MTUQ: A High-Performance Python Package for Moment Tensor Estimation and Uncertainty Quantification
Understanding the underlying source mechanisms of earthquakes, volcanic eruptions, nuclear explosions, and other seismogenic geological events is fundamental to seismology. Characterizing the source parameters might entail evaluating its location, onset time, moment tensor (fault orientation, source type, magnitude), and source-time function. Evaluating these coupled parameters requires quantification of uncertainties to provide meaningful interpretation. Here, we present the latest developments of MTUQ, an open-source python package for Moment Tensor estimation and Uncertainty Quantification in 1D and 3D Earth models. MTUQ uses mpi4py for distributed parallelism, and its parallel grid-search routines have been tested on HPC systems for efficient exploration of the parameter space. Synthetic seismograms are obtained on the fly from precomputed 1D green’s function databases (FK, Axisem/Instaseis) or 3D green’s functions computed with SPECFEM3D. We use waveform-based misfit and allow for custom misfit functions to evaluate the best-fitting source parameters. MTUQ also benefits from advanced sampling and inversion methods (Hamiltonian Monte Carlo and Covariance Matrix Adaptation-Evolution Strategies) to improve time efficiency and enable tackling ambitious problems that consider joint inversion of larger parameter sets, such as multiple moment tensors (or forces), hypocenters, and source-time function. We showcase recent applications of the code, including the result of a study of the Hunga-Tonga volcanic eruption of January 2022 and an overview of the first workshop—online and free—dedicated to the code.
Session: Earthquake Source Parameters: Theory, Observations and Interpretations [Poster]
Type: Poster
Room: Ballroom
Date: 4/18/2023
Presentation Time: 08:00 AM (local time)
Presenting Author: Julien Thurin
Student Presenter: No
Additional Authors
Julien Thurin Presenting Author Corresponding Author jthurin@alaska.edu University of Alaska Fairbanks |
Jochen Braunmiller jbraunmiller@usf.edu University of South Florida |
Felix Rodriguez Cardozo felixr1@usf.edu University of South Florida |
Lian Ding liangding86@gmail.com University of Toronto |
Qinya Liu qinya.liu@utoronto.ca University of Toronto |
Amanda McPherson ammcpherson@alaska.edu University of Alaska Fairbanks |
Ryan Modrak rmodrak@lanl.gov Los Alamos National Laboratory |
Carl Tape ctape@alaska.edu University of Alaska Fairbanks |
|
MTUQ: A High-Performance Python Package for Moment Tensor Estimation and Uncertainty Quantification
Category
Earthquake Source Parameters: Theory, Observations and Interpretations
Description