A combinational Model for Estimating Urbanized Runoff

Abstract

The aim of this paper is to develop a combinational model of ANN-SWMM for estimation of runoff in urban catchment. The ANN-SWMM model is an alternative approach for runoff estimation in areas where field data is hard to gather. To achieve this purpose, we consider the SWMM as the physical model of the urban catchment and estimate the hydrometric parameters. Then we fed the outputs of SWMM model into an ANN model and estimate the value of runoff. In order to find the most fitted ANN model a variety of ANN architectures are constructed and examined. The result of this research shows the best fitted ANN architecture include 2 hidden layers with eight and nine neurons respectively, tansig activation function in both layers, Levenberg-Marquardt training algorithm, and a vector data consists of catchment perimeter, channel length, slope, runoff coefficient, and rainfall intensity.