Long-Lead Forecasting of Monthly Rainfall Using Large Scale Climate Signals and Statistical Disaggregation Models

Monsoon, one of the most dynamic climate systems, controls rainfall variation in some countries in Asia such as India, Bangladesh, Pakistan, and Iran. It delivers the main component of annual rainfall in these regions. In this study, the relationship between Iran monsoon rainfalls and some large scale climate signals such as SLP (Sea Level Pressure) and SLP differences over certain effective regions have been examined. The correlation coefficient between various combinations of climate signals with the rainfall at different time lags demonstrates some significant correlations which are used to identify the predictors for the rainfall forecasting. In this paper, a fuzzy rule model has been developed to predict the six-month rainfall in the southeastern part of Iran. Then the long-lead forecasting rainfall is disaggregated to a monthly scale using statistical disaggregation models and by considering the historical share of each month from total 6 month precipitations. In real time, the predicted value of each month during the 6 month time horizon is modified once the observed precipitation in any of the prior months becomes available using the Bayesian Theory. This way the accuracy of prediction will be significantly increased as we approach the last 3 months of the forecast period. The proposed model makes adequate lead time for estimation of the water resources potential for mid-term planning. Since the planning scale of water resources planning is usually monthly, disaggregated rainfall could increase the accuracy of operating schemes in the study area.