Using satellite imagery for stormwater pollution management with Bayesian networks

Abstract

Urban stormwater runoff is the primary source of many pollutants to Santa Monica Bay, but its monitoring and modeling is inherently difficult and often requires land use information as an intermediate process. Many approaches have been developed to estimate stormwater pollutant loading from land use. This research investigates an alternative approach, which estimates stormwater pollutant loadings directly from satellite imagery. We proposed a Bayesian network approach to classify a Landsat ETM + image of the Marina del Rey area in the Santa Monica Bay watershed. Eight water quality parameters were examined, including: total suspended solids, chemical oxygen demand, nutrients, heavy metals, and oil and grease. The pollutant loads for each parameter were classified into six levels: very low, low, medium low, medium high, high, and very high. The results provided spatial estimates of each pollutant load as thematic maps from which the greatest pollutant loading areas were identified. These results may be useful in developing best management strategies for stormwater pollution at regional and global scales and in establishing total maximum daily loads in the watershed. The approach can also be used for areas without ground-survey land use data.