The Generation of Flow-Representative Relative Permeabilities from Experimental Data to Model Viscous Fingering in CO2 Storage

Accurate modelling of CO₂ storage and the design of carbon capture and storage (CCS) scenarios require reliable CO₂/water relative permeabilities (Relperms) for numerical simulation. These functions govern gas transport, plume evolution, and the risk of CO₂ breakthrough in wells. However, our recent work (Larki et al., 2025a) demonstrated that published conventional CO₂/water Relperms fail to generate viscous fingering, even under strongly viscous-dominated conditions. This exposes a key limitation of standard Relperm models for CCS performance assessment.

To address this issue, we developed a novel methodology to generate “viscous fingering Relperms” for the CO₂/water system using numerical reconstruction of core flood experiments. The approach adapts the method of Sorbie et al. 2020), originally proposed for water–oil systems. Here, it is applied to the experimental data of Bennion and Bachu (Bennion et al., 2006), who reported CO₂/water Relperms at 1378 kPa (200 psi) and 20,000 kPa (2900 psi). Despite adverse viscosity ratios (μ _w/μ _CO2 = 36.47 and 8.85 for the low- and high-pressure cases, respectively), their experimentally derived Relperms did not exhibit fingering behaviour. The workflow consists of three steps: (i) selecting published conventional Relperms derived under standard 1D core-scale assumptions; (ii) numerically reproducing the corresponding unsteady-state core flood to generate synthetic pressure-drop and production data; and (iii) re-simulating the experiment in 2D (or 3D) to history-match the synthetic data while allowing fingering to develop, thereby generating modified “fingering” Relperm functions, following the methodology of Sorbie et al. 2020) and subsequent applications (Beteta et al., 2022a, 2022b, 2024a; Salmo et al., 2022). The resulting Relperms match the synthetic experimental responses and produce pronounced viscous fingering patterns consistent with the adverse mobility ratios.

Finally, applying Rapoport's scaling theory (Rapoport, 1955), we upscale the core-scale model to the field-scale aquifer sector. The simulations demonstrate persistence of fingering morphology across scales, with preserved dimensionless pressure and production behaviour, confirming the importance of fingering-aware Relperms in CCS modelling.