Résumé
Thau lagoon is a large Mediterranean coastal lagoons and it supports traditional shellfish farming activities. It has been subject to eutrophication leading to major anoxic events associated with massive mortalities of shellfish stocks. Since the 1970s, improvements have been made to wastewater treatment systems, which have gradually led to oligotrophication of the lagoon. The aim of our study was to determine how the decrease in nutrient inputs resulted in major ecological changes in Thau lagoon, by analysing five decades of time-series (1970–2018) of observations on pelagic and benthic autotrophic communities. We were able to identify two periods during the oligotrophication process. Period 1 (1970–1992) was considered a eutrophic period, characterised by the shift from seagrass dominance to dominance of red macroalgae. Period 2 (1993–2018), characterised by improved eutrophication status, was further divided into three: a transition phase (1993–2003) during which the water column continued to recover but the benthic community lagged behind in recovery and in partial resilience; a regime shift (2003–2006), after which the water column became oligotrophic and seagrass began to recover (2007–2018). Considering anoxia crises as indicators of ecosystem resilience and resistance, we used a generalised linear model to analyse meteorological and environmental data with the aim of identifying the triggers of summer anoxia over the study period. Among the meteorological variables studied, air temperature had the strongest positive effect, followed by the period and wind intensity (both negative effects) and by rainfall in July (positive effect). The risk of triggering anoxia was lower in period 2, evidence for the increasing resistance of the ecosystem to climatic stress throughout the oligotrophication process. At the ecosystem scale and in the long term perspective, the ecological gains related to oligotrophication are especially important in the context of climate change, with more frequent and severe heat waves predicted.