M. Nasseri - Department of Civil Engineering, Isfahan University of Technology, Isfahan, Iran
K. Asghari - Department of Civil Engineering, Isfahan University of Technology, Isfahan, Iran
M. J. Abedini - Department of Civil Engineering, Shiraz University, Shiraz, Iran

Simulation of rainfall field plays important roles in water resources studies.River training works and design of flood warning systems are usually confronted by the fact that historical rainfall data are insufficient and sparse in spatial domain for analysis and decision-making purposes. Both internal and external characteristics of rainfall field depend on many factors including: pressure, temperature, wind speed and its direction.Recent advanced in artificial intelligence and in particular those techniques aimed at converting input to output for highly nonlinear, non-convex and dimensionalized processes such as rainfall field, provide an alternative approach for development of rainfall forecasting model. Artificial Neural Networks (ANNs), which perform a nonlinear mapping between inputs and outputs, are such a technique.Current literatures on ANNs show that selection of network architecture and its efficient training are major obstacle for their daily usage. In this paper, both feed-forward and recurrent type networks will be developed to simulate the rainfall field and an algorithm so called Radial Basis Function (RBF) coupled with Genetic Algorithm (GA) will be used to train the networks. The technique will be implemented to forecast rainfall for a number of lead-time using rainfall hyetograph of recording rain-gauges in Fars province. Cross-validation will be used to evaluate the prediction performance of the developed technique. Implication of such approach in real-time rainfall forecasting will be highlighted.

Artificial Neural Networks, Genetic Algorithm, Rainfall Forecasting