Coastal ecosystems, the interfaces between land and sea, are subject to climate change and high anthropogenic pressure. Consequently, most coastal waters are prone to eutrophication. The phytoplankton requires a special attention because of its role of primary producer in marine ecosystems. Recently, Harmful Algae Bloom outbreaks has raised concern worldwide. The thesis aim was to describe and to measure the temporal responses and causalities of the phytoplankton community structure, with the occurrence of a harmful algae, under global changes. To do so, the ecological niche concept and a statistical method were adapted. The Within Outlying Mean Indexes was proposed to refine the Outlying Mean Index analysis by combining its properties with the K-select analysis species marginality decomposition. The subniche dynamics of the species composing the community were studied under environmental conditions hosting low (L) and high (H) Phaeocystis spp. abundance. Subset H was characterized by a large Phaeocystis spp. niche and a high diatom diversity. In subset L, Phaeocystis spp. was subject to great biological constrain suspected to be caused by diatom competition for resources. The phytoplankton diversity productivity was stronger at a seasonal scale than on the long-term. The resource imbalance had no direct link with productivity in the long-term. The long-term invasive species success and its impact on productivity is favored by successive cold years with high resource imbalance which rise the number of small species and its bloom. I finally discussed on the methodological improvements, the potential use of the trait-based approach, and possible experimental set-ups to support the thesis results.