Sensitivity experiments are performed using the NCEP regional spectrum model (RSM) for selected cases to examine the impact of the surface data sets such as vegetation cover, vegetation type and soil type, better representation of evaporation (E) and soil moisture initialization on precipitation over the central and the southwestern United States. The horizontal resolution is 50-km for the United States and 30-km for the Southwest. The simulations were performed for the 1988 drought case in June, 1993 floods over the central United States in July, the 1999 July wet monsoon month and a dry period in July 1995 .

A better representation of soil moisture in a regional model improves the simulations of surface temperature and precipitation (P). This can be achieved by using realistic vegetation cover and better parametrization of evaporation directly from bare soil (Edir) and accurate soil moisture initialization. Soil wetness has direct effect on surface temperature locally. Wet (dry) soil results in the increase (decrease) in E , which is balanced by the decrease (increase) in sensible heat. That results in cooler (warmer) surface temperature.

The feedback between E and P is more complicate and nonlinear. In addition to the effect of boundary layer physics and instability, the changes in low level flow associated with the changes in E play a strong role in improving P. If the changes in E are weak and not able to generate large changes in low level circulation, then the changes in P are small and the impact is limited to the area that significant differences in soil moisture occur. If the changes in E are strong enough to alter low level circulation, then the associated changes in moisture fluxes improve P. P can feedback to E. Then, the impact is not limited to the local area, where the difference in soil moisture originally locates . The accurate soil moisture initialization improves P in general, but the impact is local and is weaker than the accurate surface condition data sets and Edir.