The eastward winds that form each winter in the Arctic stratosphere are intermittently disrupted by planetary-scale waves propagating up from the surface in events known as stratospheric sudden warmings. It is shown here that following roughly half of these sudden warmings, the winds take as long as three months to recover, during which time the polar stratosphere evolves in a robust and predictable fashion. These extended recoveries, termed here Polar-night Jet Oscillation (PJO) events, are relevant to understanding the response of the extratropical troposphere to forcings such as solar variability and climate change. They also represent a possible source of improvement in our ability to predict weather regimes at seasonal timescales. Four projects are reported on here. In the first, the approximation of stratospheric radiative cooling by a linear relaxation is tested and found to hold well enough to diagnose effective damping rates. In the polar night, the rates found are weaker than those typically assumed by simplified modelling studies of the extratropical stratosphere and troposphere. In the second, PJO events are identified and characterized in observations, reanalyses, and a comprehensive chemistry-climate model. Their observed behaviour is reproduced well in the model. Their duration correlates with the depth in the stratosphere to which the disruption descends, and is associated with the strong suppression of further planetary wave propagation into the vortex. In the third, the response of the zonal mean winds and temperatures to the eddy-driven torques that occur during PJO events is studied. The collapse of planetary waves following the initial warming permits radiative processes to dominate. The weak radiative damping rates diagnosed in the first project are required to capture the redistribution of angular momentum responsible for the circulation anomalies. In the final project, these damping rates are imposed in a simplified model of the coupled stratosphere and troposphere. The weaker damping is found to change the warmings generated by the model to be more PJO-like in character. Planetary waves in this case collapse following the warmings, confirming the dual role of the suppression of wave driving and extended radiative timescales in determining the behaviour of PJO events.

The eastward winds that form each winter in the Arctic stratosphere are intermittently disrupted by planetary-scale waves propagating up from the surface in events known as stratospheric sudden warmings. It is shown here that following roughly half of these sudden warmings, the winds take as long as three months to recover, during which time the polar stratosphere evolves in a robust and predictable fashion. These extended recoveries, termed here Polar-night Jet Oscillation (PJO) events, are relevant to understanding the response of the extratropical troposphere to forcings such as solar variability and climate change. They also represent a possible source of improvement in our ability to predict weather regimes at seasonal timescales. Four projects are reported on here. In the first, the approximation of stratospheric radiative cooling by a linear relaxation is tested and found to hold well enough to diagnose effective damping rates. In the polar night, the rates found are weaker than those typically assumed by simplified modelling studies of the extratropical stratosphere and troposphere. In the second, PJO events are identified and characterized in observations, reanalyses, and a comprehensive chemistry-climate model. Their observed behaviour is reproduced well in the model. Their duration correlates with the depth in the stratosphere to which the disruption descends, and is associated with the strong suppression of further planetary wave propagation into the vortex. In the third, the response of the zonal mean winds and temperatures to the eddy-driven torques that occur during PJO events is studied. The collapse of planetary waves following the initial warming permits radiative processes to dominate. The weak radiative damping rates diagnosed in the first project are required to capture the redistribution of angular momentum responsible for the circulation anomalies. In the final project, these damping rates are imposed in a simplified model of the coupled stratosphere and troposphere. The weaker damping is found to change the warmings generated by the model to be more PJO-like in character. Planetary waves in this case collapse following the warmings, confirming the dual role of the suppression of wave driving and extended radiative timescales in determining the behaviour of PJO events.

L'étude de la variabilité climatique dans les moyennes et hautes latitudes est très complexe, principalement en raison des nombreux mécanismes physiques mis en jeu. Cette variabilité climatique résulte de deux contributions majeures : la variabilité interne associée à des processus internes au système climatique et la variabilité forcée qui est liée aux forçages externes, qui peuvent être d'origine naturelle (comme le volcanisme, les aérosols naturels) ou anthropique (GES, aérosols anthropiques). Ces forçages externes jouent un rôle important sur le climat et sa variabilité. Le défi de la recherche climatique est de comprendre leurs effets sur le climat et leurs rôles face à la variabilité interne. Cette thèse a comme objectif une meilleure compréhension des rôles respectifs de la variabilité interne et des différents forçages externes sur la variabilité passée de la circulation atmosphérique dans les deux hémisphères et de l'activité synoptique associée, ainsi que de leur évolution pour le climat futur en utilisant des ré-analyses atmosphériques et des simulations issues du modèle de climat CNRM-CM5. Nous nous sommes intéressés dans un premier temps à l'étude des changements des modes annulaires dans les hémisphères nord et sud, appelés respectivement NAM (Northern Annular Mode) et SAM (Southern Annular Mode). Nous montrons que la tendance positive observée du SAM sur la période 1960-1990 en été austral est bien reproduite par le modèle. Cependant, celle-ci est reproduite lorsque la diminution d'ozone stratosphérique et l'augmentation des gaz à effet de serre sont toutes deux prescrites au modèle, alors que certaines études suggéraient qu'elle était principalement expliquée par la seule présence de l'ozone. Nous nous sommes ensuite focalisés sur les changements de circulation \linebreak atmosphérique dans l'hémisphère nord. Ceux-ci sont plus complexes que dans l'hémisphère sud. En effet, l'augmentation des GES provoque un réchauffement global différent suivant l'altitude : maximum dans la haute troposphère tropicale ainsi qu'aux hautes latitudes en surface. Ce dernier est en réalité expliqué par la fonte de la banquise Arctique. De nombreux processus physiques et de nombreuses rétroactions sont mis en jeu et rendent la compréhension des changements compliquée. Pour cela, nous avons mis en place un protocole expérimental avec le modèle couplé CNRM-CM5 pour séparer l'effet direct des GES (réchauffement de l'atmosphère) de l'effet indirect (fonte de la banquise Arctique et rétroaction sur l'atmosphère). Cette fonte est responsable de l'augmentation des flux de chaleur entre l'océan et l'atmosphère qui perturbe ensuite la circulation atmosphérique. Nous montrons que la diminution de glace de mer Arctique pourrait être à l'origine de la réponse barocline dans l'hémisphère nord. Enfin, la dernière partie de cette thèse a fait l'objet d'une étude plus spécifique, puisque l'on s'est intéressé aux évolutions passée et future des dépressions atmosphériques nord Atlantique. De nombreuses incertitudes persistent en raison de processus complexes mettant en jeu la vapeur d'eau, le gradient méridien de température et la stabilité statique. Nous retrouvons la réponse tripolaire dans le RCP8.5, mentionnée dans le 5\up{ème} rapport du GIEC, qui correspond à une diminution dans la partie sud du domaine Atlantique nord et la mer Méditerranée, une augmentation sur les îles Britanniques et la Scandinavie et une diminution à l'est du Groenland. Nous montrons que ce signal est peu robuste sur la période dite historique en raison de la forte variabilité interne associée aux systèmes dépressionnaires. / Climate variability in mid and high latitudes is very complex due to numerous physical mecanims implied. This climate variability can be decomposed into 2 components : the internal variability associated with internal processes and the forced variability linked to the external forcings which can be natutal (volcanism, natural aerosols) or anthropogenic (greenhouse gases, anthropogenic aerosols). These external forcings play a crucial role on the climate and its variability. The challenge in the climate research is to understand their effects on the climate and their roles relatively with the internal variability. The objective of this thesis is a better understanding of the respective roles of internal variability and forced variability on the past and future atmospheric circulation in both hemispheres characterized by the annular mode and the synoptic activity associated using atmospheric reanalysis and experiments performed with the coupled climate model CNRM-CM5. First, we focus on the annular mode changes in both hemispheres, named the NAM (Northern Annular Mode) and the SAM (Southern Annular Mode). We show that the observed positive trend of the SAM in the 1960s in austral summer is well reproduced by the climate model. However, contrarily to other studies which suggest that this positive trend can be explained by only stratospheric ozone depletion, it is reproduced in the CNRM-CM5 model when the ozone depletion and greenhouse gases (GHG) increase are both prescribed. Then, we investigate the changes in the Northern Hemisphere atmospheric circulation. These are more complex than in the Southern Hemisphere. Indeed, the increase of GHG in the atmosphere causes a general global warming maximum in the tropical high troposphere and over the pole at the surface which is mainly explained by Arctic sea ice loss. So the understanding of the changes is very complex due to several physical processes and retroactions. Thus, we have conducted a protocol with the coupled climate model CNRM-CM5 in order to assess the respective role of Arctic sea ice loss and GHG increase. Arctic sea ice loss is reponsible for an increase in the heat flux between the atmosphere and the ocean which modify the atmospheric circulation. We show that Arctic sea ice loss can cause the baroclinic response in the Northern Hemisphere. Finally, the last part of the thesis is the study of past and future changes in the North Altantic storm-tracks. There are still sereval uncertainties because of the complex processes involving the water vapour, the meridional temperature gradient and the static stability. We find the tripolar response, already found by other studies, consisting of a significant decrease in the south of the basin and over the Mediterranean sea, a small increase over the British Isles, and a decrease east of the Greenland. We show that the signal in the historical period is not robust, due to large chaotic variability associated with storms.