GeoBUS – Geological Modeling by Bayesian Updating of Scalar Fields

In structural geological modeling, interfaces between different rock units are typically represented as sharp boundaries. However, geophysical methods such as electrical resistivity tomography (ERT) often depict these interfaces as smooth transitions of geophysical properties. This representation can obscure the precise locations of geological features that are critical for utilizing the models. To address this challenge, we present GeoBUS (Geological modeling by Bayesian Updating of Scalar fields), a novel workflow for creating probabilistic geological models that seamlessly incorporates information from probabilistic geophysical inversion results through Bayesian updates.

The GeoBUS workflow consists of three key steps: (1) We employ the Kalman Ensemble Generator (KEG) to invert geophysical data and generate probabilistic images in terms of a geophysical property. (2) We perform implicit structural geological modeling by creating an ensemble of scalar fields based on interface point information for geological units while accounting for associated uncertainties. This ensemble serves as the foundation for our probabilistic model. (3) We use the same subsurface discretization as in geophysical inverse modeling and assign probabilistic scalar field values to each cell based on our ensemble of geological models to build a discrete prior for a second KEG application. Based on literature-derived probability density functions for geophysical properties across different geological units, we formulate a corresponding likelihood. The KEG then returns a discretized posterior scalar field that integrates petrophysical likelihoods with geophysical inversion data. We demonstrate this novel workflow for ERT data and simple 2D structural geological models.