Characterisation of explosive volcanic eruptive processes from deposit interpretation is key to long-term volcanic hazard and risk assessment, particularly for large explosive eruptions which occur relatively infrequently and whose deposits are often transient in the geological record. Of particular importance is estimation of eruption size, or magnitude, which is typically determined by measurement and interpretation of tephra fall deposits. The accuracy of individual tephra thickness measurements is analysed using the Fogo A deposit, Sao Miguel, Azores. Statistical analysis shows measurements are highly clustered spatially and the area represented by a single measurement ranges between 0.5 and 10 km2 . Tephra thickness measurement error (9 %) is much less than error associated with natural variance (10 - 40 %). The volume uncertainty related to thickness uncertainty for the Fogo A deposit is estimated to be 1.3 %, equivalent to a volume of 0.02 km2 , but this value is highly dependent on the number of measurements. Uncertainties in tephra volumes were further investigated by application of a statistical model to thickness data using cubic B-splines under tension. Volume uncertainties are correlated with number and spatial distribution of data points and decrease from as much as 40 % for 30 measurements to 10 % for 120 measurements. Large explosive eruptions are commonly multiphase, with each phase resulting in different hazards. Generally, proximal deposits are used to interpret eruption dynamics and hazard. Analysis of tephra from the Magnitude 8 Campanian 19nimbrite (Cl) eruption from distal deep-sea sediments show that thickness trends are better preserved in the deep sea than in sub-aerial environments. Grain size trends are consistent regardless of depositional environment. Plinian ash is identified to distances of 1500 km and within 150 - 900 km constitutes 40 ± 5 % of the deposit volume. The grain size information is used to provide estimates of the total grain size distribution of the Campanian Plinian and co-ignimbrite phases. The Plinian grain size distribution is highly dependent on the samples used, while the co-ignimbrite distribution does not change. The co-ignimbrite TGSD is much finer than the Plinian TGSD or the TGSD determined using all of the deposits. Results for both eruption phases vary little with method used. Results have implications for tephra dispersal modeling and highlight a need for the eruption phases to modeled separately. Because grain size trends are similar for all co-ignimbrite deposits, for real time modelling purposes, a total grain size distribution such as that presented here may be used, with intensity, magnitude and local wind speeds likely to be the more variable factors Ash transport and deposition is studied in a fluid mechanical model of umbrella clouds based on gravity current dynamics controlled by buoyancy forces and interaction with the wind. Deposit thinning trends in medial-to-distallocations are presented and results related to conditions near the source. Resulting deposition trends are complex and have implications for isopach production methods.

This thesis presents the results of laboratory experiments, theoretical analysis, field work and petrological analysis conducted to study the rheology and dynamics of lava flows. The rheology of suspensions of cubic crystals in viscous liquids was investigated with a series of experiments, consisting of the release of a fixed volume of fluid inside a horizontal channel. A Herschel-Bulkley rheology was assumed and the consistency K and the shear rate exponent n of this constitutive equation were calculated using the evolution of the flow front; the yield strength was calculated using the final shape of the flow. Results show an increase in K when crystal content increases. The mixtures start to show a shear thinning behaviour at 4>-0.3 with n values going from approximately 1 (Newtonian behaviour) to 0.5 at 4>=0.6. Yield strength was detected at the same 4> as the beginning of shear thinning behaviour and increases with a power-law relationship with crystal content. Suspensions with bimodal crystal sizes show a dramatic decrease of the apparent viscosity compared to unimodal suspensions, especially at the higher total crystal concentrations. Further experiments, with the same types of fluids (with approximately Herschel- Bulkley rheology) in a sloping channel with a constant flux rate were approximated with a simplified 2-D model. The results show an excellent agreement between the theory and the experiments. Additionally, experiments with two fluids, where a low viscosity fluid (syrup) is injected in a fixed volume of more viscous material (syrup plus crystals) suggest that the front advance and height is controlled mainly by the rheology of the fixed volume fluid. The simplified 2-D theory was applied to real lava flows. Three dynamical regimes were considered: A Newtonian viscous regime, a yield strength-dominated regime, and a crust-dominated regime. The resulting equations were applied to published data from eruptions of 10 lava flows with a range of compositions and conditions. Comparisons of the fits of the models to the data reveal that short-lived, high effusion rate eruptions typical of Hawaiian lava flows, are dominated by the internal viscosity of the lava, whereas low effusion rate or long lived eruptions are dominated by the yield strength in the growing crust. Finally, eruptions with very high initial crystal contents are dominated by the internal yield strength. The developed rheological model and 2-D theory were applied to the 2002 lava flow deposits at the NE flank of Etna volcano to reconstruct the variations in flow rate and advance of the flow front. Analyses of samples were used to estimate crystal content and liquid viscosity variations with distance. Yield strength variations were estimated from levee width measurements. The modelled flow rate and advance of the flow front compare well with the measured data, with a mean flow rate of 19.5 m3/s. The method developed in this thesis provides the prospect of using measurements of flow dimensions together with analysis of samples from the deposits to estimate the rheology of the flow and reconstruct variations of eruption parameters.

A key goal in volcanology is to develop long-term hazard and risk models for future volcanic events. For low frequency, large magnitude Plinian eruptions these assessments are based on the geological record of past eruptions. However, many deposits are poorly preserved and reveal very little information about the deposit. This thesis aims to develop methods for extracting additional information from these poorly preserved deposits, focussing on the estimation of erupted volume and eruption column height, and constrain uncertainties in the values found. A key development in this thesis is the quantification of a standardised field technique for characterising maximum clast-measurements, which are used to estimate plume heights. Bayesian inversion of the different techniques shows that the current recommendation for standardising maximum-clast measurements, the 50th percentile of 20 clasts, produces accurate results for reconstructing plume height. Plume height is a key parameter used to model future eruption scenarios, therefore the uncertainties associated with this parameter need to be quantified. Here a method has been developed that quantifies this uncertainty, dependent on the preservation and exposure of the deposit. An alternative statistical method to objectively determine the volume of a fall deposit without the production of isopach maps is presented. Integration of a log-linear regression model for thickness measurements with distance from the vent is applied to field measurements without any prior interpretation. Data and model uncertainty is accounted for using Bayesian methods. The eruption volumes calculated correspond well, and are within 10 %, to those previously determined by alternative methods. The number and distribution of thickness and maximum clast measurements from fall deposits can limit the amount of information gained from the deposit. Application of a fixed point variogram to field measurements identified the dependency of the deposit structure, with respect to the vent location, on spatial distribution of measurements. Results suggest that measurements need to be evenly distributed across a deposit, not concentrated along a dispersal axis. The methods developed in this thesis provide objective ways to calculate parameters from fall deposits and constrain the uncertainty in these parameters when limited data is available. This is important for modelling the long-term hazard and risk for future volcanic events.

The rift volcano Nyamuragira in D.R. Congo regularly produces large- volume lava flows resulting from flank fissure eruptions. The eruptions are usually fed by near-vertical dykes whose orientation and location are controlled by local and regional stress. Fissuring occasionally occurs within the summit caldera, but dyke emplacement and fissuring is usually confined to the flanks along preferential zones of weakness which radiate away from the 2 x 2.2 km caldera. Space-borne Synthetic Aperture Radar Interferometry (InSAR) was used to measure eo-eruptive and inter-eruptive surface deformation at the volcano between 1996 and 2010 (8 eruptions). The largest line-of-sight (LOS) displacement due to dyke emplacement (42 cm) was recorded during the 2002 eruption. Previously unreported displacements have been measured for the 1998, 2001 and 2004 eruptions. Numerical modelling of the 2006 and 2010 eruptive events was tried. The . orientation and size of the dykes is, however, poorly constrained, and the nature of subsurface connectivity with the caldera is not known. Both dykes were emplaced on the southern flank and are aligned with a NNW-trending fracture zone running between Nyamuragira and nearby Nyiragongo. Two methods using regression analyses on time-series data were devised to model and remove lava flow subsidence signals from interferograms. Subsidence signals> 3cm/year have been measured and are a function of time and lava thickness. Linear rate subsidence models were found to be appropriate for most lava flows. Detailed mapping of the recent lava flows of Nyamuragira has also better constrained their location and spatial extent. By stacking interferograms we obtained mean deformation maps of the volcano revealing inter-eruptive deformation: 1. Uplift within the Eastern Pit Crater and inflation of the summit prior to the 2010 eruption; 2. Post- eruptive deflation centred on the 2010 eruption site; 3. Long-period subsidence beneath the Western Crater and rifring of the caldera and 1 immediate flanks; 4. Long-period subsidence centred on the 2006 eruptive vent thought to be associated with visco-elastic relaxation of a cooling magma body; 5. Similar subsidence fields centred on the 1998 and 2002 eruptive vents; 6. Anomalous subsidence associated with the 1991-93 lava flow; 7. The existence of an apparently stable, fault-bounded, and dyke- resistant block of Precambrian crust beneath the NW flank of the volcano, probably related to the Western Border Fault.

Historically, volcanic research has been focussed primarily upon the geological hazard of eruptions while the exposure of the population at risk is often understudied. However, contemporary disaster risk management requires the analysis of vulnerability and hazard exposure, which is imperative at Volcan de Colima (VdC), Mexico, due to the predicted, large-magnitude eruption forecasted to occur before 2025. The methods used to gauge social vulnerability included the development and application of proxies to census records, the undertaking of a building vulnerability survey and the spatial mapping of civil and emergency infrastructure. Hazard exposure was assessed using primary modelling of laharic events and the digitisation of secondary data sources detailing the modelled extent of pyroclastic flows and tephra deposition associated with a large-magnitude (VEl 5) eruption at VdC. The undertaking and analysis of a risk perception. survey of the population enabled an understanding of the cognitive behaviour of residents towards the volcanic risk. In comparison to the published hazard map, the GIS analysis highlighted an underestimation of lahar hazard on the western flank of VdC and the regional tephra hazard. Vulnerability analysis identified three communities where social deprivation is relatively high, and those with significant elderly and transient populations near the volcano. Furthermore, recognition of the possibility of an eruption in the near future was found to be low across the study region. These results also contributed to the analysis of emergency management procedures and the preparedness of the regional authorities. This multidisciplinary research programme demonstrates the success of applying a GIS platform to varied integrative spatial and temporal analysis. Furthermore, ascertaining the impact of future activity at VdC upon its surrounding populations permits the evaluation of emergency preparedness and disaster risk reduction strategies.

As magma ascends towards the surface and the surrounding pressure from the rock drops, diluted gases (mainly C02, H20 and S02) are released and reach the surface prior to the magma itself. Accurate measurements of volcanic gases are thus essential in order to acquire an overview of the magmatic processes of a volcano in collaboration with geophysical data with the end goal of forecasting eruptions and managing volcanic crises. The standard approach for volcanic S02 flux measurements has been to traverse, by car, beneath the plume, with a UV spectrometer measuring overhead S02 concentrations. Recently, a number of compact and inexpensive commercially available UV spectrometers have become standard in the application. The first part of this thesis details the first intercomparison between these different devices, defining which attributes and models are optimal. Error sources e.g., from thermal effects and stray light, and proposed mitigative actions, are described. The second part of the thesis details the development of a freely downloadable stand-alone application (UVolc), which is compatible with the best identified spectrometers. This user friendly code, with meticulously written instructions, enables volcanologists, the vast majority of whom have little expertise in optics, to perform their own S02 flux measurements. Given that the associated hardware is inexpensive (;:::: €5k), this is particularly significant in the many developing countries where monitoring budgets are low, but risks high. This software incorporates routines designed to characterise and correct the errors identified during the intercoparison of the spectrometers, for improved accuracy. Finally, a novel upgrade to these spectrometers, based on UV camera technology, is described. These units image the plume, providing gas concentration profile maps, and so fluxes once per second, cf. 2-3 data per week, from the manual traverses. A rigorous study into optimum operational protocols, including validation, via intercomparison with conventional UV spectrometer observations, has been conducted on Mt. Etna and Stromboli in Italy. This approach is set to revolutionise volcanology, reflecting a step change in our capacity to understand volcanic processes, particularly transient explosions, and the first possibility of non-aliased corroboration with the other sources of volcanic data, from geophysics, which are acquired at 1 Hz. The latter has long been an aim of observational volcanology, allowing the prospect of far more holistic insights than available from the individual series in isolation.

Finite element method (FEM) models using the commercially available software package FASTFLO and traditional low-level computational programming methods have been used to consider the Peleean lava dome extruded on Soufriere Hills Volcano, Montserrat. Phenomenological nonNewtonian models are used to describe the complex rheology of the lava and time-dependent processes. Four research areas are studied: endogenous dome growth, the transition from endogenous to exogenous dome growth, conduit flow dynamics and the stability of structures containing lava domes. An elasto-viscoplastic FEM model is used to analyse the rheological gradients and the growth and evolution of an endogenous lava dome. The transition from endogenous to exogenous dome growth is an important process in Peleean lava domes. It is found that the development of shear planes within the conduit and lava dome ultimately govern this process and that the lava must be non-Newtonian for shear planes to form. Temporal non-linearities observed in the extrusion rate for Soufriere Hills Volcano are analysed in conduit flow models. Accelerating extrusion rates can be explained by a change in magma chamber over-pressure and a 'rheological memory' associated with the magma. Non-Newtonian magma may form unstable shear planes due to gas over-pressure in the conduit and it is thought that this process is responsible for hybrid seismicity and oscillatory flow. Model results show a first order fit to the cyclicity behaviour observed. The graZritational instability of lava domes may be influenced by viscous relaxation. Instability models show a possible time-delayed collapse mechanism as well as enhanced front lobe toppling. Numerical models are only as useful as the data used to constrain them and due to a paucity of data these models should be regarded as qualitative rather than quantitative.

Analysis of 240 samples for a wide suite of major and trace elements enable geochemical maps of the extrusive series of the Troodos Massif using ArcView 3.2 (a Geographic Information Systems application). Proxies for mantle source depletion, degree of partial melting, lithospheric component, subduction-derived input and fractional crystallisation form the basis of the maps. No substantial difference in mantle source depletion or subduction-derived component was detected between the three stratigraphically-defined lava units (Basal Group, Lower Pillow Lavas and Upper Pillow Lavas). Variation between the lava units is a function of the degree of partial melting of the mantle source and the degree of fractional crystallisation of the resulting magma, both of which are generally highest in the Basal Group and lowest in the Upper Pillow Lavas. Proximity to the paleo-transform fault has three effects on lava composition prior to component addition. Lavas within 10 km of the fault become increasingly primitive. The mantle source becomes strongly depleted on approximately the same scale. The lithospheric (OIB-like) component is also greatest close to the fault. The subduction-derived component is greatest in the north west of the Troodos Massif, in crust generated at the Solea Graben, the oldest of the grabens within the Troodos Massif. The younger Mitsero and Larnaca Grabens have progressively lower subduction-derived components. Taking into account rotation of the Troodos Massif since its formation, the areas of strongest subduction enrichment originally lay to the north. Therefore, the paleo-subduction zone where the Paleotethys ocean floor was being subducted beneath the Troodos supra-subduction spreading centre probably lay to the north of Cyprus, with a southerly dip.

History has shown that the combination of active volcanism and human populations can lead to devastating consequences, which at their most extreme have resulted in the collapse of entire societies. However these losses have not only been attributed to the impact of volcanic hazards but also the management strategies put in place to try to mitigate them, criticised for the detrimental long-term socio-economic impacts they have had on communities involved. General risk management theory argues that the failure of risk management strategies can be in part due to the poor risk assessment methodologies used to inform decision-making. ‘Insufficient’ or ‘inaccurate’ data is often attributed to disciplinary biases, a weakness in methodological tools and a focus on top down prescriptive approaches lacking in participation from those living with the risks in question. This thesis, acknowledging these broader debates, examines whether the way in which volcanic risk is assessed is fully representative of the complexities of the relationship between society and volcanoes. Using an empirical study of communities living on the Galeras volcano in Colombia, it compares the public experience of risk with the way in which it is interpreted and measured within Volcanology. Results of the study show that whilst previous volcanic risk assessments have been strong in their ability to capture data on volcanic hazards, assessment methods have been significantly weaker in their ability to address the threat of other non-volcanic hazards, social vulnerability and the social, economic and cultural value of the volcano environment. This thesis argues that a more sustainable approach to volcanic risk management is dependent upon risk assessments methodologies being developed that combine both the analytical frameworks of Volcanology with the experiential influences that drive the attitudes and behavior of the communities in question.

Active volcanoes pose a significant natural hazard. In order to evaluate the hazard it is important to reconstruct the history of such volcanoes to understand the frequency, style of eruption and the areas typically affected by the explosive eruptions. This thesis focuses on deciphering the volcanic and magmatic record for one of the most productive volcanoes in southern Chile, Volcán Mocho-Choshuenco. Work presented in the thesis establishes a detailed record of the explosive activity during the last 18 kyrs, constructed using field observations and geochemical analyses of the eruption deposits. Using a multi-technique approach Mocho-Choshuenco is shown to be one of the most explosive, frequently active and hence hazardous volcanoes in Chile. This high-resolution eruptive record provides new constraints on the underlying causes of spatial and temporal variability in arc volcanism. Temporally, the record gives new understanding into the response of arc volcanoes to deglaciation; clear temporal variation in eruption flux, eruption size and magma composition are observed. This time-varying behaviour is hypothesised to reflect changes in the crustal plumbing system, and magma storage timescales in response to removal of an ice-load. It demonstrates that deglaciation can drive changes in eruption behaviour at arc volcanoes; however the response is more complex and subtle than settings where decompression melting dominates. Spatially, Mocho-Choshuenco has a high number and density of scoria cones that have erupted relatively primitive magmas but nonetheless with a wide range of magma compositions. For some of the 'classical' slab and mantle geochemical tracers the erupted magmas span the complete range seen in this part of the arc. The tight temporal and spatial constraints provided by the analysed samples, coupled with recent advances in numerical modelling of magma transport through subduction zones, enable new hypotheses for interpreting the signatures of mafic arc magmas to be defined.