Back in 2020, Google kicked off a project to crowdsource signals from Android phones that an earthquake might be imminent.
The system leveraged the accelerometers inside these devices to gather seismic data, combine it to detect patterns consistent with an earthquake, and deliver timely alerts to people who might be affected.
Three years on, researchers who worked on the project have published a paper in Science this week that shows their earthquake detection tech worked pretty well – proving its utility in places that don't have seismic stations that can study and deliver local alerts.
The system worked by detecting sudden increases in the acceleration of a phone that are consistent with ground motions in earthquakes. Once triggered, the phone would send a signal to a central server, which would listen for similar signals from other phones in the same area and around the same time. When the observed data met a high enough confidence level, an earthquake was officially declared.
Over three years of operation, the crowdsourced system deployed across 98 countries detected an average of 312 earthquakes a month, with magnitudes ranging from a low M 1.9 to a dangerously high M 7.8. According to users' feedback, 85% of people who received an alert felt shaking, and as many as 36% of survey respondents reported receiving an alert before a shake.
That's not bad for a system that doesn't require a whole lot of local infrastructure to help keep people safe. The researchers had deployed this system across 2.5 million devices via pre-installed software that ships with all Android phones, which meant it was switched on by default and didn't require inputs from owners to use.
The accelerometer-based crowdsourced system is a more modern version of a 2015 effort called ShakeAlert, in which "the GPS receivers in a smartphone can detect the permanent ground movement (displacement) caused by fault motion in a large earthquake."
This tech also tied into the Android Earthquake Alerts system, which broadcasts warnings directly to phones in areas where seismologists expect shakes to occur, along with instructions on how to take cover.
This sort of public infrastructure not only helps us learn more large earthquakes and prevent widespread hazards, but also provides aggregated observations of shaking over multiple earthquakes to inform regional models. The built-in feedback system that people can use to indicate whether they felt a quake can also enhance that data, and potentially give us a better chance at spotting future quakes before they happen.
Source: Science