THE ELECTRICAL-CONDUCTIVITY OF THE OCEANIC UPPER MANTLE

Previous inversions of sea-floor magnetotelluric (MT) sounding data have predicted upper mantle electrical conductivities which are more than an order of magnitude higher than laboratory measurements of the conductivity of olivine would suggest, and controlled source-field electromagnetic (CSEM) soundings require a lithospheric mantle conductivity of 3 x 10(-5) Sm-1, which is so low that the electromagnetic (EM) coast effect would produce more anisotropy in MT soundings than is observed. We address these issues by constructing an olivine mantle model for conductivity and examining the inversion of MT data from three sea-floor sites, and show that the incompatibilities can be largely resolved if the effects of oceans and coastlines are considered. Our mantle model is based on recent measurements of olivine conductivity, the conductivity of tholeiite melt, a thermal model for the upper mantle based on lithospheric cooling and the temperature of alpha --> alpha + beta olivine transition, the pyrolite model of mantle petrology, and conductivities derived from CSEM sounding. We propose Archie's Law with exponent 2 and interconnected tubes as realistic lower and upper bounds for the effect of partial melt on rock conductivity, and Archie's Law with exponent 1.5 as the preferred estimate. The 1-D forward response of this model is not compatible with observed sea-floor MT data. Three data sets presented by Oldenburg (1981) pass a test for one-dimensionality based on the size of the residuals when fit with Parker's D+ algorithm, but two of the three soundings fail a test for independence of residuals. We also find that the presence of the upper mantle 'high-conductivity zone' previously inferred from these data is highly dependent on data misfit and not required when the misfit criterion is relaxed a reasonable amount. Re-inversions of the MT data produce models which are incompatible with our petrological model of mantle conductivity. However, by adding an ocean with various coastlines of simple geometry to our petrological model and solving the forward 3-D problem using thin-sheet analysis we predict MT responses which are distorted in a manner that is remarkably similar to the observed data.