L'augmentation du bilan de masse en surface de la calotte polaire Antarctique causée par celle des chutes de neige est la seule contribution négative à l'élévation du niveau de mer attendue dans le courant du 21ème siècle dans le cadre du réchauffement climatique causé par les activités humaines. La régionalisation dynamique de projections climatiques issues de modèles couplés océans-atmosphère est la méthode la plus couramment utilisée pour estimer les variations futures du climat Antarctique. Néanmoins, de nombreuses incertitudes subsistent suite à l'application de ces méthodes, en particulier en raison des biais conséquents sur les conditions océaniques de surface et sur la circulation atmosphérique aux hautes latitudes de l’Hémisphère Sud dans les modèles couplés.Dans la première partie de ce travail, différentes méthodes de corrections de biais des conditions océanique de surface ont été évaluées. Les résultats ont permis de retenir une méthode quantile-quantile pour la température de surface de l'océan et une méthode d'analogues pour la concentration en glace de mer. En raison de la forte sensibilité du climat future Antarctique aux variations de couverture de glace de mer dans l'Océan Austral, les conditions océaniques issues de deux modèles couplés, NorESM1-M et MIROC-ESM, présentant des diminutions d’étendues de glace de mer hivernales largement différentes (-14 et -45%) ont été retenues. Les conditions océaniques provenant d'un scénario RCP8.5 de ces deux modèles ont été corrigées afin de forcer le modèle atmosphérique global ARPEGE.Par la suite, ARPEGE a été utilisé dans une configuration grille-étirée, permettant d'atteindre une résolution horizontale de 40 kilomètres sur l'Antarctique. Il a été contraint aux limites par les conditions océaniques de surface observées et celles issues des simulations historiques des modèles NorESM1-M et MIROC-ESM pour la période récente (1981-2010). Pour la fin du 21ème siècle (2071-2100), les forçages océaniques originaux et corrigés issus de ces deux derniers modèles ont été utilisés. L'évaluation pour le présent a permis de mettre en évidence, la capacité du modèle ARPEGE de reproduire le climat et le bilan de masse de surface Antarctique ainsi que la persistance d'erreurs substantielles sur la circulation atmosphérique y compris dans la simulation forcée par les conditions océaniques observées. Pour le climat futur, l'utilisation des forçages océaniques MIROC-ESM corrigés a engendré des augmentations supplémentaires significatives à l'échelle continentale pour les températures hivernales et le bilan de masse annuel.Enfin, ARPEGE a été corrigé en ligne, à l'aide d'une climatologie des termes de rappel du modèle issus d'une simulation guidée par les réanalyses climatologiques. L'application de cette méthode sur la période récente a très largement amélioré la modélisation de la circulation atmosphérique et du climat de surface Antarctique. L'application pour le climat futur suggère des augmentations de températures (+0.7 à +0.9 C) et de précipitations (+6 à +9%) supplémentaires par rapport à celles issues des scénarios réalisés sans correction atmosphérique. Le forçage de modèles climatiques régionaux ou de dynamique glaciaire avec les scénarios ARPEGE corrigés est à explorer au regard des impacts potentiellement importants pour la calotte Antarctique et sa contribution à l'élévation du niveau des mers. / The increase of the Antarctic ice-sheet surface mass balance due to rise in snowfall is the only expected negative contribution to sea-level rise in the course of the 21st century within the context of global warming induced by mankind. Dynamical downscaling of climate projections provided by coupled ocean-atmosphere models is the most commonly used method to assess the future evolution of the Antarctic climate. Nevertheless, large uncertainties remain in the application of this method, particularly because of large biases in coupled models for oceanic surface conditions and atmospheric large-scale circulation at Southern Hemisphere high latitudes.In the first part of this work, different bias-correction methods for oceanic surface conditions have been evaluated. The results have allowed to select a quantile-quantile method for sea surface temperature and an analog method for sea-ice concentration. Because of the strong sensitivity of Antarctic surface climate to the variations of sea-ice extents in the Southern Ocean, oceanic surface conditions provided by two coupled models, NorESM1-M and MIROC-ESM, showing clearly different trends (respectively -14 and -45%) on winter sea-ice extent have been selected. Oceanic surface conditions of the ``business as usual" scenario (RCP8.5) coming from these two models have been corrected in order to force the global atmospheric model ARPEGE.In the following, ARPEGE has been used in a stretched-grid configuration, allowing to reach an horizontal resolution around 40 kilometers on Antarctica. For historical climate (1981-2010), the model was driven by observed oceanic surface conditions as well as by those from MIROC-ESM and NorESM1-M historical simulation. For late 21st century (2071-2100), original and bias corrected oceanic conditions from the latter two model have been used. The evaluation for present climate has evidenced excellent ARPEGE skills for surface climate and surface mass balance as well as large remaining errors on large-scale atmospheric circulation even when using observed oceanic surface conditions. For future climate, the use of bias-corrected MIROC-ESM oceanic forcings has yielded an additionally significant increase in winter temperatures and in annual surface mass balance at the continent-scale.In the end, ARPEGE has been corrected at run-time using a climatology of tendency errors coming from an ARPEGE simulation driven by climate reanalyses. The application of this method for present climate has dramatically improved the modelling of the atmospheric circulation and antarctic surface climate. The application for the future suggests significant additional warming (~ 0.7 to +0.9 C) and increase in precipitation (~ +6 to +9 %) with respect to the scenarios realized without atmospheric bias correction. Driving regional climate models or ice dynamics model with corrected ARPEGE scenarios is to explored in regards of the potentially large-impacts on the Antarctic ice-sheet and its contribution to sea-level rise.

In this thesis, we investigated modes of intraseasonal variability (ISV) observed in the Indian monsoon rainfall and how these modes modulate rainfall over India. We identified a decreasing trend in the intensity of low-frequency intraseasonal mode with increasing strength in synoptic variability over India. We also made an attempt to understand the reason for these observed trends using numerical simulations. In the first part of the thesis, satellite rainfall estimates are used to understand the spatiotem-poral structures of convection in the intraseasonal timescale and their intensity during boreal sum-mer over south Asia. Two dominant modes of variability with periodicities of 10–20-days (high-frequency) and 20–60-days (low-frequency) are found, with the latter strongly modulated by sea surface temperature. The 20–60-day mode shows northward propagation from the equatorial In-dian Ocean linked with eastward propagating modes of convective systems over the tropics. The 10–20-day mode shows a complex space-time structure with a northwestward propagating anoma-lous pattern emanating from the Indonesian coast. This pattern is found to be interacting with a structure emerging from higher latitudes propagating southeastwards. This could be related to ver-tical shear of zonal wind over northern India. The two modes exhibit variability in their intensity on the interannual time scale and contribute a significant amount to the daily rainfall variability in a season. The intensities of the 20–60-day and 10–20-day modes show significantly strong inverse and direct relationship, respectively, with the all-India June–September rainfall. This study also establishes that the probability of occurrence of substantial rainfall over central India increases significantly if the two intraseasonal modes simultaneously exhibit positive anomalies over the region. There also exists a phase-locking between the two modes. In the second part of the thesis, we investigated the changing nature of these intraseasonal modes over Indian region, and their association with extreme rainfall events using ground based observed rainfall. We found that the relative strength of the northward propagating 20–60-day mode has a significant decreasing trend during the past six decades, possibly attributed to the weakening of large-scale circulation in the region during monsoon. This reduction is compensated by a gain in synoptic-scale (3–9 days) variability. The decrease in the low-frequency ISV is associated with a significant decreasing trend in the percentage of extreme events during the active phase of the monsoon. However, this decrease is balanced by a significant increasing trend in the percentage of extreme events in break phase. We also find a significant rise in occurrence of extremes during early- and late-monsoon months, mainly over the eastern coastal regions of India. We do not observe any significant trend in the high-frequency ISV. In the last part of the thesis, we used numerical simulations to understand the observed changes in the ISV features. Using the atmospheric component of a global climate model (GCM), we have performed two experiments: control experiment (CE) and heating experiment (HE). The CE is the default simulation for 10 years. In HE, we prescribed heating in the atmosphere in such a way that it mimics the conditions for extreme rainfall events as observed over central India during June– September. Heating is prescribed primarily during the break phase of the 20–60-day mode. This basically increases the number of extremes, majority of which are in break phase. The design of the experiment reflects the observed current scenario of increased extreme events during breaks. We found that the increased extreme events in the HE decreased the intensity of the 20–60-day mode over the Indian region. This reduction is associated with a reduction of rainfall in active phase and increase in the length of break phase. A reduction in the seasonal mean over India is also observed. The reduction of active phase rainfall is linked with an increased stability of the atmosphere over central India. Lastly, we propose a possible mechanism for the reduction of rainfall in active phase. We found that there is a significant reduction in the strength of the vertical easterly shear over the northern Indian region during break–active transition phase. This basically weakens the conditions for the growth of Rossby wave instability, thereby elongating break phase and reducing the rainfall intensity in the following active phase. This study highlights the redistribution of rainfall intensity among periodic (low-frequency) and non-periodic (extreme) modes in a changing climate scenario, which is further tested in a modeling study. The results presented in this thesis will provide a pathway to understand, using observations and numerical model simulations, the ISV and its relative contribution to the Indian summer monsoon. It can also be used for model evaluation.