Data integration for rapid flood mapping and impact assessment

Abstract

Efficient flood disaster response and recovery need rapid flood mapping and impact assessments. Increasing amounts of data are available in near real-time from a growing diversity of sensors and data sources. Therefore we need to improve our capabilities of extracting and integrating knowledge from heterogeneous data sources. Using the June flood 2013 in Dresden as a use case, this study investigates whether the combination of multiple data sources provides more reliable and more accurate inundation depths maps for rapid flood impact assessments. In this example, we use remote sensing-based flood masks, water level observation and inundation depth-estimates extracted from photos shared in social media streams for inundation mapping. We assess the reliability and accuracy of the inundation maps and compare them with a combined multi-source inundation map as well as an official reference map. The results indicate that the combination of different data sources helps to close gaps in individual data sources. Further, it has a compensatory effect, which results in more reliable and more accurate inundation maps.