Inversion of geological maps for fault geometry

Lachlan Grose and Laurent Ailleres. ( 2021 )
in: 2021 RING Meeting, ASGA

Abstract

The geometry and kinematics of faults recorded on geological maps are usually inferred from limited observations, such as the observed stratigraphic offset, lineations and/or regional tectonic knowledge. The process of interpreting fault geometries and kinematics is an inverse problem, where the geologist is required to identify the best fit model (fault geometry and kinematics) given observations and geological knowledge. The interpreted geometry is usually a single non-unique solution with little to no indication of the associated confidence/uncertainty. One of the challenges for characterising the fault geometry from a map pattern is that multiple combinations of fault kinematics and fault surface geometries result in a similar observed map pattern. LoopStructural is an open source 3D modelling library, where faults can be modelled from a combination of parameters describing the fault surface geometry, fault slip direction, fault ellipsoid geometry and fault displacement magnitude. Faults are added to the implicit function that represents the faulted geological objects by un-faulting the observations of the faulted object and allowing the geological object to be interpolated without the fault. In this study, we use the fault model within LoopStructural as a forward model and define a likelihood function for determining how well the fault describes the un-faulting of the observations. The likelihood function is incorporated into a Bayesian inversion scheme where the geologist's knowledge can be used to constrain the prior distributions of the fault parameters. Preliminary results from the Hamersley region in Western Australia show comparable map patterns can be produced for a range of parameter combinations – e.g. decreasing the displacement magnitude as the fault slip vector becomes more horizontal.

Download / Links

BibTeX Reference

@inproceedings{GROSE_RM2021,
 abstract = { The geometry and kinematics of faults recorded on geological maps are usually inferred from limited observations, such as the observed stratigraphic offset, lineations and/or regional tectonic knowledge. The process of interpreting fault geometries and kinematics is an inverse problem, where the geologist is required to identify the best fit model (fault geometry and kinematics) given observations and geological knowledge. The interpreted geometry is usually a single non-unique solution with little to no indication of the associated confidence/uncertainty. One of the challenges for characterising the fault geometry from a map pattern is that multiple combinations of fault kinematics and fault surface geometries result in a similar observed map pattern. LoopStructural is an open source 3D modelling library, where faults can be modelled from a combination of parameters describing the fault surface geometry, fault slip direction, fault ellipsoid geometry and fault displacement magnitude. Faults are added to the implicit function that represents the faulted geological objects by un-faulting the observations of the faulted object and allowing the geological object to be interpolated without the fault. In this study, we use the fault model within LoopStructural as a forward model and define a likelihood function for determining how well the fault describes the un-faulting of the observations. The likelihood function is incorporated into a Bayesian inversion scheme where the geologist's knowledge can be used to constrain the prior distributions of the fault parameters. Preliminary results from the Hamersley region in Western Australia show comparable map patterns can be produced for a range of parameter combinations – e.g. decreasing the displacement magnitude as the fault slip vector becomes more horizontal.   },
 author = { Grose, Lachlan AND Ailleres, Laurent },
 booktitle = { 2021 RING Meeting },
 publisher = { ASGA },
 title = { Inversion of geological maps for fault geometry },
 year = { 2021 }
}