Potential predictability of seasonal mean river discharges is investigated based on an offline river ow model forced with outputs obtained from an SST-forced ensemble seasonal atmospheric forecast using a global climate model. The ensemble consists of six-member integrations with different atmospheric initial conditions with the same sea surface temperature. The variance ratio based on the analysis of the variance is used as an index of the potential predictability. The potential predictability of the river discharges in the selected three river basins of tropical Asian Monsoon regions is largely higher than that of the net water ux into the ground (P-E). This is responsible for river discharges that is obtained by integration of the surface and subsurface runoff along the river routing network of a given subbasin area since the integration of P-E over a given subbasin area has a minor in uence on the potential predictability.