Simulation of catastrophic volcanic debris avalanches from source to impact : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Earth Science at Massey University, Palmerston North, New Zealand

Abstract

Volcanic debris avalanches are rapid, complex gravitational flows triggered by the collapse of unstable portions of volcanic edifices. Their significance as a major volcanic hazard was only fully recognised after the 1980 Mount St. Helens eruption, in which a cryptodome emplacement triggered a catastrophic flank collapse (2.5 km³) and a large lateral blast. Real-time observations and eyewitness accounts from this event greatly advanced understanding of large-volume debris avalanche processes (Glicken, 1996). Subsequent research has focused on better characterising these processes and improving hazard predictions. These collapses represent destructive phases within a volcano’s lifecycle and depend on factors such as edifice morphology, eruptive activity, and external conditions, including rainfall or earthquakes. Some volcanoes, such as Mount Taranaki (New Zealand), show recurrent collapse behaviour (Zernack, 2021). Although their frequency varies, debris avalanches can be highly destructive, as they can exhibit very large volumes, extreme mobility, and high velocities. Such mobility is commonly enhanced by the incorporation of water and sediment, which alters the flow rheology from dry frictional to fluidised, as seen in the 1998 Casita collapse (Nicaragua), where a small volume (200,000 m³) failure rapidly transformed into a deadly lahar, resulting in 2,500 fatalities. Events of this magnitude occur worldwide every few years, underscoring their persistent threat to nearby populations and infrastructure (Siebert, 1992; McGuire, 1996). Debris avalanches are inherently multi-stage phenomena, comprising pre-failure, failure, and post-failure phases (van Asch et al., 2007). Each stage involves distinct processes, from quasi-static deformation to fully dynamic flow, complicating robust hazard estimation. Most attempts to model debris avalanches (del Potro et al., 2013; Ball et al., 2018; Reid et al., 2010; Procter et al., 2021) focus on simulating a single stage, treating it independently from the others. Failure is usually assessed with slope-stability analyses, from limit-equilibrium methods (LEM) to continuum-mechanics approaches, based on material properties, pore pressures, and slope geometry (e.g., Gu et al., 2015; Acevedo et al., 2021; Ji et al., 2025). Post-failure dynamics is commonly modelled with depth-averaged approaches (e.g., Sheridan et al., 2010; Mead and Magill, 2017; Charbonnier et al., 2015; Ding et al., 2023; Wolff et al., 2025), ranging from single-phase models treating the avalanche as a homogeneous mass to multi-phase models capturing solid-fluid interactions. Attempts to couple these stages remain limited, and comprehensive physics-based simulations are rare, often relying on empirical simplifications or producing semi-quantitative outputs (e.g., Mergili et al., 2012; Horton et al., 2013). The central hypothesis of this thesis is that accurate debris avalanche predictions require modelling of the entire multi-stage process, since assumptions made in one stage can strongly influence subsequent stages. To address this, the thesis develops an integrated approach combining slope-stability analysis with depth-averaged flow modelling, aiming to improve the quantification of debris avalanche hazard, demonstrated and validated using the well-documented 2012 Te Maari debris avalanche in New Zealand. Finite-element modelling of the pre-failure conditions at Te Maari indicates that hydrothermal alteration and elevated pore pressures are the primary drivers of the collapse. The interaction of these conditions also reveals the potential for progressive failure, involving two sliding blocks. Overall, the results show that failure predictions require detailed evaluation of pre-failure conditions, including well-constrained mechanical properties and groundwater distribution. Post-failure flow dynamics are examined using the depth-averaged, two-phase model D-Claw to assess the influence of initial parameters on flow mobility. Assessing the post-failure stage is a crucial step in landslide risk analysis, as it helps delineate areas potentially impacted by the event. Simulations reveal that the internal permeability regime strongly controls runout through its effect on excess pore pressure build-up. The results also highlight the influence of topographic confinement on energy dissipation and flow behaviour, as well as the occurrence of equifinality, where different parameter combinations can produce similar inundation extents. For Te Maari, both low- and high-permeability combinations reproduced the observed runout, indicating that flow behaviour is not dominated by a single parameter but emerges from the complex interaction of multiple factors. Building on these findings, a coupled, physically-based approach integrating LEM slope-stability analyses with depth-averaged flow modelling is developed and validated against observations from Te Maari to quantify spatial probabilities of flow inundation. Results show that the framework can effectively reproduce the complex, channelised Te Maari flow, with several simulations achieving high performance scores. A Factor of Safety (FOS)-based weighting function converts deterministic stability outputs into relative failure probabilities, accounting for uncertainty in the most likely slip surface. Exceedance maps derived from this approach better delineate observed hazard zones, demonstrating improved hazard prediction. The transferability of this approach is also demonstrated using the Ruapehu case study, illustrating its applicability to a distinct volcanic environment and over a larger spatial extent. By encompassing a broader range of potential failures and reducing dependence on past events, this application represents an important advance towards probabilistic volcanic debris avalanche hazard assessment. Together, these results demonstrate that integrating pre-failure assessment, failure mechanics, and post-failure flow modelling provides a robust framework for understanding and predicting debris avalanche behaviour. The approach reproduces observed events with high fidelity and captures key uncertainties and interactions that govern avalanche dynamics. While the framework represents a major advance towards probabilistic volcanic landslide hazard assessment, several avenues remain for future research. These include exploring more parameter variability, implementing more sophisticated FOS-based weighting schemes, reducing computational demands for large-scale or multi-scenario assessments, and incorporating additional impact metrics such as flow velocity or dynamic pressure. Pursuing these directions will further enhance the predictive capability of the framework and its applicability for risk mitigation and planning in diverse volcanic contexts.