TV meteorologists may soon be adding lightning forecasts to their weather reports. A computer model now in development predicts where lightning will likely strike up to an hour in advance. That amount of forewarning could help emergency managers plan for and respond to lightning damage, said the model developers.
The idea was sparked by a meeting between lightning experts and scientists working with big data and data mining.
The predictive algorithm in its current form evaluates ongoing storm conditions in an area as small as 60 square kilometers, compares them to a large archive of past storm data, and generates the probability of cloud-to-ground lightning in that area.
The model’s development was sparked by a meeting between lightning experts and scientists working with big data and data mining, said Kristin Calhoun, lead project scientist on the model and alert system. After the meeting, “it clicked with me that we could actually use these techniques to predict lightning,” said Calhoun, a research scientist at the University of Oklahoma’s Cooperative Institute for Mesoscale Meteorological Studies in Norman and the National Oceanic and Atmospheric Administration’s (NOAA) National Severe Storms Laboratory (NSSL), also in Norman.
Calhoun and her colleagues are now devising a public lightning alert system to be powered by the predictive model. She discusses the new model and alert system today at the 2017 AGU Fall Meeting in New Orleans, La.
Closing a Hazard Prediction Gap
According to the National Weather Service, lightning strikes kill 30 people and injure 270 more, on average, each year in the United States. In 2016, U.S. lightning strikes and subsequent fires caused more than $825 million in property damage to homes, farms, and crops. However, no public early-warning system for lightning strikes exists, Calhoun said. Her team is seeking to fill that need.
The lightning prediction model considers current in-cloud and cloud-to-ground lightning conditions gathered by ground-based instruments, as well as local radar data and environmental conditions near an immediate storm. It also considers geography; lightning-generating storms in Florida, for example, behave differently than storms in Oregon, Calhoun explained. Future versions of the system may also incorporate satellite data of lightning activity.

This time-lapse video, recorded in April 2017 by the Geostationary Lightning Mapper aboard NASA and NOAA’s GOES-16 satellite, tracks a lightning storm’s path across the United States. Lightning-tracking satellites provide information about storm trajectories and lightning occurrence that might further fine-tune a forecast model. Credit: NOAA, NASA, Lockheed Martin, GOES-16, GLM
Sander Veraverbeke, an assistant professor in remote sensing and wildfire analysis at Vrije Universiteit Amsterdam in the Netherlands, welcomes the advances toward a lightning alert system. “An early-warning forecast for lightning could be of great value,” said Veraverbeke, who is also an associate project scientist at the University of California, Irvine. He added that time- and location-specific lightning forecasts “could help [wild]fire managers in their fire risk and vulnerability assessments” and “may significantly increase fire extinction success.”
Forecasts from Forests

Data science powers the model’s predictive capabilities. Well informed about current storm conditions in an area, the model sifts through data from almost 100,000 past storms using a machine learning algorithm called a random forest. This method creates hundreds of “trees,” each one a possible yes or no lightning outcome based on whether similar storms produced cloud-to-ground lightning. The algorithm evaluates this forest of decision trees collectively to arrive at a single probability of lightning striking the ground in the area under evaluation.
Forecasters who experimented with the model were able to generate predictions 15, 30, 45, and 60 minutes into the future. Tests of the predictor, which is part of NSSL’s Forecasting a Continuum of Environmental Threats (FACETs) program, occurred during the laboratory’s spring Hazardous Weather Testbeds in 2016 and 2017.
In those experiments, the researchers trained forecasters to predict lightning in current conditions and in complex test cases based on past storms. After some training, the forecasters could keep pace with the rapidly changing storm data and lightning predictions, according to Calhoun.
Streamlining the System

For now, however, the strike alert system lacks user-friendliness. At times, the model’s effectiveness depended on a forecaster’s ability to quickly acclimate to the surge of information, said Calhoun. What’s more, the output maps generated by the model are very technical. Before broadcast meteorologists and public alert systems incorporate the lightning forecaster, the team will simplify the maps and prediction language to be more user-friendly.
To that end, Calhoun explained that her team is working with social scientists and broadcast meteorologists to test different visualizations of the lightning prediction maps and find the most effective way of alerting the public to possible danger. She and her colleagues are hoping to have a streamlined version of the system ready for emergency managers in about 2 years and a public-friendly alert system ready a few years later.
—Kimberly M. S. Cartier (@AstroKimCartier), News Writing and Production Intern
Citation:
Cartier, K. M. S. (2017), New model predicts lightning strikes; alert system to follow, Eos, 98, https://doi.org/10.1029/2017EO088591. Published on 11 December 2017.
Text © 2017. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.