A family escapes rising water

Laura Lin was working from home in rural Indiana when heavy rain began flooding her yard. She and her family escaped safely, but the town received over 8 inches of rain in just a few hours.

“I look over to where the barn is over there, and I see pieces of my wood floating”

Floods are deadly and getting worse

Floods are the second-deadliest weather event in the US and deadliest globally. Just 6 inches of fast-flowing water can knock an adult down, and a warming climate means more extreme rainfall events.

Warnings came too late

Lin didn't receive flood warnings until she was already at a neighbor's house. By the time alerts went out, the entire town was already flooded.

“All the people in town were already flooded when they started sending out alerts”

Meet TACLS, a new flood prediction tool

TACLS, the Transient Artifact and Continuous Learning System, uses satellites and machine learning to spot areas at risk of flooding sooner. It helps the National Weather Service issue better flash flood alerts.

The potential to save lives

Ivory Small, a science officer at the National Weather Service, says TACLS could help save lives by enabling faster warnings. Without it, a storm could kill people. With it, warnings can go out sooner.

“It will help you save lives”

How the NWS currently issues alerts

Weather forecast offices use rain gauges, stream gauges, satellites, and weather radars to monitor flood conditions. There are 122 forecast offices across the US and its territories.

Three levels of flood alerts

A watch means flooding is possible and comes 12 to 48 hours early. An advisory means minor flooding. A warning means life-threatening danger. Flash floods are defined as flooding that occurs in under 6 hours.

Current systems have limits

Satellites provide better detail over oceans than land. Tracking rainfall as it happens doesn't give enough lead time. Deserts and rural areas have fewer rain gauges, leaving gaps in coverage.

TACLS uses a different approach

TACLS relies on the Global Navigation Satellite System, which is primarily used for earthquake prediction. The GNSS network measures water vapor in the atmosphere to help predict dangerous floods.

Real-time moisture data improves forecasts

TACLS provides real-time data about precipitable water in the atmosphere. This helps forecasters determine if storms are moving at predicted speeds and make better warning decisions.

Machine learning powers the system

A UCSD graduate student developed the machine learning side using long short-term memory architecture, which works well for weather patterns that change over time. The system was trained on years of atmospheric data.

Handling false positives

Machine learning can produce false alarms when a single sensor shows a strong signal. TACLS suppresses isolated detections and only flags patterns seen across multiple nearby stations.

TACLS assists, doesn't replace forecasters

TACLS is designed to provide another data source for human forecasters to use when deciding whether to issue alerts. It's not meant to automate the warning process.

Already in use in California

TACLS is currently being used in Los Angeles and San Diego weather forecast offices. A newer version with improved graphics is being loaded onto the National Weather Service system.

Rolling out nationwide soon

The updated TACLS with detailed maps and graphics will be available to all National Weather Service offices in the second half of October. This should help speed up alert times nationwide.

Potential for global expansion

TACLS will likely be used most in the Western US where most GNSS sensors are located. But the system could work elsewhere too, as long as there are GNSS sensors and local weather data available.

Ongoing research remains critical

Weather officials say every storm teaches them something new. Continued research after major events helps them understand what happened and why, improving future forecasts.

What residents need most

Lin hopes for earlier warnings that give people time to evacuate before they're trapped. Faster alerts could give residents the crucial minutes they need to reach safety.

“If I got an alert that was like, 'You need to get out now if you're in lower-lying ground'”