GeoBUS: A Probabilistic Data Assimilation Framework for Geological Interpretation of Geophysical Inversion Results

Probabilistic geophysical inversion methods increasingly provide ensembles of subsurface physical property models, offering valuable insight into data-driven uncertainty. However, the geological interpretation of such inversion results remains challenging, as uncertainty is typically quantified in terms of the physical parameter space rather than in terms of geological structure. Translating probabilistic inversion outcomes into consistent geological models is therefore still often performed in an ad hoc or deterministic manner.

We present the probabilistic data assimilation framework GeoBUS for the geological interpretation of geophysical inversion results. GeoBUS operates on scalar-field-based implicit geological models and treats geological structures as uncertain quantities that can be updated using observational constraints. The framework is independent of the specific geophysical inversion algorithm used and can assimilate probabilistic inversion results alongside other sources of geological information.

We test GeoBUS using a synthetic case study. A reference geological model is defined and used to generate corresponding electrical resistivity tomography data, which are subsequently inverted using a probabilistic inversion scheme to obtain an ensemble of resistivity models. These inversion results are then assimilated in GeoBUS through petrophysical consistency relationships, yielding posterior ensembles of geological scalar fields that can be directly compared to the known reference model for validation of the workflow.

In a second step, we extend the study by sequentially assimilating additional geological information. In this example, borehole interface depths are incorporated to illustrate how GeoBUS naturally accommodates heterogeneous observations and progressively reduces structural uncertainty. This demonstrates the flexibility of the framework and its potential for bridging the gap between probabilistic geophysical inversion and geological modeling in applied geophysics.