Precursory seismicity in earthquake prediction

A groundbreaking study by Társilo Girona, a research assistant professor at the University of Alaska Fairbanks (UAF) Geophysical Institute, suggests that advance notice of major earthquakes could be possible by detecting precursory seismicity—low-level tectonic activity occurring days to months before significant quakes. This study, published in Nature Communications and co-authored by geologist Kyriaki Drymoni of Ludwig-Maximilians-Universität in Munich, used machine learning to identify patterns in the buildup of seismic activity. Their analysis focused on two notable earthquakes: the 2018 magnitude 7.1 Anchorage earthquake in Alaska and the 2019 Ridgecrest earthquake sequence in California.

The researchers developed a computer algorithm to analyze seismic datasets from earthquake catalogs. Their algorithm identified abnormal seismic patterns, revealing that regions affected by the Anchorage and Ridgecrest quakes experienced about three months of unusual low-magnitude seismicity before the main events. In these areas, roughly 15% to 25% of the regions had experienced seismic activity with magnitudes under 1.5, indicating that precursory seismicity may be a key indicator for predicting larger quakes.

The Anchorage earthquake, which struck on November 30, 2018, caused significant damage to infrastructure, highlighting the need for effective early warning systems. Girona and Drymoni’s algorithm showed that the likelihood of a major earthquake occurring within 30 days increased sharply to about 80% three months prior to the Anchorage quake, and reached 85% just days before the event. A similar pattern was observed for the Ridgecrest earthquake sequence, where precursory seismicity was detected approximately 40 days before the earthquakes.

The scientists propose that increased pore fluid pressure in faults is a likely geologic cause of the low-magnitude seismicity preceding major earthquakes. Pore fluid pressure, which refers to the pressure of fluids within rocks, can influence fault slip if the pressure exceeds the frictional resistance between fault blocks. This leads to uneven variations in the regional stress field, causing abnormal seismic activity. These findings suggest that understanding and detecting these subtle shifts in precursory seismicity could improve earthquake forecasting.

Machine learning played a critical role in this research by allowing the analysis of vast seismic datasets, which traditional methods might overlook. Girona noted that seismic networks produce enormous amounts of data, and advanced computing techniques such as machine learning can help identify patterns that may signal an impending earthquake.

The researchers acknowledge that while their findings are promising, challenges remain in earthquake forecasting. The algorithm will need further testing in real-time scenarios, and it may not be applicable to regions without historical seismic data for training. Moreover, forecasting earthquakes poses ethical and practical challenges. False alarms could cause unnecessary panic or economic disruptions, while missed predictions could lead to devastating consequences. Nonetheless, Girona emphasizes the potential for accurate earthquake forecasting to save lives and reduce damage by providing early warnings and enabling timely evacuations.

By incorporating precursory seismicity into earthquake prediction models, this research represents a significant step forward in seismic science, with the potential to transform how we understand and respond to earthquake risks.

https://scitechdaily.com/new-ai-model-could-predict-major-earthquakes-months-before-they-happen