Rainfall Modeling and Forecasting using Neural Networks: A Case Study of Zab Watershed

Rainfall forecasting involves a rather complex nonlinear data pattern; there are lots of novel forecasting approaches to improve the forecasting accuracy. Due to their powerful capability and functionality, ANNs provide an alternative approach for many engineering problems that are difficult to solve by conventional approaches. In this study the applicability of artificial neural networks (ANNs) to forecasting, estimation of the monthly rainfall data belonging to the Zab watershed was investigated. Throughout the study a multi-layer perceptron network was used as the ANN structure. In the forecasting part the average monthly variables of the minimum and maximum relative humidity, minimum temperature, pressure and rainfall of the watershed level stations during the 27 year statistical period (1986-2013) were collected and adjusted. Then, to predict the monthly rainfall of 2010-2012, the modeling software were applied. Therefore, the measurement indices of the network’s performance including the SPSS and MATLAB were determined, and the average square of the error squares, average error squares, average absolute error, relative error percentage, and the correlation coefficient were calculated. The results of this studys indicated that then maximum error of this model with the actual data in the Piranshahr’s and Sardasht’s stations are 7.30 and 8.13 respectively.