A biointegrity index (IBI) for coldwater streams of Western Oregon and Washington

We developed, tested, and applied an index of biological integrity (IBI) for fish and amphibian assemblages in coldwater streams of the Oregon and Washington Coast Range. A probability sample of 104 wadeable sites was quantitatively sampled for fish and amphibian assemblages as well as physical and chemical habitat from 1994 to 1996. Natural gradients and anthropogenic disturbances were assessed by examining digital data for catchment-scale road density and vegetation cover, along with site-scale physical and chemical habitat data. A set of 109 candidate metrics was evaluated for variance properties, redundancy, and responsiveness to multiple measures of disturbance, resulting in the selection of eight metrics for the index. The IBI itself was subsequently evaluated for variance and responsiveness to disturbance, then compared against an independently selected set of 101 reference sites that had minimal anthropogenic disturbance. Our IBI was fairly precise, with an among-stream variance-residual variance (sampling time, measurement error, and crew error) ratio of 4.0 (indicating a theoretical maximum correlation of 0.80 between the IBI and a predictor variable with a similar ratio). The IBI was significantly correlated with multiple estimates of anthropogenic disturbance, and reference sites had significantly higher IBI scores than the nonreference sites. Low IBI scores were associated with low bed stability, instream cover, and riparian cover and structural complexity but high percent fine substrate, road density, and human disturbances of riparian areas. Applying this IBI we assessed fish assemblage condition in the Coast Range, inferring our results to all mapped (1:100,000-scale) wadeable streams in this region. Using 2 SDs from the reference sites mean as a biological criterion, 45% of stream kilometers (10,646 km) were classified as impaired. High IBI scores clustered near national parks and wilderness areas. Our results indicate that minimally disturbed regional reference sites and probability survey designs, which produce statistically representative sites and inferences with known confidence bounds, offer considerable advantages over least disturbed reference sites and hand-picked survey sites for making regional ecological assessments.