NASA's TACLS System Uses Machine Learning to Improve Flash Flood Forecasting

NASA's Jet Propulsion Laboratory, in collaboration with the University of California, San Diego (UCSD) and the National Oceanic and Atmospheric Administration (NOAA) National Weather Service, has developed the Transient Artifact and Continuous Learning System (TACLS) to enhance flash flood warnings through machine learning and satellite data analysis. The system leverages data from Global Navigation Satellite System (GNSS) satellites, which detect water vapor in the atmosphere by measuring signal delays as satellite transmissions reach Earth. TACLS processes this atmospheric moisture data using machine learning algorithms trained on more than 30 years of historical GNSS observations to automatically identify unusual increases in water vapor that may indicate imminent flooding.

TACLS operates through two integrated components: an analytical back-end that uses machine learning to process satellite data and identify areas at risk for flooding, and user-friendly visualization software (MGViz) that displays findings for human meteorologists to evaluate. The system performs in near real-time, generating forecasts in as little as fifteen minutes. A key innovation of TACLS is its anomaly detection capability, which distinguishes between artifacts—false features or distortions in data—and transients, time-sensitive physical events such as heavy precipitation that warrant meteorological analysis and potential warnings.

In simulation testing using data from severe weather events between 2017 and 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 (ESTO) and incorporates several JPL innovations, including algorithms from the Domain-agnostic Outlier Ranking Algorithms program and the Time-series Forecasting, Evaluation, and Deployment program. The visualization component is based on the Multi-Mission Geographic Information System, originally created at JPL for NASA's Mars exploration missions.

Meteorologers at the National Weather Service are currently working to integrate TACLS into their existing flash flood forecasting systems for Southern California. According to Yehuda Bock, Distinguished Researcher at UCSD's Scripps Institution of Oceanography and principal investigator for TACLS, the goal is to provide meteorologists with a decision-support tool that helps them issue flash flood warnings more effectively. By automating the detection of unusual atmospheric moisture patterns in large satellite datasets, TACLS enables human analysts to focus their expertise on interpreting flagged events and making informed decisions about issuing warnings or advisories.