Extended Abstract (672K)9.10
Forecast evaluation of a mixed-physics ensemble
Matthew S. Wandishin, Univ. of Arizona, Tucson, AZ; and S. L. Mullen, D. J. Stensrud, and H. E. Brooks
The aim of ensemble forecasting is to get a probability density function (pdf) of a forecast field of interest instead of the traditional approach of determining a best guess for that field at the valid forecast time. There are two major aspects to this problem: obtaining an appropriate pdf for the analysis and correctly modeling the evolution of that pdf through the forecast period. The former aspect of the problem has been much more extensively studied, typically through the use of the perfect model assumption in which modeling error is neglected. Obviously, this assumption is not valid and just as multiple realizations of the analysis are needed in an attempt to envelop the true initial state of the atmosphere, in the same way multiple realizations of the evolution of the analysis pdf are needed in order to envelop the final state of the atmosphere. (The terms initial and final are used here in the context of a forecast cycle.)
One method for achieving different realizations of the atmospheric evolution is for different ensemble members to use different physical parameterizations. This method is employed in a nine-member MM5 ensemble that has been run on 43 cases from the spring of 1999. The nine members consist of the possible combinations of three planetary boundary layer (PBL) schemes and three convection parameterizations (CP). With this configuration, the relative effect of the PBLs and CPs, in general, on the evolution of the model atmosphere can be examined, along with possible differences in how a particular CP interacts with the different PBLs. Preliminary results from the evaluation of precipitation forecasts from the mixed-physics ensemble will be presented, including that the PBL scheme appears to play a dominant role in whether the model produces precipitation while the CP plays a dominant role in determining whether the model produces substantial precipitation amounts.
Session 9, Mesoscale predictability and ensembles
Wednesday, 1 August 2001, 9:30 AM-1:44 PM
Previous paper