Atmosphere Ocean Science Friday Seminar

---

Quantum Mechanical Data Assimilation

Speaker: David Freeman, CAOS

Location: Online

Date: Friday, April 9, 2021, 4 p.m.

Notes:

When modelling chaotic dynamical systems, notably in climate prediction models, the use of deterministic local parametrization can cause a loss in the model’s ability to capture important dynamics of the original system. To circumvent this issue, stochastic non-local parametrization schemes can be implemented. We propose such a parametrization scheme incorporating a data assimilation method based on quantum mechanics and employing use of the Koopman operator. Given a system in which some component of the state is unknown, this method involves defining the system as being in a time-dependent “quantum-state” which influences the tendency of a random draw of the unknown component of the classical state. We analyze the results of this methodology applied to the Lorenz 63’ system and show that this method can preserve large-scale chaotic dynamics.