JWSS - Journal of Water and Soil Science

---

In Iran, about 75% of national rice production is supplied in Gilan and Mazandaran proviences which have the highest amount of precipitation. Seasonal prediction of rainfall induces significant improvement on yield production and on preventing climate hazardz over these feritle areas. Canonical correlation analysis (CCA) model was carried out evaluates the possibility of the prediction of winter rainfall according to the states of ENSO events. The time series of (southern oscilation index (SOI) and SST (sea surface temperature) over Nino's area (Nino's SST) are used as the predictors, and precipitation in Bandar Anzali and Noushahr are used as the predictands. Emperical orthogonal functions (EOF) were applied for reducing the number of original predictors variables to fewer presumably essential orthogonal variables. Four modes of variations (EOF1, EOF2, EOF3, EOF4) which account for about 92% of total variance in predictors field were retained and the others were considered as noise. Based on the retained EOFs and precipitation time series, the canonical correlation analysis (CCA) was carried out to predict winter precipitation in Noushahr and Bandar Anzali. The results indicated that the predictors considered account for about 45% of total variance in the rainfall time series. The correlation coefficents between the simulated and observed time series were significant at 5% significant level. For 70% of events the anomalies of observed and simulated values have the same sign indicating the ability of the model for reasonable prediction of above or below normal values of precipitation. For rainfall prediction, the role of Nino's SST (Nino4 in particular) was found to be around 10% more influential than SOI. .

---

Since the fluctuations of the Persian Gulf Sea Surface Temperature (PGSST) have a significant effect on the winter precipitation and water resources and agricultural productions of the south western parts of Iran, the possibility of the Winter SST prediction was evaluated by multiple regression model. The time series of PGSSTs for all seasons, during 1947-1992, were considered as predictors, and the time series of MSSTs during 1948-1993, as the prrdictand. For the purpose of data reduction and principal components extraction, the principal components analysis was applied. Just the scores of the first four PCs (PC1 to PC4) that accounted for the total variance in predictor field were considered as the input file for the regression analysis. For finding the dependency of each principal component to the first time series of the PGSST, the Varimax rotation analysis was applied. The results have indicated that PC1 to PC4 respectively are the indicator of temperature changes during winter, autumn, Spring and Summer. According to the regression model, the components of PC1, PC2 and PC4 were significant at 5% level. But the components of PC3 was insignificant. The results indicated that the significant variables are held accountable for the 33.5% of the total variance in the winter PGSSTs. It became obvious that for the prediction of the winter PGSST, the PGSST during the winter of the last year has a particular importance. At the next stage, autumn and summer temperature have also a role in prediction of winter PGSST.