Evolving infrasound detections from Bogoslof volcano, Alaska: Insights from atmospheric propagation modeling

Abstract

Bogoslof volcano, a back-arc volcano in Alaska’s Aleutian arc, began an eruptive sequence in mid-December 2016 that ended in late August 2017, with 70 individual eruptive episodes. Because there were no local seismic or infrasound stations on the island, the Alaska Volcano Observatory (AVO) relied on distant geophysical networks and remote sensing techniques to assess activity during the eruption. AVO maintains six infrasound arrays to monitor activity along the Aleutian arc: Adak, the Island of Four Mountains, Okmok, Akutan, Sand Point, and Dillingham. Eruption detection at infrasound arrays is subject to local as well as mesoscale meteorological conditions that vary greatly over both short and long timescales. Infrasound detections from the array nearest to Bogoslof (Okmok), with a latency of about 3 min, played a crucial role in monitoring activity during the eruption. Despite the relative proximity of the Okmok array to Bogoslof (60 km), infrasound detections were not uniformly observed with only about two-thirds of the events successfully detected. The farthest array at Dillingham (816 km) detected approximately half of the explosive events, with all other arrays detecting less than half of the events. We compare observations with infrasound propagation model predictions, using both normal mode and parabolic equation forward models, to interpret the variation in detections of the 70 explosive events across the AVO infrasound network. The forward models utilize the newly created, publicly available AVO-G2S atmospheric reconstruction using numerical weather predictions data for the lower atmosphere, coupled with upper atmosphere empirical models of wind speeds and temperature. We find that long-range detections (> 100 km) of Bogoslof events are largely aligned with seasonal variability in favorable propagation conditions, while regional detections (< 100 km) are less consistent with propagation modeling. Understanding the output of numerical models in comparison to past observations will facilitate their use in future operational settings for AVO and other observatories.