How was the GLM lightning binary grid defined?

[blaylockbk]

This question is related to my paper Comparison of Lightning Forecasts from the High-Resolution Rapid Refresh Model to Geostationary Lightning Mapper Observations

In your analysis, did you compare the HRRR LTNG directly to the GLM "Flash_extent_density" or was there additional work that needed to be done? You indicate that HRRR data became binary using a threshold of 0.04 flashes km-1 but no information about the GLM. I just want to make sure I'm doing my analysis correctly.

[blaylockbk]

There are likely better tools available today for converting GLM data to a gridded product than when I prepared that paper. I did not use the flash extent density. Instead, I did my best to map the GLM events onto the HRRR grid and used that as my binary field of lightning/no-lightning. Specifically, the nearest HRRR grid point to each GLM event's latitude/longitude was labeled as "lightning". And since the GLM detection resolution (~14km) is coarser than the HRRR grid (3 km), the HRRR grid points in the neighboring 5x5 box around each labeled event were also labeled as having an event (this gave the event on the HRRR grid roughly the same area as the GLM detection resolution). My interpretation of the GLM products is that flashes tell you how much lightning there is, but the events tell you the spatial extent of the lightning. We felt the GLM events product was more appropriate to use with the fraction skill score since we were concerned about the spatial area of convection rather than how much lightning was actually produced.