Source Parameters of the 2005 Mw 7 . 2 Miyagi-Oki , Japan , Earthquake as Inferred from Teleseismic P-Waves

We investigate the fault parameters of the 2005 Miyagi-Oki (Japan) earthquake using duration variations of teleseismic P-waves. The results show that the earthquake has a thrust-type mechanism and a seismic moment of 4.46 × 1019 Nm. Rupture directivity analysis suggests that the earthquake occurred as a result of a bilateral faulting on the fault plane with a strike of 247°, a dip of 17° and a slip of 125°. The optimal rupture azimuth, measured counterclockwise from the strike on the fault plane, is 170° (or 350°). The rupture length and average source duration are estimated to be 73.4 km and 14.5 sec, respectively. Thus the rupture velocity is 2.53 km sec-1 (~0.57 times the value of S-wave velocity), which is lower than the value for other similarly sized earthquakes. This implies that the 2005 Miyagi-Oki earthquake was probably a slow event. Consequently, there may have been less release of high-frequency seismic energy, leading to lower radiated seismic energy and radiation efficiency (~0.32 0.48). In other words, relatively larger fracture energy occurred during earthquake faulting in addition to the heat due to friction. The ratio of the static stress drop to the apparent stress (> 2.0) also suggests that the earthquake can be modeled as a frictional overshoot in a stress model, which implies the transformation of a lower percentage of strain energy into seismic-wave energy during the process of earthquake rupturing.


InTrOducTIOn
Fault parameters and rupture directivity for earthquakes provide important information for understanding the physical behavior of earthquake ruptures (Ben-Menahem 1961;Kanamori and Heaton 2000).Generally, the focal mechanism is determined using moment tensors inversion (e.g., Dziewonski et al. 1981;Sipkin 1994); the estimation of the rupture directivity of an earthquake is primarily based on variation of the source duration with a station azimuth and source finiteness theory (cf.Ben-Menahem 1961;Chung and Kanamori 1980).Several methods are employed to derive source-duration variations with an azimuth, for example, using empirical Green's function and forward wave-form modeling (e.g., Kanamori and Stewart 1976;Chung and Kanamori 1980;Ruff and Kanamori 1983;Ammon et al. 1993).Macroscopic source parameters (including radiated seismic energy, fracture energy and rupture velocity, etc.) are also helpful to understand earthquake rupture mechanisms (Kanamori 1994;Kanamori and Heaton 2000;Venkataraman and Kanamori 2004).Radiated seismic energy can be determined either from teleseismic P-waveforms or from a source time function (e.g., Vassiliou and Kanamori 1982; Kikuchi and Fukao 1988;Choy and Boatwright 1995;Pérea-Campos and Beroza 2001;Bilek et al. 2004;Venkataraman and Kanamori 2004).In general, the ratio (E s /M o ) of radiated seismic energy (E s ) to seismic moment (M o ) is related to the dynamics of earthquake faulting; hence E s /M o can be employed to obtain some indication of the condi-tion of stress during an earthquake (Wyss and Brune 1968;Kanamori and Heaton 2000;Hwang et al. 2001;Kanamori and Rivera 2004).A large E s /M o indicates a rapid drop in friction during earthquake faulting, while a relatively low E s /M o indicates that friction drops more slowly (Kanamori and Heaton 2000).However, there is high divergence in the E s /M o derived in previous studies because of uncertainty in the estimated radiated seismic energy.For example, Kikuchi and Fukao (1988) estimated the E s /M o for large earthquakes to be 5.0 × 10 -6 ; Choy and Boatwright (1995) found a global average E s /M o of 1.6 × 10 -5 ; Vassiliou and Kanamori (1982) and Kanamori and Heaton (2000) found the E s /M o for large earthquakes to be 5.0 × 10 -5 ; Pérea-Campos and Beroza ( 2001) obtained an E s /M o value of 4.7 × 10 -6 for reverse faulting earthquakes; Bilek et al. ( 2004) found a relatively lower E s /M o of 5.6 × 10 -6 ; while Hwang et al. (2001) estimated a relatively larger E s /M o of 1.0 -3.0 × 10 -4 for the 1999 Chi-Chi earthquake from near-field seismograms.Moreover, usage of a finite frequency bandwidth usually leads to underestimation of radiated seismic energy (Boore 1986;Di Bona and Rovelli 1988;Ide and Beroza 2001;Wang 2004;Wang and Huang 2007).In spite of the difficulty in estimating radiated seismic energy, it is known that the E s /M o of strike-slip earthquakes is significantly different from that of reverse and normal earthquakes, i.e., E s /M o is dependent on focal mechanism (e.g., Choy and Boatwright 1995;Pérez-Campos and Beroza 2001).Finally, it should be noted that tsunami earthquakes have a lower E s /M o than do other types (Newman and Okal 1998;Bilek et al. 2004).In addition, the relationship between the static stress drop and apparent stress also can indicate the stress mechanism, a partial-stress drop model or a frictional overshoot mechanism (Zúñiga 1993).
On 16 August 2005, a large earthquake (M w = 7.2) occurred offshore near Miyagi Prefecture in northeastern Japan.This earthquake had a focal depth of 36 km as reported by USGS.The epicenter is displayed in Fig. 1.Also shown in Fig. 1 are the station distribution and aftershocks that occurred over the following 2 months.The location of the 2005 Miyagi-Oki earthquake was close to the epicenter of an earthquake, which occurred in 1978 (Okada et al. 2005;Yagi 2005).Earthquakes with M w ≥ 7.0 have occurred repeatedly in northeastern Japan, at an interval of approximately 30-year (cf.Yamanaka and Kikuchi 2004).Okada et al. (2005) analyzed the rupture process and suggested that the 2005 Miyagi-Oki earthquake possibly caused rupturing in the source area of the 1978 Miyagi-Oki earthquake.They also found that the coseismic slip distribution more or less followed the dip-direction of the fault.However, Yagi (2005) and the ERI report (EIC Seismological Note: No.168) showed that the coseismic slip was distributed along the strike-direction of the fault.Furthermore, the aftershocks of the 2005 Miyagi-Oki earthquake (see Fig. 1) occurred mainly along the dip-direction of the fault.They mostly surrounded the main area of the rupturing as derived by Yagi (2005).In other words, both Yagi's results and the coseismic slip distributions in the ERI report seem to agree well with the asperity source model (e.g., Ruff and Kanamori 1983).In this study, we attempt to infer the source mechanism and parameters of the Miyagi-Oki earthquake by utilizing the far-field P-waveform inversion procedure proposed by Lin et al. (2006).With this inversion process, we can simultaneously derive the fault plane solution, the seismic moment, and the rupture directivity of the earthquake.The estimated fault parameters can then be used to calculate the macroscopic source parameters, which give us a better understanding about the physical behavior of rupture during the 2005 Miyagi-Oki earthquake.

METhOd
The method proposed by Lin et al. (2006) is employed in this study.It differs from both the moment tensor inversion method (e.g., Dziewonski et al. 1981;Sipkin 1994) and empirical Green's function (Ammon et al. 1993).Traditionally, for a shallow earthquake, a synthetic far-field seismogram at a particular receiver can be expressed in the following form [for details, see Kanamori and Stewart (1976) and Okal (1992)]: ( ) u t P is the synthetic far-field seismogram; M o is the seismic moment; α h , β h , and h t are the P-wave velocity, S-wave velocity and density in the source area, respectively; g(Δ) denotes the geometrical spreading factor as a function of epicentral distance (Δ) and focal depth; r is the radius of the Earth (about 6371 km); R P , R pP , and R sP are the radiation patterns for the P-, pP-and sP-waves (cf.Kanamori and Stewart 1976;Aki and Richards 1980); V pP and V sP denote the reflection coefficients of the P-to P-wave and the S-to P-wave at the free surface (cf.Aki and Richards 1980); i h and j h are the take-off angles of the P-and S-waves just leaving the source and i o is the incident angle of the P-wave in relation to the free surface; ( ) C i P o is the free surface effect at the receiver (cf.Aki and Richards 1980); f(t) stands for a triangular source time function and its duration represents the source-process time; t P , t pP , and t sP are the travel times for the P-, pP-and sP-waves, which are calculated according to a given Earth velocity model; Q(t) is the attenuation filter of the Earth based on Azimi's law (e.g., Yoshida 1988); I(t) is the instrumental response.Figure 2 shows a sketch of generating a synthetic far-field seismogram, exclusive of the instrumental response.
After taking into account some corrections to the seismogram, we can rewrite Eq. ( 1) in the following form: where ( ) u t c P is the corrected seismogram, not inclusive of the instrumental effect, geometrical spreading, the free surface effect, etc.The unknown parameters in Eq. ( 2) are M o , R P , R pP , and R sP , that is, the seismic moment and radiation patterns.For convenience in solving for the unknown parameters for each station, Eq. ( 2) can be expressed in matrix form with n data points and 3 unknown parameters The iasp91 velocity model (Kennett and Engdahl 1991) is used to calculate the travel time of seismic-waves as well as several parameters.The attenuation filter Q(t) is calculated by Azimi's law and using t * = 1.0 (travel time  over quality factor) for the P-wave (cf.Okal 1992).The parameters to be solved in Eq. ( 3) are M o R P , M o R pP , and M o R sP , which are the so-called pseudo radiation patterns (Lin et al. 2006).The pseudo radiation pattern is defined as the product of seismic moment (M o ) and radiation pattern (R P , R pP , R sP ).Equation ( 3) can be solved by the standard leastsquares technique (cf.Menke 1984).The inversion process in Eq. ( 3) has two solutions.One is the source duration; the other is the pseudo radiation pattern.In other words, we find the corresponding pseudo radiation pattern when giving a source time function with a fixed source duration.Thus, we search for a succession of source time functions with various source durations.The appropriate source duration and pseudo radiation patterns are obtained for each station by minimizing misfits between the observed and synthetic seismograms.The fault plane solution and seismic moment are derived afterwards from a grid search of the strike, dip and slip angles of the fault by making a comparison between the theoretical and pseudo radiation patterns at an interval of 1° (also refer to Lin et al. 2006).The rupture feature of the earthquake, i.e., unilateral or bilateral faulting mechanism, can be judged through rupture directivity analysis based on the source duration variations with station azimuth (cf.Ben-Menahem 1961;Chung and Kanamori 1980;Lin et al. 2006).From the viewpoint of the source finiteness theory of Ben-Menahem (1961), the relationship between observed source duration and several fault parameters for an earthquake with unilateral faulting can be expressed as: and for bilateral faulting can be written as: where T is the source duration observed at a given station, L is the rupture length, V R is the rupture velocity, V C is the P-wave velocity in the source area, and Θ is the angle between the rupture direction and a ray taking off from the source.The rupture direction is measured counterclockwise from the strike of the fault on the fault plane.Through rupture directivity analysis, the fault parameters can be easily estimated by using the aforementioned equation through the least-squares technique.

dATA
The IRIS (Incorporated Research Institutions for Seismology) Data Management Center collected hundreds of seismograms which have been recorded by several seismic networks during the 2005 Miyagi-Oki earthquake.This pro-vides a wealth of high-quality seismic data from which to analyze the rupture behavior of the 2005 Miyagi-Oki earthquake.In order to eliminate the interference of depth phases (such as the PcP-wave) and multipathing waves (such as the PP-wave), only seismograms recorded at epicentral distances of 30° -90° are used.We extract the 50-sec-long P-waves, including the first 5 sec of the P-arrival and the 45 sec after the P-arrival, from the original seismogram.In the inversion, the travel times of the P-, pP-, and sP-waves, geometrical spreading, take-off angle, free surface effect and reflection coefficients are calculated according to the iasp91 velocity model (Kennett and Engdahl 1991).The P-wave velocity, S-wave velocity and density at the source area of the 2005 Miyagi-Oki earthquake are 8.04, 4.47 km sec -1 and 3.30 g cm -3 , respectively.Most of stations are located in the first and second quadrants (Fig. 1), but the station coverage does not greatly affect the eventual results obtained in this study, especially for the determination of fault plane solution (cf.Lin et al. 2006).

rESulTS
Initially, we search for the minimum misfit between the observed and synthetic seismograms from a succession of source duration to obtain the best source duration for each station.Meanwhile, the pseudo radiation patterns for each station are inverted using Eq.(3). Figure 3 shows several examples of the inversion process.The focal mechanism can be derived by a grid search of the status of the fault at an interval of 1°, noting when the misfit between the theoretical and pseudo radiation patterns has a minimum value (Fig. 4a).The fault plane solutions for the best double couple (strike, dip, slip) are (247°, 17°, 125°) and (31°, 76°, 79°), respectively, and the seismic moment is 4.46 × 10 19 Nm.
As shown in Fig. 3, we obtain the source duration observed in each station using Eq. ( 3).The maximum difference of the source duration in Fig. 3 is 4.3 sec (in fact, maximum difference larger than 10 sec), which is significant and indicates the rupture directivity of source.Hence, the fault parameters of the 2005 Miyagi-Oki earthquake can be determined by the rupture directivity analysis [i.e., from Eqs. (4) or ( 5)].We test two source models (the unilateral faulting and bilateral faulting models) to determine the appropriate faulting model and fault plane.In this procedure, we find the minimum misfit (i.e., best linear fitting as shown in Fig. 5) for a combination of bilateral faulting and fault plane (247°, 17°, 125°) when searching a series of Θ [rupture azimuths as indicated in Eqs. ( 4) and ( 5)].Thus, the optimal rupture azimuth is estimated at 170° (or 350°) measured counterclockwise from the strike of the fault on the fault plane (Fig. 5).The earthquake rupture apparently follows along the strike of the fault.Taking the P-wave velocity in the source area to be 8.04 km sec -1 (cf.Chung and Kanamori  1976, 1980;Lin et al. 2006;Warren and Shearer 2006), we estimate the rupture length to be 73.4 km from the slope as shown in Fig. 5.The average source duration (intercept in Fig. 5) and rupture velocity are 14.5 sec and 2.53 km sec -1 , respectively.
Assuming that L = 2W (L = rupture length, W = rupture width) for large earthquakes (cf.Geller 1976), we can estimate the average dislocation of the 2005 Miyagi-Oki earthquake to be 0.33 m over the whole fault plane using the following formula: M o = μDA, where M o is the seismic moment, μ is the rigidity (5.0 × 10 10 Nm -2 ), A is the area of the fault plane, and D is the average dislocation.Based on Brune's circular source model (Brune 1970), the static stress drop (Δσ) is estimated to be 7.7 bars.In addition, the radiated seismic energy (E s ) is calculated from the average source duration using the method of Vassiliou and Kanamori (1982) to be 5.85 × 10 13 Nm.Thus the ratio of radiated seismic energy to seismic moment (E s /M o ) is about 1.31 × 10 -6 .We also estimate the apparent stress (σ a ) at 0.65 bars using: σ a = μ(E s /M o ) (Wyss and Brune 1968).The estimated source parameters are listed in Table 1.

dIScuSSIOn
Before inversion through Eq. ( 3), the original seismograms are normalized by a factor of 8.0 × 10 19 Nm.Thus, the inverted pseudo radiation patterns in Fig. 3 are also normalized by this factor.The use of various normalizing factors does not cause differences in results when determining the fault plane solution and the seismic moment (cf.Lin et al. 2006).Because the seismic moment is a constant observed for each station, the pseudo radiation pattern varying with station azimuth represents the variation of radiation pattern in connection with the status of the fault.The focal mechanism obtained in our study is similar to that obtained in several other studies (Harvard CMT; USGS; ERI; Yagi 2005), as shown in Fig. 4b.From the rupture directivity analysis, it can be inferred that the 2005 Miyagi-Oki earthquake was a bilateral faulting event with a rupture direction of 170° (or 350°), almost following the strike of the fault on the fault plane.The models of the coseismic slip distribution from the teleseismic P-waves (Yagi 2005; ERI's EIC Seismological Note: No.168) also indicated that the rupture distribution mainly followed along the strike-direction on the fault plane and the large slips close to the epicenter of the main shock.This implies the presence of bilateral extension, which is consistent with our study.In contrast, our results differ slightly from the work of Okada et al. (2005), indicating relatively larger ruptures in the dip-direction on the fault.
Nonetheless, the aftershocks of the 2005 Miyagi-Oki earthquake were distributed along the dip-direction, where a region of weak slip is displayed (Yagi 2005;ERI's EIC Seismological Note: No.168).From the viewpoint of the asperity source model, aftershocks would occur mostly around the region of the main rupture of the fault plane (e.g., Ruff and Kanamori 1983).The slip models from Yagi (2005) and ERI seem to better interpret the connection between the coseismic slip model and the distribution of the aftershocks.The estimated rupture length is about 73.4 km, which is also comparable with both Yagi's result (~90 km) and the ERI report (~60 km), but differs from that obtained by Okada et al. (2005).Our estimate of rupture length is slightly different from Yagi (2005) and ERI and might be regarded as the effective rupture length of the earthquake fault.The average source duration ~14.5 sec is slightly smaller than the results of Yagi (2005), ERI or the Harvard CMT (18 ~ 25 sec).The rupture velocity for this earthquake is estimated to be 2.53 km sec -1 , ~0.57Vs (Vs = 4.47 km sec -1 in the source area), also slower than that for similarly sized earthquakes (~0.8 -0.9 times the S-wave velocity).A slow rupture velocity might convey information suggesting the need for a tsunami warning.
The estimated E s /M o , 1.31 × 10 -6 coincides with that found by Pérez-Campos and Beroza (2001) for reverse earthquakes, but is still lower than global observations (Choy and Boatwright 1995) and about 4 times as small as that given in a USGS report.It is likely that the E s (Table 1) estimated from the simple source time function in this study may have been underestimated as a result of the elimination of highfrequency energies (e.g., Kikuchi and Fukao 1988;Bilek et al. 2004).This is the so-called finite frequency bandwidth limitation, which is a problem when estimating the radiated seismic energy from seismic waveforms (Boore 1986;Di Bona and Rovelli 1988;Ide and Beroza 2001;Wang 2004;Wang and Huang 2007).However, when calculating the E s from the source duration by using the method of Vassiliou and Kanamori (1982), one cannot introduce the finite frequency bandwidth limitation to estimates of E s .Multiple event analysis might improve the estimation of E s to some degree when using Vassiliou and Kanamori's method (e.g., Kikuchi and Fukao 1988;Hwang et al. 2008;Hwang and Wu 2009).
The estimated static stress drop, 7.7 bars, corresponds to the value for interplate earthquakes (Kanamori and Anderson 1975), but differs from that (17.0 bars) estimated from a fault area of 60-km × 40-km and a seismic moment of 8.2 × 10 19 Nm based on the ERI report.The estimated apparent stress (0.65 bars) of the 2005 Miyagi-Oki earthquake agrees well with that for reverse events (~1.5 bars) found by Pérez-Campos and Beroza (2001).Kanamori and Rivera (2004) stated that E s /M o is related to the cube of the rupture velocity and static stress drop.A relationship among the minimum E s /M o (i.e., the lower bound of E s /M o ), the rupture velocity and the static stress drop de- rived by Kanamori and Rivera (2004) can be formulated as h, where μ the shear modulus, β the S-wave velocity, Δσ the static stress drop, and V R the rupture velocity.When Δσ = 7.7 bars, V = 2.53 km sec -1 , β = 4.47 km sec -1 and μ = 6.6 × 10 10 Nm -2 in our study, the value of / E M min s o ^h is calculated to be about 1.84 × 10 -6 .This is close to our estimate of E s /M o .If we use E s = 3.8 × 10 14 Nm from the USGS report, and M o = 8.2 × 10 19 Nm and Δσ = 15 bars from the ERI report, the estimated rupture velocity is about 3.08 km sec -1 , ~0.69 Vs (Vs = 4.47 km sec -1 in the source area), from the above-mentioned formula (Kanamori and Rivera 2004).This value is also smaller for this earthquake than those for large earthquakes (~0.9 Vs) (Kanamori and Heaton 2000;Kanamori and Rivera 2004).Additionally, estimates of rupture velocity seem to be slow from the slip distribution of Yagi (2005) and the ERI report.The two models show that the 2005 Miyagi-Oki earthquake is a bilateral faulting event and has source duration of 25 sec.From their fault dimensions, the average rupture velocity is probably 1.5 -2.0 km sec -1 .Hence, we suggest that the low E s and E s /M o might be due to the occurrence of slow rupturing during the earthquake faulting, resulting in the suppression of high-frequency energies.In addition, it should be noted that our results seem to be similar to those indicating tsunami earthquakes (Newman and Okal 1998;Bilek et al. 2004).On the other hand, the estimated E s /M o approximates that of an underthrusting type of earthquakes occurring in shallow subduction zones (Pérez-Campos and Beroza 2001;Bilek et al. 2004).The slow rupture velocity also leads to smaller radiation efficiency, ranging from ~0.32 -0.48, obtained from simple energy considerations and a Model III crack (cf.Venkataraman and Kanamori 2004).The lower radiation efficiency reflects the fact that relatively higher fracture energy occurred during the 2005 Miyagi-Oki earthquake relative to the radiated seismic energy.
Providing that the rise time is 0.1 -0.2 times as large as the whole source duration (Kanamori and Anderson 1975;Geller 1976), we also estimate the average particle velocity associated with the dynamic stress drop to be ~11 -22 cm sec -1 , which is much lower than the value for samesized earthquakes (Kanamori 1994).The low particle velocity suggests a low dynamic stress drop; in other words, high dynamic friction occurred during the 2005 Miyagi-Oki earthquake faulting.This meant that most of the strain energy due to the stress relaxation was transformed into heat.Furthermore, the lower apparent stress suggests a lower seismic efficiency for the Miyagi-Oki earthquake (Kikuchi and Fukao 1988).The apparent stress is smaller than half of its static stress drop (cf.Zúñiga 1993); this implies that the frictional overshoot stress model can be used to interpret the rupture behavior of the 2005 Miyagi-Oki earthquake.In this case, a lower percentage of strain energy will be transformed into seismic-wave energy.

cOncluSIOnS
Large earthquakes frequently occur in northeastern Japan where their rupture areas overlap and cover each other (Yamanaka and Kikuchi 2004).Okada et al. (2005) noted that the 2005 Miyagi-Oki earthquake ruptured the source area of the 1978 event.We investigate the source parameters of the 2005 Miyagi-Oki earthquake from teleseismic P-waves inversion based on the method of Lin et al. (2006).Our results indicate that this earthquake is a bilateral faulting event, and the findings coincide with those from several other studies (Yagi 2005; ERI's report).The macroscopic source parameters from our study suggest that the low particle velocity, low dynamic stress drop, low E s , and low E s /M o of the 2005 Miyagi-Oki earthquake were due to the obstruction of the 1978 event.In other words, it would take more energy to drive the faulting during the rupturing of the 2005 Miyagi-Oki earthquake in the locked source area where the 1978 event occurred.For this reason, the 2005 Miyagi-Oki earthquake had a relatively larger fracture energy, low dynamic stress drop, and low rupture velocity.Because (σ a /Δσ) < 0.5 (cf.Zúñiga 1993), our results also suggest that the stress model of frictional overshoot can be used to interpret the rupture of the 2005 Miyagi-Oki earthquake.
Acknowledgements We extend our thanks for the use of the facilities of the IRIS Data Management System, and specifically the IRIS Data Management Center who gave us access to the waveform and metadata required for this study.The IRIS DMS is funded by the National Science Foundation, specifically by the GEO Directorate through the Instrumentation, under Cooperative Agreement EAR-0004370.We also give special thanks to the two anonymous reviewers for critical suggestions in improving the readability of the manuscript.This study was funded by the National Science Council, ROC, under Grant Nos.NSC91-2119-M-432-001 and NSC92-2119-M-238-002.

Fig. 1 .
Fig. 1.(a) Distribution of the available stations (triangles) and the epicenter of the 2005 Miyagi-Oki earthquake (star).(b) Map showing the distribution of aftershocks (circles) with M w > 2.0.

Fig. 3 .
Fig. 3. Map showing the decision of the best source duration according to Lin et al. (2006).The bottom diagram for each station shows the estimate of the best source duration obtained from a successive search of source durations between 1 and 50 sec at 0.1 sec intervals.The arrow denotes the best source duration obtained from the grid search.The top diagram shows a comparison of the observed (solid line) and synthetic (dashed line) P-wave seismograms.Also shown are the station code and pseudo radiation pattern (M o R P , M o R pP , M o R sP ) determined from the best source duration.

Fig. 4 .Fig. 5 .
Fig. 4. (a) The best fault plane solution found through a grid search of the status of an earthquake (at an interval of 1°) obtained by comparing the theoretical and pseudo radiation patterns.The solid line indicates the best linear regression and its slope (~0.557) stands for the normalized seismic moment (by 8.0 × 10 19 Nm).(b) Comparison of the fault plane solution and seismic moment obtained in our study with those obtained from USGS, Harvard CMT, ERI and Yagi (2005).The corresponding fault plane solution and seismic moment are also shown below each beach ball.

Table 1 .
Source parameters estimated in this study.
* The rupture azimuth is measured counterclockwise from the strike of the fault on the fault plane.