Well grounded? The assumptions behind CO2 storage in climate change mitigation pathways

Context

To combat climate change, we need to consider how economies, energy systems, land use and the climate itself interact over decades. To do this we turn to advanced computer models called Integrated Assessment Models (often abbreviated to IAMs), which map out a host of different pathways society could take to meet a defined climate target. Give an IAM a temperature goal and it works out the carbon budget that goes with it, then searches for the most efficient combination of technologies and policies that stays inside that budget. The results and understandings learned from IAMs are hugely impactful: underpinning IPCC reports, prominent scientific publications and ultimately feeding into policy.

One thing that many IAM pathways agree on is that we will need an assortment of different technologies and behavioural changes to meet climate targets. These include reducing emissions at source, reducing demand for carbon-intensive products, but crucially, also isolating residual emissions from the atmosphere. Storing carbon dioxide (CO₂) deep underground is a way to achieve this, dealing not just with residual emissions (say, from industrial processes or dispatchable power) but also enabling technologies that remove CO₂ already in the atmosphere. During the last IPCC assessment reporting cycle, most pathways submitted by the modelling community agreed that global CO₂ storage efforts needed to ramp up quickly, from 1 Gigatonne (Gt) per year in 2030 to 7 Gt per year in 2050 and 15 Gt per year in 2100 (Figure 1). This represents an eye-watering increase from current levels, with just 0.045 Gt stored worldwide in 2023: translating to a scale-up of more than 300-fold.

What we did and found

Whether the technology can scale up this fast has been questioned by some groups; others have asked whether the rates in IAM pathways are realistic under current assumptions. We set out to try and better understand the latter, which is a challenging task because IAMs are complex models, exhibit different degrees of transparency and are continually being updated and improved (a moving target!). We turned to model documentation and manuals, scientific articles, various GitHub repositories and, crucially, colleagues within the IAM community, without whom this task would have been even harder. That last point mattered more than we expected: source code was publicly available for four of the six models we studied, but for all but one, the input files we actually needed could only be obtained by asking the modelling team directly. From this we were able to unravel what levers six particular IAMs currently pull to ensure their carbon dioxide storage predictions are feasible.

We discovered that IAMs deploy sophisticated means to constrain CO₂ storage projections. All the models we studied essentially used three mechanisms: limits on storage potential (the maximum amount that can be stored in a region), scale-up limits (the maximum amount that can be stored per year) and cost assumptions (how much a tonne of CO₂ costs to store). The second of these is genuinely new, and encouraging: four of the six models have added annual limits since the last IPCC report, in direct response to criticism that their earlier projections grew implausibly fast.

But we also discovered there were tangible differences in both the numbers and methodological approaches used to define these. Storage potential limits vary drastically, even for a single country or region - for North America alone, the limits adopted range from 230 to 8,333 Gt, but this speaks to disagreement in the underlying literature base (Figure 2). Scale-up limits are often single values that are held constant and don’t account for non-linearities often associated with the deployment of new technologies. Cost assumptions were largely dated and underestimated relative to the latest understanding: most models assume $2-20 per tonne stored, drawn from studies over a decade old, while recent assessments that account for monitoring requirements reach $35 per tonne, and those that account for pressure build-up in some instances reach $147 (Figure 3).

What are the implications of our findings?

From these learnings we extracted some recommendations for the IAM and broader scientific community. Firstly, IAM predictions are snapshots: the models are continually being updated, and modelling results are quickly superseded. Every model we examined has changed since the runs that are still most widely quoted, and several of those changes act to reduce projected deployment. Secondly, the scale-up rates required during the last IPCC reporting cycle are not compatible with the recent literature in this space. Thirdly, more reporting of project costs would be beneficial if we are to build realistic cost assumptions into longer-term models; when one group revisited storage costs in their own model, projected global storage fell by over 100 Gt. Fourthly, annual limits should be anchored on how quickly an industry can realistically grow, rather than on how much space exists underground. Two of the models currently set their rate limits as a share of regional storage potential, yet technoeconomic and supply-chain factors are increasingly understood to be the binding constraint. Finally, most models require CO₂ to be stored in the region where it was captured, so emerging cross-border arrangements - such as those between Japan, South Korea and prospective storage hosts in Australia and Southeast Asia - are invisible in these pathways.

Underneath all of this sits an important point. The constraints inside these models are only as good as the data the CO₂ storage community hands over. Project-level storage costs are barely published. Regional storage assessments are rarely produced in a form a modelling team can pick up and use. Where they have been - the recent global storage assessment by Gidden et al. is the clearest example - models began adopting them almost immediately. Getting that pipeline right will do more for confidence in IAM projections than any single number we could argue over.