Three-component Seismograms Research Articles

Study of seismogram envelopes based on scattering by random inhomogeneities in the lithosphere: a review

Abstract Instead of the precise examination of individual phases in seismograms, we will focus on the relation between random inhomogeneities in the lithosphere and the systematic change in seismogram envelopes. Recent developments in the study of high-frequency seismogram envelopes are briefly reviewed. The smooth amplitude decay of coda envelopes of local earthquakes first attracted seismologists' interest in scattering and was investigated based on the single-scattering assumption with a uniform distribution of point-like scatterers. Subsequently, the importance of multiple scattering was recognized and theoretical models were introduced. As a bridge between conventional seismology and the scattering approach, a hybrid model in which a large impedance contrast plane was embedded in a scattering medium was proposed for the interpretation of a reflected S phase in the coda for shallow events around a volcanic front in Japan. Next, three-component seismogram envelopes of a small local earthquake were synthesized using the Born approximation in inhomogeneous media. The S coda is excited in all directions from the hypocenter. The P coda appears even in the nodal direction of the P-wave radiation mostly owing to SP scattering near the hypocenter. Both pseudo P and S phases appear even on the null axis. These synthetic seismogram envelopes are in harmony with characteristics of observed seismogram envelopes of local earthquakes. Scattering due to random inhomogeneities quantitatively explains the frequency dependence of S-wave attenuation. From the analysis of seismogram envelopes around direct S waves of earthquakes with intermediate hypocentral distances, it was recently found that the time lag of the maximum amplitude arrival from the S-wave onset increased and the envelope broadened with travel distance. Diffraction due to slowly varying velocity inhomogeneities is modeled using the parabolic approximation, and is used to explain the envelope around the S-wave arrival observed at intermediate distances. A nonlinear inversion of the full seismogram envelope is used for the evaluation of randomness. Thus, the analysis of seismogram envelopes offers us useful information on the randomly inhomogeneous structure of the lithosphere complementary to the array observation analysis of individual phases.

Automated phase picker and source location algorithm for local distances using a single three-component seismic station

Procedures for automatic location of local events by using single-site, three-component seismogram records are described. Epicentral distance is determined from the time difference between P- and S-onsets. For onset times estimates a special phase picker algorithm is introduced. Onset detection is accomplished by comparing short-term average with long-term average after multiplication of north, east and vertical components of recording. For epicentral distances up to 100 km, errors seldom exceed 5 km. The slowness vector, essentially the azimuth, is estimated independently by using the Christoffersson et al. “polarization” technique, although a- priori knowledge of the P-onset time gives the best results. Differences between “true” and observed azimuths are generally less than 12°. Practical examples are given by demonstrating the viability of the procedures for automated three-component seismogram analysis. The results obtained compare favourably with those achieved by a miniarray of three stations.

The Removal of Free Surface Interactions From Three-Component Seismograms

SUMMARY The effect of the free surface can be removed from three-component seismic recordings to recover the incident upgoing wavefield, if the slowness and azimuth of the current wavefront are known as a function of time. For a single threecomponent station it is usually possible to estimate an azimuth for an event from the first arriving P-waves, but slowness estimates are less reliable when more than one wavetype is presented in the seismic wavetrain. However, the free surface correction operators are generally slowly varying functions of slowness and so some error in slowness can be tolerated. Effective approximations for the removal of the free surface effects can be made for hard rock sites to cover slowness bands for the main regional phases Pn, Pg, Sn and Lg. By applying these operators in turn over group velocity windows appropriate to the particular phases, the relative amplitude of the P, SV and SH contributions to the wavefield can be estimated. Because the free surface amplification effects have been removed, the amplitudes can be compared directly and provide useful constraints on the radiation characteristics of the source. This procedure is therefore helpful for developing discrimination measures for different classes of sources.

Three-component analysis of regional seismograms

Abstract Both phased array techniques for single-component sensors and vectorial analysis of three-component recordings can provide estimates of the azimuth and slowness of seismic phases. However, a combination of these approaches provides a more powerful tool to estimate the propagation characteristics of different seismic phases at regional distances. Conventional approaches to the analysis of three-component seismic records endeavor to exploit the apparent angles of propagation in horizontal and vertical planes as well as the polarization of the waves. The basic assumption is that for a given time window there is a dominant wavetype (e.g., a P wave) traveling in a particular direction arriving at the seismic station. By testing a range of characteristics of the three-component records, a set of rules can be established for classifying much of the seismogram in terms of wavetype and direction. It is, however, difficult to recognize SH waves in the presence of other wavetypes. Problems also arise when more than one signal (in either wavetype or direction) arrive in the same window. The stability and robustness of the classification scheme is much improved when records from an array of three-component sensors are combined. For a set of three-component instruments forming part of a larger array, it is possible to estimate the slowness and azimuth of arrivals from the main array and then extract the relative proportions of the current P-, SV-, and SH-wave contributions to the seismogram. This form of wavetype decomposition depends on a model of near-surface propagation. A convenient choice for hard-rock sites is to include just the effect of the free surface, which generates a frequency-independent operation on the three-component seismograms and which is not very sensitive to surface velocities. This approach generates good estimates of the character of the S wavefield, because the phase distortion of SV induced by the free surface can be removed. The method has been successfully applied to regional seismograms recorded at the medium aperture Warramunga array in northern Australia, and the two small arrays NORESS and ARCESS in Norway, which were designed for studies of regional phases. The new wavefield decomposition scheme provides results in which the relative proportions of P, SV, and SH waves as a function of time can be compared without the distortion imposed by free surface amplification. Such information can provide a useful adjunct to existing measures of signal character used in source discrimination.

Computation of complete waveforms in general anisotropic media-results from an explosion source in an anisotropic medium

SUMMARY An algorithm of generating complete waveforms in general anisotropic media (21 constants) is presented. Complete waveforms have been computed including a point-source in the anisotropic medium and the reflection and transmission properties of individual interfaces. A recursive scheme of scatterer operators and the numerical wavenumber integration technique are used. The 6 X 6 system matrix A of each layer is derived. Each element of A is expressed in such a manner that the wavenumber could be used as an inner loop (vectorizing loop). A numerical algorithm is used to compute simultaneously the eigenvalues and eigenvectors of A. The upgoing and downgoing eigenvalues are separated according to a radiation condition and the properties of the z-component Poynting vector. Filon integration is applied to evaluate the integrals over horizontal wavenumbers. Since wave propagation in an anisotropic medium is a 3-D problem, it requires an enormous number of computations. To reduce the computation time, the wavenumber is used as the deepest loop. An efficient computer code has been developed for vectorization in the super computer. Both source and receivers can be placed at arbitrary depths. Three-component synthetic seismograms are computed from an explosion in the anisotropic medium. Examples are calculated for half-space multilayered anisotropic media. Theoretical seismograms show that radiation patterns are strongly affected by the anisotropy. For example, an explosion in an anisotropic medium with horizontal axes of symmetry generates P-, SV-, and Rayleigh waves, all whose amplitudes are varying with azimuth, and significant transverse (SH) waves at certain azimuths. These effects could be explained by the non-spherical behaviour of the source due to the directional dependence of elastic moduli. The waveforms are further complicated by the reflection and conversion phenomena of waves propagating in anisotropic media.

On shear-wave splitting in the Los Angeles basin

Three-component seismograms at the three USC stations, PVP, GFP and DHB, have been examined. Most earthquakes, with magnitudes ranging from 1.4 to 5.0, within a period from 1985 to 1988, show evidence of shear-wave splitting. The preferred polarization of the first split-shear wave arrivals at PVP is nearly in N-S which is consistent with both regional maximum horizontal compressive stress direction and local subsurface fault strike, showing that shear-wave splitting is caused by liquid-filled cracks or fractures associated with the N-S faulting. The polarizations of first shear wave arrivals at GFP are roughly divided into two almost perpendicular directions, ENE-WSW and NNW-SSE, which are parallel or perpendicular to the strike of the geology or topography near the station. Because GFP is near the foothills of Santa Monica Mountains, the shear-wave arrivals may be disturbed by topographic irregularities and by subsurface dipping interfaces. Two examples at DHB clearly display shear-wave splitting. Their polarizations of shear wave are in the direction of N-S, which agree with the fragmentary surface and fracturing direction there. From some relatively reliable delay times, the crack densities at three stations are given, that is, 0.025 at PVP, 0.020 at GFP and 0.045 at DGB. No systematic change of shear-wave polarization is discovered in this study.

Evidence for the splitting of shear waves from waveform and focal mechanism analyses

Abstract A temporary seismograph station was sited above the aftershock area of an intermediate crustal earthquake (M = 4.9) in the Kinki district, Japan. The results from analysing digital three-component seismograms and numerically determining focal mechanisms of aftershocks have provided clear evidence for the splitting of shear waves attributed to crustal anisotropy. Faster shear waves from most of the aftershocks are polarized WNW-ESE, regardless of their azimuths and incident angles. Such shear wave motions are often significantly inconsistent with the focal mechanisms of the aftershocks determined using P-wave first motions. For many of the events, slower shear waves, polarized orthogonally to the faster shear wave direction (NNE-SSW), are observed to arrive about 0.1 s later. Two vectors showing particle motions of faster and slower shear waves in the horizontal plane are summed, to reconstruct polarizations of shear waves before splitting. For events with sufficiently high signal-to-noise ratios of shear waves and well determined focal mechanisms, the reconstructed polarizations are consistent with the focal mechanisms within the range of uncertainty. Such consistency is not clearly detected for the other events, possibly because of large noise effects and poorly constrained fault-plane solutions The direction of faster shear wave polarization (WNW-ESE) generally coincides with that observed at three other stations located within 15 km of the temporary station. In the study area, the fabric of crustal anisotropy is almost uniform over the scale of 15 km, but appears to exhibit small local fluctuations.

Evidence from shear-wave splitting for the restriction of seismic anisotropy to the upper crust

The existence of anistropy in the propagation speed of seismic waves through the Earth's crust has been attributed to the non-random alignment of microcracks in the crust1. For sparsely distributed, parallel vertical cracks, perturbation theory predicts the splitting of a shear wave into two orthogonally polarized waves2–4. Here we present extremely clear evidence for shear-wave splitting in three-component seismograms from crustal and sub-crustal microearthquakes recorded in the Shikoku district of Japan. Quantitative analysis of travel-time differences between split shear waves leads to the conclusion that the observed anisotropy is limited to the upper crust, shallower than 10–15 km. This model is consistent with recent results from reflection seismology5, and implies that microcrack density or orientation changes from the upper to the lower crust.

Wavefield decomposition using ML-probabilities in modelling single-site 3-component records

SUMMARY This paper presents a new approach to the analysis of three-component digital seismograms. Earlier approaches used techniques such as Principal Components to estimate particlemotion using models of P and S waves. In this paper the Maximum-Likelihood (ML) estimator is preferred because this allows the use of X2-probabilities to test whether energy of a specific wave type (P, S, Love or Rayleigh) is present. In addition, this analysis allows the joint estimation of azimuth of approach and in cases of P- and SV-waves also apparent angle of incidence (and, hence, information on apparent velocity). For a single three-component seismogram, the covariance matrix provides only six independent observations, thus restricting analysis to rather simple wave models. The technique works satisfactorily for P-waves whereas shear and surface-wave models sometimes prove cumbersome to handle due to correlation between radial and transverse components reflecting complex propagation characteristics in inhomogeneous media. This technique has been tested on synthetic data and in such cases works perfectly for all wave types. An important aspect of this work has been the visual display of the probability and velocity information as functions of time and azimuth. Displaying the data in this form provides information on the ray path in a manner similar to analysis performed by seismic arrays. Practical examples on a variety of siesmic data are given to illustrate the viability of the technique: the data cover a broad spectrum of frequencies and applications from broad-band teleseismic data (1 Hz), regional seismic data (10-40 Hz), seismic profiling data (125 Hz) and VSP (500 Hz) recordings.

Event location at any distance using seismic data from a single, three-component station

This paper demonstrates the usefulness of a novel technique for analysis of single-site, three-component seismogram records for reliable epicenter determinations for even small seismic events at any distance. Parameters that may be extracted by this technique are P -phase identifications in terms of x 2 probabilities and the associated slowness vectors. Using data from short-period and broadband seismometers installed in the new NORESS (Norway) array, single-site event locations are demonstrated: azimuth estimates are found directly from the slowness vector, while epicenter distance estimates are derived from differential arrival times between Pn, Pg, Sn , and Lg ( Sg ). Simultaneous inversion for crustal model and hypocenter data indicate a crustal thickness beneath NORESS of 33.5 km. If Sn and/or Lg are well-recorded, focal depth estimates are often accurate to within a few kilometers. Locating teleseisms is problematic, since secondary phases are not fully readily identifiable, and hence distance can be estimated only by conversion of the measured apparent phase velocity. For distances out to approximately 1000 km, event location errors as compared to network solutions seldom exceed 50 km and are due mainly to errors in azimuth estimates. With Sn phases available, focal depth errors appear to be of the order of 4 km. For teleseismic P waves (short-period), location errors occasionally exceeded 7° and were due mainly to poor distance estimates. Using broadband records from relatively strong events, location errors were about 1°. Further refinements seem feasible by using seismicity information and/or recognition of stationarity in the P -wave decomposition pattern for events within the same region. At least in certain cases, such approaches reflect genuine location refinements as demonstrated.

Observation and modelling of fault-zone fracture seismic anisotropy -- II. P-wave polarization anomalies

Summary. In Part I of this paper we modelled shear-wave splitting observed in crystalline rock bordering an active, normal fault-zone at Oroville, California, with Cervený's ray-tracing system applied to anisotropic heterogeneous media using Hudson's formulation of elastic constants for a medium containing aligned cracks. In Part II we use the ray-tracing results of Part I to quantitatively interpret P-wave polarization anomalies observed in the three-component seismograms recorded in the Oroville fault zone. We show that the eigenvectors of the first-order Christoffel tensor defined by the ray-tracing slowness vector and Hudson's first-order anisotropic corrections to the isotropic elastic tensor correctly account for P-wave first motion that deviates from the ray vector.

Complex polarization analysis of particle motion

Knowledge of particle motion polarization aids in identifying phases on three-component seismograms. The scheme of Montalbetti and Kanasewich (1970) is extended to analytic three-component seismograms, where the imaginary part of the signal is the Hilbert transform of the real part. This scheme has only one free parameter, the length of the time window over which the polarization parameters are estimated, so it can be applied in a routine way to three-component data. The azimuth and dip of the direction of maximum polarization and the degree of elliptical polarization as a function of time for the seismograms are obtained. Polarization analysis of strong motion data from the 1971 San Fernando earthquake aids in the discrimination between wave types, which is important for the understanding of the complicated earthquake-induced shaking observed in basins. Most arrivals are incident on the receivers in the direction of the back-azimuth to the epicenter, which suggests that despite the complicated motions, two-dimensional finite difference methods are sufficient to understand the effect on seismic waves of the Los Angeles and San Fernando basins (Vidale and Helmberger, 1986b).

Observations of volcanic tremor at Mount St. Helens in April and May 1980

Abstract High-amplitude, nearly monochromatic volcanic tremor was observed before, during, and after the 18 May 1980 eruption of Mount St. Helens. The tremor often remained stationary for periods of several minutes or more. For the period before and immediately following the main eruption, spectra calculated from three-component seismograms contained a nearly constant dominant frequency of 1.1 to 1.5 Hz. This contrasts with a dominant frequency of 1.7 to 2.1 Hz calculated for the tremor associated with the 25 May eruption. Amplitude decay with distance obeys the form A ∼ r n , where A is the tremor amplitude and r is the observation distance from the source. The exponent n , was determined in a least-squares sense to be −1.5 ± 0.4, −1.5 ± 0.3, and −1.7 ± 0.3, before and shortly after the main eruption on 18 May 1980 and 24 May 1980, respectively. The exponent for low-frequency volcanic earthquakes before the main eruption was −1.6 ± 0.3, a value similar to that of the volcanic tremor and suggests that the source of the volcanic tremor is shallow as is the source of the low-frequency earthquakes. The spectra of the tremor periods are very similar to those of the low-frequency volcanic earthquakes. There are cases where volcanic tremor appears as a continuation of a low-frequency earthquake. These observations suggest that the volcanic tremor and the low-frequency volcanic earthquakes are strongly interconnected and indeed may well share a common source mechanism. Based on both signal stationarity and spectral amplitude decay consistency, we suggest, in common with others, that the monochromatic nature of volcanic tremor is mainly due to a source effect, while the path and station effects are relatively minor. The tremor amplitude changes as a result of variations in the source driving force. The frequency shift for the 24 May 1980 tremor may be due to a source dimension change .

Shear-wave polarizations in the Peter the First Range indicating crack-induced anisotropy in a thrust-fault regime

Summary. Three-component seismograms of small local earthquakes recorded in the Peter the First Range of mountains near Garm, Tadzhikistan SSR, display shear-wave splitting similar to that previously observed near the North Anatolian Fault in Turkey. The Peter the First Range is in a region of compressional tectonics, whereas the North Anatolian Fault is a comparatively simple strike-slip fault. Detailed analysis of the Turkish records suggests that the splitting is diagnostic of crack-induced anisotropy caused by vertical microcracks aligned parallel to the direction of maximum compression. Preliminary examination of paper records from Garm shows that most shear waves arriving within the shear-wave window display shear-wave splitting, and that the polarizations of leading shear-waves are consistently aligned in a NE/SW direction. The area is complicated and the tectonics are not well-understood, but the NE/SW direction is approximately perpendicular to the compressional axis in many of the fault-plane mechanisms of the earthquakes. These earthquakes are usually at depths between 5 and 12 km, although there are some deeper events nearby. Parallel shear-wave polarizations, such as those observed, are expected to indicate the strike of nearly vertical parallel microcracks, which would be aligned parallel to the direction of maximum compression. Thus the shear-wave polarizations in the Peter the First Range indicate that the directions of principal stress are reversed in the rock above the earthquake foci where thrust faulting is taking place.

Source parameters from locally recorded aftershocks of the 9 January 1982 Miramichi, New Brunswick, earthquake

AbstractSource parameters are derived from 111 locally recorded three-component seismograms of 40 aftershocks (magnitude = 1.0 to 3.5) of the 9 January 1982 Miramichi, New Brunswick, earthquake. This data set greatly expands the set of three-component data recorded at small source-to-receiver distances in eastern North America. Preliminary analysis showed that some aftershock spectra were modified by strong site effects. The need to understand and account for site effects, when analyzing the aftershock data in terms of simple source models, led to a two-part data analysis. First, site effects were modeled as a seismic-wave resonance in a thin layer (0 to 10 m) of low-seismic-impedance glacial till over a high-seismic-impedance bedrock half-space. Second, we calculated source parameters from SH-wave seismograms and amplitude spectra, incorporating what had been learned about site resonances. Correcting for obvious site effects partially succeeded in reducing scatter in the source-parameter determinations but some site-dependent discrepancies remain. Seismic moments range from 1 × 1018 to 5 × 1020 dyne-cm and corner frequencies from 5 to 45 Hz. Brune-model stress drops primarily range between 5 and 300 bars; stress drop apparently increases with increasing seismic moment. Trends in apparent stress calculated from radiated energy and seismic moment parallel trends in stress drop. Corner frequencies from these locally recorded data are greater than those derived from regionally recorded Lg waves for earthquakes of comparable moment. This result contradicts a recently proposed M0ω4 = constant source-scaling law based primarily on Lg data.

Inversion of arrival-times in a region of dilatancy anisotropy

Shear-wave splitting has been identified in many three-component seismograms from two separate field experiments on a section of the North Anatolian Fault in North-West Turkey. These observations are consistent with shear-wave propagation through a zone of extensive-dilitancy anisotropy. A preliminary attempt has been made to confirm this interpretation by simultaneously inverting suites of arrival-times for hypocentral locations and for parameters describing an anisotropic halfspace. Although the inversion procedure is not globally convergent, it is possible to recognize the true solution by systematically varying the initial conditions. Applied to selected data sets, the inversion defines several anisotropic models that fit the data significantly better than a simple isotropic model, and display the anisotropy required by the shear-wave splitting. However, most of these anisotropic models are not superior when they are used to individually locate events in a much larger data set. However, for each experiment, there is a single model that produces clearly superior locations for the larger data sets than those of other anisotropic or simple isotropic models. Both models display similar velocity variations which are characteristic of propagation through distributions of biplanar cracks displaying orthorhombic symmetry. The principal axes of the two models are oriented in similar directions and are within 20° of the principal axis of regional stress derived from fault-plane solutions. The solutions indicate low velocities close to the tensional axis, as would be expected in extensive-dilatancy anisotropy.

Depth dependence of source parameters at Monticello, South Carolina

Abstract Three estimates of stress differences, which include Brune stress drop, stress drop from rms of acceleration (arms), and the apparent stress, have been determined for 13 earthquakes at Monticello, South Carolina, a site of reservoir-induced seismicity. Data for nine of the events come from digitally recorded three-component seismograms at four or five stations that were deployed around the Monticello Reservoir in May and early June 1979. The data from the other four events come from a strong-motion accelerograph located on the dam abutment at the southwest end of the reservoir. Estimates of the seismic moment (Mo) range from 4.6 × 1017 to 3.4 × 1020 dyne-cm (S waves) and radiated energy from about 1011 to 3 × 1016 dyne-cm (S waves). Brune stress drops ranged from 0.5 bars to about 90 bars and show a strong dependence on depth (focal depths range from 0.07 to 1.4 km) and a moderate dependence on Mo. Arms stress drops from the direct S-wave span a similar range of values and also exhibit a strong dependence on depth. Apparent stresses are usually lower than the other estimates of stress differences by a factor of 2 to 4. Seismic stress differences are highest in the topmost 0.2 to 0.3 km, a depth range for which the in situ measurements of stress and pore pressure suggest that the rock is in a state of incipient failure. In this depth range, where the four largest events occurred, the stress drops are of the same order as the ambient shear stress. These data suggest that at Monticello, where pore fluids have a strong influence on the failure process, the largest stresses released seismically are in regions most conducive to failure and that the seismic efficiencies for events at Monticello are larger than have been reported for other tremors in different tectonic settings.

The anisotropic reflectivity technique: anomalous reflected arrivals from an anisotropic upper mantle

Anomalous body wave arrivals due to anisotropy in a subcrustal layer are investigated using synthetic seismograms. Previously, seismograms of such anomalous phases could only be calculated for plane wavefronts, which simulated teleseismic waves. An extension of the reflectivity technique to anisotropic structures now permits the use of a curved wavefront from a local or regional source. Explosion and double couple sources at the surface are used to generate P- and SV-waves, which are incident on simple stratified models of an anisotropic upper mantle. The anisotropic subcrustal layer possesses the same degree of velocity anisotropy as has been found under Western Germany. Anomalous P to SH converted waves, and shear wave splitting, are evident on the synthetic seismograms. Characteristics of the anomalous arrivals which depend on the velocity—depth structures of the models are discussed. It is shown how the velocity contrasts at the base of an anisotropic layer may allow shear wave splitting to remain undetected even for a high degree of anisotropy. These results from simple models seem to indicate that anomalous arrivals on individual three-component seismograms are seldom likely to be of large enough amplitude to be useful for the analysis of upper mantle anisotropy.

Premonitory seismic-wave phenomena before earthquakes near Fairview Peak, Nevada

abstract Anomalous variations in three different seismic processes have been observed before an earthquake of magnitude 4 on May 14, 1971 in the Fairview Peak region of central Nevada. The data used are the three-component seismograms from Tonopah (TNP) and Battle Mountain (BMN) as well as the vertical-component seismograms from several other seismographic stations. The observed precursory phenomena are (1) reorientation of the compressive stress axis: evidence for this is based on clear reversals of first motion of Pg at certain stations together with reversal of wave form of the SV component of Sg at BMN; (2) vertical migration of hypocenters: the PmP phase is often distinctly observed in central Nevada and the observed temporal variations in t(PmP)−t(Pg) at TNP indicate generally upward migration of foci before the main event; (3) changes in the extent of S-wave splitting: a large precursory increase is observed along the tensile stress direction and a small premonitory decrease along the compression direction. Similar but unidentical results have been obtained before another earthquake of magnitude 4 whereas all events of magnitude 3.5 or larger have been preceded by anomalous S-wave splitting along the tension direction. The various observed pre-earthquake processes appear to be interrelated and may be explained in terms of recent laboratory and theoretical results when applied to the tectonics of central Nevada. It seems highly desirable to attempt simultaneous observations of anomalous changes in more than one seismic process.

小金井における脈動観測結果について (III)

The most popular means to obtain the direction from which microseismic waves approach are the tripartite method and the method based on the orbital motion of the ground particle. However, it has been clarified by the theoretical researches on the Rayleigh-type dispersive waves that even among the waves having apparently Rayleigh-type orbits, there can exist two kinds of waves, one undergoing retrograde motion and the other undergoing progressive motion.In the present study, the writer examined the winter microseisms that were recorded during the period from November 1964 to February 1965. The directions of approach of these microseisms were determined by the tripartite method, and also the orbital motions of the ground surface particles were drawn on the basis of the three-component seismograms, so as to compare the results of the two different methods. In consequence, 221 waves were recognized to have typical Rayleigh-type orbits, about 40% of which have undergone retrograde motion and about 20% are of progressive motion. Thus, it has become clear that the directions from which microseismic waves approach should not be determined merely by their orbital motion, because it would involve a great danger.