Researchers hope to offer early earthquake warnings around the globe using innovative crowdsourcing approach
By Nancy D. Lamontagne
Earthquakes remain impossible to predict, but earthquake early warning systems can give people seconds to minutes of notice before shaking reaches their location. This bit of advanced warning can save lives by giving people time to take cover or stop driving.
Although some earthquake-prone areas such as California have networks of seismic sensors that measure vibrations in the earth and can provide life-saving early warnings, many developing countries don’t have adequate ground-based sensors.
These countries do, however, have millions of smartphone users.
Richard Allen, Ph.D., and his team at the Berkeley Seismological Laboratory want to use these smartphones to build a dense worldwide network of earthquake sensors that could not only bring early earthquake warnings to regions without established seismic networks, but also provide faster and more accurate warnings in areas with traditional seismic networks. As a first step, they have worked with Deutsche Telekom’s Silicon Valley Innovation Center to develop an Android app called MyShake that uses the built-in accelerometers on today’s smartphones to sense earthquakes.
The team quickly learned that realizing their vision would require not only good science and technology, but also sensitivity to the needs and desires of the global network of community members on which the project would depend.
“When we first came up with this idea to record seismic waveforms on phones it seemed simple — we were just using a different type of sensor,” explained Allen, who has been working on early-warning systems with traditional seismic networks for some time. “The reality is that you have to not only perform the seismology, you have to also develop the technology and get the message out there to people that they can and should download this app.”
Thanks to media interest surrounding publication of a scientific paper on the algorithm used in MyShake and a talk Allen gave on MyShake at a science and technology meeting, the app was downloaded 150,000 times in the first two months after its release.
Even though the app doesn’t yet provide early warnings, people who have downloaded it are helping the researchers work toward that goal by allowing their phones to collect data. The data is sent automatically to the Berkeley Seismological Laboratory, where researchers compare phones’ accelerometer readings to data from traditional seismic monitors during real earthquakes. These comparisons allow the team to validate and fine-tune the algorithm they developed to distinguish cell phone motion caused by an earthquake from other forms of motion.
The researchers are pleased with the early results of this crowdsourcing effort. They far exceeded their expectation of 10,000 downloads and successfully gathered smartphone data on a magnitude three earthquake in Los Angeles—a surprising level of sensitivity given that they were initially only expecting to be able to detect earthquakes of magnitude five and higher.
“There are many facets to address to make a citizen science project like this work,” said Allen. “Recognizing what all of those facets are and figuring out where to spend our limited resources was the real challenge.”
One of the primary challenges in developing MyShake was how to gather enough data to detect earthquakes without using too much of the phone’s resources. “As a seismologist, I would love to have the phones recording data and sending this data 24 hours a day,” Allen said. “However, we had to strike the right balance by thinking very carefully about what information we really needed and how to collect that data without impacting users in a way that would cause them to turn the app off.”
Managing phone resources
To keep power and data consumption low, the MyShake app defaults to a low-power mode when the phone is in motion, such as when it is picked up or carried around. Once the phone is at rest, the app monitors for earthquakes more carefully, and if it detects movement that fits the profile of an earthquake, it sends that information to servers at the Berkeley Seismological Laboratory. This earthquake “trigger” allows the real-time detection of earthquakes.
The phone also records five minutes of motion to capture the complete waveform for the earthquake. This includes one minute of data that precedes the earthquake — to provide a picture of what was happening before the trigger — and four minutes of data after the earthquake trigger. The five minutes of data is uploaded to the laboratory’s server only when the phone is connected to Wi-Fi and plugged into power, so as to not sap the phone’s battery, data usage or memory. The waveform data, which is used for further research and refinement of the detection algorithm, typically arrives to the server within 24 hours of the event. Separating the upload of the trigger event from the waveform data was key to keeping power and data usage as low as possible.
Another technical challenge lies in seamlessly connecting the app with the servers. “It takes a ton of resources to provide a smooth connection between a simple, usable app and the array of info systems that support the back end,” Allen said. “It is complicated, but we are working very hard to make it work for this seismology application.”
The researchers are currently refining their algorithm based on data already gathered with MyShake with a goal of launching an early alert version of the app in the next year or two. They also plan to attract more users by translating the MyShake user interface into other languages and developing an iPhone version.
Additional resources on MyShake:
Nancy D. Lamontagne is a freelance science communicator and a contributing writer for Creative Science Writing and the Thriving Earth Exchange.