Data-driven model reduction for slowing climate change

Abstract

Climate and weather models anticipate future climate and weather based on numer ical solutions of a set of time-dependent partial differential equations given beginning circumstances that describe meteorological field values of the atmosphere at a certain point in time. Even the most powerful computers can’t keep up with the computing needs of today’s models. Models are requiring more calculations as they grow to rep resent a more thorough knowledge of atmospheric dynamics. Concerns regarding the viability of a desired solution for a given model on a certain hardware platform stem from the high expenses of both model creation and computational resources. The goal of this project is to reduce these vast, complex models to smaller, simpler mod els that can properly describe the behaviour of the original process under a range of operating situations. Model Order Reduction (MOR) is a time-saving, systematic tech nique for creating a dynamical system that evolves in a much smaller area while main taining similar response characteristics to the original system. Reduced order models might be utilised as efficient surrogates for the original model in bigger simulations, substituting it as a component. The Loewner framework for model reduction of linear systems will be used in this regard.