Sites Inria

Version française



Using statistics for the good of the environment

In order to understand properly the consequences of global warming for the environment, relevant models, and methods of analysing those models have to be developed. Clémentine Prieur, professor at the University of Grenoble and a member of the AIRSEA research team, talks about managing the uncertainty inherent in the models. She has just been awarded the Blaise Pascal prize from the Académie des Sciences for her work.

Why is uncertainty such an important issue, particularly in the study of models of the ocean or atmosphere?

Clémentine Prieur : When we study those environments, we use partial differential equations. In these equations, some of the parameters are little known, such as the initial conditions or the topography of the ocean floor (bathymetry). Where this is the case we try to reduce the uncertainty of the most influential parameters, or to propagate it to the evaluation of quantities of interest, linked for example to energy. The models we study, discretized in time and space, are generally very large. As a result the computing times can quickly become vast, hence the interest in developing smaller models. These smaller models must reach the following compromise: they must be faithful to the original model but they must also keep the evaluation cost low.
This is important because when we provide results to the scientific community, we must always specify the error bounds.

What are currently the challenges for climate simulation?

Clémentine Prieur : There are a number of challenges. Size, as I've just mentioned, is one.
The adaptation of methodologies and algorithms. On another level, with most climate models it's difficult to reproduce the internal variability of the climate, and extreme events. In West Africa that variability has translated as, for example, a dramatically long period of drought lasting nearly 30 years. It seems important to improve the feedback between models and observations in order to improve forecasts and the evaluation of the measures of risk associated with extreme events.
The work involves a number of different disciplines, including programming, statistics (my speciality) and also hydrology and glaciology.

What have your latest studies been looking at?

Clémentine Prieur : I've recently been working on a study concerning water level stabilisation in a canal. Fluid dynamics in canals are described by the Saint-Venant equations. Levels are controlled at the ends, by opening gates upstream and downstream. The aim of our work is to find out which parameters (friction, opening of gates, etc.) have the greatest effect on water level.
The Blaise Pascal prize from the Académie des Sciences was awarded for my work on quantifying uncertainty for models using partial differential equations.

What does the prize mean to you?

Clémentine Prieur : Well, it was totally unexpected! It was awarded to me for the work I've done at the forefront of a number of scientific disciplines. I'm a fervent believer in the need to build bridges between these different disciplines, which are not used to talking one other!

Brief biography

  • 1998 : Clémentine Prieur enters ENS Cachan in Rennes.
  • 1999 - 2001 : thesis on the study of dependence for stochastic processes, at the University of Cergy-Pontoise
  • 2002 - 2008 : Senior Lecturer at INSA, Toulouse
  • 2008 : joins the MOISE team, now AIRSEA, at Inria Grenoble - Rhône-Alpes, as Professor
  • 2015 : awarded the Blaise Pascal prize by the Académie des Sciences

Keywords: Blaise Pascal prize - Academy of Sciences - Environment - Uncertainty - Clémentine Prieur - Modelling - Statistics - Applied mathematics