Prediction of Pollutant Load in Stormwater Runoff from an Urban Residential Area

Abstract

The pollutant load in urban stormwater runoff has become a major concern with respect to the impact on the water quality of receiving streams and detention ponds. Common pollutants of concern in the runoff are suspended solids and nutrients (nitrogen and phosphorus). Past research efforts have attempted to use pollutant build-up equations to predict the increase in pollutant mass available for wash-off within a catchment. The total pollutant mass is generally expressed as a linear or exponential function of time between runoff events. Such functions are generally coupled with a wash-off equation, which predicts the varying extent to which a specific pollutant is transported out of the catchment during a runoff event. These simple models are not completely adequate, but they provide a reasonable starting point for further investigation into stormwater pollutant load generated from an urban residential catchment.In this paper, statistical models are proposed to predict the total mass of specific pollutants removed with stormwater runoff from an urban residential catchment. The statistical models are based upon analysis of data collected during an on-going research program in Saskatoon, Saskatchewan. In this program, a small, primarily-residential urban catchment was monitored and sampled for six summers (1994–1999). Stormwater outflow from the catchment and rainfall intensity at three locations in and around the catchment were recorded during rainfall events. Also, during each event, a timed series of water samples was withdrawn from the catchment outflow and analyzed for suspended solids and nutrient concentrations.Various input parameters to the statistical models were tested to assess their significance. These parameters included variables describing the precipitation event generating the runoff, the antecedent precipitation event, and the dry period between current and antecedent events. The proposed models were formulated with the most significant input parameters and calibrated using the existing data set. At present, individual water quality parameter models result in prediction errors ranging from 2.3–88% of the observed pollutant load.