Square-Root Variable Metric based elastic full-waveform inversion – Part 1: Theory and validation
KAUST DepartmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
Earth Science and Engineering Program
Extreme Computing Research Center
Physical Science and Engineering (PSE) Division
KAUST Grant NumberUAPN#2605-CRG4
Online Publication Date2019-05-17
Print Publication Date2019-08-01
Permanent link to this recordhttp://hdl.handle.net/10754/652934
MetadataShow full item record
AbstractFull-waveform inversion (FWI) has become a powerful tool in inverting subsurface geophysical properties. The estimation of uncertainty in the resulting Earth models and parameter trade-offs, although equally important to the inversion result, has however often been neglected or became prohibitive for large-scale inverse problems. Theoretically, the uncertainty estimation is linked to the inverse Hessian (or posterior covariance matrix), which for massive inverse problems becomes impossible to store and compute. In this study, we investigate the application of the square-root variable metric (SRVM) method, a quasi-Newton optimisation algorithm, to FWI in a vector version. This approach allows us to reconstruct the final inverse Hessian at an affordable storage memory cost. We conduct SRVM based elastic FWI on several elastic models in regular, free-surface and practical cases. Comparing the results with those obtained by the state-of-the-art L-BFGS algorithm, we find that the proposed SRVM method performs on a similar, highly-efficient level as L-BFGS, with the advantage of providing additional information such as the inverse Hessian needed for uncertainty quantification.
CitationLiu Q, Peter D, Tape C (2019) Square-Root Variable Metric based elastic full-waveform inversion – Part 1: Theory and validation. Geophysical Journal International. Available: http://dx.doi.org/10.1093/gji/ggz188.
SponsorsThe authors are grateful to editor Jean Virieux and two anonymous reviewers for improving the initial manuscript. This work was supported by the King Abdullah University of Science & Technology (KAUST) Office of Sponsored Research (OSR) under award No. UAPN#2605-CRG4. Computational resources were provided by the Information Technology Division and Extreme Computing Research Center (ECRC) at KAUST.
PublisherOxford University Press (OUP)
Except where otherwise noted, this item's license is described as This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.