Reservoir-Scale Caprock Seal Integrity for CO2 and Hydrogen Storage: A Critical Review of Upscaling Methods and Uncertainty Propagation

Reservoir-scale caprock seal integrity during geological CO2 and underground hydrogen storage (UHS) depends on processes occurring well below the scale at which a safety assessment typically treats mineral dissolution and precipitation, capillary entry pressure at grain contacts, and stress-dependent pore structure. This review addresses one question precisely: how is finer-scale hydromechanical alteration translated into defensible continuum-scale parameters and reservoir-scale predictions, and what information is lost in that translation? We distinguish pore-scale observation, corescale measurement, constitutive closure, and reservoir-scale prediction as categorically different evidence types rather than treating them as interchangeable. Published experimental and simulation studies consistently document localized, order-ofmagnitude permeability increases along preferential dissolution pathways at small scale; the small number of studies that have attempted the full upscaling chain against an actual site report comparatively small net bulk change over decades to a century. We treat this as an open scale-dependence problem rather than a resolved discrepancy, and evaluate candidate explanations, including volume averaging, mineral buffering, connectivity and percolation thresholds, and flow-regime dependence via the Damköhler and Peclet numbers, against what published evidence demonstrates versus what remains a hypothesis. CO2 and H2 are treated as physically distinct problems, compared on molecular size, interfacial tension, capillary entry pressure, and reactivity, before their reservoir-scale outcomes are compared. We propose a physics-consistent workflow connecting characterization to reservoir-scale seal-integrity prediction with explicit uncertainty propagation and identify where physics-informed machine learning could prospectively, not presently, contribute. Claims not supported by verified evidence are flagged explicitly rather than filled with plausible-sounding literature.