Unpacking uncertainty in Carbon Dioxide Removal requirements

Deep uncertainty about the costs and resource limits of carbon dioxide removal (CDR) options challenges the design of robust portfolios. To address this, we identified key uncertainties in CDR pathways and developed the CDR-SPEC model, a mixed-integer linear optimization model for cost-optimal and time-dependent CDR portfolios with endogenous treatment of technology cost dynamics. Within this framework, we sampled the option space to explore the impact of input parametric uncertainty on the composition and performance of CDR portfolios. The resulting database contains detailed information about how varying combinations of uncertainty conditions trigger the implementation of different CDR portfolios. What is missing is an understanding of which factors drive large variability in the outcomes, where outcome is understood as any metric of performance, without making assumptions about their desirability.

To shed light on this, we recur to the concept of entropy, a measure of the uncertainty in a distribution. We use this as a proxy for guiding an exploration strategy that aims at maximising the amount of information gained about a desired outcome, providing a comparative assessment of each uncertain parameter’s contribution to an outcomes distribution. This assessment shows that among all parameters represented in CDR-SPEC, cumulative (i.e., from 2020 to 2100) removal requirements (CRR) drives the largest entropy reductions across a series of outcomes. The interpretability of this result is nevertheless challenged by the multitude of uncertainty dimensions this parameter englobes: it represents both scenario uncertainty (i.e., how much abatement takes place for different greenhouse gases) and climate response uncertainty (i.e., potential additional CDR incurred when considering beyond-median warming outcomes). Unpacking these two dimensions bundled under CRR would allow highlighting the relative impact of key uncertainties in the science that informs CDR-deployment policies.

Representing all three dimensions of uncertainty (i.e., CDR-specific, scenario and climate response uncertainty) requires expanding our understanding of the impacts of different CDR approaches on global temperatures under varying assumptions on how the earth system responds to emissions. This could be achieved by, for a fixed illustrative mitigation pathway and set of CDR-specific parameters, iterating the results from the CDR portfolio analysis in a simple climate emulator until the removals required for climate stabilisation and the removals delivered by the CDR portfolio converge. For many illustrative mitigation pathways and sets of CDR-specific parameters, this results in a database which disentangles all three dimensions of uncertainty mentioned above.