The US National Weather Service (NWS) issues more than 4,000 flash flood warnings each year, and until now, every single one of them has triggered a wireless emergency alert—the blaring notifications that pop up on your phone to warn of severe weather in your area.
But starting in January, the NWS will cut the number of flood alerts Americans receive by about 80%, and only send notifications for potentially life-threatening floods that fall under the agency’s highest threat categories.
Part of the motivation for muzzling alerts came from a recent deluge of complaints from users angry that their phones woke them up in the middle of the night to warn of floods that never materialized. As smartphones bring emergency alerts to our pockets and nightstands, authorities are struggling to figure out how to balance their mission to keep the public informed with the risk of sending so many notifications that people start to tune them out.
“We don’t want to desensitize people to our warning messages,” said NWS hydrologist Daniel Roman. “We want to limit the number of wireless emergency alerts they’re getting so that they can focus on taking life-saving action.”
Meanwhile, authorities are overhauling the wireless emergency alert system for all hazard warnings to make them more informative, better targeted and—they hope—less frustrating for users.
California’s subtle shake
On Jan. 3, Los Angeles debuted an early warning app for earthquakes called ShakeAlert, with the goal of giving residents crucial seconds to prepare before the ground begins to tremble. Almost immediately, it set off a public uproar for sending too few alerts.
Originally, the app was programmed to warn users if seismic sensors detected an earthquake at an intensity of at least 4 on the Modified Mercalli intensity (MMI) scale—shaking that is noticeable, but not strong enough to damage buildings. On July 5, the Ridgecrest earthquake shook much of Los Angeles at an intensity of about MMI 3, which feels like the tremor from a passing truck. It passed just below the threshold for triggering a ShakeAlert notification.
Many residents were incensed that they felt an earthquake but didn’t get an alert. (“Super weird and useless for my blinds and walls to shake and get zero notifications,” one Twitter user mused. “What’s the point of the app?”) They raised such a ruckus that, even though early warnings are less likely to arrive in time for minor earthquakes, Los Angeles authorities lowered the threshold for sending notifications to MMI 3.
But as earthquake researchers Elizabeth Cochran and Allen Husker argue in an article published Thursday (Nov. 21) in Science, there’s a risk in over-alerting for relatively minor events. They point to the example of Mexico City, which has a well-regarded earthquake early warning system. In one 2017 incident, the system sent an alert for an earthquake most residents didn’t feel—causing one man to hurt himself by jumping out of a window to escape a building he feared would collapse. Media reports also linked the false alarm to two heart attack deaths, which are harder to corroborate.
“Though overreactions may be uncommon,” the authors argue, “even a very small percentage of a population responding imperfectly to an alert can have grave consequences.”
Crying wolf in Colorado
In Colorado, avalanche authorities have internally researched their track record on issuing false alarms. “The results suggest that we do cry wolf a little bit,” said Ethan Greene, director of the Colorado Avalanche Information Center. “Even though we definitely think about message fatigue, we err on the side of raising the red flag because the consequences of missing some of these events are life-threatening.”
But Mark Roulston, a meteorologist-turned-data-scientist who builds weather prediction models for Hivemind, says that emergency forecasters should take seriously the potential risks of crying wolf. Over time—as in the famous fable—false alarms might make the public less likely to heed authorities’ warnings.
“It’s all very well having a very skilled forecast,” Roulston said, “but if you botch the job of communicating it to the end user then you could destroy a lot of value in that last step.”
Part of that communication, Roulston argues, should involve telling users about the uncertainty in hazard forecasts. Colorado authorities, for example, grade the threat of an avalanche on a scale of five increasingly risky levels. Roulston says that unequivocal forecasts—like the old NWS alerts that just told users there was a flood warning in their area with no explanation of its likelihood—are more likely to wear down people’s trust.
The other half of communication is making sure that people know what to do once you’ve alerted them to an emergency. Robert de Groot, who works with the US Geological Survey on ShakeAlert, says West Coast residents should already have the basic “drop, cover, and hold on” earthquake response drilled into their minds, and emergency managers take pains to educate people about what to do before, during, and after a quake.
“There are always going to be outliers where people jump out of a building or run into the street and injure themselves,” de Groot said. “What we really want to rely on is that whatever actions people would take if they start to feel shaking, we would assume they would take identical actions if they receive an alert, but sooner.”
Designing better alerts
Authorities hope they can address some of these communication challenges with a suite of changes to emergency alerts that will allow them to pack much more information into notifications and target them more precisely.
Starting Nov. 29, the US Federal Emergency Management Agency announced that wireless emergency alerts would quadruple in size from 90 characters to 360 characters. The NWS plans to use the extra space to give users detailed instructions in English and Spanish about what to do next. When visibility drops during a dust storm, for example, notifications will coach users to “be ready for sudden drop to zero visibility,” “pull far off the road and put your vehicle in park,” and “turn the lights off and keep your foot off the brake.”
The following day (Nov. 30) the US Federal Communications Commission has mandated that all new phones come equipped to deliver geo-targeted emergency alerts that only appear on devices within a tenth of a mile of the warning area. Currently, any phone connected to a cell tower near the danger gets an alert.
Overall, the NWS aims to cut down on the number of alerts it sends out while making them more digestible through its Hazard Simplification Project. Starting in January, phones will only buzz for floods that fall under the agency’s “considerable” or “catastrophic” categories, and all hazard warnings will be reformatted into simple bullet points.
In the meantime, emergency managers continue to tinker with alerts that will better capture people’s attention. Back in California, Robert de Groot says ShakeAlert’s designers are debating how their early warning notifications should sound. The tone should be distinct from ringtones, message chimes, and other disaster alerts, at an audible frequency for people of all ages, and convey a sense of urgency without being too startling. He thought they might end the debate by borrowing Japan’s earthquake alert noise—a chipper trill somewhere between a bicycle bell and a doorbell—but others thought it wasn’t alarming enough.
“This is a new thing,” de Groot said. “There are going to be growing pains. We need to make improvements. But we’re committing resources to make it happen.”