Search

Tectonic tremors occurring on subducting plate boundaries are known to migrate at various timescales and migration speeds. Spatiotemporal patterns of tremor migration are key to investigating the rupture growth of slow earthquakes. However, spatiotemporal patterns are not sufficiently simple to visually define tremor migrations. This study developed a space‐time Hough transform to objectively extract tremor migrations. The space‐time Hough transform enables the extraction of multiple tremor migrations with various durations, migration directions, and migration speeds. We applied this method to a catalog of tremors for the period from 2012 to 2014, which was determined from the data analysis of a dense seismic array deployed in the Kii Peninsula, Southwest Japan. We successfully extracted 1,010 tremor migrations with durations ranging from 10 min to 24 hr. Along‐strike migrations propagating southwestward were predominant in the northeastern part of the Kii Peninsula, whereas those propagating northeastward were principal in the southwestern part. Regarding the along‐dip direction, tremor migrations propagating in the up‐dip directions were predominant in the deep part, and those propagating in the down‐dip directions were principal in the shallow part. The patterns of along‐strike migrations were related to the distribution of tremor energies, suggesting that tremor migrations may be controlled by heterogeneous structures of frictional properties on the plate interface. We further found that the migration speed is proportional to the inverse of the square root of the duration. This relation implies that a diffusion process controls the growth of fault ruptures behind tremor migrations. Plain Language Summary: Tectonic tremors, which are continuous seismic signals with a predominant frequency of 2–8 Hz, have been discovered in subduction zones worldwide. The tremor activity continues for several hours to days, and source locations migrate at speeds ranging from 10 to 1,000 km/day. Tremor migrations provide us with key to understanding the associated fault ruptures. However, their durations, migration directions, and migration speeds have variations that are too large to be characterized visually. Therefore, it is necessary to objectively extract the characteristics of tremor migrations. We developed a space‐time Hough transform to extract tremor migrations and applied it to data from a tremor catalog in the Kii Peninsula, Southwest Japan. We extracted 1,010 tremor migrations with durations ranging from 10 min to 24 hr. The tremor migrations consist of long‐duration migrations at slow speeds and short‐duration migrations at high speeds. The migration speeds decreased as the durations increased, suggesting that there was a diffusion process behind tremor migrations. We identified four areas where tremor migration was active. Migration patterns are different in each area and are related to the distribution of tremor energies. The results suggest that the migration patterns may reflect the frictional properties of the plate interface. Key Points: We developed a new method using a space‐time Hough transform to objectively extract tremor migrationsTremor migrations show fine structures in which multiple short‐duration tremor migrations occur during long‐duration tremor migrationThe relation between migration speed (Vmi) and duration (T) follows Vmi ∝ T−0.5 [ABSTRACT FROM AUTHOR]

We present corrections to the expressions for surface-wave Green's tensors shown in a recent paper by Malkoti et al. [ABSTRACT FROM AUTHOR]

Recent developments of distributed acoustic sensing (DAS) techniques and rotational seismometers enable us to record strain and rotational seismograms. This prompts us to adapt the methods used for traditional translational seismograms to strain and rotational seismograms. Recently, we extended the formulation of the spatial autocorrelation (SPAC) method to strain, rotation and tilt records. According to seismic interferometry for translation seismograms, cross correlations have clear connections to Green's functions under the isotropy and equipartition of noise wavefields. In this study, we clarify similar connections for strain and rotation seismograms. Because we extend the formulation in the frequency domain, we actually study the connection between the cross-spectral matrix of strain and rotation at two receivers and the Green's tensor. First, we provide a proof under a general framework by simply extending the proof for translation to that for strain and rotation. The proof shows the following results: (1) The cross-spectral matrix of (i, j)-component strains at two receivers is found to be proportional to the strain Green's tensor at one receiver for the sum of (i, j)- and (j, i)-components of the moment tensor source at the other receiver. (2) The cross-spectral matrix of (i, j)-component rotations at two receivers is found to be proportional to the rotation Green's tensor at one receiver for the difference of (i, j)- and (j, i)-components of the moment tensor source at the other receiver. Necessary assumptions for the proof are the isotropy and equipartition of the wavefield. We then check this proof with specific calculations for surface waves observed on the free surface. The proof is confirmed for the isotropic incidence of random-phase surface waves with a specific ratio between Rayleigh and Love wave energies. Seismic interferometry for strain and rotation was already studied. However, the connection of cross correlations to the Green's function for strain and rotation seismograms is discovered by this study for the first time. A specific proof for body waves that is assured by the general proof will be our next study. [ABSTRACT FROM AUTHOR]

Small-scale seismic velocity heterogeneity has been studied through the calculation of peak amplitude ratio as a means to quantify the strength of seismic wave scattering at volcanoes in Japan. This ratio is defined as the ratio of the maximum (peak) P wave energy in the transverse component seismogram envelope over that of the three-component sum seismogram envelope (transverse + radial + vertical). According to the previous study using Japan's Hi-net seismometer network, the peak ratio is observed to be larger near the (active) quaternary volcanoes. However, these Hi-net stations are not positioned on the volcanoes themselves. This study systematically examines the peak ratios at 47 active volcanoes across Japan, using seismometers operated by the Japan Meteorological Agency (JMA). Analyses were performed at four frequency bands: 0.5–1, 1–2, 2–4, and 4–8 Hz. We found that the JMA stations yield higher peak ratios than the Hi-net stations. Their differences are statistically significant at the 99.9% confidence level in all frequency bands. We also examined the differences between the ground surface and borehole stations of the JMA network. The former shows larger peak ratios, and for most frequency bands, the differences are also statistically significant at the 99.9% confidence level. This suggests an intensification of small-scale medium heterogeneities especially at shallow depths at active volcanoes, and that scattering might have been enhanced at the very shallow parts. [ABSTRACT FROM AUTHOR]

The spatial autocorrelation (SPAC) method has been applied to ambient seismic noise measured by arrays of translational seismometers for inverting phase-velocity dispersion curves of Rayleigh or Love waves for shallow S -wave velocity structure. Recently, it is becoming possible to observe wave spatial gradients such as strain, rotation and tilt owing to the development of dense seismic networks and improving measurement technologies. Therefore, it is desirable to extend the formulation of the SPAC method to strain, rotation and tilt. This study presents analytical expressions of cross-spectra and coherence of the strain, rotation and tilt components that are measured on the free surface. According to the results, both Rayleigh and Love waves contribute to most components of strains. The exceptions are the areal strain and the vertical axial strain (e zz ) on the free surface that are affected by only Rayleigh waves. Only Rayleigh waves contribute to the tilts and rotations around the horizontal axes on the free surface, too. On the other hand, only Love waves contribute to the rotation around the vertical axis. Therefore, different kinds of wave spatial gradients are helpful to separate Rayleigh and Love waves correctly. For practical applications, the analytical expression for the radial axial strain (e rr ) component will be applied directly to distributed acoustic sensing data measured with straight sections of a fibre-optic cable. On the other hand, dense observations of rotation and tilt may still be difficult to carry out at present. However, an application of analytical formulations in this study to arrays of at least several three-component rotational seismometers is attractive because it enables us to separately estimate the phase velocities of Rayleigh and Love waves. [ABSTRACT FROM AUTHOR]

This study performs detection and location of deep low-frequency tremors beneath the Kii Peninsula, Nankai subduction zone, Japan, using data from a dense seismic array. We set a quantitative threshold for the tremor detection based on the Gaussian distribution of waveform coherence among stations to distinguish between tremor signal and noise. Then, we estimated slownesses and arrival azimuths by applying the Multiple Signal Classification method and performed hypocentre determination by backprojection onto the plate interface. During the analysed period of 2 yrs from July 2012 to July 2014, we detected  25155 tremor events which are approximately 2.2 times more than those in the AIST-tremor catalogue based on the conventional envelope correlation method ( 11475 events). This improvement is due to the detection of even weak amplitude tremors owing to the array analysis scheme, which uses recorded waveform's phase information. Our high-resolution tremor catalogue leads to the following findings of the tremor activity in the Kii Peninsula. A few tremor episodes show continuous updip migration in the initial stage. The updip migration speed depends on whether the tremor migrates in the along-strike direction after the updip migration: approximately 1.0 km hr –1 for tremors that develop into the along-strike migration, and approximately 0.5 km hr–1 for those that do not. Estimated radiated energies are found to increase gradually with propagation into the updip direction. These findings suggest the depth dependence of the frictional strength on the plate interface. Furthermore, we noticed that accumulated radiated energies are positively correlated with the updip migration speed, and this relationship implies that fluid migration might play a role in the updip migration. [ABSTRACT FROM AUTHOR]

Locating change regions of seismic velocities and seismic scattering properties associated with volcanic activities and earthquakes is important for structural monitoring. To increase such applications, we propose to use sparse modeling to estimate spatial distributions of seismic scattering property changes. The sparse modeling is an inversion technique that enables us to estimate model parameters from a small data set with sparsity condition such as ℓ1 norm regularization. We apply this technique to seismic ambient noise cross‐correlation functions from 17 Hi‐net stations around the epicenter of the 2008 Iwate‐Miyagi Nairiku, Japan, earthquake (Mw=6.9). We compute waveform decoherences at the 0.5–1 Hz band and invert the waveform decoherences for the spatial distributions of seismic scattering property changes. Just after the main shock, the largest change occurred at the south of the epicenter, and the maximum change of the scattering coefficient in this region is estimated to be 0.032 km−1. The result from an ordinary linear least squares inversion with the ℓ2 norm regularization is almost consistent with that from the sparse modeling. Moreover, we confirm the superiority of sparse modeling in imaging with smaller data sets. Only five seismic stations that are deployed near the epicenter so as to surround the change regions are necessary to retrieve the result from 17 stations. On the other hand, in the case of the ℓ2 norm regularization, we need at least 15 stations. The sparse modeling will be helpful to estimate the spatial distribution of seismic scattering property changes from a small data set. Key Points: We detected seismic scattering property changes associated with the 2008 Iwate‐Miyagi Nairiku, Japan, earthquakeThe largest change was located to the south of the epicenter using 17 seismic stationsThis change region was also retrieved from only five seismic stations by using sparse modeling [ABSTRACT FROM AUTHOR]

We present a passive method to estimate the scattering and intrinsic absorption parameters of Rayleigh waves. Envelopes of seismic ambient noise cross‐correlation functions are calculated using the temporal flattening technique, and these parameters are estimated by modeling their space‐time distributions using radiative transfer theory. The scattering mean free paths are estimated to be 1.6–2.4 km at 0.5–1 Hz, 1.6–2.0 km at 1–2 Hz, and 1.6–3.2 km at 2–4 Hz at the Sakurajima volcano, Japan. Intrinsic absorption parameters are 0.04–0.06 s−1 at 0.5–1 Hz, 0.08–0.09 s−1 at 1–2 Hz, and 0.16–0.18 s−1 at 2–4 Hz. We validate these values by conducting a comparative study using active shot records at Sakurajima. Our passive method provides an effective way to measure scattering and intrinsic absorption parameters in various regions. Plain Language Summary: In recent years, small‐scale heterogeneous structures in the Earth's crust have been estimated in some regions using the data from natural earthquakes or active seismic experiments. However, these estimation methods are not applicable in regions where seismicity is low or active seismic experiments are not conducted. Therefore, we present a passive method to estimate these parameters using seismic ambient noise data. We apply our method to data at the Sakurajima volcano, Japan, and confirmed its validity by comparing our results with those from active seismic experimental data recorded at Sakurajima. Our passive method may be useful for imaging small‐scale heterogeneous structures in the shallow part of the Earth's crust in various regions. Key Points: We developed a passive method to estimate the scattering and intrinsic absorption parameters using ambient noise cross‐correlation functionsThis method is validated by comparing our results with those obtained from active shot records at Sakurajima volcano, JapanOne‐bit normalization and spectral whitening are not recommended for the estimation of scattering and intrinsic absorption parameters [ABSTRACT FROM AUTHOR]

Microcracks or micropores in rocks cause the elastic moduli to change with the applied strain owing to the nonlinear elasticity of the geomaterial, which causes temporal changes in the seismic wave velocity. Thus, variations in seismic wave velocity can be used as a proxy for understanding the strain or stress variations in the crust, which are crucial for figuring out the dynamics of the fault zones and volcanic domains. According to the theory of nonlinear elasticity, the second- and third-order elastic constants and strain tensors contribute to the strain derivative of seismic wave velocity changes. Although laboratory experiments have estimated third-order elastic constants for rock samples, the in situ values of those constants for the crust are difficult to obtain. In this study, seismic velocity changes were investigated in different directions of tidal deformations to provide constraints on the third-order elastic constants in the shallow crust by applying a seismic interferometry method to ambient noise records. We observed that negative velocity changes were of larger magnitude in the station-pair direction parallel to the tidal strain's direction. Nonlinear elasticity in shallow crust may cause anisotropic velocity variations in response to tidal deformations. Our results highlight the use of velocity change measurements in different directions of tidal strain to constrain nonlinear elastic parameters on a field scale. [ABSTRACT FROM AUTHOR]

Coda-wave interferometry is a technique to detect small seismic velocity changes using phase changes in similar waveforms from repeating natural or artificial sources. Seismic interferometry is another technique for detecting seismic velocity changes from cross-correlation functions of ambient seismic noise. We simultaneously use these two techniques to clarify seismic velocity changes at Sakurajima volcano, one of the most active volcanoes in Japan, examining the two methods. We apply coda-wave interferometry to the records of repeated active seismic experiments conducted once a year from 2011 to 2014, and seismic interferometry to the ambient seismic noise data. We directly compare seismic velocity changes from these two techniques. In coda-wave interferometry analyses, we detect significant seismic velocity increases between 2011 and 2013, and seismic velocity decreases between 2013 and 2014 at the northern and eastern flanks of the volcano. The absolute values are at a maximum 0.47 ± 0.06% for 2-4 Hz, 0.24 ± 0.03% for 4-8 Hz, and 0.15 ± 0.03% for 8-16 Hz, respectively. In seismic interferometry analyses, vertical-vertical cross-correlations in 1-2, 2-4, and 4-8 Hz bands indicate seismic velocity increases and decreases during 3 years of 2012-2014 with the maximum amplitudes of velocity change of ±0.3% for 1-2 Hz, ±0.4% for 2-4 Hz, and ±0.2% for 4-8 Hz, respectively. Relative velocity changes indicate the almost annual change. These periodical changes are well matched with volcano deformation detected by GNSS receivers deployed around the volcano. We compare the results from coda-wave interferometry with those from seismic interferometry on the shot days and find that most of them are consistent. This study illustrates that the combined use of coda-wave interferometry and seismic interferometry is useful to obtain accurate and continuous measurements of seismic velocity changes. [ABSTRACT FROM AUTHOR]

Recently, coda-wave interferometry has been used to monitor temporal changes in subsurface structures. Seismic velocity changes have been detected by coda-wave interferometry in association with large earthquakes and volcanic eruptions. To constrain the spatial extent of the velocity changes, spatial homogeneity is often assumed. However, it is important to locate the region of the velocity changes correctly to understand physical mechanisms causing them. In this paper, we are concerned with the sensitivity kernels relating travel times of coda waves to velocity changes. In previous studies, sensitivity kernels have been formulated for two-dimensional single scattering and multiple scattering, three-dimensional multiple scattering, and diffusion. In this paper, we formulate and derive analytical expressions of the sensitivity kernels for three-dimensional single-scattering case. These sensitivity kernels show two peaks at both source and receiver locations, which is similar to the previous studies using different scattering models. The two peaks are more pronounced for later lapse time. We validate our formulation by comparing it with finite-difference simulations of acoustic wave propagation. Our formulation enables us to evaluate the sensitivity kernels analytically, which is particularly useful for the analysis of body waves from deeper earthquakes. [ABSTRACT FROM AUTHOR]

We estimate high-frequency (4-20 Hz) energy release due to the 2016 Kumamoto, Japan, earthquake sequence, within a time period from April 14 to 26 through envelope inversion analysis applied to the Hi-net continuous seismograms. We especially focus on energy releases after each of the April 14 M 6.5 and the April 16 M 7.3 earthquakes. The cumulative energy release from aftershocks of the April 14 event reaches 60% of that from the April 14 event itself by the lapse time of 27 h (pre-April 16 period). On the other hand, the cumulative energy release from aftershocks of the April 16 event reaches only 11 and 13% of that from the April 16 event itself by the lapse times of 27 h and 10 days (post-April 16 period), respectively. This discrepancy in the normalized cumulative energy release (NCER) indicates that the April 14 event was followed by much larger relative aftershock productivity than the April 16 event. Thus, NCER would provide information that reflects relative aftershock productivity and ongoing seismicity pattern after a large earthquake. We also find that the temporal decay of the energy release rate obeys the power law. The exponent p of the power-law decay is estimated to be 1.7-2.1, which is much larger than the typical p value of the Omori-Utsu law: slightly larger than 1. We propose a simple relationship given by p = βp/ b, where p value, b value of the Gutenberg-Richter law, and β value of the magnitude-energy release relationship are combined. [ABSTRACT FROM AUTHOR]

For monitoring temporal changes in subsurface structures I propose to use auto correlation functions of coda waves from local earthquakes recorded at surface receivers, which probably contain more body waves than surface waves. Use of coda waves requires earthquakes resulting in decreased time resolution for monitoring. Nonetheless, it may be possible to monitor subsurface structures in sufficient time resolutions in regions with high seismicity. In studying the 2011 Tohoku-Oki, Japan earthquake (Mw 9.0), for which velocity changes have been previously reported, I try to validate the method. KiK-net stations in northern Honshu are used in this analysis. For each moderate earthquake normalized auto correlation functions of surface records are stacked with respect to time windows in the S-wave coda. Aligning the stacked, normalized auto correlation functions with time, I search for changes in phases arrival times. The phases at lag times of <1 s are studied because changes at shallow depths are focused. Temporal variations in the arrival times are measured at the stations based on the stretching method. Clear phase delays are found to be associated with the mainshock and to gradually recover with time. The amounts of the phase delays are 10 % on average with the maximum of about 50 % at some stations. The deconvolution analysis using surface and subsurface records at the same stations is conducted for validation. The results show the phase delays from the deconvolution analysis are slightly smaller than those from the auto correlation analysis, which implies that the phases on the auto correlations are caused by larger velocity changes at shallower depths. The auto correlation analysis seems to have an accuracy of about several percent, which is much larger than methods using earthquake doublets and borehole array data. So this analysis might be applicable in detecting larger changes. In spite of these disadvantages, this analysis is still attractive because it can be applied to many records on the surface in regions where no boreholes are available. [ABSTRACT FROM AUTHOR]

SUMMARY Spatial autocorrelation (SPAC) method is formulated in dissipative media for one-, two- and three-dimensional (1-D, 2-D and 3-D) scalar wavefields based on the generalized wave equation or the generalized telegraph equation. A rather straightforward derivation is possible by using a close mathematical relation between the SPAC method and seismic interferometry, though a model set-up should be modified by including attenuation. For 3-D cases, the normalized cross spectrum of a scalar wavefield in a dissipative medium is found to be different from that in a non-dissipative medium merely by an exponentially damping term. However, expressions for 1-D and 2-D cases are not as simple as 3-D cases meaning that not only amplitude but also phase in the normalized cross-spectrum are modified by attenuation. The SPAC expressions derived are considered to be applied to rather homogeneous distribution of attenuation. For the estimation of attenuation, the SPAC method needs to be used for larger station separations than the traditional SPAC method does. Analysis of the SPAC expressions for 2-D cases shows that the conjecture of Prieto et al. (2009) is not strict but approximately good for small attenuation. This study will provide a theoretical basis to estimate not only phase velocity but also attenuation from analysis of ambient noises. [ABSTRACT FROM AUTHOR]

It has been demonstrated theoretically and experimentally that Green's function between two receivers can be retrieved from the spatial correlation of an ambient noise field. While studies on the spatial correlation of ambient noises have been mainly conducted in the frequency domain, the retrieval of Green's function has been done in the time domain. Recently, it was pointed out that both studies could be connected via the inverse Fourier transform. In this paper, a systematic study of theoretical relations is conducted between the spatial correlation and the Green's function of wave equation in 1-, 2- and 3-D random scalar fields under an assumption that mutually uncorrelated plane waves are incident on two receivers from various directions. Different relation is obtained for a different dimensional case: Integration, Hilbert transform and differentiation have to be operated on the normalized spatial correlation to retrieve Green's function for 1-, 2- and 3-D, respectively. It is reconfirmed that the Green's function can be retrieved from the spatial correlation between two receivers only for isotropic random fields. The effect of the anisotropy of incident waves on the retrieval of Green's function is evaluated analytically. Even for the anisotropic incident waves, it is possible to retrieve Green's function from the spatial correlation by taking an azimuthal average in 2-D and a spherical average in 3-D. [ABSTRACT FROM AUTHOR]

Recently, an assertion has been verified experimentally and theoretically that Green's function between two receivers can be reproduced by cross-correlating the records at the receivers. In this paper, we have theoretically proved the assertion for 1-D media with the free surface by using the Thomson–Haskell matrix method. Strictly speaking, one side of the cross-correlation between records at two receivers is the convolution between Green's function and the autocorrelation function of the source wavelet. This study extends the geometry considered by Claerbout to two receivers vertically apart, and is a special case of the proof by Wapenaar et al. which dealt with 3-D arbitrary inhomogeneous media. However, a simple geometry in 1-D problems enables us to make the proof without any approximations and to better understand the physical background with more ease. That is the main advantage of this study. Though a 1-D geometry seems far from reality, it may be sufficient if an appropriate combination of receivers and earthquakes is selected. In fact, such a geometry is often seen in seismological observations by a vertical array of seismographs in the shallow subsurface. Therefore, we refer to a possibility that the proof in this paper is applied to the estimation of site amplification factors by using records of a vertical seismographic array. [ABSTRACT FROM AUTHOR]

The way in which cross-correlation coefficients of seismograms for closely located events change with frequency and event separation was investigated by using records of aftershocks of the 1995 Hyogo-Ken Nanbu, Japan, earthquake . For 11 master events ( M between 2.9 and 3.5) on the mainshock fault plane and smaller events ( M greater than 2.0) within 5 km of the master events, the relative hypocentre was determined with respect to each master event and cross-correlation coefficients of seismograms were calculated for a time window of 2 s including direct P waves for pairs of events in each cluster at two stations with an average hypocentral distance of 36 km and 42 km. Relative hypocentres were located within errors of several hundred metres for most of the earthquakes. Cross-correlation coefficients are high for frequencies below 8 Hz and decrease with increasing frequency above 8 Hz at both stations. It was also found that cross-correlation coefficients decay with increasing event separation. If the source–time function is assumed to be simple for small earthquakes, it may be said that dependence of cross-correlation coefficients on event separation below 8 Hz is mainly influenced by the heterogeneous structure around the earthquake source. We introduce a correlation distance where cross-correlation coefficients first become lower than 0.6 with increasing event separation. The correlation distance exhibits a clear regional difference: it is less than 4.5 km in the Awaji region and more than 4.5 km in the Kobe region. This result implies that heterogeneity of velocity structure around the earthquake source is stronger in the Awaji region than in the Kobe region. It is notable that both low- and high-frequency seismic waves radiated strongly from the Awaji region at the time of the mainshock. The correlation distance may be an indicator of heterogeneous structure around an earthquake source, which may have a close relationship to the rupture process itself. [ABSTRACT FROM AUTHOR]

Classification of volcanic tremors and earthquakes is an important part of volcano monitoring. Conventional classification technique relies on visual characterization of the amplitudes, frequency, or duration of seismic signals. In some cases, such classification is difficult to be maintained continuously, especially during high eruptive activities. In the present study, we develop a classification technique based on analyses of seismic wave correlation. The technique is based on a measure of spatial coherence using the distribution width of eigenvalues from the seismic covariance matrix that is referred to as 'spectral width'. We use the spectral width characteristics to extract the frequencies where the seismic signals are the most spatially coherent. We analyse 6-month continuous seismic data at six seismic stations deployed on Sakurajima volcano, Japan. The classification is performed every 10 min to recognize volcanic tremor and B-type earthquakes, both of which show unclear onsets and are dominant at around 1–4 Hz. Their frequencies of high spatial coherence are different from each other, providing a basis for automatic classification of both types of seismic events. Our classification results show that an increase in seismic activity, particularly volcanic tremors, occurred during high eruptive activity and is well matched with the events that are routinely determined by the Japan Meteorological Agency (JMA) based on visual examination of the data. We discuss the volcanic activities by combining our classification results with those from detections and location methods based on seismic correlation. Most of the source locations of volcanic tremor and B-type earthquakes are distributed at the same region with depths of less than 4 km beneath the active craters, which are almost consistent with the source locations of explosion earthquakes and B-type earthquakes determined from P -wave onsets in the previous studies. Also, the distribution of horizontal locations is spatially elongated at shallow depths above a dyke that is inferred to extend from Aira caldera at a depth of 7–9 km. We obtain deep and shallow sources for the tremor that occurred before and after an explosive eruption, respectively. We also obtain shallow sources beneath the active crater during a continuous tremor that accompanies increasing eruptive activity. The source locations of the classified volcanic tremor and B-type earthquakes may represent the movement of magma and gas from the shallow magma chamber at 5 km depth to the active craters. [ABSTRACT FROM AUTHOR]

We present one of the first studies on source location determination for volcanic earthquakes and characterization of volcanic subsurfaces using data from a distributed acoustic sensing (DAS) system. Using the arrival time difference estimated from well-correlated waveforms and a dense spatial distribution of seismic amplitudes recorded along the fiber-optic cable, we determine the hypocenters of volcanic earthquakes recorded at Azuma volcano, Japan. The sources are located at a shallow depth beneath active volcanic areas with a range of approximately 1 km. Spatial distribution of the site amplification factors determined from coda waves of regional tectonic earthquakes are well correlated with old lava flow distributions and volcano topography. Since DAS observation can be performed remotely and buried fiber-optic cables are not damaged by volcanic ash or bombs during eruptions, this new observation system is suitable for monitoring of volcanoes without risk of system damage and for evaluating volcanic structures. [ABSTRACT FROM AUTHOR]

Source location determination of volcanic tremor has been a challenge in seismology due to the waveform complexity and difficulties in reading P - and S -wave arrival times. We present a method for locating volcanic tremor recorded at a seismic network distributed around a volcano. The method combines the source-scanning algorithm and cross-correlation analysis. Tremor records are processed using a technique adopted from ambient seismic interferometry to obtain stacked cross-correlation functions (CCFs) for all station pairs, which are expected to show high amplitudes at the lag time that corresponds to the traveltime difference between the stations. The best seismic source location is determined from the maximum of the sum of envelope amplitudes of CCFs at predicted traveltime differences between all pairs of stations. This method does not compute theoretical amplitudes, assume an initial hypocentre location or measure the arrival times. To quantitatively evaluate the accuracy of the location determination, we examine the method by using the vertical component seismic data of volcano-tectonic (VT) earthquakes recorded at six seismic stations at Izu-Oshima volcano. The VTs have been previously located by using arrival times of P and S waves, and the hypocentres are used as the reference for evaluation and error estimation of the method. The results show that the misfit, which is the distance between our estimated sources and the references, is about 2 km or less when using CCFs at the frequency band of 4–16 Hz which contains the dominant frequencies of direct S waves. To test whether the method can be used for volcanic tremor, we simulate the tremors by combining the observed VTs that occurred randomly in time in a localized region. The simulated tremors are determined with location errors of approximately 1 km or less, when the sources of VTs are located within a distance of 1 km and CCFs are calculated for a minimum data length of about 2 min. The volcanic tremor location method we present here can be used as an alternative tool for volcano monitoring, especially to locate tremors and seismic events with no clear phase arrival. [ABSTRACT FROM AUTHOR]

It's quite important to clarify what physical parameters control the magnitude of volcanic explosion to understand magma processes in the conduit. Analyses of explosion earthquakes, which are associated with vulcanian eruptions, enable us to quantitatively evaluate dynamic process of the explosive eruptions. In the present study, we analyze hundreds of explosion earthquakes recorded at Sakurajima volcano, Japan, during 2012–2013. Applying a spectral ratio method that can retrieve the source spectral information without being disturbed by heterogeneous structure to three-component seismograms recorded at three stations operated by Japan Meteorological Agency, we examine scaling relations in explosion earthquakes. To obtain an averaged feature, we group the observed explosion earthquakes into five classes (I, II, III, IV, V) according to their maximum amplitudes and calculate the spectral ratios between the five classes. Because vulcanian eruptions continue to effuse volcanic ash for several to tens of minutes, spectral amplitude ratios are calculated by shifting a time window of 10 s for five successive 10s-long time windows from the onset to coda wave. The results show that the spectral amplitude ratios are characterized by the low and high corner frequencies: about 2 Hz and 3.5 Hz, respectively. With increasing the magnitude of explosion earthquakes, amplitude ratios at low and high frequency ranges, which are 0.5–2 Hz and 3.5–10 Hz, respectively, increase, while the corner frequencies do not significantly change. Since the explosion earthquakes at Sakurajima volcano occur at a same crater without changing the vent radius, the observed spectral amplitude ratio change may be attributed to pressure changes in the conduit. Another finding is that the amplitudes of spectral ratio gradually decrease with lapse time keeping the corner frequencies. Such a lapse time dependence of spectral ratios is completely different from those of tectonic earthquakes in which direct and coda waves show a same spectral amplitude ratio. We infer that the early part of the seismic waves of explosion earthquakes is generated by an initial rapid pressure release in the conduit while coda waves are by successive oscillations of the magma system during continuous ash emission. • Lapse time dependence of spectral ratios of volcanic explosion eartquakes is completely different from those of tectonic earthquakes. • The amplitudes of spectral ratio gradually decrease with lapse time while the corner frequencies do not change significantly. • Lapse time dependence is caused by an initial rapid pressure in the conduit followed by continuous ash emission. [ABSTRACT FROM AUTHOR]

We investigate seismic velocity changes in response to the tidal strain at Izu‐Oshima volcano, Japan, by analyzing the data of permanent seismic stations and a small seismic array to evaluate the characteristics of strain sensitivity of velocity changes. We estimate the seismic velocity changes by phase differences between cross‐correlations functions of ambient noises at the frequency of 2–4 Hz stacked for time periods with different tidal strain amplitudes. The seismic velocity changes decrease and increase during dilatation and contraction periods, respectively, when analyzing the cross‐correlations functions at early lapse times ranging from 2 to 7 s. The strain sensitivity of seismic velocity changes is estimated to be −2.1±0.2×104 at the early lapse times. However, we find that strain sensitivity of the seismic velocity changes decreases when analyzing the cross‐correlation functions at later lapse times from 7 s to 35 s. Applying an array analysis to the cross‐correlation functions, we observe apparent velocities of about 1 km/s at the early lapse times and those of higher than 1 km/s at the late lapse times. Since the group velocity of Rayleigh waves is 1.1 km/s at Izu‐Oshima volcano, the apparent velocities at the late lapse times may indicate the scattered or reflected body waves incident from a deeper region. Decrease of strain sensitivity with the lapse times therefore results from the emergence of body waves on the late lapse times. These results highlight the need to pay attention to wave types of cross‐correlation functions and their paths to interpret seismic velocity changes. Key Points: Seismic velocity decreases and increases during dilatation and contraction periods, respectively, at lapse times from 2 to 7 sArray analysis enables us to understand wave properties of cross‐correlations functions accuratelyStrain sensitivity of seismic velocity changes decreases with lapse times because scattered body waves contaminate the cross‐correlation functions [ABSTRACT FROM AUTHOR]

We investigate the stress sensitivity of velocity changes at Izu-Oshima in Japan using seismic interferometry method. We calculate cross correlation functions (CCFs) of ambient noises recorded by four seismic stations on the active volcano from 1 January 2012 to 31 December 2015 at 0.5-1 Hz, 1-2 Hz, and 2-4 Hz. Applying moving time windows to calculate cross spectrum between daily CCFs and reference CCF, which is the average for the all observation period, we compute daily velocity changes. The obtained velocity changes vary from −1% to 3% with a dominant period of about 1 year. We calculate areal strains by using Global Navigation Satellite Systems data at the volcano. It is evident that the velocity changes are well correlated with the areal strain changes whose linear trend is removed. The stress sensitivity of velocity changes, which are obtained from the observed velocity changes and areal strains, is lower at 0.5-1 Hz than that at higher frequencies: (7.1 ± 1.3) × 10−8 Pa−1 at 0.5-1 Hz, (1.4 ± 0.1) × 10−7 Pa−1 at 1-2 Hz, and (1.3 ± 0.1) × 10−7 Pa−1 at 2-4 Hz. Modeling the velocity changes of Rayleigh wave propagating in layered structures, we find that the observed velocity changes are concentrated in the upper 1 km of the structure. Since similar frequency dependency is recognized among the stress sensitivity of velocity changes reported in previous studies, we conclude that the velocity changes are localized in the shallow depth because of the increase of confining pressure. [ABSTRACT FROM AUTHOR]

Investigating spatio-temporal changes in crustal structure is a key to understanding the dynamic processes in the earth's crust associated with large earthquakes and volcanic activities. The dense seismic network in the northeast part of Japan (Tohoku region) allows us to apply seismic interferometry to continuous records of ambient noise data for studying crustal changes related to the 2008 M w 6.9 Iwate-Miyagi Nairiku earthquake and the 2011 M w 9.0 Tohoku-oki earthquake. We track changes in both seismic scattering properties and velocities over five years and spatially localize them based on the coda wave decorrelation and travel time sensitivity kernels calculated from the solution of the radiative transfer equation. The 2008 Iwate-Miyagi Nairiku earthquake caused significant changes in both seismic scattering properties and seismic velocities in the vicinity of the epicenter subjected to dynamic or/and static stress perturbations. Changes in scattering properties associated with the 2011 Tohoku-Oki earthquake around some active volcanoes in the Tohoku region are likely related to geofluid migrations triggered by strong vibrations during the large earthquake. The decrease in seismic velocity exists mainly along the eastern coastal region and extends to the central Quaternary volcanic chains under the combined dynamic and static stress changes by the earthquake. The waveform correlation is possibly sensitive to the fluid content and might be a useful indicator of fluid migration in the crust. This is the first comprehensive demonstration and comparison of the spatio-temporal changes of seismic scattering properties and velocities at long-term and regional scales under the influence of two large earthquakes. • Large earthquakes caused changes in subsurface structure in northeastern Japan. • Fluid migration causes co-seismic changes in scattering properties near volcanoes. • Both dynamic and static stress affect co-seismic velocity drops. [ABSTRACT FROM AUTHOR]

We introduce the single-station cross-correlation (SC) technique of processing ambient seismic noise and compare its results with the established cross-correlation (CC) and autocorrelation (AC) techniques. While CC is the correlation of the signals of two seismic stations with each other and AC is the correlation of a signal with itself, SC is the correlation of two different components of a single three-component seismic sensor. The comparison of the three different correlation techniques shows that CCs give the best results at frequencies below 0.5 Hz and that SCs give the best results at higher frequencies. In all three processing techniques, ambient seismic noise is correlated in order to reconstruct the Green's function describing the wave propagation between the first and the second sensor. By relating the coda parts of the daily Green's functions with the long-term reference Green's functions, shear wave velocity changes are determined. Here, we apply this technique to the data of 20 seismic stations in the surroundings of the fault zone of the Iwate-Miyagi Nairiku earthquake (MW = 6.9), which occurred on 2008 June 13, UTC (2008 June 14, Japan Standard Time) in the northern part of the Japanese island Honshu. The data range from 2008 January to 2011 June and therefore include the Tohoku earthquake (MW = 9.0), which occurred on 2011 March 11, off the coast of northern Honshu. The data are analysed in five different frequency ranges between 0.125 and 4.0 Hz. The data show coseismic velocity changes for both earthquakes followed by a post-seismic velocity recovery. In general, the coseismic velocity changes increase with frequency. For the Iwate-Miyagi Nairiku earthquake, the strongest velocity changes occur close to the fault zone. Quickly recovering coseismic velocity changes can be separated from changes not recovering during the study period. For the Tohoku earthquake, the complete area is affected by coseismic velocity changes. A modelling of the depth of the coseismic velocity changes indicates that the Iwate-Miyagi Nairiku earthquake can be explained either by large shallow velocity changes or by small, but deep changes. For one station, the observations can only be explained by assuming deeper changes. For the Tohoku earthquake, the modelling shows that different parts of the study area are affected in different ways, some showing shallow changes, others deeper changes. Furthermore, seasonal velocity variations occur, which are compatible for the different stations above 0.5 Hz, with velocity maxima in autumn. [ABSTRACT FROM PUBLISHER]

SUMMARY Using ambient seismic noise for imaging subsurface structure dates back to the development of the spatial autocorrelation (SPAC) method in the 1950s. We present a theoretical analysis of the SPAC method for multicomponent recordings of surface waves to determine the complete 3 × 3 matrix of correlations between all pairs of three-component motions, called the correlation matrix. In the case of isotropic incidence, when either Rayleigh or Love waves arrive from all directions with equal power, the only non-zero off-diagonal terms in the matrix are the vertical-radial (ZR) and radial-vertical (RZ) correlations in the presence of Rayleigh waves. Such combinations were not considered in the development of the SPAC method. The method originally addressed the vertical-vertical (ZZ), RR and TT correlations, hence the name spatial autocorrelation. The theoretical expressions we derive for the ZR and RZ correlations offer additional ways to measure Rayleigh wave dispersion within the SPAC framework. Expanding on the results for isotropic incidence, we derive the complete correlation matrix in the case of generally anisotropic incidence. We show that the ZR and RZ correlations have advantageous properties in the presence of an out-of-plane directional wavefield compared to ZZ and RR correlations. We apply the results for mixed-component correlations to a data set from Akutan Volcano, Alaska and find consistent estimates of Rayleigh wave phase velocity from ZR compared to ZZ correlations. This work together with the recently discovered connections between the SPAC method and time-domain correlations of ambient noise provide further insights into the retrieval of surface wave Green's functions from seismic noise. [ABSTRACT FROM AUTHOR]