ProbSevere v3

ProbSevere v3 is a set of four machine-learning models that predict the probabilities of severe hail (≥ 1″ diameter), severe wind gusts (≥ 50 kt), tornadoes, and any of those hazards in any identified convective storm in the contiguous United States (Quick Guide). ProbSevere v3 became operational at NOAA in August 2025.

Training

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

Model predictors

Publications and presentations

• ProbSevere Version 3: Improved Exploitation of Data Fusion and Machine Learning for Nowcasting Severe Weather, WAF (2024)
• NWS Southern Region Science Circle (April 2022)
• ProbSevere v2 publications and presentations:
  • NOAA ProbSevere v2.0–ProbHail, ProbWind, and ProbTor, WAF (2020)
  • The NOAA/CIMSS ProbSevere Model: Incorporation of Total Lightning and Validation, WAF (2018)
  • An Empirical Model for Assessing the Severe Weather Potential of Developing Convection, WAF (2014)
  • Evolution of Severe and Nonsevere Convection Inferred from GOES-Derived Cloud Properties, JAMC (2013)
  • A Satellite-Based Convective Cloud Object Tracking and Multipurpose Data Fusion Tool with Application to Developing Convection, J. Tech. (2012)
  • Advances in Extracting Cloud Composition Information from Spaceborn Infrared Radiances–A Robust Alternative to Brightness Temperatures. Part I: Theory, JAMC (2010)
  • Preliminary Evaluation of a Fused Algorithm for the Prediction of Severe Storms, AMS Annual Meeting (2014)
  • NWS Great Lakes SOOs talk (2018)
  • NWS Great Lakes SOOs talk (2016)
  • NWS Southern Region talk (2016)
  • NWS Western Region talk (2016)
  • NWS Central Region talk (2015)