Forecasting of Monthly rainfall using teleconnection climate indices using of artificial neural network and statical models (Case study: Sheshde and gharebolagh adjacent stations)

Many of the meteorological variables such as precipitation, strongly depend on the large scale atmospheric and ocean surface circulations.In the current study, the effect of climatic signals on the average monthly rainfall of the adjacent stations of Sheshdeh and Gharebolagh area was investigated during the statistical period twenty five years from 1985 to 2009. The regression and neural network models were used for simulation of precipitation. Then correlation of the climatic signals with precipitation were evaluated in different modes without and with delays of 3, 6, 9 and 12 months. Among twenty climate signals the most important of them include NINO1.2, NINO3 and WHWP with correlation coefficient in confidence level of 95%, (61, 45, and 33 respectively) were selected. Results showed that maximum correlation of climate indices via precipitation was associated with a delay of 6 months. The result of models simulation showed that the artificial neural network model has more accurate compare to other statistical models.This model is able to simulate the amount of precipitation according to the fluctuations with a correlation coefficient 66% and root mean square errors of 1.38. Finally, forecasting was done with coefficient of determination 44 % for five years by artificial neural network. Therefore, according to the importance of precipitation and serious water crisis in the region, identifying the parameters affecting on precipitation and the long-term forecasts precipitation is necessary to manage the water resources.