Inversion of magnetotelluric data in an anisotropic domain

The aim of this study was to develop a set of tools -- algorithms that would be translated into computer programs at a later stage -- to invert anisotropic magnetotelluric data within a framework that integrates multi-disciplinary information relative to the subsurface. The initial stages of research was aimed at understanding and summarizeing the available modeling strategies for the magnetotelluric method, to identify an inversion strategy that was both accurate and effective, and to explore a method to allow information from different disciplines to be used in the inversion process. Over a year was spent testing the suitability of a selected genetic algorithm to be used in solving the inverse problem. Even though this genetic algorithm was successfully used in different inverse problems with magnetotelluric data, it was not successful here because of the lack of efficiency in the framework that was used. The final part of this study was dedicated to the development, testing and appraisal of newly-developed codes based on the concept of "mutual information", information shared between two images that can be quantified in a probabilistic sense. Preliminary studies relating to synthetic tests were performed, and the results analyzed from a numerical perspective. A real anisotropic one-dimensional dataset from the DIE magnetotelluric station deployed in Central Germany was inverted. The resulting model was consistent with the most up-to-date models from independent research, surpassing these models by effectively constraining the amount of anisotropy needed to fit the data. These results made the electrical conductivity values obtained in the studied region compatible with laboratory measurements. The anisotropic two-dimensional approach is demonstrated on a test model, and shown to be very effective at elucidating conductivity structures.

Rights

Attribution-NonCommercial-NoDerivs 3.0 Ireland

cbn

Collections

University of Galway Theses (PhD Theses)