Abstract : Europe witnessed a spate of record-breaking warm seasons during the 2000s, inducing particularly strong societal and environmental impacts. This thesis aims to contribute to the understanding of physical mechanisms responsible for such extremes, in order to anticipate their responses to future climate change. Involved processes are assessed by both statistical data-analysis of observations and climate projections and regional modeling experiments. First we show that while the inter-annual European temperature variability appears driven by disturbances in the North-Atlantic dynamics, the recent warming is likely to be dissociated with potential circulation changes. This inconsistency climaxes during the exceptionally mild autumn of 2006. Recent warm surface conditions in the North-Atlantic ocean seem to substantially contribute to the European warming in autumn--winter, while more local processes appear predominant in spring--summer (e.g., influence of soil moisture). Then multi-model analysis, based on climate projections of the International Panel on Climate Change, show that the inconsistency noticed over recent decades is likely to continue in future years. In particular, it appears that climate change does not affect the internal variability of the North-Atlantic dynamics in a significant manner. Thus future temperature extremes should be associated with circulations similar to those observed during recent episodes. In this context the European winter of 2009/10 can be considered as a cold extreme in a warming climate.