Harvesting direct damage data after the July 2021 flash floods in Wallonia, Belgium: a pilot study

Damage modelling generally remains less mature than other components of the flood risk modelling chain. Existing damage models were calibrated for particular regions and flood types, but their transferability to other contexts is scientifically challenging. The calibration of damage models is currently hampered by a lack of reliable empirical data.

In July 2021, devastating floods occurred in several catchments in North-West Europe. Germany and Belgium were particularly hit. This communication presents ongoing initiatives to collect consistent hazard, vulnerability and impact data in the catchment of river Vesdre, Belgium, where up to 30% of annual precipitation was measured in just 48 h. During the night of July 14-15, 2021, the peak discharge in river Vesdre exceeded by a factor 2 to 4 the 100-year flood. At most stations, the observed flow was the highest on record. This led to tremendous tangible and intangible impacts, including 100+ buildings washed out by the flow. Considerable surprise effects also played a part.

A large-scale field survey is being designed to collect, structure and analyse data on flow variables, building characteristics and experienced damage in the housing sector. For the sake of ensuring the quality of data, particularly on technical aspects such as type of building material or level of ground flood compared to street level, a field survey was preferred to an online questionnaire or phone-based interviews.

As an onset for the large-scale flood damage survey, a pilot study was undertaken in November 2021. With the help of a group of students, about 400 buildings were visited to interview the inhabitants. For only 30 % of them, it was possible to get in contact with the inhabitants, as many buildings were still not inhabited due to the extent of the damages. Only for four buildings out of ten, the respondent was willing to conduct the interview, primarily due to lack of time or lack of interest. Other invoked reasons include trauma, language difficulties, or Covid quarantine. It was noticed that building characteristics are easier to obtain when the inhabitant owns the building and that respondents express a strong expectation on receiving feedback on the study outcomes.

Based on collected data, the mean damage per building is 55 k€. However, although the interviews were conducted four months after the event, only half of the participants were able to provide monetary estimates of the damages. The heating system was by far the most affected component due to its usual positioning in the building basements.

The upcoming large-scale survey will enable calibrating suitable flood damage models and provide new insights into damage mechanisms to contribute to improved cost-effectiveness analyses for risk reduction measures.