Someday, the West Coast will, out of nowhere, have a very bad day. Perhaps the ground will split open across 180 miles, from Los Angeles to the Salton Sea, in the worst local upheaval in more than 150 years. Or perhaps the 600-mile-long scar off the coast of Washington, where a hunk of crust called the Juan de Fuca plate is constantly being shredded, will give way.
Whichever fault gives way first, scientists know the nightmare scenario will happen eventually—the only question is when. If it happens tomorrow, there won’t be any warning. “The way I know that an earthquake is happening is I feel shaking,” Robert-Michael de Groot told Newsweek. And if people were going to know about shakes in advance, de Groot would be one of them, since he works with the U.S. Geological Survey's program to create earthquake warnings.
On September 8 in Mexico City, a different scenario played out: A siren rang out more than a minute before the shaking began. When a second large quake struck just a week and a half later, residents weren’t as lucky—they got five seconds or less notice before the shaking started. Even still, every second could theoretically help people protect themselves.
If it sounds like the sort of system that would be useful in the earthquake-prone American west, well, USGS agrees. They have spent years developing a system called ShakeAlert, designed in the long run to offer an early warning of an earthquake, like the systems used in Mexico and Japan.
These systems don’t predict earthquakes, a feat scientists currently deem impossible. Instead, they exploit a quirk of how earthquakes work. The shaking people feel on the surface is the result of waves on energy flowing through the Earth—but those waves move much more slowly than the ones into which we encode our communications. If there’s enough distance between the source of the earthquake and the humans to be warned, the signal can arrive moments before the shakes do.
But right now, ShakeAlert is underfunded. In fact, President Donald Trump requested the project be tabled entirely in his administration’s fiscal year 2018 budget plan, although Congress has reinstated funding. Despite the struggles, it’s still being built, albeit more slowly than USGS hoped, and they are testing applications with a few targeted industries. “It doesn’t mean that the system is not up and running currently,” de Groot said. “We’re making great progress.”
ShakeAlert has two layers. First, there’s a network of sensors to pinpoint seismic activity. USGS has calculated it needs 1,675 stations spread across California, Oregon and Washington to do the job right. Currently 859 stations are installed and operational, each at a cost of about $60,000. Each is a rubber tub buried underground, with antennae perched on top to connect with the system’s processing center. It isn’t just cost that slows installations—each individual station needs to go through its own environmental permitting process. De Groot says they’ve been concentrating on installing stations near cities first, and that California’s network should be about three quarters complete by the end of this year.
Each station feeds into processing centers, the second layer of ShakeAlert. Here, the actual alert is created, a data file with the earthquake’s location, time and size. Then ever-improving algorithms crunch incoming data to produce estimates of just how much shaking will occur in a given location. And that’s an important constraint of ShakeAlert—it’s not about the magnitude of the earthquake itself, it’s about the shaking that results from it. Scientists measure that using a system called the Modified Mercalli Intensity Scale, which measure above-ground impact and ranges from intensity I, too gentle to even be felt, to intensity X, destroying buildings.
This is where ShakeAlert begins to interact with people and institutions on the ground. In early 2016, it began working with public and private organizations to turn those calculations into actions. Right, now USGS is focusing on partnerships in five sectors—utilities, transportation, education, healthcare and emergency management.
For each application, a group sets a local threshold for what intensity of shaking it should flag. Then, they can do something about it. The Bay Area light rail system is one of the longest-running partners; they want to use the alerts to stop trains before an earthquake begins shaking them, either automatically or with human intervention.
They can test the system with scenarios and hypotheticals, but real earthquakes are better for seeing how the system responds. “It’s always nice to have a real earthquake, but not too big, right?, de Groot said.” “Every earthquake that happens allows us to learn more.” They successfully got the message when a magnitude 4.4 quake hit near Berkeley on January 4, but because it was just after 2:30 a.m. local time, there weren’t any trains on the rails, so the test didn’t go any farther than the alert.
It will likely be a while before the public receives any alerts directly, however. The key problem for a full roll-out of a public alert system is technology. “There’s currently no way to get a message quickly enough from our system to the public because none of our systems were built for speed,” de Groot said. USGS is working with colleagues in Mexico and Japan to try to speed up the chain of communications. In the long run, they’d like to explore a host of alert mechanisms, whether television, cell phone or radio.
Most of all, they want to make sure the alerts are backed by social science to make sure they truly help people respond to the situation. That means ensuring alerts are unmistakable and that people know what to do when they get one. “The beauty of a hurricane warning or a tornado warning is you have some time to do something about it,” de Groot said. Not so with earthquake warnings, which operate on a scale of seconds rather than minutes or days.
And as recent mistaken nuclear missile and tsunami warnings have highlighted, they also need to find a way to handle false alarms—whether those are triggered accidentally or represent real threats that simply don’t come true. De Groot talks about safeguarding the system from, say, cyberattacks, but also suggests reframing false alarms as practice runs for when the inevitable does come.
Even if ShakeAlert is running public alerts by that point, there will likely be more fine-tuning to the system. That’s because there’s only so much scientists can do while the ground is still quiet. “For what we want to do with early warning, it’s hard to simulate those very very large events,” de Groot said. “Earthquakes are very complex organisms.”