Arrival Traveltime Tomography Research Articles

Weighted Adjoint-state Crosshole Traveltime Tomography

The adjoint-state method of first arrival traveltime tomography, featured with implicit ray tracing calculation, was used to compute the gradient of the misfit function without introducing Fréchet derivatives. However, the majority of velocity gradients in the adjoint-state method focus on the high velocity area due to high raypath density. Thus, the current adjoint-state method faces challenges of imaging low velocity anomalies in the near-surface by crosshole seismic acquisition. In this paper, we proposed a weighted adjoint-state Eikonal equation-based crosshole traveltime tomography method. A preconditioning operator is provided to fix the ray path density influence. The weighted method was tested on a homogeneous model, a checkerboard model, and a reduced-scale Marmousi model. Both misfit cures and velocity logs are used to make a comparison between the weighted method and the traditional method. The results show the weighted method has better resolution, faster convergence speed, and lower final residual, which demonstrate the method can mitigate the ray path effect and reduce the inverted velocity error of low velocity anomalies.

New insights on the origin of Barra volcanic ridge system, offshore Ireland: A long-distance influence of the Iceland mantle plume

The Iceland Plume has significantly impacted the North Atlantic region. Igneous edifices and large seamounts in the Rockall region were linked to Iceland plume pulsations during the Late Cretaceous to Mid-Eocene. In the south of the Rockall Basin a chain of northwest-southeast trending volcanic ridges known as the Barra Volcanic Ridge System (BVRS) have been identified on seismic and magnetic datasets. However, the timing, morphology, extent, and emplacement mechanism of the BVRS are not well understood. To investigate the volcanic ridge system, we analyze a 360-km-long pre-stack time migrated seismic section along the margin, which covers the southern extent of BVRS. The high velocity (4.5–4.8 km/s) obtained by using first arrival travel time tomography of downward continued data set, high magnetic anomaly and the typical morphology of the ridges suggest that they could be volcanoes of basaltic compositions. The integration of gravity and seismic data constrain the crustal structure and thickness, which indicates that the crust could be as thin as 4 km beneath the BVRS, requiring a maximum stretching factor >6. Folded compressional structures and lava flows within the Early Paleocene to Mid-Eocene sediments draping the volcanic ridges, suggest that these volcanic intrusions possibly have occurred post Early Paleocene age, which is approximately the same time-period when Iceland mantle plume arrived in the Rockall region. These results suggest that the BVRS developed because of pre-existing lithospheric configuration in the Rockall Basin and could be one of the southernmost volcanic provinces of the North Atlantic Igneous Province (NAIP).

Supervirtual Refraction Interferometry in the Radon Domain

Accurate picking of seismic first arrivals is very important for first arrival travel time tomography, but the first arrivals appearing at far offsets are often more difficult to pick accurately due to the low signal-to-noise ratio (SNR). The conventional supervirtual refraction interferometry (SVI) method can improve the SNR of first arrivals to a certain extent; however, it is not suitable for seismic data that interfered by strong noise. In order to better process the first arrivals at far offsets with serious noise interference, we propose a modified method, in which SVI implemented in the Radon domain (RDSVI) due to the cross-correlation in the Radon domain have a better effect. According to the kinematic characteristics of first arrival refractions, SVI is performed in the linear Radon domain. Both synthetic data and field data demonstrate the proposed method can enhance the effective signal and attenuate the strong noise simultaneously, so as to significantly improve the SNR of the first arrival data. Meanwhile, the RDSVI method is tested on the first arrival data with missing traces, which proves that this method can overcome the influence of abnormal traces and is suitable for the reconstruction of sparsely sampled seismic data.

Seismic Imaging of Dante's Domes Oceanic Core Complex From Streamer Waveform Inversion and Reverse Time Migration

AbstractEarly arrival traveltime tomography and full waveform inversion were conducted on downward continued streamer seismic data at Dante's Domes oceanic core complex (OCC), providing unprecedented details of shallow P wave velocity structure. Together with reverse time migration images, seafloor morphology, in situ geological samples, magnetic and gravity data, the seismic constraints are used to infer the lithological distribution along the seismic profiles. Based on the striking similarity in velocity structure beneath the corrugated domes with other OCCs and drilling results from Atlantis Massif, we confidently reconfirmed the Southern Dome as dominantly gabbroic rocks, and the Northern Dome as serpentinized peridotites. A series of isolated gabbroic bodies embedded in the diabase and basaltic layers is observed in the breakaway zone, suggesting that the initiation of Dante's Domes OCC occurred over a long period during which there were several failed attempts to form a long‐lived detachment fault. This early development of the OCC probably occurred under a regime of alternating magma starvation and magma replenishment. The predominantly gabbroic section, beneath the Southern Dome and extending to termination, indicates the OCC has been created with relatively high magma flux. We also imaged distinct shallow subseafloor reflections which are also termed as D reflectors underneath the corrugated domes. The location of the D reflectors is similar to those in the Atlantis Massif, with depths well correlated with the top of exhumed gabbroic bodies and the discontinuities in the D reflectors between gabbroic bodies. Our findings contribute to the understanding of processes controlling the OCCs initiation and evolution at slow spreading ridges.

Wavefield reconstruction inversion with weak-total-variation constraint based on first arrival traveltime tomography

Wavefield reconstruction inversion with weak-total-variation constraint based on first arrival traveltime tomography

Potential of Surface-to-Tunnel Seismic Tomography to Detect Vertical Faults: Application to the Tournemire Underground Research Laboratory, France

The Tournemire Underground Research Laboratory (Aveyron, France), developed by the French Institute for Radiological Protection and Nuclear Safety, is composed of a tunnel excavated in an argillaceous geological layer belonging to a limestone–argillite–limestone sub-horizontal sedimentary sequence. Sub-vertical fault zones are intercepted by drifts and boreholes excavated from the tunnel in the argillaceous layer at a depth of about 250 m. These fault zones are characterized by velocities lower than those of the embedding rocks. In this study, we assess the capacity of first arrival traveltime tomography to detect and characterize such fault zones in a surface-to-depth transmission configuration with a limited angle aperture. The source at the surface is a sledgehammer and turns out to be very repetitive. The source trigger is transmitted to the acquisition bay via a VHF transmitter. Receivers are located at depth in underground works. First arrivals are clear and traveltime picking is straightforward. Before inverting real data, we conduct synthetic tests with the real acquisition geometry to assess limits and validity of first arrival traveltime tomography to detect a lower velocity fault zone using different initial configurations and inversion constraints during the inversion. We show that introducing a priori information on velocities in the shallow subsurface, limestones and argillaceous layers and constraints on maximum velocities greatly helps improving results and subsequent interpretation. Velocities obtained from the real data display a main lower velocity zone that extends from the galleries in the argillaceous layer right up to the surface in limestones. The location and extension of this lower velocity zone are consistent with fault and fracture zones mapped from geological observations.

First arrivalQtomography based on an adjoint-state method

AbstractUnder the assumption of invariant ray path in a weakly dissipative (high quality factor Q) subsurface medium, a tomographic inversion approach composed of two cascading applications of first arrival traveltime and Q tomography is proposed for compensating amplitude loss caused by near-surface anomalies, such as unconsolidated soils or the overburden gas cloud. To improve the computational efficiency, these two related tomography methods were adopted with an adjoint-state technique. First, arrival traveltime tomography will be performed to provide an inverted velocity model as one of the inputs for the following first arrival Q tomography. Then, the synthetic first break generated by the inverted velocity model will be used as a stable guidance of accessing the scopes of first arrival waveforms in the time domain where the potential attenuated time information is contained. The attenuated time will be estimated through a logarithmic spectral ratio linear regression corresponding to frequency-dependent propagation responses of different wave types. All these estimated attenuated times will be applied with reference signals to generate synthetic attenuated seismic data in the time domain, and their discrepancies with real data will be evaluated using similarity coefficients. The ones with larger values will be selected as optimal attenuated time inputs for the following Q tomographic inversion. Examples of both synthetic and field data reveal the feasibility and potential of this method.

Towards large‐scale stochastic refraction tomography: a comparison of three evolutionary algorithms

ABSTRACTThe main goal of this study is to assess the potential of evolutionary algorithms to solve highly non‐linear and multi‐modal tomography problems (such as first arrival traveltime tomography) and their abilities to estimate reliable uncertainties. Classical tomography methods apply derivative‐based optimization algorithms that require the user to determine the value of several parameters (such as regularization level and initial model) prior to the inversion as they strongly affect the final inverted model. In addition, derivative‐based methods only perform a local search dependent on the chosen starting model. Global optimization methods based on Markov chain Monte Carlo that thoroughly sample the model parameter space are theoretically insensitive to the initial model but turn out to be computationally expensive. Evolutionary algorithms are population‐based global optimization methods and are thus intrinsically parallel, allowing these algorithms to fully handle available computer resources. We apply three evolutionary algorithms to solve a refraction traveltime tomography problem, namely the differential evolution, the competitive particle swarm optimization and the covariance matrix adaptation–evolution strategy. We apply these methodologies on a smoothed version of the Marmousi velocity model and compare their performances in terms of optimization and estimates of uncertainty. By performing scalability and statistical analysis over the results obtained with several runs, we assess the benefits and shortcomings of each algorithm.

First arrival traveltime tomography using supervised descent learning technique

First arrival traveltime tomography is a highly non-linear inverse problem. Regularization is often used to constrain and stabilize the model reconstruction. It can be considered as adding prior information into the inversion. But not all prior information can be done this way because some cannot be expressed in rigorous mathematical forms. Moreover, the inversion algorithms based on gradient-descent are often trapped in local minima. All these have a negative impact on the accuracy of reconstruction.In this work, we apply the supervised descent method (SDM) for 2D traveltime tomography. SDM is a machine-learning algorithm that contains two steps. In the training stage, SDM learns a set of descent directions from a training dataset containing prior knowledge. In the prediction stage, it uses both the learned descents and the data residual to accomplish model reconstruction. In this paper, we present the formulations of the training and prediction for SDM, and introduce two methods to solve the regularized objective function in the online prediction stage. The descent learning technique provides a new perspective to combine machine-learning-based inversion and classical gradient-based inversion, and offers a flexible way to incorporate both uncertain prior knowledge and physical modeling process into deterministic inversion. It can help the inversion to skip local minima and achieve fast convergence. The numerical results show that the accuracy and speed of SDM inversion can be enhanced compared with classical gradient-based methods. In addition, the learning ability of this descent learning technique is also validated.

A parallel competitive Particle Swarm Optimization for non-linear first arrival traveltime tomography and uncertainty quantification

Seismic traveltime tomography is an optimization problem that requires large computational efforts. Therefore, linearized techniques are commonly used for their low computational cost. These local optimization methods are likely to get trapped in a local minimum as they critically depend on the initial model. On the other hand, global optimization methods based on MCMC are insensitive to the initial model but turn out to be computationally expensive. Particle Swarm Optimization (PSO) is a rather new global optimization approach with few tuning parameters that has shown excellent convergence rates and is straightforwardly parallelizable, allowing a good distribution of the workload. However, while it can traverse several local minima of the evaluated misfit function, classical implementation of PSO can get trapped in local minima at later iterations as particles inertia dim. We propose a Competitive PSO (CPSO) to help particles to escape from local minima with a simple implementation that improves swarm's diversity. The model space can be sampled by running the optimizer multiple times and by keeping all the models explored by the swarms in the different runs. A traveltime tomography algorithm based on CPSO is successfully applied on a real 3D data set in the context of induced seismicity.

Ensuring very shallow-water sediment properties: case study from Capo Granitola harbour, Sicily (Italy)

We present high-resolution Vp models of the Capo Granitola harbor, Sicily (Italy) obtained by first arrival traveltime tomography. Seismic data were collected along four hydrophone arrays on the sea-bottom and via a Watergun as seismic source, in order to plan dredging operations in the harbor. Using a hydrophone spacing of 2.5 m and shot spacing of 5 m, very high resolution quality data were recorded. Seismic tomography expands existing knowledge of the harbour subsoil with a penetration of about 20 m, illuminating the Lower Pleistocene bedrock (Marsala calcarenites) that corresponds to high-Vp regions (Vp > 4.5 km/s). Low Vp (1.8–4.5 km/s) deposits belonging to terraced calcarenites (Upper Pleistocene in age) are also well imaged; they are about 8 m thick and lie below loose sand deposits (Vp = 1.5 km/s). The substratum has an articulated morphology; Vp images unravel small steps in the basement probably related to structural discontinuities (e.g., faults). Processing data with 3D techniques enables images of the structure and the thickness of the lithotypes to be reconstructed, thus leading to large-scale, realistic estimates of the total quantity of material to be excavated or dredged. Tomographic profiles permit clear discrimination of the soft sediment above the basement and thus allow the determination of the total volume of sediment above the seismic bedrock, estimated at about 265,000 m3.

Automatic early arrival traveltime tomography and its applications

We have developed a method of macrovelocity inversion that does not require explicit picking of either common-image point gathers or first breaks. The method uses head waves, diving waves, and wide-angle reflections in seismic data (collectively early arrival energies) for accurate estimation of velocity and anisotropy parameters. In this method, seismic data are first decomposed into Gaussian packets. Packets associated with early arrival energies are selected and used as input to a tomography solver. The outputs of the solver are velocity and Thomsen’s anisotropy parameters, or any of their combinations. Using information contained in the packets, we can correctly model the early arrival energies (first breaks and/or other refractions). The workflow is fully automatic and can be used in a batch processing environment with minimum human intervention. We have tested the method on synthetic and field data sets. In one synthetic test, we were able to reduce traveltime residuals of diving waves from 400 to 5 ms and recover anisotropic model parameters that are sensitive to early arrival traveltimes. In another synthetic test, we were able to recover a large shallow low-velocity anomaly with a very simple starting velocity model. The first field data set was for a shallow marine seismic data project. We were able to obtain a better shallow velocity model using our method than when using a legacy approach. In the second field data test, we applied our method on a deepwater data set from a dual-coil acquisition, with full-azimuth and long-offset coverage. Our method can correctly model early arrival energies recorded at long offsets and use them in the iterative inversion such that better estimation of velocities and anisotropy parameters in shallow sediments can be achieved. We have tested different starting models for the inversion. We are able to get very similar results, suggesting that our method is not sensitive to the accuracy of a starting model.

Frequency-dependent traveltime tomography using fat rays: application to near-surface seismic imaging

Frequency-dependent traveltime tomography does not rely on the high frequency assumption made in classical ray-based tomography. By incorporating the effects of velocity structures in the first Fresnel volume around the central ray, it offers a more realistic and accurate representation of the actual physics of seismic wave propagation and thus, enhanced imaging of near-surface structures is expected. The objective of this work was to apply frequency-dependent first arrival traveltime tomography to surface seismic data that were acquired for exploration scale and near-surface seismic imaging. We adapted a fat ray tomography algorithm from global-earth seismology that calculates the Fresnel volumes based on source and receiver (adjoint source) traveltime fields. The fat ray tomography algorithm was tested on synthetic model data that mimics the dimensions of two field data sets. The field data sets are presented as two case studies where fat ray tomography was applied for near-surface seismic imaging. The data set of the first case study was recorded for high-resolution near-surface imaging of a Quaternary valley (profile length<1km); the second data set was acquired for hydrocarbon search (profile length>10km). All results of fat ray tomography are compared against the results of classical ray-based tomography. We show that fat ray tomography can provide enhanced tomograms and that it is possible to recover more information on the subsurface when compared to ray tomography. However, model assessment based on the column sum of the Jacobian matrix revealed that especially the deep parts of the structure in the fat ray tomograms might not be adequately covered by fat rays. Furthermore, the performance of the fat ray tomography depends on the chosen input frequency in relation to the scale of the seismic survey. Synthetic data testing revealed that the best results were obtained when the frequency was chosen to correspond to an approximate wavelength-to-target depth ratio of 0.1.

Crustal structures from the Wuyi-Yunkai orogen to the Taiwan orogen: The onshore-offshore wide-angle seismic experiments of the TAIGER and ATSEE projects

Knowledge of the crustal structure is important for understanding the tectonic framework and geological evolution of southeastern China and adjacent areas. In this study, we integrated the datasets from the TAIGER (TAiwan Integrated GEodynamic Research) and ATSEE (Across Taiwan Strait Explosion Experiment) projects to resolve onshore-offshore deep crustal seismic profiles from the Wuyi-Yunkai orogen to the Taiwan orogen in southeastern China. Three seismic profiles were resolved, and the longest profile was 850km. Unlike 2D and 3D first arrival travel-time tomography from previous studies, we used both refracted and reflected phases (Pg, Pn, PcP, and PmP) to model the crustal structures and the crustal reflectors. In total, data from 40 shots, 2 earthquakes, and approximately 1,950 stations were used; 15,612 arrivals were selected among three transects. Using these data, we determined the complex crustal evolution since the Paleozoic era , involving the closed Paleozoic rift basin in central Fujian, the Cenozoic extension due to the South China Sea opening beneath the coastline of southern Fujian, and the on-going collision of the Taiwan orogen. The shape of the Moho, which also reflects the crustal evolution, can be summarized as follows: ~30km deep to the west of Fujian, deepening toward central Fujian (~35km), becoming shallower toward the Taiwan Strait (~28km), deepening again toward the mountain belt of Taiwan (~42km), and becoming shallower toward the Pacific Ocean (~10km).

New Insights from Seismic Imaging Over the Youanmi Terrane, Yilgarn Craton,Western Australia

The general crustal structure of the Youanmi Terrane – a tectonic unit within the Neoarchean Yilgarn Craton – and its relationship to ore deposits are of major interest in Western Australia. To image geological structures, the Geological Survey of Western Australia in collaboration with Geoscience Australia acquired three deep crustal 2D seismic profiles over the region in 2010. In this study, we have reprocessed a 120km long section of one of the seismic profiles, line 10GA-YU1. We reprocessed the data with focus on shallow reflections down to 2 s (about 6km depth). A cross-dip analysis was performed to investigate the dip of the reflections in cross-line direction. The cross-dip analysis showed that there is no significant cross-dip influence, confirming that a 2D interpretation is a valid assumption for this dataset. The seismic image of the reflections was improved after applying dip-moveout correction. Finally, the reflections were migrated to their “true” locations in the subsurface. To better understand the physical properties of the subsurface, we also tried First Arrival Travel-Time Tomography. The tomography results show that low velocities correlate with shear zones and that high velocity zones are common where mineral deposits are located close to the surface.

Successful application of joint reflection/refraction tomographic inversion in a shallow water marine environment

We show how refraction tomography (also called first arrival travel time tomography) helps to produce more accurate and detailed depth velocity models below a shallow seafloor. We do not use refractions by themselves to build a complete shallow velocity model. In our proposed workflow, refraction tomography complements standard reflection tomography and the priority remains with the reflections to guarantee stability of the solution and to avoid uncertainties associated with refracted or diving waves in complex media.We use wave equation modelling to calculate synthetic gathers and estimate the travel time mismatch between real and synthetic first arrivals. It leads to a robust workflow which can be easily introduced into production depth-velocity processing.We show how this joint reflection/refraction velocity inversion works using a real 1000sq.km 3D marine seismic dataset acquired in an area where the water depth varies from 20 m to 1100 m.

Full waveform inversion of seismic reflection data from the Forsmark planned repository for spent nuclear fuel, eastern central Sweden

The Swedish Nuclear Fuel and Waste Management Company (SKB) has been carrying out extensive studies at the planned repository for spent nuclear fuel at the Forsmark site in the eastern part of central Sweden since 2002. Identification of subhorizontal to gently dipping seismic reflections is especially important since these may represent transport routes for radionuclides. Studies have shown that such reflections can be generated by water filled fracture zones that have a lower velocity than the surrounding bedrock. Lithological changes, that is, mafic sills, may also be responsible for reflections in some cases. At the Forsmark site, it is difficult to distinguish fracture zones from mafic sills in the standard reflection seismic processed sections. However, since mafic sills usually have a positive velocity contrast with the background velocity field compared to fractures zones that have a negative one, the two possibilities could be differentiated if we could reconstruct the underground velocity field. Seismic full waveform inversion has the potential to perform this reconstruction, allowing us to discriminate between fractures zones and mafic sills. In this study, we apply a 2-D waveform inversion code on crooked line data sets acquired at the Forsmark site. This implies we are dealing with a 3-D geometry. We handle this problem by applying 3-D to 2-D coordinate projections. First, we perform a synthetic benchmark test with a similar geometry to that of the projected real data. We test both amplitude and phase inversion and phase only inversion on the synthetic data. The results show that the phase only inversion has fewer artefacts and is more stable. After successful application on the synthetic data, we apply the phase only waveform inversion on the real data. The resulting velocity fields show more details compared with the starting model based on first arrival traveltime tomography. Time domain synthetic data sets generated from the final velocity fields show a good match between the early arriving phase of the real data and the synthetic data sets. Checkerboard tests are also applied on the final models for resolution checks and indicate good resolution down to 200 m in the central parts of the profiles. Based on a comparison between the Kirchhoff pre-stack depth migrated sections and the final waveform inversion results we conclude that the reflections investigated most likely are generated at low-velocity fracture zones.

Variable damping constraint tomography and its application in VSP Data

First arrival travel time tomography has achieved wide application. However, tomographic resolution is insufficient because geometry constraints cause rays to be unevenly distributed in the velocity model. The variable damping constraint method adopts uneven priori information to match uneven data distribution which can lessen the correlation between velocity correction values and ray coverage density. In this paper, we combine the variable damping constraint with a smoothness constraint which is added into the regularization equations in velocity inversion to avoid instability caused by only using the variable damping constraint method. The alpha-trimmed-mean filter is used to smooth and denoise intermediate results in the velocity inversion process. We use the LSQR algorithm to enhance the convergence rate and suppress error propagation in solving linear equations. In this paper, we apply the proposed tomographic method to perform velocity inversion using VSP data. The application in recovery test of the checkerboard model and velocity inversion of real VSP data show that the variable damping constraint method can improve tomographic quality because it can solve the effects of uneven ray coverage. In addition, the examples show that the tomographic result near geophones is much more reliable than other areas in the velocity model.

Finding a realistic velocity distribution based on iterating forward modelling and tomographic inversion

Tomography is like a photograph that was taken by a camera with blurred and defective lenses that deform the shapes and colours of objects. Reporting quantitative parameters derived from tomographic inversion is not always adequate because tomographic results are often strongly biased. To quantify the results of tomographic inversion, we propose a forward modelling and tomographic inversion (FM&TI) approach that aims to find a more realistic solution than conventional tomographic inversion. The FM&TI scheme is based on the assumption that if two tomograms derived from the inversion of observed and synthetic data are identical, the synthetic structure may appear to be closer to the real unknown structure in the ground than the inversion result. However, the manual design of the synthetic velocity distribution is usually time-consuming and ambiguous. In this study, we propose an approach that automatically searches for a probabilistic model. In this approach, a synthetic model is iteratively updated while taking into account the bias of the model in previous stages of the FM&TI performance. Here, we present an example of synthetic modelling and real data processing for an active source refraction data set corresponding to a marine profile across the subduction zone in Chile at about 32°S latitude. A key feature of the model is a low-velocity channel above the subducted oceanic crust, which was defined in the synthetic model and expected in the real case. The conventional first arrival traveltime tomography was barely able to resolve this channel. However, after several iterations of the FM&TI modelling, we succeeded in reconstructing this channel clearly. In the paper, we briefly discuss the nature of this low-velocity subduction channel, and we compare the results with other studies.

Three-dimensional seismic refraction tomography of the crustal structure at the ION site on the Ninetyeast Ridge, Indian Ocean

SUMMARY We have performed a 3-D seismic refraction tomography of a 48 × 48 km2 area surrounding ODP site 757, which is planned to host an International Ocean Network (ION) permanent seismological observatory, called the Ninetyeast Ridge Observatory (NERO). The study area is located in the southern part of the Ninetyeast Ridge, the trail left by the Kerguelen hotspot on the Indian plate. The GEOMAR Research Centre for Marine Geosciences and the Federal Institute for Geosciences and Natural Resources acquired 18 wide-angle profiles recorded by 23 ocean bottom hydrophones during cruise SO131 of R/V Sonne in spring 1998. We apply a first arrival traveltime tomography technique using regularized inversion to recover the 3-D velocity structure relative to a 1-D background model that was constructed from a priori information and averaged traveltime data. The final velocity model revealed the crustal structure down to approximately 8 km depth. Resolution tests showed that structures with approximately 6 km horizontal extent can reliably be resolved down to that depth. The survey imaged the extrusive layer of the upper crust of the Ninetyeast Ridge, which varies in thickness between 3 and 4 km. A high-velocity anomaly coinciding with a positive magnetic anomaly represents a volcanic centre from which crust in this area is thought to have formed. A pronounced low-velocity anomaly is located underneath a thick sedimentary cover in a bathymetric depression. However, poor ray coverage of the uppermost kilometre of the crust in this area resulted in smearing of the shallow structure to a larger depth. Tests explicitly including the shallow low-velocity layer confirmed the existence of the deeper structure. The heterogeneity of the upper crust as observed by our study will have consequences for the waveforms of earthquake signals to be recorded by the future seismic observatory.