Item

Abstract

Flash floods cause high numbers of casualties and enormous economic damage. Good knowledge of the damage processes is crucial for the implementation of effective flash flood risk management. However, little is known about the damage processes that occur during flash floods, despite their severity. To gain more knowledge, independent data collection initiatives were carried out in the affected areas of Belgium and Germany after the 2021 floods. The resulting datasets include 420 damaged residential buildings in the Vesdre valley in Belgium, 277 in the Ahr valley in Rhineland-Palatinate (Germany) and 332 in North Rhine-Westphalia (Germany). A total of 30 potential damage-influencing variables were harmonized across the regions, providing valuable insights into hazard characteristics, the vulnerability of exposed assets, the coping capacity of inhabitants, and socio-economic factors. Machine learning-based analysis reveals the significant importance of hazard variables, such as water depth and sediment transport, particularly for building damage. In addition to these, exposure (living area) and physical vulnerability factors (building type and wall type) also play a role in determining building damage across the affected regions. For content damage, besides water depth and living area, socio-economic vulnerability (ownership status of the building) and emergency measures were found to be important predictors. These key drivers of building and content damage from flash floods can be utilized to develop more accurate damage models, thereby improving flash flood risk assessments, enhancing risk communication, and supporting better preparedness strategies.