Incorporating water quality into land use scenario analysis with random forest models

Abstract

Emerging research has begun to document the nuanced ways that urban form can influence water quality in urban areas. To facilitate the greater consideration of water quality by planning practitioners, this paper illustrates a two-step method to predict the water quality performance of land use scenarios through the presentation of a case study in the Huron River watershed in Michigan, USA. First, random forest models are used to relate 38 urban form variables to three water quality outcomes within the watershed: total suspended solids (TSS), total phosphorus (TP), and Escherichia coli ( E. coli) concentrations. Second, the calibrated random forest models are used to predict the water quality performance for three land use scenarios for a local jurisdiction. The case study illustrates how even scenarios describing additional urbanization can result in predicted improvements to water quality. The methods contribute to the greater consideration of water issues in urban planning practice.