We present results of marine MT acquisition in the Alboran sea that also incorporates previously acquired land MT from southern Spain into our analysis. The marine data show complex MT response functions with strong distortion due to seafloor topography and the coastline, but inclusion of high resolution topography and bathymetry and a seismically defined sediment unit into a 3-D inversion model has allowed us to image the structure in the underlying mantle. The resulting resistivity model is broadly consistent with a geodynamic scenario that includes subduction of an eastward trending plate beneath Gibraltar, which plunges nearly vertically beneath the Alboran. Our model contains three primary features of interest: a resistive body beneath the central Alboran, which extends to a depth of $150 km. At this depth, the mantle resistivity decreases to values of $100 Ohm-m, slightly higher than those seen in typical asthenosphere at the same depth. This transition suggests a change in slab properties with depth, perhaps reflecting a change in the nature of the seafloor subducted in the past. Two conductive features in our model suggest the presence of fluids released by the subducting slab or a small amount of partial melt in the upper mantle (or both). Of these, the one in the center of the Alboran basin, in the uppermost-mantle (20-30 km depth) beneath Neogene volcanics and west of the termination of the Nekkor Fault, is consistent with geochemical models, which infer highly thinned lithosphere and shallow melting in order to explain the petrology of seafloor volcanics.

No abstract

No abstract

Summary Bayesian inversion of magnetotelluric (MT) data is a powerful but computationally expensive approach to estimate the subsurface electrical conductivity distribution and associated uncertainty. Approximating the Earth subsurface with 1D physics considerably speeds-up calculation of the forward problem, making the Bayesian approach tractable, but can lead to biased results when the assumption is violated. We propose a methodology to quantitatively compensate for the bias caused by the 1D Earth assumption within a 1D trans-dimensional Markov chain Monte Carlo sampler. Our approach determines site-specific likelihood functions which are calculated using a dimensionality discrepancy error model derived by a machine learning algorithm trained on a set of synthetic 3D conductivity training images. This is achieved by exploiting known geometrical dimensional properties of the MT phase tensor. A complex synthetic model which mimics a sedimentary basin environment is used to illustrate the ability of our workflow to reliably estimate uncertainty in the inversion results, even in presence of strong 2D and 3D effects. Using this dimensionality discrepancy error model we demonstrate that on this synthetic dataset the use of our workflow performs better in 80% of the cases compared to the existing practice of using constant errors. Finally, our workflow is benchmarked against real data acquired in Queensland, Australia, and shows its ability to detect the depth to basement accurately.