Improvements in the analysis of the convective storm environment for the 2nd generation Rapid Refresh: impact on RAP and nested HRRR forecasts

The initial version of the Rapid Refresh (RAP) mesoscale analysis and prediction system was implemented at the National Centers for Environmental Prediction in May of 2012. Numerous enhancements to the Gridpoint Statistical Interpolation (GSI) based RAP analysis have been made since the RAP code was frozen for the initial NCEP implementation. These changes have included 1) adaptive application of surface moisture increments throughout the depth of the model boundary layer, 2) modification of soil temperature and moisture fields based upon the lowest model level analysis increments, 3) limits on the magnitude of precipitable water innovations and more accurate accounting for model vs. observed terrain elevation, and 4) conservation of virtual potential temperature when saturating/evaporating regions of cloud building/removal. The combined effect of these changes has been a substantial improvement in the RAP analysis of the convective storm environment and in the skill of subsequent nested HRRR runs.

The improvement in RAP analyses and short-range forecasts of the convective storm environment is important for all operational users and includes an improvement in the sometimes too dry RAP boundary layer structure and associated CAPE analysis. These changes are especially important for the Storm Prediction Center, where forecasters closely scrutinize convective storm environment fields (T, Td, CAPE, etc.) on an hourly basis. The improvement in the nested HRRR fields is quite apparent in aggregate statistical score comparisons between the 2011 real-time HRRR forecasts initialized with the 1st generation RAP analysis / forecast system and retrospective HRRR forecasts initialized with the 2nd generation RAP system (especially a reduction in the overprediction of storm coverage). Qualitative examination of the individual forecast cases indicates improved HRRR storm location and structure from the 2nd generation RAP. Additional work is ongoing to systematically isolate the various aspects of the RAP 2nd generation enhancements and other possible improvements (via controlled retrospective runs) to enable better documentation of the individual and combined contributions to the RAP and HRRR forecast skill improvement. At the conference, I will describe, in detail, the changes and their contribution to both the RAP and nested HRRR forecast improvement from both objective skill score and qualitative case study perspectives.