In a study led by University of Alaska Fairbanks scientist Társilo Girona, it was discovered that it is possible to predict major earthquakes with days or even months of warning by identifying prior low-level tectonic unrest over large areas. This research, published in Nature Communications, utilized machine learning techniques to analyze seismic activity data leading up to two major earthquakes in Alaska and California. The team found that abnormal low-magnitude seismicity occurred across 15% to 25% of Southcentral Alaska and Southern California in the months before the earthquakes.
The study focused on two major earthquakes: the 2018 magnitude 7.1 Anchorage earthquake and the 2019 Ridgecrest, California, earthquake sequence. By analyzing seismic activity with magnitudes below 1.5, the researchers were able to identify patterns of unrest preceding these major earthquakes. They found that the probability of a major earthquake occurring increased significantly three months before the Anchorage earthquake and about 40 days prior to the onset of the Ridgecrest earthquake sequence. The authors attribute this precursor activity to a significant increase in pore fluid pressure within faults, altering the faults’ mechanical properties and leading to abnormal seismicity.
According to Geologist Kyriaki Drymoni, the uneven variations in the regional stress field caused by increased pore fluid pressure in faults are responsible for the abnormal, low-magnitude seismicity observed before major earthquakes. This research highlights the potential for machine learning to revolutionize earthquake research by analyzing large seismic datasets and identifying meaningful patterns that signal impending seismic events. The authors emphasize the importance of testing their algorithm in near-real-time situations and caution against using it in new regions without training it with historical seismicity data from those areas.
While accurate earthquake forecasting has the potential to save lives and reduce economic losses by providing early warnings for evacuations and preparations, it also raises ethical and practical concerns. False alarms can lead to unnecessary panic and economic disruption, while missed predictions can have catastrophic consequences. Despite these challenges, the researchers believe that advancements in machine learning and high-performance computing can play a transformative role in improving earthquake forecasting and preparedness efforts. By understanding the precursory activity associated with major earthquakes, scientists hope to provide valuable insights that could lead to more effective early warning systems.