Moment tensor estimation and uncertainty quantification using mtuq, instaseis, obspy and pymc

Characterizing the source process for nuclear explosions, earthquakes, volcanic and geothermal events, induced events, landslides, and related phenomena is a fundamental component of seismology. This characterization may include origin time, hypocenter, magnitude, orientation, full moment tensor, force, and source time function. Inverting for these parameters involves strong tradeoffs that motivate the use of robust uncertainty quantification methods. We present validation tests and early results from MTUQ, a new software package designed for such use cases (github.com/uafseismo/mtuq). MTUQ provides distributed parallel grid search capabilities based on mpi4py and Monte Carlo inversion capabilities based on pymc. Obspy is used for basic data structures and, rather than having to perform forward modeling on-the-fly, synthetics can be generated from precomputed AxiSEM databases using instaseis. The parallel grid search routines have been well tested, and the MCMC routines are currently under development. After establishing these capabilities, we intend to apply MTUQ to a variety of regional seismology and nuclear monitoring problems.