The distributed hydrological model provides an effective means for runoff simulation and hydrological response to land use change. SWAT model is widely used to evaluate the hydrological response of land use change. However, the response of model's parameters to land use change and its effects on the simulation of runoff have not been well investigated. In this study, Shanmei Reservoir watershed located in the southeastern China was took as a research area and SWAT models for the simulation of the annual, monthly, daily runoff were constructed under the land use conditions of 1995, 2005 and 2015, respectively. The optimal values set for the sensitive parameters under three land use conditions were determined with the method of SWAT-CUP, respectively. The three optimal values sets of parameters then were applied to analyze the effects of land use change on the simulation of different time scales runoff. The results showed that:(1) the land use pattern in the study area changed greatly from 1995 to 2015. The garden and urban land were increased while the forest was reduced significantly. The SWAT models well simulated the different time scales runoff process for the three land use conditions of 1995, 2005 and 2015 with the Nash-Suttcliffe (NS) was above 0.62 and the coefficient of determination (R²) above 0.78. (2) The sensitive parameters were same for the three land use conditions which included CN2, SOL_AWC, SOL_K, CANMX, ESCO,GW_DELAY and OV_N. The values of the sensitive parameters were varied with the change of land use and the variation was coincident with the physical meaning of sensitive parameters. The values of CN2, SOL_K, CANMX, ESCO and OV_N in the study area were gradually increased from 1995 to 2015, while the values of SOL_AWC and GW_DELAY gradually decreased. The reason for the changing trend of sensitive parameters was that the land use in the study area was changed partly from forest to garden, pasture, orchard and urban land from 1995 to 2015, resulting in the decrease of the forest area. (3) The change of the value of sensitive parameters, caused by land use change, had little influence on the simulation of the annual and monthly runoff but significantly on the daily runoff. The influences on the daily runoff increased with the increase of the amplitude of land use change.