Received May 11, 1989; Accepted July 10, * To whom correspondence should be addresseḍ

J. Phys. Earth, 38, 163-177, 1990 Teleseismic P- Wave Travel Time and Amplitude Anomalies Observed in Hokkaido Region, Japan Ichiro Nakanishi1,* and Yoshinobu Motoya2 1Department of Geophysics and 2Research Center for Earthquake Prediction, Faculty of Science, Hokkaido University, Sapporo 060, Japan P-wave travel time and amplitude anomalies from 59 teleseismic earthquakes are analyzed for 17 stations of a seismic network in Hokkaido region, Japan. Relative travel time and log amplitude variations within the network are parameterized by an azimuth-independent term (average), and first and second azimuthal terms. For the travel time, the azimuth-independent terms, which represent shallow near-station velocity anomalies, show an obvious geographical variation and a significant correlation with gravity (Bouguer) anomalies. The azimuth-independent terms vary from - 0.73 to 0.58 s. The slow directions of the first azimuthal terms for stations surrounding the Hidaka Mountains point consistently toward an area near the axis of the mountains, suggesting the existence of a low-velocity zone in the uppermost mantle beneath the Hidaka Mountains. For the amplitude, the azimuth-independent terms have large positive values (amplification) at stations where the geological setting is young, i.e., Quaternary. The first azimuthal terms for stations surrounding the Hidaka Mountains show some indication of large amplitudes for the incidence azimuth in the direction of the mountains. If the azimuth-independent terms are subtracted from the observed values, the travel time residuals and log amplitudeshow some correlation, suggesting the focusing and defocusing of rays by the heterogeneity in the uppermost mantle beneath the network. 1. Introduction Teleseismic P travel times have often been analyzed to investigate the structure of the crust and uppermost mantle beneath seismic arrays. Aki et al. (1977) inverted P travel time residuals to recover 3-D structure beneath the NORSAR array. Since then the 3-D inversion method has been applied to P-residual data from seismic arrays in many areas of the world. Amplitude variations were used to investigate the velocity heterogeneity beneath NORSAR in conjunction with the travel times by Haddon and Husebye (1978) by applying a parabolic approximation of the wave equation. Thomson and Gubbins (1982) analyzed the NORSAR amplitude data by using the bending method of ray tracing. To the author's knowledge, such studies of teleseismic P travel times as presented in this paper have been rare for the Japanese Islands, except Maki (1982), who studied Received May 11, 1989; Accepted July 10, 1990. * To whom correspondence should be addresseḍ 163

164 I. Nakanishi and Y. Motoya P-residuals reported in the JMA bulletins. For the Japan region, most inversion studies have used P travel times of local earthquakes (Aki and Lee, 1976) to investigate the crust and upper mantle structure beneath seismic networks. In this paper we measure P-wave travel times and amplitudes from teleseismic events recorded by a local seismic network of Hokkaido University and parameterize their variations within the network by azimuth-independent terms (averages) and azimuth-dependent terms. P travel times measured from four JMA high-gain seismographs (EMT 76) in the Hokkaido region as reported in the JMA bulletins are also analyzed for comparison. We compare travel time and log amplitude variations to seek an indication of focusing and defocusing effects caused by velocity heterogeneity. 2. Data We use short-period vertical-component seismograms recorded by a seismic network of the Research Center for Earthquake Prediction of Hokkaido University (RCEP). Figure 1 shows the locations of the stations. The network whose configuration is shown in this figure has operated since April 1985. The seismograms have been digitally recorded and stored on optical disks. For the travel time study, we also analyze data from four JMA (Japan Meteorological Agency) high-gain seismograph (EMT 76) stations (Fig. 1) as reported in the JMA bulletins. We listed earthquakes of magnitudes greater than 5.5 at teleseismic distances according to reports of the PDE (Preliminary Determination of Epicenters). Based on the examination of the seismograms reproduced from the optical disks by a computer of the RCEP, we selected the records on which both onset time and peak-peak amplitude of the first P-arrival were accurately measured. Figure 2 shows the epicenters of the earthquakes in a map of equidistant-azimuthal projection. As seen in the figure, the azimuthal coverage is not homogeneous. In the period range of this study (April 13, 1985-September 12, 1986), no earthquake was selected in azimuth ranges of ENE-ESE and NNW-NNE. Since the areas of these teleseismic azimuth ranges are seismically inactive, this situation of the azimuthal coverage will not be significantly improved if we extend the study period. 2.1 Travel time measurement We measured P-wave arrival times on the magnified plots of seismograms. The initial onsets of P-waves are well correlated among the stations. This helps us read accurately the arrival times on such noisy records as those from station HSS and TOI. The large noise amplitude is consistent with the amplification of P waves observed at these stations, which will be shown later (see Figs. 1 and 8(a)). The measurement accuracy is generally better than 0.1 s. The travel times are calculated by using source parameters taken from the PDE, and are corrected for station elevation and the Earth's ellipticity (Dziewonski and Gilbert, 1976). 2.2 Amplitude measurement We measured peak-peak amplitudes of the first cycle of P-waves on the seismograms J. Phys. Earth

Teleseismic P-Wave Travel Time and Amplitude Anomalies 165 Fig. 1. Map showing locations of Hokkaido University (solid circles) and JMA (solid triangles) stations. The shaded area indicates the Hidaka Mountains (H.M.). appropriately magnified so as to be able to measure the amplitude accurately. No correction is made for source mechanism. 2.3 Relative travel time residual We calculate P-residuals from the Jeffreys-Bullen Table (1940). These residuals include the effects of not only the heterogeneity in the crust and upper mantle beneath the seismic network but also errors in source parameters and velocity deviations from the Jeffreys-Bullen Table for paths from the sources to the bottom of the upper mantle beneath the seismic network. To minimize the effects of the latter two terms on the P-residuals, we subtract the average of the P-residuals for each event to calculate relative P-residuals Ė (hereinafter referred to as P-residuals): 2.4 Relative log amplitude anomaly We take the logarithm of observed amplitude, on the basis of the observation of Ringdal et al. (1972) that the amplitude data have a lognormal distribution. Vol. 38, No. 2, 1990

166 I. Nakanishi and Y. Motoya Fig. 2. Map showing epicenters of teleseismic events. Equidistant-azimuthal projection is used. The map center is the center of the seismic network (42.85, 142.58 ). The concentric circles correspond to epicentral distances of 30, 60, 90, and 100. Since the observed amplitudes are affected by the earthquake magnitude, we calculate the average log amplitude for each event and subtract it from the observed amplitudes to obtain relative log amplitudes A (hereinafter referred to as log amplitudes). 3. Azimuthal Variation of Travel Time and Amplitude Anomaly 3.1 Travel time residual Figure 3 presents the P-residuals Ė as a function of azimuth for two selected stations. The azimuth is measured clockwise from the north. The P-residuals of each station seem to show a slow azimuthal variation, though there exists some scatter about the slow variation. For example, at station IWN (see Figs. 1 and 3(b)) P-waves from the NE have a negative residual of about -1 s, but those from the SSW show a slight positive residual. From this observation we may expect the existence of a high-velocity J. Phys. Earth

Teleseismic P-Wave Travel Time and Amplitude Anomalies 167 Fig. 3. P-residual as a function of azimuth. The solid curve is the least-squares fit of equation Ė=a0+a1 cos(ľ-ľ1)+a2 cos 2(Į-Į2) to the observed P-residuals (open circles). The solid straight line is the constant term A0. (a) ESH. (b) IWN. zone to the NE or of a low-velocity zone to the SSW in the uppermost mantle beneath an area near IWN. Another interpretation is an inclination of Moho to the SSW. We analyze these data from 17 stations by adopting a procedure originally suggested by Cleary and Hales (1966) and recently used by Dziewonski and Anderson (1983) in their study of P arrival times from the Bulletins of the International Seismological Centre. We assume the P-residual r at a station has the form Ė = A0 + A1 cos(ľ - ľ1) + A2 cos 2(Į - ľ2), where ľ is the azimuth from the station to an event, and the azimuths 1 and ľ2 denote the slow directions. In the second azimuthal term, the ľazimuth (02 + 180 ) is another slow direction. We make a least-squares fit of the above equation of azimuthal variation to the P-residuals for each station. Table 1 lists the five parameters A0, A1, A2, ľ1, and ľ2 thus determined, standard deviation before (s1) and after (s2) the fit, and the variance Vol. 38, No. 2, 1990

I 168. Nakanishi and Y. Motoya reduction (vr =1 (s2/si)2). In Fig. 3, the least-squares fit of the equation to, the data is indicated by the solid curve. Also indicated by the horizontal straight line in the figure is the azimuth-independent term A0. In the Hokkaido region, the JMA operates high-gain seismographs at four stations (Fig. 1). As a comparison we analyze travel time data from these seismographs, as reported in the JMA bulletins, in the same way as described above in conjunction with RCEP data. The result is presented in Table 1 and Fig. 4. Figure 4(a) and (b) compare the azimuthal variations of P-residuals for stations AIB and ASAJ, which are located close to each other (Fig. 1). Figure 4(c) and (d) are for stations MYR and HOOJ (Fig. 1). As can be seen in the figure, the JMA stations and the RCEP stations show similar azimuthal variations. In Fig. 4, however, it is clear that the first and second azimuthal terms are generally larger for the JMA stations than for the RCEP stations. These apparent larger values of the azimuthal terms for the JMA stations may be caused by the larger scatters in the observed travel times. Figure 5(a), (b), and (c) show the geographical variations of the azimuthindependent terms (A0), the first azimuthal terms (A1 cos(ľ - ľ1)), and the second azimuthal terms (A2 cos 2(Į - ľ2)), respectively. The azimuth-independent terms show an obvious geographical variation as can be seen in Fig. 5(a). The Pacific side of the Kurile arc part of Hokkaido shows negative Table 1. P travel time residual. J. Phys. Earth

Teleseismic P-Wave Travel Time and Amplitude Anomalies 169 Fig. 4. Comparison of P-residual Vol. 38, the neighboring No. 2, 1990 RCEP and JMAstations. AIB (a) and ASAJ (b). MYR (c) and HOOJ (d).

170 I. Nakanishi and Y. Motoya Fig. 5. Map showing the geographical variations of P-residuals of the RCEP and JMA stations. (a) Azimuth-independent term. The size of the symbols is proportional to the anomaly according to the scale indicated. (b) First azimuthal term. The arrow points toward the slow direction and its length is proportional to the anomaly according to the scale indicated. (c) Second azimuthal term. The bar attached to the station location is parallel to the slow direction and its length is proportional to the anomaly according to the scale indicated. anomalies (NMR, AKK, TES, URH, IWN, MYR), the Japan Sea side of that part of Hokkaido shows positive anomalies (KNP, AIB, TOI), the south-western side of the Hidaka Mountains shows positive anomalies (ERM, KMU, MUJ), and the southwestern peninsula region shows no anomalies (HSS, IMG, KKJ) except ESH. Utsu (1975) and Maki (1982) obtained azimuth-independent residuals by analyzing P-wave travel times observed in the Japanese Islands from nearby deep earthquakes and teleseismic earthquakes, respectively. Since they obtained the residuals from low-sensitivity JMA stations, with the exception of KMU (Kamikineusu) and HSS (Misumai), it is difficult to make a direct comparison of Fig. 5(a) and their results, and only an overall pattern must be compared. Looking at Fig. 5(a) of the present paper, Fig. 6 of Utsu, and Fig. 10 of Maki, we find that the residuals of the three studies J. Phys. Earth

Teleseismic P-Wave Travel Time and Amplitude Anomalies 171 show a similar geographical variation. There is one difference between the present and the previous studies. Station IWN (Iwanai), which is located on the northeastern side of the Hidaka Mountains (Fig. 1), shows a negative residual of - 0.56.s (Table 1). The residuals obtained by Utsu and Maki for OBI (Obihiro), which is located 32 km to the NNE from IWN, indicate a positive value of about 0.5 s. The negative anomaly suggests the existence of a high-velocity zone in the crust right beneath IWN, which is probably related to the structure and history of the Hidaka Mountains. The positive anomaly of OBI may be caused by a thick sedimentary layer as evidenced by a large amplification factor at this station (Okada and Kagami, 1978). The most striking feature in the map of the first azimuthal terms (Fig. 5(b)) is that the slow directions for the stations surrounding the Hidaka Mountains (ERM, KMU, MUJ, HIC, IWN, and MYR) point toward the axis of the mountains. This observation may suggest the existence of a low-velocity zone in the uppermost mantle beneath the mountains. Takanami (1982) and Miyamachi and Moriya (1984) inferred a steeply dipping low-velocity zone beneath the western side of the mountains based on threedimensional inversions (Aki and Lee, 1976) of local earthquakes. The second azimuthal terms (Fig. 5(c)) are generally small in amplitude, except ERM, ASAJ, and HOOJ, and do not show any obvious geographical variations. This might be due to insufficient azimuthal coverage of the epicentral distribution (Fig. 2). The azimuth-independent terms may have their origins in the crust just beneath the stations. The gravity anomalies are generally interpreted in terms of crustal structure. Thus we can expect a correlation between both quantities. The azimuth-independent terms may be correlated with the Bouguer anomalies (Geographical Survey Institute, 1955; Mori, 1965) as shown in Fig. 6. In the figure we also notice that two data points (250 and 200 mgal) seem to deviate from the correlation. Those stations (NMR and AKK) are located on the most Pacific side of eastern Hokkaido. The gravity anomaly of this region has been well recognized as the highest in Japan (Geographical Survey Institute, 1955; Mori, 1965). 3.2 Log amplitude anomaly Figure 7(a) and (b) are the log amplitude anomalies A as a function of azimuth for two stations surrounding the Hidaka Mountains. We analyze the log amplitude data A assuming a form similar to that used for the P residuals: A = A0 + A1 cos(ľ - ľ1) + A2 cos 2(Į - ľ2), where ľ is the azimuth from the station to an event, and the azimuths 1 and ľ2 denote the largest amplitude directions. The results of a least-squares ľ fit of this equation to the observed log amplitudes are shown in Table 2 and Fig. 7. Figure 8(a) shows the azimuth-independent terms of the log amplitude variations. Since the amplitude is expected to be affected by local structure more strongly than the travel time, we study the relation of the amplitude with the local geology near the stations, which is summarized in Table 3. Here we point out an obvious feature that the places at which the stations with large azimuth-independent terms are located are characterized by young geology, i.e., Quaternary for HSS, TES, and IMG, and Neogene for TOI. Station MUJ shows no large amplitude anomaly in spite of the young geology reported near the station. A detailed comparison of site response with local geology is Vol. 38, No. 2, 1990

172 I. Nakanishi and Y. Motoya Fig. 6. Azimuth-independent P-residuals versus Bouguer anomalies for the RCEP stations. The Bouguer anomalies are based on the data of the Geographical Survey Institute (1955) and Mori (1965). beyond the scope of this paper. Figure 8(b) presents the first azimuthal terms. We notice that for the three stations surrounding the Hidaka Mountains (HIC, MUJ, and MYR) the directions of large amplitudes point toward the mountains. For these stations the first azimuthal terms are the largest of the three terms (see Table 2). The other three stations in the Hidaka region (IWN, KMU, and ERM) do not show these features. This observation seems to be consistent with the results of the first azimuthal terms of the travel time residuals, from which focusing of rays by a low-velocity zone beneath the Hidaka Mountains is expected. 4. Correlation between Travel Time and Amplitude Anomaly The above results of azimuthal variations of travel time and log amplitude at the stations surrounding the Hidaka Mountains suggest the occurrence of focusing of rays by a low-velocity zone beneath the mountains. To investigate this point more closely we correlate the travel time residuals with the log amplitude anomalies. Figure 9(a) shows a plot of original travel time residuals vs. log amplitude anomalies. This figure J. Phys. Earth

Teleseismic P-Wave Travel Time and Amplitude Anomalies 173 Fig. 7. Log amplitude anomaly as a function of azimuth. The solid curve is the least-squares fit of equation A=A0+Ai cos(ľ-ľ1)+a2 cos 2(Į-Į2) to the observed log amplitudes (open circles). The solid straight line is the constant term A0. (a) HIC. (b) MYR. does not seem to suggest any significant correlations between them. Comparison of Fig. 8(a) and Table 3 may suggest the amplitudes are strongly affected by local geology. The amplification or deamplification due to the local geology is expected to cause azimuth-independent amplitude variations, because the incidence angles of rays from teleseismic events are close to the vertical at the surface, especially in the case of a low-velocity sedimentary layer. Figure 9(b) shows a similar plot of the correlation between the travel time and amplitude but with the correction of A0 terms. There seems to exist some, correlation between the two observations. Table 4 lists the correlation coefficients for all cases of the corrections. Vol. 38, No. 2, 1990

174 I. Nakanishi and Y. Motoya Table 2. Log amplitude anomaly. Fig. 8. Map showing the geographical variations of log amplitude anomalies of the RCEP stations. (a) Azimuth-independent term. The size of the symbol is proportional to the anomaly according to the scale indicated. (b) First azimuthal term. The arrow points toward the large amplitude direction and its length is proportional to the anomaly according to the scale indicated. J. Phys. Earth

Teleseismic P-Wave Travel Time and Amplitude Anomalies 175 Table 3. Geology at stations, Fig. 9. Plots of P-residual versus log amplitude. r refers to the correlation coefficient. (a) Original P-residual and log amplitude. (b) Those corrected for A0 terms. 5. Conclusions We have studied the azimuthal variations of the travel time residuals of teleseismic P-waves observed in Hokkaido, Japan. The azimuth-independent terms show an obvious Vol. 38, No. 2, 1990

176 I, Nakanishi and Y. Motoya Table 4. Correlation coefficient r of travel time residual and log amplitude anomaly. geographical variation and a significant negative correlation with gravity (Bouguer) anomalies. This variation may be caused by the heterogeneity in sedimentary layers and crust. The first azimuthal terms of 6 stations surrounding the Hidaka Mountains are characterized by the slow directions pointing to the axis of the mountains. This observation suggests the existence of a low-velocity zone in the uppermost mantle beneath the mountains. We have analyzed the amplitude variations in the same way as the travel times. The azimuth-independent terms seem to be closely related to. the local geology of the station sites. The stations located above Quaternary (Recent) deposits exhibit large amplification. The first azimuthal terms of 3 stations surrounding the Hidaka Mountains show that the rays from the directions of the mountains are amplified. The travel time and amplitude anomalies show a correlation if we correct for their azimuth-independent terms. This result may be explained by the focusing and defocusing due to velocity anomalies beneath the network. This work was supported by a Grant in Aid from the Ministry of Education, Science and Culture of Japan (No. 62540285). REFERENCES Aki, K. and W. H. K. Lee, Determination of three-dimensional velocity anomalies under a seismic array using first P-arrival times from local earthquakes. 1. A homogeneous initial model, J. Geophys. Res., 81, 4381-4399, 1976. Aki, K., A. Christofferson, and E. S. Husebye, Determination of the three-dimensional seismic structure of the lithosphere, J. Geophys. Res., 82, 277-296, 1977. Cleary, J. and A. L. Hales, An analysis of the travel times of P waves to North American stations, in the distance range 32 to 100, Bull. Seismol. Soc. Am., 56, 467-489, 1966. Dziewonski, A. M. and D. L. Anderson, Travel times and station corrections for P waves at teleseismic distances, J. Geophys. Res., 88, 3295-3314, 1983. Dziewonski, A. M. and F. Gilbert, The effect of small, aspherical perturbations on travel times and a reexamination of the corrections of ellipticity, Geophys. J. R. Astron. Soc., 44, 7-17, 1976. J. Phys. Earth

Teleseismic P-Wave Travel Time and Amplitude Anomalies 177 Geographical Survey Institute, Gravity survey in Japan: (I) Gravity survey in Hokkaido district, Bull. Geograph. Surv. Inst., 4, Part 2, 23-99, 1955. Geological Survey of Hokkaido, Geology and resources of Hokkaido, Japan I: Geological map of Hokkaido (1: 600,000), 1980. Haddon, R. A. W. and E. S. Husebye, Joint interpretation of P-wave time and amplitude anomalies in terms of lithospheric heterogeneities, Geophys. J. R. Astron. Soc., 55, 19-43, 1978. Jeffreys, H. and K. E. Bullen, Seismological Tables, British Association for the Advancement of Science Gray Milne Trust, London, 1940. Maki, T., Residuals of teleseismic P-wave travel times observed in Japan, Bull. Earthq. Res. Inst., 57, 151-191, 1982. Miyamachi, H. and T. Moriya, Velocity structure beneath the Hidaka Mountains in Hokkaido, Japan, J. Phys. Earth, 32, 13-42, 1984. Mori, T., Gravity anomalies in the Konsen plain, Geophys. Bull. Hokkaido Univ., 14, 59-71, 1965 (in Japanese with English abstract). Okada, S. and H. Kagami, A point-by-point evaluation of amplification characteristics in Japan on 1-10 sec seismic motions in relation to deep soil deposits, Trans. Archit. Inst. Jpn., No. 267, 29-38, 1978 (in Japanese with English abstract). Ringdal, F., E. S. Husebye, and A. Dahle, Event detection problems using a partially coherent seismic array, NTNF/NORSAR Techn. Rep., No. 45, 1972. Takanami, T., Three-dimensional seismic structure of the crust and upper mantle beneath the orogenic belts in southern Hokkaido, Japan, J. Phys. Earth, 30, 87-104, 1982. Thomson, C. J. and D. Gubbins, Three-dimensional lithospheric modelling at NORSAR: Linearity of the method and amplitude variations from the anomalies, Geophys. J. R. Astron. Soc., 71, 1-36, 1982. Utsu, T., Regional variation of travel-time residuals of P waves from nearby deep earthquakes in Japan and vicinity, J. Phys. Earth, 23, 367-380, 1975. Vol. 38, No. 2, 1990