Impact of multi-hazard interactions on risk assessment

Multi-hazard process chains are complex event involving multiple hazardous processes. Common examples are co-seismic landslides and simultaneous triggering of flash floods and landslides due to extreme precipitation. Due to the complexity, and lack of reliable modelling methods, relatively little is understood about the impact of multi-hazard interactions on risk assessment. We present an integrated multi-hazard analysis of impact from tropial cyclones on Dominica. Through general extreme value analysis, design storms with return periods between 1 and 100 years were produced. Using the LISEM model, hydrology, flash floods, slope stability, slope failure and landslide/debris flow runout was simulations for each potential event. The model was calibrated with high accuracy on the 2017 impact of Hurricane Maria. The results provide counter-intuitive insights into how multi-hazard interactions influence events. In a significant number of cases, increased trigger intensity does not equate to increased impact, or surface hazard intensity. For tropical cyclones in particular, insurance rules or policy-related documents can be based on the return period of the triggering event (e.g. hurricane peak wind velocity). Secondly, many types of interactions shaped the dynamics of the simulated events, as well as the impact from real events on Dominica. Dilution of mass movements increases runout, and additional drag leads to a very strong increase in deposition near urbanized areas. Investigation of physical processes in such situations should strongly consider the relevance of these interactions. Finally, the simulation ensemble highlights the spread in model outcomes due to intrinsic uncertainties. While these are significant due to complexity of the model, they are insignificant compared to the errors made in traditional approaches where hazards are considered in isolation. While this is partly due to the high intensity of the triggering events on Dominica, it provides evidence that the recent developments in integrated multi-hazard models have reached discussion of applicability