Mathematical modeling of first flush in highway storm runoff using genetic algorithm.

A mechanistic model was developed to predict the highway runoff pollutographs during precipitation events. Pollutants were assumed to be in two phases, attached to the pavement surface and mobile in the runoff water. Detachment and reattachment of contaminants were considered as rate-limited processes and the detachment rate was assumed to be a function of flow velocity by a power expression. The build-up of pollutants on the surface during the dry period between the storm events was also included in the model. Using measured pollutographs from three highly urbanized highway sites in Los Angeles, California, a hybrid genetic algorithm was used to estimate the model parameters for four different constituents including total suspended solids, oil and grease, dissolved copper, and particulate copper. The model was then validated by predicting pollutographs for the same site for a different monitoring season. Results revealed that estimated model parameters were different for particle-associated and dissolved constituents. These findings suggest different mechanisms governing the detachment of particle-associated and dissolved constituents from the surface. The results have also indicated that there is a linear build-up of pollutants during the dry period and the removal of pollutants during dry periods was insignificant. From these findings, it has been concluded that either the removal rate during the dry period is small or it is not proportional to the concentration of pollutants accumulated on the surface. In general, the build-up contaminant concentrations from the build-up model followed the same trend. However, in some cases the estimated and measured pollutograph did not closely match that may be due to some unknown factors affecting the build-up rates in the current model.