No description available.

No description available.

No description available.

No description available.

Open-vent activity at volcanoes of low-silica composition, such as Stromboli (Italy), Villarrica (Chile), Mt. Erebus (Antarctica), Masaya (Nicaragua), is characterised by persistent passive gas emission and recurrent mild explosive outgassing. Four styles of bubble bursting activity have been recognised in such volcanoes: seething magma, small short-lived lava fountains, strombolian explosions and gas puffing. Whilst strombolian explosions are perhaps the most common among these volcanoes, seething magma and small lava fountains have been observed only at the surface of active lava lakes. At Villarrica, one of the two case study volcanoes of this thesis, seething magma consists of continual bursts of bubbles up to a few meters in diameter, with varying strength over the entire surface of the lava lake. Small lava fountains, seen as a vigorous extension of seething magma, commonly lasts 20-120 s and reach 10-40 m high above the lava free-surface. Strombolian explosions exhibit a wide range of behaviour. For instance, they can last for less than a second in a single bubble burst that erupts mainly bombs, as seen at the lava lake of Mt. Erebus and Villarrica volcanoes, or for more than 30 seconds accompanied by large amounts of ash, as seen at Stromboli and Mt. Etna volcanoes. At Stromboli, the second case study volcano, gas puffing consists of small but repetitive bubble bursts with a generally table eruption frequency in the range 0.2-1.2 s⁻¹. More vigorous explosive phenomena, such as hundreds-metres high lava fountains or very strong (paroxysmal) explosions, may occur during eruptions or episodes of elevated activity. Multi-parameter monitoring offers a fuller recognition and understanding of the processes governing the volcanic activity at this type of volcano. For instance, correlations between seismicity and visual observations at Villarrica volcano indicate that the seismic tremor is mostly caused by explosive outgassing.

Groundwater is often an essential natural water resource on active volcanic islands. It can also interact with the volcanic system to generate hazards and exploitable geothermal reservoirs. Unfortunately, in volcanic island arc settings an understanding of the fundamental hydrology is lacking. Using field observation and numerical models, this study explores the hydrology of the active Caribbean volcanic arc island of Montserrat and aims to improve the current understanding of hydrology on Montserrat and island arc volcanoes in general. Recharge models for Montserrat, presented here, suggest that 10%-20% of annual rainfall is recharged to the groundwater system. Predicted recharge is strongly seasonal and spatially distributed, associated with rainfall and land use variations. Permeability measurements of core samples reveal a great range of permeabilities, between 1 x 10-18 and 1 x 10-12 m2, for lavas and volcaniclastic material. A lack of surface water is suggestive of relatively high permeability across the island. However, > 20 springs, with yields up to 19 L/s at elevations between 200 and 400 m, on the extinct volcanic complex of Centre Hills indicate the presence low permeability material. With recharge on the order of 0.24 mlyr such low permeability units must be reasonably laterally continuous. A suite of 250 TOUGH2 3D flow simulations, together with observations from Montserrat, suggest that the high elevation springs are supported by a low permeability central core of intrusive volcanics and associated (hydro ) thermally altered material. The numerical models generally struggle to reproduce high total spring discharge, suggesting that additional processes contribute to the observed high spring yields. A contribution of deeper waters to the Centre Hills springs has the potential to explain the observed high yields. This hypothesis is supported by field observations; an increase in temperature (up to 29°C) correlates with elevated specific conductivity and a decline in chlorofluorocarbon (CFC) concentrations in the southern Centre Hills springs, indicative of the presence of a deeper, older groundwater component. A final suite of more generic numerical models, applying the insights gained elsewhere in this study, reveal that temporal variations in recharge can generate gravity changes > 100 μGal. These simulations demonstrate the importance of understanding hydrology when performing and interpreting geophysical surveys that monitor active volcanoes.

Researchers in volcanic infrasound have proposed links between the acoustic signals of industrial jets and those recorded during sustained plume-generating volcanic eruptions (Matoza et aI., 2009; Fee et aI., 201Ob). If valid, this link offers a scaling law between the peak frequency of the volcanic signal and the exit velocity, a key parameter in determining the behaviour of eruption columns and predicting the dispersal of volcanic products. This PhD explores the validity and practicality of applying the engineering aeroacoustic results to volcanology through laboratory experiments with a range of nozzle geometries and a field case study. PIV analysis of a series of laboratory jet experiments demonstrated significant differences between the near to intermediate fields (NIP) of jets exiting from convergent nozzles, such as those commonly reported in the engineering literature, and those exiting from straight and divergent nozzles. As the NIP is the main region for sound generation, changes in its flow have implications for jet noise generation. Spectral analysis and adaptive beamforming of the acoustic signals of these jet flows showed them to be dominated by internal noise sources. Though the identification of differing noise sources was achieved, retrieval of the jet operating conditions via the standard empirical spectra was not possible. These results have important implications for the interpretation of infrasonic signals from sustained volcanic eruptions. The complex non-convergent geometry and large diameter of volcanic vents mean that high levels of internal volcanic noise are expected in infrasound from sustained explosive volcanic eruptions; As a substantial component of infrasound generated by a sustained eruption might not be from the plume itself, it is important to separate generation regions of different sources before inferring source parameters from infrasound data. Such separation requires a good understanding of the propagation path. Results from a multi-array field study at Sakurajima volcano, Japan, demonstrated the influence of local topography on the recorded infrasound signals. These influences require robust modelling before the vertical location of volcanic infrasound signals can be identified. This thesis serves as a reminder of the infancy of the infrasonic monitoring in terms of sustained events and identifies key goals for future laboratory experiments.

This thesis investigates the strengths, and potential pitfalls, of grouping observations from multiple volcanoes to understand magmatic and volcanic activity, and ultimately, to forecast eruptive activity and volcanic hazards. Crucial to the analysis and discussion is the principle of exchangeability, where volcanoes are analysed on an "all other things being equal" basis. Specifically, this research concerns issues regarding how volcanoes should be grouped together and what techniques can be used to analyse hierarchically structured data. Firstly, the thesis considers this from a statistical perspective, by introducing the methodology of hierarchical Bayes. The results suggest that simply aggregating multiple eruptive records may inaccurately capture the frequency of eruption magnitudes and volcanic phenomena. Secondly, analysis of historical eruptive activily at lava-dome building volcanoes indicate thal volcanoes of the same broad classification can be subdivided based upon empirical evidence, including the pattern of eruptive activity, monitoring observations and petrological data. However, based on further analysis, experts struggle to interpret these data independently in long-term forecasts: an expert elicitation technique is used to forecast repose duration at Soufriere Hills Volcano, with experts unable to identify behaviour that is atypical for a particular volcano. Survival analysis is investigated as a technique to analyse multiple eruptive records. Results indicate that whilst volcanoes may exhibit similar patterns of eruptive activity, they cannot necessarily be fitted using the same statistical model, suggesting it is not correct to simply group datasets that appear similar. Finally, the 0ppOltunity to use exchangeability to understand magmatic processes is discussed, by exploring differences in the pattern of behaviour at volcanoes of the same broad class. This allows the behaviour of lava-dome building volcanoes to be examined using well -established and novel paradigms for magmatic systems.

Fundamental questions persist regarding volcanic conduit erosion during sustained Plinian events, despite the clear control of conduit geometry on eruptive behaviour. This study has addressed this gap through a combination of detailed field work on the Avellino Plinian eruption at Vesuvius with a focus on lithic components, and modelling of stresses in wall rocks during steady eruptions. Together, the two approaches enable the interpretation of the processes and sources of lithic generation. The Avellino deposits provide a natural laboratory with a well-characterized subsurface from which lithic source depths are inferred. The lithic abundance, componentry and grain size distribution vary significantly with distance from the vent and between Plinian phases with different eruption rates. Conduit implosion is inferred to be the key mechanism involved in conduit erosion during the Avellino eruption, with secondary roles played by vent spalling/collapse and viscous shear below magmatic fragmentation producing xenoliths. The variations in component Total Grain Size Distributions (TGSDs) are interpreted to be controlled by both pre-existing weaknesses and fragmentation mechanism. Based on the lithic volumes and subsurface stratigraphy the upper, lava-hosted conduit radius increased by 167%, and the lower, carbonate-hosted conduit increased sharply at the fragmentation depth. The pressures predicted by conduit flow models mean that the wall-rock surrounding the conduit is fractured during the eruption, which would significantly reduce its strength and increase its permeability. The amount of conduit implosion predicted from analytical and Finite Element models can be many times more voluminous than the initial conduit and increases with deeper magmatic fragmentation, which increases conduit underpressure, and with conduit size. The modelled volume of failed conduit wall is greater for elliptical than cylindrical geometries, and is reduced considerably for 'hybrid' models, which transition from elliptical at depth to a cylindrical geometry. The combined, field and modelling work constrains how the conduit evolved during the Avellino eruption, and has broad implications for our understanding of eruption dynamics.

A comprehensive understanding of the origin, nature, and significance of volcanic unrest is currently missing, but of fundamental imp0l1ance to communities living with the threat of volcanic hazards. My thesis addresses this shortcoming from a volcanic deformation angle. By incorporating a range of multi-disciplinary data, I have developed new integrated models of volcanic deformation using Finite Element Analysis (FEA) that are suitable for use in both forward and inverse modelling approaches. They are consistent with independent geophysical observables and provide detailed insight on volcanic processes during unrest crises. I have applied these models to three volcanoes. A study of Uturuncu volcano, Bolivia, highlighted the importance of subsurface structure and time-dependent source processes in explaining both the spatial and temporal deformation patterns. The combined results alluded to a diapiric-type ascent of magma. At Cotopaxi volcano, Ecuador, I used novel inversion models employing FEA to elucidate the location and volume of a magmatic intrusion during an aseismic, and non-eruptive, unrest episode. The models also provided insights into observable signals that could be associated with future intrusive or eruptive activity. My analysis of the persistent inflation at Aira caldera, Japan, during an ongoing emptive phase at Sakurajima volcano, used inverse Finite Element models to, for the first time, quantify the statistical significance of including topography and subsurface heterogeneity in deformation models. Additional models results were used to identify the rate, timing, volume, location and mechanism of magma supply, as well as the timescales that could be associated with increases in future eruptive activity. Together, these results highlight how models with more plausible, and geophysically consistent, components can improve our understanding of the mechanical processes affecting volcanic unrest and geodetic eruption precursors. They provide a framework to help advance emption forecasting and risk mitigation.