Contribution à l'étude de la dynamique du thon rouge Atlantique - Approches écologiques et océanographiques

[1] The spatial and temporal dynamics of Atlantic bluefin tuna, Thunnus thynnus, are here investigated through data analysis and statistical modelling. The local, regional and global scales are considered successively. A phenomenological approach is favoured, as well the extended use of image processing tools and estimation theory techniques. Their potential role in partially observed or ill-defined problems in marine ecology is presented and discussed. The acute need for frequency-explicit methods is in particular stressed out. Hypothesis testing and significance testing of ecological space/time processes are further discussed. [2] The first chapter depicts how surface fronts can structure the pelagic habitat: dedicated detection and mapping tools are introduced, along with an example in the North-East Atlantic and the Mediterranean regions. The influence of such heterogeneity on top predators is then discussed and illustrated with a case study in the gulf of Lions: high resolution datasets derived from aerial surveys and satellite radiometry are merged for this purpose. Monte-Carlo methods are shown to be of great value to assess the significance of observed spatial patterns. [3] The second chapter introduces an individual-based view of the pelagic habitat, through the use of archival tags. Three major drawbacks of this approach are first listed, i.e. (i) the large geolocation errors linked to hardware and software issues, (ii) the non-linearity and ill-posedness of the initial estimation problem, and (iii) the lack of accurate and continuous 3D ancillary data sources. A Monte-Carlo method is proposed to address these problems, and applied first to a controlled case. A real world application is then presented, involving pop-up tags and 3D temperature outputs from the Mercator model in the Gulf Stream area. Some specific technical points and issues to resolve are pinpointed. [4] In the third part of this work, an extended time series analysis is conducted on the long-term fluctuations in the bluefin tuna population, as derived from coastal trap catches in the Mediterranean sea. Linear and non-linear general techniques are employed to uncover deterministic features in these series. A direct and delayed density-dependence is in particular extracted from the high frequency component of the data. Possible endogenous and exogenous causes for the low and high frequency are discussed, as well as the need for new significance tests in studies involving long term changes.