RESUMEN
In order to find the optimal design of first flush diverter, this study shifts the focus of first flush research from the existence of first flush phenomenon to utilization effect of the phenomenon. The proposed method consists of four parts: (1) key design parameters, which describing key structure of first flush diverter rather than first flush phenomenon; (2) continuous simulation, which replicating the uncertainty by using the full scope of runoff events that might occur over the years analyzed; (3) design optimization, through an overlapped contour graph of key design parameters and key performance indicators that are relevant to but different from conventional indicators describing first flush phenomena; (4) event frequency spectra, which presenting the diverter's behavior at daily temporal resolution. As an illustration, the proposed method was used to determine design parameters of first flush diverters for roof runoff pollution control in the northeast of Shanghai. The results show that annual runoff pollution reduction ratio (PLR) was insensitive to buildup model. This greatly reduced the difficulty of buildup modeling. The contour graph was useful in finding the optimal design, i.e., the optimal combination of design parameters that could meet PLR design goal with most concentrated first flush on average (quantified by MFF). For instances, the diverter could achieve PLR = 40% with MFF >1.95, and PLR = 70% with MFF = 1.7 at most. Pollutant load frequency spectra were generated for the first time. They showed that a better design reduced pollutant load more stably while diverting less volume of first flush within almost each runoff day.