Industrial and social development of mountain regions in the south part of Siberia requires reliable estimation of the river flow. The runoff is mostly formed by rainfalls and melting. Scarce network of hydrometeorological stations and their location in valleys make interpretation of precipitation spreading in high mountains disputable. Therefore, mathematic modeling of runoff volume is of prime importance. Complex approach to mountain small catchments runoff estimation is suggested. River flow modeling is based on the "Hydrograph" model created by J. Vinorgadov. The object of research is the river Us and Amyl basins (right tributaries of Yenisei, Krasnoyarsk district, southern Siberia, Russia) which represent typical runoff conditions for the middle and low mountains of the West Sayan. The multicomponent modeling data based on the cadastral, archive, hydrometeorological information was formed. It includes both meteorological (daily values of precipitation, air temperature, vapour pressure deficit) and hydrological data (daily discharges from 2000 to 2010 at 7 river sites). Additional field research data obtained from expeditionary investigations was suggested to be taken into consideration. Watersheds landscape descriptions, hydrometeorological data and soil hydraulic parameters were obtained. Original mesoscale landscape map based on fieldwork data served the basis for defining the areas with similar runoff forming conditions. Several runoff forming complexes (RFC) inside basins under study were identified. Obtained soil and vegetation characteristics such as soil texture, hydraulic conductivity and plant cover for each RFC were used as parameters for modeling. The method of pluviometric gradients calculation based on the daily precipitation data analysis for warm and cold periods is offered. Specific height dependent precipitation distribution was detected. The convergence of modeling results and observation dates was estimated by using daily Nash-Sutcliffe criteria. To sum it up, the inclusion of the pluviometric gradients and RFC characteristics for modeling resulted in significant improvement for the convergence of calculated and obtained hydrographs.