# Examination of source-model construction methodology for strong ground-motion simulation of multi-segment rupture during 1891 Nobi earthquake

## Abstract

We examine the construction methodology of a source model for strong ground-motion prediction of scenario earthquakes with a long active-fault zone including a multi-segment rupture case. For the multi-segment rupture event, different seismic moments are given by applying two different source-model construction methodologies: (1) a methodology based on seismological scaling relationships and (2) a methodology in an active-fault study based on the cascade model (Working Group on California Earthquake Probabilities, 1995). The difference between these methodologies is whether or not a slip of each segment depends on total rupture length. We investigate which construction methodology is more appropriate for the construction of the source model. We use the two methodologies to construct characterized source models for strong ground-motion simulations of the 1891 Nobi earthquake, which was a multi-segment rupture event occurring in the Nobi active-fault system. We evaluate the constructed source models by comparing the seismic intensities obtained from simulated waveforms with those estimated from the questionnaire-based intensities determined by Muramatu and Kominami (1992). The characterized source model constrained by seismological scaling relationships can give a more adequate distribution of the estimated seismic intensities. We also suggest that the Gifu-Ichinomiya fault might be considered as a source fault to explain the damage distribution on the Nobi Plain.

### Key words

Multi-segment rupture event source-model construction methodology seismological scaling relationship cascade model 1891 Nobi earthquake## 1. Introduction

One of the most important issues in applied seismology is the establishment of a methodology for strong groundmotion prediction. The construction of a source model that can treat a variety of earthquake scenarios is the key for achieving reliable strong ground-motion prediction. A characterized source model (Irikura and Miyake, 2001) has recently been used as a source model for strong groundmotion prediction for scenario earthquakes (e.g., Iwata and Miyake, 2004; the Headquarters for Earthquake Research Promotion, 2005). Irikura and Miyake (2001) suggested a methodology for constructing the characterized source model based on the empirical relationships of the total rupture area and the combined asperity area with the seismic moment.

To develop a methodology for the construction of the characterized source model, it is essential to simulate the strong ground motion of actual earthquakes using such a model and to compare the results with the observation records. For example, the Headquarters for Earthquake Research Promotion (HERP) (2002, 2008a) has been continuously validating methods by comparing observed and synthetic strong ground motions and seismic intensities for, e.g., the 2000 Western Tottori Prefecture earthquake (*M*_{J} 7.3) and the 2005 West Off Fukuoka Prefecture earthquake (*M*_{J} 7.0). However, the source-fault lengths of those events are short and a validation of a multi-segment rupture event in a long active-fault zone has not yet been carried out.

For multi-segment ruptures, researchers have discussed whether or not the repetition of the same dislocation rupture characterizes each of the segments of a long active-fault zone (e.g., Sieh, 1996; McGill and Rubin, 1999; Kaneda and Okada, 2008). The repetition of the same dislocation rupture for each segment is not considered in existing construction methodologies for the characterized source model for strong ground-motion prediction, such as the methodology developed by HERP (2008b). However, several activefault studies have supported the concept of repetition of a similar slip along a segment (e.g., Sieh *et al.*, 1996; Kondo et al., 2005). Therefore, in the construction methodology we need to consider the behavior of each segment involved in a multi-segment rupture. Here, we consider behavioral segments that are bounded by changes in the slip rates, recurrence intervals, elapsed times, sense of displacement, creeping versus locked behavior, and fault complexity (McCalpin, 1996).

For the 1891 Nobi earthquake (*M* 8.0) in Japan, the results of field investigations (e.g., Matsuda, 1974) and paleoseismological studies (e.g., Awata et al., 1999; Yoshioka et al., 2001) indicated that three behavioral segments ruptured along the Nukumi fault, the Neodani fault, and the Umehara fault. This means that the 1891 Nobi earthquake was a multi-segment rupture event. Although all the waveform records in the source region were clipped, seismic intensities based on the responses to a questionnaire survey are available at more than 1000 sites for estimating the strong ground-motion distribution.

In this study, we consider alternative construction methodologies for the characterized source model based on scaling from the entire rupture length and from behavioral segment lengths. We give multiple characterized source models for the 1891 Nobi earthquake as an example of a multi-segment rupture event. We compare the estimated seismic intensities from ground motions simulated by the characterized source models and those estimated from the questionnaire-based intensities determined by Muramatu and Kominami (1992). We also discuss the preferred construction methodology for the characterized source model for strong ground-motion prediction in a multi-segment rupture event.

Researchers have discussed the rupture of the Gifu-Ichinomiya fault, which did not show clear evidence on surface faulting, in the 1891 Nobi earthquake (e.g., Muramatu, 1963; Mikumo and Ando, 1976; Aichi Prefecture, 1998, 2000; HERP, 2001; Fukuyama et al., 2007; Nakano et al., 2007). Based on the results of the strong ground-motion simulations, we also evaluate whether predictions using synthesis have the resolution to say whether or not the Gifu-Ichinomiya fault ruptured during the 1891 earthquake.

## 2. Previous Studies of the 1891 Nobi Earthquake

The Nobi earthquake, one of the greatest inland crustal earthquakes in Japan in recent times, occurred on October 28, 1891. It generated surface earthquake faulting along the active faults known as the Nukumi fault, the Neodani fault, and the Umehara fault (Fig. 1). Matsuda (1974) compiled field investigation reports (e.g., Koto, 1893; Omori, 1894; Bessho, 1967) and investigated traces of the surface earthquake faults and the distribution of displacements along these faults.

Kaneda and Okada (2008) performed detailed tectonic geomorphological mapping along the Neodani fault to test the behavioral model for a fault involved in a multi-segment rupture. They showed that the distributions of the leftlateral and vertical slip rates based on the estimated ages of terrace abandonment and the cumulative displacements of geomorphic markers mimic in shape those of left-lateral and vertical slips during the 1891 Nobi earthquake. They also indicated that the cumulative displacement since the L1 abandonment at 17–20 ka is only three times the displacement of the 1891 event. This means that the recurrence interval from their slip-rate analysis is longer than that revealed by trench study at the Neodani fault. They concluded that if the recurrence interval revealed by Awata et al. (1999) for the Neodani fault is correct, this fault should not be characterized by repetition of the same slip-function earthquake ruptures or the characteristic slip-function behavior. However, they also noted that characteristic spatial slip-function behavior may still be expected if some of the identified faulting events are minor slip events caused by adjacent faults such as the Ibigawa and Mugigawa faults. In this study, via the validation of our construction methodology, we discuss whether or not the same slips were repeated on each behavioral segment.

Mikumo and Ando (1976) calculated theoretical vertical displacements based on the plural fault plane models of the 1891 Nobi earthquake and compared them with the observed vertical displacements along two leveling routes across the Umehara and Gifu-Ichinomiya faults. They indicated that movements of four faults including the Gifu-Ichinomiya fault were needed to explain the observation. Nakano et al. (2007) determined the location and geometry of the Gifu-Ichinomiya fault by a grid-search method based on an inversion of the coseismic vertical displacements obtained by leveling surveys. They showed that the Gifu-Ichinomiya fault is located 5 km east of the location assumed in the model of Mikumo and Ando (1976). Fukuyama et al. (2007) evaluated the seismic moment of the 1891 earthquake by comparing calculated synthetic seismograms from a point source and finite source models with the historical seismogram recorded at the Central Meteorological Observatory in Tokyo. They showed that a comparison between the observed and synthetic seismograms appears to favor the existence of a buried fault along the Gifu-Ichinomiya line. In contrast to these results, the Aichi Prefecture (1998, 2000) and HERP (2001) used stratigraphic and seismic reflection surveys to rule out the possibility of the existence of the Gifu-Ichinomiya fault. We use the intensities estimated by Muramatu and Kominami (1992) to investigate whether the Gifu-Ichinomiya fault ruptured during the 1891 Nobi earthquake.

## 3. Characterized Source Model for Strong-motion Prediction

In the last section, we introduced the surface source fault and the distribution of the questionnaire-based seismic intensity of the Nobi earthquake for evaluating source models. We need to see how much it is possible to constrain the source model from the seismic intensity distribution. Kanda and Takemura (2005) developed source inversion using seismic intensity data. They estimated the empirical attenuation relation of the seismic intensity from recent earthquake data and applied it to the seismic intensity distribution of historical events by assuming that the finite source model consists of distributed point sources. They obtained spatial heterogeneity of source on the fault. Although their method gave information on the heterogeneity of the seismic wave radiation, our aim is to discuss the construction methodology of a seismic source model for multi-segment rupture events. Therefore, our approach is to construct a set of source models for ground-motion simulation, to estimate the ground motions according to those models, and to compare the observed and simulated seismic intensities in order to select the most appropriate model.

The characterized source model is one of the most reliable approaches for broadband strong ground-motion prediction. This model was proposed by Irikura and Miyake (2001). They first introduced a methodology for the source model construction. HERP applied this technique to explain the observations of recent disastrous earthquake events and confirmed the applicability and limitations of the method. Irikura et al. (2002) improved the method, and HERP summarized the methodology together with evaluation methods for the ground-motion simulation. This is now known as Irikura’s recipe. The procedure and parameter settings are given in the Appendix.

For strong ground-motion simulation, we use the empirical Green’s function (EGF) method proposed by Irikura (1986). Records of small events at the K-NET and KiK-net stations are used for the EGF. We simulate the ground motions for the Nobi earthquake at the strong motion recording sites, and estimate the seismic intensities from the simulation results.

## 4. Strong Ground-motion Simulation

### 4.1 Construction of characterized source models

The Research Group for Active Faults of Japan (1991) estimated the location and length of the Gifu-Ichinomiya fault based on the boundary between tectonic uplift and subsidence. Sugisaki and Shibata (1961) suggested the existence of the Gifu-Ichinomiya fault from the vertical offset of a pumice bed of 5 m in the borehole records at Ichinomiya city. Fukuwa et al. (2003) estimated the locations of structural boundaries of the Nobi plain using the vertical gradient distribution of gravity anomaly. From their results, it seems that a structural boundary exists near the location of the Gifu-Ichinomiya fault shown by the Research Group for Active Faults of Japan (1991). Therefore, we use the location and length of the Gifu-Ichinomiya fault from the Research Group for Active Faults of Japan (1991). In Table 2, the length of the Gifu-Ichinomiya fault is listed.

We assume the width of each fault from the lower cutoff depth of microseismic activities, the depth of the upper limit of the fault plane, and the dip angle assumed for each fault. We assume that the depth of the lower limit of the fault plane of each fault is same as the bottom depth of the seismogenic zone. By referring to Ito (1999), we determine the bottom depth of the zone from the distribution of the hypocenters of the small earthquakes that occurred from 1926 to 2006 around the Nobi fault zone. These hypocenters were determined by the Japan Meteorological Agency (JMA).

For inland crustal earthquakes, HERP (2008b) has assumed that the depth of the upper limit of the fault plane is equal to the top depth of the seismogenic zone. In this case, the upper limit of the fault plane of the characterized source model does not reach the ground surface even if surface earthquake faults are generated during a scenario earthquake. However, some studies on source-fault rupture models give a finite source model in which the fault plane reach the ground surface, e.g., Horikawa et al. (1996) for the 1995 Hyogo-ken Nanbu earthquake along the Nojima fault, and Wald and Heaton (1994) for the 1992 Landers earthquake. Their source models show slips in the shallower part of the fault plane at the surface earthquake faults. Since during the Nobi earthquake surface earthquake faults were generated along the Nukumi, Neodani, and Umehara faults, we assume that the depths of the upper limits of the fault planes are 0 km. We also assume that all stations at the ground surface have a depth of 0 km. Because no surface earthquake fault was generated along the Gifu-Ichinomiya fault, we assume that the depth of the upper limit of the Gifu-Ichinomiya segment is equal to the top depth of the seismogenic zone. The dip angles of all segments are assumed to be 90 degrees. The width of each fault is listed in Table 2.

In the second method which we call the segment-L model (s-L model), the seismic moment of each segment is obtained from the rupture area of that segment by applying the empirical relationship in Eqs. (A.1) and (A.2). Here, we assume that this relationship between seismic moment and rupture area is satisfied for each behavioral segment. In this method, the average slip of each segment is constant for any combination and number of segments that rupture simultaneously. The total seismic moment is obtained by summing the seismic moments for each segment. This approach is based on the cascade model proposed by the Working Group on California Earthquake Probabilities (1995). The cascade model indicates that a behavioral segment ruptures independently or along with other contiguous segments and assumes that repetition of the same slip at given point characterizes the behavioral segment. The average slip of each segment increases as the rupture length increases when the downdip width of the fault reaches saturation because of the finite thickness of the brittle zone.

*M*

_{w}7.2) for a three-segment rupture and 1.10E+20 N m (

*M*

_{w}7.3) for a four-segment rupture. Applying the second method gives a total seismic moment of 2.48E+19 N m (

*M*

_{w}6.9) for a three-segment rupture and 3.04E+19 N m (

*M*

_{w}6.9) for a four-segment rupture (Fig. 3).

Next, we set up the combined asperity area using two different methods. In the first method (method 1), short-period spectral level of acceleration source spectrum, which is obtained from a seismic moment by applying the empirical relationship in Eq. (A.3), is used to estimate the equivalent radius of asperity. We apply the short-period spectral level of the acceleration source spectrum for causative faults to estimate the equivalent asperity radius in the total-L model. In the segment-L model, we assume the short-period spectral level of the acceleration source spectrum for each segment to estimate the equivalent asperity radius for that segment. In the other method (method 2), we use the relationship that for inland crustal earthquakes the combined asperity area is specified to be about 22% of the total rupture area (Somerville et al., 1999).

*R*being the equivalent radius of the source fault area. In the s-L and s-L 22 models, we apply Eq. (A.5) to the equivalent radius of the asperity of each segment, with R being the equivalent radius of the corresponding fault area. The stress drop on asperity in each model is listed in Table 4. We also estimate the effective stress on the background area of each segment using the method of HERP (2008b).

Sugiyama et al. (2002) revealed that the distributions of coseismic displacements along surface earthquake faults are consistent with the slip distributions in the shallower part of the seismic fault plane. Iwaki et al. (2006) identified the asperities on the fault plane for fourteen earthquakes and examined the spatial relationship between surface earthquake faults and subjacent source faults. They showed that locations where the surface displacements are large correspond in many cases to the locations of asperities on source faults. We assume that the asperities are located where larger dislocations are observed by referring to the distribution (Matsuda, 1974) of the left-lateral displacements.

### 4.2 Synthesis of strong ground motions

We simulate strong ground motion using the EGF method proposed by Irikura (1986). This method has been used to simulate strong ground motions during inland crustal earthquakes such as the 1995 Hyogo-ken Nanbu earthquake (e.g., Kamae and Irikura, 1998) and the 2000 Tottori-ken Seibu earthquake (e.g., Ikeda et al., 2002). They used the recorded waveforms of aftershocks as the EGF to simulate the strong ground motions during the mainshock. In this study, we use the recorded waveforms of two small earthquakes that occurred around the Nobi fault zone to take into account the different propagation path from each segment to the strong motion stations (Fig. 1). We use the focal mechanism solutions of these small earthquakes determined by F-net (Fukuyama et al., 1998). Event 1 is an earthquake of Open image in new window in Fukui prefecture on October 5, 2004. Event 2 is an earthquake of Open image in new window in Gifu prefecture on February 16, 2006. One nodal plane of the small earthquakes is almost consistent with that of the assumed fault plane. We use the recorded waveforms of Event 1 as the EGF for the Nukumi segment fault and those of Event 2 for the Neodani, Umehara, and Gifu-Ichinomiya segment faults.

*Q*value of Satoh et al. (1994) and hypocentral distance. We determine corner frequencies of 2.0 Hz and 1.9 Hz for Event 1 and Event 2, respectively. We assume that the faults are circular cracks and estimate the radius

*r*

_{c}by referring to Brune (1970, 1971). The area of the fault for each event is calculated by

*π*

*r*

_{c}

^{2}. We assume that subfaults of the large event are square-shaped so they are equivalent in area to the small-event fault. We therefore choose a subfault size of 1.1 km by 1.1 km for the Nukumi fault segment and 1.2 km by 1.2 km for the others. We calculate static stress drops of 8.0 MPa and 3.4 MPa for Event 1 and Event 2 respectively, by referring to Brune (1970, 1971). The rise time of each event is given from the empirical relationship between seismic moment and rise time given by Somerville et al. (1999).

*M*

_{0}) and that of the small event (

*m*

_{0}): On the right-hand side,

*c*is the ratio of the average stress drops,

*N*

_{L}is the number of subfaults along the fault length,

*N*

_{W}is the number of subfaults along the fault width, and

*N*

_{T}is the ratio of the rise times of the scenario earthquake and the small earthquake.

*N*

_{L},

*N*

_{W}, and

*N*

_{T}are more than 10. Irikura and Kamae (1994) observed that the spectrum of the simulated ground motions has deeper sags in the intermediate frequency range as the number of summations for the EGF method increases. This is because of the effect of artificial enhancement of the recurrence period of the summation. To avoid these artificial sags, we introduce a twostage summation method to synthesize the waveforms. We first simulate ground motions for virtual middle-size earthquakes called Event 3 and Event 4 from the recorded waveforms of the real small earthquakes (Event 1 and Event 2). The fault sizes for Event 3 and Event 4 are 3.3 km by 3.3 km and 3.6 km by 3.6 km, respectively. The seismic moments for Event 3 and Event 4 are 1.31E+17 N m and 6.45E+16 N m, respectively. We assume that the rupture starting point is the center of the fault, and the stress drops of Events 3 and 4 are assumed to be the same as those of Events 1 and 2, respectively. As a second step, we synthesize the waveforms of the Nobi earthquake using the simulated waveforms of the virtual middle-size earthquakes. Equation (1) is satisfied for the seismic moment of asperity area of each segment and that of the small event. It is also satisfied in background slip area of each segment. Parameters for the EGF method are listed in Table 5.

We simulate the acceleration and velocity waveforms of the NS, EW, and UD components at 148 strong motion stations of K-NET (Kinoshita, 1998) and KiK-net (Aoi et al., 2000). We use the recorded waveforms of the small earthquakes for 30 s from 0.5 s before the direct *S*-wave onset. The frequency band for simulating the acceleration and velocity waveforms is 0.5 to 10.0 Hz. A cutoff frequency Open image in new window of 6 Hz is chosen in this study (e.g., Satoh, 2002).

## 5. Questionnaire-based Intensity Distribution of the 1891 Nobi Earthquake

We use the questionnaire-based intensities to estimate the seismic intensities at the strong motion stations. We assume that the questionnaire-based intensities are the same as the JMA intensity scales. The answers to the questionnaires seem to reflect the situations of the communities around the locations where the questionnaire survey was conducted. Therefore, we calculate the seismic intensities at the strong motion stations from the questionnaire-based intensities reported within 5 km. We estimate the seismic intensities in units of 1 degree by applying the inverse distance weighted mean of the questionnaire-based intensities. If there are no questionnaire-based intensities within 5 km of a strong motion station, we estimate the seismic intensity by referring to the contour map of the distribution of the questionnairebased intensities determined by Muramatu and Kominami (1992).

## 6. Results

The sum of the squared differences of the seismic intensities obtained from simulated waveforms and those estimated from the questionnaire-based intensities is the smallest in the characterized source model 16.

_{cal}) for the model 16. For this model, SIS

_{cal}values above 6 are simulated at the KNET strong motion stations at AIC001 and GIF020 in the heavily damaged region with a wooden-house damage ratio from Muramatu (1983) of more than 80%. Strong motion stations with SIS

_{cal}values above 6 are distributed from the southwestern part of Gifu prefecture to Aichi prefecture. Figure 8(b) shows the distributions of the differences of the SIS

_{cal}values for model 16 and the estimated questionnairebased scale values (SIS

_{qi}). The SIS

_{cal}values are mostly within 1 of the SIS

_{qi}values. At the strong motion stations in the heavily damaged region, the SIS

_{cal}values are consistent with the SIS

_{qi}values.

^{2}is simulated. The seismic wave radiating from the Neodani segment causes this large acceleration.

_{cal}values and the SIS

_{qi}values for models 16, 13, 22, and 4. The difference between the best model and the models assumed in Fig. 10(b), (c), and (d) is the method for estimating the combined asperity area, obtaining the total seismic moment, and deciding whether or not the Gifu-Ichinomiya fault ruptured. The sums of the squared differences for the models in Fig. 10(a), (b), (c), and (d) are 27.2, 45.4, 67.0, and 34.1. We observe from Fig. 10(c) that the SIS

_{cal}values are significantly smaller than the SIS

_{qi}values at most strong motion stations. In Fig. 10(d), the SIS

_{cal}values are inconsistent with the SIS

_{qi}values at the strong motion stations in the heavily damaged region. This means that rupturing in the Gifu-Ichinomiya segment might be necessary to explain the distribution of the damage caused to wooden houses near the Gifu-Ichinomiya fault.

## 7. Discussion

### 7.1 Comparison of simulated peak horizontal velocities (PHVs) with empirical attenuation relationship

*V*

_{s}30 of 300 m/s. An amplification correction toward this value is estimated by applying the

*V*

_{s}30 from the PS logging results of K-NET and KiK-net to the empirical relationship of Midorikawa et al. (1994). The attenuation characteristics of the PHVs in the best characterized source model are found to approximately follow the empirical attenuation relation, as shown in Fig. 11. However, the PHVs at HYG020, HYGH09, and OSKH02 are larger than expected. These stations are located in the southwestern area more than 100 km from the causative faults. At these stations, the waveforms radiating from the Neodani and Gifu-Ichinomiya segments significantly affect the results of the strong ground-motion simulation. We used the recorded waveforms of Event 2 to simulate waveforms from the Neodani and Gifu-Ichinomiya segment faults. The PHVs of the recorded waveforms of Event 2 at HYG020, HYGH09, and OSKH02 are larger than the attenuation relation obtained from PHVs at other stations. It seems that a propagation-path effect causes these larger-than-expected PHVs. Here, we confirmed that the selection of the best model is independent of whether these stations are included or not.

### 7.2 Verification of the source model 16

The characterized source model constrained by the seismological scaling relationship is better for expressing the distribution of the estimated seismic intensities. In this model, the total seismic moment is 1.10E+20 N m. From the historical seismograms recorded at the Central Meteorological Observatory in Tokyo, Fukuyama et al. (2007) estimated the total seismic moment of the Nobi earthquake to be 1.83E+20 N m (*M*_{w} = 7.5). Based on an inversion analysis of the coseismic vertical displacements, Nakano et al. (2007) estimated the total seismic moment to be 1.38E+20 N m (*M*_{w} = 7.4). These two values are slightly larger than the total seismic moment of the source model 16.

The seismic moments of each behavioral segment in the model 16 are larger than those obtained by applying the empirical relationship to the rupture area of the segment. If the empirical relationship between seismic moment and rupture area is satisfied for each behavioral segment in a single-segment rupture, our result may mean that the slip of each segment in a multi-segment rupture is different from that in a single-segment rupture.

Kaneda and Okada (2008) observed that if the earthquake recurrence interval estimated from a previous paleoseismic trench is correct, the tectonic geomorphological observations in the Neodani fault can be explained by repetition of a similarly shaped slip distribution independently of the rupture of adjacent faults, but with a variable amount of slip. Our result appears to be consistent with their result in that it has no repetition of the same dislocation rupture for the Neodani segment.

Wesnousky (2008) examined the average and maximum values of the coseismic slip as a function of the surface rupture length. Wesnousky considered the slopes of the fitted curves describing the slip increase and found that these slopes decrease as a function of the total rupture length without apparently reaching a plateau in the strikeslip earthquake data. If the relationship between the slip of a behavioral segment and the total rupture length is similar, the difference of the slip in a single-segment rupture and that in a multi-segment rupture may be small, and similar displacements may be repeated along the segment with a large rupture length. In the Nobi fault zone, the rupture length of each segment may not be enough to satisfy this relationship.

We assume that the combined asperity area is specified to be about 22% of the total rupture area in source model 16. In this model, the stress drops on asperity are 23.1 MPa for all segments. Dan et al. (2005) discussed two methods for estimating the effective stress on asperity for inland earth quakes caused by large strike-slip faults. The first method is based on the scaling law of Irikura and Miyake (2001) and the crack model. In the second method, the effective stress on asperity is 14.1 MPa independent of the earthquake size. Following Madariaga (1979), in the second method the effective stress on asperity of 14.1 MPa is estimated from the average static stress drop (e.g., 3.1 MPa from Fujii and Matsu’ura (2000)) and the ratio of the combined asperity area to the total rupture area. Dan et al. (2005) conducted strong ground-motion simulations based on characterized source models with different values for the effective stress on asperity. They compared simulated peak ground velocities with the attenuation relation of Si and Midorikawa (1999) and concluded that the effective stress on asperity from the second method is more proper. In this study, the attenuation characteristics of the simulated peak ground velocity in the source model 16 are found to approximately follow the empirical attenuation relation. This shows that the calculated stress drop on asperity of the model 16 is preferred for the Nobi earthquake.

The location of the rupture starting point in the model 16 is the northwestern end of the Nukumi segment (CASE1). However, the sum of the squared differences of seismic intensity is almost the same for CASE1 and CASE2 (southeastern end of the Nukumi segment). By comparing synthetic seismograms with recorded seismograms at the Gifu and Nagoya stations, Mikumo and Ando (1976) estimated the location of the starting point to be the northwestern end of the Nukumi fault. Muramatu et al. (2002) also estimated the starting point to be the northwestern end of the Nukumi fault based on the time interval reported by Matsuda (1974) of about 10 s to the formation of the surface fault at Seki. Our result is consistent with these previous results.

Our results show that the Gifu-Ichinomiya fault might rupture during the Nobi earthquake. The location of the Gifu-Ichinomiya segment in this study is based on that of the Research Group for Active Faults of Japan (1991). The location is nearer Nagoya City than that used by Mikumo and Ando (1976). We also assume the dip of the Gifu-Ichinomiya segment to be 90 degrees. Mikumo and Ando (1976) assumed a buried fault with a dip of 90 degrees. Nakano et al. (2007) estimated a thrust fault with a dip of 60 degrees located 5 km east of the location of Mikumo and Ando (1976). They compared the distribution of seismic intensities calculated using their best model with the distribution from Muramatu (1983) of the damage caused to wooden houses by the Nobi earthquake. However, they did not calculate seismic intensities of 7 at Gifu city where the percentage of damaged houses is more than 80%. In the model 16, we simulate a seismic intensity of 7 at Gifu city. The distribution of the strong motion stations showing a seismic intensity above 6 near the Gifu-Ichinomiya fault might be consistent with the distribution from Muramatu (1983) of the damage caused to wooden houses.

## 8. Conclusions

We examine the source-model construction methodology for the strong ground-motion prediction of scenario earthquakes in a long active-fault zone including a multi-segment rupture case. We construct 24 characterized source models of the 1891 Nobi earthquake by combining two methods for obtaining the total seismic moment, two methods for estimating the combined asperity area, three types of rupture starting points, and two cases of the multi-segment ruptures that includes or omits the Gifu-Ichinomiya segment. We conduct strong ground-motion simulation for the characterized source models using the empirical Green’s function method.

To determine the most plausible source-model construction methodology, we compare the seismic intensities calculated from simulated waveforms using the assumed source models with those estimated from the questionnairebased intensities determined by Muramatu and Kominami (1992). The sum of the squared differences between the seismic intensities obtained from simulated waveforms and those estimated from the questionnaire-based intensities is the smallest in the source model having the following characteristics: (1) the total seismic moment is given by applying the methodology based on seismological scaling relationships, (2) the combined asperity area is specified to be 22% of the total rupture area (Somerville et al., 1999), (3) the rupture starting point is the northwestern end of the Nukumi segment, and (4) the Gifu-Ichinomiya fault might rupture during the 1891 Nobi earthquake.

The attenuation characteristics of the PHVs in the best characterized source model are found to approximately follow the empirical attenuation relation of Kanno et al. (2006). In this model, the total seismic moment is 1.10E+20 N m (*M*_{w} 7.3). The seismic moments of each behavioral segment in the best characterized source model are larger than those obtained by applying the empirical relationship to the rupture area of each segment. If the empirical relationship between seismic moment and rupture area is satisfied for each behavioral segment in a single-segment rupture, our result may imply that the slip of each segment in a multi-segment rupture is different from that in a singlesegment rupture.

## Notes

### Acknowledgments

We are grateful to two anonymous reviewers for the critical reviews and suggestions on the original manuscripts. We sincerely thank Prof. Takashi Kumamoto for helpful comments. We used data from the K-NET and KiK-net strong motion networks. We also used the moment tensor solutions of two small-sized events from the F-net. We are grateful to the National Research Institute for Earth Science and Disaster Prevention (NIED) for providing the K-NET, KiK-net, and F-net. We used hypocentral information provided by Japan Meteorological Agency (JMA). Some figures have been drawn using GMT (Wessel and Smith, 1998).

### References

- Aichi Prefecture,
*Research Report of the Gifu-Ichinomiya and Yoro-Kuwana-Yokkaichi Fault Zones*, 118 pp, 1998 (in Japanese).Google Scholar - Aichi Prefecture,
*Underground Structure in the Nobi Plain*, 61–70, 2000 (in Japanese).Google Scholar - Aoi, S., K. Obara, S. Hori, K. Kasahara, and Y. Okada, New strong-motion observation network, KiK-net,
*Eos Trans. AGU*,**81**, F863, 2000.Google Scholar - Awata, Y., Y. Kariya, and K. Okumura, Segmentation of the surface ruptures associated with the 1891 Nobi earthquake, Central Honshu, Japan, based on the paleoseismic investigations,
*Interim Report on Active Fault and Paleoearthquake Researches in the 1998 Fiscal Year*, 115–130, 1999 (in Japanese with English abstract).Google Scholar - Bessho, B., Unusual tectonic movements,
*Rep. Geol., Kanazawa Univ.*,**1**, 1–364, 1967 (in Japanese).Google Scholar - Boatwright, J., The seismic radiation from composite models of faulting,
*Bull. Seismol. Soc. Am.*,**78**, 489–508, 1988.Google Scholar - Brune, J. N., Tectonic stress and the spectra of seismic shear waves from earthquakes,
*J. Geophys. Res.*,**75**, 4997–5009, 1970.CrossRefGoogle Scholar - Dalguer, L. A., H. Miyake, and K. Irikura, Characterization of dynamic asperity source models for simulating strong ground motions,
*Proceedings of the 13th World Conference on Earthquake Engineering*, No. 3286, 2004.Google Scholar - Dalguer, L. A., H. Miyake, S. M. Day, and K. Irikura, Surface rupturing and buried dynamic-rupture models calibrated with statistical observations of past earthquakes,
*Bull. Seismol. Soc. Am.*,**98**, 1147–1161, 2008.CrossRefGoogle Scholar - Dan, K., M. Watanabe, T. Sato, and T. Ishii, Short-period source spectra inferred from variable-slip rupture models and modeling of earthquake faults for strong motion prediction by semi-empirical method,
*J. Struct. Constr. Eng., AIJ*,**545**, 51–62, 2001 (in Japanese with English abstract).Google Scholar - Dan, K., T. Sato, and K. Irikura, Characterizing source model for strong motion prediction based on asperity model,
*Proceedings of the Japan Earthquake Engineering Symposium*, 555–560, 2002 (in Japanese with English abstract).Google Scholar - Dan, K., T. Muto, J. Miyakoshi, and M. Watanabe, Estimation of effective stress on asperities in inland earthquakes caused by large strike-slip faults and its application to strong motion simulation,
*J. Struct. Constr. Eng., AIJ*,**589**, 81–88, 2005 (in Japanese with English abstract).Google Scholar - Eshelby, J. D., The determination of the elastic field of an ellipsoidal inclusion, and related problems,
*Proc. R. Soc. Lond. Ser. A*,**241**, 376–396, 1957.CrossRefGoogle Scholar - Fujii, Y. and M. Matsu’ura, Regional difference in scaling laws for large earthquakes and its tectonic implication,
*Pure Appl. Geophys.*,**157**, 2283–2302, 2000.CrossRefGoogle Scholar - Fukuyama, E., M. Ishida, D. S. Dreger, and H. Kawai, Automated seismic moment tensor determination by using on-line broadband seismic waveforms,
*Zisin, Second Ser.*,**5l**, 149–156, 1998 (in Japanese with English abstract).Google Scholar - Fukuyama, E., I. Muramatu, and T. Mikumo, Seismic moment of the 1891 Nobi, Japan, earthquake estimated from historical seismograms,
*Earth Planets Space*,**59**, 553–559, 2007.CrossRefGoogle Scholar - Fukuwa, N., K. Yamaoka, M. Nakano, J. Tobita, T. Sato, and Y. Suzuki, Investigation for the mechanism of the seismic damage band on Nobi earthquake,
*Report of Grants-in-Aid for Scientific Research (Research Project Number F12480110)*, 2003 (in Japanese).Google Scholar - Geller, R. J., Scaling relations for earthquake source parameters and magnitudes,
*Bull. Seismol. Soc. Am.*,**66**, 1501–1523, 1976.Google Scholar - Horikawa, H., K. Hirahara, Y. Umeda, M. Hashimoto, and F. Kusano, Simultaneous inversion of geodetic and strong-motion data for the source process of the Hyogo-ken Nanbu, Japan, earthquake,
*J. Phys. Earth*,**44**, 455–471, 1996.CrossRefGoogle Scholar - Ikeda, T., K. Kamae, S. Miwa, and K. Irikura, Source characterization and strong ground motion simulation of the 2000 Tottori-ken Seibu earthquake using the empirical Green’s function method,
*J. Struct. Cosntr. Eng., AIJ*,**561**, 37–45, 2002 (in Japanese with English abstract).Google Scholar - Irikura, K., Prediction of strong acceleration motions using empirical Green’s function,
*Proc. 7th Japan Earthq. Eng*, 151–156, 1986.Google Scholar - Irikura, K. and K. Kamae, Estimation of strong ground motion in broadfrequency band based on a seismic source scaling model and an empirical Green’s function technique,
*Ann. Geofis.*,**37**, 1721–1743, 1994.Google Scholar - Irikura, K. and H. Miyake, Prediction of strong ground motions for scenario earthquakes,
*J. Geogr.*,**110**(6), 849–875, 2001 (in Japanese with English abstract).CrossRefGoogle Scholar - Irikura, K. and H. Miyake, Recipe for predicting strong ground motion from crustal earthquake scenarios,
*Pure Appl. Geophys.*, http://www.springerlink.com/content/424323g755p14436/fulltext.pdf, 2010.Google Scholar - Irikura, K., H. Miyake, T. Iwata, K. Kamae, and H. Kawabe, Revised recipe for predicting strong ground motion and its validation,
*11th JEEE*, 567–572, 2002 (in Japanese with English abstract).Google Scholar - Ishii, T., T. Sato, and P. G. Somerville, Identification of main rupture areas of heterogeneous fault models for strong ground-motion estimation,
*J. Struct. Constr. Eng., AIJ*,**527**, 61–70, 2000 (in Japanese with English abstract).Google Scholar - Ito, K., Seismogenic layer, reflective lower crust, surface heat flow and large inland earthquakes,
*Tectonophysics*,**306**, 423–433, 1999.CrossRefGoogle Scholar - Iwaki, H., H. Ito, N. Kitada, N. Inoue, T. Kagawa, K. Miyakoshi, K. Takemura, and A. Okada, Compiling the catalog of surface faults due to large earthquakes and subjacent source faults,
*Active Fault Res.*,**26**, 37–61, 2006 (in Japanese with English abstract).Google Scholar - Iwata, T. and H. Miyake, Ground motion simulation during a scenario earthquake based on the strong motion prediction recipe—In case of the northern part of the Biwako-Seigan faults as a hypothetical source fault—,
*J. JSNDS*,**23-2**, 259–271, 2004 (in Japanese with English abstract).Google Scholar - Kamae, K. and K. Irikura, Source model of the 1995 Hyogo-ken Nanbu earthquake and simulation of near-source ground motion,
*Bull. Seismol. Soc. Am.*,**88**, 400–412, 1998.Google Scholar - Kanda, K. and M. Takemura, Verification of the sequence of disastrous earthquakes off Miyagi prefecture based on seismic intensity data,
*Zisin Second Ser.*,**58**, 177–198, 2005 (in Japanese with English abstract).Google Scholar - Kaneda, H. and A. Okada, Long-term seismic behavior of a fault involved in a multiple-fault rupture: Insights from tectonic geomorphology along the Neodani fault, central Japan,
*Bull. Seismol. Soc. Am.*,**98**, 2170–2190, 2008.CrossRefGoogle Scholar - Kanno, T., A. Narita, N. Morikawa, H. Fujiwara, and Y. Fukushima, A new attenuation relation for strong ground motion in Japan based on recorded data,
*Bull. Seismol. Soc. Am.*,**96**, 879–897, 2006.CrossRefGoogle Scholar - Kondo, H., Y. Awata, O. Emre, A. Dogan, S. Ozalp, F. Tokay, C. Yildirim, T. Yoshioka, and K. Okumura, Slip distribution, fault geometry, and fault segmentation of the 1944 Bolu-Gerede earthquake rupture, North Anatolian fault, Turkey,
*Bull. Seismol. Soc. Am.*,**95**, 1234–1249, 2005.CrossRefGoogle Scholar - Koto, B., On the cause of the great earthquake in central Japan,
*J. Coll. Sci., Imp. Univ. Tokyo*,**5**, 295–353, 1893.Google Scholar - Kumamoto, T., A. Okada, S. Toda, K. Ueda, and Y. Ikeda, Trenching study of the Neodani fault of the Nobi active fault system at Kinbara, Gifu Prefecture in 1991 (1991 nen Nobi katsudansokei Neodani danso (Kinbara chiku) torenchi chousa),
*Active Fault Res.*,**10**, 85–91, 1992 (in Japanese).Google Scholar - Madariaga, R., On the relation between seismic moment and stress drop in the presence of stress and strength heterogeneity,
*J. Geophys. Res.*,**84**(B5), 2243–2250, 1979.CrossRefGoogle Scholar - Matsuda, T., Surface faults associated with Nobi (Mino-Owari) earthquake of 1891,
*Japan, Spec. Rep. Earthq. Res. Inst.*,**13**, 85–126, 1974 (in Japanese).Google Scholar - McCalpin, J. P.,
*Paleoseismology*, 583 pp, Academic Press, 1996.Google Scholar - McGill, S. F. and C. M. Rubin, Surficial slip distribution on the central Emerson fault during the June 28, 1992, Landers earthquake, California,
*J. Geophys. Res.*,**104**, 4811–4833, 1999.CrossRefGoogle Scholar - Midorikawa, S., M. Matsuoka, and K. Sakugawa, Site effects on strong motion records observed during the 1987 Chiba-Ken-Toho-Oki, Japan Earthquake,
*Proceeding of 9th Japan Earthquake Engineering Symposium*, E85–90, 1994.Google Scholar - Mikumo, T. and M. Ando, A search into the faulting mechanism of the 1891 great Nobi earthquake,
*J. Phys. Earth*,**24**, 63–87, 1976.CrossRefGoogle Scholar - Miyakoshi, K., S. Ogata, T. Kakuda, Y. Satake, K. Tanaka, K. Kiho, Y. Inohara, and Y. Kanaori, Activity of the Neodani fault: characteristics and analysis of fault movement at Kinbara, Gifu Prefecture,
*Industry Report number U88052, Central Research Institute of Electric Power Industry*, 1988 (in Japanese with English abstract).Google Scholar - Muramatu, I., Distribution of seismic intensity and crustal deformation in the region destroyed by the great Nobi earthquake of October 28, 1891,
*Res. Rep., Gifu Univ.*,**3**, 202–224, 1963 (in Japanese).Google Scholar - Muramatu, I., Distribution of the percentage of collapsed houses in the Nobi plain for the Nobi earthquake of 1891,
*Res. Rep., Gifu Univ.*,**7**, 867–882, 1983 (in Japanese with English abstract).Google Scholar - Muramatu, I. and M. Kominami, Collection of answers for a questionnaire investigation soon after the Nobi earthquake of 1891,
*Technical Note of the National Research Institute for Earth Science and Disaster Prevention*,**155**, 1–841, 1992 (in Japanese with English abstract).Google Scholar - Muramatu, I., T. Matsuda, and A. Okada,
*The 1891 Nobi Earthquake and Neo-dani Fault Zone*, 340 pp, Kokon Shoin, 2002 (in Japanese).Google Scholar - Nakano, M., J. Miyakoshi, and K. Yamaoka, A new model for the fault beneath the sedimentary basin in the 1891 Nobi earthquake,
*Earth Planets Space*,**59**, 13–19, 2007.CrossRefGoogle Scholar - Nakata, T., K. Shimazaki, Y. Suzuki, and E. Tsukuda, Fault branching and directivity of rupture propagation,
*J. Geogr.*,**107**, 512–528, 1998 (in Japanese with English abstract).CrossRefGoogle Scholar - Okada, A., M. Watanabe, M. Ando, T. Tsukuda, and S. Hirano, Estimation of paleo-seismicity in the Nobi active fault system, central Japan: excavation study of the Umehara fault, central strand in the Nobi active fault system,
*J. Geogr.*,**101**, 1–18, 1992 (in Japanese with English abstract).CrossRefGoogle Scholar - Omori, F.,
*General Report on the Nobi Earthquake*, Gifu Observatory, 48–88, 1894 (in Japanese).Google Scholar - Sato, H., A. Okada, T. Matsuda, and T. Kumamoto, Geology of a trench across the Midori fault scarp, from the Nobi earthquake of 1891, central Japan,
*J. Geogr.*,**101**, 556–572, 1992 (in Japanese with English abstract).CrossRefGoogle Scholar - Satoh, T., Radiation pattern and fmax of the Tottori-ken Seibu earthquake and the aftershocks inferred from KiK-net strong motion records,
*J. Struct. Constr. Eng., AIJ*,**556**, 25–34, 2002 (in Japanese with English abstract).Google Scholar - Satoh, T., H. Kawase, and T. Sato, Statistical spectral characteristics for engineering bedrock waves in which local site effects of surface geology are removed—Based on the ground motion records of small and medium earthquakes observed in the boreholes in Sendai—,
*J. Struct. Constr. Eng., AIJ*,**462**, 79–89, 1994 (in Japanese with English abstract).Google Scholar - Si, H. and S. Midorikawa, New attenuation relationships for peak ground acceleration and velocity considering effects of fault type and site condition,
*J. Struct. Constr. Eng., AIJ.*,**523**, 63–70, 1999 (in Japanese with English abstract).Google Scholar - Sieh, K., The repetition of large-earthquake ruptures,
*Proc. Natl. Acad. Sci.*,**93**, 3764–3771, 1996.CrossRefGoogle Scholar - Somerville, P. G., K. Irikura, R. Graves, S. Sawada, D. Wald, N. Abrahamson, Y. Iwasaki, T. Kagawa, N. Smith, and A. Kowada, Characterizing crustal earthquake slip models for the prediction of strong ground motion,
*Seismol. Res. Lett.*,**70**, 59–80, 1999.CrossRefGoogle Scholar - Sugisaki, R. and K. Shibata, Geochemical study on ground water (I): subsurface geology and aquifers in the Nobi plain,
*J. Geol. Soc. Jpn.*,**67**, 335–345, 1961 (in Japanese with English abstract).CrossRefGoogle Scholar - Sugiyama, Y., H. Sekiguchi, Y. Awata, Y. Fukushima, and K. Shimokawa, Correlation analysis between active fault parameters, and heterogeneous source characteristics,
*Report for Study on the Master Model for Strong Ground Motion Prediction Toward Earthquake Disaster Mitigation*, 119–129, 2002 (in Japanese).Google Scholar - The Headquarters for Earthquake Research Promotion, Evaluation of Gifu-Ichinomiya fault zone, http://www.jishin.go.jp/main/chousa/01jan2/index.htm, 2001 (in Japanese).Google Scholar
- The Headquarters for Earthquake Research Promotion, Examination of strong ground motion prediction method based on recorded seismograms for the 2000 Western Tottori Prefecture earthquake, http://www.jishin.go.jp/main/chousa/02oct_tottori/index.htm, 2002 (in Japanese).Google Scholar
- The Headquarters for Earthquake Research Promotion, Evaluation of strong ground motion for scenario earthquake in the Takayama and the Ohara fault zone, http://www.jishin.go.jp/main/kyoshindo/04sep_takayama/setsumei.pdf, 2004 (in Japanese).Google Scholar
- The Headquarters for Earthquake Research Promotion, Evaluation of strong ground motion for scenario earthquake in the Yamasaki fault zone, http://www.jishin.go.jp/main/kyoshindo/05jan_yamasaki/setsumei.pdf, 2005 (in Japanese).Google Scholar
- The Headquarters for Earthquake Research Promotion, Examination of strong ground motion prediction method based on recorded seismograms for the 2005 West Off Fukuoka Prefecture earthquake, http://www.jishin.go.jp/main/kyoshindo/08apr_fukuoka/index.htm, 2008a (in Japanese).Google Scholar
- The Headquarters for Earthquake Research Promotion, Strong ground motion prediction method (“Recipe”) for earthquakes with specified source faults, http://www.jishin.go.jp/main/kyoshindo/08apr_kego/recipe.pdf, 2008b (in Japanese).Google Scholar
- The Research Group for Active Faults of Japan,
*Active Faults in Japan*, 437 pp, University of Tokyo Press, 1991.Google Scholar - Wald, D. and T. Heaton, Spatial and temporal distribution of slip for the 1992 Landers, California, earthquake,
*Bull. Seismol. Soc. Am.*,**84**, 668–691, 1994.Google Scholar - Wells, D. L. and K. J. Coppersmith, New empirical relationships among magnitude, rupture length, rupture width, rupture area, and surface displacement,
*Bull. Seismol. Soc. Am.*,**84**, 974–1002, 1994.Google Scholar - Wesnousky, S. G., Displacement and geometrical characteristics of earthquake surface ruptures: Issues and implications for seismic-hazard analysis and the process of earthquake rupture,
*Bull. Seismol. Soc. Am.*,**98**, 1609–1632, 2008.CrossRefGoogle Scholar - Wessel, P. and W. H. F. Smith, New, improved version of the Generic Mapping Tools released,
*Eos Trans. AGU*,**79**, 579, 1998.CrossRefGoogle Scholar - Working Group on California Earthquake Probabilities, Seismic hazards in Southern California; Probable earthquake, 1994 to 2024,
*Bull. Seismol. Soc. Am.*,**85**, 379–439, 1995.Google Scholar - Yoshioka, T., Y. Awata, K. Shimokawa, H. Ishimoto, M. Yoshimura, and K. Matsuura, Paleoseismological study of the Nukumi fault, the 1891 Nobi earthquake fault system, central Japan,
*Annual Report on Active Fault and Paleoearthquake Researches*,**1**, 97–105, 2001 (in Japanese with English abstract).Google Scholar