Abstract
Accelerating global changes and human impacts threaten the survival of fish communities worldwide, which are critical to the functioning of marine ecosystems and fisheries-dependent populations. Only an effective and rapid monitoring of fish communities at small and large scales to understand their distributions, assembly rules and impacts of human and environmental pressures can enable the implementation of optimal conservation measures. Environmental DNA (eDNA) is a recent method that has been demonstrated to be effective at local and regional scales for studying coastal fish communities. This method makes it possible to overcome some biases induced by conventional monitoring methods (fishing, diving, cameras). The goal of this thesis is to use eDNA to study fish distributions at various spatial scales in response to environmental, geographic, and socioeconomic factors, and then to feed conservation planning approaches. I first compared frequently used bioinformatics tools for eDNA data analysis, identified the best programs and pipelines, and constructed an optimal pipeline for identifying species contained in a sample, in the case of a complete reference database. However, at a large scale, the genetic reference databases are largely incomplete for the 12S mitochondrial gene we use, and thus do not allow for taxonomic assignment of all eDNA fragments. The other studies in this thesis therefore rely on a method of clustering sequences into molecular taxonomic units (MOTUs). From a large-scale dataset sampled in three oceans (Indian, Pacific, Atlantic), I compared estimates of coral reef fish diversity obtained with eDNA and with visual census data, at the scale of several bioregions. I demonstrated that eDNA estimated a higher diversity of families and MOTUs than visual census, while recovering the known major distribution patterns (longitudinal gradient, isolation of the Caribbean fauna). Then, on a global scale, I investigated the influence of environmental, geographical and socio-economic factors on several alpha and beta diversity indices, using more than 500 eDNA samples collected in 11 regions of the world. The results show a dominant effect of the environment (temperature and productivity) on alpha and beta diversity, but also a decrease of these diversities in areas close to human populations, and in particular in countries depending on marine resources. Finally, at a regional scale in New Caledonia, by combining eDNA with more conventional methods (baited cameras and acoustic echosounder), I estimated and modeled several metrics of fish diversity on deep outer slopes and seamounts, down to 600m depth, which I then integrated into three-dimensional conservation planning. These results indicate high richness, abundance and biomass on shallow and isolated seamounts, as well as on the deep slopes of islands and atolls far from urbanized areas, which should be prioritized in conservation plans. All this thesis work demonstrates the utility of eDNA metabarcoding to study fish distribution at fine and large spatial scales, to study the impact of environmental and socio-economic conditions on the diversity and distribution of fish communities, and to inform managers on priority areas for conservation.