In this study, impacts of land cover changes (LCC) on the simulated surface variables over South Korea are examined using a fifth-generation mesoscale model (MM5) coupled with a land surface model. Two types of experiments are performed for July 2003 through consecutive reruns of 48-hours every 24 hours. The control experiment (CNTL) uses the original U. S. Geological Survey (USGS) land cover map, while the KLCV (Kongju's Land Cover) experiment uses the newly classified land cover map, especially over South Korea. The National Center for Environmental Prediction Final Analysis data are used to provide boundary conditions. Land cover change from savanna to mixed forest slightly reduced the albedo, but significantly increased the soil moisture availability and roughness length over South Korea. The increased roughness length in the KLCV systematically decreased wind speed irrespective of simulation time. The increased soil moisture availability in the KLCV also contributes to the reduction of diurnal variation of air temperature. As a result, the systematic overestimation of air temperature, mixing ratio, and wind speed shown in the CNTL are improved about 7.2, 7.5 and 11.8%, respectively, in the KLCV experiment. The improvements in the surface variables are very consistent with the simulation time. This study shows that realistic description of a land cover map can significantly improve the simulation skill of surface variables.