Journal
GEOPHYSICAL JOURNAL INTERNATIONAL
Volume 199, Issue 2, Pages 1205-1213Publisher
OXFORD UNIV PRESS
DOI: 10.1093/gji/ggu329
Keywords
Inverse theory; Probability distributions; Magnetotellurics; Non-linear electromagnetics
Categories
Funding
- National Natural Science Foundation of China [41204081, 41174103]
- China Postdoctoral Science Foundation [2012M510293]
Ask authors/readers for more resources
Correlated data errors (noise) are common in magnetotelluric (MT) data, but MT inversions typically neglect error correlations without investigating the impact of this simplification on inversion results. This paper examines effects of neglecting frequency-and spatially correlated noise on MT inversion, based on a nonlinear Bayesian formulation which quantifies the uncertainties of inversion results in terms of marginal posterior probability densities and credibility intervals. To do so, data with frequency-and spatially correlated noise of differing degrees are generated for several layered (1-D) synthetic cases. Bayesian MT inversions are carried out for these data sets with and without accounting for error correlation (i.e. applying full and diagonal covariance matrices, respectively, in the inversion), and the results are compared. For cases with noise that is strongly correlated over frequency or space, parameter uncertainties estimated using the diagonal-covariance simplification (neglecting error correlations) are found to often be significantly underestimated compared with results computed using the full covariance matrix.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available