AI Reshaping the Future of Earthquake Forecasting

When a series of earthquakes shook the Greek island of Santorini this year, sending residents and tourists scrambling, seismologist Margarita Segou turned to QuakeFlow—an AI-powered algorithm—to analyse the unfolding crisis.

Developed by Stanford University, QuakeFlow leverages machine learning and cloud computing to detect and process seismic activity with greater accuracy and speed than traditional methods, said Segou, who works for the British Geological Survey.

By harnessing AI, Segou identified 1,500 small tremors beginning in December 2024—subtle warnings often missed by conventional techniques—before seismic activity surged into a full-blown crisis on 26 January, culminating on 2 February.

Segou explained:

"When looking at large events (in Santorini), researchers saw a repeating pattern – seismicity came in pulses, starting with a magnitude-four quake, followed by a five, followed again by a four until the system relaxed."

She added:

"We are filling in the gaps, seeing the connection between the magnitude fours."

The devastating earthquake in Southeast Asia, centered in Myanmar, serves as a stark reminder of the challenges in predicting the timing, location, and magnitude of seismic events.

While AI has already transformed storm and flood forecasting, seismologists are now using it to uncover previously undetected earthquakes.

This growing trove of data not only enhances hazard assessment but also holds the potential to save lives.

Christopher Johnson, a scientist at the Los Alamos National Laboratory, told the Thomson Reuters Foundation:

"AI has revolutionised our ability to detect more and more small earthquakes that typically would fall below the signal to (background) noise level. Even though we have sensors in certain areas ... you could always bet that there were events that were missed."

Forecasting a Specific Earthquake is a Challenge

Earthquakes occur when stress builds up and is suddenly released within the Earth's crust, but the intensity of this stress doesn’t always dictate the magnitude of the quake.

Despite decades of research, scientists have yet to identify a definitive precursor—an event signalling an impending earthquake.

Instead, both seismologists and communities rely on historical patterns to assess earthquake risk in specific regions.

Johnson said:

"We have hazard estimates on the order of 30 years. So if you live next to a fault, you know there's a chance of an earthquake in your lifetime. If you have an earthquake anywhere, there is an increased probability of another earthquake occurring within that vicinity within a short amount of time that could rupture into something larger."

In New Mexico, Johnson’s team is leveraging Meta’s Wav2Vec-2.0, an AI model originally designed for speech recognition, to detect earthquakes with unprecedented accuracy.

By analysing seismic waveforms—patterns of vibrations over time—the AI outperforms traditional methods in tracking real-time fault movements.

He noted:

"You want to know when, where and how big. Trying to solve all of those at once is not where we are."

Yet, pinpointing the exact time and location of an earthquake remains the ultimate challenge in seismology.

This pursuit is becoming even more urgent as climate change may be influencing global seismic activity.

Research from the GFZ German Research Centre for Geosciences suggests that rising sea levels exert additional pressure on tectonic faults, potentially accelerating their seismic cycles and increasing the frequency of earthquakes worldwide.

Ample and Necessary Resources Needed for Earthquake Forecasting

AI is revolutionising earthquake detection, but its effectiveness still depends on high-quality seismic data—something not all countries can provide.

Wealthier nations like the US and China have the infrastructure to support dense seismometer networks, while earthquake-prone regions such as the Philippines and Nepal often lack the necessary resources, said Johnson.

To bridge this gap, researchers are turning to alternative solutions, including accelerometers in smartphones and tablets.

Google, for example, has integrated earthquake warnings into its Android operating system, launching in California in 2020 and expanding to India in 2023.

Now a global feature, the system analyses data from millions of smartphones, detecting tremors in real time and sending alerts within seconds—giving users crucial moments to prepare before impact.

Beyond detection, AI-driven technology is enhancing disaster response.

In India, researchers developed the Uttarakhand State Earthquake Early Warning System, an app-based platform that issues alerts and provides location data to help coordinate rescue efforts.

Segou stated:

"A second is an eternity in an earthquake."

Early warnings can minimise damage and save lives by allowing high-speed trains to slow down or critical medical procedures to be paused before a quake strikes.

Looking ahead, machine learning could play an even greater role in seismic risk management.

By analysing vast amounts of data, AI may soon help governments refine hazard assessments and implement more effective safety measures.

She pointed out:

"We shouldn't always be in panic mode.”

While earthquakes remain unpredictable, AI offers a promising path to reducing their devastation.

Segou concluded:

"We are living a revolution in understanding the Earth."