Science and Technology News
China University of Science and Technology uses machine learning to reveal the rupture patterns of large earthquakes around the world
2022/4/24
Viewed：
Source: China University of Science and Technology News
Li Zefeng, a researcher from the University of Science and Technology of China, used machine learning to summarize the focal time function characteristics of more than 3000 earthquakes with magnitudes of 5.5 and above in the world, and to show the similarity and diversity of global earthquake rupture processes in a panorama, deepening the understanding of earthquake energy release patterns, and having implications for earthquake early warning. The results were published in Geophysical, an internationally renowned geoscience journal, as "A generic model of global earthquake rupture characteristics revealed by machine learning. Research Letters.
Earthquake is one of the important natural disasters faced by human society. In the past 20 years, the world's major earthquakes have caused nearly 1 million casualties and countless economic losses. There are many kinds of earthquake rupture processes, and the objective measurement of their similarities and differences is helpful to understand the seismic physical processes and the early prediction of earthquake magnitude. However, previous studies either superimposed the average rupture process of multiple earthquakes, unable to measure the range of global seismic differences, or based on the statistics of certain rupture characteristics, unable to achieve a systematic comparison of the entire rupture process.
Figure 1: Global seismic source time function distribution in variational autoencoder implicit space (a) and reconstructed global seismic rupture mode manifold (b).
Researcher Li Zefeng used the Variational Autoencoder in deep learning to compress the source time function of more than 3000 medium and large earthquakes in the world in two-dimensional space and reconstruct the model, and showed the global seismic moment release mode and quantity distribution (Figure 1). It is found that the major earthquakes are mainly simple ruptures, while the complex ruptures are rare, and the distribution laws of two special types of earthquakes are revealed, namely, the escape mode in which the energy release is concentrated in the late ruptures and the complex earthquakes in which the energy release is divided into multiple times. It is found that the energy release mode of the major earthquakes has a weak magnitude dependence, which provides useful enlightenment for the predictability of the final magnitude in earthquake early warning. This study follows the development of the focal time function clustering method jointly researched by Li Zefeng's team and Harvard University in 2021, and is also one of the series of research results of the team's recent efforts to apply artificial intelligence to scientific discovery (AI for Science).
The thesis links: https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1029/2021GL096464
(School of Earth and Space Sciences, Department of Scientific Research)
Li Zefeng, a researcher from the University of Science and Technology of China, used machine learning to summarize the focal time function characteristics of more than 3000 earthquakes with magnitudes of 5.5 and above in the world, and to show the similarity and diversity of global earthquake rupture processes in a panorama, deepening the understanding of earthquake energy release patterns, and having implications for earthquake early warning. The results were published in Geophysical, an internationally renowned geoscience journal, as "A generic model of global earthquake rupture characteristics revealed by machine learning. Research Letters.
Earthquake is one of the important natural disasters faced by human society. In the past 20 years, the world's major earthquakes have caused nearly 1 million casualties and countless economic losses. There are many kinds of earthquake rupture processes, and the objective measurement of their similarities and differences is helpful to understand the seismic physical processes and the early prediction of earthquake magnitude. However, previous studies either superimposed the average rupture process of multiple earthquakes, unable to measure the range of global seismic differences, or based on the statistics of certain rupture characteristics, unable to achieve a systematic comparison of the entire rupture process.
Figure 1: Global seismic source time function distribution in variational autoencoder implicit space (a) and reconstructed global seismic rupture mode manifold (b).
Researcher Li Zefeng used the Variational Autoencoder in deep learning to compress the source time function of more than 3000 medium and large earthquakes in the world in two-dimensional space and reconstruct the model, and showed the global seismic moment release mode and quantity distribution (Figure 1). It is found that the major earthquakes are mainly simple ruptures, while the complex ruptures are rare, and the distribution laws of two special types of earthquakes are revealed, namely, the escape mode in which the energy release is concentrated in the late ruptures and the complex earthquakes in which the energy release is divided into multiple times. It is found that the energy release mode of the major earthquakes has a weak magnitude dependence, which provides useful enlightenment for the predictability of the final magnitude in earthquake early warning. This study follows the development of the focal time function clustering method jointly researched by Li Zefeng's team and Harvard University in 2021, and is also one of the series of research results of the team's recent efforts to apply artificial intelligence to scientific discovery (AI for Science).
The thesis links: https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1029/2021GL096464
(School of Earth and Space Sciences, Department of Scientific Research)
