Captured Runoff Prediction Model by Permeable Pavements Using Artificial Neural Networks

Abstract

Industrialization has degraded water resources over the last century due to increasing stormwater runoff. Increased impervious area utilization has substantially reduced infiltration into the ground, raised flood risk, and transferred contaminant materials into water bodies. Sustainable stormwater management is needed in order to manage runoff in urban areas and prevent pollution from water resources. Low-impact development (LID) practices, such as permeable pavement, manage a large volume of surface runoff during rain events and prevent combined sewer systems from overflowing. Predicting the captured runoff volume from watershed area by permeable pavements provides useful resources to achieve more efficient designs. Artificial neural network (ANN) models have been developed to predict the captured runoff with higher accuracy. The ANN models relate rainfall parameters and site characteristics to the stored runoff volume. A comprehensive database is obtained from the recorded data of the monitored two permeable pavements over a 2-year period. The performances of the ANN-based models are analyzed and the results demonstrate that the accuracy of the proposed models is satisfactory as compared to the measured values. Sensitivity analyses are conducted to calculate the relative importance of the studied parameters on the stored runoff. It was concluded that the ANN models are accurately predicting the stored runoff during different rain events and site characteristics. The ANN models consider the contributing parameters and provide precise volume estimation in comparison with the linear model. The results of the prediction models are useful to schedule the efficient maintenances and achieve better permeable pavement performance.