Résumé
We introduce a new SAR based flood extent mapping algorithm enabling systematic and automatic monitoring of water bodies at large scale in near real time. The algorithm is both efficient and robust, especially in areas where flood events are not sporadic but long lasting (e.g. monsoon-related floods). It is based on the regular processing of subsequently acquired pairs of Sentinel-1 images. The algorithm has been developed in the framework of the ESA-funded e-DRIFT project, with the aim to respond to the needs of the disaster risk financing sector in Southeast Asia. The approach has been validated after intensive testing over different areas of interest in South East Asia i.e. Myanmar and Laos, where risks associated with flooding are currently not well-known. Moreover, the algorithm is implemented on a virtual platform that efficiently handles large collections of Sentinel-1 data from all the orbits and dates available over areas of interest affected by floods. The output of this near real-time system are reliable and useful input data for the parametric modelling carried out by the (re-)insurance companies, allowing them to better anticipate risk of natural disasters.