LightningCast

The ProbSevere LightningCast model uses images of visible, near-infrared, and long-wave infrared channels aboard GOES ABI to predict the probability of lightning in the next 60 minutes (QuickGuide). LightningCast has been used extensively by NWS forecasters since 2021.

Training

• GOES-R Hazardous Weather Testbed Blog posts
• CIMSS Satellite Blog posts

Download training slides

GRLevelX placefiles (v2.60+ required)

These placefiles are all parallax-corrected with a constant cloud-height assumption of 9 km.

• GOES-East CONUS sector
• GOES-East Meso1 sector
• GOES-East Meso2 sector
• GOES-West CONUS sector
• GOES-West Meso1 sector
• GOES-West Meso2 sector

Publications and presentations

• The Impact of Radar Reflectivity Data in a Satellite-based Lightning Nowcasting Model, WAF (2025)
• NOAA’s ProbSevere LightningCast: Progress and Updates toward Transition-to-Operations (AMS Annual Meeting 2024)
• Applying ProbSevere LightningCast (invited, NWS SOO meeting 2023)
• ProbSevere LightningCast for decision-support services (AMS Annual Meeting 2023)
• NOAA’s ProbSevere LightningCast: Artificial intelligence for nowcasting lightning from geostationary satellites (EUMETSAT 2022)
• ProbSevere LightningCast: A Deep-Learning Model for Satellite-based Lightning Nowcasting, WAF (2022)
• NESDIS Satellite Book Club (2021)
• NWS Southern Region Science Circle (2021)
• Using GOES ABI and Deep Learning to Nowcast Lightning (Monday Jan 11; Joint 2.5), AMS Annual Meeting (2021)
• Lightning Prediction in the Atlantic offshore region, NOAA AI Workshop (2020)

Notebooks

• Using machine learning to train a lightning nowcasting model

Verification

These verification metrics, for LCv1 (ABI-only predictors) were computed from the GOES-16 CONUS sector, every 10 minutes, for approximately all of 2023.