Publication 167

Magnetotelluric Inversion based on Mutual Information

Eric Mandolesi and Alan G Jones


Joint inversion of different geophysical datasets is becoming a more popular and powerful tool, and it has been performed on data sensitive both to the same physical parameter and to different physical parameters. Joint inversion is undertaken to reduce acceptable model space and to increase sensitivity to model parameters that one method alone is unable to resolve adequately. We examine and implement a novel hybrid joint inversion approach. In our inversion scheme a model – the reference model – is fixed, and the information shared with the subsurface structure obtained from another method will be maximized; in our case conductivity structures from magnetotelluric (MT) inversion. During inversion, the joint probability distribution of the MT and the specified reference model is estimated and its entropy minimized in order to guide the inversion result towards a solution that is statistically compatible with the reference model. The powerful feature of this technique is that no explicit relationships between estimated model parameters and reference model ones are presumed: if a link exists in data then it is highlighted in the estimation of the joint probability distribution, if no link is required, then none is enforced. Tests performed verify the robustness of this method and the advantages of it in a 1D anisotropic scenario are demonstrated. A case study was performed with data from Central Germany, effectively fitting an MT dataset from a single station within as minimal an amount of anisotropy as required.


Geophysical Journal International, accepted 03 July, 2014. [PDF preprint]

MTNet Home Page Alan's Home Page

Alan G Jones / 03 July 2014 /