← Back to Sessions

WITHDRAWN Volcanic Eruption Forecasts Through Seismic Data Assimilation: The 2023 Paroxysms of Shishaldin Volcano, Alaska

Description: 

WITHDRAWN Establishing connections between geophysical observables and subsurface processes is critical for detecting volcanic unrest and anticipating eruptions. One of the most important observables to monitor pre-eruptive volcanic activity is tremor, a more or less persistent ground vibration often recorded at active volcanoes. Tremor may manifest or alter its characteristics before eruptions; for instance, variations in the dominant frequency, the appearance or disappearance of overtones, and changes in seismic amplitude have been observed before eruptions. Nevertheless, similar variations can also occur during non-eruptive periods or when volcanic activity decreases. Consequently, this raises important questions: How does tremor reflect subsurface overpressure? Can we deduce alterations in pressure underneath volcanoes by monitoring tremor? This study introduces a novel data assimilation framework that integrates seismic data, a physics-based model of volcanic tremor, an optimization approach utilizing genetic algorithms, and time series modeling based on neural networks to track volcanic pressure changes and to produce physics-based eruption forecasts. Specifically, our physics-based model (an extension from Girona et al., 2019; https://doi.org/10.1029/2019JB017482) presumes that tremor occurs when gas randomly enters shallow levels of the volcanic plumbing system, accumulates temporarily in the subsurface, and transfers to the surface via permeable flow. The performance of our framework is examined by analyzing the recent 13 paroxysms of Shishaldin Volcano, Alaska, between July and November in 2023; these paroxysms are ideal for testing our approach as they were heralded by variations in tremor. Our framework suggests that the recent Shishaldin paroxysms resulted from a combination of magma ascent, escalating gas flux, and conduit sealing, leading to pressure increases of up to several MPa and increases in eruption probability within hours to the events. This data assimilation strategy exhibits promising potential to produce physics-based eruption forecasts from seismic tremor data in near-real-time.

Session: Multidisciplinary Approaches for Volcanic Eruption Forecasting - I

Type: Oral

Date: 5/2/2024

Presentation Time: 02:45 PM (local time)

Presenting Author: Társilo

Student Presenter: No

Invited Presentation: 

Authors