Transformative research is needed to realize qualitative improvements in the modeling of climate. The investigators are carrying out an interdisciplinary program of direct statistical approaches to climate modeling that has the potential to dramatically improve our understanding of how different climate processes interact over short and long time scales and small and large spatial scales. The program lays the foundations for a new class of climate models that may complement or even eventually replace existing climate models. It has two parallel tracks. First, the investigators are developing methods for the direct computation of climate statistics that may eventually replace conventional numerical simulations, leading to computationally more efficient climate models. Second, they are developing improved representations of low clouds and an improved understanding of cloud interactions with large-scale circulations, leading to more accurate climate models and reducing the currently large uncertainties in the representation of clouds in them.
The methods for the direct computation of climate statistics are based upon cumulant expansions in which fast modes are integrated out. This refocuses attention on the evaluation and understanding of low-frequency modes, which dominate climate variability. Improved representations of low clouds are developed based on probabilistic closures of cloud and subgrid-scale dynamics, which can be systematically improved in accuracy. These are integrated into a general circulation model (GCM) that couples the clouds to the large-scale flows, allowing a systematic study of their interactions. The work extends past work on the direct computation of statistics in relatively simple models of geophysical flows to GCMs of increasing complexity and realism, while improving the accuracy of the statistical methods and of the GCMs themselves.
