Techniques for estimation of storm-runoff loads, volumes, and selected constituent concentrations in urban watersheds in the United States

Abstract

Urban planners and managers need information on the quantity of precipitation and the quality and quantity of run off in their cities and towns if they are to adequately plan for the effects of storm runoff from urban areas. As a result of this need, four sets of linear regression models were developed for estimating storm-runoff constituent loads, storm-runoff volumes, storm-runoff mean concentrations of constituents, and mean seasonal or mean annual constituent loads from physical, land-use, and climatic characteristics of urban watersheds in the United States. Thirty-four regression models of storm-runoff constituent loads and storm-runoff volumes were developed, and 31 models of storm-runoff mean concentrations were developed . Ten models of mean seasonal or mean annual constituent loads were developed by analyzing long-term storm-rainfall records using at-site linear regression models. Three statistically different regions, delineated on the basis of mean annual rainfall, were used to improve linear regression models where adequate data were available . Multiple regression analyses, including ordinary least squares and generalized least squares, were used to determine the optimum linear regression models . These models can be used to estimate storm-runoff constituent loads, storm-runoff volumes, storm-runoff mean concentrations of constituents, and mean seasonal or mean annual constituent loads at gaged and ungaged urban watersheds. The most significant explanatory variables in all linear regression models were total storm rainfall and total contributing drainage area. Impervious area, land-use, and mean annual climatic characteristics also were significant in some models. Models for estimating loads of dissolved solids, total nitrogen, and total ammonia plus organic nitrogen as nitrogen generally were the most accurate, whereas models for suspended solids were the least accurate. The most accurate models were those for application in the more arid Western States, and the least accurate models were those for areas that had large mean annual rainfall.