Climate uncertainty as an integral part of integrated assessment models

Cost-benefit integrated assessment models (IAMs) such as the Dynamic Integrated model of Climate and the Economy (DICE) are often used to assess the social cost of carbon (SCC), the marginal damage arising from each additional ton of emitted CO₂. The climate component of such IAMs has recently come under increased scrutiny. Alongside ensuring that economists are getting climate dynamics correct, the uncertainty in the climate system should be embraced, as it greatly influences the appropriate SCC and CO₂ emissions mitigation pathway.

We use DICE, replacing its native climate module with the Finite-amplitude Impulse Response (FaIR) model (v2.1). FaIR is assessed to be fit-for-purpose for evaluating emissions projections from IAMs by the IPCC, and has an advantage over the native DICE module in that carbon cycle feedbacks are included. The FaIR emulator has been calibrated to CMIP6 models and constrained such that its projections are consistent with historical global mean temperature change, atmospheric CO₂ concentration and ocean heat content, and IPCC Sixth Assessment Report assessed uncertainty ranges for equilibrium climate sensitivity (ECS), transient climate response and non-CO₂ effective radiative forcing, constructing a 1000-member posterior ensemble from a 1.5 million member prior. Three ensembles are produced: a Nordhaus “socially optimal” ensemble with median 2100 warming of around 2.8°C, somewhat consistent with current Nationally Determined Contributions; a 2°C-consistent ensemble; and a 1.5°C-consistent ensemble. We update the economic and climate baseline in DICE/FaIR to 2023 and use a 3-year model timestep. The three scenarios are constructed solely by modifying the discount rate.

The influence of climate uncertainty is profound, having a factor of 5 uncertainty (5-95% range) in the social cost of carbon for a 1.5°C consistent ensemble, and a factor of 3 uncertainty in the business as usual case. There is also a very strong positive correlation between the SCC and the ECS, which re-confirms earlier analysis that reducing climate system uncertainty can realise net present economic benefits by guiding appropriate choices for the SCC. 

Alongside calculating a SCC for the year 2023, DICE/FaIR computes probabilistic projections of socially “optimal” CO₂ pathways for each scenario that also show substantial variation depending on the climate configuration (for example, -14 to +11 GtCO₂/yr in 2050 for the 1.5°C ensemble) but are broadly consistent with findings from the IPCC Sixth Assessment Working Group 3 report in the median case (such as global net zero emissions required in the 2050s to meet 1.5°C). The range of socially optimal emissions pathways consistent with a specific temperature threshold also highlights a climate-socioeconomic feedback: if climate sensitivity is high, mitigation efforts must be strong to limit future warming and climate damages. This feedback, while implicitly included in cost-benefit IAMs such as DICE, are not typically present in process-based IAMs used to construct emissions scenarios for use by the IPCC or climate models such as the Shared Socioeconomic Pathways. We claim that including climate and climate uncertainty in these process-based IAMs will improve emissions scenarios.