[S-SS05] Advancements in Regional Seismic Networks: Operations, Applications, and Development

[SSS05-P05] mtuqorg: a modular approach to moment tensor estimation and uncertainty quantification with MTUQ

*Julien Thurin¹、Ryan Modrak⁴、Amanda McPherson¹、Félix R Rodríguez-Cardozo²、Jochen Braunmiller²、Liang Ding⁵、Qinya Liu³、Jonas A Kintner⁴、Jinyin Hu⁶、Hrvoje Tkalčić⁶、Carl Tape¹ (1.University of Alaska Fairbanks、2.University of South Florida、3.University of Toronto、4.Los Alamos National Laboratory、5.Geological Survey of Canada、6.Australia National University)

キーワード：Moment tensor, Inversion, Uncertaity quantification, Seismic sources

Seismic moment tensors, along with point forces, are used to describe various source mechanisms, including tectonic earthquakes, human-made explosions, and mass movements. Accurately estimating these parameters and rigorously quantifying uncertainties are crucial for enhancing our understanding of seismic sources and reducing ambiguity in their interpretation. The open-source Moment Tensor and Uncertainty Quantification (MTUQ) code provides flexible tools for point-source inversion using seismic waveforms and first-motion polarities, with a focus on balancing computational efficiency and robust uncertainty estimation. We showcase the code's capabilities by revisiting the DPRK nuclear test of September 2017 to evaluate the influence of various data sources on the misfit space, both using a conventional grid-search approach and leveraging the fast Covariance Matrix Adaptation Evolution Strategy (CMA-ES) method. We also demonstrate the modularity of the code by supplementing the CMA-ES optimization with the "increase population restart" strategy, which helps map out the misfit space more efficiently and illuminate local minima. MTUQ accommodates a wide range of Green's function databases (both 1D and 3D and receiver side Strain Green's Function), and integrates with high-performance computing architectures, supporting workflows from single-core optimizations to large-scale grid searches spanning billions of grid points. MTUQ also offers tools for visualizing solutions and uncertainties, facilitating better interpretation of inversion results and aiding reproducibility. New users can engage easily through workshops, online resources, and open development channels, ensuring that MTUQ remains a robust and evolving resource for seismic source characterization and forensic seismology.