Flood peaks and volumes have decreased in Daqinghe River Basin in recent years. To search for the effects of underlying surface on flood decrease, Zijingguan watershed in Daqinghe River Basin was selected as a study area, and the statistical analysis of the underlying surface change was done, including land use change and soil moisture storage capacity variation. The remote sensing land use data of 1980 and 2000 were used, and the main change is forest increase, but grassland and arable land decrease. We selected several large floods to evaluate soil moisture storage capacity, and it is 110 mm before 1980, whereas 120 mm after 1980. A hydrological model was established considering underlying surface according to the hydrologic characteristics in Daqinghe Basin, in order to quantify the effects of underlying surface on floods. This model considers different types of land use, and the runoff generation was calculated for each land use type. The infiltration excess and saturation excess runoff generation mechanism was used. Finally, it was calibrated and verified by the historical hydrological data, and the Nash coefficients are about 0.70. It means that this model is consistent with the fact, and it can be used to analyze flood response to underlying surface change in the Zijingguan watershed. We simulated several floods under scenarios of 1980 and 2000, respectively. The results showed that both flood peak and volume decreased as a result of underlying surface change, and the effects are small to large floods, about 5% decrease for 10 years return period. It is shown that transfer of land use types has minor effect on floods, and the increase of soil moisture storage capacity is the main cause of decrease of flood peaks and volumes.