NATIONWIDE REGRESSION MODELS FOR PREDICTING URBAN RUNOFF WATER QUALITY AT UNMONITORED SITES¹

Abstract

ABSTRACT: Regression models are presented that can be used to estimate mean loads for chemical oxygen demand, suspended solids, dissolved solids, total nitrogen, total ammonia plus nitrogen, total phosphorous, dissolved phosphorous, total copper, total lead, and total zinc at unmonitored sites in urban areas. Explanatory variables include drainage area, imperviousness of drainage basin to infiltration, mean annual rainfall, a land‐use indicator variable, and mean minimum January temperature. Model parameters are estimated by a generalized‐least‐squares regression method that accounts for cross correlation and differences in reliability of sample estimates between sites. The regression models account for 20 to 65 percent of the total variation in observed loads.