Will the IPCC’s Sixth Assessment Report Rely on Increasingly Overheated Models?

In a letter published last week in the journal Nature, Jiang Zhu and Christopher Poulsen of the University of Michigan and Bette Otto-Bliesner of the National Center for Atmospheric Research caution that models used in the forthcoming Sixth Assessment Report of the United Nations Intergovernmental Panel on Climate Change (IPCC) may have unrealistically high equilibrium climate sensitivity estimates. This arcane-sounding issue is a very big deal.
 
Equilibrium climate sensitivity (ECS) is the long-term change in global mean surface temperature (GMST) after a doubling of atmospheric carbon dioxide (CO2) concentration. ECS is a key input to climate change impact assessments. The more sensitive the climate, the larger the potential warming from a given increase in CO2 concentration, hence the larger the potential impacts on weather patterns, sea levels, crop yields, economic growth, biodiversity, and human health.
 
Far from being “settled,” climate sensitivity remains an active area of investigation and debate. As Zhu and his colleagues note, “ECS is poorly constrained with a ‘likely’ range of 1.5°–4.5°C, which has remained nearly unchanged since the Charney report 40 years ago.”
 
An organization called the World Climate Research Program runs a “climate model inter-comparison project” (CMIP) to develop, test, and finalize the ensemble of models used by the IPCC to assess climate change impacts. The IPCC’s 2013 Fifth Assessment Report used CMIP5 models. The Sixth Assessment Report will use CMIP6 models.
 
Zhu et al. report that 10 in 27 of the currently available CMIP6 models “have an ECS higher than the upper end of [the 1.5°-4.5°C] range,” in contrast to just 2 in 28 CMIP5 models. For example, the ECS in the Community Earth System Model version 2 (CESM2) is 5.3°C.
 
To test the realism of CESM2, the authors ran it with CO2 concentrations three times higher than pre-industrial level, or 855 parts per million. They then compared the model’s projected global mean surface temperature to proxy data estimates of GMST during the Early Eocene Climate Optimum, a period when CO2 concentrations of approximately 1,000 ppm persisted for millions of years.
 
The CESM2-estimated global mean surface temperature was 37.5°C, or “5.5°C greater than the upper end of proxy temperature estimates for the Early Eocene Climate Optimum.” Moreover, the modeled tropical land temperature exceeded 55°C, “which is much higher than the temperature tolerance of plant photosynthesis and is inconsistent with fossil evidence of an Eocene Neotropical rainforest.”
 
The authors note that the CESM2 model “is among the best-performing CMIP6 models based on mean pattern correlations of a variety of climate fields.” In other words, the model hind-casts 20^th century climate patterns better than most. Yet its high ECS estimate is “incompatible with [the] known Eocene greenhouse climate.”
 
The authors draw two conclusions from those results. First, though their analysis is limited to CESM2, they “expect that other models with similarly high ECS may also be biased too warm when driven by high levels of atmospheric CO2.” Second, their study “illustrates that the development and tuning of models to reproduce the instrumental record does not ensure that they will perform realistically at high CO2.”