Global Sensitivity Analysis-based Design of Low Impact Development Practices for Urban Runoff Management Under Uncertainty

In this paper, a new methodology is developed for urban runoff management based on global sensitivity analysis of the storm water management model (SWMM) considering uncertainties. The variogram analysis of response surface (VARS) model is utilized for sensitivity analysis of the SWMM parameters by combining the runoff simulation model of the SWMM with VARS. Three model efficiency metrics, namely Nash-Sutcliffe efficiency metric for the runoff, NSE metric for the logarithm of the runoff, and percent bias in simulating runoff are used to evaluate SWMM outputs and rank its parameters. The reliability of the obtained rankings of parameters is evaluated by developing a bootstrapping-based strategy to estimate confidence intervals for the calculated sensitivity values. A multiobjective optimization model is integrated with the calibrated SWMM, to select optimum scenarios of low impact development-best management practice (LID-BMP). To take into account the rainfall uncertainty, design storm hyetograph is stochastically derived using Monte Carlo analysis and Huff curves (Huff in Water Resour Res 3(4):1007-1019, 1967; Time distributions of heavy rainstorms in Illinois, State of Illinois Department of Energy and Natural Resources, Illinois, 1990). Finally, a socially acceptable LID-BMP scenario out of a set of non-dominated solutions is obtained using the Nash bargaining theory. The proposed method is applied to an urban watershed Iran. The resulted LID-BMPs could decrease runoff volume and pollution load by 24% and about 74%, respectively.