Scientists at the University of Alaska Fairbanks have developed a new method that can accurately forecast earthquakes months in advance.
Led by Assistant Research Professor Társilo Girona from the UAF Geophysical Institute, the researchers analyzed two significant earthquakes in Alaska and California: the 7.1 magnitude earthquake in Anchorage in 2018 and the series of earthquakes in Ridgecrest, California in 2019, which ranged from magnitudes of 6.4 to 7.1.
Earthquake in Alaska (USA) November 2018. (Photo: Reuters).
The scientists discovered unusual low-intensity seismic activity. They noted that approximately three months prior to each earthquake studied, there was unusual low-intensity seismic activity occurring across about 15% to 25% of the South Central Alaska and Southern California regions.
The research indicates that the precursory instability before major earthquakes is primarily recorded through seismic activity with magnitudes below 1.5.
“Our research demonstrates that advanced statistical techniques, particularly machine learning, have the potential to identify foreshock signals of major earthquakes by analyzing data from earthquake catalogs,” Girona stated.
The authors developed an algorithm, a set of computer instructions that teach the program to interpret data and make predictions to detect unusual seismic activity.
Using the program trained with their data, Girona and co-author Kyriaki Drymoni found that for the Anchorage earthquake, the likelihood of a major earthquake occurring within 30 days or less increased sharply to about 80% roughly three months before the November 30 earthquake.
This probability rose to approximately 85% just a few days before it occurred. They also found similar probabilities concerning the Ridgecrest earthquake series starting about 40 days before the sequence began.
The researchers propose a geological cause for the low-intensity foreshock activity, citing a significant increase in fluid pressure within the faults.
“The increase in fluid pressure in the faults leads to major earthquakes that alter the mechanical properties of the faults, resulting in uneven changes in the regional stress field,” Drymoni explained.
Girona noted that modern seismic networks generate massive datasets, which, if analyzed correctly, could provide valuable insights into precursory signals of seismic events.
“This is where advancements in machine learning and high-performance computing can play a transformative role, allowing researchers to identify meaningful patterns that may signal an impending earthquake,” Girona commented.
The algorithm developed by the researchers will soon be tested in near-real-time scenarios to address potential challenges in earthquake forecasting.
