Probabilistic Inversion of AEM Data with an Explicit Choice of Prior Information
T.M. Hansen and B. Minsley
Event name: Second European Airborne Electromagnetics Conference
Session: Forward Modelling and Inversion I
Publication date: 03 September 2017
Info: Extended abstract, PDF ( 1.38Mb )
Price: € 20
Multiple approaches exist that allow inverting airborne electromagnetic (AEM) data into subsurface models of resistivity. Here we present an alternative approach to inversion of AEM data, based on probabilistic approach to inverse problems where known states of information is combined into a posterior probability density (the solution to the inverse problem) using the concept of conjunction of information. Two types of information is integrated as part of the inversion: The observed electromagnetic data with associated model of uncertainty, and an explicit choice of a priori information based on existing expert knowledge. The choice of prior model is here completely detached from the quantification of the data (and uncertainty of the data), and there is therefore no manual regularization to consider. We demonstrate the methodology on an airborne electromagnetic data from, using a variety of Gaussian based a priori models with non-Gaussian 1D marginal distributions.