The full understanding of the dynamics in flow and pollutant concentrations in combined sewer systems is an important issue in the management and design of these systems.
While high resolution data on water levels and discharge is often available and used in model calibration, water quality data in sufficient quality and resolution is often lacking. Conventionally obtained samples (e.g. via automated samplers and lab analysis) cannot account for the full dynamics of pollutant concentrations encountered in sewer systems, e.g. first flush occurrence in wet weather conditions. Moreover, the lack of field data is a critical aspect in modeling, with serious consequences for model calibration (Bertrand-Krajewski, 2007).
In Graz, Austria a sewer online monitoring station has been operated at a combines sewer overflow (CSO) at the outlet of an urban catchment area since 2002. Flow meters are installed in the inflow and the overflow channel of the CSO. A submersible UV/VIS spectrometer probe measures continuously CODeq, T(D)OCeq and TSS concentrations with an interval of 1 minute directly in the overflow chamber (Gruber, 2005).
Two models of the catchment were set up in previous studies: an aggregated hydrological model in the software SMUSI (TU Darmstadt, Germany) and recently a detailed hydrodynamic model in SWMM 5 (U.S. EPA). Both models were coupled with an optimization algorithm based on evolution strategies allowing automated model calibration (Muschalla, 2008). The SMUSI model was calibrated against discharge and pollutant concentrations (CODeq). The SWMM model, set up in late 2009 was calibrated against discharge so far.
Based on the experiences from the measurement station and the results obtained from the simulation models this contribution details on
i. the set up of the measurement station and the experiences obtained from long term operation. The challenges in maintenance, operation and probe calibration are addressed and the limits of in-situ sewer monitoring discussed.
ii. the comparison of simulation results obtained for discharge simulation with the two models. Furthermore, the quality of the SMUSI model in COD prediction is briefly discussed.