NOAA’s ProbSevere LightningCast: Progress and Updates toward Transition-to-Operations

ProbSevere LightningCast is a lightning-nowcasting model in NOAA’s ProbSevere portfolio of machine-learning models. LightningCast uses GOES-R data and deep learning methods to predict the probability of lightning at any location in the next hour. LightningCast was trained using GOES-16 Geostationary Lightning Mapper (GLM) flash-extent density as truth data, which can depict the spatial extent of flashes. LightningCast has been evaluated in several testbeds in the National Weather Service, and has been selected to transition to NOAA/NESDIS operations. This presentation will highlight the status of that effort, including training activities, and provide an update on recent developments on the use of LightningCast data. Furthermore, the Cooperative Institute for Meteorological Satellite Studies (CIMSS) Community Supported Processing Package for Geostationary Satellites (CSPP-GEO) will be incorporating the LightningCast software package to be released in early 2024, enabling users within the GOES-R and Himawari satellite fields of view to routinely produce LightningCast probabilities on user-defined domains.