Multifunctional retention pond for stormwater management: A decision-support model using Analytical Network Process (ANP) and Global Sensitivity Analysis (GSA)

Abstract

The ever-growing impervious and contaminated surfaces in urban areas result in severe floods, degraded waterways, and stormwater. Water-sensitive strategies and water-sensitive urban design and planning can manage stormwaters through multifunctional retention ponds; however, urban professionals seek an assessment model that evaluates and quantifies the performance of multifunctional retention ponds on stormwater management. This research has developed the Urban Retention Pond Index (URPI) assessment model, a universal multi-layered decision-support tool that constitutes three criteria (C1. Geotechnical functions, C2. Water quality and treatment, and C3. Structural and physical landscaping functions), and twenty sub-criteria. Employing Analytical Network Process (ANP) has determined the weights of indicators, which formulated the URPI index. The ANP result indicated soil investigation (WC1.1 = 0.170), soil retention (WC2.1 = 0.156), and infiltration rate (WC1.2 = 0.108) could extensively impact to the performance of multifunctional retention ponds. To validate the model, it was implemented in the Boneyard Creek retention pond using the Weighted Sum Method. The assessment analysis assigned grade A to this site, meaning, Boneyard Creek pond manages stormwater mainly through Soils Investigation (WC1.1 = 0.150), Soil retention (WC2.1 = 0.144), and Infiltration Rate (WC1.2 = 0.091). Furthermore, Global Sensitivity Analysis (GSA) was conducted to analyze the URPI model's input–output uncertainty and effect of variations, through a series of methods; Cumulative Distribution Functions (CDF), Probability Density Function (PDF), Scatterplot-Histogram Plot, Box-Whisker Plot, and Parallel Coordination. GSA could support the dominant controls and robust decision-making of the URPI model. GSA results determined that model outputs are empirically distributed with minor regression variance to the theoretical distribution. Most of the outputs fall within the intervals where the mean and median are more significant than the mode. The multiple regression analysis has shown that the three criteria are positively and linearly correlated. The Box-Whisker plots revealed the behaviors of the four mentioned measures are similar. Notably, the Box-Whisker standard error plots indicated the minor errors of the outputs in the whole network of the URPI model. Meanwhile, the Parallel Coordination indicated the largest centrality degrees by the spillway and landscape habitat retention sub-criteria and the largest Eigenvector centralities by soil retention and soil investigation sub-criteria in the whole network.