Detecting water bodies in nearly real time over the whole Africa is a challenge addressed by the Geoland-2 project (GMES). This information was found critical by end-users for natural resource management as well as for environmental monitoring. Non-permanent water bodies dynamic is also expected to be affected by climate change and requires appropriate system development for operational monitoring. In the framework of Geoland-2 project, a water bodies detection algorithm has been developed in two version. The Version1 (1km of spatial resolution) is the result of combination of three existing products in order to have the best information available. The Version2 (250m of spatial resolution) is the detection result of a method based on a standardized multispectral transformation (developed by Pekel et al., 2010), the new algorithm is suited to detect the water occurrence in various ecological conditions and appears to be more generic than classification approaches. Implementation of this algorithm allowed processing multi-year 250 m surface reflectance time series acquired by the MODIS instruments. Both versions are designed to be operational in nearly real time and to be externally validated.