New artificial intelligence has appeared, which can predict not only large earthquakes, but also small shifts of the earth’s crust.
Strong earthquakes are hard to miss, but rare. Meanwhile, small shifts of the crust pass imperceptibly, which occur constantly. They occur on the same faults as stronger earthquakes, and with the participation of the same mechanisms. These “microquakes” are fertile ground for research on the development of changes in the earth’s crust.
If we improve our ability to detect and locate these very small earthquakes, we can get a clearer idea of how earthquakes interact or propagate along a rift, how they occur, and even how they end.
Gregory Beroza, a Stanford eophysicist and one of the authors of the article
Mousavi began working on technology to automate earthquake detection. He previously studied daily gathers in Memphis. A few years later, after joining Beroza’s lab at Stanford, he began to think about how to solve the problem of detecting cortical vibrations using machine learning.
In their paper, the authors explain that they have developed a new model for detecting very small earthquakes with weak signals. They call it an earthquake transformer.
According to Mousavi, the model is built on PhaseNet and CRED and “includes the algorithms that scientists make manually.” In particular, Earthquake Transformer mimics the way human analysts look at a set of movements as a whole and then hone in on a small piece of interest.
This technology could allow analysts to focus on extracting information from a more complete catalog of earthquakes. The new algorithm saves researcher’s time, which they are going to spend on analyzing the nature of the earthquake.