A global earthquake alert system developed by researchers at Google and the University of California, Berkeley now delivers early warnings through Android phones to users in 98 countries. The system uses built-in motion sensors in smartphones to detect earthquakes and provide alerts to nearby populations. It operates without the need for dedicated hardware, making it accessible in regions lacking traditional seismic infrastructure.
Phones as Sensors in Real-Time Networks
Android phones include accelerometers that detect motion. When stationary devices register movements consistent with seismic activity, they transmit data to Google's servers. If enough phones in the same area report similar signals, the system identifies the event as an earthquake. According to the study, this method enables earthquake detection in areas without access to conventional seismic networks.
The system operates through Google Play Services, which runs on approximately 70% of smartphones globally. From April 2021 to March 2024, it detected 11,231 earthquakes. Among those, 85% matched events listed in official seismic catalogs, confirming the method’s reliability.
Alert Accuracy and Performance Over Time
Early magnitude estimates produced by the system showed a median error of 0.50, which was reduced to 0.25 by the end of the study period. These improvements brought its accuracy closer to that of Japan’s ShakeAlert system, which is among the most advanced national early warning systems.
The alert system uses two types of warnings: "TakeAction" and "BeAware." TakeAction alerts override all phone settings and display instructions for strong shaking. BeAware alerts appear as regular notifications and are intended for areas where shaking is predicted to be lighter.
The study notes that the system’s design prioritizes timely alerts while minimizing false positives. Three false alerts were recorded during the three-year evaluation period.
Performance During Major Earthquake Events
During the February 2023 earthquake in Turkey, which had a magnitude of 7.8, the system initially underestimated the event. The early estimate reported a magnitude of 4.5. However, subsequent updates to the detection algorithm showed that the same event could have been detected within 6.3 seconds, and the estimated magnitude would have been 7.4.
In a separate case, during a magnitude 5.7 earthquake in Nepal, the system issued more than 10 million alerts, and some users received the warning up to 60 seconds before the shaking began.
User Feedback and Reach
A survey of 1.55 million users who received alerts showed that 85% found the alerts very helpful, regardless of whether they felt ground movement. The system recorded that 36% of users received alerts before shaking, 28% during, and 23% after the event.
Among users who received TakeAction alerts, 28% reported taking protective action, such as dropping and covering. According to the study, this rate of response is higher than what was observed in similar warning systems.
Trust in the alerts remained high, with 84% of users saying their trust increased after experiencing the system. Fewer than 0.1% of users chose to disable alerts, based on the study’s data.
Expanded Access to Earthquake Warnings
Before the Android-based system was implemented, earthquake early warnings reached about 250 million people worldwide. As of the end of the study period, alerts were available to more than 2.5 billion people, many of whom reside in countries with limited seismic monitoring infrastructure.
The system has been used to detect earthquakes in locations without permanent seismic stations, including offshore events as far as 100 kilometers from coastlines, provided the magnitude exceeded 4.5.
According to the study, the system "increases coverage in regions where building dedicated networks is not financially viable", and it can be used to enhance existing hazard models and post-event assessments.
Operational Considerations and Limitations
The study recognizes that the system performs best in areas with high smartphone density. It also reports a limitation in detecting very large earthquakes exceeding magnitude 7.5, due to the complexity of modeling and the nature of strong ground motions. Device models, building types, and regional network configurations also affect performance.
The Android system does not replace traditional warning networks but serves as a complementary solution, especially in underserved regions.
Read next: Study Uncovers Gender-Based Disparities in Career Advice from Popular AI Models
