GREMLIN: GOES Radar Estimation via Machine Learning to Inform NWP

Imagery from the Geostationary Operational Environmental Satellite (GOES) has been a key element of U.S. operational weather forecasting since 1975. The latest generation, the GOES-R Series, offers new capabilities to support the need for high-resolution rapidly refreshing imagery for situational awareness. Advances in artificial intelligence (AI) / machine learning (ML) have enabled new approaches for using GOES observations in numerical weather prediction (NWP). This presentation will provide an update on the status of GOES Radar Estimation via Machine Learning to Inform NWP (GREMLIN). GREMLIN is a machine learning approach to data fusion of Advanced Baseline Imager (ABI) radiances and Geostationary Lightning Mapper (GLM) flashes to produce synthetic radar reflectivity fields.

This presentation will focus on three areas of recent GREMLIN development. First, the use of GREMLIN to initialize convection over the ocean in the NASA-Unified Weather Research and Forecasting (NU-WRF) model will be described. The effect of microphysics scheme on the relationship between radar reflectivity and hydrometeor latent heating will be discussed. Second, the presentation will report on implementation of real-time GREMLIN processing on GOES Full Disk sectors on NOAA’s GeoCloud. This will touch on the impact of different grid geometry and lightning predictors on GREMLIN predictions. Third, the presentation will review results and feedback from demonstration of GREMLIN at the 2024 Hazardous Weather Testbed. The perceived value of GREMLIN as a function of radar coverage quality is elucidated. Differences between GOES-East and GOES-West storm depictions will be examined, in addition to cases where GREMLIN provides short-term predictive value. Finally, planned near-term GREMLIN enhancements will be briefly outlined.