Carbon sequestration: what a CO2 storage site screening study must answer on capacity, injectivity and containment

A CO₂ storage site screening study has to answer three questions before anyone spends money on a well: how much CO₂ the rock can hold (capacity), how fast it can be pushed in without overpressuring the rock (injectivity), and whether it will stay there for centuries (containment). A site that fails any one of them is not a storage site, however good the other two look. This guide sets out what each question needs, which public frameworks define it, and where a screening study usually goes wrong.

Why carbon sequestration screening starts with three questions

Geological storage puts CO₂ into porous rock deep enough that it is dense: depending on the geothermal gradient, from about 800 m down injected CO₂ is in a dense supercritical state, with a liquid-like density, so far more of it fits in the same pore space than as a gas [1]. The IPCC's special report lists the ways it stays there: beneath an impermeable caprock, as a residual phase trapped in the pores, dissolved in the formation water, adsorbed on organic matter in coal and shale, and turned into carbonate minerals [1]. For carefully selected and well-managed sites, the report considers it likely that 99 % or more of the injected CO₂ will be retained for 1,000 years [1]. The weight is on carefully selected: screening is how you select.

Regulators frame it the same way. Under the EU CCS Directive a formation may only be chosen as a storage site if there is no significant risk of leakage and no significant environmental or health risk, judged by characterising the storage complex against the criteria of its Annex I [2]. Annex I has three steps: data collection, a three-dimensional static geological earth model, and then the dynamic behaviour, sensitivity and risk of the store [2]. A screening study is the first, cheaper pass through the same three steps.

CO2 storage capacity estimate: volumetric first, dynamic next

The first capacity number is volumetric: the area of the formation, times its thickness, times its porosity, times the density of CO₂ at reservoir conditions, times a storage efficiency factor for the fraction of pore space the CO₂ can actually reach. The US Department of Energy methodology, used for national and regional atlases, is built this way and samples the efficiency factor with a Monte Carlo method to give a range rather than a single figure [3].

The efficiency factor carries most of the uncertainty. Bachu's review found published values from under 1 % to over 10 %, so no single number can be applied everywhere; volumetric estimates suit regional evaluations and should use the low end of that range, while site-scale estimates need dynamic numerical simulation that includes operational and regulatory limits [4]. The same review notes that storage efficiency is first limited by pressure and only later by space, and that in open saline aquifers a storage operation is most likely pressure-limited for its whole life [4]. In such aquifers the binding constraint is usually the pressure the rock can take, not the pore volume the plume could fill.

How the number is reported matters as much as the number. The SPE CO₂ Storage Resources Management System (SRMS), approved in 2017, classifies storable quantities by project maturity and states their uncertainty, in the way the petroleum industry reports reserves and resources [5]. A screening study should say which class its estimate sits in and give a low, best and high case, not a single tonnage.

Injectivity and pressure build-up

Injectivity is how much CO₂ a well can take per day at a pressure the rock tolerates. It depends on permeability times thickness near the well, on how the layers connect vertically, on the water the CO₂ has to push aside, and on how the pressure dissipates through the wider formation. The EU Directive asks the dynamic model to cover possible injection rates, the pressure and temperature of the formation over time as the cumulative amount grows, and the risk of fracturing the storage formation and the caprock [2].

Two things often go wrong at screening stage. First, the pressure limit is set from a regional gradient when a local stress measurement would give a different answer. KAI's Utah FORGE geothermal case shows how much that matters for any injection project: with the regional stress gradient the highest rate the well pair can sustain for 30 years below the minimum stress is 9.8 kg/s (P10–P90 9.4–10.7), but with the minifrac measurements from a neighbouring well taken as measured it falls to 4.9 kg/s. These are simulated results for hot water in granite, not CO₂, but the lesson transfers: lead with the cautious limit and say which stress data it rests on.

Second, vertical permeability is taken from core plugs. In KAI's Zeerijp-3A case, 78 vertical–horizontal plug pairs give a median kv/kh of 0.95, while ripple-laminated bedding calibrated on the same plugs and upscaled by flow simulation gives kv/kh of 1 : 20 to 1 : 360 at 1 m for 20–40 % mud (the bedding is simulated, the plugs are real). For CO₂, which rises under buoyancy, the direction of the error is fairly clear. A model given the plug value lets the CO₂ rise quickly to the caprock and gather in one thin tongue beneath it. With bed-scale kv/kh, mud laminae act as baffles: the CO₂ rises more slowly, spreads sideways beneath each baffle in stacked layers, and usually contacts more rock and brine, which favours residual and dissolution trapping but changes the plume's footprint. Weaker vertical communication can also concentrate the pressure rise in the injected layers and lower the injectivity of a thin completion. How large these effects are, and whether capacity rises or falls overall, depends on the site and has to be tested by simulation, but a model built on plug kv/kh gets the plume shape and the pressure wrong.

Caprock integrity screening and containment

Containment is the question regulators weigh most. A screening study should check, at least:

• The seal. Is there a continuous, low-permeability caprock over the whole area the plume and the pressure front will reach, and is its capillary entry pressure high enough for the CO₂ column expected beneath it?
• Faults. Which faults cut the seal, how they are oriented in today's stress field, and how much extra pressure would bring them closer to slip.
• Wells. Every old well that penetrates the seal inside the pressure footprint, with its age and abandonment record. The Directive names abandoned or poorly sealed wells as a leakage risk the dynamic model must address [2], and the raised pressure that could drive fluids up such a well can extend many kilometres from the injector, further than the plume itself [6].
• Secondary containment. Whether a second seal or a dissipating formation sits above the primary one [2].

Caprock integrity also depends on the operation: cooling of the rock near the well and the pressure rise both change the stress on the seal, so containment cannot be judged from the static geology alone.

CCS reservoir simulation: what the dynamic model must show

Annex I of the EU Directive gives a useful list of what carbon sequestration reservoir modelling has to report, even outside Europe: the extent of the CO₂ over time, its flow and phase behaviour, trapping mechanisms and rates, storage capacity and pressure gradients, the risk of CO₂ entering the caprock, displacement of formation fluids, changes in fluid chemistry, and induced seismicity and surface elevation change [2]. It also asks for a sensitivity characterisation: several simulations that vary the uncertain parameters, so the result is a range, not a single realisation [2].

That last point is where most CCS reservoir simulation at screening stage is weakest. A single deterministic model built on averaged properties will give one plume and one pressure curve, and no sense of how wrong either could be. An ensemble of geological models, each consistent with the wells, gives P10–P90 bands for plume extent and pressure, and shows which unknown matters most and is worth a well or a test to reduce.

Monitoring (MMV) from the first day

Measurement, monitoring and verification (MMV) is not something to design after injection starts. The Directive's Annex II requires a monitoring plan for the baseline, operational and post-closure phases, and, when observations deviate significantly from the predictions, a recalibration of the 3D model by repeating the modelling and risk steps [2]. So the screening model should be built from the start to be updated: the same model that predicts the plume should be the one that is later checked against pressure gauges, seismic surveys and surface deformation.

A screening checklist

1. Depth, temperature and pressure: is the CO₂ dense at reservoir conditions?
2. Capacity as a range, with the efficiency factor justified, and its SRMS class stated [5].
3. Injectivity per well and the number of wells needed for the target rate, with permeability from logs and core and an honest kv/kh.
4. Pressure build-up over the injection period compared with the fracture pressure of reservoir and seal, using local stress data where they exist. Annex I asks for fracture pressure and the risk of fracturing but sets no numeric margin [2]; use the margin the regulator in your jurisdiction applies.
5. Seal continuity, faults and legacy wells inside the pressure footprint, not only the plume footprint.
6. An ensemble of models with P10–P90 for plume and pressure, and the main sensitivities.
7. A baseline and monitoring concept that the model can later be calibrated against.

What KAI offers today and what it does not

To be plain: KAI has not carried out a CO₂ storage project and has no published CO₂ storage case. In KAI-GeoEnergy, CO₂ and hydrogen storage studies are in build or available on request; the live preview today is the Reservoir Lab for geothermal doublets, which runs in the browser and queues full simulations on the KAI engine. Fault-slip, subsidence and caprock modules are also still in build.

What KAI has published, with open notebooks, are the methods a storage screening study leans on:

• Permeability where there is no core. In the Zeerijp-3A case, three rock types put 96 % of plugs within a factor of 3 of the measured permeability, and a log-based prediction tested blind puts 60 % within a factor of 3 and 93 % within a factor of 10, with a P10–P90 band. Run it on your wells with KAI-Subsurface.
• Vertical permeability from plug to bed. The Plug-to-Block Lab builds the bedding and upscales it by flow.
• Facies between wells as an ensemble. The Heterogeneity Lab conditions facies realisations on every log, on Dutch BRO borehole data and TNO GeoTOP.
• Flow, heat and stress with uncertainty. The Utah FORGE case runs a 20-member ensemble forward 30 years and reports P10–P90, including the injection limit set by the minimum stress (simulated).

KAI works remotely with your team and your data. It does not drill, hire equipment or run field acquisition, and it does not replace the operator's or regulator's own assessment. If you have a candidate site, read how KAI approaches subsurface energy or see what KAI-GeoEnergy does today, then tell us what data you hold and we will say what we can do now.