In cold region of China, problems exist in the study of hydrological models, including lack of basic data, low spatial and temporal resolution of available data and the difficulty of quantitative calculating evapotranspiration, snow cover and snow melting. In response, grid-based Sacramento (GSAC) model was selected, different land cover change scenarios were set, effects of land cover change on runoff were analyzed quantitatively, and runoff in dry season was predicted. Results showed that in simulating the daily runoff of Baoqing River Watershed in cold region with model which had parameters calibrated, the average determination coefficient was 0.7329 during the periods of calibration and validation, which well reproduced the daily runoff of this watershed during 1997—2001;returning farmland to grassland or forest can both increase the annual runoff, with the growth rate between 3.09% and 44.85%, and the growth rate of returning farmland to grassland had more influence on runoff;the steepness of land returned to grassland or forest had negative correlation with the runoff at the watershed export section, i.e., the smaller the slope was, the greater the flow was;the average runoff of farmland with steepness over 15° after being returned to grassland was similar with the ecological water demand in this area.