Uncertainty Assessment in Long Term Urban Drainage Modelling

The thesis covers a general introduction and a combilation of seven papers on the subject of uncertainty assessment in urban drainage modelling. Urban drainage models applied for design and analysis of urban drainage systems are uncertain due to uncertainty in parameter assessment and especially on the rainfall inputs. In order to handle the uncertainties three different stochastic approaches are investigated applying a case catchment in the town Frejlev: (1) a reliability approach in which a parameterization of the rainfall input is conducted in order to generate synthetic rainfall events and find the probability of system failures (defined as either flooding or surcharge of manholes or combined sewer overflow); (2) an application of the Generalized Likelihood Uncertainty Estimation methodology in which an event based stochastic calibration is performed; and (3) long term Monte Carlo simulations with the purpose of estimating the uncertainties on the extreme event statistics of maximum water levels and combined sewer overflow volumes in drainage systems.

The thesis concludes that the uncertainties on both maximum water levels and combined sewer overflow volumes are considerable, especially on the large return periods, and even within the return periods specified in the design criteria. If urban drainage models are based on standard parameters and hence not calibrated, the uncertainties are even larger. The greatest uncertainties are shown to be the rainfall input and the assessment of the contributing area.