Regression models for explaining and predicting organochlorine pesticide concentrations in whole fish from U.S. streams

AGRO 145

Lisa H. Nowell,, Charles G. Crawford,, Naomi Nakagaki,, Gail P. Thelin,, and David M. Wolock, U.S. Geological Survey, Placer Hall, 6000 J Street, Sacramento, CA 95819
Empirical regression models were developed to predict organochlorine pesticide concentrations in whole fish from unmonitored U.S. streams on the basis of fish lipid content and watershed characteristics. Models were developed for DDT compounds, chlordane compounds, and dieldrin using whole-fish data collected from 650 streams nationwide by the U.S. Geological Survey's National Water Quality Assessment (NAWQA) Program. The most important explanatory variables were fish lipid content and various watershed characteristics, including past agricultural-pesticide use intensity, surrogate variables representing past termiticide use, population density, and forested land where past pesticide use was likely minimal. Variables representing fish taxa or geographic regions were of secondary importance. These models typically explained 50-70 percent of the variability in pesticide concentrations measured in whole fish. Only one model (p,p'-DDT) was improved substantially when the measured pesticide concentration in bed sediment from the same streams was included as an explanatory variable.