← Back to Sessions

Distant Seismic Monitoring of a Volcanic Earthquake Swarm Near the Manuʻa Islands, American Samoa, with Deep-learning and Template-matching Event Detection

Description: 

From July to October 2022, an earthquake swarm occurred near the Manuʻa Islands in eastern American Samoa. The swarm started in late July, with peak activity in August, before decreasing in late September and has fortunately been quiet since. Earthquake information was limited to felt shaking reports from local residents sent to the National Weather Service Pago Pago office, because of a lack of seismometers in the region when the swarm started. Residents were concerned about the unusual frequent shaking, making it essential to understand the source of these earthquakes in rapid response mode.

We show how modern seismological methods provided initial information and situational awareness of the evolving Taʻū volcano earthquake swarm, with limited available data. Using continuous seismic data from one Global Seismographic Network station located ~250 km west of Taʻū volcano, we applied the EQTransformer deep-learning model (Mousavi et al., 2020) to automatically pick P and S phase arrivals for earthquakes between 2022-07-01 and 2022-08-12, when seismic activity began and intensified. We used S-P times from events matching felt reports to identify additional likely Taʻū volcano swarm events, while excluding events from unrelated sources such as the Tonga Trench. Waveforms from these events were used in template-matching to detect additional events. Starting 2022-08-13, when continuous Raspberry Shake seismometer data from Taʻū Island became available, short (<3 second) S-P times output by EQTransformer constrained the swarm location to the northern submarine realm of Taʻū volcano. The Raspberry Shake data was essential in determining Taʻū volcano as the correct source of the swarm, ruling out the previously assumed Vailuluʻu seamount as a potential source. On 2022-08-20, the USGS Hawaiian Volcano Observatory commenced operational response seismic monitoring in American Samoa, estimating magnitudes of 2 to 4 for most subsequent earthquakes. Methods such as deep-learning and template-matching can improve monitoring and forecasts of swarms and aftershock sequences far from seismic stations.

Session: Advances in Operational and Research Analysis of Earthquake Swarms - I

Type: Oral

Date: 5/3/2024

Presentation Time: 02:30 PM (local time)

Presenting Author: Clara

Student Presenter: No

Invited Presentation: 

Authors