Recovering lithology geometries via geophysical data inversion based on a generalized level set approach

Abstract:

Geophysical data inversion is particularly useful in reducing the gaps in sparse geological data and knowledge. While optimization or Bayesian approaches allow exploring an ensemble of solution models fitting the data to some measurement error level, they quickly become computationally expensive. In addition, as with deterministic approaches, they often rely on a smooth conceptual representation of the rock properties. Level-set inversions are deterministic approaches that offer the advantage of preserving sharp boundaries between geological units. However, existing level-set algorithms generally only consider two level-set functions, lack topological control over the recovered model, and typically require hundreds to thousands of iterations to converge.

In this work, we develop a new technique relying on level-set functions that does not present these restrictions while, like most inversions using clustering, are capable of considering an arbitrary number of geological units. It relies on the definition of a specific level-set function per geological unit and on the use of the signed-distance away from the geological unit interfaces. While we apply it to the inversion of gravity data from the Yerrida Basin, Australia and from the Pyrénées, France, the mathematical framework is general. It is straightforward to apply it to magnetic data or to the joint inversion of potential field data, and can easily be extended to the inversion of travel-time data and gravity.

Acknowledgement

We acknowledge the support from the ARC-funded Loop: Enabling Stochastic 3D Geological Modelling consortia (LP170100985) and DECRA (DE190100431). The work has also been supported by the Mineral Exploration Cooperative Research Centre whose activities are funded by the Australian Government's Cooperative Research Centre Programme. This is MinEx CRC Document 2020/38.