Can AI help us predict earthquakes?

Kyoto, Japan -- Predicting earthquakes has long been an unattainable fantasy. Factors like odd animal behaviors that have historically been thought to forebode earthquakes are not supported by empirical evidence. As these factors often occur independently of earthquakes and vice versa, seismologists believe that earthquakes occur with little or no warning. At least, that's how it appears from the surface.

Earthquake-generating zones lie deep within the Earth's crust and thus cannot be directly observed, but scientists have long proposed that faults may undergo a precursory phase before an earthquake during which micro-fracturing and slow slip occur. Yet, despite their obvious potential, exactly how these processes could enable prediction of a main shock remains unclear. Furthermore, observational studies have suggested that small and large earthquakes appear indistinguishable during the beginning of their rupture, raising doubts about the usefulness of short-term precursors.

These difficulties have prompted interest in the use of machine learning to search for potentially predictive fault signals. Machine learning models have demonstrated an ability to predict stick-slip laboratory earthquakes  in small, centimeter-scale experiments, but this approach has not yet been applied to larger, more complex systems that more closely mimic natural faults.

This motivated a team of researchers at Kyoto University to take this approach to the next level. They tested machine learning-based prediction on meter-scale laboratory earthquakes, where the physical processes and timescales more closely mirror those operating in the Earth's crust.

"We applied an advanced machine learning technique to data collected from a meter-scale rock-friction experiment that generates stick-slip laboratory earthquakes along with numerous acoustic emission events, which are tiny foreshock signals produced as the fault nears failure," says first author Reiju Norisugi.

The team's analysis revealed that machine learning models trained on the foreshock data alone can accurately detect subtle signals that emerge just before the onset of these laboratory earthquakes. To better understand how machine learning could achieve such accuracy, the team compared its performance with physics-based numerical simulations that reproduce the experimental data. They discovered that the key predictive factor is the evolution of shear stress on creeping -- or slowly slipping -- regions of the fault.

"These localized stress changes provide more diagnostic information than the average stress across the fault, which underscores the importance of monitoring spatially heterogeneous fault-slip behavior," says team leader Yoshihiro Kaneko.

By demonstrating that machine learning can detect and interpret subtle physical changes that precede rupture in realistic, large-scale laboratory fault systems, this study helps to bridge the gap between laboratory research and natural fault systems. Moreover, the team has established a physics-based framework for understanding the final stages of fault loading, representing an important step toward short-term earthquake forecasting in nature.

###

The paper "Machine learning predicts meter-scale laboratory earthquakes" appeared on 30 October 2025 in Nature Communications, with doi: 10.1038/s41467-025-64542-4

About Kyoto University

Kyoto University is one of Japan and Asia's premier research institutions, founded in 1897 and responsible for producing numerous Nobel laureates and winners of other prestigious international prizes. A broad curriculum across the arts and sciences at undergraduate and graduate levels complements several research centers, facilities, and offices around Japan and the world. For more information, please see: http://www.kyoto-u.ac.jp/en