11:00 AM - 1:00 PM
[SSS10-P04] Comparison of short-period source spectral level estimated by three methods
Keywords:short-period source spectral level , strong motion predicton, scaling, crustal earthquake, rise time, slip velocity
Acceleration source spectral value from corner frequency to fmax is called “short-period source spectral level” (Dan et al., 2001). The short-period source spectral level A is one of important parameters for strong motion predictions. There are three methods to estimate A of large earthquakes. We compare the scaling relations of M0 and A for crustal earthquakes estimated by the three methods.
The first method of estimating A is the method proposed by Dan et al.(2001). The A is calculated from slip-velocity functions estimated by waveform inversions based on multi-time window analysis. In this method, maximum slip velocity of each subfault is defined as the average slip velocity in the time from 10 to 70 % of final slip. We call this the effective maximum slip velocity (Veff) in this study. The A is calculated using Veff in the method by Dan et al.(2001). Dan et al.(2001) proposed the M0-A relation (A=2.46×1017M01/3) using 12 crustal earthquakes with Mw5.6-7.2 in the world before 1995. We calculate A by the Dan et al.’s method using source inversion models on the website of SRCMOD (Mai et al., 2014), Asano and NIED. The A of 12 crustal earthquakes with Mw5.5-7.1 in Japan after 1997 and the 1999 Chi-Chi earthquake (Mw7.6) is about 1/3 to 1/4 of the M0-A relation by Dan et al.(2001). This difference is caused by the difference of Veff between source inversion models used by Dan et al.(2001) and this study. The maximum of Veff of each earthquake used by Dan et al.(2001) is from 2 to 20 m/s. On the other hand, the maximum of Veff of each earthquake used by this study is from 0.5 to 2 m/s. Source inversion models used by Dan et al.(2001) were estimated using one to a few time-windows. Source inversion models used by this study were estimated using about 5 to 10 time-windows. Resultantly, the rise time of the source inversion models used by Dan et al.(2001) is about 1/3 to 1/4 of those used by this study. It is interpreted that Veff of earthquakes used by Dan et al.(2001) becomes large to estimate slip by using short rise-time.
The second method of estimating A is the spectral inversion method. We estimate A of 14 crustal earthquakes with Mw5.5-7.0 in Japan by the spectral inversion method using K-NET, KiK-net and S-net strong motion records. Then we compile M0-A data of 14 earthquakes in this study and 18 earthquakes by Satoh(2010), Satoh and Tsutsumi (2012) and Satoh (2017). The M0-A relation is calculated assuming that the A is proportional to M01/3 using data of 30 earthquakes with Mw5.7-7.0. The obtained relation is A=2.31×1017 M01/3 which is almost the same to the relation by Dan et al.(2001). The A of dip-slip, strike-slip, and normal fault earthquakes are 1.13 times, 0.71 times and 0.70 times of A by Dan et al.(2001), respectively.
The third method of estimating A is to calculate it from broadband source models estimated by the empirical Green’s function method. We survey 14 previous broadband source models of 10 crustal earthquakes in Japan and calculate the A. Then we compile them with A of 10 earthquakes surveyed by Satoh (2010). The M0-A relation is calculated assuming that A is proportional to M01/3 using the data of 20 earthquakes with Mw5.7-7.0. The obtained relation is A=2.36×1017 M01/3 which is almost the same to the relation by Dan et al.(2001). The A of dip-slip and strike-slip earthquakes are 1.06 times and 0.91 times of A by Dan et al.(2001), respectively.
It is found that the method by Dan et al.(2001) is not applicable to estimate A using recent waveform source inversion models. On the other hand, the M0-A relation by Dan et al.(2001) agrees well with M0-A relations estimated by the spectral inversion method and the broadband source models for crustal earthquakes with Mw5.7-7.0 in Japan.
Acknowledgements: We use K-NET, KiK-net and S-net records by NIED, F-net by NIED and unified hypocenter information by JMA. This study is a part of cooperative research of 12 electric power companies.
The first method of estimating A is the method proposed by Dan et al.(2001). The A is calculated from slip-velocity functions estimated by waveform inversions based on multi-time window analysis. In this method, maximum slip velocity of each subfault is defined as the average slip velocity in the time from 10 to 70 % of final slip. We call this the effective maximum slip velocity (Veff) in this study. The A is calculated using Veff in the method by Dan et al.(2001). Dan et al.(2001) proposed the M0-A relation (A=2.46×1017M01/3) using 12 crustal earthquakes with Mw5.6-7.2 in the world before 1995. We calculate A by the Dan et al.’s method using source inversion models on the website of SRCMOD (Mai et al., 2014), Asano and NIED. The A of 12 crustal earthquakes with Mw5.5-7.1 in Japan after 1997 and the 1999 Chi-Chi earthquake (Mw7.6) is about 1/3 to 1/4 of the M0-A relation by Dan et al.(2001). This difference is caused by the difference of Veff between source inversion models used by Dan et al.(2001) and this study. The maximum of Veff of each earthquake used by Dan et al.(2001) is from 2 to 20 m/s. On the other hand, the maximum of Veff of each earthquake used by this study is from 0.5 to 2 m/s. Source inversion models used by Dan et al.(2001) were estimated using one to a few time-windows. Source inversion models used by this study were estimated using about 5 to 10 time-windows. Resultantly, the rise time of the source inversion models used by Dan et al.(2001) is about 1/3 to 1/4 of those used by this study. It is interpreted that Veff of earthquakes used by Dan et al.(2001) becomes large to estimate slip by using short rise-time.
The second method of estimating A is the spectral inversion method. We estimate A of 14 crustal earthquakes with Mw5.5-7.0 in Japan by the spectral inversion method using K-NET, KiK-net and S-net strong motion records. Then we compile M0-A data of 14 earthquakes in this study and 18 earthquakes by Satoh(2010), Satoh and Tsutsumi (2012) and Satoh (2017). The M0-A relation is calculated assuming that the A is proportional to M01/3 using data of 30 earthquakes with Mw5.7-7.0. The obtained relation is A=2.31×1017 M01/3 which is almost the same to the relation by Dan et al.(2001). The A of dip-slip, strike-slip, and normal fault earthquakes are 1.13 times, 0.71 times and 0.70 times of A by Dan et al.(2001), respectively.
The third method of estimating A is to calculate it from broadband source models estimated by the empirical Green’s function method. We survey 14 previous broadband source models of 10 crustal earthquakes in Japan and calculate the A. Then we compile them with A of 10 earthquakes surveyed by Satoh (2010). The M0-A relation is calculated assuming that A is proportional to M01/3 using the data of 20 earthquakes with Mw5.7-7.0. The obtained relation is A=2.36×1017 M01/3 which is almost the same to the relation by Dan et al.(2001). The A of dip-slip and strike-slip earthquakes are 1.06 times and 0.91 times of A by Dan et al.(2001), respectively.
It is found that the method by Dan et al.(2001) is not applicable to estimate A using recent waveform source inversion models. On the other hand, the M0-A relation by Dan et al.(2001) agrees well with M0-A relations estimated by the spectral inversion method and the broadband source models for crustal earthquakes with Mw5.7-7.0 in Japan.
Acknowledgements: We use K-NET, KiK-net and S-net records by NIED, F-net by NIED and unified hypocenter information by JMA. This study is a part of cooperative research of 12 electric power companies.