Prediction of Monthly Average Temperature through Artificial Neural Network Multi Layer Perceptron (MLP)

Prediction of temperature as one of the most important climate parameters in different management areas and natural water resources, droughts, environmental studies, flood risk, food shortages, development of pests and diseases, transportation and etc., of special importance in determine future policy for the optimization of resources and spending costs, control and prevent crisis and has use of resources. In this study,through information monthly average temperature of Sanandaj Synoptic Stations in 38-year statistical period (2001-1964), as input Multilayer Perceptron network, the monthly average temperature was predicted during the years (2005-2002) to determine error model. For this purpose, used the features and functions available in environment programming MATLAB software, advantage was taken. Then the performanceevaluation model by statistical criteria, including regression and correlation relationships between observed and predicted values of temperature and addressed the relative mean error percent. The results show good efficiency and acceptable accuracy of artificial neural networksin predicting the temperature. So that the correlation coefficient equal to0/99 and the mean percentage error of the model with 1/97 percent., ie prediction is Network true, the temperature difference of less than one degree Celsius temperature Therefore, using this method, temperature conditions can be defined beforehand, and involved water and natural resources management.