In order to produce reliable information of climate change over North-East Asia and South Korea, sensitivity analysis on the land surface models and cumulus parameterization schemes have been performed using the regional climate model, RegCM4. Four types of sensitivity experiments (BATS_EM, BAT_GR, CLM_EM, and CLM_GR) were performed based on the combination of two physical parameterization schemes: land surface models(LSM: BATS, CLM 3.5) and cumulus parameterization schemes(CPS: Grell, MIT-Emanuel). To evaluate the simulation results objectively, APHRODITE, GPCP, NCEP-Reanalysis2 and ground observation data over South Korea were used. In general, RegCM4 well simulated seasonal and spatial variations of temperature. However, regardless of combination of parameterization schemes, RegCM4 systematically underestimated temperature during spring but overestimated it during winter. Also, RegCM4 overestimated the temperature in high-latitudes but underestimated it in low-latitudes. In general, BATS (bias: -0.58~-0.28 ℃, RMSE: 2.55~3.23 ℃) showed better simulation skills than CLM 3.5 (bias: -1.48~-1.24 ℃, RMSE: 2.68~3.48℃) relatively. RegCM4 also well simulated temporal and spatial variations of precipitation. Regardless of LSM, Grell simulated less rainfall than observation over North-East Asia in summer. On the other hand, MIT-Emanuel overestimated precipitation for the all LSMs but relatively well simulated the seasonal pattern of East-Asia summer monsoon in summer. As the results, more heavy rainfall (greater than 10 mm/day) were simulated when the MIT-Emanuel scheme was used than Grell. The simulation skills of temperature were not significantly different according to the combination of LSMs and CPSs. On the other hand, simulation skills of precipitation clearly impacted by the combination of LSMs and CPSs, in particular, choice of CPS.