Chinese seismologists are developing artificial intelligence models that integrate multiple sources of monitoring data to assess earthquake risks across different time scales, a leading researcher said on Monday, cautioning that the technology cannot reliably predict when and where an earthquake will strike.
Meng Lingyuan, head of the earthquake forecasting department at the China Earthquake Networks Center, said at a media event organized by the Ministry of Emergency Management that researchers had built an AI-based forecasting model in an earthquake science experimental zone spanning parts of Sichuan and Yunnan provinces.
The model integrates earthquake catalogs, seismic waveforms and observations collected through different monitoring methods to generate probability-based assessments, Meng said.
"Artificial intelligence is a tool that helps us work, not a fortune teller," she said.
Accurate earthquake forecasting remains difficult because scientists must assess three variables before an event occurs: its timing, location and magnitude.
Geological structures and the quality of observation records also vary across regions, limiting how easily a model developed in one area can be applied elsewhere, she explained.
During China's 15th Five-Year Plan period (2026-30), researchers hope to test the models in more regions and across more scenarios, identify their capabilities and limitations, and improve physics-based forecasting methods, Meng said.
The work also extends to assessing risks after major earthquakes, when rescuers and residents need timely information about potential aftershocks.
Following a magnitude 6.8 earthquake in Dingri county, Xizang autonomous region, on Jan 7, 2025, Meng's team monitored seismic activity for more than 30 days, conducting intensified analyses and consultations roughly every two hours.
The team worked for more than 700 hours and produced over 100 special reports on seismic activity to support assessments of evolving risks, she said.
The experience also prompted adjustments to seismic assessment methods. Meng said related technological advances were subsequently applied following a magnitude 7.0 earthquake in Drachen, Xizang, this year.
Meng recalled meeting a resident after a 2021 earthquake in Yangbi, Yunnan, who asked whether it was safe to return home to collect belongings.
The encounter showed her that technical assessments must be translated into advice people can understand and act on.
Her priority, she said, is to make earthquake information available earlier, improve the precision of assessments and turn expert findings into practical safety guidance.
Despite advances in data analysis, Meng emphasized that researchers must remain clear about scientific uncertainty rather than promise exact predictions.
