You are here:
Predicting macroinvertebrate MMI for geographic targeting
Citation:
VAN SICKLE, J. AND M. H. WEBER. Predicting macroinvertebrate MMI for geographic targeting. Presented at 59th Annual Meeting of the North American Benthological Society, Providence, RI, May 22 - 26, 2011.
Impact/Purpose:
The US Environmental Protection Agency surveys the ecological conditions of streams across broad regions.
Description:
The US Environmental Protection Agency surveys the ecological conditions of streams across broad regions. We wish to identify specific streams in poor condition, as well as their regional extent. To identify such streams in Idaho, Oregon and Washington we built multiple regression models to predict macroinvertebrate, multimetric index (MMI) scores, using data from the EMAP-West survey. Driven by GIS data layers, the models predict MMI for 116,000 supposedly perennial reaches in the NHDPlus network. Due to low R-squared, our models make imprecise and unreliable predictions for individual NHDPlus reaches. However, the average of predicted MMI scores across several adjacent reaches has greater precision, assuming independence of model errors. We averaged our reach-scale predictions over the progressively larger subnetworks of adjacent reaches that are defined by 12, 10 and 8-digit hydrologic units (HUCs). We then assessed the MMI condition of each HUC as being either poor, not poor, or indeterminate, based on the average prediction and its confidence interval. The resulting maps show how model-based identification of streams in poor condition faces a tradeoff between high reliability and high geographic resolution