Monthly Precipitation Prediction by Artificial Neural Networks (Case study: Mashhad synoptic station) Prévision de précipitations mensuelles à l'aide de réseaux de neurones artificiels (étude de cas de la station synoptique de Mashhad)

Several ANN models were developed to prediction of monthly precipitation data in Mashhad synoptic station. From the total 636 monthly precipitation data (from 1958 to 2008), 580 data has been used for training networks and the rest selected randomly has been used for validation of the models. To extract the precipitation dynamic of this station by ANN, a new approach of three-layer feed-forward perceptron network with back propagation algorithm was used. The sensitivity of the prediction accuracy to the content and length of input layer was investigated. Based on the most suitable parameters, two structures M531 and M741 have been selected. Statistical properties were calculated to examine the performance of the models and it was found that in the best model of monthly prediction, the correlation coefficient (R), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), 0.93, 0.99 mm, 6.02 mm, respectively.