SCRIPPS INSTITUTION OF OCEANOGRAPHY FACULTY CANDIDATE SEMINAR -
DATE: May 28th, Tuesday, 12:45 p.m.
LOCATION: Eckart 227
TITLE: Turbulence matters: Emerging opportunities for numerical climate dynamics in the era of exascale computing and deep learning.
In short -- turbulence matters. I will show pilot results from a new form of multiscale climate model that reduces our heavy reliance on assumptions of turbulent parameterization. The result is the world’s first global atmospheric model to permit explicit atmospheric boundary layer eddies. Reassuringly, this increases realism of the local dynamics supporting simulated low clouds. A first look at the emergent global climate dynamics reveals a cloud response to surface warming surprisingly similar to other climate models with parameterized physics, and a major muting of aerosol’s impact on cloud radiative forcing, suggesting interesting new buffering dynamics occur when turbulence is resolved. I will conclude with the current exciting technical outlook, focusing on the now demonstrable promise of GPU-powered exascale computers, and the recent successes of modern deep learning methods. Machine learning especially proves to be a skillful emulator of explicitly simulated moist turbulence and, importantly, an intriguing new tool for dynamical inquiry of models with realistic complexity and high degrees of freedom.