A revolutionary study recently published in the journal Nature Communications by the University of Alaska Fairbanks (UAF) proposes an innovative approach to earthquake prediction using machine learning. The research, led by Dr. Tarsilo Girona from UAF and Dr. Kyriaki Drimoni from Ludwig-Maximilian University in Munich, is based on data collected from two significant earthquakes: the 7.1 magnitude events that struck Anchorage in 2018 and California in 2019.
The research team observed unusual seismic activity patterns in the earthquake epicenters, which they believe may have preceded the seismic events. Using machine learning models and data analysis from seismic sectors, the researchers identified abnormal activity, primarily small earthquakes (with magnitudes of 1.5 or less), which occurred three months prior to each major quake.
In the case of the Anchorage earthquake, the algorithm developed by the researchers identified an 80% likelihood of a major earthquake occurring about a month before the event, with the likelihood rising to 85% in the week before the quake. These remarkable findings suggest that earthquake prediction might be possible, with the anomalous seismic activity linked, according to the researchers, to fluid pressure within the region’s geological structures. Changes in this pressure can alter the mechanical properties of geological plates, leading to the observed unusual seismic phenomena.
Despite the research team’s cautious optimism, they acknowledge that the method still requires further testing on various historical data before it can be practically applied. They emphasize that to avoid panic or economic damage, clear guidelines must be developed between decision-makers and the scientific community for responsible use of this prediction system.
