Supervised deep learning models have become a popular choice for seismic phase arrival detection. However, they don’t always perform well on out-of-distribution data and require large training sets to aid generalization and prevent overfitting. This can present issues when using these models in new monitoring settings. In this work, we develop a deep learning model for automating phase arrival detection at Nabro volcano using a limited amount of training data (2498 event waveforms recorded over 35 days) through a process known as transfer learning. We use the feature extraction layers of an existing, extensively-trained seismic phase picking model to form the base of a new all-convolutional model, which we call U-GPD. We demonstrate that transfer learning reduces overfitting and model error relative to training the same model from scratch, particularly for small training sets (e.g., 500 waveforms). The new U-GPD model achieves greater classification accuracy and smaller arrival time residuals than off-the-shelf applications of two existing, extensively-trained baseline models for a test set of 800 event and noise waveforms from Nabro volcano. When applied to 14 months of continuous Nabro data, the new U-GPD model detects 31,387 events with at least four P-wave arrivals and one S-wave arrival, which is more than the original base model (26,808 events) and our existing manual catalogue (2,926 events), with smaller location errors. The new model is also more efficient when applied as a sliding window, processing 14 months of data from 7 stations in less than 4 hours on a single GPU.

Firn densification profiles are an important parameter for ice-sheet mass balance and palaeoclimate studies. One conventional method of investigating firn profiles is using seismic refraction surveys, but these are limited to point measurements. Distributed acoustic sensing (DAS) presents an opportunity for large-scale seismic measurements of firn with dense spatial sampling and easy deployment, especially when seismic noise is used. We study the feasibility of seismic noise interferometry on DAS data for characterizing the firn layer at the Rutford Ice Stream, West Antarctica. Dominant seismic energy appears to come from anthropogenic noise and shear-margin crevasses. The DAS cross-correlation interferometry yields a noisy Green’s function (Rayleigh waves). To overcome this, we present two strategies for cross-correlations: (1) hybrid instruments – correlating a geophone with DAS, and (2) selected stacking where the cross-correlation panels are picked in the tau-p domain. These approaches are validated with results derived from an active survey. Using the retrieved Rayleigh wave dispersion curve, we inverted for a high-resolution 1D S-wave velocity profile down to a depth of 100 m. The inversion spontaneously retrieves a “kink” (velocity gradient inflection) at ~12 m depth, resulting from a change of compaction mechanism. A triangular DAS array is used to investigate directional variation in velocity, which shows no evident variations thus suggesting a lack of deformation in the firn. Our results demonstrate the potential of using DAS and seismic noise interferometry to image the near-surface and present a new approach to derive S-velocity profiles from surface wave inversion in firn studies.

Icequakes, microseismic earthquakes at glaciers, offer critical insights into the dynamics of ice sheets. For the first time in the Antarctic, we explore the use of fibre optic cables as Distributed Acoustic Sensors (DAS) as a new approach for monitoring basal icequakes. Fibre was deployed on the ice surface at Rutford Ice Stream, in two different configurations. We compare the performance of DAS with a conventional geophone network for: microseismic detection and location; resolving source and noise spectra; source mechanism inversion; and measuring anisotropic shear-wave splitting parameters. The DAS arrays detect fewer events than the geophone array. However, DAS is superior to geophones for recording the microseism signal, suggesting the applicability of DAS for ambient noise interferometry. We also present the first full-waveform source mechanism inversions using DAS anywhere, successfully constraining the horizontal stick-slip nature of the icequakes. In addition, we develop an approach to use a 2D DAS array geometry as an effective multi-component sensor capable of accurately characterising shear-wave splitting due to anisotropy of the ice fabric. Although our observations originate from a glacial environment, the methodology and implications of this work are relevant for employing DAS in other microseismic environments.

Uturuncu volcano in southern Bolivia is something of a “zombie” volcano – presumed dead, but showing signs of life. The volcano has not erupted in 250 kyr, but is exhibiting unrest in the form of ground deformation, seismicity, and active fumaroles. Elucidating the subsurface structure of the volcano is key for interpreting this recent unrest. Magnetotelluric measurements revealed alternating high and low resistivity anomalies at depths <10 km beneath the volcano, with a low-resistivity anomaly directly beneath Uturuncu. A key question is, what is the nature of this anomaly? To what extent is it partial melt, a hydrothermal brine reservoir, or a mature ore body? Knowing the density of this anomaly could distinguish between these scenarios, but existing density models of the area lack sufficient resolution. To address this issue, we collected additional gravity measurements on the Uturuncu edifice with 1.5 km spacing in November 2018. Gradient analysis and geophysical inversion of these data revealed several features: a 5 km diameter, high density anomaly beneath the summit of Uturuncu (1 – 3 km elev.), a 20 km diameter ring-shaped negative density anomaly around the volcano (-3 – 4 km elev.), a NNE trending, positive density anomaly northwest of the volcano (0 – 4 km elev.), and a NW trending, negative density anomaly to the southeast. These structures often (but not always) align with resistivity anomalies, features in new seismic tomography models, and relocated earthquake hypocenters. Based on a joint analysis of these data, we interpret the positive density anomaly as a crystallizing dacite pluton, and the negative density ring anomaly as a zone of hydrothermal alteration. Earthquakes around the edges of the crystallizing pluton may represent escaping fluids as the magma cools. The high density anomaly to the northwest likely represents a solidified pluton, and the low density anomaly to the southeast may represent a fractured fault zone. We posit that the alternating zones of high and low resistivity anomalies represent zones of low and high fluid/brine content, respectively. Based on this analysis we suggest that the unrest at Uturuncu is unlikely to be pre-eruptive. This study shows the value of joint analysis of multiple types of geophysical data in evaluating volcanic subsurface structure.

The crust and upper mantle structure of the Greater and Lesser Antilles Arc provides insights into key subduction zone processes in a unique region of slow convergence of old slow-spreading oceanic lithosphere. We use ambient noise tomography gathered from island broadband seismic stations and the temporary ocean bottom seismometer network installed as part of the VoiLA experiment to map crustal and upper mantle shear-wave velocity of the eastern Greater Antilles and the Lesser Antilles Arc. We find sediment thickness, based on the depth to the 2.0 km/s contour in the Grenada and Tobago basins up to 15 km in the south, with thinner sediments near the arc and to the north. We observe thicker crust, based on the depth to the 4.0 km/s velocity contour, beneath the arc platforms with the greatest crustal thickness of around 30 km, likely related to crustal addition from arc volcanism through time. There are distinct low velocity zones (4.2-4.4 km/s) in the mantle wedge (30-50 km depth), beneath the Mona Passage, Guadeloupe-Martinique, and the Grenadines. The Mona passage mantle anomaly may be related to ongoing extension there, while the Guadeloupe-Martinique and Grenadine anomalies are likely related to fluid flux, upwelling, and/or partial melt related to nearby slab features. The location of the Guadeloupe-Martinique anomaly is slightly to the south of the obliquely subducted fracture zones. This feature could be explained by either three-dimensional mantle flow, a gap in the slab, variable slab hydration, and/or melt dynamics including ponding and interactions with the upper plate.

Uturuncu volcano in southern Bolivia is a member of a distinctive class of volcanoes – systems that show unrest despite not having erupted in the Holocene. Uturuncu has not erupted in 250 kyr, but has been deforming (uplift with a moat of subsidence) for several decades, along with seismic swarms and active, sulfur-encrusted fumaroles. Our work builds on previous geophysical imaging at Uturuncu by jointly analyzing multidisciplinary datasets, focusing on imaging the shallow (<15 km depth below surface) structure of the system with geophysical and geochemical data. Whereas previous research pointed to andesite melt at depths >15 km depth, results were ambiguous as to what proportions of melts vs. brines are present at shallower depths. Identifying fluids (melt, brine, etc.) and structures at shallow depths is key for evaluating the hazard potential of the volcano and understanding the source of the unrest. We present new results from gravimetry, seismology (hypocenter relocation, seismic velocity and attenuation tomography), gas geochemistry, and InSAR observations. The results point to an extensive and active hydrothermal system extending ~20 km laterally and ~10 km vertically from Uturuncu, with possible connections at depth to the deeper magmatic system. A combined view of the new density, seismic velocity and attenuation models, and the existing resistivity model is crucial for revealing key features of the hydrothermal system: a vapour-rich conduit beneath Uturuncu (low resistivity/high attenuation column extending from 1.5 to 12.5 km depth), an extensive alteration zone surrounding Uturuncu (complex zone of annular shaped anomalies surrounding Uturuncu from 1.5 to 12.5 km depth), and a possible zone of sulfide deposition just below the western flank of Uturuncu at 1.5 km depth (high density/low resistivity/high attenuation). High fluxes of diffuse CO2 degassing at sub-magmatic temperatures and a small area directly above a low resistivity anomaly subsiding from 2014 to 2017 show that the hydrothermal system is currently active. Analyzed jointly, this multidisciplinary data set suggests that current activity within the shallow structure at Uturuncu is dominated by hydrothermal, rather than magmatic processes.

The combined gravitational pulls from the moon and the sun result in periodical tidal stresses at rates potentially exceeding the tectonic ones. Yet, tidal triggering of earthquakes in critically stressed faults is still under debate and controversial results have been obtained, depending upon specific physical properties and geological settings. Although no universal triggering pattern between earthquakes and tides has been observed in oceanic environments, previous research implies relation between increased seismicity rates and low tides at particular sites at fast-spreading ridges in the Pacific. We present a dataset of 4719 microearthquakes (-1.4≤ML≤4.0) recorded by an Ocean Bottom Seismometer (OBS) network at the slow-spreading equatorial Mid-Atlantic Ridge from March 2016 to February 2017. We use a single-station template matching technique to focus on a small volume, spreading within a ~5km radius from the station. The origin time of the events and their epicentral location is sufficiently determined for a robust comparison with the ocean tides. Our analysis suggests a significant correlation between seismic potential and tidal forces, with the majority of events occurring during or towards low tides, i.e., during maximized extensional stress and maximized extensional stress rate. The tidal dependence of magnitude distribution is also investigated. Although the b-values are generally lower at low tides, the differences are not sufficiently large to achieve statistical significance. However, seismic bursts (enhanced activity rate clusters), occurring at rates above the reference seismicity, are exclusively initiated at extensional stress rates. Coulomb stress modelling implies that slip is promoted during low tides at low-angle normal faults. Local morphology, seismicity distribution and focal mechanisms suggest the existence of high angle faults at shallower depths. Coulomb modelling suggests slip on these faults should not be triggered at low tides unless another factor is considered. One possibility is the presence of a shallow magma chamber. Such a chamber has also been suggested by previous seismic imaging results. Overall, the result yields new insight into magmatic – tectonic cycles and seismicity triggering at mid-ocean ridges.

Hydraulically-forced crevassing is thought to reduce the stability of ice shelves and ice sheets, affecting structural integrity and providing pathways for surface meltwater to the bed. It can cause ice shelves to collapse and ice sheets to accelerate into the ocean. However, direct observations of the hydraulically-forced crevassing process remain elusive. Here we report a new, novel method and observations that use icequakes to directly observe crevassing and determine the role of hydrofracture. Crevasse icequake depths from seismic observations are compared to a theoretically derived maximum-dry-crevasse-depth. We observe icequakes below this depth, suggesting hydrofracture. Furthermore, icequake source mechanisms provide insight into the fracture process, with predominantly opening cracks observed, which have opening volumes of tens to hundreds of cubic meters. Our method and findings provide a framework for studying a critical process, key for the stability of ice shelves and ice sheets, and hence rates of future sea-level rise.

Uturuncu volcano is situated in the Bolivian Andes, directly above the world’s largest crustal body of silicic partial melt, the Altiplano-Puna Magma Body (APMB). Uturuncu last erupted 250,000 years ago, yet is seismically active and lies at the centre of a 70 km diameter uplifted region. Here, we analyse seismicity from 2009 to 2012. Our earthquake locations, using a newly developed velocity model, delineate the top and bottom of the APMB, reveal individual faults, and reconcile differences in depth distribution between previous studies. Spatial clustering analysis of these earthquakes reveals the orientations of the faults, which match stress orientations from seismic anisotropy. Earthquake b-values derived from moment magnitudes (1.4) differ significantly from those using local magnitude measurements (0.8). We suggest that, if possible, moment magnitudes should always be used for accurate b-value analysis. We interpret b-values > 1 in terms of fluid-enhanced seismicity. Shallow seismicity local to Uturuncu yields b-values > 1.1 with some temporal variation, suggesting fluid migration along pre-existing faults in a shallow hydrothermal system, likely driven by advection from the APMB. Intriguingly, events deeper than the APMB also yield large b-values (1.4), mapping the ascent into the lower crust of fluids originating from a subducting slab. Cumulatively, these results provide a picture of an active magmatic system, where fluids are exchanged across the more ductile APMB, feeding a shallow, fault-controlled hydrothermal system. Such pathways of fluid ascent may influence our understanding of arc volcanism, control future volcanic eruptions and promote the accumulation of shallow hydrothermal ore deposits.

The lithosphere-asthenosphere system is fundamental to our understanding of mantle convection and plate tectonics. Seismic and electromagnetic methods are our primary means of determining its structure and physical properties. These independent constraints with different sensitivities to Earth’s properties hold promise for understanding the system. Here we use the shear velocity model from Rayleigh waves and the MT based resistivity model from near the equatorial Mid-Atlantic Ridge. Cross-plots of the models suggest a linear or near-linear trend that is also in agreement with petrophysical predictions. We therefore map the MT model to a new shear-wave starting model using the petrophysical relationship, which is then used to re-invert for shear-wave velocity. The resulting shear-wave velocity model fits the phase velocity data, and the correlation coefficient between the shear velocity and resistivity models is increased. Much of the model can be predicted by expectations for a thermal half-space cooling model, although some regions require a combination of higher temperatures, volatiles, or partial melt. We use the petrophysical predictions to estimate the melt fraction, melt volatile content, and temperature structure of the asthenospheric anomalies. We find up to 4% melt, with the lowest resistivities and shear velocities explained by up to 20% water or 20% CO2 in the melt or ~1% nearly pure sulfide melt, depending on the set of assumptions used. Melt is required in punctuated anomalies over broad depth ranges, and also in channels at the base of the lithosphere. Melt in the asthenosphere is dynamic, yet persistent on geologic time scales.

The gravitational pulls from the moon and the sun result in tidal forces which influence both Earth’s solid and water mass. These stresses are periodically added to the tectonic ones and may become sufficient for initiating rupture in fault systems critically close to failure. Previous research indicates correlations between increased seismicity rates and low tides for mid-ocean, fast-spreading ridges in Pacific ocean. Here, we present a microseismicity dataset (4719 events) from an Ocean Bottom Seismometer (OBS) network at the equatorial Mid-Atlantic Ridge, suggesting a significant correlation between seismic potential and tidal forces. We show that low as well as decreasing ocean water level results in elevated seismicity rates and lower b-values, translated into considerably increased probabilities of stronger event occurrence at or towards low tides. In addition, seismic bursts (enhanced activity rate clusters), occurring at rates fairly above the reference seismicity, are exclusively present during either high extensional stresses or high extensional stress rates. Our results exhibit remarkable statistical significance, supporting the previous findings for tidal triggering at low tides within normal-faulting regimes and extending the range of observations to slow-spreading ridges. Observed triggering of slip on low angle faults at low tides is predicted by Coulomb stress modelling. The triggering of slip on high angle faults observed here, is not easily explained without another factor. It may be related to fatigue and/or the presence of a shallow magma body beneath the ridge, as suggested by previous seismic imaging in the region.

Oceanic Transform Faults (TF) comprise first order discontinuities bounded between mid-ocean ridge spreading centres. TF mainly accommodate strike slip motion, separating lithospheric plates of different age and thermal structure. Oceanic TF are intriguing in that they do not produce earthquakes as large as might be expected given their long length, with seismic slip corresponding only to a small fraction of the total tectonic slip. The relative geologic simplicity of oceanic TF means that they are an important analogue for more hazardous continental TF, with high potential for improving insights into the earthquake cycle. We investigate the earthquake properties along Chain, a ~300 km long TF in the equatorial MAR by combining both microseismic and teleseismic data. We use the ~1-year microseismicity data (total of 812 events) gathered during the PI-LAB (Passive Imaging of the Lithosphere-Asthenosphere Boundary) experiment and EURO-LAB (Experiment to Unearth the Rheological Lithosphere-Asthenosphere Boundary). We perform cluster analysis in multi-dimensional phase space, consisting of various seismic (epicentral coordinates, magnitude) and geophysical (gravity anomalies, bathymetry, tidal height) parameters. We investigate potential triggering mechanisms, including tidal, static and dynamic stresses. We extend our analysis back in time by considering stronger earthquakes (MW>~5.0) from Global Centroid Moment Tensor (GCMT) since 1976. We find three unique, 50-100 km long clusters or segments from our analysis going from east to west, separated by seismic gaps. Microseismic activity is highest at the eastern segment of Chain where there is the largest positive flower structure, negative rMBA gravity anomaly but very few M>5.5 events. The western segment has reduced seismicity rates relative to the eastern, and is associated with a positive rMBA and a few small flower structures. The central segment is bounded between two seismic gaps and demonstrates relatively high activity rates in the middle. Our result suggests that trans-pression of highly altered mantle/crust and/or high pore pressure due to hydrothermal fluid circulation in the eastern flower structure enhances seismic activity. Overall, we find the existence of consecutive locking and creeping segments, with some of the patches exhibiting hybrid behaviour, potentially causing their sporadic activation/reactivation.