Probabilistic cascading multi-hazard risk assessment methodology using graph theory, a New Zealand trial

Abstract

Probabilistic multihazard risk assessment from natural hazard is still a challenge today. Current limitations are the number of different hazards that can be included in the assessment, the capacity to output detailed spatial results and access to data for inverse fitting models. A novel quantitative multi-hazard framework is proposed which permit to model the interactions and triggering of perils of different types. Franz Josef on the West Coast of South Island, New Zealand, is used as a case study where the interactions and impacts of earthquake, rainfall, landslide, landslide dam and flooding hazards on the road network, stopbanks and housing is stochastically quantified. The results show that losses are earthquake-dominated, while rarer events show an influence of landsliding on the losses. Due to the complex interaction between hazards allowed by the modelling framework, while initial perils reach a “loss ceiling” with reducing probabilities, it also supercharges secondary perils that can cause greater and greater losses. Combining a network framework with iterative simulations is shown to provide inherent advantages by its discrete nature. In-depth exploration of the data outputs grants insight into the interconnected disaster system on a granular level. The risk assessment herein pointed out the seriousness of the multihazard threat to the Franz Josef township.