There are a large number of research vessels and fishing vessels equipped with echosounders plying the world ocean, making continual observations of the ocean interior. Developing data collation programmes (e.g. Integrated Marine Observing System) and automated, repeatable analyses techniques enable the upper c. 1,200 meters of the world ocean to be sampled routinely, and for their characteristic deep scattering layers (DSLs) to be compared. Deep scattering layers are comprised of zooplankton (e.g. euphausiids) and fish, particularly myctophids or lantern fish, and comprise the majority of sub-surface biomass. Here we present, by the analysis of a global acoustic dataset, a mesopelagic biogeography of the sea. This was accomplished by (i) the collation and processing of a global active acoustic dataset, (ii) the development of a standardised and automated method of sound scattering layer (SSL) extraction and description, (iii) the derivation of the environmental drivers of DSL depth and biomass, (iv) the definition of a mesopelagic biogeography based on the drivers of DSL metrics and (v) the prediction, using output from the NEMO-MEDUSA-2.0 coupled model, of how the metrics and biogeography may change by 2100. Key findings include, the development of the Sound Scattering Layer Extraction Method (SSLEM) the inference that primary production, water temperature and wind stress are key drivers in DSL depth and biomass and that mesopelagic fish biomass may increase by 2100. Such an increase is a result of increased trophic efficiency from the shallowing of DSLs and rising water temperatures, suggesting, that as the climate warms the ocean is becoming more efficient. The biophysical relationships and biogeography derived here, serve to improve our understanding of mesopelagic mid-trophic level dynamics in open-ocean ecosystems. This will aid both fisheries and conservation management, which now adopt more holistic approaches when monitoring and evaluating ecosystem health and stability.

Du fait de l’isolement géographique de l’océan austral et des conditions météorologiques qui y règnent, la collecte de données océanographiques par les moyens conventionnels est particulièrement couteuse dans cette région du globe. Pour pallier à ces difficultés, l’utilisation d’enregistreurs électroniques embarqués sur des prédateurs marins, et notamment sur les éléphants de mer austraux, s’est révélée être une approche intéressante. Les femelles éléphants de mer s’alimentent en grande partie de petits poissons méso-pélagiques bioluminescents, les myctophidés. Leur forte abondance et leur comportement de migration nycthémérale confèrent à ces organismes un rôle écologique de première importance dans l’océan austral. Cependant, la distribution spatiale des proies de l’éléphant de mer, et les processus physiques et biologiques qui influent sur la dynamique de cette distribution sont encore mal connus. Cette thèse se propose d’enquêter sur ces sujets à partir des données à haute fréquence d’échantillonnage collectées par des femelles éléphants de mer. L’analyse du comportement de plongée des femelles éléphants de mer, en relation avec les variations du taux de rencontre de proie dont les données d’accélération permettent d'avoir une estimation, est au cœur des différents travaux développés dans cette thèse. Les résultats obtenus suggèrent notamment, (1) une diminution de l’abondance des proies avec la profondeur ainsi qu’une homogénéisation de leur distribution, (2) une distribution en couches de cette ressource, (3) l’intervention de contraintes verticales délimitant l’étendue verticale de ces couches et ainsi capable de moduler leur densité, (4) au sein des couches, une dispersion relativement importante des proies en comparaison de la portée des capacités de perception du prédateur, et enfin (5) un rôle significatif des tourbillons méso-échelle et de leur bordure sur la structuration de la distribution des ressources alimentaires de l’éléphant de mer durant l’été austral. / Because of the remoteness and harsh meteorological conditions of the southern ocean, data sampling is more costly in this area. Use of electronic devices attached to marine predators (Bio-logging), such as southern elephant seals, has emerged as an interesting approach to cope with this problem. Female southern elephant seals primarily feed on small bioluminescent meso-pelagic fishes called myctophids. Because of their large abundance and of their diel vertical migration behaviour, these organisms have a major ecological importance in the southern ocean. However, the spatial distribution of the elephant seals prey, as well as the bio-physical processes affecting the dynamics of this distribution, are still poorly known. This thesis intends to investigate this issue using high sampling frequency bio-logging data collected by female southern elephant seals. This work is based on the analysis of elephant seals diving behaviour in relation to changes in the occurrence of prey encounter events detected from acceleration data. Our results suggest that (1) prey abundance decreases with depth and that their distribution tend to standardize, (2) prey are distributed into layers, (3) vertical constraints could modulate the prey density by acting on the vertical spread of these layers, (4) prey items are well dispersed in comparison to the perception range of elephant seals, and finally (5) that meso-scale eddies, notably their edges, play a structuring role in the prey distribution during the austral summer.