2:00 PM - 2:15 PM
[SSS11-19] Improvements to the Stochastic Green's Function Method: Part 2 Comparison with Recorded Accelerations considering Effects of Moho Reflection and Layered Crust Structure
Keywords:Stochastic Green's Function Method, Wavenumber Integration Method, Refrection Waves from Moho and Layered Crust Structure, Aftershock of 2019 Off-Yamagata Prefecture Earthquake, Strong Motion Records of KiK-net
1. Introduction
Following the previous study (Hisada, 2024), we investigate the effects of the layered structure of the crust and mantle on short-period strong-motion in the Stochastic Green's Function method, and compare the results with strong-motion records.
2. Comparison of Strong Motions with Different Green's functions considering Moho reflection and crustal layer structures
The target earthquake is the aftershock (Mj 4.0) of the 2019 Off-Yamagata Prefecture earthquake shown in Figure 1 and Table 1. First, we compute strong ground motions using the average deep ground structure from the epicenter to the strong-motion observation sites. The material properties of the under-ground structure in Table 1 consists of the four layers from the upper crust to the mantle (Earthquake Headquarters, 2012), and the thickness of each layer is the average of the layers below the epicenter and two KiK-net observation points, YMTH13 and YMTH11. As shown in Table 1, we use two-types of Q values: constant Q values (Earthquake Headquarters, 2012), and Q values dependent on frequency (Sato, 2006). We use the source model of Hisada (2008), which radiates P, SH, and SV waves independently. The source spectrum is Brune's ω2 model, and the radiation coefficient is isotropic (P wave: 0.52, S wave: 0.63). The horizontal one component of the three waveforms obtained is divided by √2 and three-types of waves are simply added. The phase spectrum is generated as a pulse wave with zero phase at all frequencies to enable waveform tracking. We compute strong motions using Wavenumber integration method up to 12.5 Hz and filtered through a trapezoidal bandpass filter with corner frequencies at 0.3, 0.4, 10, and 12.5 Hz.
Figure 2 shows velocity waveforms and Fourier amplitude spectra of Transverse, Radial, and UD at epicentral distance of 100 km (using f-dependent Q values). There are four types of waveforms: “Deep structure” is a model using all four layers in Table 1, “Without Vs4.5” is a model without the fourth layer, “Without Vs4.5&3.8” is a model without the fourth and third layers, and “Uniform structure” is a model using only direct waves in infinite homogeneous full-space of the second layer (the source layer). In the “Deep structure” model, repeated reflections appear from the mantle and lower crust, but in the “Without Vs4.5” model, the Moho reflections in the latter half of the waveform disappear, and in the “Without Vs4.5&3.8” model, there are no reflections from the lower crust, the waveform amplitude is small, and its duration is short. The waves in “Uniform structure” shows relatively large amplitude, but this is due to the fact that the transmission coefficient from the source layer to the first layer, which decreases with distance, is not taken into account (Hisada, 2024). In the Fourier amplitude spectrum, “without Vs4.5” has a slightly smaller amplitude than “Deep structure,” and “without Vs4.5&3.8” and “Uniform ground” have about 1/3 the amplitude.
Figure 3 shows the comparisons among the observed and calculated accelerations for YMTH13 (epicentral distance 30.8 km) and YMTH11 (epicentral distance 96.7 km). The structure model is a connection between the deep structure (Earthquake Headquarters 2012) and the surface structure at the KiK-net stations. Two types of Q values, constant Q value and f-dependent Q value, are used for the structure deeper than the seismic bedrock, and a simple constant Q value (Qs=Vs/15, Qp=1.7Qs Qp=1.7Qs) for the shallower structures. We assume the phase spectrum was to be random phase for all frequencies, and use the two types of time envelope functions: Boore's (1983) and the empirical formula that duration increases with Mj and epicenter distance (Sato et al., 1994). Figure 3 shows four waveforms, where “Observation B” is the wave observed by KiK-net, “Boore+Q0” is the Boore envelope function and constant Q value model, “Sato+Q0” is the Sato envelope function and constant Q value model, and “Sato+Q=91f^1.03” is the calculation result using Sato envelope function and f-dependent Q value. The start time of the waveforms are the origin time of the earthquake. The simulation results for the two stations show slightly later arrival times and larger amplitudes. At YMTH13, which is closer to the epicenter, Boore function is close to the envelope of the main motion, while at the distant YMTH11, Sato function reproduces the long duration of the observed records well. A more detailed interpretation on the effects of reflected waves from deep structure and the surface ground, including other stations, will be given on the day of the presentation.
Following the previous study (Hisada, 2024), we investigate the effects of the layered structure of the crust and mantle on short-period strong-motion in the Stochastic Green's Function method, and compare the results with strong-motion records.
2. Comparison of Strong Motions with Different Green's functions considering Moho reflection and crustal layer structures
The target earthquake is the aftershock (Mj 4.0) of the 2019 Off-Yamagata Prefecture earthquake shown in Figure 1 and Table 1. First, we compute strong ground motions using the average deep ground structure from the epicenter to the strong-motion observation sites. The material properties of the under-ground structure in Table 1 consists of the four layers from the upper crust to the mantle (Earthquake Headquarters, 2012), and the thickness of each layer is the average of the layers below the epicenter and two KiK-net observation points, YMTH13 and YMTH11. As shown in Table 1, we use two-types of Q values: constant Q values (Earthquake Headquarters, 2012), and Q values dependent on frequency (Sato, 2006). We use the source model of Hisada (2008), which radiates P, SH, and SV waves independently. The source spectrum is Brune's ω2 model, and the radiation coefficient is isotropic (P wave: 0.52, S wave: 0.63). The horizontal one component of the three waveforms obtained is divided by √2 and three-types of waves are simply added. The phase spectrum is generated as a pulse wave with zero phase at all frequencies to enable waveform tracking. We compute strong motions using Wavenumber integration method up to 12.5 Hz and filtered through a trapezoidal bandpass filter with corner frequencies at 0.3, 0.4, 10, and 12.5 Hz.
Figure 2 shows velocity waveforms and Fourier amplitude spectra of Transverse, Radial, and UD at epicentral distance of 100 km (using f-dependent Q values). There are four types of waveforms: “Deep structure” is a model using all four layers in Table 1, “Without Vs4.5” is a model without the fourth layer, “Without Vs4.5&3.8” is a model without the fourth and third layers, and “Uniform structure” is a model using only direct waves in infinite homogeneous full-space of the second layer (the source layer). In the “Deep structure” model, repeated reflections appear from the mantle and lower crust, but in the “Without Vs4.5” model, the Moho reflections in the latter half of the waveform disappear, and in the “Without Vs4.5&3.8” model, there are no reflections from the lower crust, the waveform amplitude is small, and its duration is short. The waves in “Uniform structure” shows relatively large amplitude, but this is due to the fact that the transmission coefficient from the source layer to the first layer, which decreases with distance, is not taken into account (Hisada, 2024). In the Fourier amplitude spectrum, “without Vs4.5” has a slightly smaller amplitude than “Deep structure,” and “without Vs4.5&3.8” and “Uniform ground” have about 1/3 the amplitude.
Figure 3 shows the comparisons among the observed and calculated accelerations for YMTH13 (epicentral distance 30.8 km) and YMTH11 (epicentral distance 96.7 km). The structure model is a connection between the deep structure (Earthquake Headquarters 2012) and the surface structure at the KiK-net stations. Two types of Q values, constant Q value and f-dependent Q value, are used for the structure deeper than the seismic bedrock, and a simple constant Q value (Qs=Vs/15, Qp=1.7Qs Qp=1.7Qs) for the shallower structures. We assume the phase spectrum was to be random phase for all frequencies, and use the two types of time envelope functions: Boore's (1983) and the empirical formula that duration increases with Mj and epicenter distance (Sato et al., 1994). Figure 3 shows four waveforms, where “Observation B” is the wave observed by KiK-net, “Boore+Q0” is the Boore envelope function and constant Q value model, “Sato+Q0” is the Sato envelope function and constant Q value model, and “Sato+Q=91f^1.03” is the calculation result using Sato envelope function and f-dependent Q value. The start time of the waveforms are the origin time of the earthquake. The simulation results for the two stations show slightly later arrival times and larger amplitudes. At YMTH13, which is closer to the epicenter, Boore function is close to the envelope of the main motion, while at the distant YMTH11, Sato function reproduces the long duration of the observed records well. A more detailed interpretation on the effects of reflected waves from deep structure and the surface ground, including other stations, will be given on the day of the presentation.