Regression Model for Explaining and Predicting Concentrations of Dieldrin in Whole Fish from United States Streams

A regression model was developed for estimating dieldrin
concentrations in whole fish from United States streams using
watershed characteristics, including estimates of past agricultural
use intensity. Dieldrin concentrations were measured
in composite whole-fish samples collected at about 650 sites
nationwide during 1992–2001, as part of the U.S. Geological
Survey’s National Water-Quality Assessment Program.
Important explanatory variables in the model include (1) fish
lipid content; (2) the estimated historical agricultural use
intensity of aldrin and dieldrin (which was calculated from
regional application rates developed using U.S. Department
of Agriculture data on pesticide use by farmers in 1966
and harvested acreage from the 1964 and 1969 Census of
Agriculture); (3) a weighted termite-urban score (which
represents past use for termite control and was developed
from estimates of urban land use in the basin, combined with
the national distribution of subterranean termite density); and
(4) forested land in the basin (where past aldrin and dieldrin
use was expected to be minimal).