Soil moisture affects the surface energy balance, planetary boundary layer development, and cloud formation. Simulation of soil moisture with fine spatial resolution is needed to account for the interactions of soil moisture with variations in vegetation and soil properties. In this project, the Mesoscale Meteorological Model version 5 (MM5) was used to simulate soil moisture at the Walnut River Watershed in Kansas. The 0.8 deg x 1.0 deg domain is covered mostly by grass and agricultural crops, and the soil texture is silt loam and silt clay loam. The simulation, carried out for April and May 1997, included three soil wetness levels. April 29 to May 1 was dry, May 10 to May 12 was intermediate, and May 20 to May 22 was wet because of heavy precipitation on May 19. To simulate soil moisture, initial conditions are crucial. Data processed by the National Weather Service Eta model with a resolution of 40 km were selected to provide the initial soil moisture conditions. MM5 provided estimates of precipitation, as well as longwave and shortwave radiation, to its land surface model, which is a hydrologic and soil thermodynamic model. The average modeled soil moisture was compared to measurements at eight observation stations. The soil moisture content in the top 10 cm of soil matched very well for the three time periods. Findings indicated that soil moisture is sensitive to initial conditions when soil is relatively dry but not sensitive when soil is wet. Soil moisture during rain events was simulated for May 17, 18, and 19. The runs started at (1) 0000 local time on May 19 and continued for one day, (2) 0000 hr on May 18 for 2 days, (3) 0000 on May 17 for 3 days, and (4) 1200 hr on May 18 for 1 and a half days. The comparison showed that for runs 2 and 3 the surface soil moisture content increased realistically during rain. In contrast, for run 1 modeled and observed soil moisture differed significantly mostly because the amount of modeled precipitation was less than half the observed amount. This discrepancy was associated with poor simulation of cloud amount in run 1 one-third that for runs 2 and 3 during the rain events, which might have been caused by inaccurate initial soil moisture conditions. The soil moisture spatial distribution was simulated by run 1 every 3 hr on May 19. The minimum appeared at 0300 hr, and the maximum appeared at 1800 hr because of the heavy precipitation.