Bayesian surface wave dispersion inversion of glaciated environments

• Main Authors: Lanteri, A., Gebraad, L., Zunino, A., Klaasen, S., Jonsdottir, K., Hofstede, C., Eisen, O., Zigone, D., Fichtner, A.
• Format: Conference Object
• Language: English
• Published: 2023
• Subjects: Greenland; Vatnajökull; Ice Sheet; Iceland
• Online Access: https://gfzpublic.gfz-potsdam.de/pubman/item/item_5017627

We present a probabilistic approach to the inversion of surface wave dispersion data from glacial environments. This is intended to (i) assess non-linearity and non-uniqueness, and (ii) properly quantify resolution and trade-offs. For this, we use seismic data from Distributed Acoustic Sensing (DAS) experiments deployed on the Vatnajökull ice sheet located on Grímsvötn volcano in Iceland, and the Northeast Greenland Ice Stream (NEGIS). Our method is based on a regularisation-free Bayesian inference approach, implemented using a Hamiltonian Monte Carlo (HMC) algorithm. Exploiting derivative information for efficient sampling of high-dimensional model spaces, HMC approximates the posterior probability densities of all model parameters. Applied specifically to multi-mode surface wave dispersion measurements, HMC yields probabilistic models of 1-D anisotropic stratified media parameterised in terms of the P-wave velocities Vpv and Vph, the S-wave velocities Vsv and Vsh, the anisotropy parameter η, and density ρ. The benefits of this approach, not only from a glaciological perspective, include regularisation-free estimates of firn and ice properties, models that are not a priori biased by the exclusion of all parameters except S-wave speed, and some level of direct access to the vertical density profile.