As scientists work to determine why some of the latest climate models suggest the future could be warmer than previously thought, a new study indicates the reason is likely related to challenges simulating the formation and evolution of clouds. The new research, published in Science Advances , gives an overview of 39 updated models that are part of a major international climate endeavor, the sixth phase of the Coupled Model Intercomparison Project (CMIP6) . The models will also be analyzed for the upcoming sixth assessment report of the Intergovernmental Panel on Climate Change (IPCC).
A subset of these models has shown a higher sensitivity to carbon dioxide – that is, more warming for a given concentration of the greenhouse gas – than older models, though a few showed lower sensitivity as well. The end result is a greater range of model responses than any preceding generation of models, dating back to the early 1990s. If the models on the high end are correct and Earth is truly more sensitive to carbon dioxide than scientists had thought, the future could also be much warmer than previously projected. But it’s also possible that the updates made to the models between the last intercomparison project and this one are causing or exposing errors in their results.
In the new paper, an international author team led by Dr Gerarld Meehl from the National Center of Atmospheric Research (NCAR), Boulder, USA, with contributions from DLR-IPA scientists systematically compared the CMIP6 models with previous generations and cataloged the likely reasons for the expanded range of climate sensitivity.
Researchers have traditionally evaluated climate model sensitivity with the equilibrium climate sensitivity (ECS), a metric that has been used since the late 1970s. It measures the temperature increase after atmospheric carbon dioxide is instantaneously doubled from preindustrial levels and the model is allowed to run until it stabilizes. Through the decades, the range of ECS values has stayed remarkably consistent – somewhere around 1.5 to 4.5 °C – even as models have become significantly more complex. For example, the models included in the previous phase of CMIP last decade, known as CMIP5, had ECS values ranging from 2.1 to 4.7 °C. The CMIP6 models, however, have a range from 1.8 to 5.6 °C, widening the spread from CMIP5 on both the low and high ends.
Model developers have further analysed their models during the last year to understand why ECS has changed. For many groups, the answers appear to come down to clouds and aerosols. Cloud processes unfold on very fine scales which has made them challenging to accurately simulate in global-scale models in the past. In CMIP6, however, many modeling groups added more complex representations of these processes. The new cloud capabilities in some models have produced better simulations in some ways, showing better agreement with observations. However, clouds have a complicated relationship with climate warming – certain types of clouds in some locations reflect more sunlight, cooling the surface, while others can have the opposite effect, trapping heat. Aerosols, which can be emitted naturally from volcanoes and other sources as well as by human activity, also reflect sunlight and have a cooling effect. But they interact with clouds too, changing their formation and brightness and, therefore, their ability to heat or cool the surface.
Many modeling groups have determined that adding this new complexity into the latest version of their models is having an impact on ECS. According to Gerald Meehl, lead author of the new study from the National Center for Atmospheric Research in Boulder (USA), this isn’t surprising. “When you put more detail into the models, there are more degrees of freedom and more possible different outcomes,” he said. “Earth system models today are quite complex, with many components interacting in ways that are sometimes unanticipated. When you run these models, you’re going to get behaviors you wouldn’t see in more simplified models.”
One reason scientists continue to use ECS is because it allows them to compare current models to the earliest climate models. But researchers have come up with other metrics for looking at climate sensitivity along the way, including a model’s transient climate response (TCR). To measure that, modelers increase carbon dioxide by 1% a year, compounded, until carbon dioxide is doubled. While this measure is also idealized, it may give a more realistic view of temperature response, at least on the shorter-term horizon of the next several decades.
In the new paper, the authors also compared how TCR has changed over time since its first use in the 1990s. The CMIP5 models had a TCR range of 1.1 to 2.5 °C, while the range of the CMIP6 models only increased slightly, to 1.3 to 3.0 °C. The change in TCR range is more modest than with ECS, which could mean that the CMIP6 models may not perform that differently from CMIP5 models when simulating temperature over the next several decades.
For the calculation of the climate sensitivity metrics ECS and TCR, the authors of the study used the Earth System Model Evaluation Tool (ESMValTool), an open-source diagnostic and performance metric tool that is developed by an international consortium with more than 80 institutions under the lead of DLR-IPA. The ESMValTool allows for a comprehensive and routine evaluation of Earth system models that is urgently needed in order to facilitate the analysis of the complex models and large volume od data in the CMIP archive.
Figure 1: Climate sensitivity evaluated using the metrics ECS (equilibrium climate sensitivity) and TCR (transient climate response) over the years. Blue and red bars show the assessed range by the different Assessment Reports (ARs) of the Intergovernmental Panel on Climate Change (IPCC). For ECS, the assessed range did not decrease since 1979 (1.5 to 4.5 °C). Orange and green bars show the modeled ranges from climate models of the Coupled Model Intercomparison Project (CMIP). For CMIP5 and CMIP6, individual models are shown as black numbers. Compared to CMIP5, the model range of CMIP6 increased on the high and the low end. (Graphics produced with ESMValTool: DLR, CC-BY3.0)
 Meehl, G. A., Senior, C. A., Eyring, V., Flato, G., Lamarque, J. F., Stouffer, R. J., Taylor, K. E., and Schlund, M. (2020). Context for interpreting equilibrium climate sensitivity and transient climate response from the CMIP6 Earth system models. Science Advances, 6(26), eaba1981.
 Eyring, V., Bony, S., Meehl, G. A., Senior, C. A., Stevens, B., Stouffer, R. J., & Taylor, K. E. (2016). Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geoscientific Model Development, 9(5), 1937-1958.