47ºF

New forecast model to predict severe storms looks at lightning, clouds

Could improve accuracy and timing of weather warnings

Microburst storms over Georgia detected by the experimental model predicting high probabilities for severe storms over the strongest looking clusters.
Microburst storms over Georgia detected by the experimental model predicting high probabilities for severe storms over the strongest looking clusters.


JACKSONVILLE, Fla. – Computers can't see severe storms the way human forecasters detect weather threats. 

Satellite features provide visual clues for meteorologists making severe weather predictions. 

A forecaster can spot cloud formations which are precursors to severe weather in ways computers traditionally have been unable to resolve, but these visual advantages over computer nowcasting may shrink with new experimental models designed to anticipate the probability of intense thunderstorms. 

Researchers from NOAA and the University of Wisconsin-CIMSS have developed an experimental model to predict which storms have the potential to go severe based on images from the GOES 16 lightning mapper and Advanced Baseline Imager fields.

The model is based on machine learning where it analyzes visual imagery of lightning and unique cloud features that signal dangerous weather like: overshooting tops, enhanced-V features, thermal couplets, above-anvil cirrus plumes, strong brightness temperature gradients and texture from visible satellite channels.

WATCH: Imaging shows storm forming

The model is very experimental and is not yet running in real-time, but someday it may provide earlier notice of developing or decaying intense storms especially in remote areas.


About the Author: