Improving Spread Characteristics in a Convection Allowing Ensemble

A recurring problem in convection allowing ensemble model systems is the tendency for underdispersion for day 1 and day 2 forecasts (0-36 h). Numerous work has addressed this issue and has shown some promise in alleviating the issue of underrepresenting model error through variations in physics schemes and tendencies. However, this issue is often addressed only from the point of state variables. This work uses a 40 member WRF-DART convection allowing ensemble (dx=4 km) to sample a 10-day convectively active period during May 2015 across the south and central U.S. plains to assess how varied physics improves the representation of model error. This is done by assessing ensemble performance of state variables, rainfall probability thresholds, and convective rainfall objects. Not surprisingly, initial results suggest improved ensemble dispersion as physics parameterizations are varied in both the parent and child domains. Full results, including spatio-temporal disaggregation will be presented as well as initial results of physics varied only by parameterization type (e.g. PBL, microphysics) to assess which contributes most to underdispersion in convection allowing ensembles.