Daily Rainfall Forecasting for Mashhad Synoptic Station using Artificial Neural Networks

In this paper, we have utilized ANN (Artificial Neural Network) modeling for daily rainfall forecasting in Mashhad synoptic station. To achieve such a model, we have used daily rainfall data of March as a month with high humidity and May and December as months with medium humidity from 1986 to 2010 for this synoptic station. First, the Hurst rescaled range statistical (R/S) analysis is used to evaluate the predictability of the collected data. Then, to extract the precipitation dynamic of this station using ANN modeling, a new approach of three-layer feed-forward perceptron network with back propagation algorithm is proposed. Using this ANN model as a black box model, we have realized the hidden dynamics of rainfall through the past information of the system. The approach employs the gradient decent algorithm to train the network. Trying different parameters, some structures including GS 531 and GS 651 for March, GS 521 and GS 681 for May and GS 571 and GS 631 for December, have been selected which give the best estimation performance. Performance statistical analysis of the obtained models shows that in the best chosen model of daily forecasting, the correlation coefficient (R), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) are 0.89, 0.14(mm) and 1.15(mm) for March, 0.85, 0.14(mm) and 1.16(mm) for May and 0.86, 0.15(mm) and 1.17(mm) for December, respectively which presents the effectiveness of the proposed models.