Optimal Capacity of a Stormwater Reservoir for Flood Peak Reduction

Stormwater reservoirs are used to control the runoff volume and to attenuate the peak flow rate from urban catchments. The required capacity of a stormwater reservoir and its dimensions are calculated with the help of rainfall-runoff models, using a time series of observed rainfall or a design rainfall. However, many studies have revealed that the reliability of models of urbanized basins is far from perfect. This study presents an optimization of the capacity of a stormwater reservoir in a small urbanized basin, taking into account the predictive uncertainty of the rainfall-runoff model. An analysis of the model uncertainty using the generalized likelihood uncertainty estimation (GLUE) framework allowed for an assessment of the optimal capacity of the stormwater reservoir required to ensure a given safety level. Such a probabilistic approach is compared with a deterministic approach, which is commonly used in practice, where the effect of uncertainty of the rainfall-runoff model is usually neglected. The numerical experiment showed that stochastic and deterministic approaches are not equivalent as the calculated capacities of the reservoirs were different. The deterministic solution leads to an overestimation of the reservoir capacity from 20 to 150%, compared to the mean capacity obtained using a stochastic formulation. The results indicate that the uncertainty of the runoff model might potentially be a significant factor in stormwater reservoir design.