Photo credit: www.sciencedaily.com

Understanding the Earthquakes of American Samoa: A Innovative Approach

From late July to October 2022, the inhabitants of the Manu’a Islands in American Samoa experienced frequent tremors, stirring worries about potential volcanic activity or a tsunami risk.

Despite the ongoing seismic disturbances, an earthquake catalog from this region revealed no seismic activity due to the absence of a monitoring network, which hampered scientists’ efforts in identifying the underlying cause of the tremors.

Faced with the pressing need for information, Clara Yoon from the U.S. Geological Survey and her research team devised an alternative method to ascertain the nature of the seismic events. They utilized machine learning, paired with a method known as template matching, analyzing data collected from a remote seismic sensor located 250 kilometers from the affected islands.

In an article published in The Seismic Record, Yoon and her colleagues detail how they successfully monitored the earthquake swarm using the data from a single station, in conjunction with first-hand accounts from local residents, until permanent seismic stations were established in American Samoa in late August and early September 2022.

The non-eruptive volcanic earthquake swarm was first detected about 15 kilometers off the coast of Ta’ū Island in July 2022, coinciding with the geological activity associated with the Pacific tectonic plate passing over a volcanic hotspot in the South Pacific.

Initially, the only insights into the phenomenon came from residents’ experiences of the shaking, which occurred several times daily, lasting only a few seconds each time.

As Yoon explained, “American Samoa had no instrumental geophysical monitoring at the time these earthquakes began. Consequently, even basic information regarding the source of the tremors—key for emergency planning and public safety—was unavailable.”

To fill this information gap, the researchers accessed data from a remote seismic station situated on Upolu, Samoa, which is part of the Global Seismographic Network, allowing for near-real-time data downloads through the EarthScope data center.

However, the seismic signals from the American Samoa swarm were challenging to detect at the far-off station. To overcome this, Yoon and her team employed a deep-learning algorithm known as EQTransformer along with template matching techniques to discern small earthquakes amidst the disruptive seismic noise.

“EQTransformer identified numerous earthquakes with locations aligned with eastern American Samoa, the largest coinciding with local felt reports,” Yoon noted. “These accounts from residents, submitted to the National Weather Service, were critical in confirming that the earthquakes we detected with EQTransformer were indeed those experienced by the community.”

The research team was able to compile a new earthquake catalog for this series of events, enabling them to analyze both the initiation and peak activity levels of the swarm. Further investigations with affordable Raspberry Shake sensors deployed in August 2022 quickly pinpointed the swarm’s epicenter.

Although the swarm concluded by October 2022 without leading to an eruption, researchers concluded that it was likely associated with the movement of volcanic magma beneath the earth’s surface.

Yoon highlighted that their single-station approach could offer valuable insights in other under-monitored regions worldwide, where seismic hazards remain ill-defined, particularly in offshore areas vulnerable to tsunamis or tectonic plate-related earthquakes.

She emphasized that the largest earthquake recorded in the swarm, measuring 4.5 in magnitude, would likely have eluded detection by global seismic networks.

“Had local residents not reported the frequent shaking, this earthquake swarm in American Samoa might have completely gone unnoticed,” Yoon reflected. “There are numerous unidentified seismic sources and phenomena awaiting discovery, potentially unveiled through extensive applications of advanced machine learning techniques in seismology.”

Source
www.sciencedaily.com