Characterising last millennium climate variations using new statistical methods : Paleoclimate reconstructions and integration within a general circulation climate model

• Main Author: Michel, Simon
• Other Authors: Environnements et Paléoenvironnements OCéaniques (EPOC), Observatoire aquitain des sciences de l'univers (OASU), Université Sciences et Technologies - Bordeaux 1 (UB)-Institut national des sciences de l'Univers (INSU - CNRS)-Centre National de la Recherche Scientifique (CNRS)-Université Sciences et Technologies - Bordeaux 1 (UB)-Institut national des sciences de l'Univers (INSU - CNRS)-Centre National de la Recherche Scientifique (CNRS)-École Pratique des Hautes Études (EPHE), Université Paris sciences et lettres (PSL)-Université Paris sciences et lettres (PSL)-Centre National de la Recherche Scientifique (CNRS), Université de Bordeaux, Didier Swingedouw
• Format: Doctoral or Postdoctoral Thesis
• Language: French
• Published: HAL CCSD 2020
• Subjects: Decadal climate variability; Paloclimates; Last millennium climate; Physical oceanography; Machine learning; Data assimilation; Variabilité climatique décennale; Paléoclimats; Climat du dernier millénaire; Océanographie physique; Apprentissage statistique; Assimilation de données; [SDU.STU.CL]Sciences of the Universe [physics]/Earth Sciences/Climatology; North Atlantic
• Online Access: https://theses.hal.science/tel-04246131; https://theses.hal.science/tel-04246131/document; https://theses.hal.science/tel-04246131/file/MICHEL_SIMON_2020.pdf

Climate variability strongly impacts our lives in many ways. The observed growth in global temperatures, forced by anthropogenic greenhouse gases emissions, has experienced of period of hiatus between 1998 and 2012. Causes at the origin of such a sequence are subject to intensive scientific controversies, notably concerning the relative roles of climate variability intrinsic to the climate system and external radiative forcings. In order to address this questions, it is necessary to better understand decadal modes of variability resulting from the global scale organisation of the climate system. However, the only 150 years of instrumental observations are not enough to rigorously caractérise their nature, variability and dynamics.This thesis aims at improving our knowledge of multidecadal climate variability by producing reconstructions of climate indices and grids over the last millennium. For doing so, up-to-date statistical methods are used, namely machine learning technics, which are applied to paleoclimate data coming from natural archives (tree rings, ice cores.). The recently developed PAGES 2k database provides more than 700 of these records, and will thus be abundantly used for training the proposed statistical models.Objective mathematical metrics show that a non linear technique, namely the random forest, generally produces more robust results than the usual linear technics. We thus use the random forest method to reconstruct variations of the preferential mode of North Atlantic sea surface temperatures (SST) de l'Atlantique Nord, namely the Atlantic Multidecadal Variability (AMV), which is notably related to the Atlantic Meridional overturning circulation (AMOC). This reconstruction suggests that changes in this circulation, which occured at the end of the 12th century, has probably been the catalyser of an early onset of the little ice age, a relatively cold period of the last millennium. The strong volcanic activities from the 13th, 15th and 19th centuries have nevertheless been identified as the ...