Atmosphere Ocean Science Colloquium

A Statistical Emulator for Geophysical Flows

Speaker: Andre Souza, MIT

Location: Warren Weaver Hall 1302

Date: Wednesday, October 26, 2022, 3:30 p.m.

Synopsis:

In this talk, we discuss a method for constructing emulators of turbulent flow statistics. By representing the flow as a continuous time Markov process, we capture the transition probabilities to different regions of phase space associated with the turbulent dynamics. The resulting transition rate matrix is a finite volume discretization in function space and approximates the Perron-Frobenius operator of the underlying dynamical system. We illustrate the method on an oceanic simulation of topographic meanders in a zonal current. Furthermore, we show that the approach generalizes to other statistically stationary systems, such as an idealized atmospheric simulation and canonical examples of chaos. ‚Äč