PDE and applications: Of Lorenz Butterflies and other animals: atmospheric predictability from dynamical systems and machine learning perspectives
- Date: –11:15
- Location: Ångströmlaboratoriet, Lägerhyddsvägen 1 64119
- Lecturer: Gabriele Messori
- Organiser: Matematiska institutionen
- Contact person: Kaj Nyström
Abstract: Atmospheric flows are characterized by chaotic dynamics and recurrent large-scale patterns. These two characteristics point to the existence of an atmospheric attractor, defined by Lorenz as: “the collection of all states that the system can assume or approach again and again, as opposed to those that it will ultimately avoid”. By leveraging recent developments in dynamical systems theory, we can diagnose local properties of states on the atmospheric attractor, which provide us with powerful insights into atmospheric predictability. Atmospheric predictability can also be studied by using machine learning algorithms. These are often likened to a "black box", where the user knows what goes in and what comes out but is not privy to the box's internal workings. In this sense, machine learning sits at the opposite end of the spectrum to dynamical systems theory. In my talk, I will argue that the two approaches are both valuable for the study of atmospheric predictability, and that they provide largely complementary information. I will mainly focus on practical applications to idealised and real-world atmospheric datasets, rather than on technical or theoretical considerations, but I will nonetheless attempt to introduce the basic concepts underlying the dynamical systems framework.