After the preliminary site selection step for geological storage of CO₂, assessing site performance remains the main objective. Site performance assessments provide us with information on whether the site is safe and how much CO₂ can be stored. But how does site performance assessment work?

The main tool for site performance is modelling and simulation of the deep subsurface. To get started on this, we first need to frame the starting point: the local geological context. The site needs to be described as detailed as possible in computational terms, especially the reservoir layer used to store the CO₂. This is the first model representation of the storage complex, or the baseline or reference dataset. As data on underground characteristics are present but never complete, there will be uncertainties in this early characterisation.

Next, we try to figure out what happens when CO₂ is injected into the reservoir over a longer time under various scenarios. We do this by telling the preliminary model that CO₂ is injected, which then predicts the storage behaviour. By applying different injection scenarios, we can anticipate the range of actual behaviour as exhaustively as possible. We figure out likely CO₂ migration pathways and trapping mechanisms, but also identify whether there are possible ways for leakage to occur out of the reservoir. In this way, we get a first idea of the risk profile and capacity of the CO₂ reservoir. Remember, at this point no CO₂ has been injected in the reservoir yet. Only if the site characterisation and performance assessment indicate that under all imaginable circumstances CO₂ injection is safe and leakage-free, should CO₂ injection actually be permitted.

Although site performance models are rather complex, there are two essential parameters that it needs to simulate well: the reservoir pressure and the extent of the CO₂ plume -- and these are fundamentally dependent upon the nature of the local geology. The essential step for these parameters is to compare actual monitoring results with the outcome of the modelling.

Reservoir pressure is a system driving force and can induce stress variations in the reservoir that potentially could cause reactivation in a pre-existing fault, or for a fracture to develop or open, thereby leading to undesirable leakage pathways. Reservoir modelling and simulation will identify threshold pressures to avoid fault reactivation. Downhole pressure measurements can be used to calibrate the reservoir modelling, but other techniques are also possible. At the In Salah project in Algeria, InSAR satellite imaging measuring ground surface displacement was used to indirectly verify the 3D field pressure evolution, leading to a good fit between the computed performance assessment (pressure induced geomechanical effects) and the actual measurements (see Figure 1). 

As the reservoir pressure is affected beyond the limits of the CO₂ plume,  other techniques can be used to identify where the CO₂ actually resides within the reservoir. The extent of the CO₂ plume modifies the ground acoustic impedance in a more direct way than pressure. This characteristic is used in seismic monitoring, which picks up acoustic changes in the underground. Results from such seismic monitoring conducted over time can be used to map the evolution of the CO₂ plume. This technique has been used very successfully in the Sleipner project in Norway, where early site performance models were adjusted substantially to match the monitoring results. Monitoring enables better dynamic modelling development leading to better prediction for both the short and the long term, and site assessment and monitoring are therefore an inseparable pair – for responsible CO₂ storage, you need both.

In practice, early performance assessment scenarios tend to 'pragmatic' predictions to reduce risks and exclude leakage. After a series of iterations fed by new monitoring inputs, models are improved and become more and more realistic. Absolute certainty on long-term site performance can never be achieved, but through the combination of performance and risk assessment with increasingly better monitoring data we can get as close as possible.