Tuesday, 13 January 2004
An investigation of IHOP convective system predictability using a matrix of 19 WRF members
A matrix of 19 WRF members created using different physical scheme combinations has been run for 8 IHOP convective cases. The cases chosen had significant rainfall observed and/or forecasted in the IHOP domain over the central United States. For each case, three different treatments of convection were used: the KF scheme, the BMJ scheme, and the use of no convective parameterization. For each of these 3 choices, 3 different microphysical schemes were used, Lin et al, NCEP-5 class, and Ferrier. Within these 9 configurations, two different planetary boundary layer schemes were used, MRF and Eta. This 18 member matrix was supplemented with one additional member (KF with MRF PBL and NCEP-5 class microphysics) using the thermal diffusion surface physics scheme instead of the OSU scheme used for the full 18 member matrix. Subjective analysis of rainfall forecasts indicates that the greatest variability in the forecasts comes from changes in the choice of convective scheme, although noticeable impacts also occur from changes in the microphysics or PBL scheme. The Eta PBL scheme seems to be more moist and slightly cooler than the MRF scheme, which impacts convective system development. The Lin et al. microphysics scheme typically results in the most rainfall, with the NCEP-5 class producing the least. The surface physics scheme has limited impacts on the forecast. An objective evaluation of the role of the changes is being performed using the Ebert McBride Contiguous Rain Area (CRA) technique. Results of evaluations using both traditional point-to-point objective skill measures and the CRA-specific parameters will be discussed. In addition, the use of the matrix for ensemble forecasting will be examined.