En poursuivant votre navigation sur ce site, vous acceptez le dépôt de cookies dans votre navigateur. (En savoir plus)
Portail > Offres > Offre UMR8001-AURFIS-001 - Post-doctorat (projet IMPT) : méthodes statistiques pour améliorer les modèles de climat (H/F)

Postdoctoral position (IMPT project) : statistical methods to improve climate models (M/F)

This offer is available in the following languages:
Français - Anglais

Date Limite Candidature : mercredi 2 février 2022

Assurez-vous que votre profil candidat soit correctement renseigné avant de postuler. Les informations de votre profil complètent celles associées à chaque candidature. Afin d’augmenter votre visibilité sur notre Portail Emploi et ainsi permettre aux recruteurs de consulter votre profil candidat, vous avez la possibilité de déposer votre CV dans notre CVThèque en un clic !

Faites connaître cette offre !

General information

Reference : UMR8001-AURFIS-001
Workplace : PARIS 13
Date of publication : Wednesday, December 22, 2021
Type of Contract : FTC Scientist
Contract Period : 18 months
Expected date of employment : 1 April 2022
Proportion of work : Full time
Remuneration : Between 2743 and 3897 euros gross per month depending on experience
Desired level of education : PhD
Experience required : Indifferent


An important part of the uncertainty in climate models comes from fine-scale processes, which are not directly resolved. Thus, they are represented by parameterizations, whose role is to estimate the impact of "subgrid" processes on the large-scale flow. These parametrizations, as they have been developed so far, mix physics and pragmatism, and remain unconstrained, making only very limited use of observations.

This research project will use statistical learning methods to capture the relationship between the large-scale flow and internal gravity waves, which are a key subgrid process for stratospheric circulation. This will take full advantage of data from recent stratospheric balloon campaigns coordinated by the Laboratoire de Météorologie Dynamique (LMD), and guide the development of gravity wave parameterizations in climate models.

The main tasks will be to implement learning methods to estimate key quantities of gravity waves, as observed by the balloons, from information from meteorological models describing the large-scale flow. The work performed will answer the following questions:
- what fraction of the gravity waves can be determined, what fraction should be considered as "stochastic" ?
- which elements of the large-scale flow provide the most information ?
- what are the most efficient methods in this problem ?


We have unique sets of observations providing precise quantification of gravity waves (momentum fluxes, frequencies, phase velocities), from long duration stratospheric balloon campaigns. These observations will provide the target variables. A description of the large-scale flow, as it could be simulated by a climate model, will be obtained by the European Center for Medium-Range Weather Forecasts (ECMWF). The main activities will be:
- Extract from this "low resolution" description of the flow explanatory variables co-located with the balloon observations. This preliminary work is already underway.
- Apply statistical learning methods to estimate the target variable from the explanatory variables. Parametric and non-parametric regression methods will be used, the former being attractive for their interpretability, the latter for their performance. For each method, a careful calibration will be needed for optimal performance.
- Explore the sensitivity of the results to the amount of information provided (resolution, number of variables), to the target quantity (momentum flux, wave characteristics), and to the methods used (linear or polynomial regressions, k-nearest neighbor methods, kernel estimators, regression trees and associated ensemble methods, support vector machines...).
- Improve the performances of the learning methods by more advanced strategies, such as aggregation procedures, or a preliminary clustering step on the inputs.


This offer is for a young PhD (M/F) in the field of applied mathematics and statistical learning methods. Knowledge of atmospheric and climate sciences, and/or fluid mechanics, will be an advantage, but not essential. Familiarity with numerical experiments and programming for the analysis of large datasets (e.g. in R or Python) will be appreciated. A taste for collaborative work, interactions and interdisciplinarity will be positive points.

Work Context

The context of this postdoctoral fellowship, founded by the Institut des Mathématiques pour la Planète Terre (IMPT), is an approximately five-year collaboration between Aurélie Fischer, from the Laboratoire de Probabilités, Statistique et Modélisation (LPSM, University of Paris), and Riwal Plougonven, from the Laboratoire de Météorologie Dynamique (LMD, Ecole Polytechnique). The postdoctoral fellow will work at LPSM. At the beginning of the post-doc, he or she will come a few times to LMD, in Palaiseau, to establish a minimal knowledge base on climate modeling and relevant atmospheric processes. Thereafter, regular three-way interactions via videoconference will allow for continuous monitoring and discussion of any issues that arise, including the interpretation of geophysical data. More frequent face-to-face meetings may be scheduled at LPSM in Paris to complete this arrangement. The postdoctoral fellow will be encouraged to present his or her results, publish them, and participate in meetings, colloquia or conferences that will be beneficial to his or her network and career.

Additional Information

The date of hiring is indicative.

We talk about it on Twitter!