[Local sensitivity and its stationarity analysis for urban rainfall runoff modelling].

Sensitivity analysis of urban-runoff simulation is a crucial procedure for parameter identification and uncertainty analysis. Local sensitivity analysis using Morris screening method was carried out for urban rainfall runoff modelling based on Storm Water Management Model (SWMM). The results showed that Area, % Imperv and Dstore-Imperv are the most sensitive parameters for both total runoff volume and peak flow. Concerning total runoff volume, the sensitive indices of Area, % Imperv and Dstore-Imperv were 0.46-1.0, 0.61-1.0, -0.050(-) - 5.9, respectively; while with respect to peak runoff, they were 0.48-0.89, 0.59-0.83, 0(-) -9.6, respectively. In comparison, the most sensitive indices (Morris) for all parameters with regard to total runoff volume and peak flow appeared in the rainfall event with least rainfall; and less sensitive indices happened in the rainfall events with heavier rainfall. Furthermore, there is considerable variability in sensitive indices for each rainfall event. % Zero-Imperv's coefficient variations have the largest values among all parameters for total runoff volume and peak flow, namely 221.24% and 228.10%. On the contrary, the coefficient variations of conductivity among all parameters for both total runoff volume and peak flow are the smallest, namely 0.