Analysis and predictive models of stormwater runoff volumes, loads, and pollutant concentrations from watersheds in the Twin Cities metropolitan area, Minnesota, USA.

Urban nonpoint source pollution is a significant contributor to water quality degradation. Watershed planners need to be able to estimate nonpoint source loads to lakes and streams if they are to plan effective management strategies. To meet this need for the twin cities metropolitan area, a large database of urban and suburban runoff data was compiled. Stormwater runoff loads and concentrations of 10 common constituents (six N and P forms, TSS, VSS, COD, Pb) were characterized, and effects of season and land use were analyzed. Relationships between runoff variables and storm and watershed characteristics were examined. The best regression equation to predict runoff volume for rain events was based on rainfall amount, drainage area, and percent impervious area (R2 = 0.78). Median event-mean concentrations (EMCs) tended to be higher in snowmelt runoff than in rainfall runoff, and significant seasonal differences were found in yields (kg/ha) and EMCs for most constituents. Simple correlations between explanatory variables and stormwater loads and EMCs were weak. Rainfall amount and intensity and drainage area were the most important variables in multiple linear regression models to predict event loads, but uncertainty was high in models developed with the pooled data set. The most accurate models for EMCs generally were found when sites were grouped according to common land use and size.