Open AccessEvidence of Bermuda Hot and Wet Upwelling From Novel Three‐Dimensional Global Mantle Electrical Conductivity Image
Author(s) -
Li Shiwen,
Weng Aihua,
Zhang Yanhui,
Schultz Adam,
Li Yabin,
Tang Yu,
Zou Zonglin,
Zhou Zikun
Publication year - 2020
Publication title -
geochemistry, geophysics, geosystems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.928
H-Index - 136
ISSN - 1525-2027
DOI - 10.1029/2020gc009016
Subject(s) - geology , covariance , geophysics , earth's magnetic field , geodesy , mathematics , statistics , physics , quantum mechanics , magnetic field
A model of electrical conductivity in the mid–mantle transition zone was obtained with improved constraints. An L 1 ‐norm regularization inversion algorithm is proposed here that reduces the influence of noisy data on three‐dimensional geomagnetic depth sounding inversion from C‐ responses. Here the regularization is implemented by using an L 1 ‐norm to measure the predicted data error, which is normalized by the C‐ response covariance, but an L 2 ‐norm is used to measure the regularization term associated with model parameters. The limited‐memory quasi‐Newton method (L‐BFGS) is used to invert for the three‐dimensional electrical conductivity model. The model is discretized by curved rectangular prisms in spherical coordinates. Sensitivity tests show that for good‐quality data contaminated by Gaussian noise, L 1 inversion, which could perform as well as L 2 inversion, can adequately recover the main features of the electrical conductivity structure within the region of data coverage. When data errors are drawn from an exponential distribution, L 1 inversion obtains relatively reliable reconstruction of the electrical structure, even when the noise level is comparable to that of actual C‐ responses. C‐ responses from 129 low‐latitude and midlatitude geomagnetic observatories are inverted using L 1 ‐norm minimization of the data error. The resulting model reveals an electrically conductive feature in the lower mantle transition zone and upper lower mantle that is broadly coincident with that found in previous studies. The reduced influence of data with large variances on L 1 ‐norm misfits, along with inclusion of responses estimated from more observatories, makes L 1 inversion more clearly identify these deep conductive features while identifying previously obscured anoconductive zones. A feature of particular interest is the high electrical conductivity anomaly beneath the Bermuda‐Sargasso Sea region in the mid–mantle transition zone and the uppermost lower mantle. Rock physics analysis indicates that the anomaly is most possibly caused by the wet upwelling material with excessive ~650 K higher temperature, suggesting a narrow tail with a broad head.