Instabilities in Neural Autoregressive Models of Geophysical Flows

Explore the challenges and instabilities in neural autoregressive models applied to geophysical flows in this 27-minute conference talk. Delve into the complexities of modeling Earth's climate dynamics, focusing on atmospheric and oceanic processes that span multiple temporal and spatial scales. Examine how energy transfers across these scales impact our understanding of the energy cycle and the difficulties in quantifying these transfers. Learn about the limitations of current numerical models in resolving turbulent mixing and instabilities, and the subsequent reliance on parameterizations. Gain insights into the open question of energy-consistent closure in ocean and climate models, and understand the need for combining analytical theories, numerical models, and observational measurements to advance our comprehension of geophysical fluid dynamics.

(In)stabilites in Neural Autoregressive Models of Geophysical Flows by Ashesh Chattopadhyay