Infrasonic scattering by fine-scale atmospheric sound-speed structure causes variability in signals generated by repeated explosions in Hukkakero, Finland, recorded by the IS37 microbarometer array at 321 km distance. The sample covariance matrix and beam pattern of the second eigenmode (array measures) are sensitive to changes in wavefield structure below array resolution. The directional energy distribution is approximated by a skewed Gaussian distribution, where the inter-sensor coherence decay is inverted for the beamwidth and the skew is indicated by the second-eigenmode beam pattern. With increasing lapse time in the scattered wave train, the beamwidth increases over both apparent velocity and backazimuth, and the energy distribution is skewed with fluctuating directionality over backazimuth. Progressive over-prediction of the second-eigenmode magnitude by the best-fit Gaussian distribution throughout the wave train implies an increase in the mutual directional coherence of the wavefield—i.e., phase correlation in different look directions—relative to the beamwidth. Multi-scale entropy and spectral-amplitude correlation (single-sensor measures) indicate a time-variant mixture of ballistic and diffusive arrivals, often with similar amplitudes. These array and single-sensor measures should be considered when inverting for fine-scale sound speed structure in the middle atmosphere, extending beyond the standard analysis of infrasonic temporal and spectral content.

Seismic signals generated by near-surface explosions, with sources including industrial accidents and terrorism, are often analysed to assist post-detonation forensic characterization efforts such as estimating explosive yield. Explosively generated seismic displacements are a function of, amongst other factors: the source-to-receiver distance, the explosive yield, the height-of-burst or depth-of-burial of the source and the geological material at the detonation site. Recent experiments in the United States, focusing on ground motion recordings at distances of $< 15\,$ km from explosive trials, have resulted in empirical models for predicting P-wave displacements generated by explosions in and above hard rock (granite, limestone), dry alluvium, and water. To extend these models to include sources within and above saturated sediments we conducted eight explosions at Foulness, Essex, UK, where $\sim 150\,$ m thicknesses of alluvium and clay overlie chalk. These shots, named the Foulness Seismoacoustic Coupling Trials (FSCT), had charge masses of 10 and 100 kg TNT equivalent and were emplaced between 2.3 m below and 1.4 m above the ground surface. Initial P-wave displacements, recorded between 150 and 7000 m from the explosions, exhibit amplitude variations as a function of distance that depart from a single power-law decay relationship. The layered geology at Foulness causes the propagation path that generates the initial P-wave to change as the distance from the source increases, with each path exhibiting different amplitude decay rates as a function of distance. At distances up to 300 m from the source the first arrival is associated with direct propagation through the upper sediments, while beyond 1000 m the initial P-waves are refracted returns from deeper structure. At intermediate distances constructive interference occurs between P-waves propagating through the upper sediments and those returning from velocity-depth gradients at depths between 100 and 300 m. This generates an increase in displacement amplitude, with a maximum at $\sim 800\,$ m from the source. Numerical waveform modelling indicates that observations of the amplitude variations is in part the consequence of high P- to S-wave velocity ratios within the upper 150 m of saturated sediment, resulting in temporal separation of the P and S arrivals. We extend a recently developed empirical model formulation to allow for such distance-dependent amplitude variations. Changes in explosive height-of-burst within and above the saturated sediments at Foulness result in large P-wave amplitude variations. FSCT surface explosions exhibit P-wave displacement amplitudes that are a factor of 22 smaller than coupled explosions at depth, compared to factors of 2.3 and 7.6 reported for dry alluvium and granite, respectively.

Over the last decade there has been an international effort to find methods to recover and digitize recordings from historical earthquakes and explosions that occurred during the 1950’s through to the 1980’s. Making these recordings accessible in digital format offers opportunities to study what signatures are encoded in the data, and to apply state-of-the-art techniques and methods to historical data. In this study we employ unsupervised machine learning to cluster historical teleseismic waveforms from nuclear explosions conducted at the former USSR Degelen test site, in Kazakhstan, recorded at seismic arrays in the UK (EKA), Canada (YKA), Australia (WRA) and India (GBA). In particular, we use two unsupervised algorithms to cluster waveforms using shape-based clustering: kernel k-means and k-Shape. The algorithms clearly split waveforms into distinct clusters that are spatially related, even when waveform differences are subtle, and we show with local and teleseismic numerical simulations that the clusters are related to the topography. The topography at the Degelen test site has characteristic wavelengths of 2-4 km and local simulations highlight that the seismic wavefield is trapped in reverberating mountain peaks. The location of the explosion is crucial in determining which section of the mountain range reverberates, influencing the outgoing wavefield. Teleseismic waveform simulations confirm that it is this superposition of energy leaving the reverberating peaks that results in the observed teleseismic waveform differences we observe.

SUMMARY The United Kingdom (UK) experiences low-to-moderate levels of seismicity; only 12 onshore earthquakes with local magnitude (ML) ≥ 4.0 have been recorded in the past 20 yr. It is therefore difficult to estimate moment magnitude (Mw) using conventional moment tensor inversion for the majority of UK seismicity, resulting in limited reliable estimates of Mw. To address this, we calibrated coda envelopes at 16 broad-band seismic stations distributed across the UK to produce an Mw catalogue for 100 events with Mw ≥ 2.13 that occurred since 2006. This was achieved using the open-source Coda Calibration Tool, which requires independent source parameter estimates for calibration. For 13 UK events between 2006 and 2022, we used spectral modelling to estimate apparent stress (0.32–1.74 MPa) and moment tensor inversion to estimate Mw (3.35–4.52). These independent source parameters formed a subset of the inputs into the final calibration, which used seismic data from 33 events with coda-derived values of 2.57$\le $ Mw$\le $4.49. The resultant coda calibration parameters were applied to 67 further events (Mw ≥ 2.13). The coda envelopes exhibit slow seismic coda decay across the UK, with significant energy up to 20 Hz, consistent with other regions of low tectonic activity. This Mw catalogue, and the application of the calibration to future UK seismic events, will be useful for both assessing seismic hazard and event characterization.

ABSTRACT On 7 July 2011, a series of accidental explosions occurred in the town of Abadan, Turkmenistan. The explosions were observed at local distances (11 and 23 km) at the International Monitoring System (IMS) seismic array, GEYT, and at the Central Asian Cross-border Network (CAREMON), seismic station, ASHT. A total of 30 individual seismic events could be identified. IMS infrasound arrays also observed these explosions at longer ranges, for example, from 1374 km (I31KZ, Kazakhstan) to 4307 km (I48TN, Tunisia). These local and long-range observations allow us to compare yields estimated from the seismic data, 0.05–45 tonnes (trinitrotoluene [TNT] equivalent; using P-wave and Rg amplitude relationships) to those estimated from the infrasound data, 4–90 tonnes (using the dominant period). Given the long propagation distances to the closest IMS infrasound arrays, the signal durations are long and individual events in the series cannot be identified in the infrasound signals. Comparison of the seismic and infrasonic yield estimates suggests that the infrasonic dominant periods are likely associated with the largest explosions in the series. In addition to the seismic arrivals observed locally, air-to-ground coupled waves exhibiting downward first motions consistent with an initial positive blast overpressure are also observed at both GEYT and ASHT. The peak-to-peak amplitude of the air-to-ground coupled waves increases with increasing yield, whereas the period of these signals is yield independent. These amplitude observations suggest that air-to-ground coupled arrivals could be better exploited within event analyses and provide a further constraint upon yield. Fully exploiting these events of opportunity, that have multiphenomenology observations, will help to further improve our understanding of how energy from near-surface explosions partitions into the ground and atmosphere and, therefore, improve our understanding of future events.

We apply a hybrid method that couples global Instaseis databases (van Driel et al., 2015) with a local finite-difference code WPP (Nilsson et al., 2007) to study the 1960s-1980s nuclear explosions located at the USSR Degelen mountain test site. Observed teleseismic P waves (up to 2 Hz) display strong near-source signatures, yet the relative importance of contributing factors – such as explosion depth and yield, scattering from near-source topography and geological heterogeneities, as well as non-linear effects – are not well understood. An analysis of teleseismic waveforms suggest that these features are dependent on the source location within the Degelen mountain range, while depths and yields do not show a consistent effect. We therefore propose that the change in signal characteristics on teleseismic waveforms is related to the mountainous topography in the source region and we turn to deterministic hybrid modelling to test the effect of Degelen topography at teleseismic distances. Despite simplistic modelling assumptions, we achieve an excellent fit with the observed waveforms. Amplitudes are in good agreement and many observed features are reproduced by synthetic seismograms at 2 Hz, highlighting the importance of near-source 3-D effects on long-range wave propagation. Hybrid modelling of more realistic high-frequency scenarios could ultimately lead to waveform-based constraints on explosion locations, for example via grid-search methods or more advanced learning algorithms, or even improve nuclear discrimination methods.

A research concept that was first presented on a poster at a Comprehensive Nuclear-Test-Ban Treaty (CTBT) conference in 2009 has resulted in a fully-fledged operational software product named NET-VISA (Network Processing Vertically Integrated Seismic Analysis), which performs the Network Processing step of the automatic processing at the International Data Centre (IDC). It has become one of the tools of the waveform analysts to review and improve the IDC standard automatic third Standard Events List (SEL3) bulletin and produce the Reviewed Events Bulletin (REB), the only global seismo-acoustic bulletin. The basic scientific concepts imbedded into NET-VISA are briefly summarized in this paper but the emphasis is on the process of adopting, developing, adapting, testing, and operationalizing the initial prototype. Extensive off-line testing has shown that one of the expected benefits of NET-VISA was the significant reduction in missed events rate compared to the current Network Processing software, Global Association (GA). NET-VISA also finds some events previously missed by manual review. Starting in July 2017, and as of January 2020, NET-VISA generates an automatic bulletin, called VSEL3, in parallel to SEL3, which is also used by analysts since January 2018. Tracing the origin of the REB events has confirmed the significant reduction in missed events when complementary VSEL3 events are reviewed in addition to the SEL3 events. If sufficient confidence is established, NET-VISA has the potential to replace GA in producing the standard automatic SEL3 bulletin.

In this study, seismic and hydroacoustic signals from underwater explosions in 2001, 2008, and 2016 near Florida are analyzed. These 10,000 lb chemical explosions were detonated by the United States Navy to validate the ability of new classes of ships to withstand explosions. For many of the explosions, the ground-truth (GT) epicenters are known. These epicenters are used to improve the accuracy of the locations of explosions with no GT data by performing a relative relocation using a Bayesian hierarchical seismic-event locator. Seismic and hydroacoustic signals are also used to characterize the underwater explosion sources. Bubble pulse modulations, characteristic of underwater explosions, are identified at seismic stations in the United States, and the observed bubble pulse frequency is consistent with published GT information. The absence of clear modulations in the spectra caused by reverberations in the water column means that the depth of the explosions in the water and hence the trinitrotoluene (TNT) equivalent charge weight of the explosion cannot be resolved from the frequency of the bubble pulse modulations. Published estimates of the local magnitudes M-L and the known charge weights of these explosions are compared with data from previous underwater explosions. A relationship between charge weight and M-L from previous well-calibrated explosions detonated in the Dead Sea is shown to provide reasonable estimates of the charge weight once corrected for the salinity of the seawater near Florida. Hydroacoustic signals from the Florida underwater explosions are also observed as H phases on hydrophone sensors near Ascension Island. The bubble pulse is not observed as clearly at the hydrophone sensors at Ascension Island possibly as a result of signal distortion in the shallow water close to Florida. This has implications for event identification using hydrophone stations and demonstrates the importance of combining seismic and hydroacoustic observations.

The Zagros mountain belt has an unusually large discrepancy between seismic and geodetic strain rates, implying very large aseismic release of strain. However, the spatial and depth relationship between seismic and aseismic deformation is poorly understood and controversial, with important implications both for understanding the role of aseismic deformation in regions of continental convergence and for characterizing seismic hazard for large urban populations in this region. Two recent earthquake sequences in 2008 and 2012 provide us with an ideal opportunity to use geodetic and seismological data to address this topic, not only for thrust faulting but also for rarer strike-slip faulting in the Zagros. These aftershock sequences occurred on the southeast border between Iran and Iraq. We use Interferometric Synthetic Aperture Radar (InSAR) to obtain a mechanism for the 2008 mainshock and observe significant aseismic slip accompanying the earthquake. This aseismic slip occurred along strike from the seismic asperity and approximately doubled the seismic moment release, demonstrating that aseismic slip plays an important role for strike slip as well as thrust faulting in the Zagros. Depths are calculated by inverting surface-wave amplitude spectra and depth phase observations, demonstrating that all events occur at depths less than 12 km, supporting the view that seismicity predominately occurs in the sedimentary cover in the Zagros. Using both the calculated depths and InSAR location, we relocate the aftershocks to reveal two distinct clusters, with the 2012 cluster occurring in the vicinity of the Zagros foredeep fault, challenging the traditional view that this region is aseismic.

Recordings made at five broadband seismometers, deployed in central London during the summer of 2015, reveal the wideband nature (periods T of between 0.01 and 100 s) of anthropogenic noise in a busy urban environment. Temporal variations of power spectral density (PSD) measurements suggest that transportation sources generate the majority of the noise wavefield across the entire wideband, except at the secondary microseismic peak (2 < T < 6 s). The effect of road traffic is greatest at short periods (T < 0: 4 s) for which acceleration noise powers are similar to 20 dB larger than the new high-noise model; at T similar to 0: 1 s daytime root mean square acceleration amplitudes are 1000 times higher in central London than at an observatory station in Eskdalemuir, Scotland. Overground railways generate observable signals both at short periods (T < 0: 3 s), which are recorded in close proximity to the tracks, and at very long periods (T > 20 s), which are recorded across the city. We record a unique set of signals 30 m above a subway (London Underground) tunnel interpreted as a short-period dynamic component, a quasi-static response to the train moving underneath the instrument and a very long period (T > 30 s) response to air movement around the tunnel network. A lowvelocity clay and sand overburden tens of meters thick is shown to amplify the horizontal-component wavefield at T similar to 1 s, consistent with properties of the London subsurface derived from engineering investigations. We provide tabulated median PSD values for all stations to facilitate comparison with any future urban seismic deployments.

The Carboneras fault zone forms part of a major strike-slip fault system in SE Spain, striking NE–SW, and accommodating up to 40km displacement. It affects basement metamorphic rocks and unconformably overlying upper Miocene sediments and volcanic rocks. High-resolution shallow seismic tomographic sections were made across the fault zone in two localities. From the same areas, fault rocks and their wallrocks were collected for laboratory seismic velocity measurements. The laboratory data were corrected for the substantial effects of near-surface crack damage. By combining these results with geological cross sections, forward velocity models for the fault zone were constructed to compare with field seismic measurements and hence to ‘ground-truth’ the inferences made from them. These velocity/depth relationships matched moderately well with those extracted from the in-situ tomography results. Aspects of the in-situ seismic sections matched features on the forward-modelled sections, but the comparisons showed that it is important to have some degree of foreknowledge of the geology to be able successfully to interpret seismic tomography sections as an exploration tool.

We are mapping the topography of upper mantle seismic discontinuities beneath the North Atlantic and surrounding regions by using precursor arrivals to PP and SS seismic waves that reflect off the seismic discontinuities. Numerous source–receiver combinations have been used in order to collect a large dataset of reflection points beneath our investigation area. We analysed over 1700 seismograms from MW>5.8 events using array seismic methods to enhance the signal to noise ratio. The measured time lag between PP (SS) arrivals and their corresponding precursors on robust stacks are used to measure the depth of the transition zone boundaries. The reflectors' depths show a correlation between the location of known hotspots and a significantly depressed 410 km discontinuity indicating a temperature increase of 50–300 K compared to the surrounding mantle. For the 660 km discontinuity three distinct behaviours are visible: (i) normal depths beneath Greenland and at a distance of a few hundred kilometres away from known hotspots, (ii) shallower 660 km discontinuity compared with the global average value near hotspots closer to the Mid-Atlantic Ridge, and (iii) very few observations of a 660 km discontinuity at the hotspot locations. We interpret our observations as a large upwelling beneath the southern parts of our study region, possibly due to the South Atlantic convection cell. The thermal anomaly may be ponding beneath the endothermic 660 km phase transformation and likely does not extend through the top of the transition zone as a whole, except for those branches which appear as the thinner upwellings of Azores, Canaries and Cape Verde hotspots at the surface.

The 2008 M-w 6.3 Damxung earthquake on the Tibetan Plateau is investigated to (i) derive a coseismic slip model in a layered elastic Earth; (ii) reveal the relationship between coseismic slip, afterslip and aftershocks and (iii) place a lower bound on mid/lower crustal viscosity. The fault parameters and coseismic slip model were derived by inversion of Envisat InSAR data. We developed an improved non-linear inversion scheme to find an optimal rupture geometry and slip distribution on a fault in a layered elastic crust. Although the InSAR data for this event cannot distinguish between homogeneous and layered crustal models, the maximum slip of the latter model is smaller and deeper, while the moment release calculated from both models are similar. A similar to 1.6 yr post-seismic deformation time-series starting 20 d after the main shock reveals localized deformation at the southern part of the fault. Inversions for afterslip indicate three localized slip patches, and the cumulative afterslip moment after 615 d is at least similar to 11 per cent of the coseismic moment. The afterslip patches are distributed at different depths along the fault, showing no obvious systematic depth-dependence. The deeper of the three patches, however, shows a slight tendency to migrate to greater depth over time. No linear correlation is found for the temporal evolution of afterslip and aftershocks. Finally, modelling of viscoelastic relaxation in a Maxwell half-space yields a lower bound of 1 x 10(18) Pa s on the viscosity of the mid/lower crust. This is consistent with viscosity estimates in other studies of post-seismic deformation across the Tibetan Plateau.

S U M M A R Y The 2008 Mw 6.3 Damxung earthquake on the Tibetan Plateau is investigated to (i) derive a coseismic slip model in a layered elastic Earth; (ii) reveal the relationship between coseismic slip, afterslip and aftershocks and (iii) place a lower bound on mid/lower crustal viscosity. The fault parameters and coseismic slip model were derived by inversion of Envisat InSAR data. We developed an improved non-linear inversion scheme to find an optimal rupture geometry and slip distribution on a fault in a layered elastic crust. Although the InSAR data for this event cannot distinguish between homogeneous and layered crustal models, the maximum slip of the latter model is smaller and deeper, while the moment release calculated from both models are similar. A ∼1.6 yr post-seismic deformation time-series starting 20 d after the main shock reveals localized deformation at the southern part of the fault. Inversions for afterslip indicate three localized slip patches, and the cumulative afterslip moment after 615 d is at least ∼11 per cent of the coseismic moment. The afterslip patches are distributed at different depths along the fault, showing no obvious systematic depth-dependence. The deeper of the three patches, however, shows a slight tendency to migrate to greater depth over time. No linear correlation is found for the temporal evolution of afterslip and aftershocks. Finally, modelling of viscoelastic relaxation in a Maxwell half-space yields a lower bound of 1 × 1018 Pa s on the viscosity of the mid/lower crust. This is consistent with viscosity estimates in other studies of post-seismic deformation across the Tibetan Plateau.

An important tool for understanding deformation occurring within a subduction zone is the measurement of seismic anisotropy through observations of shear wave splitting (SWS). In Sumatra, two temporary seismic networks were deployed between December 2007 and February 2009, covering the fore arc between the fore‐arc islands to the back arc. We use SKS and local SWS measurements to determine the type, amount, and location of anisotropy. Local SWS measurements from the fore‐arc islands exhibit trench‐parallel fast directions which can be attributed to shape preferred orientation of cracks/fractures in the overriding sediments. In the Sumatran Fault region, the predominant fast direction is fault/trench parallel, while in the back‐arc region it is trench perpendicular. The trench‐perpendicular measurements exhibit a positive correlation between delay time and raypath length in the mantle wedge, while the fault‐parallel measurements are similar to the fault‐parallel fast directions observed for two crustal events at the Sumatran Fault. This suggests that there are two layers of anisotropy: one due to entrained flow within the mantle wedge and a second layer within the overriding crust due to the shear strain caused by the Sumatran Fault. SKS splitting results show a NNW‐SSE fast direction with delay times of 0.8–3.0 s. The fast directions are approximately parallel to the absolute plate motion of the subducting Indo‐Australian Plate. The small delay times exhibited by the local SWS (0.05–0.45 s), in combination with the large SKS delay times, suggest that the anisotropy generating the teleseismic SWS is dominated by entrained flow in the asthenosphere below the slab.

A combination of seismic refraction tomography, laboratory ultrasonic velocity measurements, and microstructural observations was used to study the shallow velocity structure of a strand of the San Andreas fault (SAF) just south of Littlerock, California. The examined site has a strongly asymmetric damage structure with respect to the SAF core. The conglomerates to the southwest show little to no damage, whereas a similar to 100m wide damage zone exists to the northeast with a similar to 50m wide zone of pulverized granite adjacent to the fault core. Seismic P-wave velocities of the damaged and pulverized granite were investigated over a range of scales. In situ seismic velocity imaging was performed on three overlapping profiles normal to the SAF with lengths of 350m, 50m, and 25m. In the laboratory, ultrasonic velocities were measured on centimeter- to decimeter-sized samples taken along the in situ profiles. The samples were also investigated microstructurally. Micro-scale fracture damage intensifies with increasing proximity to the fault core, allowing a subdivision of the damage zone into several sections. Laboratory-derived velocities in each section display varying degrees of anisotropy, and combined with microfracture analysis suggest an evolving damage fabric. Pulverized rocks close to the fault exhibit a preferred fault-parallel orientation of microfractures, resulting in the lowest P-wave velocity orientated in fault-perpendicular direction. Closest to the fault, pulverized rocks exhibit a gouge-like fabric that is transitional to the fault core. Comparison of absolute velocities shows a scaling effect from field to laboratory for the intact rocks. A similar scaling effect is absent for the pulverized rocks, suggesting that they are dominated by micro-scale damage. Fault-parallel damage fabrics are consistent with existing models for pulverized-rock generation that predict strong dynamic reductions in fault-normal stress. Our observations provide important constraints for theoretical models and imaging fault damage properties at depth using remote methods.

Knowledge of mantle flow in convergent margins is crucial to unravelling both the contemporary geodynamics and the past evolution of subduction zones. By analysing shear-wave splitting in both teleseismic and local arrivals, we can determine the relative contribution from different parts of the subduction zone to the total observed SKS splitting, providing us with a depth constraint on anisotropy. We use this methodology to determine the location, orientation and strength of seismic anisotropy in the south-central Chile subduction zone. Data come from the TIPTEQ network, deployed on the forearc during 2004–2005. We obtain 110 teleseismic SKS and 116 local good-quality shear-wave splitting measurements. SKS average delay times are 1.3 s and local S delay times are only 0.2 s. Weak shear-wave splitting from local phases is consistent with a shape preferred orientation (SPO) source in the upper crust. We infer that the bulk of shear-wave splitting is sourced either within or below the subducting Nazca slab. SKS splitting measurements exhibit an average north-easterly fast direction, with a strong degree of variation. Further investigation suggests a relationship between the measurement's fast direction and the incoming ray's back-azimuth. Finite-element geodynamic modelling is used to investigate the strain rate field and predicted LPO characteristics in the subduction zone. These models highlight a thick region of high strain rate and strong S-wave anisotropy, with plunging olivine a-axes, in the sub-slab asthenosphere. We forward model the sub-slab sourced splitting with a strongly anisotropic layer of thick asthenosphere, comprising an olivine a-axis oriented parallel to the direction of subduction. The subducting lithosphere is not thick enough to cause 1.2 s of splitting, therefore our results and subsequent models show that the Nazca slab is entraining the underlying asthenosphere; its flow causes it to be strongly anisotropic. Our observation has important implications for the controlling factors on sub-slab mantle flow and the movement of asthenospheric material within the Earth.

In recent years, an increasing number of studies have focussed on resolving the internal structure of ocean island volcanoes. Traditionally, active source seismic experiments have been used to image the volcano edifice. Here we present results using the analysis of compressional to shear (P to S) converted seismic phases from teleseismic events, recorded by stations involved in an active source experiment “TOM-TEIDEVS” (Ibáñez et al., 2008), on the island of Tenerife, Canary Islands. We supplement this data with receiver function (RF) analysis of seismograms from the Canary Islands of Lanzarote and La Palma, applying the extended-time multitaper frequency domain cross-correlation estimation method (Helffrich, 2006). We use the neighbourhood inversion approach of Sambridge, 1999a, Sambridge, 1999b to model the RFs and our results indicate magmatic underplating exists beneath all three islands, ranging from 2 to 8km, but showing no clear correlation with the age of the island. Beneath both La Palma and Tenerife, we find localized low velocity zones (LVZs), which we interpret as due to partial melt, supported by their correlation with the location of historical earthquakes (La Palma) and recent earthquakes (Tenerife). For Lanzarote, we do not sample the most recently volcanically active region and find no evidence for a LVZ. Instead, we find a simple gradational velocity structure, with discontinuities at ∼4, 10 and 18km depth, in line with previous studies.

It has become increasingly important to develop fast and accurate automatic procedures to process and fully exploit increasing large seismic data sets. Traditionally these data sets are processed manually, which requires significant amounts of both manpower and time with sometimes-variable results. We have developed a cost minimization approach to train three automatic pickers: an Sta/Lta, Tpd and the PAI-K picker at each station within a dense temporary network located in northern Chile and southern Bolivia. The optimum picking parameters for each station show regional variability and need to be adjusted individually to achieve the best performance. We developed a weighting scheme that uses four independent predictors of weight calibrated using a handpicked data subset, which mimics the picking by an expert seismologist. We use the fact that each of the three pickers highlights different properties of the observed seismic trace to combine two pickers that work in tandem. The first makes an initial pick before the second picker refines and improves the accuracy of the automatic pick. We find the tandem pickers improve the accuracy of the automatic picks when compared to the single automatic pickers. We demonstrate that following the cost minimization procedure described here the automatic picks have sufficient accuracy that they would be suitable for high-precision earthquake location, focal mechanism determination or high-resolution seismic tomography.