AI Machine Learning Advances Search for Earthquake Precursors

Earthquake forecasting remains a major challenge for disaster preparedness and insurance worldwide. In 2023, the European Research Council awarded €1.5 million to a forecasting project led by the German Research Centre for Geosciences. The project uses big data and artificial intelligence to analyze complex movements in the Earth's crust, with the aim of identifying anomalous physical signals before a mainshock and helping avert severe loss of life and property.

The German Research Centre for Geosciences published its latest findings in Nature Communications on July 13, 2026. Using unsupervised machine learning to analyze historical earthquake catalogs, the team identified abnormal concentrations of stress along faults several weeks before the 2023 Türkiye earthquake and the 2014 Chile earthquake. The pattern did not appear before the 2024 Noto Peninsula earthquake in Japan, however, underscoring the inconsistency of potential precursors and the formidable obstacles to accurate early warnings.