Underground hydrogen storage in porous formations will require seismic monitoring, and the interpretation of that monitoring rests on fluid-substitution modelling that assumes pore pressure equilibrates within a wave period. Where hydrogen and brine coexist as mesoscopic patches, that assumption fails, and the rock response becomes frequency dependent. We investigate whether including that frequency dependence materially changes the detectability interpretation that a conventional low-frequency description would produce, under conditions representative of Dutch sandstone storage targets.
We define three reference cases spanning the Dutch parameter envelope, a constant-frame Control case and two pressure-sensitive-frame cases in the style of the Rotliegend and Bunter sandstones, and establish their low-frequency Gassmann reference response. We model the frequency-dependent response with the Johnson branching-function description of mesoscopic wave-induced fluid flow, validate the implementation against exact reference solutions and the White concentric-sphere benchmark, and map how permeability and patch radius place the relaxation peak relative to the seismic band. We then embed each case at its true reservoir depth in a laterally invariant layer-cake column, generate synthetic prestack records, and migrate them to depth, comparing an all-brine state, a relaxed low-frequency state, and a dispersive state under uniform, gravity-segregated, and randomised per-layer saturation topologies, beneath both a single-block and a realistic layered Dutch overburden, and under band-limited noise at prescribed signal-to-noise ratios.
Our results show that the answer to the question we pose is conditional and casedependent. Where the relaxation peak falls inside the seismic band, which among our cases means the larger-patch Bunter-style geometry, the dispersive correction survives imaging and noise at field-attainable signal-to-noise ratios, reaches ten percent of the reference reflector amplitude, and opposes the bulk fluid-substitution response, so that a low-frequency description overpredicts the imaged monitoring signal by a factor of between 1.3 and 2.0 across 15 to 50 Hz. Elsewhere the low-frequency description is accurate to within a few percent, and the full monitoring signal remains recoverable under noise in every case, so hydrogen injection itself is seismically visible throughout. Our results describe the fluid effect alone, under a dispersion-only, multiple-free, laterally invariant treatment, and are an upper bound on what a field survey could recover; within that scope, we identify where in Dutch-reservoir parameter space frequency-dependent rock physics is worth carrying, and where a conventional low-frequency workflow is sufficient. ... Read more