Skip to Main content Skip to Navigation

Déformation des cycles saisonniers de variables climatiques

Abstract : Climate normals are usually calculated as averages over a 30-year observational period. In the context of a changing climate, these normals, even re-evaluated frequently, are "lagging behind" the current climate. The primary objective of this thesis is the estimation of non- stationary normals, in order to acquire an unbiased reference of the present climate. A good property for such normals, which are considered at the daily timescale, is to possess a certain regularity, both on the seasonal component and the secular change. For this reason, normal estimation shall be based on smoothing spline techniques such as proposed by Azaïs et Ribes (2016). The proposed modeling, inspired by a "pattern scaling" assumption, enables the study of the seasonal drift due to climate change. Beyond the mean value (normals), climate considered at a given date and location can be characterized by its distribution of possible values. A natural extension of normal estimation, would be to estimate the entire distribution with regularity constraints on its shape. We thus propose to address this question in a regularized quantile regression framework. In that respect and for a given climate variable, we obtain a detailed description of climate, at a given localization, and of its annual cycle. Both of these aspects lead to a re-examination of the RKHS (Reproducing Kernel Hilbert Space) and quantile regression theories. The complexity of the considered models is equally considered for normals and distribution and is meticulously examined. Lastly, we propose to exploit the preceding results to revisit and improve the description of future and past changes in climate, for example via the use of climate analogues.
Document type :
Complete list of metadatas
Contributor : Abes Star :  Contact
Submitted on : Friday, January 22, 2021 - 12:19:26 PM
Last modification on : Saturday, January 23, 2021 - 3:30:04 AM


Version validated by the jury (STAR)


  • HAL Id : tel-03118539, version 1


Alix Rigal. Déformation des cycles saisonniers de variables climatiques. Climatologie. Université Paul Sabatier - Toulouse III, 2020. Français. ⟨NNT : 2020TOU30112⟩. ⟨tel-03118539⟩



Record views


Files downloads