Identifying Least Impacted Streams and Modeling Reference Conditions for Fish IBI Development

Bioassessment depends on the measure of ecological resources and a benchmark (i.e. reference condition) against which to compare measured conditions.  Most reference sites are classified using best professional judgment or a restrictive set of criteria such as water/habitat quality, or species presence, often under representing larger streams.  With the increased use of spatial modeling techniques to generate expectations for streams, the inclusion of larger streams to the reference pool is vital to generating accurate predictions. This study focuses on the development of a method for identifying reference sites prior to sampling and encourages a sliding scale of criteria in order to select reference sites across a stream size continuum. We used Boosted Regression Tree modeling and known reference sites to generate probabilities of being a reference site given anthropogenic variables for all wadeable streams (8-400 Km²) in West Virginia.  We also used Principal Components Analysis to characterize landscape stressor gradients, identifying least impacted streams across a full range of stream sizes. We demonstrate how this process can provide a more complete set of reference site conditions that are not overly biased towards small streams improving IBI development.