Since the catastrophic CO2 eruption in 1986, Lake Nyos has been investigated in detail by several research groups. However, no signs of double-diffusive convection were observed before December 2002, when a set of 26 well-mixed layers with thicknesses of 0.2-2.1 m and sharp interfaces in between were discovered at 53-74 m depth. Such pronounced steps are a characteristic feature of double-diffusive convection of the diffusive regime. A temperature time series measured at 62 m depth indicates that the double-diffusive convection started in the second half of March 2002. The trigger was most probably the cooling at the top of this layer caused by relatively strong seasonal convective mixing down to 52.5 m depth during the dry season in February 2002. The heat fluxes calculated by the heat budget method and the thicknesses of the layers agree within the uncertainties with the values expected from the double-diffusive flux laws. The heat fluxes increased by an order of magnitude since the establishment of the double-diffusive convection and reached values comparable to the heat input by a source of warm and CO2-enriched water to the deepest zone of the water column. In contrast, the CO2 fluxes caused by double diffusion Lire negligible compared to the input by this source. Because the double-diffusive heat fluxes were higher in the upper layers of the staircase compared to the lower ones, the temperature gradient between 60 and 75 m depth approximately doubled from March 2002 to December 2002, whereas the total dissolved solids gradient remained almost constant during this period. Consequently, this process is reducing the stability of the staircase and could potentially lead to a complete homogenization of this zone within a few years. It cannot be excluded that a similar double-diffusive event could have been the trigger of the CO2 eruption in 1986. (C) 2004 Elsevier Ltd. All rights reserved.