Extraction of Flood-Modelling Related Base-Data from Multi-Source Remote Sensing Imagery

Flooding is one of the most destructive natural hazards, accounting for over a third of all disaster damage worldwide. In particular inless developed countries (LDCs) this is typically attributed to poor planning, lack of warning systems and limited awareness of thehazard. A number of flood risk models have been developed, but have as yet contributed little to mapping and quantifying the risk inLDCs, for several reasons. In addition to limited human and technical capacity, these models require considerable amounts ofcurrent spatial information that is widely lacking, such as landcover, elevation and elements at risk basedata. Collecting those withground-based methods is difficult, but remote sensing technologies have the potential to acquire them economically. To account forthe variety of required information, data from different sensors are needed, some of which may not be available or affordable.Therefore, data interchangeability needs to be considered.Thus we test the potential of high spatial resolution optical imagery and laser scanning data to provide the information requiredto run such flood risk models as SOBEK. Using segmentation-based analysis in eCognition, Quickbird and laser scanning data wereused to extract building footprints as well as the boundaries of informal settlements. Additionally, a landcover map to provideroughness values for the model was derived from the Quickbird image. These basedata were used in model simulations to assess their actual utility, as well as the sensitivity of the model to variationsin basedata quality. The project shows that existing remote sensing data and image analysis methods can match the inputrequirements for flood models, and that, given the unavailability of one dataset, alternative images can fill the gap.