Inundated Area Delineation Using MODIS Data: Towards a Global Scale Geo-Database of Flood Events

The availability of global and accurate information is the primary factor affecting the possibility of planning and managing effective disaster response strategies, above all in less developed countries. The second determinant factor that avoids the full spreading of remote sensing technologies is cost-effectiveness and steadiness of results. This paper illustrates a straightforward method for rapid retrieval of inundation maps at regional and global scale by processing MODIS data with the Spectral-Temporal Principal Components Analysis and Digital Terrain Model filtering. Case studies are presented for three different vulnerable regions in developing countries struck by a severe river flood during the last year (2005, from spring to fall): India, Pakistan and Romania. For all the events studied it was obtained an overall accuracy greater than 95% and a kappa coefficient grater than 0.70, demonstrating this methodology is very accurate in mapping inundated areas. Moreover, the integration with vector data (such as roads, railways or urbanized areas) may be used to fast detect infrastructure damages at regional and global scale. This work is the first step to develop a global geo-database of flood-affected areas, a basic tool for helping public administrators in efficiently managing natural hazards. This is especially useful for less developed countries, which unfortunately suffer the heaviest damages because of the high density of population and the scarcity of prevention and rapid response strategies