Assimilation of Deformation Data for Eruption Forecasting: Potentiality Assessment Based on Synthetic Cases

Abstract

In monitoring active volcanoes, the magma overpressure is one of the key parameters used in forecasting volcanic eruptions. This parameter can be inferred from the ground displacements measured on the Earth's surface by applying inversion techniques. However, in most studies, the huge amount of information about the behavior of the volcano contained in the temporal evolution of the deformation signal is not fully exploited by inversion. Our work focuses on developing a strategy in order to better forecast the magma overpressure using data assimilation. We take advantage of the increasing amount of geodetic data [i.e., Interferometric Synthetic Aperture Radar (InSAR) and Global Navigation Satellite System (GNSS)] recorded on volcanoes nowadays together with the wide-range availability of dynamical models that can provide better understanding about the volcano plumbing system. Here, we particularly built our strategy on the basis of the Ensemble Kalman Filter (EnKF). We forecast the temporal behaviors of the magma overpressures and surface deformations by adopting a simple and generic two-magma chamber model and by using synthetic GNSS and/or InSAR data. We prove the ability of EnKF to both estimate the magma pressure evolution and constrain the characteristics of the deep volcanic system (i.e., reservoir size as well as basal magma inflow). High temporal frequency of observation is required to ensure the success of EnKF and the quality of assimilation is also improved by increasing the spatial density of observations in the near-field. We thus show that better results are obtained by combining a few GNSS temporal series of high temporal resolution with InSAR images characterized by a good spatial coverage. We also show that EnKF provides similar results to sophisticated Bayesian-based inversion while using the same dynamical model with the advantage of EnKF to potentially account for the temporal evolution of the uncertain model parameters. Our results show that EnKF works well with the synthetic cases and there is a great potential in using the method for real-time monitoring of volcanic unrest.

Controls on explosive-effusive volcanic eruption styles

“…Seismicity, deformation and gas measurements may be used to interpret conditions in the magma plumbing system. For instance, long period earthquakes, seismic velocity changes and ‘drumbeat’ seismicity have been used to detect magmatic ascent 109 , 119 , 147 ; pre-eruptive, InSAR and tilt data were linked to the rate of pressure change and resulting explosivity of an eruption 148 , 149 ; and increases in CO 2 relative to SO 2 phases have been recorded before some explosive eruptions 150 , 151 . However, those monitoring techniques require much refinement before volcanologists can, in near real-time predict future behaviours of a given volcano.…”

Integrating Reservoir Dynamics, Crustal Stress, and Geophysical Observations of the Laguna del Maule Magmatic System by FEM Models and Data Assimilation

“…The EnKF solves the Bayesian update problem using a Monte Carlo method (Evensen, ). We follow the same strategy developed in previous studies using the EnKF method to model volcanic deformation (Figure ; Gregg & Pettijohn, ; Zhan & Gregg, ; Zhan et al, ; Bato et al, , ; Albright et al, ). The EnKF analysis updates the physical model ( A a ) by where A is a matrix containing the parameters and states of all ensemble models whose covariance matrix is P e .…”

“…The EnKF solves the Bayesian update problem using a Monte Carlo method (Evensen, 2003). We follow the same strategy developed in previous studies using the EnKF method to model volcanic deformation (Figure 3; Gregg & Pettijohn, 2015;Bato et al, 2017Bato et al, , 2018Albright et al, 2019). The EnKF analysis updates the physical model (A a ) by…”

Possible deep connection between volcanic systems evidenced by sequential assimilation of geodetic data

“…While several studies have used real geodetic data and applied variants of Kalman Filter as an optimization or statistical interpolation tool to solve problems in volcanology in the past 9 – 11 , this study is the first one to apply sequential data assimilation based on a dynamical model as proposed by ref. 12 using a real dataset recorded on a volcano.…”