Influence of model parameter uncertainties on decision- making for sewer system management

Decisions on the rehabilitation of a sewer system are usually based on a single computation of CSO volumes using a time series of rainfall as system loads. Therefore, uncertainty in knowledge of model parameters is not taken into account. Moreover, statistical uncertainties are left aside. This paper presents the effect of uncertainties in model parameters on overall model results taking into account statistical uncertainties. It could even be argued that it does not matter whether the predictions from the model are uncertain. What matters is whether the decisions informed by these predictions are insensitive to such uncertainties. As an example the sewer system of ‘De Hoven’ (the Netherlands) is used. CSO volumes per storm event are computed using Monte Carlo simulations. In each Monte Carlo simulation of 1000 runs a different combination of fixed and variable model parameters is used. Probability distributions of computed CSO volumes are estimated taking into account the model uncertainties involved. The extent to which uncertainty in individual model parameters influences model results is quantified using relative confidence intervals of the statistical parameters of estimated distribution functions. The results show that the uncertainty in overall model results decreases the most if contributing areas are exactly known. Also the gain of model calibration is shown.