How the Android service is likely to look.

Your mobile phone could soon warn you of earthquakes. How does it work?

Google is launching its Android Earthquake Alerts System in New Zealand and Greece but what information does it draw on, and how useful will it be for an quake-prone country such as ours? Mirjam Guesgen explains. 

What does this earthquake alert system do?

Android phones will operate as earthquake detectors and show a warning alert that shaking is about to happen. 

How much warning people will get depends on how far you are from where the earthquake originates. If you’re further away you might get 10 seconds or more warning, if you’re only a few kilometres from the source, that warning might only be a second or two.

How is that different to what we already have?

New Zealand doesn’t have an earthquake early warning system in place. That’s the reason this country and Greece were chosen to trial the Android project. 

Currently, people can access information about hazards like earthquakes through GeoNet’s website or app but only after a quake or other hazard has been detected. GeoNet doesn’t send out alerts to phones before or in the first few seconds of an earthquake occurring. 

Will the Android system trigger an alert for all quakes?

No. Phones will get a Be Aware alert for earthquakes that produce weak shaking (3 on the Modified Mercalli Intensity Scale). Those alerts will include information on the magnitude of the quake and how far away it is. 

For earthquakes more intense than that, phones will get a Take Action alert telling people to drop, cover and hold.  

OK, but how does it work?

There are three major parts to this whole process: detection, analysis, and alert.

How does the detection work?

It relies on your phone’s movement sensor, called an accelerometer. That’s the same bit of tech that keeps videos or pictures the right way up when you tilt your phone. It can also, says Google, detect earthquake waves pulsing through the ground. 

These waves ripple out from the source of the quake and come in two forms. The first are P-waves, which travel fast but aren’t usually felt by people or cause much damage. The second are S-waves, which are the shaking part and what we experience as an earthquake. The accelerometers will register the P-waves and hopefully send you an alert before the S-waves hit.

The analysis?

If the accelerometers pick up shaking, a computer algorithm in the phone’s software figures out if that shaking is from a quake, not you jumping around or tossing your phone into your backpack. 

Once the Android algorithm decides that, yes, it’s a quake, the phone will send chunks of data, including rough location information, to Android servers in the US. The servers combine information from lots of phones about where the earthquake is to estimate how strong the shaking will be. 

And the alert?

If the servers get enough evidence that a quake’s about to hit, they’ll then fire an alert message back to people’s phones. 

That all happens before or in the first few seconds of shaking because the phone signals, transmitted by light through fibre internet cables, travel faster than earthquake waves.

But how do the algorithms know what’s a quake?

Analysis is the trickiest part of the system. “The secret sauce is the algorithm,” explained Robert-Michael de Groot, from the United States Geological Survey (USGS). Their team developed algorithms like these to process earthquake waves detected by sensors in the ground, a system they call ShakeAlert. 

De Groot’s team developed their algorithms based on vibration data collected from previous earthquakes (you can imagine it like those squiggly lines on a seismograph). For big quakes, which happen very rarely, the team created their own datasets to test whether the computer algorithm worked. 

It’s a process of trial and error, de Groot told The Spinoff. “It’s like tuning a piano, you put in inputs and see how do the outputs look like.” The USGS constantly tunes its algorithms as they get more data from real earthquakes. 

The USGS partnered with Google in the US. The USGS would detect and analyse the signals through ShakeAlert and Androids would send out messages about earthquakes to their users. Later in 2020, Android phones starting collecting information about earthquakes themselves, but they’re not yet sending out alerts. 

The rollout in New Zealand and Greece will be the first time the android system does both detecting and alerting.

What can’t the system do?

The system can’t detect earthquakes that originate offshore, at least until the earthquake waves hit land. This is because, of course, the sensors the system is using, our phones, are on land. 

That’s not an issue unique to the Android system. All the sensors are on land, including offshore islands. 

These earthquake early warning systems also aren’t great at sending a warning if there’s an earthquake starting really close to you. In those cases, the P-waves the system detects are too close timewise to the destructive S-waves to send a warning before shaking begins. 

“Part of it has to do with the time it takes for the earthquake to reach the surface but then there’s got to be that processing time of moving the information from the sensors to the processing centre and then for the alerts to go out,” said de Groot.

“That’s always going to be an issue no matter what. If you’re right on top of the earthquake, that’s just the way it goes.”

That means for quakes like the Christchurch one, which started less than 7km from the city, people probably wouldn’t get a warning before they personally felt the earthquake.  

It won’t excel, either, at detecting quakes in areas where there aren’t many people. The algorithms need to get enough data from enough phones to run properly. Google did not comment on exactly how much data is enough, but according to the company their algorithms have minimum thresholds needed for earthquake detection to be accurate. However if there aren’t many people or android users in an area, there might not be enough information to trigger an alert. 

There also remains some doubt about how good the system will be at analysing New Zealand’s own “signature” of earthquake signals. If an algorithm has been trained on particular data, say from the US, it might not be as good with different data from here. “The issue with Google is they’re rolling out the system without any understanding of the local tectonic context,” said Caroline Francois-Holden, a former GNS seismologist who is now earthquake consultant for SeismoCity.

So how useful is it really?

De Groot argues that even a few seconds warning is enough to help save damage to people, equipment and buildings. “Anything you can do to make something happen before things get really bad may mean the difference between a device being able to operate after the event is over,” he said, referring to securing important equipment in hospitals but adding the same idea applies for people.

But the alerts also give people a chance to mentally prepare for what’s about to happen. That’s a big deal when it comes to recovery after a big quake, according to social scientist Julia Becker from New Zealand’s Joint Centre for Disaster Research.

“It’s the idea that you might have a few seconds to think about what will happen before it does… We did a survey in Japan where they have an earthquake early warning system in place and a lot of people feel that it’s quite helpful to get yourself in a headspace to prepare,” she said. 

Will people use it?

It depends on whether people trust the system, according to Raj Prasanna, an expert in emergency management technology. “When it comes to technology, especially related to disaster situations, people have spent millions of dollars and these systems have failed because you haven’t got that engagement with the users. There’s a lot of trust issues and perception issues around these systems, although the systems are quite capable.”

Even if people trust the system, they might not drop, cover and hold when they get an alert, says Becker. One of the surveys her team did found that a vast majority of people in Japan felt positively about their earthquake early warning system but that few actually took protective action when they received an alert. It’s a similar situation to acceptance versus use of the Covid-19 tracer app.

“People may be embarrassed or they may have previous experience with earthquakes where it wasn’t that bad,” Becker said. She adds that’s why campaigns like “Drop, Cover, Hold” or the ShakeOut practice event are so vital to being prepared for quakes on top of any early detection system.

What about iPhone users?

The system available in New Zealand is an android exclusive, so iPhone users miss out. 

Other similar earthquake alert apps, such as MyShake, are available for Apple users overseas. Japan’s iPhone users have had an earthquake alert system built into their phones’ software since 2011.




The Spinoff is made possible by the generous support of the following organisations.
Please help us by supporting them.