From {2D} {Profiles} to {3D} {Heterogeneity}: scalable topologically constrained reconstruction.

@inproceedings{Benallal2026RM,
 abstract = {Fluid flow in subsurface reservoirs depends not only on the geometry of reservoir bodies but also on their topology, which controls fluid connectivity and drainage pathways. Static connectivity and continuity have long been studied, and numerous topological metrics have been proposed to characterise reservoir architectures under different depositional scenarios. In geothermal studies, however, subsurface characterisation is often limited by sparse well data, making seismic interpretation a key tool for reservoir analysis. Recent advances in artificial intelligence (AI) and machine learning now enable quantitative workflows capable of deriving petrophysical properties, such as porosity, directly from seismic data. At the same time, the growing demand for low-carbon energy has accelerated the acquisition of new seismic surveys. While 3D seismic cubes provide full 3D heterogeneity information, they remain costly and timeconsuming to acquire and process. An alternative approach is therefore to infer 3D reservoir architectures from a limited number of 2D seismic profiles. This challenge is addressed within the PR3DICT project, whose case study focuses on the Dogger Formation of the Paris Basin, a carbonate reservoir composed of oolitic shoals and targeted for geothermal applications, for which a full 3D seismic cube is available. This work presents approaches for identifying plausible 3D reservoir models consistent with geological knowledge and available 2D profiles, while addressing the broader question of how many 2D profiles are required to reconstruct geologically realistic and topologically consistent 3D architectures. The study aims to establish quantitative relationships between the topology observed in 2D profile networks and that of the corresponding 3D reservoir models.},
 author = {Benallal, Cheryne and Viseur, Sophie and Bonte, Damien and Issautier, Benoît and Bordenave, Aurélien and Stopin, Alexandre},
 booktitle = {2026 {RING} meeting},
 language = {en},
 pages = {162--180},
 publisher = {ASGA},
 title = {From {2D} {Profiles} to {3D} {Heterogeneity}: scalable topologically constrained reconstruction.},
 year = {2026}
}