Uncertainty Quantification of Moment Tensor Inversions due to Earth Model Uncertainty

Abstract:

The 3-D seismic structure of the solid earth is a major source of uncertainty when performing many geophysical inversions. One example is the inversion of seismic waveforms for the seismic moment tensor source time function components. The seismic moment tensor provides valuable information on how fault movement evolves during an earthquake and for nuclear nonproliferation purposes. For a given earth model, the solution to the moment tensor inversion problem is purely linear in the frequency domain. However, there is a nonlinear mapping between the earth model and the resulting seismograms (as well as seismic Green’s functions), so mapping uncertainty in the earth model to uncertainty in the solution of moment tensor inversion involves expensive 3-D simulations. Monte Carlo simulations are the most obvious first step in mapping the uncertainty in the earth to uncertainty in the source inversion solutions. However, due to the computational expense of running 1000’s of Monte Carlo 3-D simulations, we are developing methods that can significantly increase the computational efficiency of uncertainty quantification in the solution of the seismic moment tensor inversion problem. We present promising techniques including using sparse sampling, polynomial chaos methods, and various estimates of covariance functions of the resulting Green’s functions. We also compare the accuracy and computational efficiency of the new techniques in relation to Monte Carlo simulations.

Sandia National Laboratories is a multi-mission laboratory managed and operated by National Technology and Engineering Solutions of Sandia LLC, a wholly owned subsidiary of Honeywell International Inc. for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525. SAND2020-7365 A.