Coseismic and postseismic crustal deformation after the M w 8 Tokachi-oki earthquake in Japan

Transcription

1 LETTER Earth Planets Space, 56, , 4 Coseisic and postseisic crustal deforation after the M w 8 Tokachi-oki earthquake in Japan Shinzaburo Ozawa, Masaru Kaidzu, Makoto Murakai, Tetsuo Iakiire, and Yuki Hatanaka Geographical Survey Institute of Japan (Received February 9, 4; Revised June 11, 4; Accepted June 11, 4) The peranent Global Positioning Syste (GPS) array in Japan detected coseisic and postseiic deforation of the 3 Tokachi-oki earthquake (M w 8). We estiate the tie evolution of its postseisic slip, together with its coseisic slip distribution. The result shows that the postseisic slip has been occurring ainly in an area adjacent to the coseisic slips, propagating to the northeast and southwest. This suggests that, as of March 6, 4, the postseisic slip of the strongly coupled area neighboring the coseisic rupture partly released seisic oent, equivalent to an earthquake of M w 7.8. Key words: Tokachi-oki earthquake, coseisic deforation, afterslip, postseisic deforation. 1. Introduction The Pacific Plate is subducting westward beneath the North Aerican or Okhotsk Plate at a rate of 1 c/year at the Kurile Trench southeast of Hokkaido, Japan (Fig. 1). This area has been repeatedly struck by large offshore thrust earthquakes (Fig. 1(b)). The segent off the Tokachi area ruptured in the 195 M w 8. earthquake, and the probability of the next earthquake within a 3 years period fro 3 has been estiated to be 6% (Earthquake Research Coittee, 4). An earthquake of M w 8 occurred on Septeber 6, 3 (local tie) with the ruptured area siilar to the 195 event. The strike, dip, and rake angles of the coseisic rupture are inferred to be 3,, and 19, respectively, indicating that this was a typical interplate thrust earthquake (Yaanaka and Kikuchi, 3). At GPS stations near the epicenter, postseisic displaceent had exceeded 1 c by March 6, 4, following the coseisic displaceent up to 1. It is iportant to investigate where and how the postseisic slip has been occurring, because it provides inforation on the stress change over tie and frictional characteristics of the plate interface (e.g., Heki et al., 1997; Nishiura et al., ; Yagi et al., 1). In this study, we investigate the coseisic slip distribution and copare it with the spatioteporal evolution of the afterslip obtained using the tiedependent inversion ethod (Segall and Matthews, 1997).. Data and Analytical Procedure The dense GPS array (GPS Earth Observation Network; GEONET), with dual-frequency P-code receivers, of the Geographical Survey Institute of Japan (GSI) has been in operation since GPS data are analyzed using the Bernese GPS software version 4. with precise epheerides and Earth orientation paraeters fro the International GPS ser- Copy right c The Society of Geoagnetis and Earth, Planetary and Space Sciences (SGEPSS); The Seisological Society of Japan; The Volcanological Society of Japan; The Geodetic Society of Japan; The Japanese Society for Planetary Sciences; TERRA- PUB. vice for Geodynaics (IGS). Tropospheric delays are estiated at each station every three hours (Hatanaka, 3). Since the raw data include periodic (largely annual) and secular coponents, we estiated the using the data between 1999 and and reoved the as described in Ozawa et al. (4). Figure 1(b) shows the velocities during the period 1999 relative to 9541 (Fig. 1(a)). Coseisic crustal oveents were estiated as offsets in the de-trended coordinate tie series. Figure (a) shows the coseisic displaceents relative to 9541 (Fig. 1). Southeastward horizontal otion up to 1 and subsidence up to c are observed near the epicenter. Postseisic oveents see to have started iediately after the earthquake (Fig. 3). Postseisic oveents, which are also relative to 9541, are larger in the areas surrounding the epicenter rather than in the epicentral region (Fig. 3), suggesting that afterslip occurred ainly in the fault area surrounding the coseisic rupture. Figure 4 shows postseisic vertical oveents of GPS sites relative to 9541 in the period fro Septeber 6, 3 to March 6, 4. The Pacific coastal area shows uplift of up to 3 c. Considering a systeatic spatial pattern of uplifting in this region, we think the observed upheaval reflects true crustal deforation, though such a sall uplift ight be within the range of vertical observation uncertainties of GPS. This ay suggest the occurrence of postseisic slip in a deeper part of the plate interface. Figure 5 shows the de-trended tie series at selected GPS sites shown in Fig. 3(a) after the reoval of coseisic offsets. GPS sites show postseisic oveents that decay with tie. Using the coseisic deforation data, we estiated the coseisic slip distribution following the ethod of Ozawa et al. (1) based on Yabuki and Matsu ura (199). We used the plate interface configuration given in Katsuata et al. (3) (Fig. 1(b)). By representing the fault slip as a superposition of B-spline functions on the plate boundary, we estiated the optial slip distribution by balancing the good- 675

2 676 S. OZAWA et al.: CRUSTAL DEFORMATION FROM TOKACHI-OKI EARTHQUAKE 45N (a) 46N (b) 1 c/year 4N Aurian plate 44N 9541 Japan trench C 35N Phillipine Sea Pacific plate plate 3N 13E 135E 14E 145E 4N 5 k õ Tokachi area Kuril trench - c/year 14E 14E 144E 146E 148E A B Fig. 1. (a) Tectonic setting of Hokkaido, Japan. Open square shows the location of 9541 which is used as a reference point for ground otion. (b) Enlarged ap of the rectangular area in (a). Large offshore earthquakes have repeatedly occurred off the east coast of Hokkaido. A: 195 Tokachi-oki earthquake (M w 8.), B: 1973 Neuro-hanto-oki earthquake (M w 7.8), C: 1969 Chishia earthquake (M w 8.). Segentation along the trench axis is approxiately shown by solid lines. Red arrows show interseisic horizontal displaceents while color represents vertical otion relative to 9541 (1999 ) in c/year. Broken lines show isodepth contours of the plate boundary between the subducting and the overriding plates (Katsuata et al., 3). Isodepth contour interval is 1 k. 46N 1 (a) (b) 1 44N 4N -1 - c 3 Tokachi-oki c 3 Tokachi-oki 14E 14E 144E 146E14E 14E 144E 146E Fig.. (a) Observed coseisic deforation fro the 3 Tokachi-oki earthquake relative to Contours show vertical coseisic otion with an interval of 1 c with relative to Star shows the epicenter of the 3 Tokachi-oki earthquake as identified by the Meteorological Agency of Japan. (b) crustal deforation calculated using the odel in Fig. 6(a). 46N 6/SEP/3-/OCT/3(a) obs calc /OCT/3-1/NOV/3 (b) 1/NOV/3-6/MAR/4 (c) 5 c 44N 4N E 144E 148E 14E 144E 148E 14E 144E 148E Fig. 3. Observed postseisic deforation relative to Open circles show the location of the selected GPS sites whose tie series are shown in Fig. 5. (a) Septeber 6 October, 3. (b) October Noveber 1, 3 (c) Noveber 1, 3 March 6, 4. Solid arrows show observations, while white arrows represent values calculated using the odel in Figs. 6(b) (d).

3 S. OZAWA et al.: CRUSTAL DEFORMATION FROM TOKACHI-OKI EARTHQUAKE 677 obs 5 c..1 (a)15 Tie after the earthquake (days) c 14E 144E 148E Fig. 4. Observed postseisic vertical deforation for the period between Septeber 6, 3 and March 6, 4 relative to Solid arrows show observations, while contours show values calculated using the odel in Figs. 6(b) (d). Broken-line contour indicates subsidence, whereas solid-line contour shows uplift. ness of fit to the data and the soothness of fault distribution based on the Akaike s Bayesian Inforation Criteria (ABIC) (Akaike, 1974; Yabuki and Matsu ura, 199). We iposed non-negative constraints of the slip coponents assuing eastward and southward slips positive. Green s functions of surface deforation are taken fro Yabuki and Matsu ura (199). We used east-west (), north-south (), and up-down () coponents at the 85 GPS sites shown in Fig. (a). After estiating the coseisic slip distribution, we applied square-root inforation filtering (see Appendix) based on the tie-dependent inversion technique for the tie series in Fig. 5 for the period between Septeber 4, 3 and March 6, 4. In this inversion analysis, we set the constraint that aseisic fault otion is eastward and southward, as was the case for the coseisic slip estiate, using the hard constraint ethod of Sion and Sion (3). Figure 3 shows the GPS sites used in the filtering analyses. 3. Results and Discussion The estiated coseisic slips are ainly located off the coast of the Tokachi area, Hokkaido, with the epicenter deterined by the Meteorological Agency of Japan (JMA) at its eastern edge (Fig. 6(a)). Our odel reproduces the observed crustal oveents well (Fig. (b)), and is consistent with past reports of coseisic slip distributions fron seisological (e.g., Yaanaka and Kikuchi, 3; Yagi, 4; Honda et al., 4; Kaae and Kawabe, 4; Koketsu et al., 4) and tsunai wavefor (Tanioka et al., 4) studies. With regard to afterslip distribution, the slip ainly occurs in an area adjacent to the coseisic slip region over the entire tie interval (Figures 6(b) (d)). Fro Septeber 6 to October, the afterslip is distributed surrounding the coseisic slips (Fig. 6(b)). The afterslip expanded toward the northeast -. 1/9/3 1/11/3 1/1/4 1/3/ (b) /9/3 1/11/3 1/1/4 1/3/ (c) /9/3 1/11/3 1/1/4 1/3/ (d) /9/3 1/11/3 1/1/4 1/3/4 Day/Month/Year Fig. 5. De-trended tie series at the selected GPS sites in Fig. 3(a) without coseisic jups of the 3 Tokachi-oki earthquake, relative to 9541.,, and represent east-west, north-south and up-down coponents with east, north, and up being positive. Solid lines indicate values calculated using the estiated odel in Figs. 6(b) (d). Labels 1 indicate nuber of days fro Septeber 6, 3. (a) 15, (b) 11, (c) 138, (d) 53.

4 678 S. OZAWA et al.: CRUSTAL DEFORMATION FROM TOKACHI-OKI EARTHQUAKE 43N 4N 41N coseisic 6/SEP/3-/OCT/3 /OCT/3-1/NOV/3 1/NOV/3-6/MAR/4 1 5 c 5 c 5 c 4 c 4 c 4 c 4 k (a) (b) (c) (d) 4 k 4 k 4 k 14E 144E 146E14E 144E 146E 14E 144E 146E 14E 144E 146E Fig. 6. Estiated coseisic and postseisic slip distributions. Open circles represent aftershocks (Data fro the Meteorological Agency of Japan). Broken lines represent isodepth contours of the plate boundary with 1 k intervals. (a) Coseisic slip distribution. The star represents the epicenter of the 3 Tokachi-oki earthquake as identified by the Meteorological Agency of Japan. Contour interval is. (b) Postseisic slip for the period Septeber 6 October, 3. Contour interval is 4 c. (c) Postseisic slip for the period October Noveber 1, 3. Contour interval is 4 c. (d) Postseisic slip for the period Noveber 1, 3 March 6, 4. Contour interval is 4 c.

5 S. OZAWA et al.: CRUSTAL DEFORMATION FROM TOKACHI-OKI EARTHQUAKE 679 and southwest in the period between October and Noveber 1 (Fig. 6(c)). After Noveber 1, a part of it sees to have coe back to the coseisic slip areas (Fig. 6(d)). Such a teporal evolution of the afterslip is consistent with the distribution of aftershocks that ainly occurred in a peripheral area of the coseisic slip region (Shinohara et al., 4; Ito et al., 4). In particular, the northeastward propagation of the aftershock distribution is in good agreeent with the afterslip igration in Fig. 6(c). In accordance with the decay of aftershocks over tie, the oent release rate of the afterslip decreases fro Septeber 6, 3 to March 6, 4. Figures 3, 4, and 5 show that the observed and odeled crustal oveents are highly consistent. The present result supports past works (e.g., Yagi and Kikuchi, 3) that afterslip and coseisic slip distributions are copleentary. In our study, aseisic slip is suggested to have extended to a relatively deep part of the plate boundary. This suggests that the coupling area in Hokkaido extends deeper ( 7 k) than the down-dip end of the slip deficit (3 5 k) suggested in past studies (e.g., Ito et al., ; Mazzotti et al., ). In fact, Murakai and Ozawa (in preparation) proposed a coupling odel, where slip deficits extend to a deeper portion of the Hokkaido area slab ( ), based on GPS data. According to their result, the slip deficits beneath the Cape Erio and the Tokachi area are estiated to be around 4 c/year. This iplies that the cuulative slip deficit aounts up to for the period The present study shows afterslip of 4 6 c, uch less than, in these areas. This ibalance is also seen in a fact that postseisic uplift of 3 c cannot copensate for the cuulative interseisic subsidence (Fig. 1(b)) and the coseisic subsidence in the Tokachi/Erio area (Fig. (a)). However, since the postseisic oveents still go on and there exist uncertainties of c/year in the slip deficit rate in these areas, we cannot rule out the possibility that the current afterslip ay eventually release the energy accuulated over the interseisic period. It would be at least reasonable to say that the part deeper than the presently believed seisogenic depth had accuulated strain before the 3 earthquake, because this earthquake is now releasing the strain. Fro this viewpoint, the present results suggest that interplate coupling occurs deep beneath the Cape Erio and the Tokachi region, though the quantitative coparison of the slip deficit and the entire afterslip requires further studies. Although the northeastern part of the coseisic slip area is thought to be coupled strongly during an interseisic period, it reained intact over the coseisic period of the 3 Tokachi-oki earthquake. Consequently, the northeastern part of section A in Fig. 1(b) requires soe echanis to release strain energy, part of which is now being copensated by afterslip of the 3 earthquake. Since energy equivalent to that of the ain shock is often released by postseisic slip in interplate thrust earthquakes (e.g., Heki et al., 1997; Nishiura et al., ), this area ay release enough strain energy by the on-going afterslip, i.e this section ay not rupture as an ordinary earthquake. At present, the oent agnitude of the postseisic slip of the 3 Tokachioki earthquake is around 7.8. Since the oent release rate is becoing roughly constant as of March 4, it will probably take ore tie for postseisic slip to coe to an end, and this will enable us to grasp overall understanding of the 3 Tokachi-oki earthquake and its postseisic behavior. Acknowledgents. We used focal echanis solutions published by the Meteorological Agency of Japan on its web site. We are grateful to Dr. Munekae of the GSI for helpful discussion. Appendix. We used square-root inforation filter (Bieran, 1977) instead of ordinary Kalan filtering schee, incorporating spatial soothing equations that are shown later and slightly different fro those in Ozawa et al. (4) in a transition equation. We adopt the following state vector x n n = (u,v,p 1, p,...,p L ), (A.1) where u, v, and p i represent fault slip, slip velocity, and a rando walk of GPS site with station nuber of i, respectively. In this representation, GPS site is assued to ove as Brownian over tie. We odify the transition equation in square-root inforation filtering by adding spatial soothing equations that are in the third and fourth rows of the following description: W 1 1 F 1 1 F 1 MJ α ω 1 ω x n+1 n = 1 x n n MJx n n, I t (A.) where ω 1, ω, x n+1 n, and x n n, represent syste noise of an ordinary transition equation in tie, syste noise of spatial soothing, the one-step-ahead predicted state, and a state at tie n. M, J, and F represent the soothing atrix in space, the atrix used to select the slip coponent fro a state, and the transition atrix in an ordinary transition equation, respectively. The variables t, α, and I represent the lapse tie between step n and step n + 1, the soothing paraeter in space, and the identity atrix, respectively. W 1 is the upper triangular atrix with WW T equaling the covariance of syste noise in tie, where the superscript T eans a transpose atrix. 1 is the upper triangular atrix with R 1 R1 T equaling V n n which is the covariance of x n n. is a atrix with MJV n n JM T = R R T with only diagonal coponents. Fixed interval soothing is described by Ozawa et al. (4). In this research, we adopted Sion and Sion s ethod of hard constraints (3) to incorporate inequality constraints. We use the conventional forula of the log likelihood of Kalan filtering without adding ters related to the initial state, as discussed by Ozawa et al. (4). Coputation of the log likelihood under inequality constraints is described by Ozawa et al. (4). References Akaike, H., A new look at the statistical odel identification, IEEE, Trans. Auto. Control, AC-19, , Bieran, G. J., Factorization ethods for discrete sequential estiation, Acadeic Press, New York, Earthquake Research Coittee, Long ter evaluation of the seisic activity along the Chishia trench, Rep. Earthquake Res. Coittee, 4 (in press). Hatanaka, Y., Iproveent of the analysis strategy of GEONET, Bull. Geograph. Survey Inst., 49, 11 37, 3. Heki, K., S. Miyazaki, and H. Tsuji, Silent fault slip following an interplate thrust earthquake at the Japan Trench, Nature, 386, , 1997.

6 68 S. OZAWA et al.: CRUSTAL DEFORMATION FROM TOKACHI-OKI EARTHQUAKE Honda, R. et al., Ground otion and rupture process of the 3 Tokachioki earthquake obtained fro strong otion data of K-NET and KiK-net, Earth Planets Space, 56, 317 3, 4. Ito, T., S. Yoshioka, and S. Miyazaki, Interplate coupling in northeast Japan deduced fro inversion analysis of GPS data, Earth Planet. Sci. Lett., 176, ,. Ito, Y. et al., Spatial distribution of oent tensor solutions of the 3 Tokachi-oki earthquake (MJMA=8.) and aftershocks, Earth Planets Space, 56, 31 36, 4. Kaae, K. and H. Kawabe, Source odel coposed of asperities for the 3 Tokachi-oki, Japan, earthquake (MJMA=8.) estiated by the epirical Green s function ethod, Earth Planets Space, 56, 33 37, 4. Katsuata, K., N. Wada, and M. Kasahara, Newly iaged shape of the deep seisic zone within the subducting Pacific plate beneath the Hokkaido corner, Japan-Kurile arc-arc junction, J. Geophys. Res., 18, 565, doi:1.19/jb175, 3. Koketsu, K., K. Hikia, S. Miyazaki, and S. Ide, Joint inversion of strong otion and geodetic data for the source process of the 3 Tokachi-oki, Hokkaido, earthquake, Earth Planets Space, 56, , 4. Mazzotti, S., X. L. Pichon, H. Pierre, and S. Miyazaki, Full interseisic locking of the Nankai and Japan-west Kuril subduction zones: an analysis of unifor elastic strain accuulation in Japan constrained by peranent GPS, J. Geophys. Res., 15, ,. Nishiura et al., Distribution of seisic coupling on the subducting plate boundary, in northeastern Japan inferred fro GPS observations, Tectonophysics, 33, 18 38,. Ozawa, S., M. Murakai, and T. Tada, Tie-dependent inversion analysis of the slow thrust event in the Nankai trough subduction zone, southwestern Japan, J. Geophys. Res., 16, 787 8, 1. Ozawa, S., S. Miyazaki, T. Nishiura, M. Murakai, M. Kaidzu, T. Iakiire, and X. Ji, Creep, dike intrusion and aga chaber deflation odel for the Miyake eruption and the Izu islands earthquakes, J. Geophys. Res., 4 (in press). Segall, P. and M. Matthews, Tie dependent inversion of geodetic data, J. Geophys. Res., 1, , Shinohara, M. et al., Aftershock observation of the 3 Tokachi-oki earthquake by using dense ocean botto seisoeter network, Earth Planets Space, 56, 95 3, 4. Sion, D. and D. L. Sion, Aircraft turbofan engine health estiation using constrained Kalan filtering, ASME Turbo Expo 3, Atlanta, GA, paper GT , 3. Tanioka, Y., K. Hirata, R. Hino, and T. Kanazawa, Slip distribution of the 3 Tokachi-oki earthquake estiated fro tsunai wavefor inversion, Earth Planets Space, 56, , 4. Yabuki, T. and M. Matsu ura, Geodetic data inversion using Bayesian inforation criteria for spatial distribution of fault slip, Geophys. J. Int., 19, , 199. Yagi, Y., Source rupture process of the 3 Tokachi-oki earthquake deterined by joint inversion of teleseisic body wave and strong ground otion data, Earth Planets Space, 56, , 4. Yagi, Y. and M. Kikuchi, Partitioning between seisogenic and aseisic slip as highlighted fro slow slip events in Hyuga-nada, Japan, Geophys. Res. Lett., 3, 187, doi:1.19/gl15665, 3. Yagi, Y., M. Kikuchi, and T. Sagiya, Co-seisic slip, post-seisic slip, and aftershocks associated with two large earthquakes in 1996 Hyuga-nada, Japan, Earth Planets Space, 53, , 1. Yaanaka, Y. and M. Kikuchi, EIC SEISMOLOGICAL NOTE: No. 139, 3. S. Ozawa (e-ail: ozawa@gsi.go.jp), M. Kaidzu, M. Murakai, T. Iakiire, and Y. Hatanaka