3D Reconstruction of Complex Fault Systems From Volumetric Geodynamic Shear Zones Using Medial Axis Transform

Abstract

Reconstructing fault surfaces from volumetric data is a longstanding challenge in geosciences. We present a novel 3D method based on the medial axis to transform a volumetric strain-rate invariant field from long-term geodynamic simulations into fault surfaces. In these geodynamic models, faults correspond to regions of locally high values of the second invariant of the strain-rate commonly referred to as shear zones. The proposed workflow begins by normalizing the strain-rate to define fault indicator field $\xi$. An iso-surface of a chosen $\xi$ value is then extracted to form an envelope around the shear zones. Using the shrinking ball algorithm (Ma et al., 2012, https://doi.org/10.1007/s00371-011-0594-7), we compute the medial axis of this 3D envelope to generate a point cloud representing the geometric skeleton of the shear zones. We reconstruct fault surfaces by applying Delaunay triangulation followed by Laplacian smoothing. For models involving multiple intersecting faults, we perform a local principal component analysis (PCA) of the coordinates defining the medial axis and use the resulting eigenvectors to detect first-order orientation variations, enabling the separation and individualization of faults. We demonstrate the generality and robustness of the method by applying it several diverse 3D geodynamic scenarios: A single strike-slip fault, a branching strike-slip fault in a restraining bend, a dense strike-slip fault network, a rift system, and a subduction zone with a megathrust and a conjugate thrust fault.