Explosive volcanic eruptions generate far-reaching impacts, as tremendous volumes of volcanic ash are ejected into the atmosphere from a plume that can intersect airspace at all flight levels. Ash is transported long distances by wind advection and dispersed by diffusion processes before settling, presenting significant hazards—including to aviation and critical infrastructure. Accurate hazard prediction is difficult, due to high levels of uncertainty in eruption intensity and variability in atmospheric fields driving dispersal. Volcanic tephra transport and dispersal models simulate ash movement based on atmospheric forcing data and a specified eruption source term. However, due to the rarity of large eruptions and the lack of high-resolution observations during events, a single simulation cannot capture the full uncertainty in dispersion outcomes, and a probabilistic ensemble modelling approach is typically employed. Theresearch presented in thesis focuses on efficient, statistically-coherent ensemble design for operational probabilistic ash hazard forecasting. Ensemble size are commonly chosen pragmatically, based on time or computational constraints. This research introduces a statistical framework for standardising ensemble design and improving communication of hazard uncertainties. It further shows how confidence in hazard estimates can be enhanced through stratified sampling of dispersion model inputs, informed by the variability in atmospheric forcing data—offering potential computational savings. The second component of this research efficiently quantifies the combined variability in forecasts, exploiting the linearity of dispersion model solutions to demonstrate that post hoc incorporation of eruptive magnitude variability is straightforward. In operational settings, and where eruptive histories are sparse, the joint variability in plume height and eruption intensity can be characterised by Bayesian methods based on their relationship and historical data. Together, these approaches enable more complete representation of uncertainties, improving ensemble design and the robustness of probabilistic forecasts for both future planning and real-time decision-making.
The Statistical Design of Assessments of Ash Hazard Impacts from Explosive Volcanic Eruptions
Williams, S. L. (Author). 30 Sept 2025
Student thesis: Doctoral Thesis › Doctor of Philosophy (PhD)