Accelerating Bayesian microseismic event location with deep learning

We present a series of new open source deep learning algorithms to accelerate Bayesian full waveform point source inversion of microseismic events. Inferring the joint posterior probability distribution of moment tensor components and source location is key for rigorous uncertainty quantification. However, the inference process requires forward modelling of microseismic traces for each set of parameters explored by the sampling algorithm, which makes the inference very computationally intensive. In this paper we focus on accelerating this process by training deep learning models to learn the mapping between source location and seismic traces, for a given 3D heterogeneous velocity model, and a fixed isotropic moment tensor for the sources. These trained emulators replace the expensive solution of the elastic wave equation in the inference process. We compare our results with a previous study that used emulators based on Gaussian Processes to invert microseismic events. We show that all of our models provide more accurate predictions and \(\sim 100\) times faster predictions than the method based on Gaussian Processes, and a \(\mathcal{O}(10^5)\) speed-up factor over a pseudo-spectral method for waveform generation. For example, a 2-s long synthetic trace can be generated in \(\sim 10\) ms on a common laptop processor, instead of \(\sim\) 1 hr using a pseudo-spectral method on a high-profile Graphics Processing Units card. We also show that our inference results are in excellent agreement with those obtained from traditional location methods based on travel time estimates. The speed, accuracy and scalability of our open source deep learning models pave the way for extensions of these emulators to generic source mechanisms and application to joint Bayesian inversion of moment tensor components and source location using full waveforms.