Accurate representation of clouds and their atmospheric moistening and heating is a major challenge in present climate prediction models. It is this challenge that contributes to the wide spread in climate prediction.
However, exact predictions of global warming in response to growing concentrations of greenhouse gas are important for policy-makers (for example, the Paris climate agreement).
Researchers, headed by Pierre Gentine, associate professor of earth and environmental engineering at Columbia Engineering, reported their findings in a paper that was recently published online in Geophysical Research Letters (May 23). They showed that machine-learning methods can be applied to address this problem and to accurately represent clouds in coarse resolution (~100 km) climate models. This would possibly reduce the range of prediction.
This could be a real game-changer for climate prediction. We have large uncertainties in our prediction of the response of the Earth's climate to rising greenhouse gas concentrations. The primary reason is the representation of clouds and how they respond to a change in those gases. Our study shows that machine-learning techniques help us better represent clouds and thus better predict global and regional climate's response to rising greenhouse gas concentrations.
Pierre Gentine, the paper’s lead author and a member of the Earth Institute and the Data Science Institute
As a proof of concept, the researchers utilized an idealized setup (a planet with continents, or an aquaplanet) for their new approach to convective parameterization based on machine learning. Then, by training a deep neural network, they subsequently learned from a simulation that clearly represents the clouds. The team named this machine-learning representation of clouds as Cloud Brain (CBRAIN), which can accurately predict most of the cloud moistening, heating, and radiative features that are important to climate simulation.
Gentine notes, "Our approach may open up a new possibility for a future of model representation in climate models, which are data driven and are built 'top-down,' that is, by learning the salient features of the processes we are trying to represent."
The researchers also observed that the CBRAIN technique could also enhance the estimates of future temperature, since global temperature sensitivity to CO2 is strongly associated with the representation of clouds. The team used fully coupled climate models to test this technique and demonstrated extremely promising results. This shows that greenhouse gas response may be predicted using this machine learning technique.