By Steven J. Larson, Charles G. Crawford and Robert J. Gilliom
U.S. GEOLOGICAL SURVEY
Water–Resources Investigations Report 03-4047
Sacramento, California 2004
Prepared in cooperation with the
U.S. Environmental Protection Agency
NATIONAL WATER-QUALITY ASSESSMENT PROGRAM
Complete accessible text of report (1.8 MB PDF)
Front cover for report(1.1 MB PDF)
To view PDF documents, you must have the Adobe Acrobat Reader (free from Adobe
Systems)
installed on your computer.
(download free copy of Acrobat
Reader).
Regression models were developed for predicting atrazine concentration distributions in rivers and streams, using the Watershed Regressions for Pesticides (WARP) methodology. Separate regression equations were derived for each of nine percentiles of the annual distribution of atrazine concentrations and for the annual time-weighted mean atrazine concentration. In addition, seasonal models were developed for two specific periods of the year--the high season, when the highest atrazine concentrations are expected in streams, and the low season, when concentrations are expected to be low or undetectable. Various nationally available watershed parameters were used as explanatory variables, including atrazine use intensity, soil characteristics, hydrologic parameters, climate and weather variables, land use, and agricultural management practices. Concentration data from 112 river and stream stations sampled as part of the U.S. Geological Survey's National Water-Quality Assessment and National Stream Quality Accounting Network Programs were used for computing the concentration percentiles and mean concentrations used as the response variables in regression models. Tobit regression methods, using maximum likelihood estimation, were used for developing the models because some of the concentration values used for the response variables were censored (reported as less than a detection threshold). Data from 26 stations not used for model development were used for model validation.
The annual models accounted for 62 to 77 percent of the variability in concentrations among the 112 model development stations. Atrazine use intensity (the amount of atrazine used in the watershed divided by watershed area) was the most important explanatory variable in all models, but additional watershed parameters significantly increased the amount of variability explained by the models. Predicted concentrations from all 10 models were within a factor of 10 of the observed concentrations at most model development and model validation stations. Results for the two sets of seasonal models were similar. Concentration distributions derived from the seasonal-model predictions provided additional information compared to distributions derived from the annual models.
Abstract
Introduction
Purpose and Scope
Acknowledgments
Methods
Atrazine Data Used for Model Development
Sample Collection and Analysis Methods
Sampling Station Selection
Atrazine Data Used for Model Validation
Watershed Characteristics Used as Explanatory Variables
Estimation of Atrazine Use
Other Watershed Characteristics
Statistical Analysis
Calculation of Annual Percentile and Mean Concentrations
Transformation of Response and Explanatory Variables
Selection of Explanatory Variables
Analysis of Model Fit
Estimation of Confidence and Prediction Intervals
Atrazine Models
Analysis of Significant Explanatory Variables
Relation Between Predicted Concentrations and Atrazine Use Intensity
Model Performance
Model Development Stations
Model Validation Stations
Uncertainty in Model Predictions
Prediction of Annual Concentration Distributions
Model Development Stations
Model Validation Stations
Year-to-Year Variability
Seasonal Atrazine Models
Model Limitations
Summary and Conclusions
References Cited
Document Accessibility: Adobe Systems Incorporated has information about PDFs and the visually impaired. This information provides tools to help make PDF files accessible. These tools convert Adobe PDF documents into HTML or ASCII text, which then can be read by a number of common screen-reading programs that synthesize text as audible speech. In addition, an accessible version of Acrobat Reader 6.0 for Windows (English only), which contains support for screen readers, is available. These tools and the accessible reader may be obtained free from Adobe at Adobe Access.
| AccessibilityFOIAPrivacyPolicies and Notices | |
  |
|