Quadratic Multivariate Regression and Self-Organizing Feature Maps (SOM) for Fish Metrics Prediction in Ohio

Prediction of streams' biological integrity is evaluated in Ohio and other states by the Indices of Biotic Integrity (IBI) based on fish and macroinvertebrate Invertebrate Community Index (ICI) composition and density. The fish IBI consist of twelve different cumulative metrics. The calculation of these indices involves a labor and time-consuming fish and benthic macroinvertebrate sampling effort. An attempt to predict fish IBI individual metrics using SOM, followed by a Multivariate Polynomial Regression using the available habitat and land use variables corresponding to fish sampling sites was performed. Metric prediction after the clustering with SOM and the regression process was variable depending on the number of clusters and the fish metric. Correlation coefficients (r) were generally improved by using SOM prior to the regression model.