PREDICTIVE MODELING AS A METHOD FOR QUANTIFYING THE IMPACT OF ANTHROPOGENIC DISTURBANCE ON STREAM ECOSYSTEMS IN THE INTERIOR COLUMBIA BASIN.

T. Simmons¹, R. Henderson², and C.P. Hawkins¹. ¹Department of Fisheries and Wildlife and the Ecology Center, Utah State University, Logan, Utah 84322, ²Fish Ecology Unit, United States Forest Service, Logan, Utah 84321

Although empirical predictive modeling approaches to stream bioassessment are becoming increasingly popular, few studies to date have attempted to assess the sensitivity of these methods to various types of anthropogenic disturbance. Such a "calibration" may help to distinguish primary from secondary impacts on stream ecosystems and provide guidance for managers in setting mitigation or restoration priorities. Using data collected for the Interior Columbia Basin Ecosystem Monitoring Project, we have begun to approach these issues. To test the ability of RIVPACS-like predictive models to provide a robust method for assessing the biological integrity of streams in the interior Columbia basin, we constructed a preliminary model using data collected from 37 reference sites in the Salmon River basin of central Idaho from 1998-2000. Model error, defined as the SD of reference site O/E scores, is 0.148. We used this model to assess the biological integrity of 103 potentially impacted streams in the Salmon River basin. As expected, test sites showed a range of impairment, as measured by deviation of the O/E score away from 1.0. We then compared how well various estimates of land use type and intensity correlated with variation in O/E scores. Preliminary analyses suggest generally weak but consistent correlations.