The National Weather Service is deploying a new machine learning system that analyzes satellite data to predict flash floods with greater speed and accuracy. TACLS—the Transient Artifact and Continuous Learning System—feeds continuous satellite observations into algorithms designed to identify flooding conditions before water levels become critical, allowing NWS offices to issue flash flood alerts sooner than current methods permit.
Floods are the second-deadliest weather event in the United States and the deadliest globally. A warming climate is increasing the frequency of extreme rainfall events that trigger flash flooding.
The timing problem is acute. On June 9, a southern Indiana resident experienced intense rainfall in Lanesville—more than eight inches in hours—but did not receive official flood warnings until after her yard was already inundated. "We didn't get the 'get on your roof' warnings until I was already at that person's house," she said. "All the people in town were already flooded when they started sending out alerts."
Ivory Small, science and operations officer at the NWS San Diego Weather Forecast Office, said the system has direct life-or-death implications. "It will help you save lives," Small said. "Without TACLS, a storm could kill some folks, but with the system, you can put out the warning and save some folks."
The economic stakes are significant. Six inches of fast-moving water can knock an adult off their feet; 12 inches can lift a car; two feet can move trucks and SUVs. Earlier warnings reduce property damage and infrastructure losses by allowing residents and emergency responders more time to respond.
Small and Jayme Laber, senior service hydrologist for the NWS Weather Forecast Office in Oxnard, California, led the TACLS development. The system streamlines alert generation across the NWS's 122 forecast offices in the United States and its territories.
