Distribution analysis of pollutant concentrations in stormwater runoff

A distributional analysis has been undertaken of the event mean concentrat ions (EMCs) of pollutants discharged during storm events from separately sewered stormwater drainage systems in four representative French urban catchments. Six basic distributions have been tested following a review of both the literature and of actual fitting procedures which demonstrated that stormwater EMCs appear to conform to a lognormal distribution. A BASIC program compatible for IBM PC use has been developed and the goodness of fit evaluated for COD, BOOs, TSS, 2n:2 .... , NO:3 - and NNH 4 .... , in respect of two sets of urban catchments located in the Paris and Aix-en-Provence regions respectively. The optimum fits are provided by maximum likelihood methods with three of the tested distributions possessing broadly similar fitting performances: the three parameter lognormal distribution; the Frechet (Extreme Value type 2) distribution; the two parameter lognormal distribution. These three distributions showed best fits for TSS and COD EMCs. A stepwise regression analysis has also been undertaken to provide a regional di fferent'iation of the pollution parameters based on lumped hydrological and storm event characteristics. Although the resultant multiple correlation coefficients are relatively weak, the main explanatory variables confirm the significance of peak flow rates and rainfall intensity in driving the stormwater flow quality within the sewer system, especially in the case of the southern Mediterranean catchments. Antecedent dry period is only of significance in the northern Parisian catchments