Item

Abstract

Potential field methods have been widely used in a great variety of geoscience-related applications. Interpreting potential field data often involves estimating the shapes of the anomalous source bodies such as salt bodies, mineral deposits, basin geometry, etc. However, most of the existing methods for recovering the anomalous body shapes do not account for the uncertainty, making it challenging for the subsequent resource evaluation and risk analysis to properly take uncertainty into account. We have developed a trans-dimensional Bayesian framework to tackle this problem. This new framework can handle 2D/3D potential field data and complex geometries. To make Monte Carlo sampling computationally feasible in a PC, we proposed a sparse geometry parameterization strategy, which allows us to adequately approximate complex geometries using a small set of simple geometries such as ellipses, thereby greatly reducing the number of parameters to be sampled. During sampling, we randomly perturb the number, locations, sizes and orientations of the ellipses. To impose prior constraints on the top boundary of anomalous bodies, for example, from seismic imaging and magnetotellurics data inversion, we adopted a set of fixed ellipses located right on the known boundaries and oriented along the boundaries. These fixed ellipses are then connected with the randomly sampled ellipses by an alpha shape which serves as an estimate of the complex geometry of an anomalous source body. We have successfully applied our method to the SEG/EAGE salt model in both 2D and 3D, as well as the Sigsbee and Pluto salt models.