Improving flood fatality risk assessment for river flooding in the Netherlands

Implications of alternative functions and model resolution variations on mortality and fatalities in the Bommelerwaard

More Info


The number of fatalities due to a potential flood event is traditionally determined utilizing 'mortality functions’. Data of recent large-scale flooding in the Netherlands are not available since the Netherlands was successful in flood prevention. Therefore, only data from the last coastal flood event in 1953 with 1795 direct fatalities are available. The mortality functions are empirical relationships to provide mortality as a function of three explicit flood characteristics, namely water depth, flow velocity, and water level rise rate. Many more factors are included implicitly since the functions were derived from 1953 data. These underlying factors are thus based on the circumstances of the coastal flooding in 1953 and might not be representative anymore for future flood events elsewhere in the Netherlands. The three flood characteristics in current flood risk assessments are determined by means of coarse flood simulations. Since modern software is becoming more advanced, more detailed flood simulations are becoming possible. Therefore, the applicability of the mortality functions needs to be studied if finer model resolutions are used. This report presents the case study of river area the Bommelerwaard in which the validity of the 1953-based functions, possibilities for alternative functions, and finer model resolutions in hydrodynamic models are tested and analyzed with regards to their impact on flood fatality risk. A hydrodynamic model is developed using the new software program D-Flow Flexible Mesh which is able to apply finer resolutions at locations that require more detail. The different model resolutions that are tested are 100m and 25m, and 5m for the area close to the breach. The flood simulations with these model resolutions resulted in similar outcomes for the number of estimated fatalities in this case study. Overall, the 100m model is preferred because it is sufficiently able to indicate the dangerous locations, provide the order of magnitude of the flood characteristics, while it demands short computation times and matches the level of detail of the data of 1953. However, it is recommended to model the area around the breach (‘breach zone’) with finer model resolutions because the resulting higher local peak velocities are relevant for potential building collapse. For the areas around obstacles and underpasses, it is also recommended to use finer resolutions or to make use of 1D objects or fixed weirs. This study concluded that finer model resolutions at dangerous locations have an impact on the individual risk value of the neighbourhood, and this can have consequences for the maximum individual risk value and thus the overall safety standard of a large dike ring. Furthermore, the case study illustrated that compartment dikes have a significant impact on the local mortality because of the high water level rise rates just upstream. It is recommended to look into possibilities to reduce this high local mortality rate and hence, individual risk, for example by optimizing the location and number of compartment dikes or exploring the effects of openings in the dikes. Moreover, this study identified the discussion points in the current Dutch loss of life approach by a literature study, knowledge of recent flood events abroad, and loss of life approaches internationally. Alternative mortality functions are proposed based on the literature and analyzed through sensitivity analyses in the case study. It is recommended to substantiate and take into account the factors water arrival time, improved building characteristics, and age in the loss of life approach. Preventive evacuation is already taken into account in this approach, but in addition, water arrival time can be included by means of fleeing. This study shows that water arrival time has a great effect on the number of fatalities because some areas have relatively large arrival times and this enables inhabitants to flee the area. Emergency response is thereby of crucial importance. Also in 1953 this factor proved to be relevant. The improved building characteristics compared to 1953 are shown to have a limited impact on the absolute number of fatalities in this case study but it reduced the maximum value of the individual risk and is thus of relevance, especially for dike ring areas with large water depths (>2.1m) and high rise rates (>0.5m/h). Moreover, this study underlines the vulnerability of the elderly during flood events. Since the age distribution has shifted since 1953 and significantly more elderly are present in society nowadays, it is relevant to take this explicitly into account. This case study shows that correcting for age can have a significant impact on the number of fatalities. The impact on the individual risk is limited, but this depends on the spatial distribution of the elderly and should be further analyzed. Finally, the individual risk is sensitive to the configuration of the neighbourhoods. It is therefore also recommended to look into more robust approaches to determine the individual risk.