Restricted Access Section


Commissariat à l’Energie Atomique (CEA) acting on behalf of Laboratoire des Sciences du Climat et de l’Environnement (LSCE), France

CEA-LSCE Laboratoire des Sciences du Climat et de l’Environnement

The ``laboratoire des sciences du climat et l’environnement" (LSCE) is a research center sponsored by three French national research institutes: ``Centre National de la Recherche Scientifique" (CNRS), ``Commissariat à l’Energie Atomique" (CEA) and ``Université de Versailles Saint-Quentin".

Three main research topics are covered at LSCE:

(1) mechanisms to explain the natural climate variability at different time scales and the interactions among human activities, the climate and the environment,

(2) processes linked to the carbon cycle, greenhouse gases and aerosols that interact with the climate,

(3) geochronology and its associated techniques to study recent and old climates.

The LSCE research teams provide and study a variety of observations, proxies and coupled model outputs (IPSL-model). Its staff is about 250 persons.

For the last 10 years, Philippe Naveau and Pascal Yiou have been research scientists whose research interests are focused on theoretical and applied statistics, with an emphasis on applications related to environmental and climatic data sets.

First, the successful candidate will have to work with the theoretical statisticians in order to develop the mathematical properties of our proposed statistical downscaling scheme for extremes. The “Institute of Mathematics” at EPFL is an excellent place to undertake these theoretical developments. It is important to emphasize that strong interactions with applied math researchers is needed because the distribution of extreme events in a multivariate context varies greatly from the classical Gaussian distribution. Consequently, the uncertainties computed for multivariate extremes are very different from the ones obtained for the mean and the standard deviation. This is a very challenging statistical problem and a strong collaborative work with probability experts is required.

The second part of the post-doc time will focus on testing and implementing the proposed statistical procedures to geophysical data (high precipitation, etc). To promote the dissemination of our research results, statistical software about the downscaling of extremes will be posted electronically on the Worldwide Web as a package in the freely-available R statistical program at the end of this project. The post-doc will strongly contribute to the development of these algorithms. Overall, this interdisciplinary research project will allow a young researcher to gain a real expertise in both statistics and geosciences. To advice the post-doc, the permanent members of this package will also meet on a regular basis.

Attributed tasks include:

  1. Proposing and studying novel statistical methods to better assess the uncertainties when modelling the frequency and the amplitudes of extreme events in past records. This deliverable would be based on multivariate Extreme Value Theory that would be specially tailored for the analysis of heavy rainfalls and droughts in a spatio-temporal context (conditional on meaningful weather regimes).
  2. Development of extreme-value models and estimation methods that can take into account of covariate structures. Such covariates include time, spatial information or any other quantitative measure. Such methods will be validated in three steps: theoretical study of the asymptotic properties, numerical experiments on simulated data and numerical experiments on real data.
  3. Development of new statistical downscaling schemes for heavy rainfall. This deliverable would be based on Bayesian hierarchical models and/or state-space modeling techniques in compliance with Extreme Value Theory.