Ocean Bottom Seismograph Array Research Articles

Deep learning for deep earthquakes: insights from OBS observations of the Tonga subduction zone

SUMMARY Applications of machine learning in seismology have greatly improved our capability of detecting earthquakes in large seismic data archives. Most of these efforts have been focused on continental shallow earthquakes, but here we introduce an integrated deep-learning-based workflow to detect deep earthquakes recorded by a temporary array of ocean-bottom seismographs (OBSs) and land-based stations in the Tonga subduction zone. We develop a new phase picker, PhaseNet-TF, to detect and pick P- and S-wave arrivals in the time–frequency domain. The frequency-domain information is critical for analysing OBS data, particularly the horizontal components, because they are contaminated by signals of ocean-bottom currents and other noise sources in certain frequency bands. PhaseNet-TF shows a much better performance in picking S waves at OBSs and land stations compared to its predecessor PhaseNet. The predicted phases are associated using an improved Gaussian Mixture Model Associator GaMMA-1D and then relocated with a double-difference package teletomoDD. We further enhance the model performance with a semi-supervised learning approach by iteratively refining labelled data and retraining PhaseNet-TF. This approach effectively suppresses false picks and significantly improves the detection of small earthquakes. The new catalogue of Tonga deep earthquakes contains more than 10 times more events compared to the reference catalogue that was analysed manually. This deep-learning-enhanced catalogue reveals Tonga seismicity in unprecedented detail, and better defines the lateral extent of the double-seismic zone at intermediate depths and the location of four large deep-focus earthquakes relative to background seismicity. It also offers new potential for deciphering deep earthquake mechanisms, refining tomographic models, and understanding of subduction processes.

Geometry and segmentation of the North Anatolian Fault beneath the Marmara Sea, Turkey, deduced from long‐term ocean bottom seismographic observations

AbstractBoth the geometry and the depth of the seismogenic zone of the North Anatolian Fault under the Marmara Sea (the Main Marmara Fault (MMF)) are poorly understood, in part because of the fault's undersea location. We recorded 10 months of microseismic data with a dense array of ocean bottom seismographs and then applied double‐difference relocation and 3‐D tomographic modeling to obtain precise hypocenters on the MMF beneath the central and western Marmara Sea. The hypocenters show distinct lateral changes along the MMF: (1) both the upper and lower crust beneath the Western High are seismically active and the maximum focal depth reaches 26 km; (2) seismic events are confined to the upper crust beneath the region extending from the eastern part of the Central Basin to the Kumburgaz Basin; and (3) the magnitude and direction of dip of the main fault change under the Central Basin, where there is also an abrupt change in the depth of the lower limit of the seismogenic zone. We attribute this change to a segment boundary of the MMF. Our data show that the upper limit of the seismogenic zone corresponds to sedimentary basement. We also identified several seismically inactive regions within the upper crust along the MMF; their spatial extent beneath the Kumburgaz Basin is greater than beneath the Western High. From the comparison with seafloor extensometer data, we consider that these regions might indicate zones of strong coupling that are accumulating stress for release during future large earthquakes.

Active tectonics in the Gulf of California and seismicity (M > 3.0) for the period 2002–2014

We present a catalog of accurate epicenter coordinates of earthquakes located in the Gulf of California (GoC) in the period 2002–2014 that permits us to analyze the seismotectonics and to estimate the depth of the seismogenic zone of this region. For the period April 2002 to December 2014 we use body-wave arrival times from regional stations of the Broadband Seismological Network of the GoC (RESBAN) operated by CICESE to improve hypocenter locations reported by global catalogs. For the northern region of the GoC (30°N–32°N) we added relocated events from the 2011-Hauksson-Yang-Shearer, Waveform Relocated Earthquake Catalog for Southern California (Hauksson et al., 2012; Lin et al., 2007). From October 2005 to October 2006 we incorporated hypcenters located by Sumy et al. (2013) in the southern GoC combining an array of ocean-bottom seismographs, of the SCOOBA experiment, with onshore stations of the NARS-Baja array. This well constrained catalog of seismicity highlights zones of active tectonics and seismic deformation within the North America-Pacific plate boundary. We estimate that the minimum magnitude of completeness of this catalog is Mc=3.3±0.1 and the b=0.92±0.04 value of the Gutenberg-Richter relation. We find that most earthquakes in the southern GoC are generated by transform faults and this region is more active than the central GoC region. However, the northern region, where most deformation is generated by oblique faults is as active as the southern region. We used the ISC catalog to evaluate the size distribution of seismicity of these regions, and the b value of the Gutenberg-Richter relation and found that b is slightly lower in the central GoC (b=0.86±0.02) compared to the northern (b=1.14±0.04) and the southern (b=1.11±0.04) regions. We observed seismicity that occurs in the Stable Central Peninsular Province, despite the fact that significant active deformation has not been identified in this region.

High-resolution seismic constraints on flow dynamics in the oceanic asthenosphere

Convective flow in the mantle and the motions of tectonic plates produce deformation of Earth's interior, and the rock fabric produced by this deformation can be discerned using the anisotropy of the seismic wave speed. This deformation is commonly inferred close to lithospheric boundaries beneath the ocean in the uppermost mantle, including near seafloor-spreading centres as new plates are formed via corner flow, and within a weak asthenosphere that lubricates large-scale plate-driven flow and accommodates smaller scale convection. Seismic models of oceanic upper mantle differ as to the relative importance of these deformation processes: seafloor spreading fabric is very strong just beneath the crust-mantle boundary (the Mohorovičić discontinuity, or Moho) at relatively local scales, but at the global and ocean-basin scales, oceanic lithosphere typically appears weakly anisotropic when compared to the asthenosphere. Here we use Rayleigh waves, recorded across an ocean-bottom seismograph array in the central Pacific Ocean (the NoMelt Experiment), to provide unique localized constraints on seismic anisotropy within the oceanic lithosphere-asthenosphere system in the middle of a plate. We find that azimuthal anisotropy is strongest within the high-seismic-velocity lid, with the fast direction coincident with seafloor spreading. A minimum in the magnitude of azimuthal anisotropy occurs within the middle of the seismic low-velocity zone, and then increases with depth below the weakest portion of the asthenosphere. At no depth does the fast direction correlate with the apparent plate motion. Our results suggest that the highest strain deformation in the shallow oceanic mantle occurs during corner flow at the ridge axis, and via pressure-driven or buoyancy-driven flow within the asthenosphere. Shear associated with motion of the plate over the underlying asthenosphere, if present, is weak compared to these other processes.

Passive-source ocean bottom seismograph (OBS) array experiment in South China Sea and data quality analyses

Long-term passive source ocean bottom seismograph (OBS) observatory is challenging due to various technical difficulties. In order to gain experience in this field, and to reveal the lithospheric structure beneath the extinct ridge in the central South China Sea (SCS), we carried out a passive source OBS array experiment, which includes 18 OBSs, in the deep portion of SCS. Here we present the instrumentation, the OBS deployment and recovery of this experiment, and more importantly, the data quality evaluated by a number of approaches. Through processing and inspecting waveforms from global, regional and local earthquakes, we find that most of recovered OBSs have good data quality with discernible main phases. The ambient noise analyses of OBS recordings show that their noise is higher than the global average, and the horizontal component is noisier than the vertical, indicating current impacts on horizontal components are more severe. In the period range of 5–10 s, there is a noise notch for the SCS OBSs, and noise levels of horizontal components are comparable to the vertical. This feature, which is not seen at OBS stations in open ocean, suggests the distant sources for double frequency microseism in this marginal sea are not significant. In addition, we successfully determined the orientations for 7 OBSs by investigating their Rayleigh wave polarizations; and we demonstrated the dispersion feature of Rayleigh waves through the frequency-time analysis. Finally, we summarized lessons learned from this experiment regarding the passive source OBS investigations in SCS.

Depth of the 660‐km discontinuity near the Mariana slab from an array of ocean bottom seismographs

High frequency records of deep Mariana earthquakes from a dense array of ocean bottom seismographs deployed in the Mariana arc and back‐arc regions are stacked and searched for the phases P660p and S660p to constrain the depth of the 660‐km discontinuity near the Mariana slab. Results of the high‐resolution study suggest that around 18°N the 660‐km discontinuity lies at about 710–730 km (±14 km) depth within or in the vicinity of the slab core. In the region seismicity ceases at ∼620 km depth. This implies that, although tomographic images show the Mariana slab penetrating into the lower mantle, deep seismicity in the region terminates ∼100 km above the base of the transition zone. These findings and similar observations in Tonga argue that factors other than the phase transition at the base of the upper mantle may control the maximum down‐dip extent of the deep seismogenic region in the slab.

Breathing of the seafloor: Tidal correlations of seismicity at Axial volcano

Tidal effects on seafloor microearthquakes have been postulated, but the search has been hindered by a lack of continuous long-term data sets. Making this observation is further complicated by the need to distinguish between Earth and ocean tidal influences on the seafloor. In the summer of 1994, a small ocean-bottom seismograph array located 402 microseismic events, over a period of two months, on the summit caldera of Axial volcano on the Juan de Fuca Ridge. Harmonic tremor was also observed on all instruments, and Earth and ocean tides were recorded on tiltmeters installed within the seismometer packages. Microearthquakes show a strong correlation with tidal lows, suggesting that faulting is occurring preferentially when ocean loading is at a minimum. The harmonic tremor, interpreted as the movement of superheated fluid in cracks, also has a tidal periodicity.

Detailed subduction structure across the eastern Nankai Trough obtained from ocean bottom seismographic profiles

To investigate the deep crustal structure of the Philippine Sea Plate at its northern margin, we performed a seismic refraction and wide‐angle reflection survey around the eastern Nankai Trough by using an ocean bottom seismograph array with an air gun source. We derived a crustal structure model across the trough, from the Shikoku Basin to the continental slope, that explains not only our seismic data but also previously published gravity data. The origin of the Zenisu Ridge, a conspicuous topographic high along the oceanward slope of the Nankai Trough, is still a matter of argument. There has been some controversy as to whether the igneous activity of the Izu‐Ogasawara Arc or the seafloor spreading of the Shikoku Basin is responsible for the formation of the ridge. Although the crustal thickness beneath the ridge is between 8 and 11 km, slightly thicker than typical oceanic crust, our structure model clearly indicates that the Zenisu Ridge has an oceanic crust and its structure is very similar to that of the Shikoku Basin. Beneath the south flank of the Zenisu Ridge, the Moho shows an offset of 5 km in depth. This offset may correlate with the recently proposed nascent subduction boundary and subduction‐collision tectonics of this area. The velocity structure beneath the continental slope appears characteristic for a well‐developed accretionary wedge bounded by the continental upper crust of the Japan Island Arc to the northwest, and subducting oceanic crust which can be traced beneath the accretionary wedge and the continental upper crust.

Tectonic setting of the Azores Plateau deduced from a OBS survey

The studies of Azores seismicity generally show shocks with either normal faulting or right-lateral strike-slip along the ESE direction, compatible with a eastward relative motion of the Eurasian (EU) relative to the African (AF) plate. However, the 1 January 1980 earthquake was interpreted as a clear left-lateral strike-slip shock along the N150E direction. This pattern is difficult to explain in terms of the relative motion between the EU, AF and North American (NA) plates: all available models for the present day movement of this triple junction fail to explain the regional variability in the stress conditions of the area. Here we present data from a 34-day long Ocean Bottom Seismograph array deployment. We show that the seismicity is distributed along a band aligned with the island chain itself, and is concentrated along several faults with an approximate N150E strike, cutting the Azores plateau in all the area covered by the OBS network. The combination of these new results with other geophysical data permits us to conclude that the tectonic setting of the Azores plateau is characterised by the existence of two sets of faults, in the N120E and N150E directions, defining several crustal blocks, whose relative motion accommodates the interaction of the three megaplates. The deformation of these tectonic blocks is probably driven by the shear between the EU and AF plates. This model explains well the spatial variability of the stress conditions in the Azores domain, the combination of dextral and sinistral strike slip mechanisms and the observed seismotectonics of the Azores islands.

Interplate seismic activity near the northern Japan Trench deduced from ocean bottom and land-based seismic observations

The off-Sanriku area, the landward slope area of the northern Japan Trench, is characterized by its active seismicity which is directly related to the subduction process of the Pacific plate. In July–August 1992, a seismic swarm activity was observed near the trench. Five earthquakes of M > 6 occurred successively within 2 weeks at the beginning of the activity. We deployed an ocean bottom seismograph (OBS) array in the epicentral region to make an in situ observation of the spatial distribution of these events and to study the characteristics of the swarm activity in detail. We relocated 357 events from 23 July to 26 August with focal depth errors less than 5 km by using both the OBS and the land network data. The swarm events are not distributed homogeneously over the plate boundary but form several clusters. A remarkable concentration of the earthquakes was found in the vicinity of the hypocenter of the largest event during the OBS observation ( M 6.2, 29 July); this formed the most active cluster. Almost all the earthquakes, including the M 6.2 event, occurred at depths of about 15 km and were concentrated along a thin plane dipping to the west at 5°. The location of the plane coincides well with that of the plate boundary derived from previous seismic explorations. We also relocated epicenters of the off-Sanriku events by introducing station corrections to the land network data during the period when the OBS array was absent. The corrections were estimated from the travel time residuals of the hypocenter relocation using the OBS and the land network data. The result of the epicenter relocation strongly indicates the existence of a zone of extremely low seismic activity beneath the inner trench wall, where low seismic velocity, mechanically weak, unconsolidated sediments overly the subducting Pacific plate.

Seismic anisotropy in the mantle beneath an oceanic spreading centre

BENEATH an active mid-ocean ridge, the mantle upwells in response to the divergence of the newly formed plates, leading to high temperatures and pressure-release melting below the ridge axis. The width of the upwelling region and the amount of melting depend on mantle rheology1–5, but all models predict a maximum decrease in seismic velocity at the ridge axis. It has also been suggested, however, that the alignment of anisotropic minerals by shear in the upwelling mantle will increase seismic velocity for rays travelling subvertically through the upwelling zone6,7. Here we report the observation of a consistent pattern of anomalously early P-wave arrival times at an array of ocean-bottom seismographs deployed across the axis of the southern Mid-Atlantic Ridge: P-waves from distant earthquakes arrive earlier at stations near the axis than at those further away. Our results are consistent with a model of anisotropy in which the degree of mineral alignment is greatest directly beneath the ridge axis, and significant anisotropy extends tens of kilometres from the axis.

Coherence lengths of seafloor noise: Effect of ocean bottom structure

Results are reported from an experiment, conducted in 1987, in which an ocean bottom seismograph array of 150-m aperture and minimally redundant design was used to record the ambient noise in deep water off the California coast. The minimum interelement spacing among the nine instruments was 8 m. The coherence lengths observed imply that the noise field in the 0.05- to 5-Hz band are fundamental and higher-mode Rayleigh/Stoneley/Scholte waves and the relative amplitudes of the modes imply that the excitation occurs within 20 km of the array. These observations imply that the noise energy is scattered into the seafloor waveguide at the boundaries of the sediment pond in which the array was sited. The implications for sub-bottom sensors are discussed.

Seismic tomography of Jasper Seamount

A vertical section of the interior structure of Jasper Seamount was modeled using a spectral tomographic inversion of P wave travel times. An array of ocean bottom seismographs (OBSs) deployed over the seamount detected the arrivals from a series of ocean bottom shots. A reference velocity model reveals that average compressional velocities within the seamount are similar to those found within Kilauea and are consistently slower than velocities at equivalent depths in typical oceanic crust. This suggests Jasper Seamount has a high average porosity. Perturbations from the reference model were imaged by tomographic inversion. A high velocity zone within the northwest flank of the seamount may result from dikes associated with a radial rift or from a shallow solidified magma reservoir. A low velocity summit may result from shallow, explosive eruptions. The tomographic model is consistent with the results of gravity, magnetic and dredging analyses.

A seismicity gap beneath the inner wall of the japan trench as derived by ocean bottom seismograph measurement

An ocean bottom seismograph (OBS) array covered the area of the Japan Trench about 40°N off the Tohoku District of northeastern Honshu, Japan. The OBS array was operated for two weeks in June 1981. Arrival time data of 8 OBSs were used to study a spatial distribution of microearthquakes around the trench. During the observation period 400–500 events with a signal-to-noise ratio greater than 2 were detected by each OBS. The arrival times of 223 of these events could be picked. 38 earthquakes were located in the area of the Japan Trench within the latitude range of 38.5°–41.0°N and the longitude range of 142.0°–146.0°E, and about a hundred events were located beneath or near Honshu Island.S-P time distribution recorded by each OBS showed seismic activities to exist on both sides of the trench axis, while the seismic level was low beneath the inner wall of the trench. Epicentral distribution showed a seismicity gap beneath the inner wall of the trench: the area of the landward activity continued eastward up to the position where the water depth is about 4000 m. Seismic activity existed beneath the trench axis and the seaward wall of the trench. The seismicity gap suggested from the 1980 OBS observations in the vicinity of the present study area was clearly confirmed by the present observations.The OBS array enabled us to determine the focal depth of microearthquake activity with better resolution than was possible from land observations. The OBS data showed that the landward activity beneath the continental slope was concentrated in the very shallow part of the lithosphere at a depth less than 30 km. Foci started to deepen westward with a very low dip angle of less than 10°, which was the most shallow part of the upper plane of the double seismic zone.The focal depth distribution showed that events just beneath the trench axis occurred to depths slightly greater than 50 km, while the foci of the seaward activity were concentrated in the uppermost lithosphere at depths less than 30 km. Seismic cross-sections clearly show that the activity beneath the trench axis and seaward wall of the trench is separated from the landward activity, which is a shallow extension of the double seismic zone, by the seismicity gap.

Recent faulting and microearthquakes at the intersection of the Vema Fracture Zone and the Mid‐Atlantic Ridge

New air gun reflection profiles, 3.5‐kHz reflection profiles, and microearthquake data recorded by an array of ocean bottom seismographs form the basis for this study of the transition from a spreading center to a major transform fault. Disturbances of the thick, normally flat‐lying, turbidite deposits provide indications of recent vertical motions. At the western intersection of the fracture zone with the median valley there is a depression in the sediments that represents the southerly extension of the median valley into the fracture zone valley. The depression is terminated abruptly on the south by the active transform fault, which acts as a locus for vertical as well as horizontal displacement. Flat‐lying, undisturbed sediments terminate abruptly at the fault. The western boundary of the depression is much broader and is characterized by a series of slumplike steps. To the west, there is little or no evidence for uplift or tilting of sediments which might indicate vertical recovery of the crust as it spreads away from the depression. This suggests that uplift and recovery out of the depression is episodic in nature and has been inactive over the last million years along the western boundary. To the east there is clear evidence of uplift and tilting of sedimentary layers. A basement ridge emerging from the sediments is currently being uplifted and rotated in a manner analogous to processes responsible for the creation and cancellation of median valley relief. The transition between the spreading center and the transform fault appears to take place within 1–2 km. The width of the transform fault just east of the depression is less than a kilometer. Microearthquakes were located and displayed by new methods that directly account for nonlinearities associated with small arrays. Microearthquakes located by three or more ocean bottom seismometers show that the greatest seismic activity occurs along the eastern walls of the median valley, at the basement ridge, in the eastern portion of the depression and in the crestal mountains. Very little activity is associated with the western edge of the transform depression and the trace of the transform fault.

Separation of high‐frequency PN phases and mantle refracted P phases at distances between 6° and 18° in the western Pacific by ocean bottom seismograph array

Observations of high‐frequency (HF) Pn phases by an oceanbottom seismometer (OBS) array were made at distances between 6° and 18° in the north‐western Pacific. At this distance range, mantle‐refracted P phases arrive earlier than HF‐Pn phases. These two phases are successfully separated for near surface earthquakes by applying appropriate filtering, since the dominant frequencies of HF‐Pn phases are higher than 6 Hz and the frequencies of mantle‐refracted P phases are rather low, about 3 Hz. The apparent velocities of HF‐Pn phases in this distance range are estimated to be from 8.3 to 8.1 km/sec or even lower, while those of mantle‐refracted P waves are from 8.4 to 8.6 km/sec. The reported value of 8.33 km/sec for Hf‐Pn phases appears to be an average of both HF‐Pn and mantle‐refracted P phases. Therefore, these observations suggest that a proposed additional HF‐Pn wave guide for the higher velocity of 8.33 km/sec is questionable.

Spatial distribution of microearthquakes beneath the Japan Trench from ocean bottom seismographic observations

The area of the Japan Trench, about 40°N off the Tôhoku District of north-eastern Honshu, Japan, was for the first time extensively covered by an ocean bottom seismograph (OBS) array. Seven OBSs were available for analyses of natural earthquakes and the observation period lasted two weeks during 1980 July and August. During this period, each OBS registered between 300 and 600 earthquakes. About 100 events were simultaneously recorded by at least three OBSs. Hypocentres were calculated by using a generalized inversion technique. Sixty-five events, mostly with magnitudes (M) from 1.5 to 3.0, which were not located by a land network, were determined to be within or near the area covered by the OBS array. The OBS array had nearly uniform detectability for earthquakes with M over 2.0 within the latitude range of 39.5° — 41.5°N and the longitude range of 143° — 146.5°E. Within this area there is a highly active region of microseismicity beneath the seaward wall of the Japan Trench. The seaward activity is concentrated within 100 km of the trench axis and is low beneath the Northwest Pacific Basin. A seismicity gap was found just westward of the trench axis, beneath the landward wall of the trench at a water depth of between 6000 and 4000 m. Seismic activity continues landward from the region west of the gap. The OBS array enabled us to determine the focal depth of microearthquake activity with better resolution than was previously possible. The results show that the microearthquake activity is concentrated within the upper part of the oceanic lithosphere. On the seaward side, most events occur at a depth of less than 30 km beneath the trench wall, while the activity on the landward side starts to deepen westward exactly at the trench axis. This paper reports the first direct observations to clarify the spatial distribution of microearthquakes associated with the bending and/or subducting of the oceanic plate just beneath the Japan Trench axis.

An earthquake sequence studied with ocean-bottom seismographs

Summary. A tripartite ocean-bottom seismograph array at the junction of the East Pacific Rise and Rivera Fracture Zone recorded an eathquake sequence, consisting of three main shocks (mB= 4.3, 4.3 and 4.8) and numerous aftershocks from the fracture zone, in the distance range 35–50 km. Delineation of the rupture zones by aftershocks indicates that the first two main shocks took place on overlapping fault areas, while the third occurred over a fault area separated from the first by several kilometres. Both rupture zones were about 4 km long. Surface wave spectra indicate a shallow (about 3 km below the sea floor) source, as does OBS array phase velocity data. The seismic moments, obtained from teleseismic surface wave data, of 1.3, 2.1 and 2.8 × 1023 dyn cm, with the fault areas as delineated by aftershocks, imply a stress drop of about 8 bars for the main shocks. Aftershock sequences of each of the main shocks are similar, with a b-value of about 0.65. Teleseismic P travel times are similar to those from near-surface sources in Nevada.

Ocean-bottom seismograph observations on the Mid-Atlantic Ridge near 45 N

Summary. This paper describes the results obtained from the deployment of an array of ocean-bottom seismographs on the Mid-Atlantic Ridge near 45° N. Seismicity is restricted to a zone only 5 km wide near the centre of the Median Valley. The large scatter of epicentres determined using teleseismic data can be reconciled with this zone. The observed activity includes aftershock and swarm-like sequences but in terms of b value these, and teleseismically observed events, belong to the same population. The results are in agreement with a caldera collapse model of the Median Valley with movement restricted to the innermost faults.