Abstract
Flood is causing devastating damages every year all over the world. One way to improve the readiness of the different stakeholders is by providing flood extent and depth maps promptly after the disaster, preferably in an automated way to reduce costs. The availability of these flood maps becomes particularly vital to assist the local authorities to plan rescue operations and evacuate the premises promptly. In the event of flooding, a clear cloud-free image acquired instantaneously, is necessary to have a synoptic view of the affected area. In this context, remotely-sensed images are suitable to map inundations, particularly when harsh climatic conditions are encountered and the access to the affected site is impractical. Moreover, satellite-borne Synthetic Aperture Radar (SAR) sensors have been extensively used in the last two decades to monitor many flooding events by taking advantage of their ability to operate independently of the sunlight, and in cloudy conditions which are common during inundations. In the majority of previous studies working with SAR images for the detection of floods, the inundation extent is essentially the only information extracted. Although, for certain applications like the assessment of the damages caused, additional inundation characteristics are needed to give a thorough analysis of the inundation hazard, like the water level. The major advantage space-borne acquisitions have specifically over gauging stations is the global-availability and the spatial-continuity of their data. A semi-automated process was proposed in this thesis to estimate the flood depth locally in the vicinity of an inundated building from a pair of high-resolution SAR images using Genetic Algorithms followed by the inversion of an urban backscattering model. One potential application of this method is to assist insurance companies in the assessment of the damages incurred by buildings and structures in flooded urban areas. Another way satellite SAR imagery can support decision makers in increasing the preparedness, is through flood extent maps. The online web application presented in this thesis addresses the issue of the flood extent mapping from SAR images promptly following the disaster, using a supervised classifier trained automatically without any intervention from the human user. This web application allowed to delineate the extent of the flooding, and managed to reach an accurate classification of the SAR image in a reasonable time. The important advantage to emphasize is the fact that the whole process is quick and automatic, which makes it useful to assist response authorities and the affected communities during emergency situations.