New AI Model May Predict Major Earthquakes Months in Advance

Breakthrough AI Model Could Forecast Major Earthquakes Months Ahead

An earthquake is the shaking of the Earth's surface caused by a sudden release of energy in the Earth's crust. This release occurs due to the movement of tectonic plates, which are large sections of the Earth's crust that float on the semi-fluid mantle beneath. When these plates grind against each other or collide, stress builds up until it is released as seismic waves. These waves propagate through the Earth and cause the ground to shake, which can result in damage to structures, landscapes, and cause significant human and economic impacts. Earthquakes can vary in magnitude from minor tremors to major quakes with severe consequences.

Recent research highlights the potential of using machine learning to predict major earthquakes months before they occur. Conducted by Társilo Girona at the University of Alaska Fairbanks, the study suggests that early warnings of days to months might be possible by detecting low-level tectonic activity preceding significant quakes. This analysis, which focuses on the 2018 Anchorage and 2019 Ridgecrest earthquakes, utilized machine learning to identify abnormal seismic patterns across large areas.

The research, published in Nature Communications on August 28, shows that abnormal low-magnitude seismic activity, often below magnitude 1.5, can signal impending major earthquakes. Girona’s study analyzed data from these events and found that the probability of a major earthquake increased significantly in the months leading up to the quakes. For the Anchorage earthquake, the likelihood of a major event in the next 30 days rose to around 80% three months prior and 85% in the days before the quake. Similar patterns were observed for Ridgecrest.

Girona and co-author Kyriaki Drymoni of Ludwig-Maximilians-Universität Munich propose that increased pore fluid pressure in faults may drive these precursor activities. This pressure changes the mechanical properties of faults, leading to variations in regional stress fields and abnormal seismicity.

While promising, the effectiveness and ethical implications of such predictive technology remain under discussion.

Machine learning is significantly advancing earthquake research, according to Társilo Girona.

"Modern seismic networks generate vast amounts of data that, when analyzed correctly, can reveal valuable insights into earthquake precursors," Girona noted. "Advancements in machine learning and high-performance computing can transform this data into meaningful patterns that might indicate an impending earthquake."

However, the authors caution that their algorithm needs further testing in near-real-time scenarios to tackle challenges in earthquake forecasting. They stress that the algorithm should not be used in new regions without first training it with that area's historical seismic data.

Girona highlights the critical and controversial nature of earthquake forecasting. "Accurate forecasts could save lives and reduce economic losses by providing early warnings for timely evacuations and preparations," he said. "Yet, the inherent uncertainty in forecasting also poses ethical and practical issues. False alarms can cause unnecessary panic and economic disruption, while missed predictions can have severe consequences."

Reference: “Abnormal low-magnitude seismicity preceding large-magnitude earthquakes” by Társilo Girona and Kyriaki Drymoni, August 28, 2024, Nature Communications.