The WRF Lightning Forecast Algorithm: Recent Updates and Extension to Forecasts of CG Lightning

Recent work with the WRF Lightning Forecast Algorithm (LFA, McCaul et al. 2009) has included intercomparison of results from a matrix of WRF runs designed to reveal the sensitivity of the algorithm to systematic changes in choice of microphysics and planetary boundary layer (PBL) physics schemes. In these WRF runs (McCaul et al. 2014), eight distinct microphysics schemes were paired with three distinct PBL schemes, and 24-h runs at 4 km grid spacing were conducted for approximately 15 different warm season storm days at Mobile, AL (KMOB) and Houston, TX (KHGX). One of the configuration pairs involved the WRF Single Moment 6-species (WSM6) microphysics and Mellor-Yamada- Janjic (MYJ) PBL schemes, which are routinely used by the daily WRF runs made by the National Severe Storms Laboratory (NSSL) in Norman, OK, and which are found to validate well with the original LFA configuration results from the 2009 paper. Another configuration pair involves the Thompson 2-moment microphysics and Mellor-Yamada Nakanishi-Niino (MYNN) PBL scheme, as used in the hourly High-Resolution Rapid Refresh (HRRR) WRF runs performed by NOAA Global Services Division in Boulder, CO. Intercomparison of the results indicates a need to recalibrate the LFA for the HRRR WRF by boosting the prognosed flash rate densities by a factor of approximately 2.16.

Meanwhile, examination of both cloud-to-ground (CG) and total lightning flash rate densities in storms, derived from Earth Networks Total Lightning Network (ENTLN) data in the continental U. S., in conjunction with a large database of radar-derived gridded storm parameters such as Vertically Integrated Liquid (VIL), has shown compelling evidence that CG flash densities may be susceptible to prediction by the LFA. At the storm cell level, the fraction of flashes which are identified as CG exhibits a very strong negative correlation (-0.92) with binned classes of VIL. Work is currently underway to explore how best to exploit this strong statistical relationship to allow inclusion of a CG flash rate density product in the LFA, in addition to the total flash rate density which is currently prognosed.