NP Nonlinear Processes in Geosciences
The 2019 Outstanding Student Poster and PICO (OSPP) Awards is awarded to Rem-Sophia Mouradi for the poster/PICO entitled:
A combined orthogonal decomposition and polynomial chaos methodology for data-based analysis and prediction of coastal dynamics (Mouradi, R.-S.; Goeury, C.; Thual, O.; Zaoui, F.; Tassi, P.)
Click here to download the poster/PICO file.
Rem-Sophia Mouradi is a PhD researcher in the hydrodynamics department of EDF R&D (Electricité De France, Chatou, France) and CERFACS (Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique, Toulouse, France). In the application context of sedimentation in nearshore channels, her research focuses on the use of field data to improve predictions provided by a numerical model. This involves both purely data based statistical learning, as well as data assimilation technics, to work jointly with the numerical model.
The work presented in EGU2019’s poster concerned the first approach. The parameters of the study (inputs on the first hand, output on the other), were considered random variables. The structure of the interest 2D field (or output) was simplified using Proper Orthogonal Decomposition (POD). The dependency structure between the components of POD and the inputs, which are natural forcings, was built using Polynomial Chaos Expansion (PCE). This introduces a methodology for a purely data based learning algorithm that comes with the advantage of probabilistic characterization, useful for uncertainty quantification studies.