AI Reveals Ten Times More Earthquakes in Yellowstone Than Previously Recorded

Before the rise of artificial intelligence, seismologists relied on manual analysis of seismic data to detect earthquakes, a process that was slow, expensive, and limited in scope. Now, machine learning has revolutionized the field. Researchers have used AI to reanalyze historical seismic data from the Yellowstone caldera between 2008 and 2022, identifying and assigning magnitudes to ten times as many earthquakes as had previously been recorded. Bing Li, an engineer at Western University, explained in a university statement that manual methods would be impractical for such a task. "If we had to do it old school with someone manually clicking through all this data looking for earthquakes, you couldn’t do it. It’s not scalable," he said. In a study published earlier this month in the journal Science Advances, Li and his team described how they used advanced deep learning algorithms and a detailed three-dimensional velocity model to create a high-resolution earthquake catalog for the Yellowstone caldera spanning 15 years. This new catalog includes 86,276 earthquakes, significantly enhancing seismologists’ understanding of the region’s volcanic and seismic activity. Calderas, such as Yellowstone, are large depressions formed when a volcano erupts and collapses into its emptied magma chamber. These areas can generate earthquakes due to various volcanic processes, including pressure changes and fluid movement. Most of Yellowstone’s earthquakes are classified as brittle-failure events, where stress in the Earth's crust causes rocks to fracture. The study revealed that more than half of these earthquakes occur in swarms—sequences of quakes that don’t follow the typical mainshock-aftershock pattern. Li, an expert in fluid-induced earthquakes and rock mechanics, noted that while each volcano has its own characteristics, the findings from Yellowstone could offer broader insights. "The hope is that these insights can be applied elsewhere," he said. "By understanding patterns of seismicity, like earthquake swarms, we can improve safety measures, better inform the public about potential risks, and even guide geothermal energy development away from hazardous areas in regions with promising heat flow." The research also found that earthquake swarms in Yellowstone tend to occur along less developed fault structures—faults that have experienced fewer "slips" compared to more mature ones. While scientists anticipate more aftershocks on these immature faults, they have lacked a systematic understanding of how one quake triggers another within a swarm. "We can only indirectly measure space and time between events," Li explained. "But now, with this more comprehensive catalog, we can apply statistical methods to quantify and discover new swarms that we haven’t seen before, study them, and learn more about their behavior." The study highlights how even the world’s first national park continues to hold mysteries for researchers. Just recently, a new hole appeared in Yellowstone’s Norris Geyser Basin, further underscoring the dynamic and unpredictable nature of the region’s geothermal activity. The integration of AI into seismic analysis is not only transforming how scientists study earthquakes but also revealing hidden patterns that could shape future research and safety protocols in volcanic regions.