The transient climate response (TCR), defined to be the warming in near‐surface air temperature after 70 years of a 1% per year increase in CO2, can be estimated from observed warming over the nineteenth and twentieth centuries. Such analyses yield lower values than TCR estimated from global climate models (GCMs). This disagreement has been used to suggest that GCMs' climate may be too sensitive to increases in CO2. Here we critically evaluate the methodology of the comparison using a large ensemble of a fully coupled GCM simulating the historical period, 1850–2005. We find that TCR estimated from model simulations of the historical period can be much lower than the model's true TCR, replicating the disagreement seen between observations and GCM estimates of TCR. This suggests that the disagreement could be explained entirely by the methodology of the comparison and undercuts the suggestions that GCMs overestimate TCR.

No abstract