Aerial view of part of the Greenland Ice Sheet
Part of the sprawling Greenland Ice Sheet is seen in this aerial view. Credit: Tim Bocek, CC BY-NC-SA 2.0

Earth’s climate is the result of a complex network of interacting systems: air currents, ocean biogeochemistry, mountain ranges, and ice sheets, to name a few. Understanding how our climate evolves and predicting what it will look like under various scenarios require comparing and combining multiple models and simulations, each with its own strengths and focus areas. Earth science researchers are currently putting the most recent release of one such modeling system through its paces.

The second version of the Community Earth System Model (CESM) offers researchers more realistic representations of changes in Greenland’s ice sheet and interactions of agricultural crops with the Earth system, as well as detailed models of clouds and wind-driven ocean waves.

The open-source Community Earth System Model (CESM) modeling framework is used for many purposes, including investigations of past and current climate, projections of future climate change, and subseasonal-to-decadal Earth system predictions. Its latest version, CESM2, was released in June 2018, followed by several incremental releases that included additional, readily available model configurations. Compared with its predecessor, CESM2 offers researchers new capabilities, including more realistic representations of changes in Greenland’s ice sheet and interactions of agricultural crops with the Earth system, as well as detailed models of clouds and wind-driven ocean waves.

A Community Effort

CESM is a collaborative modeling effort involving researchers at the National Center for Atmospheric Research (NCAR), various universities, and other national and international research institutions. CESM2 component models include the atmosphere, land, ocean, sea ice, land ice, rivers, and ocean waves, all of which use a coupler to exchange information about fluxes (e.g., latent and sensible heat, shortwave radiation, precipitation) and states (e.g., surface temperatures) with each other.

The atmospheric component model has both a low-top and a high-top version. The low-top version accounts only for effects below the stratopause (the boundary between the stratosphere and the mesosphere) and has limited chemistry capabilities, including a relatively coarse stratospheric representation and no prognostic chemistry for ozone and other stratospheric constituents. The high-top version includes the stratosphere and extends above the stratopause. This version’s comprehensive chemistry capabilities include a better stratospheric representation, with chemical mechanisms for more than 200 species represented. CESM participants performed numerous simulations with both high- and low-top versions to support CESM’s contributions to the Coupled Model Intercomparison Project Phase 6 (CMIP6) [Eyring et al., 2016].

CESM simulations are widely used in many studies, including in national and international climate assessments.

In addition to the Diagnostic, Evaluation and Characterization of Klima (DECK) experiments (four standard simulations required for participation in CMIP), CESM2 participated in about 20 CMIP6-endorsed Model Intercomparison Project (MIP) efforts. Most of these simulations were performed using 1° horizontal spatial resolution in all component models. To provide a computationally more economical model for long timescales (e.g., for paleoclimate applications), we also conducted several DECK and MIP simulations with a version that uses a coarser 2° horizontal resolution in its atmospheric component.

Because CESM simulations are widely used in many studies, including in national and international climate assessments, it is important that the model’s main characteristics are thoroughly analyzed and documented. Thus, articles describing and analyzing these CESM2 CMIP6 experiments in detail are collected in the AGU CESM2 virtual special issue, which spans several AGU journals.

Overcoming Obstacles

We encountered two major challenges during the development of CESM2. First, our control simulations of preindustrial conditions showed the formation of unrealistically extensive sea ice cover in the Labrador Sea region. Second, global mean surface temperature time series in historical simulations displayed unrealistic cooling during the second half of the 20th century [Danabasoglu et al., 2020].

Considerable analysis of model simulations did not identify a particular culprit for the emergence of extensive sea ice cover. However, we found that extremely small perturbations in model fields could trigger this behavior in some ensemble simulations that were otherwise identical to simulations that did not show the same sea ice development. Thus, a practical solution was simply to use initial conditions for the ocean and sea ice from an ensemble member that did not produce such an extensive sea ice cover in the Labrador Sea.

Because aerosol particles can affect clouds and their radiative properties, we addressed the problem of unrealistic cooling in the simulations by introducing minor modifications to the representation of aerosol–cloud interaction processes in the CESM2 atmospheric component. These modifications were aimed at reducing the excessive influence of aerosols on the liquid water path (the total amount of liquid water between two points in the atmosphere) in comparison with observations. These changes produced an acceptable simulation of the surface temperature time series during the 20th century.

The data sets from CESM2 CMIP6 simulations are available from the Earth System Grid Federation (ESGF). To date, CESM participants have run about 1,000 CMIP6 experiments—including some tier 2 simulations for several MIPs (tier 1 represents the highest-priority experiments)—and about 600 terabytes of data have been published on the ESGF. This volume of data is roughly 7 times larger than CESM1 contributions to CMIP Phase 5.

New Features and Findings

An introduction to CESM2 [Danabasoglu et al., 2020] summarizes many new scientific and technical advances in CESM2 compared with its predecessor, CESM1. Among many others, these advances include improved representations of clouds [Golaz et al., 2002], crops [Lawrence et al., 2019], and Greenland’s evolving ice sheet [Lipscomb et al., 2019].

Illustration of winds over the Greenland Ice Sheet taken from a simulation run using the atmospheric component of CESM2
This snapshot of winds over the Greenland Ice Sheet was taken from a simulation run using a configuration of the atmospheric component of the second version of the Community Earth System Model with 1/8° spatial resolution. Wind streamlines at the lowest model level are colored according to wind speed (warmer colors indicate faster winds). Katabatic winds, in which gravity carries high-density air masses downward, can be seen accelerating down the eastern slopes of the ice sheet. This image was inspired by a visualization of winds over Antarctica by the Polar Meteorology Group at the Byrd Polar and Climate Research Center. Credit: Matt Rehme and Adam Herrington, National Center for Atmospheric Research

Compared both with available observations and with CESM1, CESM2 historical simulations show reduced biases for precipitation and shortwave cloud forcing. CESM2 simulates many aspects of the El Niño–Southern Oscillation (ENSO) well, including its dominant, multiyear timescale and associated teleconnections that affect, for example, precipitation and surface pressure and temperatures [Capotondi et al., 2020]. The same is true for the Madden-Julian Oscillation [Danabasoglu et al., 2020]. In addition, CESM2 significantly improves representations of storm tracks, Northern Hemisphere stationary waves (planetary-scale variations in atmospheric circulation), and winter blocking (obstructions of weather systems that can produce extreme weather events at lower latitudes) [Simpson et al., 2020]. Despite these improvements, CESM2 still has shortcomings. These include local precipitation biases, larger-than-observed ENSO amplitudes, thin Arctic sea ice in some simulations, and some other persistent biases such as an incorrect path of the North Atlantic Current.

The equilibrium climate sensitivity (ECS) and the transient climate response (TCR) are two properties that emerge from the coupled simulations. ECS represents the equilibrium change in global mean surface temperature after a doubling of atmospheric carbon dioxide (CO2), and TCR represents the change in global mean surface temperature around the time of CO2 doubling when CO2 increases by 1% per year.

CESM2 ECS values of 5.1°C–5.3°C are considerably higher than those produced by its previous versions, which had ECS values of about 4.0°C. Models that generate higher ECS values predict more climate warming for a given amount of atmospheric CO2 than models that generate lower ECS values. The increased ECS in CESM2 is largely due to a combination of relatively small changes in cloud microphysics and boundary layer parameters, resulting in changes in clouds and their feedbacks, particularly over the Southern Ocean but also over the tropical oceans [Bacmeister et al., 2020; Zelinka et al., 2020; Gettelman et al., 2019]. In contrast, TCR values in CESM2 remain at about 1.9°C–2.0°C, similar to those in previous versions. The consistency in this globally integrated value, however, masks significant regional differences in warming magnitude and patterns in CESM2 with respect to CESM1 [Bacmeister et al., 2020].

Interestingly, CESM2 does not appear to be alone in exhibiting an increased ECS. Indeed, one third of the CMIP6 generation models (13 out of 39) studied by Meehl et al. [2020] have ECS values higher than 4.5°C, with 6 of the models showing even higher ECS values of more than 5°C. (The Intergovernmental Panel on Climate Change has previously cited a “likely” ECS range of 1.5°C–4.5°C.) These researchers also report a range of 1.8°C–5.6°C for ECS in these new generation models. Both this range and the high ECS values in CMIP6 are significantly greater than those seen in previous generation models. Meehl et al. [2020] identified cloud feedbacks and cloud-aerosol interactions as the likely contributors to increased ECS in the new generation of models—as was the case for CESM2—with the details of sensitivities possibly differing among models.

Anticipating Advances to Come

Although CESM2 has been used primarily for CMIP6-related simulations and applications so far, its use will continue to expand for many years to come. One such effort is the new CESM2 Large Ensemble, a suite of simulations of the 1850–2100 time period, each starting with slightly different initial states. This effort is being performed in collaboration with the Institute for Basic Science Center for Climate Physics in Busan, South Korea. This 100-member model ensemble is expected to be available to the community in summer 2021. Additional ensemble members that isolate the effects of forcings will also be available. Another upcoming effort is a new set of CESM2 subseasonal-to-decadal prediction simulations.

We anticipate that some of the biases and shortcomings in CESM2 will be addressed as we move toward our next-generation model, CESM3, an effort that has already started.

The AGU CESM2 virtual special issue articles highlight many improvements in model solutions and advances in our scientific understanding with CESM2. Some early analyses also suggest that CESM2 simulations rank among the best coupled models in the CMIP6 archive in comparison with observation-based metrics (e.g., metrics associated with energy and water cycles and atmospheric dynamics) [Fasullo, 2020].

As noted above, CESM2 still suffers from some biases and shortcomings. We anticipate that some of these biases will be addressed as we move toward our next-generation model, CESM3, an effort that has already started. We plan to incorporate many advances in this next iteration in close collaboration with the Earth system modeling community. Such advances will include a new ocean model component and higher atmospheric vertical resolution available with a new dynamical core.

The virtual special issue is spread across several AGU journals, including Journal of Advances in Modeling Earth Systems, Global Biogeochemical Cycles, Journal of Geophysical Research: Atmospheres, Journal of Geophysical Research: Earth Surface, Journal of Geophysical Research: Oceans, and Geophysical Research Letters. At the time of this publication, members of the Earth system modeling community have contributed about 45 published or submitted manuscripts. And we look forward to more contributions to come, which will continue advancing this modeling effort as well as our understanding of our planet’s past, present, and potential future climate.


The CESM project is supported primarily by the National Science Foundation (NSF). This material is based upon work supported by NCAR, which is a major facility sponsored by NSF under cooperative agreement 1852977. Computing and data storage resources, including the Cheyenne supercomputer (doi:10.5065/D6RX99HX), were provided by the Computational and Information Systems Laboratory at NCAR. We thank all the scientists, software engineers, and administrators who contributed to the development of CESM2.


Bacmeister, J. T., et al. (2020), CO2 increase experiments using the CESM: Relationship to climate sensitivity and comparison of CESM1 to CESM2, J. Adv. Model. Earth Syst., 12, e2020MS002120,

Capotondi, A., et al. (2020), ENSO and Pacific decadal variability in the Community Earth System Model version 2, J. Adv. Model. Earth Syst., 12, e2019MS002022,

Danabasoglu, G., et al. (2020), Community Earth System Model version 2 (CESM2), J. Adv. Model. Earth Syst., 12, e2019MS001916,

Eyring, V., et al. (2016), Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization, Geosci. Model Dev., 9, 1,937–1,958,

Fasullo, J. T. (2020), Evaluating simulated climate patterns from the CMIP archives using satellite and reanalysis datasets using the Climate Model Assessment Tool (CMATv1), Geosci. Model Dev., 13, 3,627–3,642,

Gettelman, A., et al. (2019), High climate sensitivity in the Community Earth System Model Version 2 (CESM2), Geophys. Res. Lett., 46, 8,329–8,337,

Golaz, J.‐C., V. E. Larson, and W. R. Cotton (2002), A PDF‐based model for boundary layer clouds. Part I: Method and model description, J. Atmos. Sci., 59, 3,540–3,551,<3540:APBMFB>2.0.CO;2.

Lawrence, D. M., et al. (2019), The Community Land Model Version 5: Description of new features, benchmarking, and impact of forcing uncertainty, J. Adv. Model. Earth Syst., 11, 4,245–4,287,

Lipscomb, W. H., et al. (2019), Description and evaluation of the Community Ice Sheet Model (CISM) v2.1, Geosci. Model Dev., 12, 387–424,

Meehl, G. A., et al. (2020), Context for interpreting equilibrium climate sensitivity and transient climate response from the CMIP6 Earth system models, Sci. Adv., 6, eaba1981,

Simpson, I. R., et al. (2020), An evaluation of the large-scale atmospheric circulation and its variability in the Community Earth System Model 2 (CESM2) and other CMIP models, J. Geophys. Res. Atmos., 125, e2020JD032835,

Zelinka, M. D., et al. (2020), Causes of higher climate sensitivity in CMIP6 models, Geophys. Res. Lett., 47, e2019GL085782,

Author Information

Gokhan Danabasoglu ( and Jean-François Lamarque, Climate and Global Dynamics Laboratory, National Center for Atmospheric Research, Boulder, Colo.


Danabasoglu, G.,Lamarque, J.-F. (2021), Building a better model to view Earth’s interacting processes, Eos, 102, Published on 15 March 2021.

Text © 2021. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.