Bureau Team Wins International Earthquake Forecasting Competition

March 30, 2023
Y. Chen

In artificial intelligence (AI) and deep learning, progress is recognized through competitions. The Bureau of Economic Geology’s Yangkang Chen recently led a team, including Alexandros Savvaidis, Sergey Fomel, Dino Huang, and researchers from other institutions, to win first place among 600 international teams in the 2022 international AETA Earthquake Prediction AI Algorithm Competition hosted by Peking University Shenzhen Graduate School in China. Chen’s same team also landed second place in this competition in 2021. The research considered for the award is thought to be groundbreaking. Seismologists have long envisioned the ability to predict earthquakes, but it is only now that modern AI-based data-analysis tools allow researchers to approach a solution to the problem.

AETA defines the competition’s goals as “mining the correlation between precursor observation data and earthquakes’ three elements [occurrence, magnitude, and epicenter] through innovative algorithms, discovering abnormal signals and features related to impending earthquakes, and building earthquake prediction models based on historical observation data and earthquake catalogues, in the hope of promoting the solution of scientific problems of earthquake prediction and forecast.”

AI techniques open the door to detecting earthquake precursory signals in an unprecedented way. Previously, works that leveraged AI techniques for precursory signal detection or earthquake prediction were all retrospective studies, meaning that investigations were based on historical data sets and model tuning could be iterated until the best-fitting performance was found. In this study, the team conducted AI-based real-time earthquake prediction in both 2021 and 2022.

The team developed a data-driven model for predicting natural, destructive earthquakes (events equal to or greater than Magnitude 3.5) in a specific geographic area based on AI and big data analysis. It applied the proposed data-driven model to a real-time earthquake prediction competition lasting two years and obtained encouraging prediction accuracy (70 percent) in terms of earthquake occurrence, location, and magnitude. The accuracy ranked second in the first year of the competition (2021) and first in its second year (2022). The high score for consecutive two-year predictions among all participating teams indicates the effectiveness of Chen’s team’s AI model.

This new study sheds light on completely data-driven strategies for detecting earthquake precursors. Combining the findings from the study with physics-based models may have a profound impact on the wide applications of future earthquake prediction practices that could save hundreds of human lives worldwide. This work is now undergoing peer review.

Please join all of us at the Bureau of Economic Geology in congratulating Yangkang Chen and his talented team!


University of Texas at Austin

University of Texas

© 2021 Bureau of Economic Geology | Web Privacy Policy | Web Accessibility Policy