This talk complements my poster at the EnKF Workshop 2023 of the same name. In brief, we have developed an efficient map adaptation strategy which automatically identifies parsimonious triangular maps for ensemble transport filtering. This adaptation strategy can also detect conditional independence properties, which permits an efficient form of localization. Here, I explain the idea behind this method, and discuss preliminary results for nonlinear ensemble transport filtering in the chaotic Lorenz-96 system.
This is a recording of the presentation I gave at SIAM Annual 2022, as part of the session MS65 Tutorials for Students: Accessible Introductions to Active Research Areas. I provided the theoretical introduction, and Matthew Parno guided students through a hands-on exercise using the newly released MPart toolbox afterwards. The notebooks for the exercises can be accessed here. This talk discusses some of the applications of triangular transport methods, and covers most of the basic concepts, from the properties of the triangular structure, across different optimization strategies, to methods which allow us to overcome the curse of dimensionality.
In this talk - a shortened version of my presentation at SIAM UQ - I present our research on nonlinear ensemble transport smoothing at the International Symposium for Data Assimiliation 2022 in Fort Collins, Colorado, USA. Since this was one of the first conferences completely in-person again, no talks were recorded at the conference venue. This is a re-recording of the presentation.
In this talk, I advocate for the use of transport maps for ensemble smoothing, providing a brief introduction to transport methods, then demonstrating their application for sequential smoothing. Due to technical difficulties on-site, the live talk was unfortunately not recorded. This is a re-recording of my presentation after the conference.
This hybrid poster presents intermediate progress of my work on nonlinear smoothing with transport methods. We discuss two different smoothing strategies (joint-analysis and backward smoothers), often applied as Kalman-type algorithms. We then introduce transport methods as a pathway for nonlinear generalization of these smoothers. We demonstrate the efficacy of the resulting algorithm with the Lorenz-63 test case.
My dissertation has been nominated by my thesis committee and ultimately won the Prix Léon Du Pasquier et Louis Perrier. This is an award bestowed anually to the best dissertation within the faculty of science at the University of Neuchâtel, Switzerland.
The award is endowed with a prize money of 2000 CHF. In the corresponding interview, I was asked to answer the questions of what motivated me to pursue a PhD, how I would explain my research to laymen, and what I am currently doing during my postdoc at MIT.
My open doctoral defense at the University of Neuchâtel, recorded during the COVID-19 quarantine in December 2020. My attempt to explain non-Gaussian parameter inference and data assimilation to a general audience through the use of space rabbits.