A high resolution land data assimilation system (HRLDAS) has been developed to provide land-surface initial conditions for a continuously cycling mesoscale numerical forecasting system. The HRLDAS maintains a continuous record of land- surface thermal and moisture states, which are consistent with the mesoscale model grids, by using several observational and model-derived fields to initialize and drive the Noah land surface model. The observations used to force the HRLDAS system included Stage IV precipitation estimates (hourly, 6-hourly, and daily totals), NOAA/CPC precipitation (daily total), NOHRSC snow water equivalent (SWE), and NESDIS snow mask. The HRLDAS produces somewhat different soil moisture fields when forced with various choices of precipitation and/or SWE sources. The sensitivity of the mesoscale-model forecasts to initial soil-moisture fields generated by these various choices of HRLDAS forcing is investigated. Month-long analyses and forecasts for two mesoscale model regions are used in the study. In order to include both warm- and cold-season processes, we have chosen to conduct forecasts for August 2005 over southern New Mexico and for December 2005 over central Utah. Forecasts are verified using near-surface and precipitation observations. We also evaluate the ability of the model to represent the mean surface climate over the two regions.