JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, B11313, doi:10.1029/2009JB006665, 2010

Slip distribution of the 2003 Tokachi‐oki Mw 8.1 earthquake from joint inversion of tsunami waveforms and geodetic data F. Romano,1 A. Piatanesi,1 S. Lorito,1 and K. Hirata2 Received 4 June 2009; revised 23 July 2010; accepted 24 August 2010; published 23 November 2010.

[1] We study the 2003 Mw 8.1 Tokachi‐oki earthquake, a great interplate event that occurred along the southwestern Kuril Trench and generated a significant tsunami. To determine the earthquake slip distribution, we perform the first joint inversion of tsunami waveforms measured by tide gauges and of coseismic displacement measured both by GPS stations and three ocean bottom pressure gauges (PG) for this event. The resolution of the different data sets on the slip distribution is assessed by means of several checkerboard tests. Results show that tsunami data constrain the slip distribution offshore, whereas GPS data constrain the slip distribution in the onshore zone. The three PG data only coarsely constrain the offshore slip, indicating that denser networks should be installed close to subduction zones. Combining the three data sets significantly improves the inversion results. Joint inversion of the 2003 Tokachi‐oki earthquake data leads to maximum slip values (∼6 m) confined at depths greater than ∼25 km, between 30 and 80 km northwest of the hypocenter, with a patch of slip (3 m) in the deepest part of the source (∼50 km depth). Slip values are very low (≤1 m) updip from the hypocenter. Furthermore, the rupture does not extend on the plate interface off Akkeshi. As a significant back slip amount (∼4 m) has accumulated there since the last 1952 earthquake, this segment could rupture during the next large interplate event along the Kuril Trench.

Citation: Romano, F., A. Piatanesi, S. Lorito, and K. Hirata (2010), Slip distribution of the 2003 Tokachi‐oki Mw 8.1 earthquake from joint inversion of tsunami waveforms and geodetic data, J. Geophys. Res., 115, B11313, doi:10.1029/2009JB006665.

1. Introduction et al., 2006]. Furthermore, the event generated a large tsu- nami along the southern coast of , with runup [2] The M 8.1 Tokachi‐oki earthquake of 25 September w heights higher than 4 m at some locations (Figure 1) [Tanioka 2003 (Figure 1) has been one of the strongest and better et al., 2004a]. The tsunami was recorded along the coasts of recorded seismic events in the last 50 years in , and the both Hokkaido and islands by the Japanese tide first large interplate earthquake ever recorded by the Japanese gauge network (Figure 1). strong motion networks K‐NET and KiK‐net [Kinoshita, [4] In previous studies, the 2003 Tokachi‐oki earthquake 1998; Aoi et al., 2000]. Large earthquakes occur frequently has been analyzed following different approaches and using off Hokkaido Island (northern Japan) because the Pacific various geophysical data sets. The seismic source, for example, Plate subducts at the Kuril Trench just under the Hokkaido has been investigated inverting tsunami traveltimes [Hirata region (Figure 1). Several tsunamigenic earthquakes occurred et al., 2004], tsunami waveforms [Tanioka et al., 2004b], in the last 50 years in this region, as for example the 1952 teleseismic data [Yamanaka and Kikuchi, 2003; Horikawa, Tokachi‐oki (M8.2), the 1958 Etorofu (M8.1), the 1963 off 2004; Robinson and Cheung, 2010], strong motion data Urup (M8.1), the 1973 Nemuro‐oki (M7.7), and the 1994 [Honda et al., 2004, Aoi et al., 2008; Nozu and Irikura, 2008], Hokkaido Toho‐oki (M8.1) [Piatanesi et al., 1999; Utsu, GPS data [Miyazaki and Larson, 2008], teleseismic and 1999; Watanabe et al., 2006]. strong motion data jointly [Yagi, 2004], and finally combin- [3] Many GPS stations recorded the on land surface dis- ing GPS and strong motion data [Koketsu et al., 2004]. placement due to the 2003 Tokachi‐oki earthquake. The [5] In this study we perform, for the first time, as regards vertical component of the coseismic displacement was clearly this event, a joint inversion of geodetic and tsunami data. In captured also by three pressure gauges (PG) (two ocean ‐ general, fault planes of large subduction zone earthquakes lie bottom pressure gauges and one cable end station) installed beneath both land and ocean seafloor. Thus, tsunami wave- on the seafloor near the Kuril Trench (Figure 1) [Mikada forms and PG data can help constraining the slip distribution in the offshore zone, whereas geodetic data can mainly con- 1Department of Seismology and Tectonophysics, Istituto Nazionale di strain the slip onto the onshore zone [Satake, 1993]. There- Geofisica e Vulcanologia, Rome, Italy. 2 fore, inverting them jointly might lead to the retrieval of a Seismology and Volcanology Research Department, Meteorological better source model with respect to that estimated by inverting Research Institute, Japan Meteorological Agency, Tsukuba, Japan. data separately. Copyright 2010 by the American Geophysical Union. [6] A description of the data used, their preprocessing, the 0148‐0227/10/2009JB006665 fault parameterization, and the methods for Green’s functions

B11313 1of12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313

Figure 1. Location map of the 2003 Tokachi‐oki earthquake. Epicenter (white star) and focal mechanism are from Japan Meteorological Agency (JMA). Thin black lines are the surface projection of the subfaults used in this study. Blue triangles, green dots, and red dots are the positions of tide gauge stations, GPS stations, and ocean bottom pressure gauges (PG1, PG2, and PG3), respectively, used in the inversions (Tables S1 and S2 in the auxiliary material). In the inset, yellow squares are the aftershocks with Mw ≥ 4 in the 2 days after the main shock (U.S. Geological Survey); also indicated are the positions along the coast of Hokkaido where the runup measurements were higher than 4 m (Mabiro, Hamataiki, Horokayantou, and Oikamanai [Tanioka et al., 2004a]). calculation are provided first (sections 2, 3, and 4). Then, we recordings are provided by Japan Meteorological Agency describe the synthetic checkerboard tests used to assess data (JMA), the Hokkaido Regional Development Bureau (HRDB) resolving power for the slip distribution, both for single data of the Ministry of Land, Infrastructure and Transport, and the set inversions and for their joint inversion (section 5). Finally, Hydrographic and Oceanographic Department (HOD) of the we discuss the earthquake source characteristics of the Japan Coast Guard [Hirata et al., 2004]. In this study we use 2003 Tokachi‐oki earthquake retrieved from the inversion 16 tsunami waveforms (Table S1 in auxiliary material and (sections 6 and 7). Figure 1).1 Particular attention is dedicated in making the azimuthal coverage around the earthquake source as large as 2. Data

[7] Tsunami waves from the 2003 Tokachi‐oki earth- 1Auxiliary materials are available in the HTML. doi:10.1029/ quake were recorded at Japanese coastal tide gauges. These 2009JB006665.

2of12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313 possible, and in selecting only the recordings with a good [13] Green’s functions (horizontal and vertical coseismic signal‐to‐noise ratio. The selected records are sampled at displacements) at the GPS and PG stations, as well as the 1 min. Before using them in the inversion the tidal component initial condition for the tsunami propagation, are computed is removed by fitting it with a sum of three harmonics. using a layered Earth’s model by means of the PSGRN/ Moreover, for each waveform, we choose a time window PSCMP numerical code [Wang et al., 2006]. As input including only the first oscillations of the tsunami wave to parameters the PSGRN/PSCMP code needs P and S wave minimize the contribution of local effects on the tsunami velocities as well as density values for each of the layers. signal. Local effects, such as coastal reflections and resonance Here we use values from the vertical cross sections along of the bays could hide information on the seismic source and line F‐F′ as reported by Wang and Zhao [2005], adopting a are particularly difficult to model. model with four layers, whose parameters are listed in [8] The ground displacement associated to the earthquake Table S4 (see the auxiliary material). was recorded at the GPS stations of the GEONET network, operated by the Geographical Survey Institute of Japan (GSI). We use the GPS (Table S2 in auxiliary material and 5. Inversion Scheme and Checkerboard Figure 1) distributed over Hokkaido and northern Honshu Resolution Tests for which the coseismic offsets were estimated by Larson and Miyazaki [2008]. [14] We solve the inverse problem using a particular implementation of the simulated annealing technique called [9] The vertical component of the coseismic displacement “ ” has been recorded also on the seafloor by PGs, labeled as the heat bath algorithm [Rothman, 1986], following pre- PG1 (depth 2218 m), PG2 (depth 2210 m) as well as with a vious studies using tsunami, GPS, and strong motion data depth sensor, labeled PG3 (depth 2540 m), that is included [e.g., Piatanesi et al., 2007; Lorito et al., 2010]. For tsunami in a conductivity‐temperature‐depth meter (CTD) sensor time series, we use a cost function sensitive to both ampli- [Hirata et al., 2002]. In this work we use the vertical co- tude and phase matching [Spudich and Miller, 1990; Sen seismic static offsets at the PGs estimated by Mikada et al. and Stoffa, 1991]: 2 3 [2006] (Table S2 in the auxiliary material). Ptf 2 ðÞu ðÞt u ðÞt XN 6 O S 7 6 t 7 EmðÞ¼ 61 i 7 ð1Þ 4 Ptf Ptf 5 3. Fault Parameterization k¼1 2 2 uOðÞþt uS ðÞt ti ti [10] The source area (Figure 1) is set on the basis of the k M ≥ 4 aftershock distribution in the 2 days after the main w In equation (1) u and u are the observed and synthetic shock. We set the strike of the fault (225°) according to the O S waveforms, respectively, t and t are the lower and upper geometry of the subducting plate proposed by Bird [2003]; i f bounds of the time window, N is the number of records used the dip angle of the plate boundary gradually changes with in the inversion, and m indicates the single model. For the depth (from 12° to ∼25° downdip, Table S3 in the auxiliary geodetic data set instead, we use a standard L norm as a material) [Katsumata et al., 2003; Hirata et al., 2003]. The 2 cost function to quantify the misfit between experimental rake angle (109°) is the best estimation of Yamanaka and and synthetic data sets. Kikuchi [2003]. We discretize the fault plane (210 × 150 km) [15] Regarding the joint inversion, the global cost function dividing it into 35 subfaults with a size of 30 × 30 km each is a weighted mean of the individual cost functions. Since (Figure 1). each cost function has a very different behavior depending [11] During a preliminary step of this work, we performed on its sensitivity to the variations of each parameter, it is not some checkerboard tests with smaller subfault’s size (not straightforward to set the most appropriate weights. For shown), similar to those we describe in section 5. We con- example, during the synthetic tests described below, we cluded that, given the present data set and our inversion verified that a progressive increase of the relative weight scheme, 30 × 30 km is very close to the minimum allowed assigned to the tsunami data set with respect to the geodetic size for resolving details of the slip distribution. one results in a progressive loss of resolution for the slip onshore. On the other hand, by increasing the relative ’ weight of the geodetic data set results in a loss of resolution 4. Modeling and Green s Functions on the slip offshore. Therefore, we empirically assigned [12] We model the Green’s functions for each tide gauge similar weights to the two data sets, by means of several by using the nonlinear shallow water equations for the synthetic tests (analogous to those presented in sections 5.1, tsunami propagation [Satake, 2002]. Boundary conditions 5.2, and 5.3). are pure wave reflection at the solid boundary (coastlines) [16] We examine the resolving power on the slip distri- and full wave transmission at the open boundary (open sea). bution of both data sets separately and of their combination Equations are solved numerically by means of a finite dif- in a joint inversion, with several synthetic checkerboard tests ference technique on a staggered grid [Mader, 2001]. The [Piatanesi and Lorito, 2007; Lorito et al., 2008a, 2008b]. The initial seawater elevation is assumed to be equal to the target model is a slip distribution featuring a checkerboard coseismic vertical displacement of the sea bottom, while pattern, with slip values of 2 and 6 m alternating on adjacent the initial velocity field is assumed to be identically zero. The subfaults. Moreover, a priori information is introduced in the bathymetric grid for the computational domain (depicted in model by imposing a range of variation to the slip value, Figure 1) has 10 arc sec of spatial resolution and it was pro- which is allowed to vary from 0 to 10 m on each subfault, with vided by HOD. a 0.5 m step.

3of12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313

Figure 2. Results of the checkerboard resolution tests. (a) Slip distribution of the target model. (b, c, d, and e) Slip distributions obtained inverting tsunami data, GPS data, geodetic data (GPS and PG data), and tsunami and geodetic data jointly, respectively. The star indicates the JMA’s epicenter of the 2003 Toka- chi‐oki earthquake. For each case, only the types of data used are plotted in the corresponding panel. Blue triangles are tide gauges, green dots are GPS stations, and red dots are PGs. The tide gauges and GPS shown are only a fraction of those used for the inversions (see Figure 1). The blue ellipsis highlights a fault portion where resolution is improved by adding PG data to GPS data (see text).

4of12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313

Figure 3. Marginal distributions of the slip on each subfault for the synthetic joint inversion of Figure 2e (the order of the subfaults is given in the inset of Figure 1 and in Table S3 in the auxiliary material). Vertical blue dashed and red solid lines are the best and the average model values, respectively; horizontal black solid bars are the standard deviations. Numerical values are reported in Table S5 (in the auxiliary material).

5.1. Checkerboard Test for Tsunami Data might be sufficient for constraining the offshore slip distri- [17] In order to mimic modeling uncertainties, the syn- bution only at a coarser scale (see the auxiliary material thetic tsunami waveforms generated with the checkerboard Figure S1a). At the present scale (30 × 30 km) the slip model are corrupted by adding a Gaussian random noise distribution could be recovered only with a hypothetical with a variance that is 10% of the clean waveform amplitude configuration where the stations are immediately above and variance [Sen and Stoffa, 1991; Ji et al., 2002]. around the whole fault zone (see the auxiliary material [18] The general pattern and the slip values of the target Figure S1b). model are fairly well recovered in the best model obtained [21] Adding PG data helps to better constrain only the with tsunami data (compare Figures 2a and 2b; see also fraction of the shallowest portion of the fault plane where Table S5 in the auxiliary material). However, a more careful the PGs are located (Figure 2d). However, geodetic ob- observation reveals that slip values in the checkerboard servations off Tokachi (Hokkaido) are still insufficient in pattern are almost perfectly recovered only on the portion of number and spatial density to well constrain the slip distri- the fault far from the coast, whereas in the onshore zone the bution offshore. residuals are larger (up to 1.5 m). This supports the general 5.3. Checkerboard Test for Joint Inversion of Tsunami conclusion that tsunami data constrain the offshore better and Geodetic Data than the onshore rupture features [Satake, 1993]. [22] A joint inversion combining the two kinds of data gives satisfactory results (Figure 2e and Table S5 in the 5.2. Checkerboard Test for Geodetic Data (GPS and PG) auxiliary material), as the target model is well recovered on the entire fault plane. The combination of tsunami and [19] For GPS and PGs data, we add a Gaussian random geodetic data sets is then, in principle, efficient to constrain value to synthetic data sets, with a variance equal to the the slip distribution for this earthquake, and it might be in square of the standard deviation for each component (3 mm general a good strategy for retrieving the slip distribution of for east, 4 mm for north, and 9 mm for vertical component large tsunamigenic subduction zone earthquakes. [Larson and Miyazaki, 2008]). [23] The uncertainties on the model parameters are esti- [20] With GPS data the assumed checkerboard pattern is mated, following Piatanesi and Lorito [2007], by means of an well reproduced only on the subfaults sufficiently close to a posteriori analysis of a subset of the ensemble of models the stations and degrades farther from the coast (Figure 2c explored during the global search, for estimating the average and Table S5 in the auxiliary material). GPS data alone model and the uncertainties (see the auxiliary material).

5of12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313

Figure 4. Best model of the 2003 Tokachi‐oki earthquake slip distribution, obtained by inverting (a) tsunami data, (b) geodetic data, and (c) jointly tsunami and geodetic data. Symbols are as in Figure 2. Numerical slip values are reported in Table S6 (in the auxiliary material).

Figure 3 shows that the best slip values are very close to the hypocenter (subfaults 21, 22, 23, 27, and 28). Differently average ones and are well within the error bars. Then, the best from the tsunami data inversion, there is a maximum slip model well represents the subset of models that provide a value of 7 m close to the hypocenter (subfault 24), and a comparable good fit to the data. very shallow patch with 4.5 m of slip at the southernmost [24] However, we note that the slip parameters 11–13, 16– fault corner (subfault 35). Slip is almost absent in the north- 18, 21–23, and 26–28 are less resolved, as the correspond- western part of the fault, and this feature is even more ing distributions and standard deviations are relatively larger pronounced in this case than for the inversion of tsunami than the others. Even if we still have sufficient control on data. Cost function values (the root mean square for the the results, as shown by the best model values, we have three components of the displacement) give a measure of the larger uncertainties on this portion of the fault plane. residuals between the observed and predicted coseismic displacements at the GPS and PG stations (Figure 5b). 6. Results of Inversions: The 2003 Tokachi‐oki 6.3. Joint Inversion Earthquake [28] As the final step of the procedure, the joint inversion [25] We use both single and joint data sets inversions to of tsunami and geodetic data is carried out. The source estimate the slip distribution of the 2003 Tokachi‐oki model (Figure 4c and Table S6 in the auxiliary material) earthquake. We assess by means of several tests (not shown) shows an asperity (3.5–5.5 m) located approximately between that the maximum slip value is well below 10 m, and then about 30 and 80 km northwest of the hypocenter (subfaults we use this value as a priori upper limit as in synthetic tests 22, 23, 27, and 28). A small and deeper patch of slip is present of section 5. at about 100 km northwest of the hypocenter (3 m, subfault 21), and lower slip values (≤2 m) are distributed around the 6.1. Inversion of Tsunami Data main asperity. After comparison of the synthetic tests [26] Tsunami data inversion indicates that there is a patch (Figures 2b and 2d) to inversions of real data (Figures 4a, of high slip (6 m) located at about 30 km northwest of the 4b, and 4c) it is evident that the joint inversion model hypocenter (Figure 4a, subfault 28, and Table S6 in the comprises features of the fault portions best resolved by the auxiliary material). The largest slip concentration is con- single inversions. We indeed observe that, in the joint versely located in the deep part of the fault (between 60 and inversion model, the pattern of slip distribution close to the 100 km NNW of the hypocenter), with slip values from 4 to GPS stations is similar to that obtained by inverting only the 6.5 m (subfaults 16, 17, 18, 21, and 22). Minor patches of geodetic data (Figures 4b and 4c, subfaults 21 and 26); on lower slip (≤1.5 m) are mainly distributed in the SW portion the other hand, in the offshore part of the fault, although of the fault. The differences between observed and predicted three PGs are present, the slip distribution pattern is in fair waveforms are quantified by the global cost function value agreement with the results obtained inverting only the tsu- computed by means of equation (1), and also by cost function nami data (Figures 4a and 4c, subfaults 24, 30, and 34). values for each tide gauge (Figure 5a). [29] Our best model features a seismic moment of 1.6 × 1021 N m (using the depth‐dependent shear modulus values 6.2. Inversion of Geodetic Data listed in Table S4 of the auxiliary material), corresponding [27] The source model obtained with geodetic data to a magnitude Mw = 8.1. (Figure 4b and Table S6 in the auxiliary material) features [30] The predicted tsunami waveforms agree very well a stronger slip concentration (3.5–5 m) downdip from the with the ones observed at the tide gauges (Figure 6a). We

6of12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313

Figure 5. Goodness of fit for tsunami and geodetic data in terms of cost functions. (a) Tsunami data cost functions. Black line is the global cost function value for tsunami data inversion of Figure 4a, and black dots are the cost function values for single tide gauges (progressive numbers as in Table S1 the auxiliary material). Red line and dots are the same quantities for the tsunami cost function in the joint inversion of Figure 4c. (b) Same as Figure 5a except for the components of geodetic data inversion only (black) and for the joint inversion (red). Corresponding inversion results are given in Figure 4b and 4c, respectively; progressive station numbers are as in Table S2 in the auxiliary material. The best model cost function for the joint inversion is slightly larger than those for single data sets inversions because in the former, more data have to be fit with the same number of free parameters. just notice almost rigid phase shifting in some cases (e.g., at also implies that it might be difficult to prefer one model , Tokachikou, and Muroran), probably due to sub- over another, as they roughly fit data equally well. The fault size or relatively poor bathymetry model. Additionally, whole set of synthetic tests described above comes to the also the predicted displacements fit very well the observed rescue in this respect. Indeed, it is evident that the best ones (Figure 6b). Some discrepancies can probably be resolution over the fault plane is obtained only by the joint ascribed to the size of the subfaults, or to the a priori fixed inversion (Figure 2e), whereas single inversions are not able rake angle. to resolve slip details everywhere (Figures 2b, 2c, and 2d). [31] Best model cost functions for single data sets inver- In other words, the joint inversion has conveniently reduced sions are slightly smaller than those for the joint inversion the size of the null model space. (Figures 5a and 5b), as in the latter more data have to be fit [32] Figure 7 shows the marginal distributions for each with the same number of free parameters. However, the parameter. The best values are always within the error bars fact that the overall difference is small indicates an inter- (i.e., the standard deviation). Best values can therefore be consistency between the two data sets. Their consistency considered representatives of the subset of “good” models.

7of12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313

Figure 6. Comparison between the observed (black) and predicted (red) data sets. (a) Observed and pre- dicted tsunami waveforms. (b) Observed and predicted (left) GPS horizontal displacements and (right) GPS and PGs vertical displacements.

8of12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313

Figure 7. Marginal distributions of the slip on each subfault for the real case joint inversion of Figure 4c (the order of the subfaults is in the inset of Figure 1 and in Table S3 in the auxiliary material). Vertical blue dashed and red solid lines are the best and the average model values, respectively; horizontal black solid bars are the standard deviations. Numerical values are reported in Table S6 in the auxiliary material.

As already observed in the distributions for the synthetic test which is almost absent in any other model. The peak slip in (Figure 3), there is a subset of slip values that features rel- the shallow patch of Aoi et al. [2008] (Figure 8b) is more atively larger errors (e.g., subfaults 22, 23, 27, and 28). than 6 m, whereas the corresponding patch of Robinson and Cheung [2010] exceeds 12 m. However, Robinson and Cheung [2010] claim that this shallow patch is not a robust 7. Discussion feature of their model. Furthermore, the azimuthal coverage 7.1. Comparison of the Slip Distribution Estimated offered by strong motion stations is not optimal because in This Study With Other Models they cover only the northwestern side of the fault area [Honda et al., 2004]. We agree with these conclusions, since [33] The main asperity featured by our model is consistent we are convinced that if any such shallow and strong patch at the first order with that imaged by some other models (6–12 m) existed for this earthquake, the ensuing tsunami based on teleseismic data [Yamanaka and Kikuchi, 2003; would have been, at least locally, much more destructive. Horikawa, 2004], strong motion and GPS data [Koketsu Source depth is indeed a first‐order parameter for tsunami et al., 2004], GPS data only [Miyazaki and Larson, 2008], generation efficiency, particularly for shallow depths. Accord- and tsunami data [Tanioka et al., 2004b]. All of these models ingly, tsunami data inversions (in the present work, as well indeed feature a dominant slip patch with peak slip values as in the work by Tanioka et al. [2004b]) do not indicate any ranging from 4.5 to 7 m, while our model reaches a maximum shallow slip. slip of 5.5 m on two subfaults. [35] By comparing our slip distribution with the afterslip [34] Conversely, there are important differences with the distribution obtained by Baba et al. [2006], we observe that the models obtained by Robinson and Cheung [2010] (tele- patch with the maximum amount of coseismic slip is enclosed seismic data), Aoi et al. [2008], Nozu and Irikura [2008] by regions with large amount of afterslip (Figure 8c), con- (strong motion data), Yagi [2004] (teleseismic and strong sistently with the fact that the afterslip is expected to be motion data jointly). Actually, Yagi [2004] finds three dif- smaller where the strain release is larger. ferent asperities, one surrounding the hypocenter and shifted updip with respect to our main patch, with a peak slip value exceeding 5.5 m (Figure 8a). An almost colocated patch, 7.2. Some Implications to Next Large Interplate with peak slip exceeding 11 m, appears in the model by Nozu Earthquake off Tokachi and Irikura [2008]. The difference with Aoi et al. [2008] and [36] Satake et al. [2005] divided the subduction zone in the Robinson and Cheung [2010] models is even more striking, southernmost Kuril Trench into three segments (Tokachi, as they both find significant slip updip of the hypocenter, Akkeshi, and Nemuro segments), and Hirata et al. [2009]

9of12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313 suggested the presence of a distinct seismic gap just onto and, in comparison with Aoi et al. [2008] and Tanioka et al. the Akkeshi segment between the 2003 event and the 1973 [2004b], it includes only a small deep patch off Kushiro Nemuro‐oki earthquake (M7.7) [Tanioka et al., 2007] with a very low slip value (≤1 m). This patch of slip on the (Figure 8c). The Akkeshi segment remained unbroken in eastern end of the rupture area can then be considered 2003, as already indicated in previous studies of the 2003 negligible for a large earthquake with a magnitude 8+ as the tsunami source [e.g., Hirata et al., 2004; Tanioka et al., 2003 Tokachi‐oki event. Moreover, the afterslip of the 2003 2004b]. event terminated just at the edge of the gap area (Figure 8c). [37] The coseismically ruptured area estimated in this Therefore our results confirm that the rupture does not study is mainly concentrated downdip from the hypocenter extend on the plate interface off Akkeshi (Figure 8c). [38] The entire region (including the 2003 source zone and the Akkeshi segment) is identified by Suwa et al. [2006] as one of the most strongly coupled regions in northeastern Japan, with a back slip accumulation rate of about 80 mm/ yr. The main asperity in our source model corresponds to the area of largest back slip [Hashimoto et al., 2009]. If we take into account the average slip (∼4 m) onto this area, we can roughly account for the amount of back slip accumulated locally since the 1952 event. A similar slip amount might remain to be released on the Akkeshi segment; conversely, the accumulated stress might be accommodated by moderate earthquakes or in aseismic way [McCann et al., 1979]. Two subsequent M∼7 earthquakes occurred in this zone (Figure 8c), in late 2004, likely releasing just a fraction of the accumu- lated stress. Therefore, the plate interface off Akkeshi should not be an aseismic zone and it might be able to rupture during the next large interplate event along the Kuril Trench. 7.3. Possible Implications for a Tsunami Forecast System [39] As highlighted in section 5 and as also pointed out by Nishimura et al. [2005], offshore bottom pressure gauges can help to better constrain the source characteristics of an offshore earthquake. In case a tsunami is generated only few minutes are available for launching an alert. A real‐time tsunami forecast system then must be based on the earliest available seismic data in order to estimate a first‐order tsu- nami threat level. Thus, a denser network of offshore sensors would be a crucial improvement for a better rapid source estimation. Such a dense network might for example con- tribute to the current JMA’s tsunami early warning system, which is presently based on the maximum risk method for near coastal events [Kamigaichi, 2008].

8. Conclusions

[40] We invert both tsunami (tide gauge records) and geodetic data (GPS and PGs) to infer the slip distribution of the 25 September 2003 Tokachi‐oki earthquake. We per-

Figure 8. Comparison of the slip distribution of this work with other models and with other earthquakes locations. (a) Black contours (1.2 m intervals) show the coseismic slip distribution of the 2003 Tokachi‐oki by Yagi [2004]; (b) black contours (2.0 m intervals) show the coseismic slip distribution of the 2003 Tokachi‐oki by Aoi et al. [2008]; and (c) black contours show the afterslip distribution of the 2003 Tokachi‐oki earthquake [Baba et al., 2006], with a contour interval of 0.15 m; the slip distribution of the 1973 Nemuro‐oki earthquake [Tanioka et al., 2007] is also shown; red dots represent two M∼7 earthquakes that occurred in the suggested seismic gap off Akkeshi (blue ellipse).

10 of 12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313 form several synthetic tests in order to assess the sensitivity review, Pure Appl. Geophys., 166,77–96, doi:10.1007/s00024-008- of the data to slip distribution details. We found that tsunami 0432-7. Honda, R., S. Aoi, N. Morikawa, H. Sekiguchi, T. Kunugi, and H. Fujiwara data better retrieve the rupture features on the offshore fault (2004), Ground motion and rupture process of the 2003 Tokachi‐oki portion, whereas geodetic data on the onshore zone. There- earthquake obtained by strong‐motion data of K‐NET and KiK‐net, fore using jointly tsunami and geodetic data helps to better Earth Planets Space, 56, 317–322. image the slip distribution of this large subduction zone Horikawa, H. (2004), Fault geometry and slip distribution of the 2003 Tokachi‐oki earthquake as deduced from teleseismic body waves, Earth earthquake with respect to single data inversions. The sparsity Planets Space, 56, 1011–1017. of PG data contributes only slightly in resolving the slip on Ji, C., D. J. Wald, and D. V. Helmberger (2002), Source description of the offshore portion of the fault plane. For this reason the the 1999 Hector Mine, California earthquake; Part I: Wavelet domain inversion theory and resolution analysis, Bull. Seismol. Soc. Am., 92(4), installation of denser PG networks close to subduction zones 1192–1207, doi:10.1785/0120000916. and their integration in real‐time tsunami forecast systems is Kamigaichi, O. (2008), Tsunami forecasting and warning, in Encyclopedia suggested. of Complexity and System Science, vol. 10, pp. 9592–9618, Springer, New York. [41] The slip distribution obtained by the joint inversion Katsumata, K., N. Wada, and M. Kasahara (2003), Newly imaged shape of features a maximum slip concentration (∼6 m) about 30– the deep seismic zone within the subducting Pacific plate beneath the Hok- 80 km northwest of the hypocenter, with a patch of slip (3 m) kaido corner, Japan‐Kurile arc‐arc junction, J. Geophys. Res., 108(B12), in the deep part of the source. The position of the main 2565, doi:10.1029/2002JB002175. Kinoshita, S. (1998), Kyoshin net (k‐net), Seismol. Res. Lett., 69, 309–332. asperities and the seismic moment (corresponding to a mag- Koketsu, K., K. Hikima, S. Miyazaki, and S. Ide (2004), Joint inversion of nitude Mw 8.1) are in fair agreement with previous inversions strong motion and geodetic data for the source process of the 2003 involving other kinds of geophysical data sets, with exception Tokachi‐oki, Hokkaido, earthquake, Earth Planets Space, 56, 329–334. Larson, K. M., and S. Miyazaki (2008), Resolving static offsets from high‐ of those featuring shallow slip patches that are ruled out by rate GPS data: The 2003 Tokachi‐oki earthquake, Earth Planets Space, this study. Moreover, the slip does not extend on the plate 60, 801–808. interface off Akkeshi. This supports a previous study [Hirata Lorito, S., F. Romano, A. Piatanesi, and E. Boschi (2008a), Source process et al., 2009] that suggested the presence of a seismic gap in of the September 12, 2007, Mw 8.4 southern Sumatra earthquake from tsunami tide gauge record inversion, Geophys. Res. Lett., 35,L02310, the Akkeshi segment since the 1952 earthquake. Therefore, doi:10.1029/2007GL032661. there is the possibility of a future large interplate event along Lorito, S., A. Piatanesi, and A. Lomax (2008b), Rupture process of the 18 that segment of the Kuril trench. April 1906 California earthquake from near‐field tsunami waveform inversion, Bull. Seismol. Soc. Am., 98(2), 832–845, doi:10.1785/ 0120060412. Lorito, S., A. Piatanesi, V. Cannelli, F. Romano, and D. Melini (2010), [42] Acknowledgments. We would like to thank the Japan Meteoro- Kinematics and source zone properties of the 2004 Sumatra‐Andaman logical Agency, the Hokkaido Regional Development Bureau of the Min- earthquake and tsunami: Nonlinear joint inversion of tide‐gage, satellite istry of Land, Infrastructure and Transport, and the Hydrographic and altimetry and GPS data, J. Geophys. Res., 115, B02304, doi:10.1029/ Oceanographic Department of the Japan Coast Guard, for providing the 2008JB005974. digital tide gauge records and the bathymetric data set. We also thank Claudia Mader, C. L. (2001), Numerical Modeling of Water Waves, CRC Press, Piromallo, Stephen Monna and Giovanni Muscari for valuable suggestions, Boca Raton, Fla. Zhi Wang for providing velocity profiles data, and Toshitaka Baba for pro- McCann, W. R., S. P. Nishenko, L. R. Sykes, and J. Krause (1979), Seismic viding the postseismic distribution model. Some of the figures in this work gap and plate tectonics: Seismic potential for major boundaries, Pure Appl. were drawn using GMT software [Wessel and Smith, 1995]. Geophys., 117, 1082–1147, doi:10.1007/BF00876211. Mikada, H., et al. (2006), New discoveries in dynamics of an M8 earthquake‐ Phenomena and their implications at the using a long term monitoring cabled observatory, Tectonophysics, 426,95–105, References doi:10.1016/j.tecto.2006.02.021. ‐ Miyazaki, S., and K. M. Larson (2008), Coseismic and early postseismic Aoi, S., K. Obara, S. Hori, K. Kasahara, and Y. Okada (2000), New strong slip for the 2003 Tokachi‐oki earthquake sequence inferred from GPS motion observation network: KiK‐net, Eos Trans. AGU, 81(48), Fall ‐ data, Geophys. Res. Lett., 35, L04302, doi:10.1029/2007GL032309. Meet. Suppl., Abstract S71A 05. Nishimura, S., M. Ando, and K. Tadokoro (2005), An application of Aoi, S., R. Honda, N. Morikawa, H. Sekiguchi, H. Suzuki, Y. Hayakawa, numerical simulation techniques to improve the resolution of offshore T. Kunugi, and H. Fujiwara (2008), 3D finite‐difference simulation of ‐ ‐ fault kinematics using seafloor geodetic methods, Phys. Earth Planet. long period ground motions for the 2003 Tokachi oki, Japan, earth- Inter., 151, 181–193, doi:10.1016/j.pepi.2005.03.002. quake, J. Geophys. Res., 113, B07302, doi:10.1029/2007JB005452. Nozu, A., and K. Irikura (2008), Strong‐motion generation areas of a great Baba, T., K. Hirata, T. Hori, and H. Sakaguchi (2006), Offshore geodetic subduction‐zone earthquake: Waveform inversion with empirical data conducive to the estimation of the afterslip distribution following ’ ‐ ‐ – Green s functions for the 2003 Tokachi oki earthquake, Bull. Seismol. the 2003 Tokachi oki earthquake, Earth Planet. Sci. Lett., 241, 281 292, Soc. Am., 98(1), 180–197, doi:10.1785/0120060183. doi:10.1016/j.epsl.2005.10.019. Piatanesi, A., and S. Lorito (2007), Rupture process of the 2004 Sumatra‐ Bird, P. (2003), An updated digital model of plate boundaries, Geochem. Andaman earthquake from tsunami waveform inversion, Bull. Seismol. Geophys. Geosyst., 4(3), 1027, doi:10.1029/2001GC000252. – ’ Soc. Am., 97(1A), S223 S231, doi:10.1785/0120050627. Hashimoto, C., N. Akemi, T. Sagiya, and M. Matsu ura (2009), Interplate Piatanesi, A., P. Heinrich, and S. Tinti (1999), The October 4, 1994 Shikotan seismogenic zones along the Kuril‐Japan trench inferred from GPS data (Kurile Islands) tsunamigenic earthquake: An open problem on the source inversion, Nat. Geosci., 2, 141–144, doi:10.1038/ngeo421. – ‐ mechanism, Pure Appl. Geophys., 154,555 574, doi:10.1007/ Hirata, K., et al. (2002), Real time geophysical measurements on the deep s000240050244. seafloor using submarine cable in the southern Kurile subduction zone, – Piatanesi, A., A. Cirella, P. Spudich, and M. Cocco (2007), A global search IEEE J. Oceanic Eng., 27, 170 181, doi:10.1109/JOE.2002.1002471. inversion for earthquake kinematic rupture history: Application to the Hirata, K., E. L. Geist, K. Satake, Y. Tanioka, and S. Yamaki (2003), Slip ‐ 2000 Western Tottori, Japan earthquake, J. Geophys. Res., 112, distribution of the 1952 Tokachi oki earthquake (M8.1) along the Kuril B07314, doi:10.1029/2006JB004821. Trench deduced from tsunami waveform inversion, J. Geophys. Res., Robinson, D. P., and L. T. Cheung (2010), Source process of the M 8.3, 108(B4), 2196, doi:10.1029/2002JB001976. w 2003 Tokachi‐Oki, Japan earthquake and its aftershocks, Geophys. J. Int., Hirata, K., Y. Tanioka, K. Satake, S. Yamaki, and E. L. Geist (2004), The 181(1), 334–342, doi:10.1111/j.1365-246X.2010.04513.x. tsunami source area of the 2003 Tokachi‐oki earthquake estimated from ‐ Rothman, D. (1986), Automatic estimation of large residual statics correc- tsunami travel times and its relationship to the 1952 Tokachi oki earth- – – tions, Geophysics, 51, 332 346, doi:10.1190/1.1442092. quake, Earth Planets Space, 56, 367 372. Satake, K. (1993), Depth distribution of coseismic slip along the Nankai Hirata, K., K. Satake, Y. Tanioka, and Y. Hasegawa (2009), Variable tsu- Trough, Japan, from joint inversion of geodetic and tsunami data, J. Geo- nami sources and seismic gaps in the southernmost Kuril Trench: A phys. Res., 98(B3), 4553–4565, doi:10.1029/92JB01553.

11 of 12 B11313 ROMANO ET AL.: TOKACHI‐OKI 2003 JOINT SLIP INVERSION B11313

Satake, K. (2002), Tsunamis, in International Handbook of Earthquake Utsu, T. (1999), Seismicity Studies: A Comprehensive Review (in Japanese), and Engineering Seismology, Int. Geophys. Ser., vol. 81(A), edited by 876 pp., Univ. of Press, Tokyo. W. H. K. Lee et al., pp. 437–451, Academic, London. Wang, R., F. L. Martin, and F. Roth (2006), PSGRN/PSCMP—A new code Satake, K., F. Nanayama, S. Yamaki, Y. Tanioka, and K. Hirata (2005), for calculating co‐ and post‐seismic deformation, geoid and gravity Variability among tsunami sources in the 17th‐21st centuries along the changes based on the viscoelastic‐gravitational dislocation theory, Comput. southern Kuril Trench, in Tsunami Research: Case Studies and Recent Geosci., 32,527–541, doi:10.1016/j.cageo.2005.08.006. Developments, Adv. Nat. Technol. Hazards Res., vol. 23 edited by Wang, Z., and D. Zhao (2005), Seismic imaging of the entire arc of Tohoku K. Satake, pp. 157–170, Springer, Dordrecht, Netherlands. and Hokkaido in Japan using P‐wave, S‐wave and sP depth‐phase data, Sen, M., and P. L. Stoffa (1991), Nonlinear one‐dimensional seismic wave- Phys. Earth Planet. Inter., 152, 144–162, doi:10.1016/j.pepi.2005.06.010. form inversion using simulated annealing, Geophysics, 56, 1624–1638, Watanabe, T., H. Takahashi, M. Ichiyanagi, M. Okayama, M. Takada, doi:10.1190/1.1442973. R. Otsuka, K. Hirata, S. Morita, M. Kasahara, and H. Mikada (2006), Spudich, P., and D. P. Miller (1990), Seismic site effects and the spatial Seismological monitoring on the 2003 Tokachi‐oki earthquake, derived interpolation of earthquake seismograms: Results using aftershocks of from off Kushiro permanent cabled OBSs and land‐based observations, the 1986 North Palm Springs, California earthquake, Bull. Seismol. Tectonophysics, 426, 107–118, doi:10.1016/j.tecto.2006.02.016. Soc. Am., 80, 1504–1532. Wessel, P., and W. H. F. Smith (1995), New version of the Generic Map- Suwa, Y., S. Miura, A. Hasegawa, T. Sato, and K. Tachibana (2006), Inter- ping Tools released, Eos Trans. AGU, 76, 329, doi:10.1029/95EO00198. plate coupling beneath NE Japan inferred from three‐dimensional dis- Yagi, Y. (2004), Source rupture process of the 2003 Tokachi‐oki earth- placement field, J. Geophys. Res., 111, B04402, doi:10.1029/ quake determined by joint inversion of teleseismic body wave and strong 2004JB003203. ground motion data, Earth Planets Space, 56, 311–316. Tanioka, Y., et al. (2004a), Tsunami run‐up heights of the 2003 Tokachi‐oki Yamanaka, Y., and M. Kikuchi (2003), Source process of the recurrent earthquake, Earth Planets Space, 56, 359–365. Tokachi‐oki earthquake on September 26, 2003, inferred from teleseismic Tanioka, Y., K. Hirata, R. Hino, and T. Kanazawa (2004b), Slip distribu- body waves, Earth Planets Space, 55, e21–e24. tion of the 2003 Tokachi‐oki earthquake estimated from tsunami wave- form inversion, Earth Planets Space, 56, 373–376. K. Hirata, Seismology and Volcanology Research Department, Tanioka, Y., K. Satake, and K. Hirata (2007), Recurrence of recent large Meteorological Research Institute, Japan Meteorological Agency, Tsukuba earthquakes along the southernmost Kurile‐Kamchatka subduction zone, 305‐0052, Japan. in Volcanism and Subduction: The Kamchatka Region, Geophys. – S. Lorito, A. Piatanesi, and F. Romano, Sezione di Sismologia e Monogr. Ser.,vol.172,editedbyJ.Eichelbergeretal.,pp.145 152, Tettonofisica, Istituto Nazionale di Geofisica e Vulcanologia, Via di AGU, Washington, D. C. Vigna Murata 605, I‐00143 Rome, Italy. ([email protected])

12 of 12