The NCAR High Resolution Land Data Assimilation System (HRLDAS) is further enhanced to provide the analysis and forecast of soil temperature and moisture as input to agriculture decision support systems (pest control, seeding, etc). Various atmospheric observations, analysis, and high-resolution land-use and soil texture fields are used to drive HRLDAS with 4-km grid-spacing for a large domain covering major agriculture areas in the U.S. for a retrospective spin-up period (2005-2006) analysis and for real-time forecast. The forecast of soil temperature and moisture from March to August 2006 and 2007, important seeding and growing season for agricultural applications, is evaluated against the observations from the Soil Climate Analysis Network (SCAN), which has more than 116 stations located in 39 states and provides hourly soil temperature and moisture at different soil depths. In addition, a number of numerical experiments are conducted to investigate the sensitivity of soil temperature and moisture forecast to data sources and model physics, which demonstrated that the HRLDAS forecast of soil conditions are sensitive to the vertical resolution of the land surface model, to the near –surface meteorological conditions, and to the treatment of surface roughness length for heat and moisture in the surface layer. One important goal is assess the degree to which the use of NASA MODIS vegetation data (Leaf area index and green vegetation fraction) in HRLDAS can improve the forecast of soil temperature and moisture.