Thesis: Estimating the sea levels of the natural marine flooding hazard to be taken into account for the protection of nuclear power plants is an important safety issue for IRSN. To estimate sea levels corresponding to high return periods, a local statistical analysis can be conducted. However, the low quality of the data, due in particular to short observation times, results in high estimation uncertainties. Additional information, regional and/or historical, is used in response to this problem. Regional information consists of evaluating data available at sites close to the reference site, and historical information consists of evaluating events observed before the beginning of recordings at the reference site. If the use of additional information strengthens the statistical adjustments, it also poses methodological difficulties .Extreme sea levels are typically estimated by fitting a statistical distribution to stochasticmeteorological skew surges, then doing a convolution with the empirical distribution of predictable astronomical high tides. The historical observations are major sea levels extracted from the archives from which the associated high tide surges can be estimated, but the exhaustiveness of the latter cannot be guaranteed. Indeed, if a skew surge coincides with a high tide of low or moderate intensity, then it does not systematically generate an extreme sea level that has left a trace in the archives. The problem related to historical information is therefore its possible non-exhaustiveness which can lead to biased estimates. Regional information generally is composed of two steps, the formation of homogeneous regions and the regional estimation. The resulting issues are therefore to choose a relevant method to form the regions and a relevant method for the regional estimation, and to take into account the possible dependencies between the sites of a region. A few studies have already attempted to respond to the problems associated with regional and historical information, but can be further improved. Moreover, the combination of regional and historical information has been treated very little up to now. The thesis therefore aims to develop methods to properly take into account regional and/or historical information for the estimation of extreme sea levels. The proposed approaches are illustrated on data from European tide gauges located on the coast of the Atlantic Ocean, the English Channel and the North Sea