Severe Weather Forecast Skill across Different Environments in an Experimental Convection-Allowing Model

The skill of severe weather forecasts using convection-allowing models (CAMs) is a function of season and region (Sobash and Kain 2017). For example, the peak in updraft helicity (UH)-based severe weather forecast skill within the CONUS occurs during the Spring (Mar-May), and is reduced during the Summer (July-Sept) and cool-season (Nov-Mar). It is hypothesized that these variations result from: 1) differences in convective-scale predictability and 2) the inadequacy of UH-based diagnostics to identify severe, but non-supercellular, forecast convection. The latter could be improved by understanding the severe weather environments where UH-based forecasts fail and identifying more appropriate surrogate diagnostics.

Here, the relationship between CAM forecast skill and environment was analyzed using 8-years of archived National Severe Storms Laboratory WRF forecasts, a daily 36-hour CAM forecast run since 2007. 24-hour surrogate severe probability forecasts (SSPFs) were produced using the mid-level (2 – 5 km AGL) UH field and verified against observed storm reports. Forecast verification statistics were aggregated by forecast CAPE and 0 – 6 km AGL bulk wind difference (06SHR). UH-based SSPFs were most skillful in environments with large CAPE and 06SHR magnitudes, specifically CAPE > 1000 J kg^-1and 06SHR > 10 m s^-1, with increased magnitudes of CAPE and 06SHR resulting in larger skill. Two environments were identified where UH-based SSPF skill was minimized: 1) low-CAPE and high-06SHR events occurring across the eastern U.S. during the cool-season, and 2) moderate-CAPE, low-06SHR events occurring in the southeastern U.S. during the warm season. In the latter case, the UH-based SSPFs possessed no skill compared to climatology. Interestingly, SSPFs in environments of low-CAPE and low-06SHR possessed relatively large skill. These events occurred almost exclusively in the lee of the Rockies, where 06SHR likely underestimates the relevant cloud-bearing shear due to high cloud bases. Together, these results motivate the development of surrogates related to severe weather in the two low-skill regimes; the potential of other surrogates in these environments will be discussed.