Sharp boundary inversion of 2D magnetotelluric data

Abstract

We consider 2D earth models consisting of laterally variable layers. Boundaries between layers are described by their depths at a set of nodes and interpolated laterally between nodes. Conductivity within each layer is described by values at a set of nodes fixed within each layer, and is interpolated laterally within each layer. Within the set of possible models of this sort, we iteratively invert magnetotelluric data for models minimizing the lateral roughness of the layer boundaries, and the lateral roughness of conductivities within layers, for a given level of data misfit. This stabilizes the inverse problem and avoids superfluous detail. This approach allows the determination of boundary positions between geological units with sharp discontinuities in properties across boundaries, while sharing the stability features of recent smooth conductivity distribution inversions. We compare sharp boundary inversion results with smooth conductivity distribution inversion results on a numerical example, and on inversion of field data from the Columbia River flood basalts of Washington State. In the synthetic example, where true positions and resistivities are known, sharp boundary inversion results determine both layer boundary locations and layer resistivities accurately. In inversion of Columbia flood basalt data, sharp boundary inversion recovers a model with substantially less internal variation within units, and less ambiguity in both the depth to base of the basalts and depth to resistive basement.

Identification of subsurface structures using electromagnetic data and shape priors

“…In (1) -(4), e and h are the electric and magnetic fields, respectively, j is the electric source current distribution, denotes the permittivity, μ denotes the magnetic permeability, and σ denotes the electric conductivity (assumed to be isotropic). Taking the curl of (1) and using (2) we can eliminate h, yielding…”

“…The main aim with the research presented in this paper is to develop a computationally efficient inversion methodology for CSEM data that is able to preserve prior information about geological strata, and we will therefore apply a model-based representation (see, e.g., [2][3][4][5][6][7][8][9]) of the unknown electric conductivity field. Alternatively, a pixel-based representation (see, e.g., [10][11][12][13]) could have been used, although the tendency for smoothing out the resulting conductivity field makes it less attractive.…”

Modeling the resolution of inexpensive, novel non-seismic geophysical monitoring tools to monitor CO2 injection into coal beds

“…This difficulty can be overcome by introduction of a priori information. We adopted the approach described by Smith et al (1999) for magnetotelluric data inversion, in which the top and base of the reservoir are known, and we invert for a smooth density variation inside the reservoir. The inversion result is a cumulative density change in the reservoir as a function of x and y coordinates.…”

Laterally Constrained Inversion of Fixed-Wing Frequency-Domain AEM Data

“…To overcome the aforementioned problems we have developed a laterally constrained inversion (Auken et al, 2002;Smith et al, 1999). In this type of inversion we invert simultaneously for several observation points and we require that lateral changes in the model parameters from one observation point to the next be small.…”