How Reliable Is Disaster Prediction Technology?

Google has admitted that its early warning system failed to do what it was built for during the 2023 earthquake in Turkey. Instead of sending out full screen alerts to warn people of strong shaking, it mostly kept silent.

Around ten million people were within reach of the strongest waves of the 7.8 magnitude quake. These people should have received the most urgent “Take Action” alerts, designed to give up to 35 seconds’ notice. Instead, Google told the BBC that just 469 of those alerts were sent out. Half a million people were sent the weaker “Be Aware” version, which does not trigger a loud noise or take over the screen.

The difference is not small. “Be Aware” messages are only meant for light shaking. “Take Action” alerts are supposed to tell people when they need to move away from danger quickly. The quieter alert would not have stood out on a locked phone. The more serious one might have saved lives.

In 2023, Google had told the BBC that the system had worked well. At that point, some people had already complained they got no warning at all. But only now has it become clear how few of the stronger alerts were actually sent during one of the most violent earthquakes in recent memory.

Google runs the alert system itself. It is active in nearly 100 countries and was meant to help places without their own warning service. But this case raises questions about how dependable that help really is when people are most at risk.

 

How Does Android Earthquake Alerts Work Again?

 

As you can remember, alert system works through sensors already built into Android phones. The same tool that flips a screen when the phone is turned sideways can pick up early signs of ground movement. When a still phone detects the fast-moving P-wave of a quake, it sends a signal to Google’s server.

The system checks thousands of phones in the area to confirm the movement. If it sees a real earthquake happening, it calculates its strength and location. From there, it sends out alerts based on how strong the shaking might be in a specific place.

 

 

There are two types of alerts. The less urgent one is called “Be Aware” and is used for weaker shaking. It shows up quietly. The more serious one is called “Take Action.” This alert flashes on screen and makes a loud noise, telling people to protect themselves. The phone needs to have location settings and either Wi-Fi or mobile data turned on to get these messages.

Google’s own engineers say the system has picked up more than 18,000 earthquakes since launch. Over 2000 of those led to alerts being sent out. They also say the system has helped spread early warnings to people in places like New Zealand, Greece, Nepal and the Philippines.

 

Is This A Reliable Tool?

 

Google tracks user feedback through surveys in the app. Out of over 1.5 million responses, 85 percent of people said they found the alerts “very helpful.” Even people who didn’t feel the quake still said they liked being warned. That suggests people trust the idea.

But the results on the ground say something else. If fewer than 500 people were given the serious alert in Turkey’s deadliest quake in years, that cannot be brushed aside. The quake was well within the system’s target zone. There was time to act. The system just didn’t give people that chance.

Google engineers also say they’ve made the alerts better over time. They track how close their early estimates are to the real strength of a quake. Their error rate has halved over three years, dropping from 0.50 to 0.25. On paper, that puts them close to what traditional seismic networks achieve.

Even so, every alert that goes out (or fails to) depends on those first few seconds. One wrong estimate can lead to a wrong message or silence. And in Turkey, it was silence for nearly everyone.

Until systems like this can be trusted in the moments that matter most, the trust built through surveys may not hold when it really counts.