Phytoplankton time‐series enable the observation of recurrent seasonal patterns on community composition, but drastic shifts in community structure are rarely observed, unless linked to anthropogenic impacts (e.g. changes in nutrient inputs). Here, we report a unique case of non‐seasonal, multiannual regime shift of phytoplankton community structure in Lake Kivu, East Africa, and explore the possible causes for such changes. We recorded phytoplankton biomass every 2 weeks over a period of 12 years (from 2002 to 2015). Phytoplankton community previously dominated by cyanobacteria and diatoms changed considerably from 2012, with a conspicuous rise in the biomass of green algae (from 3 to 33% lake‐wide), and a decline of cryptophytes and cyanobacteria along with an increase in total chlorophyll a concentration. Concomitantly, we observed significant differences in the mixed layer depth between the two periods, resulting from changes in the thermal gradient. The euphotic zone depth was also reduced in 2012–2015, probably from the higher phytoplankton biomass. Moreover, mean particulate nitrogen and phosphorus increased as well as the particulate N:P ratio. Desmids (Cosmarium laeve) now dominate the mixolimnion, probably due to atelomixis, observed in high‐frequency temperature measurements. We believe that decadal or multidecadal variability of climate on the Indian Ocean may be the reason for this regime shift in thermal stratification, with a strong impact in phytoplankton community structure. In conclusion, our study demonstrates that tropical lakes may be subject to non‐seasonal fluctuations that generate multiannual regime shifts. These regime shifts have probably never been reported previously because of the lack of long phytoplankton time‐series in low latitudes.

Lake Kivu, East Africa, is a deep oligotrophic and meromictic lake containing high amounts of dissolved methane (∼55-60 km 3 ) and carbon dioxide (∼300 km 3 ) in its deep waters. Methane harvesting for energy production began in 2015, and a monitoring programme was set up to assess possible disturbance on the ecosystem. Phytoplankton biomass and composition was assessed twice per month or monthly at 2 monitoring sites between June 2005 and December 2019, based on HPLC analysis of chlorophyll a (Chl-a) and marker pigments. This long-term series shows that significant changes occurred around 2010 in the lake phytoplankton, with a notable increase of Chl-a and changes in the assemblage toward an increase in non-motile green algae and diatoms. To assess possible changes due to methane harvesting, we compared 2 periods, 2012-2014 and 2018-2019. Chl-a concentration decreased slightly in 2018-2019 compared to the reference period of 2012-2014, and significant changes occurred in composition of the phytoplankton assemblage. In terms of relative contribution to Chl-a, diatoms increased from 26% to 46%, whereas green algae decreased ∼2-fold, from 35% in 2012-2014 to 18% in 2018-2019. Multivariate analyses showed that phytoplankton composition was influenced by seasonal and interannual variations of limnological variables related to changes in meteorological factors. To assess possible future changes due to methane exploitation, we recommend increasing sampling frequency and taxonomic resolution, as well as improving environmental data acquisition.

No abstract