Predictive modeling of stormwater runoff quantity and quality for a large urban watershed

Urban and agricultural areas have been recognized by the US EPA as major national problems due to their highly polluted stormwater runoff. Remediation of this runoff has not occurred in part because assessment of nonpoint source pollution is inherently complex and expensive. One approach to deal with this complex nonpoint source pollution problem is to improve understanding through modeling and information management using Geographic Information Systems (GIS).; In this research a predictive model for stormwater runoff volume was implemented in an ArcGIS platform based on the Rational Method and Browne's empirical relation for soil characteristics. The heterogeneity of the watershed was quantified by dividing the watershed into many small sub-areas and applying lumped parameters for each of them. Characterization of pollutant load contributions of landuse types to total loads of the upper Ballona Creek watershed was achieved through zeroth-order regularization and L-BFGS-B optimization techniques. Relative form was used in the objective function to compensate for strong contributions of high magnitude variables. Model predictions showed reasonable agreement with total Zn, TKN, and TSS loadings measured at the mass emission site for the upper Ballona Creek watershed. Two additional categories, highways and local roads, which have not been routinely used as landuse categories, were separately studied.; Best Management Practices (BMP) strategies were evaluated a typical storm event, which exceeded total zinc TMDL by over 70%. The model was used to compare optimized BMP applications to the simplest application, which would treat all areas equally. Approximately 44% removal efficiency with treatment of the entire runoff would be needed to meet the TMDL.; To show how the model can be used to improve BMP application, two subwatersheds with high leverages (4.3) were identified and Austin sand filters were simulated. By assuming typical total Zn removal efficiency of 45%, total Zn removal at the mouth of the watershed was 12.4% with the treatment of 5% of the upper Ballona Creek watershed area. If the same leverage and runoff per area are assumed for additional areas in the upper Ballona Creek watershed, about 20% of the watershed area would need to be treated to meet the TMDL. This would require about {dollar}240 million if one assumes Caltrans' predicted construction cost per unit volume of runoff.; Modified infiltration trenches were simulated for local roads and the TMDL could be met with treatment of 68% of local roads area, which is equivalent to 7% of the upper Ballona Creek watershed area. The construction cost was about {dollar}58 million, which is less than one-fourth of the cost needed for the subwatershed approach.