Statistical Comparison of RAP Analyses with MPEX Observed Soundings

During the Mesoscale Predictability Experiment (MPEX) in spring 2013, mobile sounding units representing the National Severe Storms Laboratory, Colorado State University, Purdue University, and Texas A&M University obtained 273 soundings near Great Plains convective storms and their environments. These soundings from MPEX were used to evaluate the error characteristics of Rapid Refresh (RAP) model analysis soundings, which are used frequently in severe weather forecasting operations. The errors in the analysis soundings often could not be reasonably approximated by a t_df distribution; therefore, the statistical testing focused on bootstrap resampling. Observations uncontaminated by storms were used to evaluate RAP analyses of the background environment, and those potentially modified by convection were used to assess RAP handling of the near-storm environment. The soundings were categorized by location relative to deep moist convection. This presentation discusses the statistical and practical significance of a warm bias at the surface, associated errors in derived parameters such as convective available potential energy and lifted condensation level, a high bias in 0-1-km storm-relative helicity, and other results, and the implications for the short-term prediction of severe convective weather.