Floods are amongst the most frequent disasters in terms of human and economic impacts. This study provides new insights into the frequency of loss of life at the global scale, mortality fractions of the population exposed to floods, and underlying trends. A dataset is compiled based on the EM-DAT disaster database covering the period 1975 until 2022, extending previous studies on this topic. Flood impact data are analysed over spatial, temporal and economic scales, decomposed in various flood types and compared with other natural disasters. Floods are the most frequent natural disasters up to 1000 fatalities, and flash floods lead to the highest mortality fractions per event, i.e. the number of deaths in an event relative to the exposed population. Despite population growth and increasing flood hazards, the average number of fatalities per event has declined over time. Mortality fractions per event have decreased over time for middle- and high-middle-income countries, but increased for low-income countries. This highlights the importance of continuing and expanding risk reduction and adaptation efforts.

Reliable hazard analysis is crucial in the flood risk management of river basins. For the floodplains of large, developed rivers, flood hazard analysis often needs to account for the complex hydrology of multiple tributaries and the potential failure of dikes. Estimating this hazard using deterministic methods ignores two major aspects of large-scale risk analysis: the spatial–temporal variability of extreme events caused by tributaries, and the uncertainty of dike breach development. Innovative stochastic methods are here developed to account for these uncertainties and are applied to the Po River in Italy. The effects of using these stochastic methods are compared against deterministic equivalents, and the methods are combined to demonstrate applications for an overall stochastic hazard analysis. The results show these uncertainties can impact extreme event water levels by more than 2 m at certain channel locations, and also affect inundation and breaching patterns. The combined hazard analysis allows for probability distributions of flood hazard and dike failure to be developed, which can be used to assess future flood risk management measures.

Increasing amounts of data, together with more computing power and better machine learning algorithms to analyse the data, are causing changes in almost every aspect of our lives. This trend is expected to continue as more data keep becoming available, computing power keeps improving and machine learning algorithms keep improving as well. Flood risk and impact assessments are also being influenced by this trend, particularly in areas such as the development of mitigation measures, emergency response preparation and flood recovery planning. Machine learning methods have the potential to improve accuracy as well as reduce calculating time and model development cost. It is expected that in the future more applications will become feasible and many process models and traditional observation methods will be replaced by machine learning. Examples of this include the use of machine learning on remote sensing data to estimate exposure and on social media data to improve flood response. Some improvements may require new data collection efforts, such as for the modelling of flood damages or defence failures. In other components, machine learning may not always be suitable or should be applied complementary to process models, for example in hydrodynamic applications. Overall, machine learning is likely to drastically improve future flood risk and impact assessments, but issues such as applicability, bias and ethics must be considered carefully to avoid misuse. This paper presents some of the current developments on the application of machine learning in this field and highlights some key needs and challenges.

Hazard analysis is a crucial step in flood risk management, and for large rivers, the effects of breaches need to be taken into account. Hazard analyses that incorporate this overall “system behaviour” have become increasingly popular in flood risk assessment. Methods to perform such analyses often focus on high water levels as a trigger for dike breaching. However, the duration of high water levels is known to be another important failure criterion. This study aims to investigate the effect of including this duration dependency in system behaviour analyses, using a computational framework in which two dike breach triggering methods are compared. The first triggers dike breaches based on water levels, and the second one based on both water-level and duration. The comparison is made for the Dutch Rhine system, where the dike failure probabilities are assumed to conform to the new Dutch standards of protection. The results show that including the duration as a breach triggering variable has an effect on the hydraulic loads and overall behaviour in the system, therefore influencing the risk. Although further work is required to fully understand the potential impact, the study suggests that including this duration dependency is important for future hazard risk analyses.