Competition funded PhD project about Machine Learning methods for non-stationary systems and applications to climate simulations
There is an opportunity to work on a Phd project at LSCE (Laboratoire des Sciences de l'Environnement et du Climat) in the ESTIMR (Extreme Events : Statistics, Impacts & Regionalization) team provided that the project receives fundings by the international PhD program NUMERICS. Please check the eligibility requirements before applying.
The goal of the Phd project is to develop Echo State Networks (ESN) learning algorithms methods for nonstationary systems, moving towards climate-change oriented applications. The scientific objective is to simulate realistic features of atmospheric variability at daily to seasonal timescales, as a response to slow external forcings. During the thesis, the PhD candidate will incorporate instantaneous information on the changing forcings as covariate of the ESN model and evaluate the capability of ESN in learning non stationary dynamics. Tests will be performed on both Lorenz (1963, 1996) equations, and on climate variables issued from model simulation data (e.g. CMIP and CORDEX) and reanalyses (e.g. 20CR and ERA20C).
For a detailed description of the research project: