Estimation of point and diffuse contaminant loads to streams by non-parametric regression analysis of monitoring data

In this study, point and diffuse contaminant loads to a stream are estimated using stream monitoring data. Regression analysis is conducted to separate point source loads and background contaminant loads using water quality and flow time series. Non-parametric methods are utilized for regression and auxiliary statistical tests to eliminate the effects of outliers in the data and to avoid restrictions posed by assumptions of normality. Point source loads and background concentrations are estimated directly by regression of data from a station. A second station is needed to estimate diffuse load contributions. The method is exemplified using data from a polluted stream in Turkey, the Porsuk Stream. Point and diffuse loads are estimated for six water quality parameters, chloride, dissolved solids, ammonia, nitrate, biochemical oxygen demand and total phosphorus. It is found that for many water quality parameters, contaminant loads to a stream can be estimated with reasonable accuracy.