The use of long-term on-line turbidity measurements for the calculation of urban stormwater pollutant concentrations, loads, pollutographs and intra-event fluxes.

This paper presents one of the largest databases on the quality of urban wet weather discharges measured since the development of continuous in-sewer water quality sensors in the late 1990s. Five years of continuous turbidity measurements enabled the validation of 263 and 239 rainfall events, respectively on two experimental catchments in Lyon (France), Chassieu (185 ha separate sewer) and Ecully (245 ha combined sewer). Except for high rainfall events of summer and second half of winter, analysis of database representativeness showed that all seasons were relatively well represented. As a first analysis of the database, traditional tools used in the urban drainage field were applied to assess: i) statistics and analysis of distributions of TSS and COD events loads and event mean concentrations (EMCs) and ii) the correlations between these statistics and events characteristics and iii) M(V) curves describing the intra-event mass distribution. Results showed that: i) EMCs and loads were approximately log-normally distributed, with a clear impact from wastewater contribution in Ecully, ii) EMCs are not correlated with storm event characteristics, whereas loads have shown significant correlation with key storm event variables such as total event volume, rainfall depth, maximum rainfall intensity and discharge and iii) M(V) curves dynamic could be classified in three categories, however with no clear correlation with storm event characteristics. The visual analysis of continuous time series of TSS and COD pollutographs, derived from turbidity time series showed that event pollutographs were highly variable, due to complex interacting processes during and between events, and suggests that further progress in knowledge and modelling of urban wet weather pollutant loads and pollutographs should be based on more detailed analyses of continuous time series rather, than on the traditional single event approach.