Microseismic event detection in large heterogeneous velocity models using Bayesian multimodal nested sampling

• DOI: 10.1017/dce.2021.1
• Publication Date: 2021-02-26T10:47:41Z
• Authors: Saptarshi Das; Michael P. Hobson; Farhan Feroz; Xi Chen; Suhas Phadke; Jeroen Goudswaard; Detlef Hohl
• Keywords: Microseism; Seismogram

Abstract In passive seismic and microseismic monitoring, identifying characterizing events in a strong noisy background is challenging task. Most of the established methods for geophysical inversion are likely to yield many false event detections. The most advanced these schemes require thousands computationally demanding forward elastic-wave propagation simulations. Here we train use an ensemble Gaussian process surrogate meta-models, or proxy emulators, accelerate generation accurate template seismograms from random locations. presence multiple occurring at different spatial locations with arbitrary amplitude origin time, noise, inference algorithm needs navigate objective function likelihood landscape highly complex shape, perhaps modes narrow curving degeneracies. This computational task even state-of-the-art Bayesian sampling algorithms. this paper, propose novel method detecting noise using inference, particular, Multimodal Nested Sampling (MultiNest) algorithm. not only provides posterior samples 5D spatio-temporal-amplitude real events, by inverting traces surface receivers, but also computes evidence marginal that permits hypothesis testing discriminating true vs. detection.