Abstarct: In the present research, the effect of large scale climate signals including Nino1+2, Nino3, Nino4, and Nino3, 4, MEI, TNI, AO, NAO, SOI, and PDO, SST and SLP on precipitation and discharge of Madarsoo watershed at upstream of Golestan dam was investigated. For this purpose, three raingauges and hydrometric stations with 40 and 38 years of daily rainfall and discharge data were selected. The correlation coefficients were calculated for total and maximum monthly precipitation and also average and maximum monthly discharge data.. baxsed on results of correlation analysis, different models with various combination of inputs are used for rainfall forecasting. These models are all baxsed on ANN technique. For Galikesh raingauge and hydrometric station, the maximum correlation coefficients between monthly rainfall and discharge data with Nino1+2 index calculated to be 0.4 and 0.51 with 11 and 13 months lag, respectively. Analyzing SST data, maximum correlation coefficients for precipitation data of Tangrah,Tamer and Galikesh raingauges equaled to 0.455, 0.478 and 0.44 with 6, 3 and 9 months lag, respectively and for discharge data of Tangrah,Tamer and Galikesh hydrometric stations the correlation values were 0.442, 0.576 and 0.552 with lag of 2, 7 and 15 months, respectively. Analyzing SLP data showed that the maximum correlation coefficients for total precipitation data of Tangrah,Tamer and Galikesh raingauges equaled to 0.487, 0.487 and 0.425 with 7, 9 and 9 months of lag, respectively; For discharge data of Tangrah, Tamer and Galikesh hydrometric stations the correlation values were 0.594,0.433 and 0.61, respectively all with 9 months lag. The highest correlations were obtained for maximum monthly precipitation of Tangrah, is 0.641 with a lag of 7 months at May with Eastern Mediterranean SST, for Tamer, 0.537 with a lag of 11 months at October with SLP of West Meditterian sea and for Galikesh, is 0.495 with a lag of 11 months at July with Atlas ocean SLP. For maximum monthly discharges, the highest correlation coefficients were calculated as 0.937, 0.782 and 0.926 for Tangrah, Tamer and Galikesh stations, respectively with a lag of 8 months (with SLP of Adan Gulf) at April.With regard to the results, the utilization of these signals, especially SLP and SST can be suggested as proper predictors to forecast the maximum and mean monthly precipitation and discharge over this region. In the second part, climate signal(s) with highest correlation coefficients were used as input variables of ANN model to predict the value of maximum monthly rainfall values at February, March, April, and August. Results showed the ability of MLP neural network with Levenberg-Marquardt training algorithm to predict the maximum monthly rainfall values. Comparison of observed and simulated data clearly indicated that the performance of the ANN Model 3 which uses all signals with significant correlation coefficients was very good with regards to the selected performance criteria, i.e. root mean square error, mean absolute error, correlation coefficient and Nash-Sutcliffe coefficient. For instance, the performance indices including root mean square error, correlation coefficient and Nash-Sutcliffe coefficient for maximum monthly rainfall of Tangrah rain gauge at August, was found to be 0.067, 0.95 and 0.945, respectively, for the test period. As the best results, for for Tangrah hydrometric station at August, the performance indices including RMSE, R and CNS for Tamer hydrometric station at August equaled to 0.0006, 0.999 and 0.999, respectively for the training period and equaled 0.078, 0.989 and 0.958in the test period.