You are cordially invited to attend virtually the defence of Bart de Leeuw on his thesis:
On shadowing methods for data assimilation
Promotoren: prof. dr. Jason Frank (UU)
Co-promotor: dr. Svetlana Dubinkina (VU)
Follow live video stream on: https://live.starleaf.com/ODQxMDg6MjYyMjEx
A well-known butterfly effect of chaotic dynamical systems - when an infinitesimal error in initial conditions leads to exponential growth over time - influences the accuracy of wether forecasting. Over decades researches in the atmospheric science community have been developing numerical algorithms that decrease the uncertainty in the initial conditions by means of data assimilation - when data from measurement devices is combined with a mathematical model of atmospheric flow.
In his PhD thesis, Bart de Leeuw proposes a novel data-assimilation algorithm that is based on a shadowing property of chaotic dynamical systems - when an exact trajectory of a uniform hyperbolic system can be shadowed over an infinite period of time by a very accurate approximation. By means of numerical experiments with toy atmospheric models de Leeuw shows that his method outperforms the state-of-the-art variational data-assimilation method. Furthermore, de Leeuw extends his method to a problem of trajectory estimation of coupled atmospheric-ocean models - a challenging problem for data assimilation due to different time scales of atmospheric and ocean variables.
