Spatial Estimates of Stormwater‐Pollutant Loading Using Bayesian Networks and Geographic Information Systems

Abstract

Stormwater runoff has become the primary source of many pollutants to the Santa Monica Bay watershed (California), and managing stormwater inputs to the bay has become the primary objective of new regulatory efforts. Empirical methods to estimate stormwater pollution have been developed using land-use data; however, land-use data collected from traditional ground surveys are expensive and time consuming and may not be available. This study used an alternative approach and estimated land use from satellite-image classification using Bayesian networks. The results were converted to thematic maps using a geographic information system to visualize spatial estimates of runoff coefficients of the given area, event-mean concentrations (EMCs) of target pollutants, and their pollutant loads. The stormwater-pollutant-loading maps identified areas of high annual-mass loads, which were more affected by impervious areas, because of their high runoff coefficients, rather than their EMCs. In this watershed, the major sources of nonpoint-source pollution are the multiple-family-residential (14%), commercial (7%), public (6%), industrial (3%), and transportation (7%) land uses adjacent to Marina del Rey and the Ballona wetlands. The contributions of single-family-residential (30%) and open (33%) land uses are less important.