A Statistical Model on a Stream Network for a Biological Sediment Tolerance Index in Impaired Rivers Draining Oregon's Mid-Coast Watersheds

In the central Oregon Coast there are 18 sites identified as impaired for biocriteria with fine sediment as a stressor. In order to identify watershed characteristics that are strongly related to these impairments we developed a spatial statistical model on a stream network (SSN) relating natural and management conditions to a biological sediment tolerance index. As a first step, we used random forest models to select the most important watershed characteristics for predicting the index out of an initial list of 112 characteristics at varying temporal and spatial scales. Using these selected characteristics we developed an SSN model to refine the list of the most strongly related characteristics. A primary benefit of the SSN model is its ability to account for autocorrelation between sampling sites on a stream network using hydrologic distance. The resulting model can be used to further our understanding of the relationship between anthropogenic and non-anthropogenic conditions to the macroinvertebrate response.