Simulating flash flood events: Importance of the subgrid representation of convection

The Pennsylvania State University-NCAR Mesoscale Model is used to simulate six flash flood events that occurred in the central and eastern United States. Results suggest that standard model simulations of 24-h accumulated precipitation provide little indication of the potential for flash flood-producing rainfall for most events. Often the model generates convective outflow boundaries that rapidly move away from their source region and initiate new convection well removed from the original convective activity. This self-sustaining propagation mechanism prevents long-lived heavy rainfall over a particular region. Three modifications to the Kain-Fritsch convective parameterization scheme, each designed to test model sensitivity to the convective scheme formulation, are discussed. The modifications include maximizing the convective scheme precipitation efficiency and altering the scheme's treatment of convective downdrafts. Model results using the three modifications of the convective scheme each show improvements in 24-h precipitation totals compared to simulations generated using the unmodified convective scheme. Precipitation totals appear to be especially sensitive to the way downdrafts are treated within the convective scheme. In addition, the impact of including the mesoscale details of a previously generated outflow boundary into the model initial conditions is demonstrated for one of the flash flood cases. It is shown that including the surface mesoscale details can have a substantial impact upon the magnitude and location of model precipitation maxima.