Abstract

This paper discusses the sensitivity of short- and medium-range precipitation forecasts for the central United States to land surface parametrization and soil moisture anomalies. Two forecast systems with different land surface and boundary layer schemes were running in parallel during the extreme rainfall events of July 1993. One forecast system produces much better precipitation forecasts due to a more realistic thermodynamic structure resulting from improved evaporation in an area that is about 1 day upstream from the area of heaviest rain. The paper also discusses two ensembles of 30-day integrations for July 1993. In the first ensemble, soil moisture is initialized at field capacity (100% availability); in the second ensemble it is at 25% of soil moisture availability. It is shown that the moist integrations produce a much more realistic precipitation pattern than the dry integrations. These results suggest that there may be some predictive skill in the monthly range related to the time-scale of the soil moisture reservoir. The mechanism responsible for the precipitation differences is concluded to be the result of differences in surface heating in the area 1 day upstream, impacting the atmospheric thermo-dynamic structure. Increased evaporation and reduced heating in moist soil conditions upstream result in the absence of significant boundary layer capping inversion and hence little inhibition of deep precipitating convection.