Assimilation of Pseudo Geostationary Lightning Mapper Observations Into a Storm Scale Numerical Model Using the Ensemble Kalman Filter

Total lightning observations that will soon be available from the GOES-R Geostationary Lightning Mapper (GLM) have the potential to be useful in the initialization of numerical weather models, particularly in areas where other types of observational data are sparse. Previous studies have shown promise in using total lightning data for this purpose, but have been limited mainly to nudging-type schemes to help initialize and maintain convection. To explore how effectively the assimilation of such data aids in modulating convection and producing more detailed storm characteristics, this study used the Ensemble Kalman Filter (EnKF) to assimilate pseudo-GLM observations into a storm-scale model with 1-km grid spacing for two real-data cases. These cases include the tornadic Moore, OK May 8, 2003 supercell, and an isolated, non-severe storm (June 6, 2000).

In both cases, pseudo-GLM observations were created from ground-based lightning mapping array data by generating flash event densities with a resolution approximately equal to the resolution of the GLM at its nadir. These psuedo-observations were then assimilated to produce analyses of each storm, and sensitivity tests were done to determine 1.) The optimal observation error values to use during the assimilation process; 2.) The effectiveness of different EnKF observation operator relationships, with a focus on relationships between graupel mass and total flash rates; and 3.) The utility of temporally averaging observations over a range of different time periods (with a maximum of 6 minutes) to smooth short time-scale variations in flash rates and to lower the computational cost of the assimilation process.

Results from these experiments will be presented and compared to observations of each storm as well as to EnKF analyses from each storm produced using the more established technique of assimilating radar radial velocity data.