How NWS forecasters use generative AI for innovative storm reporting

The National Weather Service (NWS) has partnered with AWS to develop a generative AI proof of concept that assists forecasters in creating local storm reports (LSRs) by automatically extracting weather and geolocation information from diverse sources.

• LSRs capture ground-level storm impacts like wind damage, flooding, and tornadoes to verify warnings and improve forecasts
• Serverless AWS architecture uses Lambda, Textract, S3, and Bedrock to process reports from multiple channels in real-time
• Verification Agent framework with five specialized AI agents cross-checks generated reports against radar data, observational stations, and geographic sources
• Human-in-the-loop design ensures forecasters review, edit, and approve all reports before finalization
• AWS handed off project code to NWS in October 2025 for internal evaluation across diverse hazards and regions

The solution enhances forecaster workflows by reducing manual report creation time while preserving scientific rigor and maintaining expert control over critical weather operations.