Model order reduction for large-scale data assimilation problems

Abstract

In addition we present projection methods for reducing the dimension of the space the state of the system resides in in weak constraint 4D-Var. We apply the control theoretic balanced truncation model reduction method, and introduce randomised projection methods, sometimes known as sketching methods to the data assimilation setting.

Numerical experiments with the linear advection-diffusion equation, the shallow water equations and the nonlinear Lorenz-95 model demonstrate the effectiveness of applying these methods when compared to solving the data assimilation problem with standard approaches.

Date of Award	19 Sept 2019
Original language	English
Awarding Institution	University of Bath
Supervisor	Melina Freitag (Supervisor) & Mark Opmeer (Supervisor)