Transdimensional inversion of well log data for {2D} modeling of clinoform structures
in: 2025 {RING} meeting, pages 161--185, ASGA
Abstract
This work extends the transdimensional inversion framework introduced by Herrero et al. (2025) to better account for complex clinoform geometries in layered reservoir models. Traditional inversion approaches often rely on fixed or simplified layer geometries manually interpreted by experts, introducing biases related to subjective geological interpretations. In contrast, the proposed method infers both the number and the shape of layers directly from well data using a reversible jump Markov chain Monte Carlo (RJMCMC) algorithm. To model the curved and dipping interfaces characteristic of clinoform systems, we implement a flexible parameterization based on Bézier curves, using the B3D package (Baville et al., 2022). This allows the interface geometry to deviate from simple linear shapes and better mimic stratigraphic structures encountered in sedimentary environments. The inversion is applied to a 2D case study combining real and synthetic elements. A clinoform-like geometry is extracted from a seismic section of the Netherlands F3 block, and permeability values are randomly assigned in the different layers to define a heterogeneous ground reference model. Well log data from multiple well locations are used to drive the inversion. Despite convergence challenges—which could be addressed through enhanced sampling strategies such as parallel tempering—the results demonstrate the potential of the method to recover both the geometry and permeability distribution of clinoform systems. These findings highlight the value of Bézierbased interface modeling in stratigraphic inversion and open the door to flexible transdimensional inversion of complex stratigraphic architectures.
Download / Links
BibTeX Reference
@inproceedings{Coutard2025RM,
abstract = {This work extends the transdimensional inversion framework introduced by Herrero et al. (2025) to better account for complex clinoform geometries in layered reservoir models. Traditional inversion approaches often rely on fixed or simplified layer geometries manually interpreted by experts, introducing biases related to subjective geological interpretations. In contrast, the proposed method infers both the number and the shape of layers directly from well data using a reversible jump Markov chain Monte Carlo (RJMCMC) algorithm. To model the curved and dipping interfaces characteristic of clinoform systems, we implement a flexible parameterization based on Bézier curves, using the B3D package (Baville et al., 2022). This allows the interface geometry to deviate from simple linear shapes and better mimic stratigraphic structures encountered in sedimentary environments. The inversion is applied to a 2D case study combining real and synthetic elements. A clinoform-like geometry is extracted from a seismic section of the Netherlands F3 block, and permeability values are randomly assigned in the different layers to define a heterogeneous ground reference model. Well log data from multiple well locations are used to drive the inversion. Despite convergence challenges—which could be addressed through enhanced sampling strategies such as parallel tempering—the results demonstrate the potential of the method to recover both the geometry and permeability distribution of clinoform systems. These findings highlight the value of Bézierbased interface modeling in stratigraphic inversion and open the door to flexible transdimensional inversion of complex stratigraphic architectures.},
author = {Coutard, Aubin and Herrero, Julien and Baville, Paul and Caumon, Guillaume},
booktitle = {2025 {RING} meeting},
language = {en},
pages = {161--185},
publisher = {ASGA},
title = {Transdimensional inversion of well log data for {2D} modeling of clinoform structures},
year = {2025}
}