing the question,
“How many stream miles/kilometers exceed the established LOC?” These
options are briefly summarized here and will be explored in more detail
in a future SAP. For now, the Agency is seeking feedback and
recommendations from the SAP on these or other viable options.

A first approach to identifying stream segments within the HUC-10
watersheds that exceed the LOC for atrazine would be to map out the
stream segments that met the initial selection criteria for monitoring
site locations (described in Section   REF _Ref181163359 \w \h 
III.B.1.c)  and assume that, as a first cut, this represents the extent
of stream segments that similarly exceed the LOC. 

  REF _Ref181414541 \h  Figure IV-1  (taken from Harbourt et al, 2004)
shows the extent of stream segments that met the sample selection
criteria with a breakdown of stream segment miles that met the criteria.


Figure   STYLEREF 1 \s  IV -  SEQ Figure \* ARABIC \s 1  1  Extent of
stream segments in the MO-02 HUC that met the sample selection criteria
(from Harbourt et al, 2004).

This approach has several limitations:

The stream segments were generated using digital elevation models
(Harbourt et al, 2004) and, thus, may not exactly correspond to actual
streams in the watershed. In addition, such an approach will be
dependent not only on the source of the data, but also on the scale.

This approach assumes that the specific criteria used for site selection
are valid predictors of atrazine loading in streams within a larger
HUC-10 watershed.

Limiting the focus to stream segments that meet the criteria fails to
address the extent to which atrazine loads that exceed the LOC continue
downstream at concentrations sufficient to exceed the LOC. It also
assumes that concentrations upstream of the identified segments are less
than those that would trigger the LOC.

Localized subwatershed characteristics are assumed to be captured by the
watershed-scale WARP vulnerability assessment approach.

The limitation regarding the synthetic nature of the stream segments and
data source/scale issues can be addressed by using a consistent
hydrography database. Since the initiation of this atrazine monitoring
study, an improved version of the National Hydrography Dataset (NHD),
called NHDPlus, has been released (HSC, 2006). NHDPlus includes an
increased set of value-added attributes and tools that allow for better
catchment characterization, stream network and flow evaluations, and
user-added capabilities (HSC, 2006). The US EPA plans to take advantage
of the catchment characteristics, flow network capabilities, and tool
development options to

Link the USGS WARP model to the NHDPlus stream segments and map WARP
values to stream segments (thus the vulnerability approach could be
applied ot the NHDPlus stream segments)

Accumulate and map stream catchment characteristics, such as land use,
to the stream segments

Evaluate subwatershed characteristics that may have an impact on
atrazine loading and identify stream segments with characteristics
similar to those monitored in the atrazine study (this is addressed in
more detail in the following section).

From The 40 Sampled Watersheds To The Larger Population Of Vulnerable
Watersheds

A challenge in addressing the third objective of the monitoring study
– identifying areas where higher atrazine exposures are likely to
occur – is to determine whether the watersheds that exceeded the LOC
in multiple years are randomly distributed within the 1,172 vulnerable
watersheds or represent a unique subset of conditions that may be
applied to other areas. This section presents two approaches the Agency
is taking to address this question.

Evaluation of WARP Parameters

A plot of the WARP values estimated for the most vulnerable 1,172 HUC-10
watersheds against the percent changes in the Steinhaus Similarity Index
(  REF _Ref181435840 \h  Figure IV-2 ) shows that the estimated 95th
percentile atrazine concentration from WARP does not distinguish between
the 10 site years that exceeded the LOC and the remaining 76 site years
that fell below the 4% SSI deviation. In fact, the only two sites (MO-01
and MO-02) that exceeded the LOC in multiple years of monitoring
represented the high and low end of WARP values. As noted in Section  
REF _Ref182299701 \w \h  III.D.2 , no apparent trend was seen between
WARP values and %SSI for the 40 monitored watersheds. However, the
relationship may not be evident because only the upper end of the plot
(the watersheds represent the upper 80th percentile of watersheds) is
available.

Figure   STYLEREF 1 \s  IV -  SEQ Figure \* ARABIC \s 1  2  Comparison
of WARP values to %SSI deviation for the 1,172 vulnerable watersheds
(sites exceeding the 4% LOC trigger are labeled)

The WARP estimates used for the watershed vulnerability assessment
represent an average of atrazine use over the years 1998 through 2002
while the monitoring data represent more recent atrazine usage (for the
years 2004 through 2006). A plot of WARP based on 1998-2002 usage data
and WARP estimates based on more recent usage data (2001-2003) shows a
fair bit of scatter around the 1-to-1 line (  REF _Ref181437888 \h 
Figure IV-3 ).

Figure   STYLEREF 1 \s  IV -  SEQ Figure \* ARABIC \s 1  3  Comparison
of original WARP values based on 1998-2002 use data with new WARP values
based on 2001-2003 use data

The variations in WARP estimates based on changes in usage data could be
expected because atrazine use is the major explanatory factor in the
regression models, accounting for 53 to 64 percent of the variability in
the monitoring data used for WARP development (Crawford, 2004). The
results illustrate the importance of atrazine use as a factor affecting
the vulnerability of water bodies to atrazine exposure. Changes in use
patterns both locally and at a larger watershed scale, could be expected
to result the variations in the WARP vulnerability estimates seen here. 

The Agency plans to explore the use of the WARP model with atrazine use
zeroed out to evaluate the intrinsic watershed vulnerability independent
of use. In additition, coupling the WARP parameters with updated
atrazine use intensities using the NHDPlus might provide a useful means
of identifying potential vulnerable areas based on both current use and
on projections of changing use (for instance, increasing acreage planted
in corn because of demands for ethanol). This tool, coupled with the
results of the monitoring study, could serve as a useful tool to
identify areas to target future monitoring efforts for atrazine.

Evaluation Of Other Soil- And Hydrology-Related Parameters

Results of the monitoring program suggest that the chemographs that
triggered the LOC had relatively high concentrations with a prolonged
period of elevated exposures (see MO-01 and MO-01 in Appendix 2). In
their preliminary evaluation of these two Missouri sites that triggered
the LOC in multiple years, Syngenta noted that the prolonged periods of
elevated exposure might be attributable to the presence of ‘claypan’
soils in the region. This was supported by additional research conducted
by the USDA in the Central Claypan Major Land Resource Area (MLRA),
where the watersheds are located (Blanchard and Lerch, 2000; Lerch and
Blanchard, 2003).

The US EPA is exploring the potential for the presence of a shallow
restrictive layer, such as but not limited to the claypan soils in the
Central Claypan MLRA, to be a driving factor that results in high
atrazine loads to water bodies for prolonged periods of time. The impact
of such drainage-restrictive layers is two-fold:

A shallow depth to a drainage-restrictive layer reduces the water
storage capacity of the soil. During rainfall events, the soil overlying
the restrictive layer becomes saturated quickly, increasing the
frequency and volume of runoff events in comparison to deeper soils with
no restrictive layer in otherwise similar conditions. If the rainfall
occurs after atrazine has been applied to the field, the runoff could
include sufficient quantities of atrazine in the runoff to result in
high concentrations. 

Subsurface drainage laterally over the restrictive layer would result in
a delayed baseflow, contributing additional loadings of atrazine over
time, prolonging the exposure period in the receiving water bodies

Syngenta proposes that the MO-01 and MO-02 are unique because of their
location in the Central Claypan MLRA and, as such, should be treated as
a separate stratum. The US EPA believes that MO-01 and MO-02 potentially
represent conditions where a shallow restrictive layer may result in
atrazine exposures that exceed the LOC. To test this, the Agency plans
to compare the extent, type, and depth of drainage-restrictive layers in
these 2 watersheds with the other 38 monitored watersheds to determine
whether distinctions are present. Where distinctions between the
watersheds are found, the US EPA plans to identify other areas where
conditions similar to those found in MO-01 and MO-02 may exist.

The USDA NRCS defines a restrictive layer as a “nearly continuous
layer that has one or more physical, chemical, or thermal properties
that significantly reduce the movement of water and air through the soil
or that otherwise provide an unfavorable root environment. Cemented
layers, dense layers, frozen layers, abrupt or stratified layers,
strongly contrasting textures, and dispersed layers are examples of soil
layers that are restrictions.” (USDA NRCS, 2007). 

A number of specific restrictive layers are defined by the USDA NRCS.
The Agency is evaluating county-level spatial soil survey data (SSURGO)
to determine the extent, type, and depth of the following kinds of
layers that can restrict water movement vertically through the soil
profile:

Abrupt textural change

Strongly contrasting textural stratification

Dense material

Duripan/ Fragipan

Cemented horizon

Petroferric/ Petrocalcic/ Petrogypsic/ Placic

Plinthite/ Ortstein

Bedrock (lithic or paralithic)

A preliminary evaluation of SSURGO data in the Claypan MLRA indicate
that soils with an abrupt textural change located at shallow depths (<50
cm below the surface) are more predominant in and around MO-01 and MO-02
than in the other monitoring sites (  REF _Ref181460376 \h  Figure IV-4
). Other drainage restrictive layers, particularly dense material,
bedrock, and fragipans, are found in some of the other monitoring sites,
but are typically deeper than what occurs in and around MO-01 and MO-02.
  REF _Ref181460376 \h  Figure IV-4  illustrates that the soil factors
that may influence atrazine loadings to receiving water bodies are
likely to be distinguishable at subwatershed scales. Additional
distinctions, such as the extent of shallow restrictive layers under
cropland, will also be considered.

Figure   STYLEREF 1 \s  IV -  SEQ Figure \* ARABIC \s 1  4  Depth to
restrictive soil layers in MO-01 and MO-02, located in the Central
Claypan MLRA.

In addition to the depth to restrictive layer, the US EPA plans to use
SSURGO data to evaluate the relative percentage of hydrologic soil
groups (particularly D or C and D soils combined), soil drainage classes
(somewhat poorly and poorly drained soils), low hydraulic conductivity
soils, and soils with a high surface runoff potential. A preliminary
evaluation of SSURGO data for the counties that intersect the monitoring
sites shows that not only are the MO-01 and MO-02 watersheds dominated
by soils in the hydrologic soil group D class (  REF _Ref181461414 \h 
Figure IV-5 ), but that the three Nebraska sites that exceeded the LOC
in one year as a result of low flow conditions also had higher
percentages of hydrologic group D soils than did the other four Nebraska
sites or other monitored sites (  REF _Ref181461424 \h  Figure IV-6 ).

Figure   STYLEREF 1 \s  IV -  SEQ Figure \* ARABIC \s 1  5  Distribution
of Hydrologic Group C and D Soils in MO-01 and MO-02, located in the
Central Claypan MLRA.

Figure   STYLEREF 1 \s  IV -  SEQ Figure \* ARABIC \s 1  6  Distribution
of Hydrologic Group C and D Soils in the Atrazine Monitoring Sites in
Nebraska, Missouri, Iowa, and western Illinois

While the preliminary evaluations suggest that potential differences in
soil and hydrology may exist between the two MO monitoring sites that
exceeded the LOC in multiple years, as well as between the three NE
monitoring sites that exceeded the LOC in one year as a result of low
flow, more extensive comparisons will require a substantial amount of
data processing and analysis. The US EPA plans a three step approach:

Compare soil/hydrologic properties among the 40 monitored watersheds,
both at the HUC-10 level and at the specific subwatershed level;

Extend the soil/hydrologic analysis to the 1,172 watersheds the
represented the most vulnerable tier of watersheds based on WARP

Because the soil/hydrologic parameters being investigated were not
explicitly part of the WARP parameters, further extend the evaluation to
the larger extent of atrazine use, in particular, the 5,860 watersheds
that intersected a county with atrazine use of greater than 0.25 lb
ai/county acre

These approaches are data and resource intensive. While SSURGO data are
now available for most of the counties in the US, they have to be
processed and compiled from individual county coverages to state and
regional coverages. Currently, the US EPA has only processed SSURGO for
the counties that encompass each of the 40 watersheds included in the
monitoring study. Before proceeding, the US EPA first intends to consult
with this SAP on the feasibility of such an assessment and to solicit
recommendations and suggestions on other parameters and/or approaches
that might be useful in extending the results of the monitoring study
beyond the 40 monitored watersheds in a manner that would be useful to
the US EPA for identifying other water bodies exceeding the LOC.

SAP Charge Questions Relating to Identifying Where Atrazine Exceedances
Are Likely to Occur

(1) 	While the monitoring study was based on a watershed vulnerability
assessment, the ultimate value is in identifying water bodies where
atrazine concentrations exceed the LOC. One approach is to use the
updated version of the National Hydrography Database (NHDPlus) and apply
the criteria used to select the monitoring locations to identify streams
that appear to have the potential to exceed the LOC.

Please comment on the strengths and weaknesses of the Agency’s
proposed approach for identifying streams within watersheds that
exceeded the LOC. 

In what ways can the preliminary approach be improved? 

Please recommend alternative approaches, if any, that may be better
suited to apply the watershed-based assessment to streams?

(2) 	In order to identify areas beyond the 40 study sites where higher
atrazine exposures are likely to occur, the Agency must determine
whether the watersheds that exceeded the LOC in multiple years are
randomly distributed within the 1,172 vulnerable watersheds or represent
a unique subset of conditions. If the latter and the conditions can be
identified, monitoring could be focused only in watersheds where those
conditions exist.The Agency has proposed evaluating WARP parameters and
other sub-watershed soil and hydrologic properties to determine the
extent to which the monitoring results can be used to identify other
water bodies exceeding the LOC.

To what extent can WARP be used to identify other watersheds of concern?
Given the influence of atrazine use on vulnerability and exposure,
please comment on whether the extrapolation should be limited to the
original 1,172 watersheds or include a broader atrazine use area? 

Please comment on the soil and hydrology parameters the Agency is
evaluating for extrapolation to vulnerable watersheds. What additional
soil and hydrologic parameters should the Agency consider?

What additional approaches to the identification of watersheds that may
have atrazine levels that exceed the LOC should the Agency consider?

  References

Abou-Waly, H., M.M. Abou-Setta, H.N. Nigg, L.L. Mallory. 1991. Growth
response of freshwater algae, Anabaena flos-aquae and Selenastrum
capricornutum to atrazine and hexazinone herbicides. Bull Environ Contam
Toxicol 46:223-229.

ARM (2007).  Aquatic Resource Monitoring web pages.   HYPERLINK
"http://www.epa.gov/nheerl/arm/"  http://www.epa.gov/nheerl/arm/ .

Bartell, S.M.  2003.  A framework for estimating ecological risks posed
by nutrients and reace elements in the Patuxent River.  Estuaries.
26(2A):385-397.

Bartell, S.M., K.R. Campbell, C.M. Lovelock, S.K. Nair, and J.L. Shaw.
2000. Characterizing aquatic ecological risk from pesticides using a
diquat dibromide case study. III. Ecological Process Models. Environ.
Toxicol. Chem. 19(5):1441-1453. 

Bartell, S.M., G. Lefebvre, G. Kaminski, M. Carreau, and K.R. Campbell.
1999. An ecosystem model for assessing ecological risks in Quebec
rivers, lakes, and reservoirs. Ecol. Model. 124:43-67.

Bartell, S.M., R.H. Gardner, and R.V. O'Neill. 1988.  An integrated fate
and effects model for estimating risk in aquatic systems.  pp.261-274.
In Aquatic Toxicology and Hazard Assessment, Vol. 10.  ASTM STP971,
American Society for Testing and Materials, Philadelphia, PA, USA.

Berard, A., T. Pelte and J. Druart. 1999. Seasonal variations in the
sensitivity of Lake Geneva phytoplankton community structure to
atrazine. Arch. Hydrobiol. 145(3):277-295.

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. 34:3315-3322.

Brock, T.C.M., J. Lahr, P.J. van den Brink, 2000. Ecological risks of
pesticides in freshwater ecosystems. Part 1: Herbicides.  Wageningen,
Alterra, Green World Research. Alterra-Rapport 088. 124 pp.

Brockway, D.L., P.D. Smith and F.E. Stancil. 1984. Fate and effects of
atrazine on small aquatic microcosms. Bull. Environ. Contam. Toxicol.
32:345-353.

Burrell RE, Inniss WE, Mayfield CI. 1985. Detection and analysis of
interactions between atrazine and sodium pentachlorophenate with single
and multiple algal-bacterial populations. Arch Environ Contam Toxicol
14:167-177.

Carder, J.P. and K.D. Hoagland. 1998. Combined effects of alachlor and
atrazine on benthic algal communities in artificial streams. Environ.
Toxicol. Chem. 17(7):1415-1420.

Carney, E.C. 1983. The effects of atrazine and grass carp on freshwater
communities. Thesis. University of Kansas, Lawrence, Kansas.

Caux P-Y, Menard L, Kent RA. 1996. Comparative study on the effects of
MCPA, butylate, atrazine, and cyanazine on Selenastrum capricornutum.
Environ Poll 92:219-225.

Carsel, R.F., Imhoff, J.C., Hummel, P.R. Cheplick, J.M., and A.S.
Donigian, Jr. 1998. PRZM-3, A Model for Predicting Pesticide and
Nitrogen Fate in the Crop Root and Unsaturated Soil Zones: Users Manual
for Release 3.0, National Exposure Research Laboratory, Office of
Research and Development, U.S. Environmental Protection Agency, Athens,
Georgia. Available:
http://www.epa.gov/ceampubl/gwater/przm3/przm3122.htm.

Crawford, C.G. 2004. Sampling Strategies for Estimating Acute and
Chronic Exposures of Pesticides in Streams.  Journal of the American
Water Resources Association (JAWRA 40(2) : 485-502.

DeAngelis, D.L., S.M. Bartell, and A.L. Brenkert. 1989. Effects of
nutrient recycling and food-chain length on resilience. Amer. Nat.
134(5):778-805. 

deNoyelles, F., Jr., S.L. Dewey, D.G. Huggins and W.D. Kettle. 1994.
Aquatic mesocosms in ecological effects testing: Detecting direct and
indirect effects of pesticides. In: Aquatic mesocosm studies in
ecological risk assessment. Graney, R.L., J.H. Kennedy and J.H. Rodgers
(Eds.). Lewis Publ., Boca Raton, FL. pp. 577-603.

deNoyelles, F., and W.D. Kettle. 1980. Herbicides in Kansas waters -
evaluations of effects of agricultural runoff and aquatic weed control
on aquatic food chains. Contribution Number 219, Kansas Water Resources
Research Institute, University of Kansas, Lawrence, Kansas.

deNoyelles, F., and W.D. Kettle. 1983. Site studies to determine the
extent and potential impact of herbicide contamination in Kansas waters.
Contribution Number 239, Kansas Water Resources Research Institute,
University of Kansas, Lawrence, Kansas. 

deNoyelles, F., Jr. and W.D. Kettle. 1985. Experimental ponds for
evaluating bioassay predictions. In: Validation and predictability of
laboratory methods for assessing the fate and effects of contaminants in
aquatic ecosystems. Boyle, T.P. (Ed.). ASTM STP 865, American Society
for Testing and Materials, Philadelphia, PA. pp. 91-103.

deNoyelles, F., Jr., W.D. Kettle and D.E. Sinn. 1982. The responses of
plankton communities in experimental ponds to atrazine, the most heavily
used pesticide in the United States. Ecol. 63:1285-1293.

deNoyelles, F., Jr., W.D. Kettle, C.H. Fromm, M.F. Moffett and S.L.
Dewey. 1989.  Use of experimental ponds to assess the effects of a
pesticide on the aquatic environment. In: Using mesocosms to assess the
aquatic ecological risk of pesticides: Theory and practice. Voshell,
J.R. (Ed.). Misc. Publ. No. 75. Entomological Society of America,
Lanham, MD.

Detenbeck, N.E., R. Hermanutz, K. Allen and M.C. Swift. 1996. Fate and
effects of the herbicide atrazine in flow-through wetland mesocosms.
Environ. Toxicol. Chem. 15:937-946.

Deutsch, C.V. and A.G. Journel. 1998. GSLIB: Geostatistical Software
Library and User's  Guide. Oxford University Press, New York. pp 369.

Dewey, S.L. 1986. Effects of the herbicide atrazine on aquatic insect
community structure and emergence. Ecol. 67:148-162.

Fairchild, J.F., T.W. LaPoint and T.R. Schwarz. 1994a. Effects of a
herbicide and insecticide mixture in aquatic mesocosms. Arch. Environ.
Contam. Toxicol. 27:527-533.

Fairchild JF, Ruessler DS, Carlson AR. 1998. Comparative sensitivity of
five species of macrophytes and six species of algae to atrazine,
metribuzin, alachlor, and metolachlor. Environ Toxicol Chem
17:1830-1834.

Fairchild JF, Ruessler DS, Haverland PS, Carlson AR. 1997. Comparative
sensitivity of Selenastrum capricornutum and Lemna minor to sixteen
herbicides. Arch Environ Contam Toxicol 32:353-357.

Fairchild, J.F., S.D. Ruessler, M.K. Nelson and A.R. Carlson. 1994b. An
aquatic risk assessment for four herbicides using twelve species of
macrophytes and algae.  Abstract No. HF05, 15th Annual Meeting.  Society
of Environmental Toxicology and Chemistry, Denver, CO.

Fairchild WL, Swansburg EO, Arsenault JT, Brown SB. 1999. Does an
association between pesticide use and subsequent declines in catch of
atlantic salmon (Salmo salar) represent a case of endocrine disruption?
Environ Health Persp 107:349-357.

Faust M, Altenburger R, Boedeker W, Grimme LH. 1993. Additive effects of
herbicide combinations on aquatic non-target organisms. Sci Total
Environ 113/114:941-951.

Forney DR, Davis DE. 1981. Effects of low concentrations of herbicides
on submersed aquatic plants. Weed Sci 29:677-685.

Gala WR, Giesy JP. 1990. Flow cytometric techniques to assess toxicity
to algae. In: Landis WG, van der Schalie WH, eds. Aquatic Toxicology and
Risk Assessment: Thirteenth Volume. Philadelphia, PA: American Society
for Testing and Materials. pp 237-246. AT 174.

Geyer H, Scheunert I, Korte F. 1985. The effects of organic
environmental chemicals on the growth of the alga Scenedesmus
subspicatus: a contribution to environmental biology. Chemosphere
14:1355-1369.

Giddings, J.M., T.A. Anderson, L.W. Hall, Jr., R. J. Kendall, R.P.
Richards, K.R. Solomon, W.M. Williams.  2000.  Aquatic Ecological Risk
Assessment of Atrazine: A Tiered, Probabilistic Approach. Prepared by
the Atrazine Ecological Risk Assessment Panel, ECORISK, Inc. Novartis
Crop Protection, Inc. Project Monitor, Alan Hosmer. Novartis Number
709-00.

Goolsby, D.A. and W.A. Battaglin, 1993. Occurrence, Distribution and
Transport of Agricultural Chemicals in Surface Waters of the Midwestern
United States, Selected Papers on Agricultural Chemicals in Water
Resources of the Midcontinental United States. Openfile Report 93-418.
U.S. Geological Survey, Denver, CO.

Gruessner, B. and M.C. Watzin.  1996. Response of aquatic communities
from a Vermont stream to environmentally realistic atrazine exposure in
laboratory microcosms. Environ. Toxicol. Chem. 15:410-419.

Gustavson, K. and S.A. Wangberg. 1995. Tolerance induction and
succession in microalgae communities exposed to copper and atrazine.
Aquat. Toxicol. 32:283-302.

Hamala, J.A. and H.P. Kollig. 1985. The effects of atrazine on
periphyton communities in controlled laboratory ecosystems. Chemosphere
14:1391-1408.

Hamilton, P.B., D.R.S. Lean, G.S Jackson, N.K. Kaushik and K.R. Solomon.
1989.  The effect of two applications of atrazine on the water quality
of freshwater enclosures. Environ. Pollut. 60:291-304.

Hamilton, P.B., G.S. Jackson, N.K. Kaushik and K.R. Solomon. 1987. The
impact of atrazine on lake periphyton communities, including carbon
uptake dynamics using track autoradiiography. Environ. Poll. 46:83-103.

Hamilton, P.B., G.S. Jackson, N.K. Kaushik, K.R. Solomon and G.L.
Stephenson. 1988. The impact of two applications of atrazine on the
plankton communities of in situ enclosures. Aquat. Toxicol. 13:123-140.

Hampton, M. Burnett, G., Carver, L.S., Harbourt, C.M., Hendley, P.,
Johnston, E.A., Perez, S., Snyder, N.J., and Trask, J.R., 2007a.  2007
Interim Report - 2004 - 2006 Data Overview - Atrazine Ecological
Exposure Flowing Water Chemical Monitoring Study in Vulnerable
Watersheds Interim Report.  Prepared by Waterborne Environmental, Inc.,
Leesburg, VA for Syngenta Crop Protection, Inc., Greensboro, NC.

Hampton, M. Prenger, J.J., Harbourt, C.M., Hendley, P., and Miller,
P.S., 2007b.  Atrazine Ecological Exposure Flowing Water Chemical
Monitoring Study in Vulnerable Watersheds: Approaches to Assessing
Potential Watershed Scale Vulnerability for Atrazine Runoff.  Prepared
by Waterborne Environmental, Inc., Leesburg, VA for Syngenta Crop
Protection, Inc., Greensboro, NC.

Hanratty, M.P. and K. Liber. 1996. Evaluation of model predictions of
the persistence and ecological effects of diflubenzuron in a littoral
ecosystem.  Ecol. Model. 90:79-95.

Hanratty, M.P. and F.S. Stay.  1994.  Field evaluation of the littoral
ecosystem risk assessment model's predictions of the effects of
chlorpyrifos.  J. Appl. Ecol. 31:439-453.

Harbourt, C.M., J.R. Trask, M.K. Matella, M.H. Ball, and J.M. Cheplick.
2004. Sampling Site Selection: Atrazine Ecological Exposure Flowing
Water Chemical Monitoring Study in Vulnerable Watersheds - Youngs Creek,
MO (MO-02). Prepared by Waterborne Environmental, Inc., Leesburg, VA for
Syngenta Crop Protection, Inc., Greensboro, NC.  Syngenta Number
T001508-03, volume 22.

Herman, D., N.K. Kaushik and K.R. Solomon. 1986. Impact of atrazine on
periphyton in freshwater enclosures and some ecological consequences.
Can. J. Fish. Aquat. Sci. 43:1917-1925.

Hinman ML. 1989. Utility of rooted aquatic vascular plants for aquatic
sediment hazard evaluation.  Ph.D. Thesis. Memphis State University,
Memphis, TN.

Hoberg JR. 1991. Atrazine Technical - Toxicity to the duckweed, Lemna
gibba G3.  SLI Report #91-1-3613. Springborn Laboratories, Inc.,
Wareham, MA.

Hoberg JR. 1993a. Atrazine Technical - Toxicity to duckweed (Lemna
gibba).  SLI Report #93-11-5053. Springborn Laboratories, Inc., Wareham,
MA.

Hoberg JR. 1993b. Atrazine Technical - Toxicity to the freshwater green
algae (Selenastrum capricornutum). SLI Report # 93-4-4751. Springborn
Laboratories, Inc., Wareham, MA.

Hoberg JR. 1993c. Atrazine Technical - Toxicity to the marine diatom
(Skeletonema costatum).  SLI Report # 93-4-4753. Springborn
Laboratories, Inc., Wareham, MA.

Horizon Systems Corporation (HSC). 2006. National Hydrography Dataset
Plus Home.   HYPERLINK "http://www.horizon-systems.com/nhdplus/" 
http://www.horizon-systems.com/nhdplus/ . 

Hughes JS, Alexander MM, Balu K. 1988. An evaluation of appropriate
expressions of toxicity in aquatic plant bioassays as demonstrated by
the effects of atrazine on algae and duckweed. In: Adams W, Chapman GA,
Landis WG, eds. Aquatic Toxicology and Hazard Assessment: 10th Volume.
Philadelphia, PA: American Society for Testing and Materials. pp
531-547.

Issaks, E.H. and R.M. Srivastava. 1989. Applied Geostatistics. Oxford
University Press, New York. pp 561.

Johnson, B.T. 1986. Potential impact of selected agricultural chemical
contaminants on a northern prairie wetland: A microcosm evaluation.
Environ. Toxicol. Chem. 5:473-485.

Jones TW, Kemp WM, Estes PS, Stevenson JC. 1986. Atrazine uptake,
photosynthetic inhibition, and short-term recovery for the submersed
vascular plant, Potamogeton perfoliatus L. Arch Environ Contam Toxicol
15:277-283.

Jurgensen, T. A. and K. D. Hoagland. 1990.  Effects of short-term pulses
of atrazine on attached algal communities in a small stream.  Arch.
Environ. Contam. Toxicol. 19:617-623.

Juttner, I., A. Peither, J.P. Lay, A. Kettrup and S.J. Ormerod. 1995. An
outdoor mesocosm study to assess ecotoxicological effects of atrazine on
a natural plankton community. Arch. Environ. Contam. Toxicol.
29:435-441.

Kallqvist T, Romstad R. 1994. Effects of agricultural pesticides on
planktonic algae and cyanobacteria - examples of interspecies
sensitivity variations. Norw J Agric Sci 13:117-131.

Kemp WM, Boynton WR, Cunningham JJ, Stevenson JC, Jones TW, Means JC.
1985. Effects of atrazine and linuron on photosynthesis and growth of
the macrophytes, Potamogeton perfoliatus L. and Pyriophyllum spicatum
L., in an estuarine environment. Mar Environ Res 16:255-280.

Kettle, W.D. 1982. Description and analysis of toxicant-induced
responses of aquatic communities in replicated experimental ponds. Ph.D.
Thesis. University of Kansas, Lawrence, KS.

Kettle, W.D., F. deNoyelles, B.D. Heacock and A.M. Kadoum. 1987. Diet
and reproductive success of bluegill recovered from experimental ponds
treated with atrazine. Bull. Environ. Contam. Toxicol. 38:47-52.

Kirby MF, Sheahan DA. 1994. Effects of atrazine, isoproturon, and
mecoprop on the macrophyte Lemna minor and the alga Scenedesmus
subspicatus. Bull Environ Contam Toxicol 53:120-126.

Kosinski, R.J. 1984. The effect of terrestrial herbicides on the
community structure of stream periphyton. Environ. Pollut. (Series A)
36:165-189.

Kosinski, R.J. and M.G. Merkle. 1984. The effect of four terrestrial
herbicides on the productivity of artificial stream algal communities.
J. Environ. Qual. 13:75-82.

Krieger, K.A., D.B. Baker and J.W. Kramer. 1988. Effects of herbicides
on stream aufwuchs productivity and nutrient uptake. Arch. Environ.
Contam. Toxicol. 17:299-306.

Lakshminarayana, J.S.S., H.J. O'Neill, S.D. Jonnavithula, D.A. Leger and
P.H. Milburn. 1992. Impact of atrazine-bearing agricultural tile
drainage discharge on planktonic drift of a natural stream. Environ.
Pollut. 76:201-210.

Lampert, W., W. Fleckner, E. Pott, U. Schober and K.U. Storkel. 1989.
Herbicide effects on planktonic systems of different complexity.
Hydrobiologia 188/189:415-424.

Larsen DP, DeNoyelles Jr. F, Stay F, Shiroyama T. 1986. Comparisons of
single-species, microcosm and experimental pond responses to atrazine
exposure. Environ Toxicol Chem 5:179-190.

Larson, S.J., C.G. Crawford, and R.J. Gilliom. 2004. Development and
Application of Watershed Regressions for Pesticides (WARP) for
Estimating Atrazine Concentration Distributions in Streams. U.S.
Geological Survey. Water–Resources Investigations Report 03-4047.
Sacramento, California. 2004.

Lerch, R.N., and P.E. Blanchard. 2003. Watershed Vulnerability To
Herbicide Transport in Northern Missouri and Southern Iowa Streams.
Environ. Sci. Technol. 37:5518-5527. 

Lynch, T.R., H.E. Johnson and W.J. Adams. 1985. Impact of atrazine and
hexachlorobiphenyl on the structure and function of model stream
ecosystems. Environ. Toxicol. Chem. 4:399-413.

Mayer FL. 1987. Acute toxicity handbook of chemicals to estuarine
organisms.  EPA/600/8-87/017. U.S. Environmental Protection Agency, Gulf
Breeze, FL.

Mayer P, Frickmann J, Christensen ER, Nyholm N. 1998. Influence of
growth conditions on the results obtained in algal toxicity tests.
Environ Toxicol Chem 17:1091-1098.

McDonald, M., Blair, R., Hale, S., et al. (2002). Environmental
Protection Agency's Environmental Monitoring and Assessment Program
(EMAP) in the 21st Century. Hydrological Science and Technology 18,
133-143.

Meyer, J.L, Strayer, D.L, Wallace, J.B., Eggert, S.L., Helfman, G.S.,
and Leonard, N.E. 2007.  The Contribution of Headwater Streams to
Biodiversity in River Networks. Journal of the American Water Resources
Association (JAWRA) 43(1): 86-103.

Montague B. 1998. Oneliner pesticide toxicity database. U.S.
Environmental Protection Agency.

Moorhead, D.L. and R.J. Kosinski. 1986. Effect of atrazine on the
productivity of artificial stream algal communities. Bull. Environ.
Contam. Toxicol. 37:330-336.

Naito, W., K. Myamoto, J. Nakanishi, and S.M. Bartell. 2003.  Evaluation
of an ecosystem model in ecological risk assessment of chemicals. 
Chemosphere 53:363-375.

Naito, W., K. Myamoto, J. Nakanishi, and S.M. Bartell. 2002. 
Application of an ecosystem model for ecological risk assesment of
chemicals for a Japanese lake.  Water Res. 36:1-14.

O'Neill, R.V., S.M Bartell, and R.H. Gardner.  1983.  Patterns of
toxicological effects in ecosystems:  A modeling approach.  Environ.
Toxicol. Chem.  2:451-461.

O'Neill, R.V., R.H. Gardner, L.W. Barnthouse, G.W. Suter II, S.G.
Hildebrand, and C.W. Gehrs.  1981.  Ecoystem risk analysis: A new
methodology.  Environ. Toxicol. Chem.  1:167-177.

Parrish R. 1978. Effects of atrazine on two freshwater and five marine
algae. Ciba-Geigy Corporation, Greensboro, NC.

Pierce, C. Robert and J. Anderson, 1992. Minnesota Rating for Potential
Leaching and Surface Runoff of Pesticides, University of Minnesota
Extension Service publication FO-05758. URL:
http://www.extension.umn.edu/distribution/cropsystems/DC5758.html,
8/5/2003.

Pratt, J.R., N.J. Bowers, B.R. Niederlehrer and J. Cairns, Jr. 1988.
Effects of atrazine on freshwater microbial communities. Arch. Environ.
Contam. Toxicol. 17:449-457.

R Development Core Team (2006). R: A language and environment for
statistical computing. Vienna, Austria: R Foundation for Statistical
Computing.

Radetski CM, Ferard JF, Blaise C. 1995. A semistatic microplate-based
phytotoxicity test. Environ Toxicol Chem 14:299-302.

Roberts SP, Vasseur P, Dive D. 1990. Combined effects between atrazine,
copper and pH on target and non-target species. Water Res 24:485-491.

Schafer H, Hettler H, Fritsche U, Pitzen G, Roderer G, Wenzel A. 1994.
Biotests using unicellular algae and ciliates for predicting long-term
effects of toxicants. Ecotox Environ Saf 27:64-81.

Schafer H, Wenzel A, Fritsche U, Roderer G, Traunspurger W. 1993.
Long-term effects of selected xenobiotica on freshwater green algae:
development of a flow-through test system. Sci Total Environ
113/114:735-740.

Seaber, P.R., Kapinos, F.P., and Knapp, G.L. 1987. Hydrologic Unit Maps:
U.S. Geological Survey Water-Supply Paper 2294, 63 p.

Snyder, N.J., Harbourt, C.M., Miller, P.S., Trask, J.R., Prenger, J.J.,
Hendley, P., and Johnston, E.A., 2007.  Atrazine Ecological Exposure
Flowing Water Chemical Monitoring Study in Vulnerable Watersheds:
Analysis of Chemograph Behavior between Grab Samples - Measurement and
Hybrid PRZM Approaches.  Prepared by Waterborne Environmental, Inc.,
Leesburg, VA for Syngenta Crop Protection, Inc.,

Stay, E.F., A. Katko, C.M. Rohm, M.A. Fix and D.P. Larsen. 1989. The
effects of atrazine on microcosms developed from four natural plankton
communities. Arch. Environ. Contam. Toxicol. 18:866-875.

Stay, F.S., D.P. Larsen, A. Katko and C.M. Rohm. 1985. Effects of
atrazine on community level responses in Taub microcosms. In: Validation
and predictability of laboratory methods for assessing the fate and
effects of contaminants in aquatic ecosystems. Boyle, T.P. (Ed.). ASTM
STP 865. American Society for Testing and Materials, Philadelphia, PA.
pp. 75-90.

Stevens, D. L., Jr., and Olsen, A. R. (1999). Spatially restricted
surveys over time for aquatic resources. Journal of Agricultural,
Biological, and Environmental Statistics 4, 415-428.

Stevens, D. L., Jr., and Olsen, A. R. (2003). Variance estimation for
spatially balanced samples of environmental resources. Environmetrics
14, 593-610.

Stevens, D. L., Jr., and Olsen, A. R. (2004). Spatially-balanced
sampling of natural resources. Journal of American Statistical
Association 99, 262-278.

Stratton GW. 1984. Effects of the herbicide atrazine and its degradation
products, alone and in combination, on phototrophic microorganisms. Arch
Environ Contam Toxicol 13:35-42.

Stratton GW, Giles J. 1990. Importance of bioassay volume in toxicity
tests using algae and aquatic invertebrates. Bull Environ Contam Toxicol
44:420-427. Page 16 of 19

Turbak SC, Olson SB, McFeters GS. 1986. Comparison of algal assay
systems for detecting waterborne herbicides and metals. Water Res
20:91-96.

University of Mississippi. 1991. Effects of atrazine on Selenastrum
capricornutum, Lemna minor and Elodea canadensis. Ciba-Geigy
Corporation, Greensboro, NC.

U.S. Department of Agriculture (USDA). 2001. 1997 National Resources
Inventory (Revised December 2000), Natural Resources Conservation
Services, Washington, DC, and Statistical Laboratory, Iowa State
University, Ames, Iowa, CD-ROM, Version 1.

U. S. Department of Agriculture (USDA). 1994. State Soil Geographic
(STATSGO) Data Base. USDA Natural Resources Conservation Service
(USDA-NRCS), National Soil Survey Center. Miscellaneous Publication
Number 1492. Revised 1994. 

U.S. Department of Agriculture, Natural Resources Conservation Service
(USDA NRCS). 2007. National Soil Survey Handbook, title 430-VI. NSSH
Section 618.49. [Online] Available:   HYPERLINK
"http://soils.usda.gov/technical/handbook/" 
http://soils.usda.gov/technical/handbook/ .  

US EPA. 2003a. Interim Reregistration Eligibility Decision for Atrazine.
Case No. 0062. US EPA Office of Pesticides Program. January 31, 2003.  
HYPERLINK
"http://www.epa.gov/pesticides/reregistration/REDs/atrazine_ired.pdf" 
http://www.epa.gov/pesticides/reregistration/REDs/atrazine_ired.pdf  

US EPA. 2003b. Revised Interim Reregistration Eligibility Decision for
Atrazine. US EPA Office of Pesticides Program. October 31, 2003.  
HYPERLINK
"http://www.epa.gov/pesticides/reregistration/atrazine/atrazineadd.pdf" 
http://www.epa.gov/pesticides/reregistration/atrazine/atrazineadd.pdf  

U.S. Geological Survey (USGS). 2006a. Hydrologic Unit Maps. URL:  
HYPERLINK "http://water.usgs.gov/GIS/huc.html. Last modified
26-Oct-2006"  http://water.usgs.gov/GIS/huc.html. Last modified
26-Oct-2006 .

U.S. Geological Survey (USGS). 2006b. National Stream Quality Accounting
Network (NASQAN). URL:   HYPERLINK "http://water.usgs.gov/nasqan/" 
http://water.usgs.gov/nasqan/ . Last updated Nov 13, 2006.

U.S. Geological Survey (USGS). 2007. National Water Quality Assessment
(NAWQA) Program. URL:   HYPERLINK "http://water.usgs.gov/nawqa/" 
http://water.usgs.gov/nawqa/ . Last updated Sep 24, 2007.

Van den Brink, P.J., E. van Donk, R. Gylstra, S.J.H. Crum and T.C.M.
Brock. 1995. Effects of chronic low concentrations of the pesticides
chlorpyrifos and atrazine in indoor freshwater microcosms. Chemosphere
31:3181-3200.

Versteeg DJ. 1990. Comparison of short- and long-term toxicity test
results for the green alga, Selenastrum capricornutum. In: Wang W,
Gorsuch JW, Lower WR, eds. Plants for Toxicity Assessment. Philadelphia,
PA: American Society for Testing and Materials. pp 40-48.

Volz, D.C., S.M. Bartell, S.K. Nair, and P. Hendley.  2007.  Modeling
the Potential for Atrazine-Induced Changes in Midwestern Stream
Ecosystems Using the Comprehensive Aquatic System Model (CASM).  Final
Report, T001403-07, Syngenta Crop Protection, Greensboro, NC.  

Walsh GE. 1983. Cell death and inhibition of population growth of marine
unicellular algae by pesticides. Aquat Toxicol 3:209-214.

Walsh GE, McLaughlin LL, Yoder MJ, Moody PH, Lores EH, Forester J,
Wessinger-Duval PB. 1988. Minutocellus polymorphus: A new marine diatom
for use in algal toxicity tests. Environ Toxicol Chem 7:925-929.

Williams, W. M., C.M. Harbourt, M.K. Matella, M.H. Ball, and J.R. Trask.
2004a. Atrazine Ecological Exposure Flowing Water Chemical Monitoring
Study in Vulnerable Watersheds Interim Report: Watershed Selection
Process.  Prepared by Waterborne Environmental, Inc., Leesburg, VA for
Syngenta Crop Protection, Inc., Greensboro, NC.  Syngenta Number
T001509-03, volume 50.

Williams, W. M., C.M. Harbourt, M.H. Ball, M.K. Matella, J.R. Trask, and
N.J. Snyder. 2004b.  Atrazine Ecological Exposure Monitoring Program
Interim Report: Supporting Spatial Data.  Prepared by Waterborne
Environmental, Inc., Leesburg, VA for Syngenta Crop Protection, Inc.,
Greensboro, NC. Syngenta Number T001509-03, volume 51.

Wolock, D.M. 2003. Saturation overland flow estimated by TOPMODEL for
the conterminous United States.  U.S. Geological Survey Open-File Report
03-264. USGS,  Reston, VA. URL:   HYPERLINK
"http://water.usgs.gov/lookup/getspatial?satof48" 
http://water.usgs.gov/lookup/getspatial?satof48 

Zagorc-Koncan J. 1996. Effects of atrazine and alachlor on
self-purification processes in receiving streams. Water Sci Technol
33:181-187.

 Appendix 1: Microcosm and Mesocosm Studies Used in CASM_Atrazine

Table   STYLEREF 1 \s  VI -  SEQ Table \* ARABIC \s 1  1  Micro- and
mesocosm studies table with Brock scores and estimated average % change
in community structure (Steinhaus similarity) of primary producers

No.	Dur-ation (d)	Test Conc (µg/L)	Single / Constant / Multiple	Brock
Effect Score	Reference(s)	Ecosystem	Results	Measurement Endpoint	Plant
Group	Recovery	> or <	AVG % change in community structure (Steinhaus
similarity) of primary producers 

1	300	500	single	5	Carney 1983; Kettle et al. 1987; deNoyelles et al.
1989; deNoyelles et al. 1994	mesocoms, experimental ponds	Decrease	cover
by emerged, floating and submerged aquatic plants	Macro	> 1 yr	>	48.4

2	300	20	single	5	Carney 1983; Kettle et al. 1987; deNoyelles et al.
1989; deNoyelles et al. 1994, deNoyelles & Kettle 1983, deNoyelles &
Kettle 1980, Dewey 1986	mesocoms, experimental ponds	Decrease	cover by
floating and submerged aquatic plants	Macro	> 1 yr	 	0.7

3	60	500	single	5	deNoyelles et al. 1982; Kettle 1982; deNoyelles et al.
1989	mesocoms, experimental ponds	Decrease / Change	14C-uptake
phytoplankton and biomass phytoplankton; all important phytoplankton
species / species composition phytoplankton	Phyto	60 - > 63 d	>	40.6

4	300	100	single	5	deNoyelles et al. 1989 Carney 1983	mesocoms,
experimental ponds	Decrease	cover by emerged and submerged aquatic
plants	Macro	> 1 yr	 	18.0

5	300	200	single	5	deNoyelles et al. 1989 Carney 1983	mesocoms,
experimental ponds	Decrease	cover by emerged and submerged aquatic
plants	Macro	> 1 yr	 	48.4

6	56	50	single	5	Fairchild et al. 1994	mesocoms, experimental ponds (2)
Change / No Effect	Chara sp. replaces Naja sp. / total biomass aquatic
plants	Macro	> 15 wks	 	8.9

7	60	80	multiple	5	Hamilton et al 1987	lake enclosure	Decrease	number.
biomass, composition	Peri	49 d	 	14.9

8	60	140	multiple	5	Hamilton et al 1987	lake enclosure	No Effect /
Change	numbers, biomass, Chl-a and C14 uptake / species composition	Peri
>56 d	 	15.4

9	73	143	multiple	5	Herman et al. 1986; Hamilton et al. 1988; Hamilton
et al. 1989	lake enclosure	Change	species composition	Phyto	>294 d	 
40.6

10	73	143	multiple	5	Herman et al. 1986; Hamilton et al. 1988; Hamilton
et al. 1989	lake enclosure	Decrease / Change	POC (slight), c14 uptake /
species composition	Peri	90 d; 14 d; >294 d; >294 d	 	40.6

11	63	182	constant	5	Jüttner et al. 1995	pond enclosures	No Effect /
Decrease	rotifer / DO, conductivity (slight); algal species [Mallomonas
sp (slight); Cryptomonas sp.] 	Phyto	>63 d; 50 d; 35 d; 56 d	 	40.6

12	63	318	constant	5	Jüttner et al. 1995	pond enclosures	Decrease	DO
(slight); conductivity (slight); rotifers (slight); algal species
[Mallomonas sp (slight); Cryptomonas sp.] 	Phyto	> 63 d; 50 d; 25 d; 35
d; >56 d;  	>	40.6

13	60	500	single	5	Stay et al. 1985	microcosms, laboratory Taub	Decrease
DO, 14C-uptake, net primary production, respiration, 14C-uptake, Chl-a
Phyto	> 53 d	>	40.6

14	60	1000	single	5	Stay et al. 1985	microcosms, laboratory Taub
Decrease	DO, 14C-uptake, net primary production, respiration,
14C-uptake, Chl-a	Phyto	> 53 d	>	40.6

15	60	5000	single	5	Stay et al. 1985	microcosms, laboratory Taub
Decrease	DO, 14C-uptake, net primary production, respiration,
14C-uptake, Chl-a	Phyto	> 53 d	>	40.6

16	21	10	single	4	Berard et al 1999 (1)	microcosm, lab stagnant	Change
Change in species composition & density	Phyto	?	<	0.4

17	7	100	single	4	Brockway et al. 1984	microcosms, lab stagnant	Decrease
Net O2 production	Phyto	> 12 d	 	7.4

18	12	500	single	4	Brockway et al. 1984	microcosms, lab stagnant
Decrease	Net O2 production	Phyto	> 12 d	>	16.4

19	12	5000	single	4	Brockway et al. 1984	microcosms, lab stagnant
Decrease	Net O2 production	Phyto	> 12 d	>	16.4

20	28	15	constant	4	Carder and Hoagland 1998 (1)	artificial streams,
continuous flow	Decrease	Algal community biovolume	Peri	> 28 d	 	0.4

21	28	150	constant	4	Carder and Hoagland 1998 (1)	artificial streams,
continuous flow	Decrease	Algal community biovolume	Peri	> 28 d	 	25.5

22	27	15	constant	4	Detenback et al 1996	artificial flow-through swamp
Decrease / Increase	DO, metabolism of periphyton in bioassays /
nutrients	Peri	?	 	0.4

23	27	25	constant	4	Detenback et al 1996	artificial flow-through swamp
Decrease / Increase	metabolism of periphyton in bioassays / nutrients
Peri	?	 	5.1

24	27	50	constant	4	Detenback et al 1996	artificial flow-through swamp
Decrease / Increase	metabolism of periphyton in bioassays / nutrients
Peri	?	 	6.0

25	27	75	constant	4	Detenback et al 1996	artificial flow-through swamp
Decrease / Increase	metabolism of periphyton in bioassays / nutrients
Peri	?	 	11.6

26	14	100	constant	4	Hamala and Kollig 1985	microcosm, lab flowing
Decrease / Change	primary production, number of species, Chl-a and
biomass of periphyton / species composition	Peri	pp 16-d; > 21 d	 	10.2

27	30	1000	single	4	Johnson 1986	microcosm, lab stagnant	Decrease	gross
primary production; biomass;	Macro	>30 d	>	25.5

28	21	10	constant	4	Kosinski 1984; Kosinski and Merkle 1984	artificial
streams, recirculating 	Decrease	gross primary productivity; biovolume
of periphyton on atrifical substrate	Peri	> 21 d	<	0.4

29	21	1000	constant	4	Kosinski 1984; Kosinski and Merkle 1984	artificial
streams, recirculating 	Decrease	gross primary productivity; biovolume
of periphyton on atrifical substrate	Peri	> 21 d; 14 d	>	25.5

30	21	10000	constant	4	Kosinski 1984; Kosinski and Merkle 1984
artificial streams, recirculating 	Decrease	gross primary productivity;
biovolume of periphyton on atrifical substrate	Peri	> 21 d	>	25.5

31	12	24	constant	4	Krieger et al. 1988	artificial streams,
recirculating 	No effect / Decrease	uptake of phosphorous, silicium and
nitrogen by periphyton / Chl-a and biomass of periphyton	Peri	?	 	3.4

32	12	134	constant	4	Krieger et al. 1988	artificial streams,
recirculating 	No effect / Decrease	silicium uptake by periphyton /
uptake of phosphorus and nitrate by periphyton (slight); Chl-a and
biomass of pheriphyton	Peri	?	 	10.2

33	3	10000	single	4	Moorhead and Kosinski 1986	artificial streams,
recirculating 	No Effect / Decrease	conductivity, alkalinity, soluble
reactive phosphorous, respiration, species compositionof periphyton
(study probably too short) / pH, net primary production	Peri	7 d	>	6.8

34	21	337	constant	4	Pratt et al. 1988	microcosms, laboratory flowing
Decrease	DO, potassium, magnesium, calcium (slight), number of species,
protien biomass and Chl-a protozoa	Peri	> 21 d	 	25.5

35	42	200	single	4	Stay et al. 1989	microcosms, laboratory Leffler
Decrease	pH and primary production	Phyto	42 - >42 d	 	40.6

36	42	500	single	4	Stay et al. 1989	microcosms, laboratory Leffler
Decrease	pH and primary production	Phyto	> 42 d	>	40.6

37	42	1000	single	4	Stay et al. 1989	microcosms, laboratory Leffler
Decrease	pH and primary production	Phyto	> 42 d	>	40.6

38	42	5000	single	4	Stay et al. 1989	microcosms, laboratory Leffler
Decrease	pH and primary production	Phyto	> 42 d	>	40.6

39	70	50	constant	3	Brockway et al. 1984	microcosms, lab continuous flow
Decrease / Increase	Net O2-production / Nitrate	Phyto	1 d	 	8.9

40	70	100	constant	3	Brockway et al. 1984	microcosms, lab continuous
flow	Decrease / Increase	Net O2-production / Nitrate	Phyto	1 d; 2 d	 
15.8

41	20	100	single	3	deNoyelles et al. 1989	mesocoms, experimental ponds
Decrease	14C-uptake and biomass phytoplankton	Phyto	20 d	 	12.8

42	20	200	single	3	deNoyelles et al. 1989	mesocoms, experimental ponds
Decrease	14C-uptake and biomass phytoplankton; biomass phytoplankton
Phyto	20 d	 	25.5

43	63	68	constant	3	Jüttner et al. 1995	pond enclosures	No Effect /
Decrease	one algal species, rotifer / DO, conductivity (slight)	Phyto	?
 	14.9

44	21	100	constant	3	Kosinski 1984; Kosinski and Merkle 1984	artificial
streams, recirculating 	No effect / Decrease	biovolume of periphyton on
artifical substrate; gross primary productivity 	Peri	< 3 d	 	12.7

45	3	100	single	3	Moorhead and Kosinski 1986	artificial streams,
recirculating 	No Effect / Decrease	conductivity, alkalinity, soluble
reactive phosphorous, respiration, species compositionof periphyton
(study probably too short) / pH, net primary production	Peri	7 d	 	5.6

46	3	1000	single	3	Moorhead and Kosinski 1986	artificial streams,
recirculating 	No Effect / Decrease	conductivity, alkalinity, soluble
reactive phosphorous, respiration, species compositionof periphyton
(study probably too short) / pH, net primary production	Peri	7 d	>	6.8

47	60	60	single	3	Stay et al. 1985	microcosms, laboratory Taub	Decrease
DO, 14C-uptake, net primary production, respiration / 14C-uptake, Chl-a
Phyto	20 - 27 d; 53 d	 	8.9

48	60	100	single	3	Stay et al. 1985	microcosms, laboratory Taub	Decrease
DO, 14C-uptake, net primary production, respiration / 14C-uptake, Chl-a
Phyto	25 - 32 d; 53 d	 	15.8

49	60	200	single	3	Stay et al. 1985	microcosms, laboratory Taub	Decrease
DO, 14C-uptake, net primary production, respiration / 14C-uptake, Chl-a
Phyto	25 - 32 d; 53 d	 	40.6

50	42	100	single	3	Stay et al. 1989	microcosms, laboratory Leffler
Decrease	pH and primary production	Phyto	7 - >42 d	 	15.8

51	12	50	single	2	Brockway et al. 1984	microcosms, lab stagnant	Decrease
Net O2-production (slight)	Phyto	> 12 d	 	4.2

52	7	20	single	2	deNoyelles et al. 1982; Kettle 1982; deNoyelles et al.
1989	mesocoms, experimental ponds	Decrease / Change	14C-uptake and
biomass phytoplankton / composition of phytoplankton species; increase
in dinoflagellates	Phyto	7 d	 	0.1

53	30	10	single	2	Johnson 1986	microcosm, lab stagnant	Decrease	gross
primary production (slight)	Macro	7 d	 	0.4

54	30	100	single	2	Johnson 1986	microcosm, lab stagnant	Decrease	gross
primary production (slight)	Macro	7 d	 	12.8

55	63	22	constant	2	Jüttner et al. 1995	pond enclosures	No Effect /
Decrease	one algal species / DO, pH, conductivity (slight)	Phyto	?	 
0.5

56	63	10	constant	2	Jüttner et al. 1995	pond enclosures	No Effect /
Decrease	one algal species / DO, pH, conductivity (slight)	Phyto	?	<	0.5

57	14	1.89	 	2	Lakshminarayana et al 1992	stream, adjacent to agri 
tile drainage	Decrease	number of species and cell numbers (during low
flow)	phyto	150 d	<	0.1

58	18	1	 	2	Lampert et al 1989	lake enclosure	Decrease	primary
production 	phyto	14 d	<	0.1

59	21	32	constant	2	Pratt et al. 1988	microcosms, laboratory flowing	No
Effect / Decrease	potassium, protein biomass, Chl-a of Protozoa / DO,
magnesium, calcium (slight)	Peri	> 21 d	 	5.2

60	21	110	constant	2	Pratt et al. 1988	microcosms, laboratory flowing	No
Effect / Decrease	potassium, calcium, number of species, protien
biomass, Chl-a of Protozoa / DO, magnesium, (slight)	Peri	> 21 d	 	12.7

61	42	20	single	2	Stay et al. 1989	microcosms, laboratory Leffler
Decrease	primary production (slight)	Phyto	1-10 d	 	0.5

62	35	5	constant	2	van den Brink et al. 1995	microcosms, laboratory
Decrease	photosynthetic activity, as indicated by higher conductivity
and alkalinity, and lower Do and pH	Phyto	NA	<	0.4

63	7	0.5	single	1	Brockway et al. 1984	microcosms, lab stagnant	No
effect	Net O2 production	Phyto	NA	<	0.1

64	7	5	single	1	Brockway et al. 1984	microcosms, lab stagnant	No effect
Net O2 production	Phyto	NA	<	0.1

65	70	0.5	constant	1	Brockway et al. 1984	microcosms, lab continuous
flow	No effect	Net O2 production	Phyto	NA	<	0.1

66	70	5	constant	1	Brockway et al. 1984	microcosms, lab continuous flow
No effect	Net O2-production / Nitrate	Phyto	NA	<	0.1

67	14	5 (3)	constant	1	Gruessner and Watzin 1996	microcosm, lab	No
effect	Chl-a of periphyton on artificial substrate	Peri	NA	<	0.1

68	20	1	multiple	1	Gustavson and Wängberg 1995 (1)	lake enclosure (4)
No effect	community tolerance	Phyto	NA	<	0.1

69	20	20	multiple	1	Gustavson and Wängberg 1995 (1)	lake enclosure (4)
No effect	community tolerance	Phyto	NA	 	0.4

70	20	10	multiple	1	Gustavson and Wängberg 1995 (1)	lake enclosure (4)
No effect	community tolerance	Phyto	NA	<	0.4

71	28	2	2 short pulses 24-hours @	1	Jurgensen and Hoagland 1990	stream
enclosures	No Effect	cell density, biomass	Peri	NA	<	0.1

72	28	30	2 short pulses 24-hours @	1	Jurgensen and Hoagland 1990	stream
enclosures	No Effect	cell density, biomass	Peri	NA	 	5.2

73	28	100	2 short pulses 24-hours @	1	Jurgensen and Hoagland 1990	stream
enclosures	No Effect	cell density, biomass	Peri	NA	 	12.8

74	63	5	constant	1	Jüttner et al. 1995	pond enclosures	No Effect	one
algal species; DO, pH, conductivity	Phyto	NA	<	0.5

75	30	25	?	1	Lynch et al. 1985	artificial streams, laboratory (5)	No
Effect	standing biomass, rate of primary production, community
respiration	Peri	NA	 	5.1

76	21	3.2	constant	1	Pratt et al. 1988	microcosms, laboratory flowing	No
effect	DO, potassium,  magnesium, calcium	Peri	NA	<	0.1

77	21	10	constant	1	Pratt et al. 1988	microcosms, laboratory flowing	No
effect	DO, potassium,  magnesium, calcium	Peri	NA	<	0.4

(1) Study not included in Brock et al (2000)

(2) Esfenvalerate added

(3) Concentration of atrazine reduced to 1ug/l by day-14

(4) Copper added

(5) Use of DMSO as solvent



 

Appendix 2: Summary of Atrazine Monitoring Data

Table   STYLEREF 1 \s  VII -  SEQ Table \* ARABIC \s 1  1  Summary of
Detection Frequencies by Monitoring Site and Year

Site	Year	Non Detects	Total Samples	Frequency of Detection	Not Analyzed
Not Sampled

IA-01	2004	3	37	91.9



	2005	14	38	63.2



IA-02	2004	4	34	88.2	1



2005	14	39	64.1

1

IL-01	2004	0	37	100.0



	2005	22	38	42.1



IL-02	2004	0	37	100.0



	2005	8	38	78.9



IL-03	2005	25	38	34.2



	2006	27	38	28.9



IL-04	2005	11	36	69.4



	2006	8	38	78.9



IL-05	2004	0	35	100.0



	2005	15	36	58.3



IL-06	2004	3	33	90.9

2

	2005	29	36	19.4



IL-07	2004	3	34	91.2



	2005	7	36	80.6



IL-08	2005	0	35	100.0



	2006	4	37	89.2



IL-09	2004	0	34	100.0



	2005	0	36	100.0



IN-01	2004	4	37	89.2



	2005	15	36	58.3



IN-02	2004	0	37	100.0



	2005	0	36	100.0



IN-03	2005	0	36	100.0



	2006	7	40	82.5



IN-04	2004	0	37	100.0



	2005	3	32	90.6

4

	2006	9	39	76.9



IN-05	2004	0	37	100.0



	2005	1	36	97.2



	2006	7	40	82.5



IN-06	2005	15	36	58.3



	2006	11	40	72.5



IN-07	2005	5	36	86.1



	2006	15	40	62.5



IN-08	2005	5	36	86.1



	2006	6	40	85.0



IN-09	2005	1	33	97.0

3

	2006	6	40	85.0



IN-10	2005	2	36	94.4



	2006	8	40	80.0



IN-11	2005	0	30	100.0

6

	2006	6	40	85.0



KY-01	2005	8	35	77.1



	2006	14	36	61.1



KY-02	2005	0	32	100.0

3

	2006	2	28	92.9	3	5

MN-01	2005	11	37	70.3



	2006	30	37	18.9

3

MO-01	2004	0	34	100.0



	2005	0	37	100.0



	2006	5	33	84.8

2

MO-02	2004	0	34	100.0



	2005	0	36	100.0



	2006	1	35	97.1



MO-03	2004	0	34	100.0



	2005	0	33	100.0

2

	2006	2	21	90.5

14

NE-01	2004	3	36	91.7	2



2005	11	36	69.4	1	1

NE-02	2005	5	36	86.1



	2006	15	40	62.5



NE-03	2004	3	38	92.1



	2005	13	37	64.9



NE-04	2005	0	22	100.0

15

	2006	1	10	90.0

30

NE-05	2005	2	24	91.7

13

	2006	7	15	53.3

25

NE-06	2004	3	36	91.7

2

	2005	11	33	66.7

4

	2006	24	27	11.1

13

NE-07	2005	5	29	82.8

8

	2006	2	3	33.3

37

OH-01	2004	9	38	76.3



	2005	11	35	68.6



OH-02	2005	14	35	60.0



	2006	8	40	80.0



OH-03	2004	16	38	57.9



	2005	22	35	37.1



OH-04	2005	8	35	77.1



	2006	20	40	50.0



TN-01	2005	0	34	100.0



	2006	13	35	62.9





Table   STYLEREF 1 \s  VII -  SEQ Table \* ARABIC \s 1  2  Summary of
Rolling Averages From Ecological Watershed Monitoring Data for
Comparison with CASM Thresholds

Site	Year	Max. concentration (ug/l)	Annual average (ug/L)



Peak	14-day	21-day	30-day	60-day	90-day

	IA-01	2004	10.0	3.7	2.6	2.1	1.2	0.9	0.3

0.2

	2005	1.2	0.5	0.4	0.3	0.3	0.3

	IA-02	2004	1.8	1.1	1.0	0.8	0.6	0.5	0.2

0.2

	2005	5.5	2.1	1.5	1.4	0.8	0.6

	IL-01	2004	13.2	6.6	5.0	4.1	2.5	1.9	0.6

0.2

	2005	0.6	0.3	0.3	0.3	0.3	0.3

	IL-02	2004	4.9	2.9	2.6	2.3	1.5	1.1	0.4

0.2

	2005	2.9	1.8	1.4	1.1	0.6	0.5

	IL-03	2005	5.6	1.8	1.3	1.0	0.6	0.4	0.2

0.1

	2006	2.5	0.9	0.7	0.5	0.3	0.2

	IL-04	2005	2.8	1.4	1.0	0.8	0.5	0.4	0.2

0.4

	2006	11.5	3.4	2.4	1.8	1.8	1.4

	IL-05	2004	22.1	7.2	5.0	3.6	2.0	1.4	0.4

0.1

	2005	1.8	0.7	0.5	0.4	0.3	0.2

	IL-06	2004	2.2	1.1	0.9	0.7	0.5	0.4	0.2

0.1

	2005	0.2	0.2	0.2	0.2	0.2	0.2

	IL-07	2004	21.8	7.0	5.3	4.2	2.4	1.7	0.5

0.2

	2005	2.3	0.9	0.7	0.6	0.5	0.4

	IL-08	2005	5.6	4.4	3.6	2.8	1.8	1.4	0.6

0.9

	2006	33.1	11.0	7.6	8.1	4.4	3.0

	IL-09	2004	13.2	8.1	6.8	6.3	4.6	3.4	1.1

0.7

	2005	16.0	6.2	4.6	3.4	2.3	1.8

	IN-01	2004	8.6	4.0	3.2	3.5	2.4	1.6	0.6

0.3

	2005	4.4	1.4	1.0	1.0	0.7	0.5

	IN-02	2004	9.3	6.3	5.0	4.5	2.8	2.1	0.7

0.7

	2005	20.3	6.3	4.7	4.3	3.0	2.1

	IN-03	2005	7.6	4.3	3.4	3.3	2.3	1.8	0.6

0.9

	2006	16.9	10.6	8.0	6.2	3.9	2.9

	IN-04	2004	78.1	23.8	16.3	12.0	6.4	4.4	1.2

0.4

0.5

	2005	8.8	3.6	2.6	2.1	1.4	1.0



2006	10.2	5.6	4.5	3.7	2.2	1.6

	IN-05	2004	28.9	14.9	15.5	11.9	7.0	4.9	1.4

1.1

1.5

	2005	17.3	7.8	5.8	4.5	4.1	3.5



2006	41.3	17.9	14.1	13.1	7.4	5.4

	IN-06	2005	7.2	2.9	2.4	1.8	1.0	0.7	0.2

0.5

	2006	9.4	4.0	3.4	2.7	1.9	1.4

	IN-07	2005	22.6	9.6	7.2	6.4	3.9	2.7	0.8

0.4

	2006	10.5	5.4	4.1	3.6	2.0	1.4

	IN-08	2005	21.1	6.9	5.5	4.9	2.8	2.0	0.6

0.9

	2006	20.7	8.9	7.6	7.7	4.4	3.1

	IN-09	2005	9.4	3.7	2.8	2.4	1.7	1.3	0.4

0.3

	2006	8.3	3.0	2.1	1.8	1.2	0.9

	IN-10	2005	12.4	6.1	4.6	4.0	2.8	2.2	0.6

0.8

	2006	16.4	7.5	6.1	6.3	3.6	2.6

	IN-11	2005	208.8	65.1	44.4	31.5	16.2	11.3	3.3

0.5

	2006	9.8	5.9	4.4	3.3	1.9	1.4

	KY-01	2005	2.2	1.5	1.3	1.2	0.9	0.7	0.2

0.4

	2006	22.4	6.9	4.8	3.6	1.9	1.3

	KY-02	2005	19.3	8.7	8.7	7.1	4.5	3.6	1.3

0.6

	2006	14.3	4.7	4.2	3.9	2.3	1.7

	MN-01	2005	5.8	2.6	1.9	1.4	0.8	0.6	0.2

0.1

	2006	0.2	0.2	0.1	0.1	0.1	0.1

	MO-01	2004	65.9	39.6	29.6	28.6	19.4	13.8	3.9

4.8

3.2

	2005	182.8	78.1	54.2	42.5	25.7	17.8



2006	82.8	48.2	41.1	31.6	17.5	12.0

	MO-02	2004	53.8	33.0	29.7	25.9	16.8	12.3	4.1

3.6

3.3

	2005	28.1	18.7	17.0	14.6	11.5	9.1



2006	43.2	34.7	31.6	27.4	15.4	11.5

	MO-03	2004	59.0	23.3	17.2	13.1	8.1	6.1	2.2

1.8

0.9

	2005	12.3	8.7	7.5	6.9	5.5	4.4



2006	3.9	2.3	2.0	1.9	1.5	1.3

	NE-01	2004	19.2	13.0	9.8	7.5	4.3	3.0	0.8

0.7

	2005	16.7	6.6	5.2	5.6	3.6	2.5

	NE-02	2005	20.7	11.4	11.8	10.7	6.3	4.3	1.1

1.3

	2006	82.0	28.6	19.5	14.1	7.3	5.0

	NE-03	2004	2.3	1.1	1.2	1.0	0.6	0.5	0.2

0.3

	2005	11.9	3.7	2.5	2.1	1.2	0.8

	NE-04	2005	36.0	36.0	34.8	27.3	17.4	11.9	3.1

0.4

	2006	4.1	4.1	4.0	3.1	1.7	1.2

	NE-05	2005	49.9	23.8	23.3	19.9	16.5	11.4	2.9

0.9

	2006	6.8	6.8	6.8	6.8	5.1	3.5

	NE-06	2004	7.7	2.8	2.8	2.1	1.7	1.6	0.4

1.1

0.1

	2005	33.1	20.6	13.9	11.4	6.0	4.1



2006	0.1	0.1	0.1	0.1	0.1	0.1

	NE-07	2005	112.2	80.0	61.9	45.2	22.7	16.6	4.5

***

	2006	***	***	***	***	***	***

	OH-01	2004	18.3	8.8	7.7	5.7	3.2	2.2	0.6

0.2

	2005	3.0	1.0	0.7	0.9	0.6	0.5

	OH-02	2005	18.1	7.1	5.1	4.0	2.9	2.1	0.6

0.6

	2006	14.0	5.9	5.8	5.2	2.9	2.1

	OH-03	2004	21.5	9.1	8.1	7.3	4.5	3.0	0.8

0.2

	2005	8.2	3.0	2.1	1.6	0.9	0.6

	OH-04	2005	20.2	8.0	5.7	4.7	2.7	1.9	0.5

0.2

	2006	6.3	2.6	2.3	1.7	1.0	0.7

	TN-01	2005	7.6	5.6	4.8	4.0	2.9	2.2	0.8

0.6

	2006	10.7	3.6	3.1	2.4	2.0	1.6

	

Table   STYLEREF 1 \s  VII -  SEQ Table \* ARABIC \s 1  3  Sites Ranked
from Lowest to Highest Based on Annual Average Flow

Site	Year	Annual Average Flow (m3/s)	Annual Average Flow (cfs)

NE 06	2005	0.0295	1.0418

MO 01	2006	0.06513	2.3001

MO 01	2004	0.0772	2.7263

NE 03	2004	0.0774	2.7334

MO 01	2005	0.08858	3.1282

NE 06	2004	0.1026	3.6233

NE 06	2006	0.1245	4.3967

NE 03	2005	0.1364	4.8170

KY 01	2005	0.1855	6.5509

NE 04	2006	0.1866	6.5898

IN 05	2005	0.2044	7.2184

IN 05	2004	0.2344	8.2778

NE 07	2006	0.2425	8.5639

IN 05	2006	0.2623	9.2631

NE 04	2005	0.2691	9.5033

IN 11	2005	0.3237	11.4315

TN 01	2006	0.3389	11.9683

IN 04	2006	0.3555	12.5545

IN 02	2005	0.4027	14.2214

KY 01	2006	0.405	14.3026

IN 04	2005	0.4182	14.7687

IL 01	2005	0.4262	15.0513

IN 10	2005	0.4313	15.2314

NE 07	2005	0.4352	15.3691

MN 01	2006	0.4661	16.4603

TN 01	2005	0.4876	17.2196

IN 03	2005	0.5519	19.4903

IL 06	2005	0.5803	20.4933

IA 02	2004	0.5893	20.8111

IL 07	2005	0.5896	20.8217

IA 02	2005	0.5971	21.0866

OH 03	2004	0.5985	21.1360

OH 03	2005	0.6065	21.4185

IL 04	2005	0.6073	21.4468

NE 05	2006	0.6329	22.3509

MO 03	2006	0.7017	24.7805

MO 02	2005	0.7322	25.8576

OH 01	2005	0.7636	26.9665

OH 02	2005	0.7761	27.4080

IL 09	2005	0.7808	27.5740

IN 08	2005	0.7816	27.6022

MO 02	2004	0.8015	28.3050

MN 01	2005	0.8021	28.3262

IN 10	2006	0.8311	29.3503

MO 02	2006	0.8407	29.6893

IL 07	2004	0.8412	29.7070

IL 06	2004	0.8502	30.0248

IN 02	2004	0.8655	30.5651

IL 09	2004	0.9329	32.9454

OH 1	2004	0.9805	34.6264

NE 05	2005	0.9973	35.2196

OH 02	2006	1.0262	36.2403

IN 03	2006	1.0295	36.3568

IN 11	2006	1.0419	36.7947

IL 03	2006	1.0674	37.6952

NE 02	2006	1.0733	37.9036

IN 08	2006	1.1999	42.3745

NE 01	2004	1.2133	42.8477

IN 04	2004	1.2165	42.9607

IL 03	2005	1.3122	46.3403

KY 02	2005	1.3156	46.4604

IA 01	2004	1.4214	50.1967

IL 02	2005	1.5879	56.0767

IN 09	2005	1.5895	56.1332

NE 01	2005	1.6282	57.4999

KY 02	2006	1.8037	63.6977

IN 01	2004	1.8082	63.8566

MO 03	2005	1.9744	69.7259

IA 01	2005	1.9808	69.9520

IL 02	2004	2.0305	71.7071

IN 09	2006	2.2518	79.5223

IL 05	2005	2.3536	83.1174

NE 02	2005	2.402	84.8266

MO 03	2004	2.4098	85.1021

IL 04	2006	2.7688	97.7802

IL 01	2004	2.8955	102.2546

IL 05	2004	2.9136	102.8938

IL 08	2006	3.5539	125.5060

IN 06	2005	3.763	132.8903

IN 01	2005	4.1299	145.8474

IL 08	2005	4.1452	146.3877

OH 04	2005	5.4524	192.5515

IN 07	2005	5.6568	199.7699

IN 07	2006	5.7476	202.9765

OH 04	2006	5.9098	208.7046

IN 06	2006	6.0462	213.5216





Table   STYLEREF 1 \s  VII -  SEQ Table \* ARABIC \s 1  4  Summary of
Rolling Average Concentrations (ppb) and SSI% for PRZM Augmented
Chemographs from Snyder, et al (2007)



PRZM Augmented

Site	Year	14-day rolling average	30-day rolling average	60-day rolling
average	90-day rolling average	SSI%

IA 01	2004	10.35	7.83	5.07	3.46	1.5

IA 01	2005	4.71	2.6	2.06	1.46	0

IA 02	2004	7.74	5.08	3.51	2.51	0.4

IA 02	2005	4.21	3.29	2.68	1.86	0.4

IL 01	2004	7.93	5.85	3.91	2.81	0

IL 01	2005	3.12	2.56	1.42	1	0.4

IL 02	2004	13.53	8.08	5.28	3.69	1.7

IL 02	2005	2.37	1.2	0.74	0.81	0

IL 03	2005	8.72	4.83	2.81	2.17	1.1

IL 03	2006	3.48	1.96	1.12	1.07	0.4

IL 04	2005	2.84	2.1	1.11	0.85	0

IL 04	2006	6.97	5.8	4.53	3.41	0.7

IL 05	2004	11.94	9.12	5.46	3.95	2

IL05	2005	3.91	2.12	1.1	0.8	0

IL 06	2004	6.86	6.04	3.76	2.64	0.73

IL 06	2005	2.39	1.14	0.81	0.56	0.3

IL 07	2004	12.45	9.23	5.49	3.78	1.6

IL 07	2005	6.78	3.55	2.01	1.79	0.9

IL 08	2005	7.25	5.44	4.27	3.21	1.2

IL 08	2006	17.43	11.34	6.61	5.34	3.6

IL 09	2004	12.47	11.05	7.92	5.61	2.4

IL 09	2005	7.3	5.35	3.7	3.51	1.4

IN 01	2004	13.27	9.02	7.75	5.39	2.9

IN 01	2005	4.56	3.2	2.67	1.88	0.8

IN 02	2004	8.34	7.2	4.63	3.26	0.8

IN 02	2005	5.84	3.96	2.91	2.13	0

IN 03	2005	4.2	3.27	2.29	1.89	0

IN 03	2006	11.32	7.59	6.48	4.75	0.7

IN 04	2004	29.99	17.87	10.61	7.23	4.7

IN 04	2005	17.57	9.19	7.6	5.37	3

IN 04	2006	11.26	6.67	4.02	2.84	1

IN 05	2004	15.2	12.44	8.22	5.86	1.4

IN 05	2005	7.97	5.63	3.99	4.04	0.1

IN 05	2006	12.3	10.64	7.05	5.36	0.9

IN 06	2005	3.44	2.12	1.12	0.89	0

IN 06	2006	5.07	3.42	2.19	1.96	0

IN 07	2005	6.8	5.4	3.59	2.58	0.5

IN 07	2006	8.76	5.65	4.41	3.09	0.5

IN 08	2005	6.93	4.92	2.79	2.31	0

IN 08	2006	11.12	9.14	5.49	4.24	1.1

IN 09	2005	11.59	6.81	5.86	4.23	2.5

IN 09	2006	4.89	3.06	2.27	1.86	0

IN 10	2005	6.02	5.21	3.99	3.08	0

IN 10	2006	11.2	8.95	5.12	3.87	1.1

IN 11	2005	45.35	22.39	12.45	8.67	5.7

IN 11	2006	7.29	5.22	3.25	2.33	0

MO 01	2004	28.1	21.79	16.66	11.9	5.7

MO 01	2005	65.31	36.89	23.2	11.9	5.7

MO 01	2006	44.39	31	19.19	13.17	7.7

MO 02	2004	31.22	25.65	16.94	12.53	5

MO 02	2005	19.61	17.04	14.4	10.98	3.9

MO 02	2006	30.27	25.23	15.33	11.25	5.4

MO 03	2004	18.91	11.48	7.93	5.95	1.6

MO 03	2005	10.47	8.53	6.63	5.17	0

MO 03	2006	5.18	3.22	2.16	1.8	0

NE 01	2004	13.38	7.77	4.45	3.18	0

NE 01	2005	8.29	7.41	5.1	3.51	0.8

NE 02	2005	10.86	9.78	6.24	4.29	0.4

NE 02	2006	28.15	15.97	8.29	5.59	3.7

NE 03	2004	7.23	4.01	2.29	1.6	1

NE 03	2005	3.88	2.35	1.33	0.95	0

NE 06	2004	5.37	3.13	2.84	2.31	0.4

NE 06	2005	14.71	9.25	5.06	3.43	0.7

NE 06	2006	1.54	1.41	0.74	0.52	0.01

KY 01	2005	1.88	1.52	1.09	0.93	0

KY 01	2006	7.4	4.95	3.18	2.33	0.4

KY 02	2005	10.73	8.34	5.25	4.16	0.4

KY 02	2006	7.53	7.11	4.11	2.91	0

MN 01	2005	5.13	3.81	2.29	1.71	0.4

MN 01	2006	2.54	1.56	1.05	0.87	0.4

OH 01	2004	8.05	6.03	3.7	2.62	0

OH 01	2005	2.08	1.67	1.15	0.91	0

OH 02	2005	7.7	4.51	4.04	3.48	0.4

OH 02	2006	7.27	6.54	4.17	3	0

OH 03	2004	8.97	7.34	6.51	4.38	1.8

OH 03	2005	4.26	3.44	1.82	1.23	0.4

OH 04	2005	7.85	6.35	4.93	3.94	1.7

OH 04	2006	5.79	4.6	3.3	2.33	0.7

TN 01	2005	4.92	3.78	3.04	2.3	0

TN 01	2006	3.62	2.67	1.85	1.62	0



 Appendix 3: Comparisons of precipitation during monitoring years to
historical averages

The monthly precipitation measurements reported for the monitoring sites
were compared to percentiles for the historical monthly totals for the
site. The historical weather data, provided to EPA by Syngenta, were
collected by the National Weather Service. The data, cited in Hampton et
al (2007a), comes from these sources:

National Oceanic & Atmospheric Administration (NOAA). 2002a. Cooperative

Summary of the Day TD3200. National Climatic Data Center, Asheville, NC.

Eastern United States, Puerto Rico, and the Virgin Islands CD, version
1.0, released

November 2002.

National Oceanic & Atmospheric Administration (NOAA). 2002b. Cooperative

Summary of the Day TD3200. National Climatic Data Center, Asheville,
N.C.

Central United States CD, version 1.0, released November 2002.

National Oceanic & Atmospheric Administration (NOAA). 2005. National
Data

Centers, U.S. Department of Commerce, National Oceanic and Atmospheric

Administration, National Environmental Satellite, Data and Information
Service

http://www.nndc.noaa.gov.

The crop planting dates in the illustrations come from USDA National
Agricultural Statistics Service (NASS) for the state and/or crop
reporting district (Hampton et al., 2007a).

IA-01, 2004-05

Watershed Location: Wolf Creek, IA

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1978-2001

Max	6.60	7.52	10.77	12.25	13.72	37.78

90th %ile	5.94	6.45	7.59	8.08	7.93	28.36

75th %ile	3.85	5.03	6.77	6.17	4.70	23.11

50th %ile	2.90	3.66	4.89	4.03	2.91	20.69

25th %ile	1.59	2.69	3.26	2.02	2.42	16.40

10th %ile	1.04	1.91	1.51	1.54	1.59	13.49

Min	0.65	0.64	0.88	0.39	0.39	10.51

Monthly totals during the monitoring study

2004	2.79	7.85	2.84	3.11	4.85	21.44

2005	2.38	3.58	9.22	3	1.68	19.86



The monthly precipitation for May 2004 and June 2005 exceeded the
historical 75th percentile for the monitoring site. Overall, the
precipitation for the remaining months generally fell within the 25th to
75th percentile range. 

The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IA-02, 2004-05

Watershed Location: Nishnabotna River, IA

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1949-2001

Max	8.19	13.29	9.99	16.18	16.63	38.71

90th %ile	6.12	8.20	7.58	9.78	6.85	30.38

75th %ile	3.53	5.46	5.79	6.29	4.74	23.84

50th %ile	2.65	4.02	4.25	4.33	3.39	20.14

25th %ile	1.81	2.98	3.00	2.63	2.08	16.57

10th %ile	1.35	1.74	2.07	1.45	1.38	14.62

Min	0.82	1.02	1.04	0.22	0.46	12.01

Monthly totals during the monitoring study

2004	1.28	10.38	3.48	3.4	2.15	20.69

2005	3.14	3.93	4.64	2.79	1.09	15.59



The monthly precipitation in 2004 fluctuated from low (10th percentile)
for April to high (>90th percentile) for May 2004; precipitation for the
remaining months fell between the 25th to 50th percentile. Monthly
precipitation totals in 2005 were roughly at the median for April
through June, dropping at or below the 25th percentile in July and
August.  

The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IL-01, 2004-05

Watershed Location: Pecatonica River, IL

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	6.26	9.22	11.48	8.98	9.59	29.60

90th %ile	5.52	6.46	7.66	7.14	6.82	24.90

75th %ile	4.09	5.24	5.15	5.21	4.37	21.96

50th %ile	3.14	3.61	3.64	3.50	3.54	19.06

25th %ile	2.11	2.16	2.78	2.54	2.48	16.42

10th %ile	1.64	1.28	1.71	1.88	1.51	12.43

Min	0.76	0.56	0.31	0.63	0.48	8.62

Monthly totals during the monitoring study

2004	2.35	5.78	5.3	5.12	2.27	20.82

2005	1.18	1.67	2.7	1.83	4.92	12.3



Monthly precipitation in 2004 was low (around the 25th percentile) in
April and August and high (ar or above the 75th percentile) in May
through July. IN 2005, monthly precipitation was consistently below the
25th percentile for April through July.

The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IL-02, 2004-05

Watershed Location: Pine Creek, IL

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1950-51, 1987-2001

Max	6.74	13.19	10.91	9.12	10.36	32.10

90th %ile	5.92	8.77	9.57	8.09	6.87	29.44

75th %ile	4.59	5.23	6.58	4.73	4.78	26.18

50th %ile	2.60	3.05	3.51	2.77	3.54	17.87

25th %ile	1.59	1.35	2.80	2.00	1.57	12.92

10th %ile	1.02	0.61	0.63	1.57	1.29	10.59

Min	0.61	0.11	0.38	0.75	0.87	8.73

Monthly totals during the monitoring study

2004	1.45	8.81	3.17	3.23	2.97	19.63

2005	1.23	1.47	1.32	2.12	2.56	8.7



Monthly precipitation in 2004 was low (around the 25th percentile) in
April, high in May (90th percentile), and at or around the 50th
percentile in subsequent months. In 2005, monthly precipitation was at
or below the 75th percentile from April through July. 

The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IL-03, 2005-06

Watershed Location: Spring Creek Watershed, IL

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.31	12.19	11.58	12.85	13.69	31.64

90th %ile	6.02	7.03	6.98	7.22	8.11	27.38

75th %ile	5.04	4.73	5.34	5.04	5.79	23.22

50th %ile	3.36	3.55	3.88	3.80	3.07	19.00

25th %ile	2.51	2.33	2.26	2.35	2.15	17.04

10th %ile	1.69	1.77	1.53	1.76	1.01	13.44

Min	1.35	0.55	0.60	0.35	0.52	11.06

Monthly totals during the monitoring study

2005	1.52	1.67	1.49	1.65	1.36	7.69

2006	5.41	1.55	0.36	3.08	5.88	16.28



Monthly precipitation for 2005 was consistently at or below the 10th
percentile for April through July. Precipitation in 2006 was above the
75th percentile in April but fell below the 10th percentile in May and
June (June was the lowest total for the record period).

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IL-04, 2005-06

Watershed Location: Iroquois River Watershed, IL

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-51, 1973-2001

Max	7.84	9.22	9.47	10.55	7.70	31.00

90th %ile	6.09	6.58	7.04	6.91	6.38	24.95

75th %ile	5.05	5.89	5.57	5.92	4.07	22.64

50th %ile	3.29	4.55	4.16	3.69	2.61	20.70

25th %ile	2.23	3.04	2.70	3.08	1.55	16.78

10th %ile	1.82	1.78	1.77	2.00	1.19	14.43

Min	1.07	0.67	0.03	1.49	0.62	8.61

Monthly totals during the monitoring study

2005	2.37	1.09	0.96	3.60	3.50	7.69

2006	5.28	5.49	1.23	4.54	4.31	16.28



Monthly precipitation for 2005 was near the 25th percentile in April and
below the 10th percentile in May and June. Rainfall was at or above the
median in July and August. In 2006, precipitation was near the 75th
percentile in all but June, which was below the 10th percentile.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IL-05, 2004-05

Watershed Location: Panther Creek, IL

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.96	9.68	11.15	12.82	9.18	35.79

90th %ile	6.20	6.62	7.99	6.96	6.40	25.24

75th %ile	5.25	4.57	5.45	5.06	4.55	23.08

50th %ile	4.03	3.78	4.00	3.83	2.78	19.19

25th %ile	2.82	2.51	2.14	2.21	1.86	16.46

10th %ile	1.69	1.67	1.55	1.60	0.92	12.83

Min	0.69	0.47	0.17	0.55	0.25	5.97

Monthly totals during the monitoring study

2004	2.02	4.92	2.76	2.85	4.67	17.22

2005	2.59	0.82	0.83	2.86	4.13	11.23



Monthly precipitation for 2004 was low (25th to <10th percentile) for
April through June. In 2005, rainfall was low (below the 25th
percentile) in April and high (75th percentile) in May and August.

The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IL-06, 2004-05

Watershed Location: Sugar Creek West Fork, IL

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.96	9.68	11.15	12.82	9.18	35.79

90th %ile	6.20	6.62	7.99	6.96	6.40	25.22

75th %ile	5.25	4.57	5.45	5.06	4.55	23.03

50th %ile	4.03	3.78	4.00	3.83	2.78	18.92

25th %ile	2.82	2.51	2.14	2.21	1.86	16.31

10th %ile	1.69	1.67	1.55	1.60	0.92	12.84

Min	0.69	0.47	0.17	0.55	0.25	5.97

Monthly totals during the monitoring study

2004	1.94	6.19	2.95	3.13	3.56	17.77

2005	1.23	1.09	0.68	2.59	3.61	9.2



Monthly precipitation for 2004 was low (<25th percentile) for April and
high (>75th percentile) for May.  In 2005, rainfall was low (below the
10th percentile) in April through June. 

The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IL-07, 2004-05

Watershed Location: Grindstone Creek, IL

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.43	11.12	9.43	14.92	8.87	31.58

90th %ile	5.26	7.31	6.78	7.88	5.78	27.55

75th %ile	4.77	5.22	5.52	4.90	4.58	22.62

50th %ile	3.63	3.69	4.34	3.62	3.13	18.83

25th %ile	2.26	2.47	2.70	2.54	2.29	16.46

10th %ile	1.88	1.76	1.43	1.40	1.55	13.88

Min	0.74	0.47	0.20	0.71	0.46	9.04

Monthly totals during the monitoring study

2004	1.92	5.91	2.38	3.16	5.98	19.35

2005	4.26	1.8	1.12	2.23	3.56	12.97



Monthly precipitation for 2004 was low (10th percentile) for April and
high (>75th percentile) for May and August.  In 2005, rainfall was low
(below the 25th percentile) in May through July. 

The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IL-08, 2005-06

Watershed Location: Horse Creek Watershed, IL

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1973-2001

Max	9.93	9.17	8.96	8.70	6.82	33.81

90th %ile	5.32	7.78	6.92	7.13	5.84	22.82

75th %ile	4.76	6.34	5.43	4.16	3.84	21.75

50th %ile	3.61	3.44	4.27	2.70	3.03	19.15

25th %ile	1.91	2.48	2.83	2.08	1.80	15.81

10th %ile	1.28	1.62	2.09	1.42	1.09	14.15

Min	0.74	0.89	0.69	1.12	0.60	6.56

Monthly totals during the monitoring study

2005	1.78	1.95	2.37	1.46	2.51	10.07

2006	3.54	1.89	1.2	1.95	5.38	13.96



Monthly precipitation for 2005 fell between 25th and 10th percentile
from April through July. While precipitation in April 2006 was near the
median, monthly totals for May through July were below the 25th
percentile.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IL-09, 2004-05

Watershed Location: Muddy Creek, IL

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	8.62	10.45	13.98	10.10	10.41	32.09

90th %ile	6.80	6.48	7.20	8.01	5.92	25.91

75th %ile	5.12	5.20	5.08	5.86	4.08	22.62

50th %ile	3.44	3.69	3.66	3.49	2.78	18.57

25th %ile	2.22	2.65	2.57	2.54	1.81	16.42

10th %ile	1.66	1.84	1.85	1.57	1.16	12.35

Min	0.22	0.81	0.70	1.26	0.52	8.86

Monthly totals during the monitoring study

2004	2.86	7.07	2.18	5.26	2.91	20.28

2005	2.45	2.16	2.72	2.54	1.93	11.8



Monthly precipitation for 2004 was high in May (90th percentile) and low
(<25th percentile) in June.  In 2005, rainfall was low (at or below the
25th percentile) throughout the sample period (April through August). 

The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IN-01, 2004-05

Watershed Location: Mill Creek, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	8.24	7.32	10.32	9.60	12.39	34.40

90th %ile	6.20	5.80	6.29	7.66	5.89	25.61

75th %ile	4.67	5.00	5.40	5.23	4.21	21.81

50th %ile	3.90	3.78	3.79	3.84	3.18	20.15

25th %ile	2.91	2.74	2.96	2.83	2.04	16.43

10th %ile	2.15	2.21	2.38	1.86	1.44	15.29

Min	1.05	0.83	1.34	0.75	0.63	9.94

Monthly totals during the monitoring study

2004	1.53	5.34	4.27	4.76	10.41	26.31

2005	1.94	1.27	3.15	4.86	1.87	13.09



Monthly precipitation for 2004 was high in May (above the 75th
percentile) and August (>90th percentile). In 2005, rainfall was low (at
or below the 10th percentile) in April, May, and August. 

The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IN-02, 2004-05

Watershed Location: Eel River, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	8.57	7.31	10.55	8.63	9.02	26.75

90th %ile	5.85	6.16	6.23	6.92	6.35	24.22

75th %ile	4.77	5.06	5.32	5.13	4.92	21.44

50th %ile	3.54	3.69	3.82	3.70	3.34	19.84

25th %ile	2.52	2.75	2.79	2.96	1.93	17.39

10th %ile	1.67	2.27	1.93	1.90	1.60	14.92

Min	1.03	0.43	0.69	1.05	0.90	12.14

Monthly totals during the monitoring study

2004	1.13	6.38	5.59	3.12	5.93	22.15

2005	1.77	2.09	1.83	4.22	2.56	12.47



Monthly precipitation for 2004 was high (above the 75th percentile) in
May, June, and August. In 2005, rainfall was low (at or below the 10th
percentile) in April, May, and June. 

The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IN-03, 2005-06

Watershed Location: Eightmile Creek Watershed, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1971-2001

Max	6.61	8.16	8.10	10.03	8.77	26.21

90th %ile	4.84	6.33	6.36	6.71	6.08	23.44

75th %ile	4.27	4.94	4.72	4.87	4.71	21.26

50th %ile	3.13	4.16	3.79	3.51	3.43	19.20

25th %ile	2.03	2.77	2.44	2.32	2.39	16.71

10th %ile	1.77	2.03	1.83	1.64	1.63	14.36

Min	0.96	0.51	0.50	0.23	0.66	10.36

Monthly totals during the monitoring study

2005	2.24	1.63	2.36	4.78	1.03	12.04

2006	4.18	5.77	3.76	3.45	2.57	19.73



Monthly precipitation for 2005 was at or below the 25th percentile from
April through June (NOTE: precipitation wasn’t recorded for the full
month of August in 2005). July was a wetter than normal month. In 2006,
monthly precipitation totals were at or above the median from April
through July.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IN-04, 2004-06

Watershed Location: Rock Creek Watershed, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.54	8.59	12.16	9.38	8.91	30.74

90th %ile	6.23	5.91	7.46	7.40	5.31	27.23

75th %ile	4.74	5.24	5.87	5.06	4.44	23.23

50th %ile	3.76	4.14	3.86	3.99	3.13	19.43

25th %ile	2.67	2.86	2.87	2.90	2.27	16.95

10th %ile	2.07	2.00	1.60	2.15	1.69	15.17

Min	1.08	0.85	0.90	0.38	0.78	9.94

Monthly totals during the monitoring study

2004	1.8	5.75	5.18	7.56	4.86	25.15

2005	1.77	1.51	3.33	7.09	2.95	16.65

2006	3.64	6.01	4.37	3.43	5.12	22.57



Monthly precipitation in 2004 was high (near the 90th percentile) in
May, July, and August. In 2005, monthly precipitation was low (at or
below the 10th percentile) in April and May and high (>75th percentile)
in July. In 2006, monthly precipitation totals were high (>75th
percentile) in May and August.

The following figures show the daily precipitation and atrazine residues
in water for 2004, 2005 and 2006. The planting season is shaded to
provide a point of reference for the likely timing of applications in
the watershed.

IN-05, 2004-06

Watershed Location: Limber Lost Creek Watershed, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1971-2001

Max	6.61	8.16	8.10	10.03	8.77	26.21

90th %ile	4.84	6.33	6.36	6.71	6.08	23.44

75th %ile	4.27	4.94	4.72	4.87	4.71	21.26

50th %ile	3.13	4.16	3.79	3.51	3.43	19.20

25th %ile	2.03	2.77	2.44	2.32	2.39	16.71

10th %ile	1.77	2.03	1.83	1.64	1.63	14.36

Min	0.96	0.51	0.50	0.23	0.66	10.36

Monthly totals during the monitoring study

2004	2.69	4.36	4.19	3.39	2.97	17.6

2005	4.35	1.34	2.99	1.99	1.18	11.85

2006	3.27	3.94	2.87	5.98	4.45	20.51



Monthly precipitation in 2004 was near the median throughout the
sampling period. In 2005, monthly precipitation was high (75th
percentile) in April and low (at or below the 10th percentile) in May,
July, and August. In 2006, monthly precipitation totals were high (75th
percentile) in July and August.

The following figures show the daily precipitation and atrazine residues
in water for 2004, 2005 and 2006. The planting season is shaded to
provide a point of reference for the likely timing of applications in
the watershed.

IN-06, 2005-06

Watershed Location: Vermillion River, North Fork Watershed, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	8.99	9.78	12.68	12.69	11.02	30.64

90th %ile	6.90	5.67	7.53	7.33	5.53	24.09

75th %ile	5.12	4.60	5.88	5.70	4.45	21.11

50th %ile	3.59	3.73	4.00	4.14	3.07	19.59

25th %ile	2.50	2.95	2.81	2.94	1.79	17.51

10th %ile	1.94	1.47	1.96	1.37	1.33	15.59

Min	1.13	0.44	0.45	0.22	0.61	6.76

Monthly totals during the monitoring study

2005	2.46	0.68	1.87	5.79	1.85	12.65

2006	4.84	3.81	1.55	4.74	2.7	17.64



Monthly precipitation for 2005 was at or below the 25th percentile from
April through June. July was a wetter than normal month. In 2006,
monthly precipitation totals were at or above the median in all months
except June, where the monthly total was below the 10th percentile.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IN-07 2005-06

Watershed Location: White River Watershed, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	9.45	7.17	10.55	10.66	8.63	30.58

90th %ile	5.71	6.67	6.14	6.35	5.06	24.50

75th %ile	4.89	5.00	4.77	5.30	3.78	21.78

50th %ile	3.68	3.81	3.66	3.78	3.14	19.39

25th %ile	2.78	2.89	2.31	2.77	2.20	15.87

10th %ile	2.09	2.19	1.71	1.70	1.50	14.74

Min	1.01	0.89	0.98	0.23	0.41	9.61

Monthly totals during the monitoring study

2005	4.86	2.25	3.98	8.11	1.73	20.93

2006	3.64	4.86	4.9	3.88	4.59	21.87



Monthly precipitation for 2005 was at or above the median except for
May, which was at the 10th percentile. (NOTE: precipitation wasn’t
recorded for the full month of August in 2005). In 2006, monthly
precipitation totals were at or above the median in all months.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IN-08 2005-06

Watershed Location: Whitewater, Nolans Fork Watershed, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1969-2001

Max	8.38	9.09	9.38	10.45	8.95	31.10

90th %ile	5.61	7.51	6.99	7.91	6.41	25.72

75th %ile	4.67	5.41	5.07	4.99	4.79	24.09

50th %ile	3.82	4.04	4.06	3.34	3.19	21.07

25th %ile	3.00	2.90	3.00	2.24	2.26	16.65

10th %ile	1.76	2.24	2.29	1.48	1.94	14.66

Min	1.03	1.08	1.09	1.09	0.52	10.85

Monthly totals during the monitoring study

2005	4.53	2.67	7.33	2.7	2.8	20.03

2006	4.31	5.88	4.97	6.34	3.58	25.08



Monthly precipitation for 2005 was above the median in April, below the
25th percentile in May, above the 90th percentile in June, and below the
median for the rest of the sample period (NOTE: precipitation wasn’t
recorded for the full month of August in 2005). In 2006, monthly
precipitation totals were at or above the median in all months.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IN-09 2005-06

Watershed Location: Raccoon Creek Watershed, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	8.39	11.12	10.54	11.80	11.22	36.45

90th %ile	6.06	7.22	6.58	7.85	7.07	26.55

75th %ile	4.99	5.69	5.56	6.34	4.54	24.20

50th %ile	3.70	4.01	4.63	4.13	3.42	21.19

25th %ile	2.52	2.79	2.94	2.83	2.20	18.22

10th %ile	1.85	1.97	1.89	2.16	1.33	16.35

Min	0.92	0.90	0.05	0.59	0.83	12.76

Monthly totals during the monitoring study

2005	5.83	2.07	4.05	7.61	1.45	21.02

2006	4.72	4.69	7.66	4.51	7.72	29.30



Monthly precipitation for 2005 was above the median in April, below the
25th percentile in May, near the 90th percentile in July, and below the
median for the rest of the sample period (NOTE: precipitation wasn’t
recorded for the full month of August in 2005). In 2006, monthly
precipitation totals were at or above the median in all months.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IN-10 2005-06

Watershed Location: Brandywine Creek Watershed, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	8.80	10.70	10.93	11.45	11.25	32.26

90th %ile	5.73	8.60	6.95	8.57	5.32	29.52

75th %ile	4.68	5.87	5.59	5.77	4.20	24.64

50th %ile	4.02	4.16	3.96	4.19	2.98	20.77

25th %ile	2.97	2.86	2.39	3.07	2.17	17.43

10th %ile	1.93	2.43	1.54	2.11	1.55	15.30

Min	0.82	1.00	0.40	0.49	1.00	12.68

Monthly totals during the monitoring study

2005	3.71	2.49	3.09	3.49	0.98	13.76

2006	4.96	4.76	5.11	4	3.35	22.18



Except for May, which fell to near the 10th percentile, monthly
precipitation for 2005 fell between the median and the 25th percentile
(NOTE: precipitation wasn’t recorded for the full month of August in
2005). In 2006, monthly precipitation totals were at or above the median
in all months.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

IN-11 2005-06

Watershed Location: Little Pigeon Creek Watershed, IN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1960-2001

Max	10.36	11.93	7.00	14.10	9.16	39.97

90th %ile	7.46	7.22	6.61	7.21	6.66	26.88

75th %ile	5.66	5.40	5.74	5.28	4.66	23.44

50th %ile	3.62	4.01	3.90	4.24	3.18	20.18

25th %ile	2.88	3.16	2.76	2.96	2.13	18.12

10th %ile	2.24	2.71	1.40	1.53	1.76	16.20

Min	0.61	1.53	0.71	1.06	0.55	10.75

Monthly totals during the monitoring study

2005	1.65	2.89	2.82	3.15	1.62	12.13

2006	5.6	6.16	5	7.09	6.71	30.56



Monthly precipitation for 2005 fell at or below the 25th percentile
(NOTE: precipitation wasn’t recorded for the full month of August in
2005). In 2006, monthly precipitation totals were above the median in
all months – in April and May, they were greater than the 75th
percentile; in July and August they were near the 90th percentile

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

KY-01 2005-06

Watershed Location: Brashears Creek Watershed, KY

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1949-2001

Max	9.19	11.91	9.37	12.65	9.26	31.84

90th %ile	7.27	6.88	7.18	7.34	5.39	27.47

75th %ile	5.23	5.50	5.57	5.94	4.10	23.10

50th %ile	3.28	4.04	4.18	4.14	3.04	20.07

25th %ile	2.48	2.93	3.12	2.50	2.02	17.09

10th %ile	1.65	2.49	1.80	1.82	1.03	15.24

Min	0.61	0.89	0.54	0.71	0.30	13.68

Monthly totals during the monitoring study

2005	2.91	3.37	1.02	1.87	1.57	10.74

2006	5.83	3.53	5.17	6.58	2.22	23.33



Monthly precipitation totals in April and May of 2005 fell between the
median and the 25th percentile; the remaining months were below the 25th
percentile (NOTE: precipitation wasn’t recorded for the full month of
August in 2005). In 2006, monthly precipitation totals were generally
above the median.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

KY-02 2005-06

Watershed Location: Twomile Creek Watershed, KY

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1957-2001

Max	12.95	11.22	7.31	8.19	9.11	32.68

90th %ile	7.09	7.62	5.62	6.67	5.94	27.02

75th %ile	5.81	5.71	4.37	4.99	3.45	23.05

50th %ile	3.79	4.37	3.50	3.61	2.84	18.75

25th %ile	3.05	3.13	2.45	2.61	1.70	15.66

10th %ile	1.50	2.40	1.98	1.58	1.44	14.59

Min	0.87	1.48	0.64	1.03	0.61	12.72

Monthly totals during the monitoring study

2005	2.81	3.26	2.55	2.85	1.07	12.54

2006	3.89	3.18	3.59	3.89	3.5	18.05



Monthly precipitation totals for 2005 fell at or below the 25th
percentile (NOTE: precipitation wasn’t recorded for the full month of
August in 2005). In 2006, monthly precipitation totals were generally
around the median, with the exception of May, which was close to the
25th percentile.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

MN-01 2005-06

Watershed Location: Whitewater, North Fork Watershed, MN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	8.19	7.53	10.25	15.04	8.88	31.93

90th %ile	5.61	5.73	8.07	8.22	6.86	25.30

75th %ile	4.20	4.91	5.93	5.98	5.35	23.46

50th %ile	2.83	3.12	4.29	4.53	3.82	19.68

25th %ile	1.69	2.59	2.63	2.78	2.78	16.14

10th %ile	1.14	1.71	1.61	1.56	1.94	13.72

Min	0.71	0.64	0.46	0.79	0.62	9.57

Monthly totals during the monitoring study

2005	2.33	2.8	4.8	2.37	3.66	15.96

2006	3.42	2.22	2.31	2.43	7.62	18



Monthly precipitation totals for 2005 were generally between the median
and the 25th percentile (NOTE: precipitation wasn’t recorded for the
full month of August in 2005). While rainfall in April of 2006 was above
the median, monthly precipitation totals for May through July were lower
than the 25th percentile.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

MO-01, 2004-06

Watershed Location: South Fabius River Watershed, MO

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	8.04	10.51	8.43	12.70	14.67	32.81

90th %ile	5.52	7.34	6.14	7.95	5.51	26.03

75th %ile	4.43	5.85	4.92	5.98	4.16	21.77

50th %ile	3.22	4.04	3.93	3.63	2.81	18.08

25th %ile	1.90	2.61	2.46	2.41	1.90	16.06

10th %ile	1.29	1.84	1.52	1.51	1.04	14.58

Min	0.76	1.66	0.35	0.44	0.05	8.64

Monthly totals during the monitoring study

2004	3.05	2.99	5.49	3.02	2.31	16.86

2005	1.55	2.54	4.4	1.52	3.91	13.92

2006	2.48	2.51	3.98	2.14	6.19	17.3



Monthly precipitation in 2004 was between the median and 25th percentile
throughout the sampling period except for June, when it exceeded the
75th percentile. In 2005, monthly precipitation was low (<25th
percentile) in April, May and July. In 2006, monthly precipitation
totals were low (<25th percentile) in May and July.

The following figures show the daily precipitation and atrazine residues
in water for 2004, 2005 and 2006. The planting season is shaded to
provide a point of reference for the likely timing of applications in
the watershed.

MO-02, 2004-06

Watershed Location: Youngs Creek Watershed, MO

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	8.44	11.20	10.86	15.22	9.73	38.47

90th %ile	6.38	8.26	7.51	8.56	5.43	28.24

75th %ile	5.29	5.94	5.82	5.01	4.41	24.58

50th %ile	3.55	4.30	4.33	3.48	2.85	20.33

25th %ile	2.19	3.09	3.11	2.02	1.60	16.83

10th %ile	1.63	2.40	1.41	1.19	1.04	13.00

Min	0.64	0.94	0.42	0.20	0.55	9.79

Monthly totals during the monitoring study

2004	0.97	4.07	1.07	4.24	10.03	20.38

2005	2.44	2.59	3.84	0.92	5.95	15.74

2006	1.65	4.94	6.22	3.73	1.13	17.67



Monthly precipitation in 2004 was low (<10th percentile) in April and
June throughout the  and high (>90th percentile) in August. In 2005,
monthly precipitation was low (<25th percentile) in April, May and July,
and high (90th percentile) in August. In 2006, monthly precipitation
totals were low (10th percentile) in April and August and high (75th
percentile) in June.

The following figures show the daily precipitation and atrazine residues
in water for 2004, 2005 and 2006. The planting season is shaded to
provide a point of reference for the likely timing of applications in
the watershed.

MO-03, 2004-06

Watershed Location: Little Sni-a-Bar Creek Watershed, MO

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	10.22	12.10	9.79	11.90	10.45	36.14

90th %ile	5.76	8.01	7.97	7.93	5.87	28.17

75th %ile	4.37	6.35	6.45	6.28	4.70	24.75

50th %ile	3.09	3.95	4.19	4.11	3.75	20.86

25th %ile	1.94	2.86	2.63	2.18	2.04	16.80

10th %ile	1.13	2.18	1.93	1.38	0.67	13.49

Min	0.63	0.89	0.76	0.19	0.22	9.88

Monthly totals during the monitoring study

2004	3.24	5.76	5.96	7.19	5.29	27.44

2005	1.92	2.96	8.13	1.92	8.04	22.97

2006	2.64	2.03	1.82	1.79	1.02	9.3



Monthly precipitation in 2004 was between the 50th and 75th percentile
in April through June and above the 75th percentile in July and August.
In 2005, monthly precipitation was low (25th percentile) in April, May
and July, and high (90th percentile) in June and August. In 2006,
monthly precipitation totals were low (at or below the 10th percentile)
in May through August.

The following figures show the daily precipitation and atrazine residues
in water for 2004, 2005 and 2006. The planting season is shaded to
provide a point of reference for the likely timing of applications in
the watershed.

NE-01, 2004-2005

Watershed Location: Wahoo Creek, NE

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1969-2001

Max	6.54	8.84	9.94	9.85	8.68	32.36

90th %ile	5.20	7.73	7.53	5.46	7.08	23.16

75th %ile	3.21	5.20	5.43	3.94	5.60	20.37

50th %ile	2.37	4.05	3.45	3.03	2.50	16.08

25th %ile	1.54	2.74	1.95	1.54	1.62	14.04

10th %ile	1.05	1.70	1.59	0.95	0.98	12.01

Min	0.34	0.85	1.02	0.11	0.53	10.02

Monthly totals during the monitoring study

2004	1.84	3.93	2.32	3.22	1.31	12.62

2005	4.22	3.86	2.88	2.97	2.07	16



The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

NE-02 2005-06

Watershed Location: Middle Loup River Watershed, NE

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.65	9.28	12.64	13.54	7.87	24.29

90th %ile	4.49	6.37	6.98	5.73	5.33	21.62

75th %ile	3.14	5.04	5.37	4.20	3.62	19.66

50th %ile	2.39	4.09	3.70	2.68	2.18	16.56

25th %ile	1.04	2.78	2.34	1.97	1.42	14.10

10th %ile	0.80	2.03	1.39	1.08	0.96	11.69

Min	0.13	0.89	0.20	0.62	0.50	8.86

Monthly totals during the monitoring study

2005	1.26	6.91	5.97	0.94	3.13	18.21

2006	1.7	0.89	2.01	2.3	5.04	11.94



Monthly precipitation totals for 2005 bounced between below the median/
25th percentile for April and July to above the 75th percentile in May
and June. Monthly rainfall totals in 2006 were near or below the 25th
percentile from April through July.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

NE-03, 2004-2005

Watershed Location: Platte River, NE

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	6.54	8.84	9.94	9.85	8.68	32.36

90th %ile	5.20	7.73	7.53	5.46	7.08	23.16

75th %ile	3.21	5.20	5.43	3.94	5.60	20.37

50th %ile	2.37	4.05	3.45	3.03	2.50	16.08

25th %ile	1.54	2.74	1.95	1.54	1.62	14.04

10th %ile	1.05	1.70	1.59	0.95	0.98	12.01

Min	0.34	0.85	1.02	0.11	0.53	10.02

Monthly totals during the monitoring study

2004	1.84	3.93	2.32	3.22	1.31	12.62

2005	4.22	3.86	2.88	2.97	2.07	16



The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

NE-04 2005-06

Watershed Location: Big Blue River, Upper Gage Watershed, NE

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	6.72	10.64	10.06	12.04	8.83	33.90

90th %ile	4.65	7.86	8.09	7.60	6.70	26.22

75th %ile	3.25	6.30	5.08	4.82	5.19	23.35

50th %ile	2.32	3.87	3.90	3.11	3.41	18.40

25th %ile	1.67	2.87	2.22	2.17	2.30	14.49

10th %ile	1.11	1.72	1.90	0.83	1.75	11.83

Min	0.16	0.00	0.05	0.00	0.02	9.87

Monthly totals during the monitoring study

2005	1.61	1.85	3.74	5.07	4.85	17.12

2006	4.8	3.55	2.26	4.03	5.05	19.69



Monthly precipitation totals for 2005 were at or below the 25th
percentile for April and May, at the median in June and near the 75th
percentile in July and August. Monthly rainfall totals in 2006 were at
or above the median in all but June, which was near the 25th percentile.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

NE-05 2005-06

Watershed Location: Muddy Creek, Nebraska Watershed, NE

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.55	10.76	9.83	26.20	12.92	42.36

90th %ile	4.59	6.11	8.20	8.06	8.13	27.11

75th %ile	3.33	5.18	5.77	6.10	5.72	22.54

50th %ile	2.43	3.97	4.50	3.65	3.02	19.92

25th %ile	1.92	2.91	2.83	2.17	2.36	16.62

10th %ile	1.12	1.95	1.75	1.72	1.21	13.76

Min	0.65	1.28	1.03	0.20	0.27	10.47

Monthly totals during the monitoring study

2005	4.9	3.15	4.23	4.24	3.23	19.75

2006	3.32	2.44	2.23	3.44	6.8	18.23



Monthly precipitation totals for 2005 were near the median, except for
April, which exceeded the 75th percentile value. Monthly rainfall totals
in 2006 were below the 25th percentile in May and June and at or greater
than the median in April, July, and August.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

NE-06 2004-06

Watershed Location: Crooked Creek Watershed, NE

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	6.00	9.87	12.72	11.47	7.99	28.07

90th %ile	3.43	7.31	7.25	7.16	5.09	22.66

75th %ile	2.89	6.03	4.76	4.77	4.43	19.72

50th %ile	1.89	3.94	3.69	3.36	2.56	16.78

25th %ile	1.38	2.58	2.23	1.87	1.72	13.78

10th %ile	0.75	1.39	1.46	0.90	1.18	11.98

Min	0.02	0.50	0.38	0.15	0.32	9.43

Monthly totals during the monitoring study

2004	1.27	2.85	3.93	5.23	0.83	14.11

2005	1.79	1.67	3.52	4.49	4.18	15.65

2006	2.57	2.79	3.15	1.54	4.62	14.67



Monthly precipitation totals for 2004 were at or below the 25th
percentile in April and May and above the 75th percentile in July. In
2005, monthly totals fell below the 25th percentile value in May but
were at or above the 50th percentile in the other sample months. Monthly
rainfall totals in 2006 were at or below the 25th percentile in May and
July and near the 75th percentile in April and August.

The following figures show the daily precipitation and atrazine residues
in water for 2004, 2005, and 2006. The planting season is shaded to
provide a point of reference for the likely timing of applications in
the watershed.

NE-07 2005-06

Watershed Location: Big Blue River, Lower Gage Watershed, NE

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.04	10.28	11.55	17.02	11.12	35.52

90th %ile	4.54	6.80	6.59	9.30	7.38	27.49

75th %ile	3.00	6.05	5.25	5.21	5.35	22.62

50th %ile	2.33	4.36	3.72	3.49	3.82	19.03

25th %ile	1.70	2.84	2.78	1.90	2.31	16.28

10th %ile	0.99	2.05	1.79	1.24	1.30	13.45

Min	0.26	1.30	0.85	0.10	0.51	10.45

Monthly totals during the monitoring study

2005	4.41	2.7	4.98	4.13	3.58	19.8

2006	3.88	1.94	2.44	2.61	5.87	16.74



Monthly precipitation totals for 2005 were near or above the median,
except for May, which was near the 25th percentile value. Monthly
rainfall totals in 2006 were below the 25th percentile in May and June,
below the median in July, and greater than the 75th percentile in April
and August.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

OH-01 2004-05

Watershed Location: Kokosing River, OH

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.52	9.55	9.14	9.59	9.24	28.69

90th %ile	5.57	6.29	6.92	6.53	5.57	24.13

75th %ile	4.52	5.20	5.21	4.94	3.99	22.80

50th %ile	3.52	3.63	4.07	3.73	2.98	18.99

25th %ile	2.55	2.93	2.72	2.95	2.10	16.17

10th %ile	1.97	1.90	1.94	1.95	1.09	14.47

Min	0.98	1.17	0.68	1.48	0.51	12.67

Monthly totals during the monitoring study

2004	3.06	7.65	4.62	2.68	4.83	22.84

2005	3.73	1.98	2.84	5.36	1.81	15.72



The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

OH-02 2005-06

Watershed Location: Licking River, North Fork Watershed, OH

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1936-2001

Max	7.81	8.52	9.96	10.19	10.50	31.79

90th %ile	5.93	6.28	7.34	6.51	5.75	26.50

75th %ile	4.89	5.35	5.95	5.17	4.64	23.03

50th %ile	3.77	4.23	4.18	4.33	3.07	20.08

25th %ile	2.76	2.63	2.57	2.83	2.28	17.34

10th %ile	1.58	1.85	1.93	2.11	1.54	15.22

Min	0.88	0.87	0.48	1.21	0.21	10.39

Monthly totals during the monitoring study

2005	3.61	2.67	1.56	2.27	1.95	12.06

2006	2.01	2.94	4.53	6.36	2.06	17.9



5

«

û

P

Ù

4

5

6

7

S

T

U

V

X

Y

‰

Š

‹

¥

¦

§

¨

©

ª

«

¬



É

Ê

Ë

Ì

Î

Ï

Ù

Ú

Û

õ

ö

÷

ø

"ø

ù

ú

û

ü

ý

-

.

/

I

J

K

M

N

O

P

Q

R

n

o

p

q

s

t

¶

·

¸

Ò

#Ò

Ó

Ô

Ö

×

Ø

Ù

Ú

Û

÷

ø

ù

ú

ü

ý

j¸

j;

j/

j¬

j

j

"

j

h

 h

h

 h

h

 h

h

 h

 h

h

h

 h

 h

h

h

E

I

Î

Î

Ï

å

æ

ç

í

ï

ð

ñ

ò

ö

Ff

Ff\

Ff

jr

hž

 há

j

晆ٹ

晆͢

晆K

kd‹

	

j 

Ff$ 

Ff

Ffö

há

jœ 

 há

؀#

#

#

#

#

#

#

	

 

 

$

0

Å

Æ

Þ

á



1

;

@

R

T

U

k

6k

l

m

s

v

w

x

y

z

|

†

Š

Ÿ

¢

¨

ª

«

µ

¾

Ä

Å

Ê

Ï

ß

è

ਁ氃愀϶NȀ

 h

h

h

 h

h

 h

 h

 h

 h

㓿ۖĀ̊l七昀Ĵ̀

h

 h

 h

㓿ۖĀ̊l七昀Ĵ̀

(

È

(

È

(

È

(

È

(

È

 h 

h 

 h 

h 

 h 

 h 

gd 

 h 

 h 

 h 

h 

 h 

gd 

h 

 h 

 h 

 h 

 h 

 h 

h¦

%

&

‹

˜

œ

ž

Ÿ

°

³

´

À

É

-

9

P

	



 

 

 

 

$

%

&

-

9

晆ᡝ-᠀

瀁؈摧㖏ô

›

œ

㓿ۖĀ̊l七昀Ĵ̀

 h 

 h 

 h 

 h 

Ff

Ff

Ff#

Ff

j¡

j

	

kdÌ

j

j

j

kd

j

	

	

	

8

:

x

z

®

°

D

E

F

G

I

J

L

M

N

]

^

q

š

›

¯

°

±

²

E

F

H

I

K

L

N

]

^

›

°

±

²

Ã

Ò

â

ò

	

5th percentile, except for April, which was near the median. Monthly
rainfall totals in 2006 remained at or below the 25th percentile in
April and May, before rising above the median in June and near the 90th
percentile in July. 

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

OH-03 2004-05

Watershed Location: Mad River, OH

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.36	9.30	8.83	13.61	7.88	32.74

90th %ile	6.06	7.31	6.68	7.66	6.44	26.88

75th %ile	5.06	5.48	5.17	6.15	4.78	23.92

50th %ile	3.41	3.90	4.01	4.34	3.07	19.63

25th %ile	2.60	3.12	2.70	2.80	1.86	16.12

10th %ile	1.70	1.98	1.52	2.21	1.13	13.46

Min	0.58	1.35	1.07	0.41	0.05	12.07

Monthly totals during the monitoring study

2004	2.46	6.27	3.6	2.16	3.28	17.77

2005	4.86	2.05	1.64	1.38	1.31	11.24



The following figures show the daily precipitation and atrazine residues
in water for 2004 and 2005. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

OH-04 2005-06

Watershed Location: Deer Creek Watershed, OH

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	7.84	10.41	9.30	9.43	8.85	32.91

90th %ile	6.30	7.44	6.71	6.23	6.22	24.69

75th %ile	4.77	5.24	5.21	5.10	4.72	22.82

50th %ile	3.53	4.20	3.71	3.98	3.19	19.00

25th %ile	2.61	3.10	2.39	2.99	2.04	16.02

10th %ile	1.72	1.86	1.46	1.95	1.36	14.41

Min	0.43	0.72	0.25	0.94	0.44	11.10

Monthly totals during the monitoring study

2005	4.47	2.21	5.53	1.32	0.55	14.08

2006	3.18	3.12	4.59	4.2	3.78	18.87



Monthly precipitation totals for 2005 were below the 25th percentile in
May, July, and August, and above the median in April and June. Monthly
rainfall totals in 2006 were below the median in April and May and above
the median in June through August. 

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

TN-01 2005-06

Watershed Location: Obion Middle Fork Watershed, TN

NWS Weather Station:

	April Total (in)	May Total (in)	June Total (in)	July Total (in)	August
Total (in)	Apr-Aug Total (in)

Historical precipitation summaries, 1948-2001

Max	8.01	8.47	7.51	7.63	6.61	30.26

90th %ile	6.12	6.41	5.63	6.14	5.19	26.34

75th %ile	4.61	4.38	4.29	4.71	2.21	22.08

50th %ile	3.48	3.04	2.66	2.62	1.40	18.84

25th %ile	2.74	2.13	2.10	1.75	0.99	16.74

10th %ile	2.02	0.41	0.93	0.64	0.15	12.15

Min	0.43	0.72	0.25	0.94	0.44	11.10

Monthly totals during the monitoring study

2005	7.33	0.47	4.58	3.35	1.82	17.55

2006	3.14	4.03	3.75	5.07	0.44	16.43



Monthly precipitation totals for 2005 went from near the 90th percentile
in April to near the minimum level in May, with June, July, and August
between the median and the 25th percentile. Monthly rainfall totals in
2006 were near the 25th percentile in April and between the median and
25th percentile for the rest of the sampling period.

The following figures show the daily precipitation and atrazine residues
in water for 2005 and 2006. The planting season is shaded to provide a
point of reference for the likely timing of applications in the
watershed.

 PAGE   

 PAGE   1 

 PAGE   

 PAGE   221 

