When Earth trembles, the Internet lights up. People reach for their phones to search for information and to tell others what they’ve experienced. Multiple seismic agencies solicit “felt reports,” and now scientists are trying to use these reports to increase safety in areas where seismic instrumentation is scarce.
Seismologist Henning Lilienkamp of the GFZ Helmholtz Centre Potsdam and his colleagues recently investigated whether they could use felt reports alone, gathered through an app called LastQuake, to quickly determine whether an earthquake was high impact.
“So much of the problem of responses is getting information out as quickly as possible.”
Felt reports are nothing new, but they’ve usually been used in conjunction with data from seismological instruments. For example, felt reports collected by the U.S. Geological Survey (USGS) through its website Did You Feel It? are among multiple data sources used to create ShakeMaps, which describe the intensity of earthquakes.
Rather than replacing current systems that detect and categorize earthquakes, “our idea is to be a bit faster” while providing a very preliminary view of what’s occurred, Lilienkamp said. Their model could one day be the basis for a system that advises emergency responders on whether they should respond immediately to quakes (red light), investigate further (yellow), or do nothing (green).
The current work is only a starting point, but researchers are excited about the eventual applications of felt reports. “So much of the problem of responses is getting information out as quickly as possible,” said earthquake seismologist Rachel Abercrombie of Boston University.
Current Systems Face Limitations
The LastQuake app is one of three types of crowdsourced data—along with Twitter and traffic to their website—that the European-Mediterranean Seismological Centre uses to detect earthquakes.
The app’s 1.4 million users, plus website visitors, can report shaking by clicking on one of 12 illustrations that describe different severity levels. Very minor shaking is depicted as two happy people watching TV, whereas a serious earthquake is depicted as horrified cartoon figures writhing on a street as buildings crumble.
Lilienkamp and his colleagues took these felt reports a step further by using them to calculate the probability that an earthquake was high impact (defined as destroying at least one building, damaging at least 50 buildings, causing at least two fatalities, or causing any documented financial losses).
The researchers were not able to reliably identify high-impact earthquakes. However, all the earthquakes that their model assigned under 1% probability of being high-impact were in fact low-impact events, suggesting the system could be used to identify some situations that don’t require a response.
The fact that high-impact earthquakes are rare presents an inherent difficulty in studying them, said seismologist Mostafa Mousavi of Stanford University. Most of the events that Lilienkamp and his colleagues analyzed were low-impact earthquakes, Mousavi said, making the model much more prone to detecting those events—perhaps sometimes even erroneously. Systems based solely on felt reports could be valuable, he added, but “it is a difficult problem.”
If felt reports can help emergency responders quickly get to where they’re needed, the idea is worth pursuing, Abercrombie said. “If you’re trying to rescue people alive, it’s much easier at the beginning.”
When high-impact earthquakes occur, usually, there’s no question as to whether help is needed. These events often produce the so-called donut effect: a ring of plentiful reports far from the epicenter, with only sparse reports where the most damage occurred, explained seismologist Ina Cecić of the Slovenian Environmental Agency. The recent earthquakes in Turkey and Syria produced this pattern, Lilienkamp said.
Even the uncertain suggestion that an earthquake warrants attention might sometimes be useful. Lilienkamp pointed to the 21 June 2022 earthquake that occurred in Afghanistan as an example. The region has little seismic instrumentation, so scientists from the European-Mediterranean Seismological Centre initially couldn’t tell whether the quake had affected residents. “It turned out that more than a thousand people died,” Lilienkamp said. LastQuake received 50 reports—the minimum necessary to run the model—within about 8 minutes, and the model likely would have suggested that a response was warranted.
Many Paths Forward
Felt data were “how seismology actually started.”
Felt reports are just one of the ways seismologists are using both high- and low-tech instruments to make earthquakes easier to detect, said Cecić. For example, Japan, Mexico, and parts of the United States have services that alert people to nearby earthquakes before the shaking reaches them. Google provides a similar service that draws on data from the accelerometers built into Android phones. Meanwhile, cheap, user-friendly earthquake detectors called Raspberry Shakes are taking the place of expensive seismometers in many parts of the world.
Adding felt reports to this arsenal is a modern revival of “how seismology actually started,” Mousavi said. Long before seismic instruments existed, people recorded their experiences with earthquakes on whatever media was available. As reporting apps become more versatile and widely used, returning to more intuitive devices may be key to increasing safety worldwide.
—Saima May Sidik (@saimamaysidik), Science Writer