�ֻ�ֱ��

New research helps detect elusive volcanic tremors

This week, scientists from the �ֻ�ֱ�� are presenting their work alongside thousands of colleagues from around the world at the 2023 American Geophysical Union fall meeting. Some of their discoveries are featured here. You can also find out more about UAF at AGU by searching for #UAFxAGU on social media platforms.

Earthquakes have a well-known and easily recognizable connection to volcanoes. The sudden onset of ground shaking can indicate near-term or imminent volcanic activity.

Volcanic tremors, however, are difficult to detect. These are semi-continuous seismic or acoustic signals, or both, that continue for just a few seconds or persist for a year or more.

They, like earthquakes, carry information about activity within and below a volcano.

Graduate student researcher Darren Tan at the �ֻ�ֱ�� Geophysical Institute is researching a way to improve tremor detection and to also classify them.

“Volcanic tremor tends to have a gradual emergent onset, which traditional earthquake detection algorithms struggle to detect,” Tan said. “So what is required for seismologists to detect them is visual scanning of spectrograms.”

Visual scanning is a time-consuming task done manually every day at the �ֻ�ֱ�� Volcano Observatory. The observatory is a joint program of the UAF Geophysical Institute, the �ֻ�ֱ�� Division of Geological and Geophysical Surveys and the U.S. Geological Survey. The USGS funds the observatory, part of which is based at the Geophysical Institute.

Tan’s research will automate that scanning through machine learning methods. 

Tan drew upon the diversity of tremor signals from the 2021-2022 eruption of Pavlof Volcano, on the �ֻ�ֱ�� Peninsula, to build an extensive dataset of labeled seismic and low-frequency acoustic spectrograms. Those spectrograms represent a variety of classifications such as tremor type, explosions and earthquakes, which are then used to train a computer model for each data type. 

The trained models will then be able to detect and classify the volcanic tremors in near real time.

Humans will still be involved in interpreting what the automation produces, however.

“We want to try it out on �ֻ�ֱ�� volcanoes but at the same time do it sensibly,” Tan said. “We don’t want to lose touch with interpretation.”

Tan’s research is funded by a National Science Foundation eruption forecasting project.