Characteristics of pollutants in highway runoff: Regression, representativeness, and first flush

Stormwater pollution control is the new environmental problem, even though it was recognized in the 1972 Clean Water Act Amendments. Impervious surfaces are the most problematic for stormwater management, and highways are especially important because of vehicular activities. To better understand the nature of highway runoff, three sites along major highways were monitored for 40 storm events over two seasons.; Water quality parameters were measured and chemical oxygen demand (COD), total suspended solids, oil and grease and two metals were selected for basic statistical analysis and correlation. The sampling distributions were skewed, and logarithmic transformations were useful to remove skewness. The logarithmically transformed concentrations were used as basic data forms for further analysis.; Regression was performed to relate the COD concentrations to the following predictors: the corresponding accumulative rainfall (CumR), the event's antecedent dry days (AtDry), and the previous event's precipitation (AtR). After evaluation a site-pooled COD regression model was developed regressing log COD with log CumR, log AtDry and log AtR and a constant. This model shows a strong agreement with the observations. The site effect is not important in the further analysis. This model structure can be applied to any parameter that correlates well with COD.; The flow-weighted average concentration, called the event mean concentration (EMC) is currently used for estimating stormwater mass loading, and total maximum daily loads are established using EMCs. EMCs can be estimated from a series of grab samples or by flow-weighted composite samplers. Simulations show that a composite sample is a better way to obtain the EMCs due to the large number of individual samples.; Simulations show that treating the entire first flush volume is a better strategy if mass removal efficiency is to be maximized. Computer simulations show that as many as forty samples may be necessary to construct a reliable mass-load curve to characterize the first flush. An improved, concentration-based first flush notation is proposed as an alternate way of defining first flush phenomenon. A hypothetical scenario shows how this new notation could benefit best management practice selection.