Clouds have a remarkable impact on Earth’s climate: not only do they reflect sunlight back to space, but they also absorb infrared radiation emitted by the Earth and thus cause a greenhouse effect. Their ability to reflect sunlight is stronger than their greenhouse effect, on average, so clouds keep Earth much cooler than it would be in their absence. As the atmosphere warms in response to increases in greenhouse gas concentrations, climate models predict changes in cloudiness that will affect their ability to cool the planet. This is an example of a climate feedback: warming causes a change in clouds that can amplify or dampen the original temperature change. Most current climate models predict a positive cloud feedback, meaning that clouds change so as to reinforce the effect of greenhouse gas increase. Despite their general agreement that cloud feedback is positive, models disagree on how strong the feedback is. This is because the processes behind cloud formation are extremely challenging to represent in climate models. Clouds are therefore the dominant source of model-to-model disagreement in climate sensitivity, the global-mean temperature increase due to a doubling of CO2 concentration.
In an advanced review recently published in WIREs Climate Change, Ceppi et al. review the representation of cloud feedback in current global climate models, discuss the driving mechanisms, and evaluate the causes of model uncertainty. Despite model-to-model differences in the exact details, climate models qualitatively agree on three important components of the global cloud response: (1) high clouds will rise, strengthening their greenhouse effect (a positive feedback); (2) tropical low clouds will decrease in coverage, reflecting less sunlight to space (a positive feedback); and (3) high-latitude low clouds will become thicker, reflecting more sunlight to space (a negative feedback). While these three main responses occur in all models, and are supported by theory and observations, and hence can be considered robust, their strengths remain highly uncertain. In particular, decreases in tropical low cloud coverage vary substantially across models, driving the inter-model spread in cloud feedback.
Most of the model uncertainty in cloud feedback is associated with simplified representations of small-scale processes such as radiation, convection, turbulence, and cloud microphysics (the processes involved in cloud droplet and ice crystal formation, growth, and precipitation). These processes are represented with varying degrees of realism, complexity, and accuracy in current climate models, and large errors exist in all models. For example, how much high-latitude clouds will thicken depends on how models represent the poorly-understood processes responsible for the freezing of droplets into ice crystals, and the removal of cloud water by precipitation. Ceppi et al. conclude that two important recent advances can be exploited to improve our model-based estimates of cloud feedback. First, increases in computing power are making high-resolution, process-resolving model experiments increasingly accessible, offering novel physical understanding that will guide future climate model development. Second, observations of cloud and radiation at an unprecedented level of spatial coverage and reliability are allowing us to better understand the relationships between clouds and their environment, providing observational constraints on model estimates of cloud feedback and climate sensitivity.
Contributed by Dr. Paulo Ceppi