Exploring non-linear relationships between landscape and aquatic ecological condition in Southern Wisconsin: A GWR and ANN approach

Recent studies have implied the importance of incorporating configuration metrics into landscape-aquatic ecological integrity research; however few have addressed the needs of spatial data while exploring non-linear relationships. This study investigates spatial dependence of a measure of aquatic ecological condition at two basin scales, and the spatial and non-linear role of landscape in explaining that measure across 92 watersheds in Southern Wisconsin. It hypothesizes that: (1) indicators of ecological condition have different spatial needs at subwatershed and watershed scales; (2) land cover composition, urban configuration, and landscape diversity can explain aquatic ecological integrity differently; and (3) global non-linear analysis improve local spatial statistical techniques for explaining and interpreting landscape impacts on aquatic ecological integrity. Results revealed spatial autocorrelation in the measure of aquatic ecological condition at the HUC-12 subwatershed scale, and artificial neural networks (ANN) were an improvement over geographically weighted regression (GWR) for deciphering complex landscape-aquatic condition relationships.