Probabilistic inversion of airborne electromagnetic data for a multidimensional earth

Abstract

The inversion of airborne electromagnetic data is inherently non-unique, especially when data uncertainties are taken into account. If one model can be found that fits the data, then it is likely that there are alternative models that fit the data equally well. The probabilistic approach introduced in this work therefore aims at exploring the posterior distribution which is the distribution of models that are in agreement with both the prior information and the data. We quantify the prior information using geostatistics and use a Markov Chain Monte Carlo technique to sample the unknown posterior distribution. Data are predicted taking lateral changes in structure along the flight path into account by employing a 2.5D forward solver. A case study using the Harmony Ni-S deposit in Western Australia shows that our set of samples of the posterior distribution provides a more complete picture of solution space than what can be achieved by non-linear iterative inversion schemes that have previously been employed. Such a picture of the subsurface can ultimately be used to mitigate exploration risk.