Transdimensional Inversion of Well Log Data for a Two-Dimensional Geological Model with Inclined Layers and Petrophysical Lateral Variability
Julien Herrero and Guillaume Caumon and Thomas Bodin. ( 2026 )
in: Mathematical Geosciences
Abstract
This study introduces a transdimensional inversion framework to quantify stratigraphic uncertainties in layered reservoir models from well data. The goal is to adaptively determine the appropriate number of subsurface model parameters. The parameters to infer correspond to (1) the number of layers , (2) the average permeability for each layer, (3) the horizontal gradient of permeability in each layer, (4) interface depths, and (5) interface slope angles. The inverse problem is formulated within a Bayesian framework, and a reversible-jump Markov chain Monte Carlo algorithm is employed to determine both the optimal number of layers and their properties. This approach is first applied to a synthetic case aiming to reconstruct a two-dimensional continuous field from two well logs of permeability, demonstrating the ability of the method to retrieve a posterior density close to the reference solution. The method is then successfully applied to real data from three well logs in the Teapot Dome oilfield, Wyoming, USA. Overall, the proposed methodology provides a unified framework through which to infer geological parameters and can be extended to integrate seismic inversion, flow data inversion (e.g., for well test interpretation), or other types of geophysical data.
Download / Links
BibTeX Reference
@article{herrero:hal-05484971,
abstract = {This study introduces a transdimensional inversion framework to quantify stratigraphic uncertainties in layered reservoir models from well data. The goal is to adaptively determine the appropriate number of subsurface model parameters. The parameters to infer correspond to (1) the number of layers , (2) the average permeability for each layer, (3) the horizontal gradient of permeability in each layer, (4) interface depths, and (5) interface slope angles. The inverse problem is formulated within a Bayesian framework, and a reversible-jump Markov chain Monte Carlo algorithm is employed to determine both the optimal number of layers and their properties. This approach is first applied to a synthetic case aiming to reconstruct a two-dimensional continuous field from two well logs of permeability, demonstrating the ability of the method to retrieve a posterior density close to the reference solution. The method is then successfully applied to real data from three well logs in the Teapot Dome oilfield, Wyoming, USA. Overall, the proposed methodology provides a unified framework through which to infer geological parameters and can be extended to integrate seismic inversion, flow data inversion (e.g., for well test interpretation), or other types of geophysical data.},
author = {Herrero, Julien and Caumon, Guillaume and Bodin, Thomas},
doi = {10.1007/s11004-025-10262-w},
hal_id = {hal-05484971},
hal_version = {v1},
journal = {{Mathematical Geosciences}},
keywords = {Well correlation ; Uncertainty quantification ; Stratigraphy ; Reservoir modeling ; Inverse problems},
month = {January},
publisher = {{Springer Verlag}},
title = {{Transdimensional Inversion of Well Log Data for a Two-Dimensional Geological Model with Inclined Layers and Petrophysical Lateral Variability}},
url = {https://hal.univ-lorraine.fr/hal-05484971},
year = {2026}
}