The methods to quantify equilibrium climate sensitivity are still debated. We collect millennial‐length simulations of coupled climate models and show that the global mean equilibrium warming is higher than those obtained using extrapolation methods from shorter simulations. Specifically, 27 simulations with 15 climate models forced with a range of CO2 concentrations show a median 17% larger equilibrium warming than estimated from the first 150 years of the simulations. The spatial patterns of radiative feedbacks change continuously, in most regions reducing their tendency to stabilizing the climate. In the equatorial Pacific, however, feedbacks become more stabilizing with time. The global feedback evolution is initially dominated by the tropics, with eventual substantial contributions from the mid‐latitudes. Time‐dependent feedbacks underscore the need of a measure of climate sensitivity that accounts for the degree of equilibration, so that models, observations, and paleo proxies can be adequately compared and aggregated to estimate future warming.

We present a model intercomparison project, LongRunMIP, the first collection of millennial-length (1,000+ years) simulations of complex coupled climate models with a representation of ocean, atmosphere, sea ice, and land surface, and their interactions. Standard model simulations are generally only a few hundred years long. However, modeling the long-term equilibration in response to radiative forcing perturbation is important for understanding many climate phenomena, such as the evolution of ocean circulation, time- and temperature-dependent feedbacks, and the differentiation of forced signal and internal variability. The aim of LongRunMIP is to facilitate research into these questions by serving as an archive for simulations that capture as much of this equilibration as possible. The only requirement to participate in LongRunMIP is to contribute a simulation with elevated, constant CO2 forcing that lasts at least 1,000 years. LongRunMIP is an MIP of opportunity in that the simulations were mostly performed prior to the conception of the archive without an agreed-upon set of experiments. For most models, the archive contains a preindustrial control simulation and simulations with an idealized (typically abrupt) CO2 forcing. We collect 2D surface and top-of-atmosphere fields and 3D ocean temperature and salinity fields. Here, we document the collection of simulations and discuss initial results, including the evolution of surface and deep ocean temperature and cloud radiative effects. As of October 2019, the collection includes 50 simulations of 15 models by 10 modeling centers. The data of LongRunMIP are publicly available. We encourage submissions of more simulations in the future.

Equilibrium climate sensitivity‐the equilibrium warming per CO2 doubling‐increases with CO2 concentration for 13 of 14 coupled general circulation models for 0.5–8 times the preindustrial concentration. In particular, the abrupt 4 × CO2 equilibrium warming is more than twice the 2 × CO2 warming. We identify three potential causes: nonlogarithmic forcing, feedback CO2 dependence, and feedback temperature dependence. Feedback temperature dependence explains at least half of the sensitivity increase, while feedback CO2 dependence explains a smaller share, and nonlogarithmic forcing decreases sensitivity in as many models as it increases it. Feedback temperature dependence is positive for 10 out of 14 models, primarily due to the longwave clear‐sky feedback, while cloud feedbacks drive particularly large sensitivity increases. Feedback temperature dependence increases the risk of extreme or runaway warming, and is estimated to cause six models to warm at least an additional 3K under 8 × CO2.

The long‐term warming from an anthropogenic increase in atmospheric CO2 is often assumed to be proportional to the forcing associated with that increase. This paper examines this linear approximation using a zero‐dimensional energy balance model with a temperature‐dependent feedback, with parameter values drawn from physical arguments and general circulation models. For a positive feedback temperature dependence, warming increases Earth's sensitivity, while greater sensitivity makes Earth warm more. These effects can feed on each other, greatly amplifying warming. As a result, for reasonable values of feedback temperature dependence and preindustrial feedback, Earth can jump to a warmer state under only one or two CO2 doublings. The linear approximation breaks down in the long tail of high climate sensitivity commonly seen in observational studies. Understanding feedback temperature dependence is therefore essential for inferring the risk of high warming from modern observations. Studies that assume linearity likely underestimate the risk of high warming.