Flash floods can strike without warning — this new technology could change that | The VergeSkip to main contentThe homepageThe VergeThe Verge logo.The VergeThe Verge logo.TechReviewsScienceEntertainmentAIPolicyNotificationsNotificationsHamburger Navigation ButtonThe homepageThe VergeThe Verge logo.NotificationsNotificationsHamburger Navigation ButtonNavigation DrawerThe VergeThe Verge logo.Login / Sign UpcloseCloseSearchLightSystemDarkTechExpandAmazonAppleFacebookGoogleMicrosoftSamsungBusinessSee all techReviewsExpandSmart Home ReviewsPhone ReviewsTablet ReviewsHeadphone ReviewsSee all reviewsScienceExpandSpaceEnergyEnvironmentHealthSee all scienceEntertainmentExpandTV ShowsMoviesAudioSee all entertainmentAIExpandOpenAIAnthropicSee all AIPolicyExpandAntitrustPoliticsLawSecuritySee all policyGadgetsExpandLaptopsPhonesTVsHeadphonesSpeakersWearablesSee all gadgetsVerge ShoppingExpandBuying GuidesDealsGift GuidesSee all shoppingGamingExpandXboxPlayStationNintendoSee all gamingStreamingExpandDisneyHBONetflixYouTubeCreatorsSee all streamingTransportationExpandElectric CarsAutonomous CarsRide-sharingScootersSee all transportationFeaturesVerge VideoExpandTikTokYouTubeInstagramPodcastsExpandDecoderThe VergecastVersion HistoryNewslettersArchivesStoreVerge Product UpdatesSubscribeFacebookThreadsInstagramYoutubeRSSThe VergeThe Verge logo.Flash floods can strike without warning — this new technology could change thatNotificationsNotificationsComments DrawerNotificationsCommentsLoading commentsGetting the conversation ready...ScienceCloseSciencePosts from this topic will be added to your daily email digest and your homepage feed.FollowFollowSee All ScienceAICloseAIPosts from this topic will be added to your daily email digest and your homepage feed.FollowFollowSee All AIReportCloseReportPosts from this topic will be added to your daily email digest and your homepage feed.FollowFollowSee All ReportFlash floods can strike without warning — this new technology could change thatSatellite data combined with machine learning gives meteorologists new tools to predict deadly floods sooner. by Megan WollertonCloseMegan WollertonPosts from this author will be added to your daily email digest and your homepage feed.FollowFollowSee All by Megan WollertonSep 18, 2026, 11:00 AM UTCLinkShareGift Image: Cath Virginia / The Verge; Getty ImagesScienceCloseSciencePosts from this topic will be added to your daily email digest and your homepage feed.FollowFollowSee All ScienceAICloseAIPosts from this topic will be added to your daily email digest and your homepage feed.FollowFollowSee All AIReportCloseReportPosts from this topic will be added to your daily email digest and your homepage feed.FollowFollowSee All ReportFlash floods can strike without warning — this new technology could change thatSatellite data combined with machine learning gives meteorologists new tools to predict deadly floods sooner. by Megan WollertonCloseMegan WollertonPosts from this author will be added to your daily email digest and your homepage feed.FollowFollowSee All by Megan WollertonSep 18, 2026, 11:00 AM UTCLinkShareGiftOn the morning of June 9th, Laura Lin was working from her home in Lanesville, a rural southern Indiana town about 15 miles from the Kentucky border. She was on a Zoom call, unaware that the heavy rain outside was beginning to flood her yard.“I look over to where the barn is over there, and I see pieces of my wood floating, and I was like, ‘What?’ And I immediately was like, ‘I have to go.’ Close my laptop, and I get my kids up, and I’m like, ‘Something’s wrong,’” Lin recalls.Lin and her family got out safely and sheltered at a neighbor’s house, but Lanesville got over 8 inches of rain within just a few hours that day, way over the threshold for what is considered “heavy” rainfall. The water was gone a few hours later.Floods are the second-deadliest type of weather event in the United States and the deadliest globally. Just 6 inches of fast-flowing water can knock an adult off their feet, 12 inches can lift a car, and 2 feet can move larger vehicles like trucks and SUVs. A warming climate means more instances of extreme rainfall in the US, creating more opportunities for floods to occur. The deadly flash flooding in Nepal in August caused by a glacier collapse shows the power of water to devastate communities worldwide.Fortunately, new tools are being developed to counter this threat. A new software called the Transient Artifact and Continuous Learning System (TACLS) uses satellites and machine learning to try to spot areas that could flood sooner and help the National Weather Service (NWS) make better decisions when it comes to issuing flash flood alerts.“We didn’t get the ‘get on your roof’ warnings until I was already at that person’s house,” Lin says. “All the people in town were already flooded when they started sending out alerts, being like, ‘Get on your roof. We’re dispatching boats.’ So it was way too late.”TACLS has the potential to help change that.“It will help you save lives,” says Ivory Small, science and operations officer at the NWS San Diego Weather Forecast Office. Without TACLS, Small says, a storm moving into your area “could kill some folks.” With TACLS, “you can put out the warning and save some folks.”The river near Laura Lin’s house in Lanesville, Indiana, flooded after a recent rain. New technology could provide residents with quicker warnings when the threat arises. Image: Megan Wollerton / The VergeFollow the waterWeather forecast offices are responsible for issuing severe weather alerts, including flash flood warnings. There are 122 of them throughout the US and its territories, says Jayme Laber, senior service hydrologist for the NWS Weather Forecast Office in Oxnard, California, which oversees the LA area. Small and Laber worked closely on the TACLS project with a broader team of scientists and researchers, and they hope this new project will streamline weather alerts. To understand TACLS, it’s important to first understand how the NWS issues flood alerts.Currently, the people working in weather forecast offices use a variety of tools to monitor flood conditions. They rely on rain and stream gauges to tell them how much water is in a certain place at any given time, Laber says. They track developing storms using satellites and weather radars. The staff monitors rainfall, but their computers also alert them when it rains more than what is considered safe for a specific area. “Based on the type of soils we have, the steepness of slopes, we have developed flash flood guidance so that we know that if we get this amount of rain in this amount of time in that area, it could result in flash flooding,” Laber says.“It will help you save lives.”— Ivory SmallAll of that, plus the years of expertise inside those buildings, determines whether or not someone presses a button to issue a flood alert for their area, Laber says.A watch is the least serious alert. It means that there’s a chance a flash flood could happen. The watch might be issued anywhere from 12 to 48 hours in advance, Laber explains. He says local emergency teams should start working on “flood protection measures” when the flood watch is issued, so there’s enough time to prepare before people are stuck and forced to find higher ground. Response teams might review evacuation plans, set up sandbags and road barriers, and identify the highest risk areas ahead of a potential flood.An advisory is the next level up. It might be flooding, “but it’s more nuisance-level. It’s not going to threaten life or property,” Laber says. A warning is the most serious alert. “That’s a life-and-death possibility of being swept off your feet [or] being run over by a car floating down the stream,” Small says.The NWS differentiates a regular flood from a flash flood based on time, says Small. If an area floods in under six hours, it’s a flash flood.The NWS has a lot of different ways to determine whether or not it needs to issue a flash flood warning, but its tech has limitations, says Yehuda Bock, TACLS project lead and research geodesist at the Scripps Institution of Oceanography at the University of California, San Diego. The satellites the NWS uses are more detailed over oceans than land, he says. And tracking rainfall as it’s happening doesn’t give you enough lead time, “because once there’s precipitation, you’re already in the event itself,” he explains.Deserts and other less populated areas typically have fewer rain gauges, Laber says. “We have a fairly dense network [of rain gauges], but it doesn’t cover every inch of the Earth.”Up in the airTACLS uses satellites and machine learning to identify places on a map — for now, just in California — that risk transitioning from rain to dangerous flash floods. But soon, the technology will be available to every NWS office. Scientists from UCSD, the NWS, and NASA worked together to create TACLS. NASA’s Earth Science Technology Office funded the project through its Advanced Information Systems Technology program.TACLS relies on the Global Navigation Satellite System (GNSS), which is primarily used to predict earthquakes, says Bock. But the GNSS network, which consists of satellites and ground sensors, also happens to provide data that can help forecasters decide if they need to issue a flash flood warning. The more water vapor, or moisture, there is in the atmosphere, the longer the delay in communication between the GNSS satellites and the sensors on the ground.The amount of moisture, or “precipitable water,” in the atmosphere tells the NWS what is actually happening in the skies leading up to a storm, Laber says.TACLS uses satellites and machine learning to identify places that risk transitioning from rain to dangerous flash floods.Before TACLS, Laber and his colleagues had to rely on the forecast, which wasn’t always accurate. TACLS provides real-time data about the precipitable water in the atmosphere to compare to the forecast. This helps them determine whether the storm is actually moving at the speed predicted by the weather models. “It helps us to make better decisions in our warning decision process,” Laber says.Even with the GNSS moisture data, forecasters still need something to analyze all of it. Bhavik Chandna, a UCSD graduate student, spent about a year developing the machine learning side of TACLS using what’s called long short-term memory architecture. According to Chandna, this specific machine learning system works especially well for things like developing storms and other weather patterns that change over time.Years of atmospheric measurements from the GNSS network, as well as data about atmospheric rivers, precipitation, and flash flood warnings, were used to train the machine learning model to understand moisture changes in the atmosphere “as a storm develops,” Chandna explains. At the same time, the machine learning system was trained to determine if the conditions are right to issue a flash flood warning.“Machine learning sucks at some things and excels at some things,” says Joel Johnson, associate professor in the department of earth and planetary sciences at the University of Texas at Austin. Johnson, who was not involved in the TACLS project, knows firsthand the challenges of working with these types of tools. “In general, it’s very easy to have artifacts in your data, false positives. A sensor goes funky, and things like that,” Johnson says. Still, he says TACLS seems “well suited” to the task of predicting flash floods.False positives do happen, Chandna says. “A real weather system usually affects an area, not just one isolated GNSS station. If one station shows a strong signal but its neighbors do not, that detection can be suppressed. If several nearby stations see the same pattern, we have much more confidence that it represents a real atmospheric signal,” he adds.Chandna says TACLS is not intended to replace the forecasters who issue weather alerts. It’s just another piece of information that can assist in their decision-making, he says.Finding a better wayTACLS software is already being used in the LA and San Diego weather forecast offices where flash flooding is a common occurrence, says Laber. “In the West, we deal with a lot more flash flooding, where it happens a lot quicker,” he explains. Today, TACLS shows them a graph with various points telling them what is going on in real time.A newer version of TACLS with improved graphics is currently being loaded onto the NWS system and is nearly done, Laber says. It will be available to all of the NWS offices in the second half of October, he adds. This iteration has a map with detailed graphics showing how much it’s raining and whether an extreme weather event is occurring.According to Laber, having more tools — like this latest update to TACLS — will further improve their decision-making when it comes to issuing flash flood alerts, and, in turn, maybe speed up alert times so people know there’s a problem sooner.Laber says TACLS will likely be used the most in the Western US because the majority of the GNSS sensors are located in places known for earthquake activity. But Chandna thinks TACLS could work well elsewhere too. As long as there are some GNSS sensors and available local weather data, “we could use the same ideology that we have in our system around different parts of the world,” he says.Small is thinking beyond TACLS. He believes ongoing research is crucial to improving how the NWS issues flash flood warnings. “We learn what’s going on a little more every time we do a storm because they’re all different,” he says. “It’s imperative that you do the research after, especially big events, and see what caused it to do what it did, and why, if it was different than what we already knew, did this occur.”Lin’s concern is Lanesville and her home. “It would have been helpful to know [water was coming] before I was locked in. I couldn’t go anywhere,” Lin says. “If I got an alert that was like, ‘You need to get out now if you’re in lower-lying ground, this is coming.’ Yeah, that would have [been helpful]. But it came in so fast.”Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.Megan WollertonCloseMegan WollertonPosts from this author will be added to your daily email digest and your homepage feed.FollowFollowSee All by Megan WollertonAICloseAIPosts from this topic will be added to your daily email digest and your homepage feed.FollowFollowSee All AIClimateCloseClimatePosts from this topic will be added to your daily email digest and your homepage feed.FollowFollowSee All ClimateReportCloseReportPosts from this topic will be added to your daily email digest and your homepage feed.FollowFollowSee All ReportScienceCloseSciencePosts from this topic will be added to your daily email digest and your homepage feed.FollowFollowSee All ScienceMost PopularMost PopularThe Apple Watch Series 12 is the start of a new wearable eraThe 2.5-hour AI-generated Odyssey movie is 2.5 hours too longInside the suddenly explosive world of AI safetyThe iPhone 18 Pro’s big camera update is all about the small gainsVideoI wore Snap’s $2,200 smart glassesThe Verge DailyA free daily digest of the news that matters most.Email (required)Sign UpBy submitting your email, you agree to our Terms and Privacy Notice. This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.Advertiser Content FromThis is the title for the native adMore in Science8Verge ScoreThe Apple Watch Series 12 is the start of a new wearable eraThe AI data center e-waste problem is huge — and getting biggerYour ‘health age’ is fakeTrump throws out power plant climate pollution rulesTrump is giving data centers a pass to polluteArizona’s lifeline for chip manufacturing is drying upThe Apple Watch Series 12 is the start of a new wearable eraVictoria SongSep 17The AI data center e-waste problem is huge — and getting biggerJustine CalmaSep 16Your ‘health age’ is fakeVictoria SongSep 16Trump throws out power plant climate pollution rulesJustine CalmaSep 14Trump is giving data centers a pass to polluteJustine CalmaSep 12Arizona’s lifeline for chip manufacturing is drying upJustine CalmaSep 11Advertiser Content FromThis is the title for the native adTop StoriesSep 17Inside the suddenly explosive world of AI safetySep 17Your robotaxi might be a narcSep 16The iPhone 18 Pro: an ambitious new camera on the marginsVideoSep 17The Apple Watch Series 12 is the start of a new wearable eraThe VergeThe Verge logo.FacebookThreadsInstagramYoutubeRSSContactTip UsCommunity GuidelinesArchivesAboutEthics StatementHow We Rate and Review ProductsCookie SettingsTerms of UsePrivacy PolicyYour California Privacy RightsYour Privacy ChoicesCookie NoticeAdChoicesLicensing FAQAccessibilityPlatform StatusPenske Media CorporationThe Verge is a part of PMX Global, LLC, a subsidiary of Penske Media Corporation.© 2026 VM Publishing, LLC. All rights reserved.Our SitesThe American PavilionARTnewsArt in AmericaArtforumArt Week NYCBeauty IncBillboardDeadlineDick Clark ProductionsThe DodoEaterFlow SpaceFNGold DerbyGolden GlobesThe Hollywood ReporterIndieWireLife is BeautifulPopsugarPunchSJ DenimRobb ReportRolling StoneSB NationSHE MediaShe KnowsSourcing JournalSoapsSporticoStyleCasterSXSWThrillistVarietyThe VergeVibeWWDNotifications DrawerThe VergeThe Verge logo.Sign in to see your notifications or create an account to join the conversation.Sign in

Satellite data combined with machine learning is being developed as a new technology to enhance the prediction of deadly flash floods, offering potential improvements to warning systems for meteorologists and emergency services. This effort is encapsulated in a software called the Transient Artifact and Continuous Learning System or TACLS, which aims to provide earlier and more accurate flood alerts. This development addresses the critical challenge highlighted by events like the flooding experienced by Laura Lin, where delayed warnings significantly impacted safety.

Currently, the National Weather Service (NWS) relies on various tools, including rain and stream gauges, satellite tracking, and established flash flood guidance based on topographical data, to monitor conditions and issue alerts. While this existing framework incorporates expertise and physical measurements, limitations exist, particularly regarding the speed of precipitation and the accuracy of data collection across diverse geographical areas. The NWS employs a tiered alert system, differentiating between watches, advisories, and warnings, with a flash flood specifically defined as an event occurring in under six hours.

TACLS attempts to overcome these limitations by utilizing satellite imagery and machine learning algorithms to identify areas at risk of transitioning from rainfall to dangerous flash floods, initially focusing on regions like California. The system leverages data from the Global Navigation Satellite System (GNSS), which provides data on atmospheric moisture or "precipitable water," as a key input. By comparing this real-time moisture data against weather forecasts, TACLS helps forecasters determine if a storm is progressing at the predicted speed and aids in making more informed warning decisions. This integration of satellite data and atmospheric moisture measurement facilitates a comparison against traditional forecasts, allowing for better situational awareness.

The machine learning component of TACLS was developed by Bhavik Chandna, who utilized a long short-term memory architecture. This model was trained on years of atmospheric measurements, data concerning atmospheric rivers, precipitation rates, and historical flash flood warnings to effectively understand how moisture changes in the atmosphere as a storm develops. The model is designed to predict not only the risk of a flash flood but also the appropriate level of warning necessary. While this technology is powerful, experts acknowledge that machine learning can introduce artifacts or false positives, necessitating careful analysis. However, scientists suggest that when corroborated with data from multiple GNSS stations, the confidence in detecting a real atmospheric signal is significantly increased.

The system is intended to assist, rather than replace, human forecasters, providing additional data to refine decision-making processes. Future iterations of TACLS are being developed to incorporate improved graphics, offering real-time visualization of rainfall and extreme weather events for NWS offices, which is expected to further accelerate alert times. Researchers emphasize that ongoing study following major weather events is imperative to continuously refine warning protocols and improve the understanding of atmospheric dynamics. Ultimately, the goal of technologies like TACLS is to leverage advanced data analysis to enhance public safety by providing timely and precise warnings against rapidly developing natural hazards.