Exploring Convergence and Its Limits in Case B of the 11th SPE Comparative Solution Project

The 11th SPE Comparative Solution Project (SPE11) was established to benchmark the ability of reservoir simulators to capture the key physical processes of geological carbon dioxide (CO₂) storage. A central question in this project was whether grid refinement could yield converged solutions for quantities of practical interest. In this work, we revisit Case B, perceived to be the easiest benchmark, using the open-source simulator OPM Flow (The Open Porous Media Initiative). Building on the most highly resolved submission to SPE11, we conduct systematic refinement studies on facies-aligned corner-point grids, extending the resolution down to the submeter scale and beyond in a localized subdomain. Convergence is assessed not for dense reporting fields, but rather for the 13 sparse measures defined in SPE11, which act as proxies for operational risks such as overburden disturbance, leakage, dissolution dynamics, and migration into low-permeability seals. We find that pressures at observation points converge robustly even on coarse grids, whereas measures related to phase partitioning and dissolution exhibit strong grid sensitivity, and in several cases, no clear convergence, even at the finest resolutions considered. In particular, the onset and development of convective dissolution fingers remain highly resolution-dependent, underscoring the challenges of resolving small-scale processes in field-scale models. These results highlight that while some operationally critical quantities can be simulated with confidence, others remain elusive under straightforward grid refinement. The findings emphasize the need for complementary approaches—such as upscaled models or proxy measures—to ensure reliable predictions for large-scale CO₂ storage operations.