Quantifying asset and visitor risk at Mt. Taranaki, New Zealand from multiple volcanic hazards

We present a probabilistic quantification of multiple volcanic hazards in an assessment of risk to visitors and assets in Egmont National Park, New Zealand. The probability of impact to proposed park infrastructure from volcanic activity (originating from Mt. Taranaki) is quantified using a combination of statistical and numerical techniques. While single (volcanic) hazard assessments typically follow a methodology where the hazard source (e.g. pyroclastic flow, ashfall, debris avalanche) is the focus and defines an area of impact, our multi-volcanic hazard assessment uses a location-centred methodology where critical locations are used to define the range of hazard sources that affect risk over park asset lifetimes. Key to this process is creating fast (i.e. linear/functional) mappings between hazard source parameters such as volume and impact parameters such as depth. These mappings can then be combined with stochastic models to find the probability of input parameters and the probability of eruptions generating these input parameters. For some hazards, such as ash fall, statistical models are available to map intensity to probability. However, mass flow hazards required the use of Gaussian process emulation to develop a computationally cheap surrogate to numerical simulations that can be efficiently sampled for probabilistic hazard assessment. This was a suitable alternative when statistical models for the hazard are unavailable. Our study demonstrates the use of these techniques to integrate stochastic and deterministic models for probabilistic volcano multi-hazard assessment.