Application of radial basis function neural networks for modeling rainfall-runoff processes: a case study of Semenyih River catchment, Malaysia

Abstract

The gradual transformation of arable lands into urbanized environments in built-up areas is common in fast developing countries like Malaysia. Such changes have a large effect on hydrologic processes in the catchment area, which eventually results in an increase of both the magnitude and frequency of floods in urban areas. Therefore there is a great need of reliable rainfall-runoff models that are able to accurately estimate the discharge for a catchment. So far various physically-based models have been developed to capture the rainfall-runoff process, but the drawback has been the estimation the several numbers of parameters which is quite difficult and time consuming. Recently, artificial intelligence tools are being used because of their capability of modeling complex nonlinear relationships. These tools have been widely used in hydrological time series modeling and prediction. Radial basis function neural network (RBFNN) is a popular artificial intelligence technique that is well used in hydrological modeling. In this study, 30 extreme rainfall-runoff events were extracted from twelve years of hourly rainfall and runoff data. An input selection method based on correlation analysis and mutual information was developed to identify the proper input combinations of rainfall and discharge antecedents. The results obtained by RBFNN model were then compared with a traditionally used statistical model known as auto-regressive moving average with exogenous inputs (ARMAX), as a bench mark. Results showed that RBFNN performance is superior then the traditional statistical model and has good potential to be used as a reliable rainfall-runoff modeling tool.