The university has developed a new AI model that can predict earthquake outcomes months before a major earthquake occurs.

Content

A scientist from the University of Alaska Fairbanks has shown that the public can receive notifications days to months in advance of a major earthquake based on the identification of previous large-scale low-level tectonic activity. This analysis primarily focuses on two significant earthquakes in Alaska and California. The new study emphasizes that it is possible to predict major earthquakes several months in advance by using machine learning to detect early signs of seismic activity. However, the effectiveness and ethical implications of this predictive technology remain subjects of debate.

大學研究發展新的AI模型,可在大地震發生前數月預測地震結果

This work is led by Assistant Research Professor Társilo Girona of the Geophysical Institute at the University of Alaska.

Giron is a geophysicist and data scientist who studies the precursors of volcanic eruptions and earthquakes. Kyriaki Drymoni, a geologist at Ludwig Maximilian University of Munich in Germany, is a co-author of the study.

This machine learning-based detection method was published on August 28 in Nature Communications.

大學研究發展新的AI模型,可在大地震發生前數月預測地震結果

大學研究發展新的AI模型,可在大地震發生前數月預測地震結果

Gerona said: "Our paper shows that advanced statistical techniques, particularly machine learning, have the potential to identify precursors of large magnitude earthquakes by analyzing datasets in earthquake catalogs."

The author has developed a computer algorithm to search for data, looking for anomalous seismic activity. The algorithm is a set of computer instructions used to guide the program in interpreting data, learning from it, and making informed predictions or decisions. Case study: Anchorage and Ridgecrest earthquakes

They focused on studying two major earthquakes: the 7.1 magnitude Anchorage earthquake in July 2018 and the 6.4 to 7.1 magnitude earthquake sequence in Ridgecrest, California in 2019. They found that about 15% to 25% of the areas in south-central Alaska and Southern California experienced approximately three months of abnormally low magnitude regional seismic activity before the two studied earthquakes, with the turmoil before the major earthquakes primarily caused by seismic activity with magnitudes below 1.5.

The Anchorage earthquake occurred on November 30, 2018, at 8:29 AM, with the epicenter located about 10.5 miles north of the city. The earthquake caused extensive damage to some roads and highways, and some buildings were also damaged.

Giron and Drymoni used their data training program to discover that the probability of a major earthquake occurring within 30 days or less during the Anchorage earthquake suddenly rose to about 80% around three months before the earthquake on November 30. Just a few days before the earthquake, the probability rose to about 85%. They also had similar probability findings for the Ridgecrest earthquake sequence starting about 40 days before it occurred.

大學研究發展新的AI模型,可在大地震發生前數月預測地震結果

Girona and Drymoni proposed the geological reasons for low-magnitude precursor activity: a significant increase in pore fluid pressure within the fault. Pore fluid pressure refers to the pressure of fluids within the rock. If the pore fluid pressure is sufficient to overcome the frictional resistance between the rock masses on either side of the fault, then high pore fluid pressure could potentially lead to fault slip.

The increase in pore fluid pressure in the fault that causes a major earthquake will change the mechanical properties of the fault, leading to uneven changes in the regional stress field. Research suggests that these uneven changes control the abnormal, precursory low-magnitude earthquakes.

Machine learning is having a significant positive impact on earthquake research, Girona said: "Modern seismic networks generate huge datasets, which, if analyzed properly, can provide valuable insights into the precursors of seismic events. This is precisely where advances in machine learning and high-performance computing can play a transformative role, enabling researchers to identify meaningful patterns that may indicate an impending earthquake."

The author points out that their algorithm will be tested in near real-time to identify and address potential challenges in earthquake forecasting. This approach should not be adopted in new areas without training the algorithm based on the historical seismic conditions of the region. Producing reliable earthquake forecasts has "very important and often controversial aspects." Accurate forecasts have the potential to save lives and reduce economic losses by providing early warnings for timely evacuations and preparations. However, the inherent uncertainty of earthquake forecasting also raises significant ethical and practical issues. False alarms can lead to unnecessary panic, economic turmoil, and a loss of public trust, while forecasting errors can have catastrophic consequences.

Summary
Research from the University of Alaska Fairbanks indicates that the public could receive warnings of major earthquakes days to months in advance by identifying low-level seismic activity. The study, led by geophysicist Társilo Girona, focuses on significant earthquakes in Alaska and California, utilizing machine learning to detect early signs of seismic activity. Published in Nature Communications, the research highlights that advanced statistical techniques can identify precursors to large earthquakes by analyzing seismic catalogs. The study examined the 2018 Anchorage earthquake and the 2019 Ridgecrest earthquake sequence, finding that approximately 15% to 25% of the regions experienced abnormal low-magnitude seismic activity three months prior to the major quakes. The researchers noted a significant increase in the probability of a major earthquake occurring within 30 days, rising to about 85% just days before the events. They attribute these precursory activities to increased pore fluid pressure within faults, which can lead to fault slip. While machine learning offers transformative potential in earthquake prediction, the authors caution about the ethical implications and uncertainties involved, emphasizing the need for careful application and testing of their algorithms in new regions.