China has developed its first large artificial intelligence model to mine troves of data to help predict earthquakes.

The model, called “DiTing”, is based on one of the biggest data sets of its kind in the world and can be used to spot earthquake signals, monitor seismic activity and support rapid responses to earthquakes, according to its developers.

“DiTing relies on massive amounts of data from the China Earthquake Observation Network and advanced artificial intelligence technology and has significantly improved the accuracy and speed of seismic signal recognition,” Chen Shi, deputy director of the China Earthquake Administration’s Institute of Geophysics, said at the model’s launch in the southwestern city of Chengdu on Sunday.

The launch of the model on Sunday coincided with the 48th anniversary of the devastating earthquake in Tangshan in northern China, which killed at least 300,000 and levelled the city.

The project is a joint effort by the National Supercomputing Centre in Chengdu, the institute and Tsinghua University.

It takes its name from a divine beast in Chinese Buddhism that can pick up signals from across the universe.

The supercomputing centre said it was the first time seismic information had been amassed on such a scale to train artificial intelligence in China, overcoming the limits of conventional or smaller models.

In all, the model can already assess 100 million parameters - or variables - to make predictions or decisions - and that total is expected to reach a billion in August, state-run Science and Technology Daily quoted the centre’s chairman, Guo Li, as saying.

The more parameters a model has, the more adept it is in capturing details in a data set and the better it can improve the overall performance.

The model has been trained on seven years worth of seismic reports from the China Earthquake Networks Centre, according to a paper from the research team published in Earthquake Science, the flagship journal of the Seismological Society of China.

The team said the information could be used as a foundation for a range of earthquake-related research, setting a “benchmark for machine learning model development and data-driven seismological research” in areas such as earthquake detection, earthquake magnitude prediction, and ground-motion prediction.

The model could also be used in fields such as mine seismic monitoring, shale gas exploitation, urban underground space structure detection and underwater earthquake monitoring.

“More than 95% of oil and gas fields discovered in the world mainly rely on seismic exploration. DiTing can deduce whether the underground contains oil and gas by learning the waveform characteristics of oil storage areas,” Wang Jianbo, deputy director of the supercomputing centre, said.

 - SCMP