A physically-based statistical forecasting scheme designed to predict South-East Asian rainfall monsoon variability and river discharge is presented. Specifically the technique is used to forecast rainfall over the Ganges, Brahamaputra, Yangtze and Mekong catchments as well as their discharge. Prediction of rainfall over smaller scale Indian states is also tested. The main objective is to forecast the intraseasonal variability since it is the most important time scale for agricultural applications. The scheme employs a wavelet banding technique and linear regression to forecast 5-day average (pentad) rainfall and discharge variability over regions of South-East Asia on 15-30 day time scales. Physical and statistical details of the predictors’ selection, as well as composites of the main variables in the intraseasonal band relative to “active” and “break” periods, are presented. In the statistical scheme each pentad is independently forecast which means that the coefficients of the forecasting expression changes with time. Based on the history of these coefficients, it is possible to determine the importance of each predictor in each period band. Since probabilistic forecasting is more useful for planning applications than deterministic forecasting, different techniques were combined with the statistical forecasting scheme to introduce uncertainty in the predictions. In general, forecasts for the period 1992-2003 show considerable skill out to 25 days in both the timing and amplitude of the intraseasonal oscillations. Although the skill decreases somewhat for the smaller forecast areas, most of the major features of the regional monsoon are forecast.