Stochastic ground-motion simulations for the 2016 Kumamoto, Japan, earthquake
© The Author(s) 2016
Received: 16 June 2016
Accepted: 5 November 2016
Published: 18 November 2016
KeywordsGround-motion simulations Stochastic finite-fault method 2016 Kumamoto earthquake
In this paper, the ground-motion simulations for the 2016 Kumamoto earthquake have been performed using the stochastic finite-fault method based on a dynamic corner frequency. Our major objectives are to determinate region-specific source, path and site parameters, to validate our finite-fault model and to simulate ground motions in severely damaged areas that do not have strong-motion stations. For these purposes, we first determined input parameters including quality factor (Q s), zero-distance kappa (κ 0), which represents the effect of rapid spectral decay at high frequencies, site amplifications in the Kyushu region and the stress drop of the mainshock. Then, the ground-motion simulations were performed and compared with the observations at all stations. Simulated PGA values were also compared with ground-motion prediction equations (GMPEs) suggested by Boore et al. (2014). Finally, we performed blind simulations for Kumamoto Castle and Minami Aso village based on our validated model, where engineered structures suffered severe damages but without ground-motion records.
The stochastic finite-fault method is widely used in ground-motion simulations of past or scenario earthquakes (Ugurhan and Askan 2010; Ghofrani et al. 2013; Safarshahi et al. 2013; Zengin and Cakti 2014; Mittal and Kumar 2015). Compared with other ground-motion simulation methods, such as the deterministic or hybrid approach, the advantages of the stochastic method are its independence of small earthquake selection and good performance at both low and high frequencies (Motazedian and Atkinson 2005). In this method, near-field ground motions, including the acceleration time series, Fourier amplitude spectra (FAS) and 5%-damped pseudo-acceleration response spectra (PSAs), can be synthesized at the frequency range of engineering interest (Atkinson et al. 2009). In particular, the 5%-damped PSA represents the maximum acceleration caused by a linear oscillator with 5% damping and a specified natural period. In this study, we employ the latest version of the finite-fault code EXSIM12 to simulate the ground motions of the 2016 Kumamoto earthquake. For each station, we simulate surface recordings with ten trials to eliminate the bias resulting from stochastic variability.
Data and processing
Information on the strong-motion stations used in this study
Epicentral distance (km)
V S30 (m/s)
Preliminary data processing is implemented for each record. First, we employ the empirical approach with an automatic scheme (Wang et al. 2011) to correct baseline for raw data. The corrected records are band-pass filtered in the frequency range of 0.1–25 Hz with a fourth-order Butterworth filter. Second, the S-wave and noise windows are extracted and tapered at both ends with a 10% Hanning-type window. Finally, FAS are computed for both S-wave and noise windows.
Input parameters for the stochastic finite-fault model of the 2016 Kumamoto earthquake
Moment magnitude (M w)
32.754°N, 130.763°E, 11 km
Strike and dip angle (°)
224 and 65
Subfault length and width (km)
5 and 2.9
Stress drop (bars)
S-wave velocity (km/s)
3.6 (beneath the volcanic areas)
3.7 (beneath the non-volcanic areas)
Atkinson et al. (2009)
Rupture propagation velocity
Atkinson and Boore (2006)
Pulsing area percentage
Atkinson and Boore (2006)
Geometric spreading, R b : b=
−1.0 (0–100 km)
−0.5 (>100 km)
Zengin and Cakti (2014)
Ground-motion duration, dR, d=
0.0 (0–10 km)
+0.16 (10–70 km)
−0.03 (70–130 km)
+0.04 (>130 km)
Atkinson and Boore (1995)
Q s = (85.5 ± 1.5)f 0.68±0.01 (beneath the volcanic areas)
Q s = (120 ± 5)f 0.64±0.05 (beneath the non-volcanic areas)
See Additional files
0.0514 ± 0.0055
S-wave velocity (V s) anomalies are taken into account in this paper. Some studies (e.g., Zhao et al. 2011; Liu and Zhao 2016) report that low-velocity zones exist beneath the volcanic areas in the Kyushu region. For example, Zhao et al. (2011) revealed a −3% V s perturbation at a depth of 5–15 km under active volcanic areas in the Kyushu region using tomography method. In this study, we assume the V s is 3.6 and 3.7 km/s beneath the volcanic and non-volcanic areas, respectively, as constrained by the tomographic results (Zhao et al. 2011) and typical crustal V s profiles (Atkinson and Boore 2006).
Significant lateral Q s heterogeneities in the Kyushu region have been reported by many studies. Low-Q anomalies are revealed by tomography method in the crust and uppermost mantle beneath active volcanic areas, such as Aso, Sakurajima and Kirishima, whereas the subducting Philippine Sea slab exhibits high-Q character (e.g., Pei et al. 2009; Liu and Zhao 2015). Given the strong lateral Q s variations in the Kyushu region, we divide all stations into volcanic area stations and non-volcanic area stations (shown in Fig. 1; Table 1). The Q s beneath the volcanic area is estimated from the borehole recordings of 44 stations located in the volcanic areas, based on Liu and Zhao (2015)’s tomographic results [Fig. S29(a) of their work]. However, the number of stations in the non-volcanic area (9) is inadequate to provide a robust estimate of Q s. Therefore, we choose to estimate the non-volcanic area Q s by combining the volcanic area Q s and Q s perturbation identified from Fig. S29(a) of Liu and Zhao (2015). Specifically, the non-volcanic area Q 0 equals to the volcanic area Q 0 divided by 1 plus the Q s perturbation. For η, we use Oth et al. (2011)’s estimate for the Kyushu region. In addition, we also calculate Q s using all stations to investigate the effect of lateral Q s variations on ground-motion simulations.
Ground-motion duration is another important path parameter to trigger seismic hazards, such as liquefaction. We use a well-known distance-dependent duration model obtained by Atkinson and Boore (1995).
Our optimized finite-fault model consists of the source, path and site parameters. They are summarized in Table 2.
Results and discussion
Region-specific parameters are determined in this paper, including the quality factor (Q s), zero-distance kappa (κ 0), site amplifications and the stress drop of the mainshock.
We determine the frequency-dependent Q s in the Kyushu region by incorporating lateral variations. The Q s −1 beneath the volcanic area is estimated and expressed in the form of Q s −1 = (0.0117 ± 0.002)f −(0.68±0.01) (Fig. 3). In the simulations, it is transformed to Q s = (85.5 ± 1.5)f 0.68±0.01. It is in good agreement with Oth et al. (2011)’s result, Q s = (91 ± 8)f 0.64±0.05, which is derived from a nonparametric spectral inversion scheme and based on a much richer dataset. Additionally, Q s beneath the non-volcanic area is also estimated and shown in the form of Q s = (120 ± 5)f 0.64±0.05. Both Q 0 are consistent with the results of an S-wave attenuation tomography performed by Pei et al. (2009). Therefore, the Q s we estimated provides a convincing proof to describe the complexity of crustal structure in the Kyushu region. In addition, we also obtain the Q s without considering lateral variations in the form of Q s = (102 ± 5)f 0.67±0.01.
Site amplification in the Kyushu region is estimated by using the corrected cross-spectral ratios technique (see Additional file 1: Figure S1). Compared with the classic standard spectral ratio technique, the “depth effect” reported by Cadet et al. (2012) is corrected. The amplified waves usually lead to destructive damage to engineered structures at the surface. Therefore, a calibrated estimate of the site response in the target region is essential not only for ground-motion simulations, but for site classification, both of which are necessary for the earthquake-resistant design of engineered structures.
The stress drop based on the USGS slip model (Hayes 2016) is calculated by minimizing the average absolute residual of observed and simulated PSAs. The optimized value of stress drop, 64 bars, is reasonable for a large event (Ugurhan and Askan 2010; Zengin and Cakti 2014). The stress drop determined here contributes to reliable estimates of simulated PSA, because it controls the spectral amplitude at high frequencies (Motazedian and Atkinson 2005).
Our finite-fault model using optimized parameters is validated by comparison between the observed and simulated acceleration time series, FAS and 5%-damped PSAs at all stations. To validate the attenuation characteristics of PGAs against distance, GMPEs suggested by Boore et al. (2014) are used. Moreover, the simulated PGAs are also compared with the observations in regional scale. Note that we use horizontal-component geometric mean at the surface as observed ground-motion variables (i.e., FAS and 5%-damped PSA), whereas the simulated ones are the averages of ten trials. For synthetic acceleration time series, we select the one with its peak value closest to the average of ten trials. Comparison of observed and simulated time series, FAS and PSAs at all stations can be found in the Additional files (see Additional files 2: Figure S2, 3: Figure S3, 4: Figure S4). To avoid lengthy comparisons, we select one station at each epicentral distance range of 0–50, 50–100 and larger than 100 km, respectively, for further discussion.
Our finite-fault model with optimal parameters is validated well through the good agreement of observations and simulations in both the time and frequency domains at all stations. It shows the capacity to perform ground-motion simulations at some critical but not-instrumented sites.
The 2016 Kumamoto earthquake caused severe casualty and building damage. Investigating the ground-motion characteristics of this earthquake is essential for seismic hazard analysis and earthquake-resistant design. In this study, we use the stochastic finite-fault method based on a dynamic corner frequency to simulate the strong ground motions of the 2016 Kumamoto earthquake.
To achieve realistic simulation results, the source, path and site parameters in the Kyushu region are calibrated. For source effect, the stress drop of the mainshock is determined to capture the characteristics of high-frequency response spectrum, with an optimized value of 64 bars. For path effect, the low-Q anomalies beneath the volcanic area are revealed by frequency-dependent attenuation relation Q s = (85.5 ± 1.5)f 0.68±0.1. Besides, the S-wave attenuation beneath the non-volcanic area is also estimated and expressed as Q s = (120 ± 5)f 0.64±0.05. For site response, κ 0 with a value of 0.0514 ± 0.0055 s in the Kyushu region characterizes linear decay trend of FAS at high frequencies accurately, contributing to PGA, spectral level of ground motions and the calibration of GMPEs in engineering seismology. In addition, site amplifications in the Kyushu region are also estimated to quantify impedance effect, while seismic waves traverse soil layers near the surface, which is necessary for ground-motion simulations and site classification. Overall, the calibration of region-specific parameters contributes to reliable ground-motion simulations.
In conclusion, the stochastic finite-fault method based on a dynamic corner frequency is able to reproduce the ground-motion characteristics of the 2016 Kumamoto earthquake. Our finite-fault model is validated by the comparison of observations and simulations at all stations and regional scale, which shows good agreements in both the time and frequency domains. Attenuation characteristics of the simulated PGAs are also captured well by the GMPEs. Finally, we simulate the ground motions for two severely damaged regions, Kumamoto Castle and Minami Aso village, with the validated model. The simulation results can be further considered as an input for slope stability analysis, landslide run-out estimation and response spectrum analysis of structures. This work provides a deep insight into seismic hazard assessment and mitigation in the Kyushu region, Japan.
LZ conceived this study, ran the simulations, performed data analysis and drafted the manuscript. GC and YW revised the manuscript. HJ carried out the strong-motion data processing. All authors read and approved the final manuscript.
Ground-motion data and site information were obtained from the KiK-net at http://www.kyoshin.bosai.go.jp/, the NIED strong-motion seismograph networks, last accessed May 1, 2016. All the figures in this paper were prepared using Generic Mapping Tools (Wessel et al. 2013). The authors would like to thank Dr. R. Wang for providing the baseline correction code. We appreciate the helpful discussions with Z. Jiang, Z. Liu, S. Zhou, W. Wang and W. Feng. In addition, the authors thank Prof. Martha Savage (the editor) and two anonymous reviewers for their helpful comments on the draft manuscript. Also, this work was financially supported by National Science Foundation of China (41474002), Grant-in-Aid for challenging Exploratory Research (Grant No. 15K12483, G. Chen) from Japan Society for the Promotion of Science and Kyushu University Interdisciplinary Programs in Education and Projects in Research Development. These financial supports are gratefully acknowledged.
The authors declare that they have no competing interests.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
- Aki K, Chouet B (1975) Origin of coda waves: source, attenuation, and scattering effects. J Geophys Res 80:3322–3342. doi:https://doi.org/10.1029/JB080i023p03322 View ArticleGoogle Scholar
- Anderson JG, Hough SE (1984) A model for the shape of the Fourier amplitude spectrum of acceleration at high frequencies. Bull Seismol Soc Am 74:1969–1993Google Scholar
- Anderson J, Quaas R (1988) The Mexico earthquake of September 19, 1985—effect of magnitude on the character of strong ground motion: an example from the Guerrero, Mexico strong motion network. Earthq Spectra 4:635–646. doi:https://doi.org/10.1193/1.1585494 View ArticleGoogle Scholar
- Atkinson GM, Boore DM (1995) Ground-motion relations for eastern North America. Bull Seismol Soc Am 85:17–30Google Scholar
- Atkinson GM, Boore DM (2006) Earthquake ground-motion prediction equations for eastern North America. Bull Seismol Soc Am 96:2181–2205. doi:https://doi.org/10.1785/0120050245 View ArticleGoogle Scholar
- Atkinson GM, Assatourians K, Boore DM, Campbell K, Motazedian D (2009) A guide to differences between stochastic point-source and stochastic finite-fault simulations. Bull Seismol Soc Am 99:3192–3201. doi:https://doi.org/10.1785/0120090058 View ArticleGoogle Scholar
- Boore DM, Stewart JP, Seyhan E, Atkinson GM (2014) NGA-West2 equations for predicting PGA, PGV, and 5% damped PSA for shallow crustal earthquakes. Earthq Spectra 30:1057–1085. doi:https://doi.org/10.1193/070113EQS184M View ArticleGoogle Scholar
- Cadet H, Bard PY, Adrian RM (2012) Site effect assessment using KiK-net data: part 1. A simple correction procedure for surface/downhole spectral ratios. Bull Earthq Eng 10:421–448. doi:https://doi.org/10.1007/s10518-011-9283-1 View ArticleGoogle Scholar
- Ghofrani H, Atkinson GM, Goda K, Assatourians K (2013) Stochastic finite-fault simulations of the 2011 Tohoku, Japan, earthquake. Bull Seismol Soc Am 103:1307–1320. doi:https://doi.org/10.1785/0120120228 View ArticleGoogle Scholar
- Hayes G (2016) Preliminary finite fault results for the Apr 15, 2016 Mw 7.0 1 km WSW of Kumamoto-shi, Japan earthquake (Version 1). http://earthquake.usgs.gov/earthquakes/eventpage/us20005iis#finite-fault. Accessed 1 May 2016
- Kaklamanos J, Baise LG, Boore DM (2011) Estimating unknown input parameters when implementing the NGA ground-motion prediction equations in engineering practice. Earthq Spectra 27:1219–1235. doi:https://doi.org/10.1193/1.3650372 View ArticleGoogle Scholar
- Konno K, Ohmachi T (1998) Ground-motion characteristics estimated from spectral ratio between horizontal and vertical components of microtremor. Bull Seismol Soc Am 88:228–241Google Scholar
- Ktenidou O-J, Chávez-García FJ, Pitilakis KD (2011) Variance reduction and signal-to-noise ratio: reducing uncertainty in spectral ratios. Bull Seismol Soc Am 101:619–634. doi:https://doi.org/10.1785/0120100036 View ArticleGoogle Scholar
- Ktenidou OJ, Cotton F, Abrahamson NA, Anderson JG (2014) Taxonomy of κ: a review of definitions and estimation approaches targeted to applications. Seismol Res Lett 85:135–146. doi:https://doi.org/10.1785/0220130027 View ArticleGoogle Scholar
- Liu X, Zhao D (2015) Seismic attenuation tomography of the Southwest Japan arc: new insight into subduction dynamics. Geophys J Int 201:135–156. doi:https://doi.org/10.1093/gji/ggv007 View ArticleGoogle Scholar
- Liu X, Zhao D (2016) P and S wave tomography of Japan subduction zone from joint inversions of local and teleseismic travel times and surface-wave data. Phys Earth Planet Inter 252:1–22. doi:https://doi.org/10.1016/j.pepi.2016.01.002 View ArticleGoogle Scholar
- Mittal H, Kumar A (2015) Stochastic finite-fault modeling of Mw 5.4 earthquake along Uttarakhand–Nepal border. Nat Hazards 75:1145–1166. doi:https://doi.org/10.1007/s11069-014-1367-1 View ArticleGoogle Scholar
- Motazedian D, Atkinson GM (2005) Stochastic finite-fault modeling based on a dynamic corner frequency. Bull Seismol Soc Am 95:995–1010. doi:https://doi.org/10.1785/0120030207 View ArticleGoogle Scholar
- Oth A, Bindi D, Parolai S, Di Giacomo D (2011) Spectral analysis of K-NET and KiK-net data in Japan, part II: on attenuation characteristics, source spectra, and site response of borehole and surface stations. Bull Seismol Soc Am 101:667–687. doi:https://doi.org/10.1785/0120100135 View ArticleGoogle Scholar
- Pei S, Cui Z, Sun Y, Toksöz MN, Rowe CA, Gao X, Zhao J, Liu H, He J, Morgan FD (2009) Structure of the upper crust in Japan from S-wave attenuation tomography. Bull Seismol Soc Am 99:428–434. doi:https://doi.org/10.1785/0120080029 View ArticleGoogle Scholar
- Safarshahi M, Rezapour M, Hamzehloo H (2013) Stochastic finite-fault modeling of ground motion for the 2010 Rigan Earthquake, Southeastern Iran. Bull Seismol Soc Am 103:223–235. doi:https://doi.org/10.1785/0120120027 View ArticleGoogle Scholar
- Ugurhan B, Askan A (2010) Stochastic strong ground motion simulation of the 12 November 1999 Düzce (Turkey) earthquake using a dynamic corner frequency approach. Bull Seismol Soc Am 100:1498–1512. doi:https://doi.org/10.1785/0120090358 View ArticleGoogle Scholar
- Wang R, Schurr B, Milkereit C, Shao Z, Jin M (2011) An improved automatic scheme for empirical baseline correction of digital strong-motion records. Bull Seismol Soc Am 101:2029–2044. doi:https://doi.org/10.1785/0120110039 View ArticleGoogle Scholar
- Wessel P, Smith WHF, Scharroo R, Luis J, Wobbe F (2013) Generic mapping tools: improved version released. EOS Trans AGU 94:409–410. doi:https://doi.org/10.1002/2013EO450001 View ArticleGoogle Scholar
- Zengin E, Cakti E (2014) Ground motion simulations for the 23 October 2011 Van, Eastern Turkey earthquake using stochastic finite fault approach. Bull Earthq Eng 12:627–646. doi:https://doi.org/10.1007/s10518-013-9527-3 View ArticleGoogle Scholar
- Zhao D, Wei W, Nishizono Y, Inakura H (2011) Low-frequency earthquakes and tomography in western Japan: insight into fluid and magmatic activity. J Asian Earth Sci 42:1381–1393. doi:https://doi.org/10.1016/j.jseaes.2011.08.003 View ArticleGoogle Scholar