NASA Deploys Machine Learning System to Enhance Flash Flood Warnings

NASA has developed a machine learning system called the Transient Artifact and Continuous Learning System (TACLS) to improve flash flood forecasting and warning accuracy. The system, created through a collaboration between NASA's Jet Propulsion Laboratory, the University of California San Diego (UCSD) Scripps Institution of Oceanography, and the National Oceanic and Atmospheric Administration (NOAA) National Weather Service, analyzes data from global satellite networks to detect unusual increases in atmospheric moisture that may indicate imminent flooding.

TACLS operates through two integrated components: an analytic back-end that uses machine learning algorithms to process satellite data from the Global Navigation Satellite System (GNSS), and user-friendly visualization software called MGViz that displays risk areas for human analysis. The system identifies anomalies in atmospheric moisture by examining signal delays from GNSS satellites—water vapor in the troposphere causes measurable delays in satellite signals that can be analyzed to calculate atmospheric water vapor concentrations. The machine learning model, trained on more than 30 years of historical GNSS data, distinguishes between data artifacts (false features) and transients (time-sensitive physical events requiring human interpretation).

Operating in near real-time with forecasts generated in as little as 15 minutes, TACLS is designed to assist meteorologists at the National Weather Service in making faster, more accurate flood warning decisions. During simulation testing using diverse severe weather events from 2017 to 2023—including atmospheric rivers, monsoonal convection, and tropical cyclone remnants—TACLS successfully captured 93 percent of issued flash flood warnings. The system was developed with support from NASA's Earth Science Technology Office and incorporates several existing JPL innovations, including algorithms from the Domain-agnostic Outlier Ranking Algorithms program and the Time-series Forecasting, Evaluation, and Deployment program.

Meteorologers from the National Weather Service are currently working to integrate TACLS into their existing forecasting systems for Southern California. The project aims to give forecasters a powerful decision-support tool that processes vast amounts of satellite data automatically, flagging potential flood risks that human analysts might otherwise overlook while maintaining human judgment as the final arbiter in issuing warnings.