All Paris-aligned pathways for limiting warming to 1.5°C require large-scale deployment of carbon dioxide removals (CDR) by mid-century. CDR technologies can offset hard-to-abate and dispersed emissions where removal may ultimately be less costly than further mitigation, and they may also enable a return of emissions to 1.5°C following an overshoot. Technical removals are particularly important because of their potential scale and ability to lock up CO₂ on a geological timescale. However, the technical CDR sector is nascent at present, with the costs of bioenergy with carbon capture and storage (BECCS) and direct air capture (DAC) very high. In the short term, this makes these technologies prohibitively expensive to integrate and scale consistent with Paris-aligned goals. Policies are therefore needed to support their development and rapidly reduce their costs.

At present, policy analysis commonly estimates the learning rates of BECCS and DAC technologies based on single factor learning curves (SFLCs). These estimates are often calculated by benchmarking against analogous technologies and are subsequently incorporated into whole-energy systems models. These analyses of learning rates are fundamental for policy design as they shape estimates of future technology costs and, consequently, the scale, timing and form of targeted public policy interventions.

This policy briefing makes the case that orthodox approaches to learning curve analysis can blind policymakers to the complexity inherent in technology learning, leading to a risk that poorly informed policies will be developed without the application of more sophisticated approaches. We propose a ‘deep uncertainty approach’ to address the fundamental flaws in orthodox learning curve analysis. We seek to establish this process in the literature, and invite others to utilise it, adapt it and critique it.