Seasonal runoff forecasting is extremely important in arid and semi-arid regions such as Central Asia. However, the reduction of the number of meteorological stations and snow observation points after the disintegration of the Soviet Union made the application of previously used methods based on conventional observations practically impossible. The use of ERA-15, ERA-40 and ECMWF reanalysis data was tested for two catchments of the region in order to circumvent the lack of conventional observations. While the results obtained for the Syrdarya River basin upstream of Chinaz are promising, they are unsatisfactory for the Amudarya River basin upstream of Kerki. Some of the possible causes are the extent of glaciations in the basin and the insufficiency of ERA-15, ERA-40 and ECMWF reanalysis data resolution for highly complex mountain terrain. The objective of our investigation is to ascertain the cause of the considerable difference in the quality of the results obtained for the Syrdarya and Amudarya Rivers. For this purpose, we test model assimilated data for seasonal forecasting for some (sub)-catchments of both rivers and investigate the dependence of the quality of the results on the location and size of the considered (sub)-catchments as well as the extent of glaciations in the basins. Also, an attempt to use ERA-15, ERA-40 and ECMWF reanalysis data downscaled with the CHRM (Climate High-Resolution Model) is carried out for some sub-catchments.