Site icon Temblor.net

Crowdsourcing Earthquake Early Warning Systems

Smartphone usage may provide a global earthquake early warning system through crowdsourced motion detection with the MyShake app.
 

By Melissa L Weber, M.A., freelance science writer (@Melwriter)
 

Citation: Weber, M. L., 2020, Crowdsourcing Earthquake Early Warning Systems, Temblor, http://doi.org/10.32858/temblor.102
 

Researchers at the Berkeley Seismology Lab, University of California, Berkeley are using accelerometers inside smartphones to crowdsource shaking data in an effort to create a global earthquake early warning system with the MyShake mobile app.

 

Users are key

Smartphone users who download MyShake become part of a citizen scientist brigade. The app monitors sudden changes in motion, which are analyzed by an algorithm that was designed to distinguish between earthquake and non-earthquake seismic waves. When earthquake motion is detected, the data is sent to the MyShake server and clustered in both time and space. These two levels of detection identify actual earthquakes in the aggregated data.

“The app runs all the time, but only sends data [to the server] when the algorithm running on the phone detects earthquake-like motion,” said Qingkai Kong, Assistant Data Science Researcher at the Berkeley Seismology Lab, and lead author on a two-article series of papers in Seismological Research. “Then, it aggregates data from all phones in one area to confirm the earthquake, since one phone may mistakenly report human activity as an earthquake.”

 

With the MyShake app, scientists are developing a global Early Earthquake Warning system using cell phone data. Credit: public domain

 

Since the app’s launch in February 2016, more than 1 million downloads have translated to more than 100,000 active users making daily data contributions. Users have successfully recorded more than 1,000 earthquakes worldwide.

 

Saving lives with a few seconds of warning

When a fault ruptures, the seismic energy travels at the speed of sound, while the data travels at the speed of light. Kong compares it to a well-known weather phenomenon: lightning followed by thunder. In this case, the warning is the lightning and the shaking is the thunder that follows.

How much does a few seconds of warning actually help people? According to Men-Andrin Meier, Caltech associate staff seismologist, a 10 to 15 second warning is good enough to provide safety advantages, like moving to avoid falling objects. Other safety measures could be triggered automatically.

 

During an earthquake drop to the ground (or lock wheels if using a wheelchair or walker), cover your head and neck, and hold on to a sturdy object if possible. Credit: Earthquake Country Alliance.

 

Improving the system with better simulations

Researchers at the Berkeley Seismology Lab are looking at the performance of MyShake in detecting earthquakes using both real data and derived simulations to create alerts to warn people before shaking reaches their location.

Two examples using data recorded by MyShake were used to illustrate the potential performance of the system: the June 2016 magnitude-5.2 earthquake in Borrego Springs, California and a magnitude-4.4 event in January 2018 in Berkeley, California. These two events were chosen because of the concentration of MyShake app users in the San Francisco Bay Area and Los Angeles area. The test results showed that if the network detection algorithm had been operational at the times of the two earthquakes, MyShake could have provided an early warning to the region.

From the actual data generated by the app, the researchers were able to develop an enhanced algorithm that increased the speed and accuracy of the processing time for the aggregated data in real time.

Next, the researchers wanted to assess the viability of a MyShake early earthquake alert in other parts of the world. They conducted simulations using historical data in earthquake-prone regions including Haiti, Chile, Nepal, and Sulawesi (Indonesia). They chose information from earthquakes since 1980 with a magnitude of greater than 4.0 and assumed a similar density of users as in the Los Angeles area (0.1% of the total population).

Their tests suggest that the most accurate and timely warnings occur during shallow, upper crust earthquakes in densely populated areas with many active MyShake users. Areas with too few users and earthquakes that occurred off the coast usually produced a poorer estimate of the magnitude of the quakes. Adjustments to the algorithm could be made based on regional differences, like the number of active app users in an area. This possible enhancement and other potential changes will be studied in the future.

 

Users are part of the challenge

Maintaining a user base large enough to aggregate the data is another challenge.

“Earthquakes are rare,” said Kong. “People lose interest and delete the app over time. If people don’t use it, reliable detection is difficult. We are constantly working on adding a more-engaging user interface and other functions to engage the users.”

 

A MyShake Alert warns app users of imminent shaking due to an earthquake. Credit: Berkeley Seismology Lab

 

MyShake is already one of the platforms that deliver California’s ShakeAlert early earthquake warning system alerts to the public. Currently, ShakeAlert uses a network of seismic monitoring stations, or seismometers to detect quakes and send alerts to warn people. MyShake could turn every user’s phone into a mini seismometer — essentially creating an inexpensive global dense seismic network.

 

Check your earthquake risk at Temblor
 

Further Reading

Kong, Q., R. Martin-Short, and R. M. Allen (2020) Toward Global Earthquake Early Warning with the MyShake Smartphone Seismic Network Part 1: Simulation Platform and Detection algorithm, Seismol. Res. Lett. XX 1-12, doi: 10, 1785/0220190177

Kong, Q., R. Martin-Short, and R. M. Allen (2020) Toward Global Earthquake Early Warning with the MyShake Smartphone Seismic Network Part 2: Understanding MyShake Performance around the World, Seismol. Res. Lett. XX 1-16, doi: 10, 1785/0220190178