Asperities and quasi-static slips on the subducting plate boundary east off Tohoku, NE Japan Akira Hasegawa1, Naoki Uchida1, Toshihiro Igarashi2, Toru Matsuzawa2, Tomomi Okada1, Satoshi Miura1 and Yoko Suwa1 1 Research Center for Prediction of Earthquakes and Volcanic Eruptions, Graduate School of Science, Tohoku University, Aramaki, Aoba-ku, Sendai, Miyagi 980-8578, Japan 2 Earthquake Research Institute, University of Tokyo, Yayoi 1-1-1, Bunkyo-ku, Tokyo 113-0032, Japan Abstract : We review recent studies on the interplate coupling and the recurrence of earthquakes along the subducting plate boundary east off northeastern (NE) Japan. Waveform inversions of a few successive earthquakes that occurred in common source areas have revealed that large slip areas of successive ruptures are in the same place (asperities) on the plate boundary. Many small earthquake sequences having similar waveforms, including an M 4.8±0.1 earthquake sequence regularly occurring with a recurrence interval of 5.35±0.53 years, have been found on the plate boundary zone. They are interpreted as caused by repeated ruptures of small asperity patches surrounded by stably sliding areas. Space-time distribution of quasi-static slip on the plate boundary was estimated from activity of these repeating earthquake sequences, which is approximately consistent with that estimated from back slip inversions of GPS data on land. These observations strongly suggest that the asperity model is applicable to the process of seismic and aseismic slip occurring on the plate boundary east off NE Japan. Three large asperities are left as unruptured patches on the plate boundary. One asperity south off Tokachi ruptured on 26 September 2003 as the M 8.0 Tokachi-oki earthquake. In the vicinity of another asperity east off Miyagi prefecture, there occurred an intermediate-depth event with M 7.1 in the subducted slab on 26 May 2003 and a shallow event with M 6.4 in the overriding plate on 26 July 2003, which might be triggered by the strong coupling at the asperity and a quasi-static slip in its surrounding area. Submitted to the SEIZE volume of Columbia University Press 1. Introduction Beneath Tohoku, northeastern Japan (NE Japan), the Pacific plate is subducting downward into the mantle at a convergence rate of 〜8 cm/year (Demets et al., 1994). It subducts at a very low angle of <10°for its initial descent to a distance 〜100 km away from the trench. Then it changes dip angle and descends with a steeper angle of 〜30° beneath the land area [Hasegawa et al., 1994; Umino et al., 1995; Hino et al., 1996; Zhao et al., 1997]. The seismic coupling coefficient is estimated to be about 25 % for the plate boundary off Sanriku (northern Tohoku) and about 10 % for the plate boundary off Fukushima (southern Tohoku) [e.g., Pacheco et al., 1993; Peterson and Seno, 1984]. This suggests that much of the interplate motion is accommodated by aseismic slip in these regions. Nonetheless, many shallow earthquakes including great earthquakes occur beneath the Pacific Ocean along the plate boundary or in its vicinity associated with the plate subduction. Figure 1 shows epicentral distribution of shallow (<60 km) earthquakes for each five-year period from 1981 to 2000. An uneven distribution pattern seems to be persistent, regardless of time. Recent studies based on data obtained by seismic observations and GPS observations on land have made an important contribution to understanding how aseismic and seismic slip proceeds along the plate boundary for these regions and their relation with each other. Moreover, waveform 1 inversion analyses for some interplate earthquakes have shown that the asperity model [Lay and Kanamori, 1981; Lay et al., 1982] is actually applicable to the plate boundary in these regions. In the present manuscript, we review these recent studies, and then discuss the present state of the interplate coupling and its relation with recent remarkable seismic activity in the northeastern Japan convergent plate boundary zone. 2. Distribution of large slip areas by successive earthquakes Two end-member models have been used to describe the interplate coupling and the recurrence of great earthquakes at convergent plate boundary zones. One is the asperity model, in which the plate boundary is divided into areas of strong frictional coupling dominated by stick-slip behavior (asperities), and intervening areas of weak frictional coupling dominated by stable sliding (stably sliding areas). The other is the uniform coupling model, in which the frictional coupling along the plate boundary is uniform. In the asperity model, asperities are persistent features. Large slip areas of successive ruptures will be in the same place (asperities) for a given section of the plate boundary. Earthquakes may rupture one or more adjacent asperities; earthquake magnitudes will be large if the total size of the ruptured asperities is large. On the contrary, in the uniform coupling model, both the recurrence of great earthquakes and their slip distributions will vary with time. A large slip area of one great earthquake will be a small slip area of the next great earthquake in a given section of the plate boundary. It has been difficult, however, to know from direct observations which model of the two is applicable to the process actually occurring on the plate boundary, since there have been extremely few examples of well-documented recurrent large earthquakes. Recently, some of large earthquakes in subduction zones reruptured the same portions of the plate boundary that failed in previous large earthquakes with modern seismic data, which provided the opportunity to compare the slip distribution for successive events. Schuwartz [1999] investigated such earthquake pairs (or triplets) rupturing overlapping regions. They are (1) the 1968 (M7.9) Tokachi-oki and 1994 (M7.6) Sanriku-oki earthquakes in NE Japan; (2) the 1971 (Mw 8.0) and 1995 (Mw 7.7) Solomon Islands earthquakes; (3) the 1963 (Mw 8.5) and 1995 (Mw7.9) Kuril Islands 2 earthquakes; and (4) the 1957 (Mw 8.6), 1986 (Mw 8.0) and 1996 (Mw7.9) Aleutian Islands earthquakes. Schwartz [1999] systematically studied spatial distribution of moment release of these earthquake pairs or triplets based on inversion of teleseismic broadband body waves and long-period surface waves. Comparisons of the spatial distribution of moment release for these successive earthquake ruptures showed two types in the pattern of recurrent fault slip. The NE Japan earthquakes and the Solomon Islands earthquakes seem to be in the first type; the 1994 Sanriku-oki and 1995 Solomon Islands earthquakes filled in areas of slip deficit left by their preceding earthquakes. The slip pattern for recurrent earthquakes of this type is not consistent with the asperity model. The second type, which is consistent with the asperity model, includes the Kuril Islands earthquakes and the Aleutian Islands earthquakes; the 1995 Kuril Islands and the 1996 Aleutian Islands earthquakes both reruptured portions of the asperity distribution defined by the preceding events, although amounts of slip at the identical asperities varied for sequential earthquake ruptures. Nagai et al. [2001] re-investigated slip distributions of the 1968 M 7.9 Tokachi-oki earthquake and the 1994 M 7.6 Sanriku-oki earthquake, and obtained a result that strongly supports the asperity model. They estimated detailed slip distributions of the two earthquakes by waveform inversions of both strong motion records at a regional network and teleseismic body wave records at global networks. The results show that the 1968 event has more than two large slip areas (asperities) and the 1994 event has one large slip area as shown in Fig. 2. Comparison of the slip distributions between the two events shows that the large slip area of the 1994 event is located exactly in the same place as one of the two large slip areas of the 1968 event. This indicates that one of the two asperities in the south ruptured by the 1968 event was ruptured again by the 1994 event. Nagai et al. [2001] estimated that the seismic coupling is almost 100 % in this common asperity. This observation indicates that asperities are persistent features and that the asperity model is a better representation to explain the recurrence of large earthquakes along the plate boundary in this region. They also re-examined aftershock hypocenters by using the same station corrections for the two events. Relocated aftershocks fringe the estimated asperities and are not distributed within the asperities for both events, which also supports the asperity model. 3 The study by Nagai et al. [2001] updated the previous result for the two large earthquakes in NE Japan by Schwartz [1999]. Use of regional strong-motion network together with teleseismic body wave data enabled to estimate slip distributions with higher resolutions. Thus they could detect the second largest asperity in the south for the 1968 Tokachi-oki earthquake, which was not detected in the previous studies using teleseismic body and surface wave data and tsunami waveform data [Hartog and Schwartz, 1996; Tanioka et al., 1996]. Consequently, recurrent fault slip for three of the four plate boundary segments investigated by Schwartz [1999] proved to be consistent with the asperity model. Yamanaka and Kikuchi [2003] have systematically studied slip distributions of large interplate earthquakes that occurred east off NE Japan for the last 70 years by inverting waveforms recorded by strong motion seismometers of Japan Meteorological Agency. They found a clear tendency that large slip areas of successive large earthquakes were in the same places on the plate boundary, which again supports the asperity model. Moreover, they found that the typical size of individual asperities in NE Japan is M 7 class and that M 8 class great earthquakes are caused when more than one asperities are ruptured simultaneously. 3. Aseismic slip and interplate coupling estimated from GPS observations Asperities slip dynamically as earthquakes, while the surrounding stably sliding areas (or conditionally stable sliding areas) slip quasi-statically during the interseismic period. If so, it is very important to know when the quasi-static slip actually occurs in the stably sliding areas of the plate boundary. Recently remarkable progress has been made on this point for the plate boundary east off NE Japan. Nishimura [2000] and Heki et al. [1997] detected large postseismic slip after the 1994 M 7.6 Sanriku-oki earthquake, which was comparable to the seismic moment of the main shock. Similarly large postseismic slip was also observed after the 1989 and 1992 Sanriku-oki earthquakes of M 7.1 and M 6.9, respectively [Kawasaki et al., 2001; Miura et al., 1993]. These are based on observations of crustal deformation by GPS, strainmeters and tiltmeters. GPS data also provide important information on the spatial distribution of the interplate coupling. Nishimura et al. [2000] estimated spatial and temporal distribution of seismic and aseismic slips along the plate boundary 4 east off NE Japan based on back slip inversions of GPS data on land. Back slip distribution estimated for the period of 1996-1999 showed a strong interplate coupling in the area off Miyagi prefecture and large forward slip in the downward extension of the focal area of the 1994 M 7.6 Sanriku-oki earthquake. Recently, Suwa et al. [2003] estimated back slip distribution on the plate boundary east off Tohoku and south off Hokkaido by the inversion method of Yabuki and Matsu’ura [1992]. Vertical components of the displacement velocity field estimated from GPS observations, in addition to horizontal components, were used for the back slip inversions. Addition of the vertical component provided more stable results of the back slip distribution. Estimated back slip distribution for the 5 year period 1997−2002 is shown by contours in Fig. 3. It clearly shows two strongly coupled areas (areas with large back slip amounts) on the plate boundary; one east off Miyagi prefecture and the other an area east off Aomori prefecture through south off Tokachi. In those areas the estimated seismic coupling for the 5 years is about 75 %−100 %. Back slip in a small area east off Miyagi prefecture slightly exceeds the plate conversion rate, but this is probably due to the estimation error. Figure 3 also shows that the back slip area is not confined to the area beneath the Pacific Ocean but also to the area beneath the land. This means that the interplate coupling extends to a depth of about 100 km almost beneath the volcanic front, although the amount of back slip is small (seismic coupling of 〜25 %). This is contradictory to our knowledge of the seismogenic depth along the plate boundary obtained from seismic observations. Interplate earthquakes occur only at depths shallower than 50-60 km in NE Japan [Hasegawa et al., 1991 and 1994; Igarashi et al., 2001]. If the interplate coupling really extends down to ~100 km depth, the coupled portion of the plate interface deeper than 50-60 km must slip episodically at sometime during the interseismic period. Aseismic slip after large interplate earthquakes (postseismic slip) is a possible candidate for this. In fact, large postseismic slip amounting to 25-70 cm was detected in the deep plate interface down to ~100 km depth on the down-dip extension of the mainshock fault after the previous M7.4 Miyagi-oki earthquake of 1978 [Ueda et al., 2001]. Similar postseismic slip on the deep portion of the plate interface has been reported for the 1938 M7.7 5 Fukushima-oki earthquake [Ueda, 2001] and for the 1994 M7.6 Sanriku-oki earthquake [Nishimura, 2000]. Alternatively, Heki [2004] interpreted that the subsidence in the forearc region of NE Japan observed by recent GPS measurements is caused by basal subduction erosion, which does not require the interplate coupling on the deep portion of the plate interface beneath the land. He estimated the rate of material loss at the plate interface (the rate of the erosion of the upper plate by the underthrusting plate) as 15 mm/yr in the depth range of 5-90 km, using vertical velocity data by GPS observations for 4 years of 1996-2000. However, the estimated rate of 15 mm/yr is significantly faster than that geologically estimated [von Huene and Lallemand, 1990]. Moreover, his interpretation leaves a problem that the basal erosion with such a high rate must have occurred for the 4 years in regions where the plate interface has been almost completely coupled (east off Miyagi prefecture and south off Tokachi), while the rate must have been very low in regions where the plate interface has been slipping quasi-statically (a region between 39°N and 40°N) . Accumulation of crustal deformation data for longer periods is needed to solve the problem. 4. Similar earthquakes off Kamaishi recurring at a regular interval Similar earthquakes of M 4.8±0.1 that regularly occur at the same location have been found on the plate boundary off Kamaishi, east off northern Tohoku [Matsuzawa et al., 1999, 2002]. The location of this peculiar earthquake sequence is shown by an arrow on a map of epicentral distribution of shallow earthquakes in Fig. 1(a). These similar earthquakes recur at a regular interval of 5.35±0.53 years (Fig. 4). They have identical low-angle thrust fault type focal mechanisms, typical of interplate events, and nearly the same waveforms as each other. Dimensions of rupture areas of the recent two events of this earthquake sequence were estimated from corner frequencies determined from acceleration seismograms recorded at a nearby station [Matsuzawa et al., 2002]. Results show source lengths of 〜1 km and source areas of 0.5−1 km2. Slip amounts for the two events were estimated to be 23−46 cm. Hypocentral relocations of the recent four events in this sequence using the homogeneous station method [Ansell and Smith, 1975] show that their 6 hypocenters, i. e. rupture initiation points, are confined to an area with a diameter of 〜0.8 km (Fig. 6) [Igarashi et al., 2000]. This earthquake sequence is interpreted to be generated by the repeated ruptures of the same asperity patch (frictionally locked patch) with a size of 〜1 km surrounded by stably sliding areas on the plate boundary. The seismic coupling ratio on that asperity amounts to almost 100 %. There are no historical records of earthquakes with magnitudes larger than six in the area surrounding this earthquake sequence, although microearthquake activity is very high here (Fig.1). This suggests that the plate boundary in this area is creeping similarly to the creeping segment of the San Andreas fault. The regular occurrence of M 4.8±0.1 earthquakes is perhaps due to the repeating slip of the isolated asperity patch surrounded by the creeping area slipping at nearly a constant rate. Based on this interpretation, the next event was expected to occur by the end of November 2001 with 99 % probability [Matsuzawa et al., 1999]. An earthquake with M 4.7 actually occurred on November 13, 2001 at the same location on the plate boundary as expected [Matsuzawa et al., 2002]. Rupture area of this 2001 event was compared with that of the last 1995 event in this sequence [Okada et al., 2003]. Hypocenters of the two events are precisely relocated by carefully picking P- and S-wave arrival times at surrounding stations and by using the homogeneous station method. Waveforms of the two events observed by a regional broad-band seismograph network were inverted to estimate moment release distribution on the fault plane by using the method of Hartzell and Heaton [1983]. The results show that the spatial extent of both rupture areas are almost the same with a length of 1−1.5 km. The rupture areas of the two events almost completely overlap with each other, although their ruptures were initiated from different points 〜200 m apart (Fig. 5). This clearly shows that the two events result from repeated slip of the same asperity patch on the plate boundary, and again strongly supports the hypothesis of persistent asperities. We do not have any unequivocal evidence to understand formation of the asperity (frictionally locked patch) on the plate boundary; however, the hypocentral distribution of the four recent events of the off Kamaishi sequence provides interesting information for speculation about this. Relative locations of hypocenters for these events are precisely located by the method of Nakamura et al. 7 [2002] using differences of S-P time between a master event and the other three events. Relocated hypocenters are plotted on EW vertical cross section and are shown in Fig. 6 [Igarashi et al., 2000]. Hypocenters, i. e. initiation points of ruptures, of these M 4.8±0.1 events are distributed on an inclined plane with a dimension of 〜0.8 km. The dip angle of this plane is about 40°, which is slightly steeper than that of the subducting plate boundary in this region (〜28°). This locally steeper dip angle at the asperity suggests a possibility that the asperity of the M 4.8±0.1 events are generated by a small protuberance existing on the top of the subducting Pacific plate with a radius of 〜1 km, as schematically shown in the inset of Fig. 6, although the difference in dip angle is not very significant if we take estimation errors of hypocenters into consideration. Focal mechanisms of the five recent events determined by P-wave polarity data have the average dip angle of 33.4°±4.3°. This value is again slightly steeper than the local dip of the plate boundary, which seems to support the above suggestion. 5. Repeating earthquakes and interplate quasi-static slip Many similar earthquakes (earthquakes with similar waveforms) were found along the plate boundary zone east off NE Japan [Igarashi et al., 2003]. Most of these events were not found within the subducting Pacific plate or within the continental North American/Okhotsk plate with a few exceptions (Fig. 7). We interpret that these similar earthquakes are caused by repeating ruptures of small asperity patches with dimensions of 0.1-1 km surrounded by stably sliding areas along the plate boundary, similarly to the M 4.8±0.1 earthquake sequence off Kamaishi already described in section 3. Figure 8 shows the spatial distribution of the proportion of similar earthquakes to ordinary earthquakes along the plate boundary and its vicinity east off NE Japan. Back slip distribution for the 5 year period shown in Fig. 6 is also shown by contours in this figure. Comparison between the two indicates that areas with back slip larger than 〜8 cm/year, i.e. 100 % coupling for this period, have no similar earthquakes. On the contrary, the proportions of similar earthquakes are relatively high for areas with small back slips. This observation seems to support the above interpretation that the 8 similar earthquakes are caused by repeating ruptures of small asperity patches surrounded by freely slipping regions (repeating earthquakes) [Igarashi et al., 2003]. If the seismic coupling within these small asperities is 100 %, we can estimate the cumulative amount of aseismic slip in the area surrounding each asperity on the plate boundary [Nadeau and McEvilly, 1999]. Nadeau and Johnson [1998] derived a scaling relation between seismic moment and seismic slip for repeating earthquakes in Parkfield and Stone Canyon, California, by assuming the average slip rate on the plate boundary based on the relative plate motion velocity and 100% seismic coupling at each asperity. We investigated whether or not this scaling relation, derived from repeating earthquakes on the transform-fault type plate boundary in California, is applicable to those on the subduction zone plate boundary in NE Japan [Igarashi et al., 2003]. The repeating earthquake sequence off Kamaishi with M 4.8±0.1 described above (section 3), a subcluster sequence with M 〜3.5 next to it, and earthquakes off Miyagi prefecture with M 〜7.5 occurring repeatedly with a recurrence interval of about 37 years [ Earthquake Research Committee of the Headquarters for Earthquake Research Promotion, 2001] are used for this analysis. Slip rates on the plate boundary surrounding these earthquake sequences are assumed to be 8 cm/year since those surrounding areas are supposed to be creeping. The results show that these earthquake sequences also satisfy the scaling relation of Nadeau and Johnson [1998] (Fig. 9). This indicates that the cumulative amount of aseismic slips on the plate boundary surrounding repeating earthquake sequences can be estimated for the NE Japan subduction zone using this scaling relation between the moment and slip. Spatial distribution of the slip rate along the plate boundary east off NE Japan was estimated by applying the scaling relation of Nadeau and Johnson [1998] to repeating earthquakes occurring there [Igarashi et al., 2003]. The estimated distribution of slip rate is almost consistent with the back slip distribution estimated from GPS data [Nishimura et al., 2000; Suwa et al., 2003]. Large forward slip areas estimated from the back slip inversions also have a large slip amount estimated from the repeating earthquake analysis. This suggests that repeating earthquakes can be used for estimating temporal and spatial distribution of the interplate aseismic slip. Recently, many more repeating earthquakes occurring along the plate boundary east off NE Japan were identified based on systematic analyses of waveform cross-correlation calculations for 9