What can we learn from global disaster records about multi-hazards and their risk dynamics?

Abstract. Recent studies have been reporting more extreme, compounding impacts from multi-hazards than from single hazard events owing to complex interrelationships of hazard, exposure and vulnerability in a multi-hazard setting. However, our current understanding of multi-hazard impacts is primarily based on case studies of individual events. To complement this, we examine the disaster records of the global emergency events database EM-DAT for the period 2000–2018 for evidence of multi-hazard risk dynamics. We develop an algorithm to identify multi-hazard events which uses the information on associated hazards as well as spatiotemporal relationships between disaster records in EM-DAT. We then perform a statistical analysis to assess potential risk dynamics in reported impacts of selected hazard pair types. We identified that twice as many hazards are part of multi-hazard events when considering a spatial overlap of at least 25 % and a time lag of at most 1 year between disaster records in addition to the information of associated hazards. These multi-hazard events account for 78 % of the total damages, 83 % of the total people affected and 69 % of the total deaths in the reported disasters. The statistical comparison indicates that there are different patterns of how impacts compound depending on the impact metric as well as the hazard type. However, as a general trend, hazard pairs seem to have at least as or more impact than two isolated single hazards. To capture the patterns and to integrate them into risk analysis and decision making, we propose the development of generic archetypes of multi-hazard risk dynamics. Despite the well-known limitations of EM-DAT related to completeness of the records as well as reliability of the impact data, which prevents detailed analyses of the data, we found the database useful for exploring high-level patterns at the global scale. Nonetheless, the uncertainties and limitations encountered highlight that future research should be directed at improving and supporting the multi-hazard and impact information in EM-DAT.