The process of the collapse of the dacitic lava dome and the development of pyroclastic flows at Unzen volcano, Japan, were studied using infrasonic, seismic and video records. Characteristic infrasonic and seismic signals were recorded corresponding to the collapse of lava blocks from the dome, the drop of blocks on the slope and the migration of pyroclastic flow on the mountain slope. Small infrasonic and seismic waves are excited when the lava dome starts to collapse. When the lava blocks fall onto the mountain slope and are fragmented, larger waves are excited. This suggests that the seismic waves are generated by the collision of pyroclastics on the mountain slope and that the infrasonic waves are excited by small fractures of the dome and the fragmentation of pyroclastics. Some of the infrasonic signals show an obvious Doppler effect, indicating that the pyroclastic flows emit infrasonic signals during their propagation. The location of dome collapse and the path of pyroclastic flows can be identified and traced by a network of low-frequency microphones. The migrating source of infrasonic signals and probably seismic signals is inferred to be located near the front of pyroclastic flows by comparison with video images. This suggests that the fragmentation of pyroclastics occurs mainly near the front of pyroclastic flows. The speed of pyroclastic flows is estimated as 10-30 m/s from the infrasonic records. The excitation of infrasonic and seismic signals is affected by the topography of the mountain slope. The infrasonic energy is almost the same order as the seismic energy but the ratio of infrasonic to seismic energies increases for larger and more mobile pyroclastic flows. This means that the development of pyroclastic flows is controlled not only by the volume of lava and gravitational force, but also by the explosivity related to the pore gases in the lava.
We observed three explosions along a 60 km profile in the central part of the Kitakami massif, northeastern Honshu, Japan. Explosion sites and most of the observation sites were located on hard-rock outcrops. Using P-wave first arrivals, the average QP along ray paths and site amplification factors were determined for the frequency range between 6 and 30 Hz based on the amplitude spectra decay with distance. QP increased proportional to fn (n_??_0.9). The difference in amplification factors among hard-rock sites was as much as a factor of five for frequencies lower than 12 Hz and became large at higher frequencies. QS and S-wave site amplification factors were obtained only for 5 and 7.5 Hz. QS was slightly larger than QP, but the difference was not significant. S-wave site amplification factors were more variable than those of P-wave among the stations.
The peak horizontal acceleration and velocity of observed records from the 1995 Hyogo-ken Nanbu earthquake are compared with those predicted from empirical attenuation relations that were derived to be applicable to near-source areas. We found that the observed peak values matched well the empirical attenuation relations. The observed peak vertical accelerations were about half the peak horizontal accelerations when less than 100 cm/s/s. But, the observed peak vertical accelerations tended to be more than half the horizontal ones beyond 100 cm/s/s, and at some sites on soft soil, the peak vertical accelerations were greater than the peak horizontal accelerations. On the other hand, most of the observed peak vertical velocities were about 40% of the peak horizontal velocities independent of the peak amplitudes. At Port Island in Kobe City, remarkable non-linear behavior of soft soil was observed in the vertical array records. The high-frequency shear waves decreased due to liquefaction and the peak horizontal accelerations decayed at the surface. On the contrary, vertical ground motion was amplified by the strong contrast of P-wave velocity in the surface soils. This fact may be one of the causes for greater peak vertical acceleration than peak horizontal acceleration.
We observed the aftershocks of the 1995 Hyogo-ken Nanbu earthquake in Awaji Island from January 19, 1995 to April 4, 1995. We developed three temporal stations, IWY, TSM, and ABK, in the northern part of Awaji Island just after the main shock. The locations of these three stations are shown in Fig. 1. Velocity-type strong-motion seismometers (VSE) were installed at the three stations and a velocity-type broad-band seismometer (STS2) was also installed only at IWY. The solid lines in Fig.l show active faults (Research Group for Active Faults of Japan, 1991). Around Awaji Island, there are many surface faults as can be seen in Fig. 1. In this paper, we estimate Qc-1, which is the decay rate of S-wave coda amplitude, using the data from each station. The main characteristic of this study is that we estimate Qc-1 in the active-fault area just after the main shock. The geological conditions of the three stations are as follows. IWY is located on rock (granite) (Huzita and Maeda, 1984), while the other stations are located on sediments. TSM is located on sediment consisting of two formations, both belonging to the Osaka Group; one is the Toshima Formation laid from the Neogene to Quaternary period and the other is the Kariya Formation. The Kariya Formation overlies the Toshima Formation. The thickness of the sediment is at least 200 m (Mizuno et al., 1990). For the ABK station, there are also two formations, both belonging to the Osaka Group; one is the Atago Formation and the other the Goshikihama Formation. The Goshikihama Formation overlies the Atago Formation. The thickness of sediment is about 200 m (Takahashi et al., 1992). The ages and thicknesses of the sediments at TSM and ABK are almost the same. The basement in this study area consists of granite. TSM is located at the south end of an active surface fault, the Nojima fault, which moved during the main shock. In this paper, we first estimate the Qc-1 values in the broad frequency range at the three stations by applying the method of Sato (1977) to our observed data. We compare Qc-1 in this area with Qc-1 estimated in another area similar to our results, and then determine the common characteristics of these two areas. Next we take a look at the differences of estimated among the three stations in Awaji Island. Since the site conditions around the stations, the spatial configuration of the stations, the hypocenters of earthquakes and the faults are different, there may be different effects on the decay rate of S-wave coda amplitude for each station. Therefore, we investigate the frequency dependence of the Qc-1 values, and compare the Qc-1 values among the three stations. Then, we discuss the effects of the faults or sediments on Qc-1. We processed the records of 74 aftershocks observed from January 20 to February 19, 1995 for estimating the Qc-1 values. The epicenters and depths of the aftershocks used in this study, which were
In association with the 1995 Hyogo-ken Nanbu earthquake, we observed changes in the telluric field by about 10 mV/km about 10 s after the origin time with some long-span electrical dipole network located at about 70 km from the epicenter. After the occurrence of the earthquake, electric and magnetic measurements such as DC resistivity, VLF-MT, self-potential, and the geomagnetic total intensity across the Nojima fault were made at several places in Awaji Island, The low resistive zone beneath the western side of the fault was found in the derived resistivity structure of the shallower portion of the fault in Hirabayashi. On the other hand, however, no significant anomaly associated with the fault structure was found from the measurements of total intensity and self-potential. Prior to the occurrence of the main shock, no geomagnetic or electric continuous stations existed near the focal area. Approximately within 1 week after the occurrence of the main shock, we set up 10 continuous observation sites for the geomagnetic total intensity in the focal area in order to detect geomagnetic changes due to stress change caused by the aftershock activity. No changes in the total intensity in association with the individual aftershock activity have been observed so far. However, systematic temporal trends in the geomagnetic differences have been observed, especially at sites in the northern half of the observation area.
We have estimated the stress tensor at two earthquake swarm areas in Japan, Nikko-Ashio, Tochigi Prefecture and off Ito, Izu Peninsula, by applying a technique that inverts P-wave polarity data for a large number of events under the following assumptions: 1) Stress field in a small region is uniform, 2) slips occur along pre-existing weak planes randomly distributed in the region, and 3) slip direction is parallel to the direction of the maximum shear stress. We inverted data collected by the 1993 Joint Seismic Observation in the Nikko-Ashio area. The estimated maximum principal stress in this area is nearly horizontal in the NW-SE direction and the minimum principal stress is sub-vertical. Moreover, the state of stress just beneath Nikko-Shirane Volcano is different from that in the surrounding areas. The orientation of the minimum principal stress is sub-horizontal and perpendicular to the strike direction of opening cracks found at the top of Mt. Nikko-Shirane, an active quaternary volcano. This local change is probably related to recent magmatic activity. We also analyzed data from the 1989 Teishi Knoll eruption area. The estimated direction of the minimum principal stress is parallel to that of the maximum extension of the crustal deformation measured during the submarine eruption at Teishi Knoll. Spatial variations in the state of stress off Ito were found to be small.
Polarization anomalies of surface waves suggest the existence of lateral variations of isotropic and azimuthally anisotropic velocity structures in the upper mantle. We investigate the polarization anomalies of fundamental-mode Rayleigh and Love waves (37 earthquakes, 128 paths) at periods of 5-30 s as recorded by a local four-station network of broadband seismometers in Hokkaido, Japan. The network has been operated by the Research Center for Earthquake Prediction of Hokkaido University since December 1988. Rayleigh waves coming from many back-azimuthal ranges show three types of particle motion anomalies, which are usually called inclined, tilted, and sloping motions. The Rayleigh anomalies observed in the data for the Vanuatu region are mainly caused by the azimuthally anisotropic structure beneath the northwestern Pacific, because the effects of the lateral eterogeneities on the inclined motions are considered to be negligible. The Love waves coming from the earthquakes located near Oregon and California, USA, show anomalous waves in the vertical and radial components. It was expected that the waves were higher-mode Rayleigh waves. We calculate synthetic waveforms with normal modes for an oceanic spherically symmetric Earth model for the August 17, 1991, earthquake off the coast of northern California, which shows significant anomalous Love waves. A comparison of the synthetic and observed waveforms suggests that the anomalous waves are not higher-mode Rayleigh waves and require the Love to Rayleigh conversion. The conversion locations concentrate in and around the Kuril trench region. The Love wave anomalies may be caused by lateral variation in the isotropic or anisotropic structures beneath the Kuril trench region.
The 1995 Hyogo-ken Nanbu earthquake caused not only casualties and damage to buildings and civil structures but also many landslides. Most of the landslides were triggered or reactivated in the northeastern part of the Rokko Mountains, Nishinomiya and Takarazuka Cities, and the northern tip of Awaji Island. Based on the analysis of aerial photographs coupled with field surveys, 674 landslides were mapped within an area of about 700 km(2). The observed landslides were mainly rock slides, rock falls, and rock/debris avalanches. Debris slides, complex slides, and slumps were also found. A few debris slides showed low apparent friction angle during motion with long runout distance. The investigations of landslide typology and distribution show that, based on the relationship between landslide frequency and distances from the assumed fault rupture zone, an attenuation trend can be observed which shows a significant decrease within 3 km, while the maximum recorded distance was 10 km. While focused on the long runout mechanism in the Takarazuka golf course landslide, an undrained loading ring shear test on saturated golf course soil was carried out and succeeded to reproduce the low apparent friction angle mobilized in the landslide.
One of the most striking features of the 1995 Hyogo-ken Nanbu earthquake was the appearance of clear surface faultings of over 9 km on Awaji Island. There are still numerous speculations on the location and characteristics of faultings in Kobe, the northeastern part of the entire fault system, because no surface faultings of large scale have ever been reported (e.g., Tsukuda, 1996). In contrast, the location of the fault plane on Awaji Island, the southwestern part of the fault system, was identified with no doubt. Clear right-lateral slips were observed on the surface along the Nojima fault, to the northwest of Awaji Island (Nakata et al., 1995; Nakata and Yomogida, 1995). Since this earthquake provided a great opportunity to study detailed characteristics of surface faultings associated with a great in-land earthquake, it is important to investigate how the surveyed fault characteristics on the surface are related to some geophysical, particularly seismological, observations. For example, the aftershock distribution was found to agree grossly with the location of the surface faultings on Awaji Island, suggesting a nearly vertical fault plane. This nearly vertical fault plane was confirmed by a recent deep borehole trench across the Nojima fault conducted by the Geological Survey of Japan (H. Itoh, personal communication). The location of subevents retrieved by teleseismic studies such as Kikuchi (1995) is also consistent with our studies on the surface breaks, as well as strong-motion studies (e.g., Sekiguchi et al., 1996; Kamae and Irikura, 1995). Yomogida and Nakata (1995) discussed several relations of field-surveyed fault characteristics with seismic data. One of the most striking results was that some broadband seismograms may be closely related with the slip distribution observed along the Nojima fault, down to the scale of 1 km. Since broadband seismograms should be strongly affected by propagation effect, with a minimum site effect because of the borehole setting, a realistic earth structure and a double-couple source must be considered in the above discussion. This study focuses on how the fine-scale slip distribution observed on the surface can explain broadband seismic observations quantitatively. Figure 1 shows the location of fault traces along the Nojima fault and the slip distribution along the surface breaks. No systematic patterns of verticalslip component were found and the surface faultings can be described well as pure right-lateral slip. The amount of lateral slip is not nearly constant along the Nojima fault but the spatial slip distribution can be seen in a wide range of scales. In a large scale, the present fault system can be expressed by five segments, as shown in Fig. 1. The length of each segment is about 2 km, and the amount of lateral slip is maximum in the mid-section of each segment and goes to zero at both ends. As described in Nakata et al. (1995), spatial variations of slip can be observed on a scale down to meters. Surface faultings are characterized not by a simple straight line, but have complex structures such as jogs, branches and en-echelons. They are more likely described as fractal (e.g., Okubo and Aki, 1984; Matsumoto et al., 1992). The fractal structures of surface faultings should be related to high-frequency seismic-wave radiation because a smooth rupture
The generalized fractal dimensions are measured for the time series based on two complete earthquake catalogues: one with M greater than or equal to 6 earthquakes occurring in the north-south seismic belt of mainland (China during the 1900-1990 period published by Ma er al. (1992) and the other with M greater than or equal to 5.5 earthquakes occurring in southern California, USA during the 1915-1994 period compiled by Press and Alien (1995). The log-log plot of C-q versus t, where C-q(t) is the generalized correlation integral and t is the interoccurrence time in years between two events, at positive q shows a linear distribution when t < t(c). D-q is the slope of this linear portion. The value of t(c) decreases from 50.1 to 39.8 years for Chinese earthquakes and fi om 50.1 to 31.6 years for southern California events as q is increased from 0 to 15. For M greater than or equal to 6 Chinese earthquakes, the well-distributed, monotonically decreasing function of D-q with increasing q would imply that such earthquakes have formed a multifractal time series. In contrast, the M greater than or equal to 5.5 southern California earthquakes might have not yet formed a complete multifractal time series or the number of these events is too small to accurately estimate the multifractal dimensions, especially for large qs. Different degrees of complexity of fault distributions in the two seismic regions might also be a factor in causing the difference in the D-q-q relations. In addition, the results also suggest that a D-q-q relation is better than the first three commonly-used values of D-q to completely represent a multifractal time series.
Earthquakes are extraordinarily complicated phenomena and deterministic prediction of the magnitude, time and place of large earthquakes is likely to be intrinsically impossible. A more simple phenomenon, the build up of deformation which leads to the earthquake, can be monitored by the behaviour of seismic shear waves, and this offers a way of forecasting the proximity of large earthquakes. Recent results, including field observations of shear waves before earthquakes, laboratory experiments with stress cells, and theoretical modelling of microscale deformation, demonstrate that long-term precursory build-up of deformation can be monitored for some years before a large earthquake. There may even be the possibility of identifying short-term precursors a few hours or days before the earthquake. This paper suggests that reliable routine earthquake forecasting would require controlled seismic experiments between a specific pattern of deep wells.
We explored the atmospheric contribution to the excitation of Chandler wobble (CW), which has spanned about 11 years beginning from September 1983. The atmospheric angular momentum (AAM) function presented by the Japan Meteorological Agency (JMA) and the wobble data set (SPACE93) are employed. We devised a wobble domain method of analysis which enables us to quantify the narrow band power of AAM. The AAM-induced wobble closely resembles the observed wobble, and wind contribution turns out to dominate over atmospheric pressure contribution in the vicinity of the Chandler frequency. When only pressure contribution is taken into account, it is insufficient, as shown in previous studies.
A new time-series analysis called "wavelet transform" is applied to measure the group velocity of surface waves as compared with that obtained by the conventional Fourier transform. We use vertical-component Rayleigh waves for both synthetic seismograms and GDSN long period data of oceanic paths. The results of this study are summarized as follows: for synthetic seismograms, moving-window analysis using the Fourier transform can measure the group velocity of the fundamental mode correctly, while the group velocity of the first-higher mode is systematically larger than the correct value. In contrast, the wavelet transform measures the group velocity of both modes precisely although the resolution in frequency may not be sufficiently high. For GDSN data propagating along the Pacific Ocean, both methods provide stable results for the group velocity of the fundamental mode in the period range of 20 to 100s. Using the Fourier transform, we obtain;he group velocities of the first-higher mode between 20 and 40 s although these values seem unreliable. In contrast, the wavelet transform can measure both modes precisely In the period range of 20 to 100s for non-shallow events and even for shallow events with relatively small noise in the data. Another advantage of the wavelet analysis is that we can specify resolving power in group velocity measurement rigorously.
In this paper, a case study has been made on the characteristics of earthquakes and explosions, mainly through the polarization method, estimating the epicentral azimuths and incident angles in the case of shallow- and deep-focus earthquakes as well as explosions based on three-component signal-to-noise ratios. We have found that it is possible to determine the epicenter and incident angle of an earthquake or an explosion by three-component single-station polarization analysis, especially using body waves like Pn, and in some cases, stable and accurate results are obtained. We made a theoretical discussion on the polarization properties of shear waves, and found that multievents which take place occasionally within explosions can be discerned by azimuth and motion pattern monitoring.
S-wave coda is a useful tool to investigate high-frequency seismic wave attenuation in the lithosphere. The generation and amplitude decay of S-wave coda were introduced by a single scattering model proposed by Aki and Chouet (1975). The amplitude decay rate with time is defined by the quality factor of the S-wave coda (QC-1), which is investigated widely at various regions by the single scattering model. It is only one-component seismograms that have been used to estimate QC-1 in most studies. This is based on the assumption that the decay of the S-wave coda amplitude recorded on a one-component seismometer is equivalent to the decay of the S-wave coda amplitude which is introduced by the single scattering model. In the single scattering model, S-wave coda is represented by the superposition of single scattered waves which are generated by scatterers distributed randomly and come from any direction to the station. We hereafter call the S-wave coda amplitude expected by the single scattering model the "real coda amplitude." The ground motion of the S-wave coda portion, which consists of the superposition of the scattered waves, has various directions with time. Since the amplitude of the S-wave coda part on a one-component seismogram represents only the projection of the real coda amplitude in the direction of the component, the amplitude on the seismogram depends on the direction of the ground motion. Therefore, it is not trivial that the amplitude decay of a one-component seismogram for a coda part gives the decay of the real coda amplitude (i.e., the assumption). If we have three-component seismograms, we can estimate the real coda amplitude more exactly by vector addition of the amplitudes of the three-component seismograms (hereafter "total vector-amplitude"). In this report, we will check the assumption using real three-component seismic data. We estimate QC-1 from a total vector-amplitude and amplitudes of one-component seismograms, and compare them to investigate QC-1 differences among these amplitudes. Here, we adopt 5 kinds of amplitudes to estimate QC-1 as a one-component amplitude: three of them are the amplitudes of each original component and the other two are the amplitudes of the radial and transverse components, which are constructed by horizontal two-component amplitudes. Furthermore, we produce the horizontal vector-amplitude, by vector addition of the amplitudes for, horizontal two-component seismograms, and measure QC-1. This QC-1 is also compared with the QC-1 measured from one-component amplitudes or total vector-amplitude.
Migration velocity analysis is investigated to obtain a long-wavelength velocity model for a complex structure. Residual wavefront curvature analysis is combined with prestack reverse-time depth migration to update the velocity model as well as the image of the subsurface structure. The imaging point for a given velocity error is solved rigorously in the case of a dipping layer. Convergence behavior is examined by three synthetic examples in an iterative scheme in which the depth and velocity are modified interactively. A stable solution can be obtained by smoothing the model space with an appropriate window length. It was found that the velocity is modified from the shallower part to the deeper part gradually, and that the number of iterations that is necessary for convergence is correlated with the ratio between the target depth and window length of the moving average. In a field example of land survey conducted around a buried fault zone, lateral variation of velocity was successfully detected. With the final velocity model and final seismic section, this method provides more reliable interpretations than the CMP stacking method.
Researchers usually assume a simple linear relation between velocity and temperature in interpreting seismological models for geodynamical inference. Seismic velocities are also susceptible to the bulk mantle composition (Jordan, 1979) and the amount of volatiles such as water (Karato, 1995) in addition to the temperature; nevertheless, this assumption is apparently good since the correlation between surface heat flow and surface wave phase velocity is known (Nakanishi and Anderson, 1984), and the pattern of global distribution of surface heat flow can be reproduced by three-dimensional tomographic models (Yan et al., 1989). One unsettled issue, though, is how to scale a velocity anomaly to a temperature anomaly. In this study, I examined the scaling between temperature and velocity in the oceanic lithosphere using recent thermal and velocity models. The thermal structure of the oceanic lithosphere is constrained by such geophysical observations as the variation of sea-floor depth and heat flow with lithospheric age (e.g., McNutt, 1995), and is independent of seismological observations. Therefore, researchers can use these models to investigate the relation between temperature and seismic velocity. Specifically, I used models for the Pacific Basin, where the anomaly of sea-floor depth and surface wave velocities were associated with the past activity of mantle plumes (e.g., McNutt and Judge, 1990). A previous estimate of this scaling was derived indirectly via partial derivatives of Love wave phase velocity with respect to shear-wave velocity (McNutt and Judge, 1990), which strongly depend on the assumed velocity structure. Instead, I used seismological models that satisfy both Love and Rayleigh wave observations and also include anisotropy. Gross seismological observations show that the shallow oceanic upper mantle is anisotropic, and no isotropic model so far has been successful in satisfying both Love and Rayleigh wave observations simultaneously. The use of an anisotropic model is thus essential in such an analysis. For the seismic velocity models, I used the transverse isotropic velocity models of Nishimura and Forsyth (1989) (hereafter referred to as NF89). They used a common method to derive five age-dependent models. Though the azimuthal anisotropy is the expected form of anisotropy in the oceanic lithosphere on which the alignment of olivine species plays a major role, if the propagation paths average out the azimuthal dependency, azimuthal anisotropic upper mantle can be modeled with an apparent transverse isotropy (NF89). Though they assumed a smooth gradient between a seismic lid and a low-velocity zone (LVZ), their model for 52-100 Ma is quite similar to a model
Earthquake focal mechanisms before and after the 1995 Hyogo-ken Nanbu earthquake have been investigated using seismic records from regional seismic networks. Before the mainshock, seismicity was very active at the Tamba Plateau, a neighboring area of the Hyogo-ken Nanbu earthquake rupture zone. In contrast, the seismicity along the Hyogo-ken Nanbu earthquake rupture zone was not so active. Most earthquakes in these regions had source mechanisms of E-W compression and were of the strike-slip or reverse-fault type. Most aftershocks along the Hyogo-ken Nanbu earthquake rupture zone have strike-slip solutions with P-axis in the E-W or ESE-WNW direction, which is compatible with the trend of aftershock distribution and the strike of active faults the same as the mainshock mechanism. Simultaneously, many other aftershocks were of the reverse-fault type with E-W compression. This area is still controlled by the regional stress field of E-W compression observed before the mainshock. Although, we could find various types of mechanisms in the aftershock sequence, some normal fault-type events were also observed in the mainshock rupture zone. We could find events of SE-NW compression, and this direction is nearly perpendicular to the trend of the mainshock rupture zone. Some aftershocks that occurred near the epicenter of the mainshock had solutions of N-S compression. The geometry of the active fault systems and/or local stress change induced by the mainshock may cause these complex features of focal mechanisms. After the mainshock, the focal mechanisms of earthquakes in the Tamba Plateau were approximately E-W compressional; the same as that before the mainshock.
In this study, an attempt was made to examine the effect of site conditions on damage distribution in the Kobe City area, after the Hyogo-ken Nanbu earthquake of Jan. 17, 1995, using measured microtremors. A preliminary measurement of microtremors was done between Feb. 18 and 21, 1995. Based on the findings of the preliminary measurement, a more detailed investigation was made between March 4 and 6, 1995. In both cases continuous-stationary measurements and mobile measurements were made. The measured microtremors in the Kobe area showed site-dependent variation in time history, spectral amplitude, spectral ratio, and predominant period. Stiff soil sites showed lower values of amplitude both in time history and spectral amplitude, while soft and deep soil formations showed higher values. The predominant periods also varied in relation to the soil formation in some of the cases. The damage distribution of building structures, especially that of low-rise structures including wooden houses, showed good relation with the characteristics of measured microtremors.
By modifying Uffen's approach, which relates the Einstein-Debye theory of solids to Lindemann's melting criterion, the relative melting temperature distributions in each major division of the Earth's interior are calculated from observed seismic velocities. With further assumptions, the melting temperature profile and three possible present temperature distributions inside the Earth are also proposed. By extrapolation, this approach yields a temperature of less than or equal to 8,000 K at the center of the Earth. On the basis of the melting temperature gradients at each major seismic discontinuity, it is concluded that only the 220- and 400-km seismic discontinuities are consistent with a phase transition possessing a positive Clapeyron slope, whereas all other seismic discontinuities are found to be consistent with either a phase transition having a negative slope or a chemical change. The outer-inner core boundary is not a simple melting phenomenon. These conclusions have been reached independently of all previous studies that used elastic and thermal properties of the Earth and of the predicted compositions for the various parts of the Earth.
