We demonstrate the feasibility of building an accurate velocity model, containing complex salt bodies, using only raw surface seismic data, advanced waveform inversion, and a start model that contains no salt and for which the assumed sedimentary velocities are systematically in error. We use ocean-bottom node data without exceptionally long offsets or unusually low frequencies, and we do not use reflection tomography, conventional migration, salt-boundary picking, explicit salt flooding or scenario testing. We are able to recover a model automatically that is sufficiently accurate for elastic conventional FWI. To achieve this, we use a selfcorrecting multi-stage inversion scheme, employing different implementations of adaptive waveform inversion as it proceeds. The earliest stages use an accelerated formulation, combined with simple constraints, to achieve the large velocity changes necessary using minimal compute. Use of Graviton 4 spot instances on the cloud further reduces cost. The final 10-Hz acoustic velocity model is both accurate and well resolved; it generates structurally simple FWI impedance reflectivity images beneath the salt and is ready for immediate multi-parameter elastic FWI run up to the highest frequencies in the dataset.

FWI iteratively refines subsurface velocity models by minimizing the misfit between observed and simulated datasets. AWI is a robust deconvolution-based misfit function for FWI that performs well with poor initial models and incomplete wave physics. Here we demonstrate that conventional 1D AWI can be extended to higher dimensions: 2D AWI for towed-streamer data and 3D AWI for oceanbottom data. These extensions are particularly effective for datasets that show significant elastic effects. For a marine towed-streamer dataset exhibiting moderate elastic behavior, combining 1D and 2D AWI enabled accurate kinematic matching, and captured steep velocity features likely to be associated with zones of enhanced simple shear, while avoiding overfitting finer details. The resulting AWI model produced simpler structures following Kirchhoff prestack depth migration, and produced significantly flatter gathers, when compared to a workflow that combined reflection tomography and conventional FWI.

Acoustic full-waveform inversion is often able to build highquality velocity models for sub-salt imaging using only acoustic wave propagation. This is achieved by adaptively matching acoustic data to the real-world elastic field data. Here we show both acoustic and elastic adaptive waveform inversion and reflection waveform inversion applied to saltaffected ocean-bottom node data from the Santos Basin, offshore Brazil. In this basin, the existence of high-velocity layered anhydrites within the salt generates elastic effects that appear sufficiently severe that full-elastic inversion is necessary in order to recover an adequate model. Combined RWI and elastic AWI are then able to generate a high-quality velocity model from raw hydrophone data that is sufficient for subsequent imaging using simple acoustic RTM.

Conventional seismic velocity model building in complicated salt-affected areas requires the explicit identification of salt boundaries in migrated images and typically involves testing of possible subsurface scenarios through multiple generations. The resulting velocity models are slow to generate and may contain interpreter-driven features that are difficult to verify. We show that it is possible to build a full final velocity model using advanced forms of full-waveform inversion applied directly to raw field data, starting from a model that contains only a simple 1D compaction trend. This approach rapidly generates the final velocity model and migrates processed reflection data at least as accurately as conventionally generated models. We demonstrate this methodology using an ocean-bottom-node data set acquired in deep water over Walker Ridge in the Gulf of Mexico. Our approach does not require exceptionally long offsets or the deployment of special low-frequency sources. We restrict the inversion so it does not use significant energy below 3 Hz or offsets longer than 14 km. We use three advanced forms of waveform inversion to recover the final model. The first is adaptive waveform inversion to proceed from models that begin far from the true model. The second is nonlinear reflection waveform inversion to recover subsalt velocity structure from reflections and their long-period multiples. The third is constrained waveform inversion to produce salt- and sediment-like velocity floods without explicitly identifying salt boundaries or velocities. In combination, these three algorithms successively improve the velocity model so it fully predicts the raw field data and accurately migrates primary reflections, though explicit migration forms no part of the workflow. Thus, model building via waveform inversion is able to proceed from field data to the final model in just a few weeks. It entirely avoids the many cycles of model rebuilding that may otherwise be required.

Arc volcanoes are underlain by complex systems of molten-rock reservoirs ranging from melt-poor mush zones to melt-rich magma chambers. Petrological and satellite data indicate that eruptible magma chambers form in the topmost few kilometres of the crust. However, very few chambers have ever been definitively located, suggesting that most are too short-lived or too small to be imaged, which has direct implications for hazard assessment and modeling of magma differentiation. Here we use a high-resolution technology based on inverting full seismic waveforms to image a small, high-melt-fraction magma chamber that was not detected with standard seismic tomography. The melt reservoir extends from similar to 2 to at least 4 km below sea level (b.s.l.) at Kolumbo-a submarine volcano near Santorini, Greece. The chamber coincides with the termination point of the recent earthquake swarms and may be a missing link between a deeper melt reservoir and the high-temperature hydrothermal system venting at the crater floor. The chamber poses a serious hazard as it could produce a highly explosive, tsunamigenic eruption in the near future. Our results suggest that similar reservoirs (relatively small but high-melt-fraction) may have gone undetected at other active volcanoes, challenging the existing eruption forecasts and reactive-flow models of magma differentiation.

The main techniques used to image the brain and obtain structural data are magnetic resonance imaging and X-ray computed tomography. These techniques produce images with high spatial resolution, but with the disadvantage of requiring very large equipment with special installation needs. In addition, X-ray tomography uses ionizing radiation, which limits their use. Ultrasound imaging is a safe technology that is delivered using compact and mobile devices. However, conventional ultrasound reconstruction techniques have failed to obtain images of the brain because of, fundamentally, the presence of the skull and the distortion that it produces on ultrasound. Recent studies have indicated that full-waveform inversion, a computational technique originally from Earth science, has the potential to generate accurate 3-D images of the brain. This technology can overcome the limitations of conventional ultrasound imaging, but a prototype for transcranial applications does not yet exist. Here, we investigate different designs of an annular array of ultrasound transducers to optimize the number of elements and rotations needed to conduct transcranial imaging with full-waveform inversion. This device uses small-diameter, low-frequency transducers that readily propagate ultrasound through the skull with good signal-to-noise ratios. It also incorporates the use of rotations to produce a high-density coverage of the target and acquire redundant traces that are beneficial for full-waveform inversion. We have built a ring of 40 transducers to illustrate that this design is capable of reconstructing images of the brain, retrieving its anatomy and acoustic properties with millimeter resolution. Laboratory results reveal the ability of this device to successfully image a 2.5-D brain- and skull-mimicking phantom using full-waveform inversion. To our knowledge, this is the first prototype ever used for transcranial-like imaging. The importance of these findings and their implications for the design of a 3-D reconstruction system with possible clinical applications are discussed.

Arc volcanoes are underlain by complex systems of molten-rock reservoirs ranging from melt-poor mush zones to melt-rich magma chambers. Petrological and satellite data indicate that eruptible magma chambers form in the topmost few kilometres of the crust. However, very few chambers have ever been definitively located, suggesting that most are too short-lived or too small to be imaged, which has direct implications for hazard assessment and modelling of magma differentiation. Here we use a high-resolution technology based on inverting full seismic waveforms to image a small, high-melt-fraction magma chamber that was not detected with standard seismic tomography. The melt reservoir extends from ~2 to at least 4 km below sea level (b.s.l.) at Kolumbo – a submarine volcano near Santorini, Greece. The chamber coincides with the termination point of the recent earthquake swarms and may be a missing link between a deeper melt reservoir and the high-temperature hydrothermal system venting at the crater floor. The chamber poses a serious hazard as it could produce a highly explosive, tsunamigenic eruption in the near future. Our results suggest that similar reservoirs (relatively small but high melt-fraction) may have gone undetected at other active volcanoes, challenging the existing eruption forecasts and reactive-flow models of magma differentiation.

This paper explores the potential for hyperscale public cloud high-performance compute (HPC) to enable efficient deployment of a semi-global approach to multi-parameter full-waveform inversion (FWI) over large areas. We introduce several novel aspects to semi-global FWI that improve convergence and suppress crosstalk, while establishing that the algorithm’s embarrassingly parallel nature is well suited for public cloud implementation. We describe how various public cloud services can be taken advantage of to reduce the cost of the inversion and provide a reference architecture for the deployment of semi-global FWI to the Amazon Web Services (AWS) platform. Finally, we apply semi-global FWI to raw data from a large-scale legacy surface seismic dataset acquired offshore Australia as part of a re-processing sequence undertaken recently. Our results demonstrate that semi-global FWI can be effectively parallelized across more than one million logical central processing units (CPUs) and is able to recover an anisotropic velocity model in a few hours and in an automated fashion.

We investigate the geometry of the metamorphic basement of the Santorini volcanic island using ambient noise data to determine the pre-Alpine/pre-volcanic bedrock structure. The geometry of pre-volcanic Santorini is important in order to constrain the recent volcanic history of the island and also to study the site-effect of the volcanic formations on seismic motions. Santorini is the most active volcano of the Southern Aegean Volcanic Arc and is the southernmost island of the Cyclades islands metamorphic complex. As a result, the volcanic material that has accumulated during the last 600+ Kyrs has been superimposed on the pre-volcanic Santorini (Cycladic) island. To map the thickness of volcanic material, we have performed a large number (>200) of single-station noise measurements in the Santorini area. Measurements were mainly performed using conventional acquisition systems (Guralp-40T 30sec seismometer and Reftek-130A digitizer). We also employed additional single-station noise data from several previous studies (Dimitriadis et al. 2006, PROTEUS Project 2015), as well as permanent stations from the Hellenic Seismological Network in the same region. HVSR curves were calculated using single-station noise data and were used to estimate the fundamental frequency, f0, as well as the corresponding maximum HVSR amplitude, A0HVSR. The majority of HVSR curves showed prominent peaks (A0HVSR locally larger than 7-8), indicating a clear impedance contrast between volcanics and metamorphic formations. To map the bedrock depth, we estimated the thickness of the upper volcanic formations using the quarter-wavelength approximation for each site. For this assessment, the average shear-wave velocity (Vs) of the volcanic formations was estimated from the inversion of several passive ambient noise array data, as well as additional constraints from selected MASW measurements. Where possible, the reliability of the spatial variation of volcanic formation thickness was checked with independent geological information. Using the digital elevation model and the volcanic formation thickness for each site of the single-station noise data, we estimated the spatial distribution of the pre-Alpine, metamorphic bedrock depth. The resulting geometry of the pre-volcanic Santorini island shows very deep basins (now filled with volcanic formations) around the pre-Alpine bedrock outcrop in the southern part of Santorini (Profitis Ilias), increasing to 100+ meters in the Kamari-Perissa basin area (southeastern Santorini) and to more than 400+ in the central (Fira-Imerovigli) and the north Santorini areas (Oia), in agreement with recent larger-scale tomographic results (Heath et al., 2019). The results are also in very good agreement with the pre-Alpine bedrock geometry independently inferred from gravity data inversion (Tzanis et al., 2019.) This research is co-financed by Greece and the European Union (European Social Fund- ESF) through the Operational Programme «Human Resources Development, Education and Lifelong Learning» in the context of the project “Strengthening Human Resources Research Potential via Doctorate Research” (MIS-5000432), implemented by the State Scholarships Foundation (ΙΚΥ), the Hellenic Foundation for Research and Innovation (HFRI) under the “First Call for HFRI Research Projects to support Faculty members and Researchers and the procurement of high-cost research equipment grant” (Project Number: 2924) and the Institute for the Study and Monitoring Of the SAntorini Volcano (ISMOSAV).

We demonstrate that an appropriate combination of adaptive waveform inversion (AWI) (Guasch et al., 2019), kinematic reflection waveform inversion (RWI) (Warner et al., 2018), and quantum particle-swarm global optimization (qPSO) (Debens et al., 2015), is able to generate accurate wellresolved velocity models from unprocessed raw field data. We begin from simple one-dimensional starting models, we use minimal human intervention, and we recover velocity models, that are both kinematically accurate and highly resolved in space, to depths that lie well below the deepest penetration of refracted arrivals. Using this approach, refractions, reflections and multiples all contribute to the quality of the final velocity model. We demonstrate the efficacy of this approach using a realistic blind synthetic dataset in 2D, and using the corresponding reflection-dominated narrow-azimuth 3D field dataset that served, in part, as the motivation and archetype for the synthetic. For the field data, we demonstrate a close match to a blind well that penetrates below the refracted arrivals. This approach can build final velocity models in a small fraction of the time required for conventional depth velocity-model building. We show three types of inversion: (1) simple vanilla FWI which evolves towards local minima, leading to mis-convergence when starting far from the true answer, (2) AWI which increase the region of convexity that surrounds the global minimum, and (3) AWI-RWI which targets residual timing errors and that has explicit sensitivity to reflection moveout and so is able to recover deep macro-model velocity reflection updates.

Summary We have run FWI to 100 Hz on raw field data from a deep-water marine towed-streamer dataset. We show that the results are similar to, and are broader bandwidth than, conventionally processed PSDM images. FWI removes multiples and ghosts from the raw data, and can produce a full-bandwidth PSDM reflectivity image in a few days without conventional processing or migration.

Summary Image-to-image translation using GANs have successfully been applied to a wide variety of problems, from mundane implementations to turn horses into zebras, to stunning synthetic media deepfakes. We explore its application in geophysics as a cost-reduction tool, and demonstrate its potential in 3D field data. We show it can be used to learn the mapping between different data flavours of interest in modern data processing workflows: acoustic/elastic for full-waveform inversion and geophone vertical-component/hydrophone-pressure for up- and down-going wavefield separation.

Detailed knowledge about geometry and physical properties of magmatic systems at arc volcanoes promises to better constrain models of magma differentiation, transit and storage in the crust, and to help assess volcanic hazard. Unfortunately, low-velocity zones associated with melt accumulation are particularly difficult to image by conventional travel-time tomography due to its limited resolving power, resulting in blurred boundaries and underestimated velocity contrasts. Here we alleviate these issues by applying full-waveform inversion (FWI) to study a magmatic system of Santorini - an active, semi-submerged volcano with a known record of large, caldera-forming eruptions. We use a 3D wide-angle, multi-azimuth seismic dataset from the recent PROTEUS experiment acquired with ca. 150 ocean-bottom/land seismic stations and ca. 14,000 air-gun shots. We implement a finite-difference immersed boundary method to simulate reflections off the caldera’s irregular topography, and pressure-velocity conversion to take full advantage of the multi-component data. We perform inversion with careful data-selection, increasing frequency up to 6 Hz, and extensive quality-control based on a phase spatial-continuity criterion. A final P-wave velocity model of the upper crust offers a high-resolution image of Santorini magmatic and hydrothermal systems with pronounced low-velocity zones due to a high melt and water content respectively. The features are better resolved and the velocity contrasts distinctly sharper than in the starting model obtained with travel-time tomography. We also recover a previously undetected low velocity anomaly of >40% beneath Kolumbo - a submarine volcanic cone to the NE of Santorini caldera. We interpret this anomaly as a magmatic sill.

Volcanic calderas are surface depressions formed by roof collapse following evacuation of magma from an underlying reservoir. The mechanisms of caldera formation are debated and predict differences in the evolution of the caldera floor and distinct styles of magma recharge. Here we use a dense, active source, seismic tomography study to reveal the sub-surface physical properties of the Santorini caldera in order to understand caldera formation. We find a ∼3-km-wide, cylindrical low-velocity anomaly in the upper 3 km beneath the north-central portion of the caldera, that lies directly above the pressure source of the 2011-2012 inflation. We interpret this anomaly as a low-density volume caused by excess porosities of between 4% and 28%, with pore spaces filled with hot seawater. Vents that were formed during the first three phases of the 3.6 ka Late Bronze Age (LBA) eruption are located close to the edge of the imaged structure. The correlation between older volcanic vents and the low-velocity anomaly suggests that this feature may be long-lived. We infer that collapse of a limited area of the caldera floor resulted in a high-porosity, low-density cylindrical volume, which formed by either chaotic collapse along reverse faults, wholesale subsidence and infilling with tuffs and ignimbrites, phreatomagmatic fracturing, or a combination of these processes. Phase 4 eruptive vents are located along the margins of the topographic caldera and the velocity structure indicates that coherent down-drop of the wider topographic caldera followed the more limited collapse in the northern caldera. This progressive collapse sequence is consistent with models for multi-stage formation of nested calderas along conjugate reverse and normal faults. The upper crustal density differences inferred from the seismic velocity model predict differences in subsurface gravitational loading that correlate with the location of 2011-2012 edifice inflation. This result supports the hypothesis that sub-surface density anomalies may influence present-day magma recharge events. We postulate that past collapses and the resulting topographical and density variations at Santorini influence magma focusing between eruptive cycles, a feedback process that may be important in other volcanoes.

At extensional volcanic arcs, faulting often acts to localize magmatism. Santorini is located on the extended continental crust of the Aegean microplate and is one of the most active volcanoes of the Hellenic arc, but the relationship between tectonism and magmatism remains poorly constrained. As part of the Plumbing Reservoirs Of The Earth Under Santorini experiment, seismic data were acquired across the Santorini caldera and the surrounding region using a dense amphibious array of >14,300 marine sound sources and 156 short-period seismometers, covering an area 120 km by 45 km. Here a P wave velocity model of the shallow, upper-crustal structure (<3-km depth), obtained using travel time tomography, is used to delineate fault zones, sedimentary basins, and tectono-magmatic lineaments. Our interpretation of tectonic boundaries and regional faults are consistent with prior geophysical studies, including the location of basin margins and E-W oriented basement faults within the Christiana Basin west of Santorini. Reduced seismic velocities within the basement east of Santorini, near the Anydros and Anafi Basins, are coincident with a region of extensive NE-SW faulting and active seismicity. The structural differences between the eastern and western sides of Santorini are in agreement with previously proposed models of regional tectonic evolution. Additionally, we find that regional magmatism has been localized in NE-SW trending basin-like structures that connect the Christiana, Santorini, and Kolumbo volcanic centers. At Santorini itself, we find that magmatism has been localized along NE-SW trending lineaments that are subparallel to dikes, active faults, and regional volcanic chains. These results show strong interaction between magmatism and active deformation.

The Large Synoptic Survey Telescope (LSST) large field of view is achieved through a three-lens camera system and a three-mirror optical system comprised of a unique 8.4-meter diameter monolithic primary/tertiary mirror (M1M3) and a 3.4-meter diameter secondary mirror (M2)(1). The M2 is a 100mm thick meniscus convex asphere. The M2 Assembly includes a welded steel cell and a support system comprised of 72 axial and 6 tangential electromechanical actuators to control the mirror figure. The M2 Assembly (including optical polishing and integrated optical testing) is being fabricated by Harris Corporation in Rochester, NY. The summary status of this system and results are presented.

The Large Synoptic Survey Telescope (LSST) is a large (8.4 meter) wide-field (3.5 degree) survey telescope, which will be located on the Cerro Pachón summit in Chile. Both the Secondary Mirror (M2) Cell Assembly and Camera utilize hexapods to facilitate optical positioning relative to the Primary/Tertiary (M1M3) Mirror. A rotator resides between the Camera and its hexapod to facilitate tracking. The hexapods and rotator have been designed, fabricated, assembled, and are currently being tested by Moog CSA. An update on these activities is provided along with a detailed discussion of the testing approach and results ranging from proof load and life testing to positioning performance. Particular emphasis is given to testing of the positioning accuracy, repeatability, and resolution of the hexapods and tracking accuracy and runout of the rotator. Verification of power off braking, heat dissipation, settling time, range of motion, and velocity requirements are also presented.

Hydrothermal circulation at mid‐ocean ridges is responsible for ~25% of Earth's heat flux and controls the thermal and chemical evolution of young oceanic crust. The heat flux of black smoker hydrothermal systems is thought to be primarily controlled by localized magma supply and crustal permeability. Nevertheless, magma chamber characteristics and the nature of crustal permeability beneath such systems remain unclear. Here we apply three‐dimensional full‐waveform inversion to seismic data from the hydrothermally active Endeavour segment of the Juan de Fuca Ridge to image the upper crust in high resolution. We resolve velocity variations directly above the axial magma chamber that correlate with variations in seismicity, black smoker heat flux, and the depth of the axial magmatic system. We conclude that localized magma recharge to the axial magma lens, along with induced seismogenic cracking and increased permeability, influences black smoker heat flux.