Pluvial flood depth mapping in urban areas using data fusion

Inundations of urban areas induced by extreme rainfall are an increasingly important driver of loss and damage. With climate change, locally heavy precipitation will occur more frequently and with greater intensity. For efficiently reducing flood impacts and informing precautionary measures rapid and reliable information on affected areas is essential. Increasing amounts of data are available from a growing diversity of sensors and data sources. The Increasing volume and velocity of data are auspicious but require improved capabilities of extracting and integrating knowledge from this wide variety of data. Using recent pluvial flood events in Budapest (Hungary), Dresden (Germany), and Braunschweig (Germany) we investigate whether the combination of data from multiple sources (remote sensing, simulation models, online media, VGI) provides more reliable and more accurate inundation depths maps to better inform the assessment and management of pluvial floods. We combine data with geospatial analysis methods and fuse the different datasets using statistical and ML-based approaches. The results indicate that the combined data sources help to close gaps in individual data sources. Further, we note a compensatory effect, which results in more reliable and accurate inundation maps.