Lac du Bourget, in the Alps, is the largest natural French lake. Following major restoration programmes during the 1970s and early 1980s, involving massive efforts to reduce nutrient loads and pollution in the lake, the water quality has improved over the past two decades. This can be inferred from the increase in the nitrate: phosphate ratio, the intensification of the "clear-water phase" (i.e. the increase in the water column transparency in spring), and the reduction in the total phosphorus and chlorophyll a concentrations. However, blooms of the filamentous, phycoerythrin-rich, non-nitrogen fixing and hepatotoxic cyanobacterium Planktothrix rubescens have occurred since 1996 and have been maintained subsequently, at least during summer and autumn periods. Nutrients (especially phosphorus) are usually thought to be one of the most important factors responsible for cyanobacterial blooms, and so the question is asked if this bloom is a paradoxical outcome of the restoration programs? Using a large set of data taken from surveys of Lac du Bourget, from the literature, and from recent laboratory experiments, and also using field data for the neighboring Lake Geneva, we propose a realistic scenario to account for the population dynamics of the cyanobacterium and the occurrence and maintenance of the bloom in Lac du Bourget. The characteristics of the lake (high water column stability, deepening of the nutrient-depleted layer during the last decade, a long water residence time), local conditions (the nutrient load and charge) and global changes (global warming) all had to be taken into account to explain this bloom. We suggest that the success of P. rubescens in Lac du Bourget is probably due to increased transparency and a longer stratified period following (1) the restriction of other phytoplankton species following reduced phosphorus, which has allowed P. rubescens to make use of organic phosphorus to improve its competitiveness; (2) warmer than average winter/spring periods allowing an earlier water stratification and in fine a competitive advantage to P. rubescens; (3) lower than average surface irradiance, which has also given the low-light preferring P. rubescens an advantage. Finally, this study highlights the importance of long-term data sets in attempting to elucidate the global causes of a major ecological problem (such as this cyanobacterial bloom) and impacts with regard to the function and use of freshwater ecosystems. (c) 2004 Elsevier B.V. All rights reserved.