ABSTRACT

Regression models for predicting atrazine concentrations in streams were updated by incorporating refined annual atrazine-use estimates and by adding an explanatory variable representing annual precipitation characteristics. The updated Watershed Regressions for Pesticides (WARP) models enable improved predictions of specific pesticide-concentration statistics for unmonitored streams.

Separate WARP regression models were derived for selected percentiles (5th, 10th, 15th, 25th, 50th, 75th, 85th, 90th and 95th), annual mean, annual maximum, and annual maximum moving-average (21-, 60-, and 90-day durations) concentration statistics. Development of the regression models involved the same model-development data, model-validation data, and regression methods as those used in the original development of WARP. The original WARP models were based on atrazine-use estimates from either 1992 or 1997. This update of the WARP models incorporates annual atrazine-use estimates. In addition, annual precipitation data were evaluated as potential explanatory variables.

The updated WARP models include the same five explanatory variables and transformations that were used in the original WARP models, including the new annual atrazine-use data. The models also include a sixth explanatory variable, total precipitation during May and June of the year of sampling. The updated WARP models account for as much as 82 percent of the variability in the concentration statistics among the 112 sites used for model development, whereas previous WARP models accounted for no more than 77 percent. Concentration statistics predicted by the 95th percentile, annual mean, annual maximum and annual maximum moving-average concentration models were within a factor of 10 of the observed concentration statistics for most of the model development and validation sites.

Overall, performance of the models for the development and validation sites supports the application of the WARP models for predicting atrazine-concentration statistics in streams and provides a framework to interpret the predictions in terms of uncertainty. For streams where direct measurements of atrazine are lacking, the updated WARP model predictions can be used to characterize the probable values of atrazine-concentration statistics for comparison to specific water-quality benchmarks.

Posted July 2009