Tracking reservoir stability through multi-data stream statistical data assimilation: Application to the 2008 eruption of Okmok, AK
Abstract
Over the past few decades increasing effort has been dedicated to monitoring active volcanoes using geodetic techniques (e.g., InSAR, campaign GPS, and continuous GPS), resulting in large datasets, each of which possesses its own characteristic temporal and spatial resolutions. To analyze these data for a given period of unrest, a variety of model-data fusion techniques have been developed ranging from static grid searches to complex Bayesian inversions that match observations to models of magma chamber dynamics. Recently, the Ensemble Kalman Filter (EnKF), a statistical data assimilation algorithm commonly used in hydrology and climatology, has been adapted to provide sequential forecasts of volcanic unrest. The flexible nature of the EnKF provides a framework for assimilating temporally and spatially disparate geodetic data into finite element models (FEMs) capable of resolving host rock stability under a wide variety of conditions. In this study, EnKF is used to analyze geodetic data from Okmok Volcano, Alaska, in the years leading up to its July 2008 eruption, simultaneously assimilating data from 4 continuous GPS stations, yearly GPS campaigns through 2005, and 18 Envisat InSAR interferograms. As new data are assimilated, the EnKF updates an ensemble of parameters (e.g., chamber position and volume change) for an elastic FEM simulating a deforming magma reservoir. At each iteration the FEM also calculates the stress evolution along the reservoir wall and within the roof rock to determine whether or not tensile and/or Mohr-Coulomb failure is indicated. The EnKF is able to track the observed deformation signal, constraining a source in broad agreement with previous studies at Okmok, while also providing an assessment of the magma system's stability. Our results indicate a marked increase in the likelihood of reservoir failure shortly before the 2008 eruption, providing a successful hind-cast of the event. Finally, we hope that the use of more advanced thermomechanical models will allow the assimilation of a wider variety of data for improved resolution, with the ultimate goal of rapidly assimilating data in real time to provide forecasts of volcanic unrest.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.G14A..03A
- Keywords:
-
- 1207 Transient deformation;
- GEODESY AND GRAVITYDE: 1217 Time variable gravity;
- GEODESY AND GRAVITYDE: 8419 Volcano monitoring;
- VOLCANOLOGYDE: 8488 Volcanic hazards and risks;
- VOLCANOLOGY