Search

Southeast Asia is arguably the most tectonically active region on the planet, fuelled to a large extent by nearly 10,000 km of ongoing subduction on its western, southern and eastern flanks that accommodates the northward motion of the Indo-Australian plate and westward motion of the Philippine Sea plate. It has hosted one of the largest earthquakes ever recorded (Mw 9.2 Sumatra-Andaman earthquake in 2004) and perhaps the most famous volcanic eruption in history (Krakatoa eruption of 1883), which profoundly affected the Earth's climate. While the western Pacific margin and Indonesian archipelago along the Sunda and Banda arcs have been well studied, the same is not true of the interior region of Southeast Asia, which includes Borneo and Sulawesi. Borneo is the 3rd largest island in the world and lies on the eastern margin of Sundaland, the continental core of Southeast Asia, but its intraplate setting means that it has no active volcanoes and little in the way of seismicity. By contrast, the adjacent island of Sulawesi features active subduction and a network of continental transform faults that give rise to high levels of earthquake activity. How this central region of southeast Asia was formed, and the tectonic relationship between Borneo and Sulawesi, is still poorly understood. The goal of this thesis is to exploit passive seismic data recorded by temporary and permanent seismic networks to image the first order crustal structure of both Borneo and Sulawesi. This will provide fresh insight into their deep structure, provenance and tectonic evolution, and how they have been impacted by recent events (e.g. opening of the South China Sea) that have clearly left their mark on surrounding regions. The passive seismic data used in this dissertation is in the form of teleseismic body wave arrivals, which interact with upper mantle and crustal structure before being recorded by stations on the surface. I utilise both receiver function analysis (RFA) and virtual deep seismic sounding (VDSS) to extract information from mode converted and reflected phases to constrain seismic properties including Moho depth, V_S and V_P 〖/V〗_S. In the case of RFA, H-k stacking, migration and inversion are separately applied, with the ensemble of results producing robust estimates of crustal thickness in particular. The primary strength of VDSS is its ability to constrain Moho depth using relative few sources, even in the presence of significant crustal complexity (e.g. thick sedimentary sequences); I therefore find this to be a particularly useful technique in some regions, such as northern Borneo, where RFA produces equivocal results due to strongly heterogeneous crustal structure. One of the main outcomes of this thesis is a new detailed map of Moho depth variations beneath northern Borneo from the application of VDSS to data from an array of 65 temporary and permanent broadband seismometers. This is the most recently active part of Borneo, having experienced at least two phases of subduction in the Miocene. Moho depths vary between ~45 km beneath the Crocker Range to less than 25 km beneath the central interior, consistent with the presence of a thick crustal root transitioning to an area of relatively thin crust that extends NE into the Sulu Sea. These results support a model of subduction polarity reversal (SPR) in Borneo, in which opening of the South China Sea led to SE-directed subduction of the proto-South China Sea beneath Northern Borneo, followed by continent-continent collision, which formed the Crocker Range, supported by a thicker crust. This was followed by subduction termination and the initiation of Celebes Sea subduction beneath northern Borneo of opposite polarity. Roll-back of the Celebes Sea slab resulted in the opening of the Sulu Sea, which extended into the interior of northern Borneo and caused localised crustal thinning, but did not proceed to rifting. In addition to the detailed study of northern Borneo, I also investigated Borneo in its entirety, albeit at a much lower resolution owing to the more limited data coverage at this scale. The primary tool in this case was RFA, which was used on data from 28 broadband stations distributed throughout Sarawak and Kalimantan, as well as a subset of stations from northern Borneo. VDSS was also applied to data recorded at a number of sites where RFA was not able to produce a well constrained Moho. The results show that on a broad scale, northern Borneo features the thickest crust, with the vast majority of Borneo having crust that is less than 35 km thick, despite the interior mountain range having an average elevation in excess of 1000 m. Some of the thinnest crust (~25 km) occurs beneath a Mesozoic accretionary complex in western Borneo, which is thought to mark the location of a previous subduction zone and crustal extension related to slab rollback. In Southwest Borneo, the Schwaner mountains are underlain by elevated V_P 〖/V〗_S, which are juxtaposed against lower Vp/Vs of the Kuching Zone. Thicker crust (~40 km) in southeast Borneo approximately overlays the Meratus Suture, and may be related to the docking of the East Java-West Sulawesi and Southwest Borneo blocks, which have East Gondwana provenance. The investigation of the crustal structure of Sulawesi made use of data from 23 permanent network stations and 18 temporary stations, with receiver function analysis applied to constrain the bulk crustal properties of Sulawesi. Crustal thickness ranges between about 20 km and 40 km, with the thinnest and thickest crust juxtaposed across the Palu-Koro fault, a > 500 km long sinistral strike slip fault in central Sulawesi. This fault - which produced an Mw 7.5 earthquake in 2015 that also unleashed a destructive tsunami - accommodates some 4 cm/yr of left-lateral motion along with clockwise motion of the North Arm of Sulawesi. Thinner crust along the western part of the North Arm of Sulawesi is likely related to roll-back along the North Sulawesi trench, while thicker crust to the east may be due to opposed subduction of the Celebes Sea and Molucca Sea straddling the land mass. The relatively simple crustal seismic models of Borneo and Sulawesi produced in this dissertation represent a solid first step towards understanding their tectonic assemblage and geological architecture. Future work might involve bringing in other datasets to aid with interpretation, such as gravity, magnetic, heat flow, geochronology, petrology etc., and using the new seismic models to help answer first order questions, such as the mismatch between predicted and observed dynamic topography in Borneo in particular. Incorporating data from a recent OBS deployment in the southern Celebes Sea and Makassar Strait may significantly improve our ability to connect the seismic structure of Borneo and Sulawesi, although it remains to be seen whether the data coverage and quality will be sufficient to substantially change the current picture.

The regions adjacent to tectonically active mountain belts are exposed to significant earthquake hazard, since the range-bounding faults produce large earthquakes, and the underlying geological structure amplifies the resulting ground shaking. The aim of this dissertation is to investigate the regional-scale controls on earthquake ground motions and seismic hazard in these settings. The first part of this dissertation describes models of the seismic wavefield produced by thrust-faulting earthquakes on mountain range fronts. The earthquake source characteristics and foreland basin structure were varied within reasonable geological bounds, and the earthquake-induced ground shaking was calculated. The earthquake source parameters were determined to be the dominant control on the amount of near-source ground shaking. However, the foreland basin structure, in particular the basin depth relative to the dominant wavelength of the seismic waves, determines the importance of dispersion as the waves propagate through the basin. These results highlight the importance of accurately determining earthquake source characteristics (particularly depth), and the underlying geological structure, during hazard assessment. These principles were then applied to study the active tectonics and seismic hazard in the north-west Himalayas. Field, satellite, and seismological observations were used to determine the fault geometry beneath the NW Himalayas and investigate the relationship between thrust faulting and folding. These results were used to construct seismic-wavefield models, to determine earthquake ground motion estimates if the Main Himalayan Thrust in the region were to rupture. These models show that peak ground velocities are extremely sensitive to minor variations in the fault geometry. Finally, the earthquake-induced building damage in foreland basins was investigated. Using seismic-wavefield modelling, alongside fragility curves for generic building types, the relationships between earthquake location, characteristics, and building damage were investigated. The results quantify the previously poorly known trade-off between earthquake location and magnitude in determining damage distributions. Additionally, the results quantify the factors that can cause over- or under-estimates of the magnitudes of historical earthquakes based on reported damage distributions, with important implications for understanding the accumulated slip deficit in continental collision zones.

Repeating earthquakes are a unique seismic phenomenon where the same fault patch regularly fails. They occur in various settings and are useful geophysical tools to monitor and image the subsurface. However, we do not understand the underlying physics that generates repeating earthquakes. I focus on two subsets of these repeaters: in Parkfield, California and in the Moon. I examine repeaters in Parkfield to test a model where the deficit between seismic slip and long-term slip on the fault is explained by partial ruptures. I find that partial ruptures can accommodate ~30% of the slip on the fault, but this is not enough to explain the slip deficit on the fault, and an alternative model may be required. I use deep moonquake observations to constrain the tectonic stress in the moon. Some deep moonquakes appear to slip backwards relative to other moonquakes. There are two models for these backwards slipping events: a large tectonic stress or a large tidal stress. To discriminate between these two models, I apply principal component analysis to the waveforms of deep moonquakes. I find the tidal stress to be the main driver of deep moonquakes and place an upper constraint of 0.1 MPa on any tectonic stress in the lunar interior. Finally, I align deep moonquakes to reduce the uncertainty of arrival times; the uncertainty can be up to ~10 seconds for P-arrivals and ~4 seconds for S-arrivals. This limits our ability to use this data to locate deep moonquakes. To reduce the uncertainty, I align deep moonquakes to a reference event using principal component analysis. The method aligns the P-arrivals of 20 - 40% of the real deep moonquakes to within 1.5 seconds of the reference event -- a significant improvement relative to the original absolute error.

The structure and composition of the Earth's interior provokes great debate within the Earth Science community. While the bulk composition and density of the Earth are largely known, many discrepancies exist between different probes of the Earth, and little is understood of the distribution of compositional heterogeneity within the Earth. Seismic discontinuities arise from jumps in velocity and density, with global discontinuities manifesting from mineral phase transitions and large-scale changes in composition like the core-mantle boundary. Widely distributed previous seismic observations of localised reflectors in the upper mantle have been termed the X-discontinuity (X; 230-350 km depth) and allude to the widespread presence of chemical heterogeneity, but the exact cause of the X is yet to be determined. Receiver function (RF) techniques using P-to-S converted waves are capitalised upon to study the X at higher spatial resolution than previously attainable. Stacking RFs in the depth and time-slowness domains provides identification of the X beneath 15 hotspots. The X appears more prevalent at hotspots compared to six reference locations. Amplitude information from recorded and synthetically modelled RFs constrains the causal mechanism of the X in mantle plume settings to the coesite-stishovite phase transition and/or the presence of carbonated silicate melt, shedding insight into the nature of mantle plumes and their role in mantle convection. I study the presence of the X beneath Africa, using continuous observations and a vote map style procedure to understand the length scales of the X, its short wavelength variation, and the presence of topography across the discontinuity. The X continues to be co-located with regions of mantle upwelling and observations appear to show a strong dependence on back azimuthal distribution, suggesting that the X may possess short wavelength topography. Finally, with the potential to add further constraints on the X and increase data coverage, I determine the feasibility of using S-to-P converted RFs for studying the X. While it is possible to record signals from the upper mantle using these RFs, the presence of interfering phases greatly limits the epicentral distance range available for study and requires careful windowing of the source for iterative deconvolution. I do not recommend this potential complimentary probe.

Seismic tomography has played a crucial role in the illumination of deep Earth structure. Most existing tomographic methods are based on seismic ray theory and hence do not fully account for the true physics of wave propagation. Recent computational advances allow us to embrace the full complexity of seismic wave propagation by accurately solving the 3-D seismic wave equation numerically. This can account for effects such as wavefront healing, interference, scattering and (de)focusing, which are often ignored or not properly captured by other methods such as ray tracing. Thus, such methodologies are particularly suitable for strongly heterogeneous regions such as Southeast Asia, where large variations in elastic parameters are likely to be present. Here, an unprecedented dataset and access to sizeable computational resources allow their application to Southeast Asia for the first time. In the first part of this thesis, a continental-scale seismic model of the lithosphere and underlying mantle beneath Southeast Asia obtained from adjoint waveform tomography (often referred to as full-waveform inversion or FWI) is presented. FWI is a non-linear imaging method, where an initial model is updated in order to minimise the difference between observed and predicted waveforms. Based on > 3,000 h of analysed waveform data gathered from  13,000 unique source-receiver pairs and filtered at periods between 20 - 150 s, isotropic P-wave velocity, radially anisotropic S-wave velocity and density are imaged via an iterative non-linear inversion that begins from a 1-D reference model. At each iteration, the full 3-D wavefield is determined through an anelastic Earth, accommodating effects of topography, bathymetry and ocean load. SASSY21, the final model after 87 iterations, appears to be robust since it is able to explain true-amplitude data from events and receivers not included in the inversion. The new model reveals detailed anomalies down to the mantle transition zone, including multiple subduction zones. The most prominent feature is the (Indo-)Australian plate descending beneath Indonesia, which is imaged as one continuous slab along the 180  curvature of the Banda Arc. The tomography confirms the existence of a hole in the slab beneath Mount Tambora and locates a high S-wave velocity zone beneath northern Borneo that may be associated with subduction termination in the mid-late Miocene. A previously undiscovered feature beneath the east coast of Borneo is also revealed, which may be a signature of postsubduction processes, delamination or underthrusting from the formation of Sulawesi. In the second part of this thesis, SASSY21 is used as a starting model to obtain a more refined image of the eastern Indonesian region, using seismic data filtered at periods from 15-150 s. In this study, the fluid ocean is accounted for explicitly by solving a coupled system of the acoustic and elastic wave equation. This is computationally more expensive but allows seismic waves within the water layer to be simulated, which becomes important at shorter periods. The effects arising from surface topography, bathymetry and the fluid ocean on synthetic waveforms become pronounced at periods ≤ 20 s. In particular, surface elevation can result in a considerable phase advance and change in amplitude of the surface wave train, and has an effect on both horizontal and the vertical seismogram components for this simulation setup. The fluid ocean results in a phase delay as well as a change in amplitudes and duration of the surface wave train, and affects both the radial and vertical components. At periods ≤ 20 s, accounting for the fluid ocean explicitly can lead to more realistic lithospheric velocities and a more refined image compared to the commonly used ocean load approximation, even at greater depths. Furthermore, it allows for an improved waveform match for source-receiver paths passing partially or entirely through oceanic regions. The final model, SASSIER22, after 34 iterations reveals a convergent double-subduction along the southern segment of the Philippine Trench, which was not evident in the starting model and transitions to a divergent system in the Molucca Sea further south. A more detailed illumination of the slab beneath the North Sulawesi Trench subduction zone reveals a pronounced positive wavespeed anomaly down to  200 km depth, consistent with the maximum depth of seismicity, and a more diffuse but aseismic positive wavespeed anomaly that continues to the 410 km discontinuity.

Machine learning (ML) has caused a fundamental shift in how we practice science, with many now placing learning from data at the focal point of their research. As the complexity of the scientific problems we want to study increases, and the amount of data generated by today's scientific experiments grows, ML is helping to automate, accelerate and enhance traditional workflows. Emerging at the forefront of this revolution is a field called scientific machine learning (SciML). The central goal of SciML is to more tightly combine existing scientific understanding with ML, generating powerful ML algorithms which are informed by our prior knowledge. A plethora of approaches exist for incorporating scientific principles into ML and expectations are rising for SciML to address some of the biggest challenges in science. However, the field is burgeoning and many questions are still arising. A major one is whether SciML approaches can scale to more complex, real-world problems. Much SciML research is at a proof-of-concept stage, where techniques are validated on simplified, toy problems. Yet, understanding how well they scale to more complex problems is essential for them to become widely applicable. This question is of central focus in this thesis. Firstly, multiple different physics-informed ML approaches are designed for three complex, real-world, domain-specific case studies taken from the fields of lunar science and geophysics, and their performance and scalability is assessed. Secondly, the scalability of physics-informed neural networks, a popular and general SciML approach, for solving differential equations with large domains and high frequency solutions is evaluated and improved. Common observations across these studies are discussed, and significant advantages and underlying limitations are identified, highlighting the importance of designing scalable SciML techniques.

Seismic tomography allows us to study the interior structure of the Earth on a range of lengthscales, with the method of waveform inversion being a promising and relatively recent development. In this work, we present new theoretical methods which may contribute to the study and solution of seismic tomography inverse problems. In particular, we set out the mathematical grounding for novel inversion frameworks which could be applied to waveform inversion. In Chapter 2, we introduce an inversion framework for determining the topography of internal boundaries jointly with volumetric parameters, using waveform inversion. The approach makes use of a referential description of elasticity to encode boundary topography within a volumetric parameter, putting topography on an equal footing with other parameters in the joint inversion. We derive expressions for sensitivity kernels in this parameterisation using the adjoint method, and calculate examples on a 2D domain. A second-order adjoint method is used to derive expressions for the application of the Hessian to a model perturbation. Chapter 3 sets out an approach for determining a suitable function space to use as a model space in geophysical inverse problems, including seismic tomography. In particular, we show that a Sobolev space is often a suitable choice, and allows us to specify a required degree of regularity for model parameters. We show that gradients in a Sobolev space can be calculated by a post-processing of a derivative, with the gradient being smoother than the derivative. The calculation of these gradients is demonstrated in different scenarios on spherical domains. The final chapter presents a simple application of the Sobolev gradient method, to linearised ray tomography in 2D. This allows us to investigate the potential impacts of the choice of function space on tomography problems more generally. A particular advantage of the method used is that the minimum-norm solution to the least squares inverse problem with exact data can be calculated.

The application of fibre optic Distributed Acoustic Sensors (DAS) is a rapidly developing field in seismology. DAS cables are cheap and non-intrusive and are capable of rapid deployment on the surface, or in permanent or semi-permanent deployment into the subsurface. As seismic arrays, DAS cables offer wide coverage at high spatial resolution allowing the simultaneous monitoring of entire wellbores. Consequently there has been growing interest in quantifying the performance of DAS arrays relative to conventional geophone arrays. This thesis aims to assess the viability of DAS as a tool for seismic monitoring and microseismic location. In contrast to conventional workflows using geophones, certain characteristics of DAS - including gauge length effects, increased spatial density and one-component measurements - cause additional considerations in how we process DAS data. In addition, continuous DAS monitoring produces huge data volumes in comparison with conventional geophone arrays. This increased data volume requires the development of automated processing methods in order to process seismic events in near real-time. This thesis investigates these differences in data characteristics in a near-surface refraction survey and in downhole arrays during microseismic monitoring. In our first study we are able to directly compare geophone and DAS seismic data through the acquisition of a near-surface seismic refraction dataset in North London in March, 2019. This allows exploration of the differences in first break arrivals in a controlled small-scale setting. In the subsequent chapters we investigate downhole monitoring of a microseismic dataset via data provided by Silixa Ltd., with the aim to identify different processing requirements and develop a best automated workflow that would allow the picking and locating of DAS data in close to real-time.

This thesis aims to present the data measurement process, waveform modelling, full waveform sensitivity measurements, and inversion workflow, required to produce a P-wave tomography of the upper mantle, including the mantle transition zone, down to 1000km depth. Improved seismic velocity models are needed: many tomography models do not include regional travel-time datasets. This is because the first arrivals identified in travel-time datasbases are not enough on their own to seismically sample the mantle transition zone. Sharp velocity gradients at the top (410km) and bottom (660km) of the mantle transition zone induce triplications in body waves. This means that there is more than one possible path between source and receiver for a given waveform: these can range from a topside reflection off the 410km discontinuity, through to refraction at the base of the 660km discontinuity. It is therefore difficult to attribute the first arrival time in a body waveform to its true path in an automated manner, as needed for tomography. This is heightened by discrepancies between the assumed reference background model for data processing and inversion, with respect to the true background model: therefore, seismic travel-times must be measured relative to synthetic data based on full waveform modelling. A limited number of regional body wave velocity models based on the full seismic waveform exist, which can provide sufficient seismic illumination of the mantle transition zone, a known barrier for subducting oceanic lithosphere. In this study I present JEABOX, a P-wave tomography of the western Pacific upper mantle, using the joint inversion of existing teleseismic P-wave travel times based on multifrequency cross-correlation, with regional P-wave measurements between 14-30 degrees epicentral distance. I calculate Born sensitivity kernels for these latter triplicated data using the MC Kernel software based on the spectral element solver, Axisem. I showcase the integration of these novel triplicated data and their sensitivity kernels into a tried-and-tested finite-frequency tomography workflow, to perform joint inversion of teleseismic data and Dahlen kernels with regional travel-times and MC Kernels, to produce a hybridised P-wave velocity model based on Axisem, the first tomography of its kind. The resulting model corroborates several major seismic structures observed across western Pacific subduction zones in previous studies, including what appears to be slab stagnation beneath the Kuril and Japan trenches, and steep subduction under the Izu-Bonin trench. In the southern hemisphere, seismic signatures imply protracted subduction beneath the Sunda-Java trench under Borneo and surrounding area. Future work will tectonically interpret the JEABOX model in full.

Microseismicity offers an opportunity to image subsurface deformation at exceptionally high spatial and temporal resolution. This may be related to a diverse range of processes, including fore- and aftershocks to destructive earthquakes on large faults, magma movement at volcanoes, or the gradual advance of glaciers. Microearthquakes at faults might also signal more exotic behaviour, including due to transient events such as fluid injections or pulses of aseismic fault creep. The study of small earthquakes has advanced significantly over the past decades, as seismology has entered the digital age. Denser networks, with larger numbers of more sensitive seismometers allow ever smaller seismic events to be detected and analysed. The study of larger numbers of earthquakes brings a raft of benefits, including more robust statistical analyses, higher temporal resolution, and the opportunity to achieve significantly greater spatial resolution by harnessing the power of relative relocation algorithms. However, the fundamental task has remained the challenge of extracting an earthquake catalogue from continuous waveform recordings. I present new software that offers a powerful, efficient, and highly automatable method to produce exceptionally complete and robust earthquake catalogues from continuous seismic data. Instead of reducing the information contained in waveforms to "picks" - discrete timestamps marking candidate phase arrival times - within QuakeMigrate the seismic data is transformed to continuous functions which are then combined across the network by migration. Retaining as much information as possible until the point of comparing the recordings from across the entire array allows phase arrivals at or below the signal-to-noise ratio to be successfully associated with events, improving both detection capability and location resolution. I use QuakeMigrate to detect and locate microseismicity in the Northern Volcanic Zone, Iceland, from 13 years of continuous seismic data recorded on a dense local network. Through comparison with a comprehensive catalogue of previously analysed events, I demonstrate that this new, automated algorithm is capable of location accuracy equal to or surpassing the widely held gold standard of manually refined earthquake locations. Within my new catalogue of more than 155,000 events, including related to magma ascent and intrusion in the lower crust, shallow geothermal activity, and seismicity triggered by fluid injection, I focus on persistent earthquake swarms that delineate a network of faults surrounding Askja volcano. Tightly constrained earthquake hypocentres, refined by cross correlation and relative relocation, reveal an intricate mesh of strike-slip faults, with fine-scaled structures resolved at length-scales of 10's of metres. This comprehensive image of the subsurface fault structure fills in the gaps between manually analysed events, and combined with tightly-constrained fault plane solutions reveals kinematics which point to a complex deformation field, and the role of elevated pore pressures or mechanical anisotropy in generating anomalous fault geometries. Comparison of the temporal evolution of the major sources of deformation with the seismicity provides evidence that deflation of a shallow magma reservoir in Askja caldera plays a principal role in driving these earthquakes. Within individual earthquake swarms, seismicity systematically migrates along fault planes, indicating the role of aseismic processes in controlling their evolution. Analysis of the migration velocities and b-value statistics points to transient creep events as the primary driving mechanism, indicating the role of unusual frictional properties. Aseismic slip has recently been recognised as an important process in releasing strain at subduction zone interfaces, as well as on faults at a range of depths both in the oceans and within the continents. Such well-resolved examples of this process are crucial in gaining a deeper understanding of the wide spectrum of fault behaviour, and the links between slow and largely silent, and fast and potentially devastating fault slip. The Bárðarbunga-Holuhraun dike intrusion provides a unique opportunity to analyse the triggering of seismic swarms due an extremely well-resolved static stress change. As the dike intruded towards Askja, a surge in swarm seismicity was observed across the network of faults studied in this dissertation. In the absence of complicating factors present in the study of stress changes caused by large earthquakes, I determine that swarms initiated at Coulomb stress increases of as small as 0.025 MPa. Furthermore, swarms do not show any seasonal tendency in occurrence, despite strong seasonal loading of the crust, which dominates measurements of relative crustal velocity changes. This further indicates that these faults experience very low effective normal stresses. Together, these observations indicate that this network of faults possess an exotic rheology. The absence of seismicity triggered elsewhere by the dike intrusion provides confidence in my observation of their spatial extent, which is consistent with earthquake catalogues derived from analogue seismic recordings extending back to the 1970s. I suggest the potential roles of fluids released at sites of vigorous hydrothermal activity in Askja volcano, and from numerous melt feeders in the lower crust, in producing the conditions that allowed these faults to develop, setting this area apart from the remainder of the Northern Volcanic Zone, which is otherwise largely aseismic during the current inter-rift period.

The plate tectonic cycle is fundamental to our dynamic Earth, encompassing the formation and evolution of new lithosphere at divergent, mid-ocean ridges, all the way to its eventual return to, and re-equilibration with, the mantle in subduction zones. I investigate the structure of the crust and upper mantle of the Earth in two regions that represent different endmembers in this cycle through the analysis of microseismicity and seismic anisotropy, seeking to learn more about the stresses in these environments and how they are manifest in the structure of the subsurface. In Iceland, new oceanic crust is accreted episodically within the volcanic rift zones that delineate the subaerial portion of the Mid-Atlantic Ridge, the divergent margin between the North American and the Eurasian plates. Northern Borneo, conversely, exhibits the tectonic signatures of not one but two terminated subduction zones, where oceanic lithosphere was once being actively recycled into the Earth's mantle. As part of this work, I have helped to deploy and service two passive seismic experiments-the Cambridge Volcano Seismology network in Iceland and the northern Borneo Orogeny Seismic Survey (nBOSS) network in Sabah-from which I have derived my results. Seismic anisotropy is manifest on a vast range of scales, from swathes of the crust and mantle, all the way down to the scale of single mineral crystals. It has the potential to inform on the dynamic state of the mantle, the structural fabric of fault zones, layering in sedimentary basins, and the distribution of partial melt in the subsurface, to name a few applications. Here, I seek new insights into the stress state and structure of nascent oceanic crust as it is accreted at a mid-ocean ridge, and the volcanic systems found therein, in Iceland. I also look to piece together the interplay between past tectonic events, subduction termination, and the present-day state of the mantle in northern Borneo. In the first part of my dissertation, I apply shear-wave splitting analysis to a microseismic catalogue (spanning the period 2008-2018) in the Northern Volcanic Zone of Iceland in order to investigate the relationship between seismic anisotropy and the tectonic stresses arising from extension and plate spreading, microseismicity, and the presence of melt. I find the upper 3-4 km of the Icelandic crust to be seismically anisotropic. Modelling of the stresses in the upper crust arising due to plate spreading and active deformation around Askja volcano, I find this anisotropic layer can be explained through a mechanism of stress-aligned microcracks within the porous crust. I then use earthquakes that occur within the lower, ductile crust-associated with the movement of melt within the volcanic plumbing system-and effective elastic media modelling to explore the role of melt in the generation of seismic anisotropy below 10 km depth, finding a strong correlation between my observations and the presence of melt as inferred from seismic tomography. I synthesise these observations with recent works examining the seismic anisotropic structure of the Icelandic crust from ambient noise observations to construct a conceptual model for the mechanisms generating anisotropy throughout the Icelandic crust. In the second part of my dissertation, I step back in terms of scale to investigate how tectonic processes related to subduction (and post-subduction) are manifest in the form of seismic anisotropy by applying shear-wave splitting analysis to teleseismic core-refracted seismic phases. My results, in conjunction with a recent tomographic model of the region, constrain the seismic anisotropy to the lithosphere and reflect recent (past ∼20 Myr) events in the tectonic history of northern Borneo. In north-west Sabah, my observations of seismic anisotropy are aligned with the orogenic belt running down the north-west coast (the Crocker Range), which formed during the termination of subduction of the proto-South China Sea. Conversely, in south-west Sabah my observations suggest extension has played an important role in the subsequent tectonic evolution of northern Borneo following subduction of the Celebes Sea and the opening of the Sulu Sea. I supplement this work with the first detailed study of seismic activity on a network of faults around Mount Kinabalu, a 4100 m tall mountain within the Crocker Range of northern Borneo. For this, I developed and used QuakeMigrate, a new Python package for the automatic detection and location of earthquakes using waveform migration and stacking. These new observations indicate ongoing extension in the region around Mount Kinabalu and allow, for the first time, for this network of faults to be mapped and explored in detail.

The lowermost mantle - D" - is a crucial thermochemical boundary layer within the Earth, exerting control over dynamic processes in both the overlying mantle and the core below. Constraining seismic anisotropy (the variation of seismic velocity with direction) within this layer could give us key insights into its dynamics and composition. In this thesis I collect a new broadband dataset of SKS-SKKS shear-wave splitting, where ca. 17% of global observations are discrepant representing a significant increase compared to previous analysis. Using this dataset I develop new techniques to study discrepant SKS-SKKS shear-wave splitting and to invert of shear-wave waveforms for the orientation and strength of seismic anisotropy. I apply both these new techniques to study seismic anisotropy in D" beneath the Eastern Pacific. Analysis of SKS-SKKS shear-wave splitting discrepancies highlight a significant region of azimuthal anisotropy in D" which is most plausibly explained by the lattice preferred orientation of post-perovskite. I then test these qualitative interpretations, which are typical of SKS-SKKS shear-wave splitting studies, by jointly inverting ScS, SKS and SKKS waveforms for seismic anisotropy in a fast shear-wave velocity anomaly beneath the Eastern Pacific. I evaluate four candidate mechanisms for seismic anisotropy in D": elliptical transverse isotropy (TI; representing for example anisotropy due to melt or layering), bridgmanite, and post-perovskite (for fabrics dominated by either [100](001) or [100](010) slip). Elliptical TI and both post-perovskite fabrics reasonably explain the input data, with predictions of horizontal flow for elliptical TI and post-perovskite [100](001) agreeing with recent mantle flow models. The techniques I have developed give seismologists powerful new tools to constrain lowermost mantle anisotropy.