NASA System Uses Machine Learning to Improve Flash Flood Warnings
NASA, in collaboration with the University of California, San Diego (UCSD) and the National Oceanic and Atmospheric Administration (NOAA) National Weather Service (NWS), has developed the Transient Artifact and Continuous Learning System (TACLS). This new software utilizes machine learning and satellite data to enhance the efficiency of flash flood forecasting for meteorologists. TACLS is designed to identify atmospheric moisture increases indicative of impending floods, providing near real-time forecasts in as little as 15 minutes.
A new system called the Transient Artifact and Continuous Learning System (TACLS) is designed to help meteorologists at the National Weather Service (NWS) forecast flash floods more efficiently. Developed through a collaboration involving NASA’s Jet Propulsion Laboratory, the University of California, San Diego (UCSD), and the National Oceanic and Atmospheric Administration (NOAA) NWS, TACLS leverages data from continuously operating satellite networks and machine learning models.
The system automatically identifies unusual increases in atmospheric moisture, which are indicators of impending flash floods. TACLS flags this evidence, suggests areas where flash flooding is probable, and presents this information through a user-friendly visualization interface for human analysts. These analysts then use the information to determine whether to issue a flash flood warning or a weather advisory. The framework operates in near real-time, capable of producing forecasts in approximately 15 minutes.
TACLS is comprised of two main components. An analytic back-end software suite uses machine learning algorithms to process satellite data and identify areas at risk of flooding. The second component is user-friendly visualization software that highlights these areas for further human review. The back-end analyzes data from satellites in the Global Navigation Satellite System (GNSS). It calculates the amount of water vapor in the atmosphere by measuring delays in satellite signals caused by tropospheric water vapor.
The machine learning model within the TACLS analytic back-end suite was trained using over 30 years of past GNSS data. It functions as an anomaly detector, tracking unusual increases in atmospheric moisture and distinguishing between data artifacts and transients (time-sensitive physical events like heavy precipitation). If a transient event is detected, the data is forwarded to the visualization software.
Simulations conducted between 2017 and 2023, which included various severe weather events such as atmospheric rivers, monsoonal convection, and tropical cyclone remnants, showed that TACLS successfully captured 93% of issued flash-flood warnings. Meteorologists from the National Weather Service are currently working to integrate TACLS into their existing forecasting systems for Southern California, aiming to provide communities with more advance notice to prepare for severe weather events.
Yehuda Bock, a Distinguished Researcher at the UCSD Scripps Institution of Oceanography and principal investigator for TACLS, stated that the system's goal is to provide meteorologists with a tool to assist in decision-making for flash flood warnings.
(Source: NASA Breaking News)
