Аннотация:Disastrous floods can be caused by an unusual combination of hydrometeorological factors that have not been observed during a long observation period. The estimation of risk of such floods by statistical methods can result in considerable errors. To solve this problem the system of physically-based models of rainfall and snowmelt runoff generation, modified for different physiographic conditions, was applied. The models describe the snow cover formation and snowmelt processes, interception by the canopy, infiltration into the soil, interaction between groundwater and surface water, surface, subsurface, groundwater and channel flow. The investigations were carried out using observation data in the catchments of the river Sosna (basin area 16 300 km2 , the
tributary of the Don), the Seim (basin area 7460 km2 , the Dnieper basin), the Vologda (basin area 1270 km2 , the tributary of the North Dvina), the Velesa (basin area 870 km2 , the tributary of the West Dvina) and the Studeniy (basin area 54 km2 , the Transcarpathian region). After the calibration and testing of the models, different combinations of factors, which can cause maximum discharges and damages, were examined. To calculate the risk of maximum rainfall flood discharges the probable maximum precipitation was estimated, under account of the size of precipitation area, the duration of rainfall and the level of groundwater. Risk of the maximum snowmelt runoff discharges was estimated taking into account the calculated probable maximum of snow water equivalent, rain precipitation during the snowmelt period and snowmelt intensity. The estimates of probable maximum discharges, obtained using physically based models, were compared with those computed by statistical methods.