The aim of this study is to determine the predictability of river discharge in several major rivers on time scale varying from weeks to a century. We investigated predictability considering relationship between SST and Ganges and Brahmaputra River discharge. On seasonal time scales, statistically significant correlations are found between monthly equatorial Pacific SST and the summer Ganges discharge with lead times of 2-3 months due to oscillations of the ENSO phenomena. In addition, there are strong correlations in the southwest and northeast Pacific. The Brahmaputra discharge shows weaker relationships with tropical SST. Strong correlations relationships are found with SST in the Bay of Bengal but these are the result of very warm SSTs and exceptional Brahmaputra discharge during the summer of 1998. When this year is removed, relationships weaken everywhere except in the northwestern Pacific for the June and July discharge. Second goal is to project the behavior of future river discharge forced by the increasing greenhouse gases and aerosols from natural and anthropogenic sources. Three more rivers, the Yangtze, Blue Nile, and Murray-Darling rivers are considered. The original precipitation output from CMIP3 project has large inter-model variability, which limits the ability to quantify the regional precipitation or runoff trends. With a statistical Quantile-to-Quantile (Q-Q) technique, a mapping index was built to link each modeled precipitation and observational discharge. We also use the climatological annual cycle to choose the ¡°good¡± models. With the same indices, the future 21st century discharges of the first four rivers are simulated under different SRES scenarios. The Murray-Darling River basin does not have the similar seasonal cycle of discharge with modeled precipitations. We choose to project basin averaged precipitations instead. The Yangtze, Ganges, Brahmaputra River mean wet season discharges are projected to increase up to 15-25% at the end of the 21st century under A1B and A2. The risks of flooding also reach to a high level throughout the time. Inter-model deviations increase dramatically under all scenarios except for COMMIT. With large uncertainty, the Blue Nile River discharge and Murray-Darling River basin annual precipitation do not suggest a sign of change on multi-model mean.