A 3D bayesian inversion for thermodynamic state using the VBRc
paper citation Paxman, G. J. G., Lau, H. C. P., Austermann, J., Holtzman, B. K., Havlin, C. 2022. Inference of the timescale-dependent apparent viscosity structure in the upper mantle beneath Greenland. Submitted to AGU Advances.
The archived version of this code for the above paper is located at https://doi.org/10.5281/zenodo.6598795.
Further development may occur on Github at https://github.com/guypaxman/vbr_bayesian_inversion.
This code requires MATLAB or GNU-Octave and the VBRc. This code also requires some data files not included in the Github repository (see the following section).
The main input datasets are available in the Zenodo repository. These include .mat files for the GLAD-M25 shear wave velocity model (Vs_Model.mat) and the QL6 1D global attenuation model (QL6_Model.mat). Also included is the VBRc parameter sweep lookup table (sweep_box.mat).
If you do not download the VBRc lookup table, it can be generated via the make_sweep.m script before running the main code for the first time (see below).
The seismic tomography data were originally accessed from the IRIS Earth Model Collaboration data products webpages:
GLAD-M25 (Vs): https://doi.org/10.17611/dp/emc.2020.gladm2500.1. PREM-QL6 (Q): https://doi.org/10.17611/DP/9991844.
The main driving scripts are
make_sweep.m: if you did not download the VBRc lookup table sweep_box.mat, you'll need to generate it. run_with_*_res.m: these scripts handle loading the observation data, loading the Vs and Q predictions from the VBRc and running the bayesian inversion. Most adjustable parameters are stored in high-level structures in these files. Note that run_with_original_res.m will run a high resolution inversion that requires significant RAM.This repository includes a minimal test suite, see ./tests/readme.md for more details.