11:45 AM - 12:00 PM
[SSS12-05] Site amplification characteristics of S-net sites and differences from K-NET/KiK-net sites based on the comparison of the site terms of GMPE
Keywords:Seafloor ground motion, S-net, Peak Ground Velocity, Site amplification, Vertical to horizontal PGV ratio
For several years, seafloor observation networks (e.g. S-net) have provided strong motion records near sources of earthquakes that occurred in subduction zones.
Some previous studies have shown some specific characteristics on seafloor records, for example, high horizontal amplifications in long period and low vertical amplifications in short period (Boore and Smith, 1999, Dhakal et al. al., 2022, 2023) compared to those on the land. It is important to understand characteristics of ground motions on the seafloor in advance, to improve an accuracy of a prediction near source regions of subduction earthquakes.
In this study, we developed GMPEs for PGV of horizontal and vertical motions using the records of S-net, K-NET and KiK-net. We discussed differences in site terms between land and seafloor sites and between buried and non-buried sites of S-net, and correlation with a thickness of sedimentary layers or a water depth in terms of horizontal and vertical motions.
To construct a GMPE data set, we picked 60 earthquakes with MW (F-net) >= 5.5 that occurred within the interplate/intraslab after an S-net installation. We set the hypocentral distance range within 300 km for MW >= 7.0 and within 200 km for MW < 7.0.
We collected S-net records for the tilt and rotation angles based on Takagi et al. (2019) and excluded the records which had been rotated or tilted more than 1 degree during an earthquake. We also excluded an earthquake with less than 3 records, and a station with less than 3 earthquakes. We used the bandpass filter of 0.2-10Hz before an integration. To avoid the effect of instrumental characteristics (hereafter referred to as N-shape (Sawazaki and Nakamura., 2020)), we treated the horizontal PGV of S-net independently for X and Y components. For the K-NET and KiK-net records, we used the geometric mean of the two horizontal components.
We set the GMPE function form as follows with reference to Si and Midorikawa (1999).
logPGVi,j=bi-log(Xsh+0.0075*100.5Mw)-kst.Xsh+Gj
A kst. is an inelastic attenuation coefficient, and is divided into two cases: the S-net or the K-NET and KiK-net. As a constraint condition for the horizontal PGV at a reference, KiK-net Kahoku, we fixed the empirical amplification factor proposed by Yamaguchi and Midorikawa (2011). For the vertical PGV, we used the horizontal reference value multiplied by the V/H ratio of the observed PGV at the reference site.
The source term and attenuation term were not significantly different from Si and Midorikawa (1999). Hereafter, we focus on the values of the site terms.
Site terms on the backarc tended to be smaller than those on the forearc (Figure. (a)). This might be because the site term was reflected by the amplitude attenuated through the low Q-zone just below the volcanic area. The site term for the horizontal motion was correlated with the depth of the layer top of the Vs >= 1400 m/s (hereafter referred to as D1400) in the J-SHIS model. Our results in the horizontal site term qualitatively correspond to the empirical relation between the site term and D1400 in Morikawa and Fujiwara (2013) (Figure (a)). The vertical site terms were also smaller on the backarc sites than on the forearc sites and correlated with D1400 (Figure (b)).
In the S-net records, the N-shape effect is not seen in the X components lower than 10 Hz (Dhakal et al., 2023), so we focus only on the X components. The horizontal site term was correlated with D1400 (Figure. (a)), and the vertical site terms were weakly correlated with D1400 (Figure. (b)). For both horizontal and vertical components, site terms on the non-buried sites were significantly larger than those on the buried sites with similar D1400 of them (Figure. (a), (b)). Both horizontal and vertical site terms were correlated with water depths (Figure. (c)). The site terms of the non-buried sites were larger than on the buried sites, because most non-buried sites have deeper water depths than buried sites.
The vertical site terms were smaller than the horizontal site terms on both the land and the seafloor (Figure. (a)), and the vertical ones to horizontal ones seemed to be different from on the sea or on the land. Comparing the histograms of the vertical to horizontal ratios of the records, the V/H ratio on the seafloor was clearly smaller than that on the land (Figure. (d)). We consider that the low vertical amplitude on the seafloor could be affected by the interference with P waves propagating in water layer (Boore and Smith, 1999). We are currently studying the effect of the water layer on the seismic motion of the seafloor.
Acknowledgments
We used the strong motion data from K-NET, KiK-net, S-net, the mechanism data from F-net, and the deep subsurface structure model from J-SHIS(NIED).
Some previous studies have shown some specific characteristics on seafloor records, for example, high horizontal amplifications in long period and low vertical amplifications in short period (Boore and Smith, 1999, Dhakal et al. al., 2022, 2023) compared to those on the land. It is important to understand characteristics of ground motions on the seafloor in advance, to improve an accuracy of a prediction near source regions of subduction earthquakes.
In this study, we developed GMPEs for PGV of horizontal and vertical motions using the records of S-net, K-NET and KiK-net. We discussed differences in site terms between land and seafloor sites and between buried and non-buried sites of S-net, and correlation with a thickness of sedimentary layers or a water depth in terms of horizontal and vertical motions.
To construct a GMPE data set, we picked 60 earthquakes with MW (F-net) >= 5.5 that occurred within the interplate/intraslab after an S-net installation. We set the hypocentral distance range within 300 km for MW >= 7.0 and within 200 km for MW < 7.0.
We collected S-net records for the tilt and rotation angles based on Takagi et al. (2019) and excluded the records which had been rotated or tilted more than 1 degree during an earthquake. We also excluded an earthquake with less than 3 records, and a station with less than 3 earthquakes. We used the bandpass filter of 0.2-10Hz before an integration. To avoid the effect of instrumental characteristics (hereafter referred to as N-shape (Sawazaki and Nakamura., 2020)), we treated the horizontal PGV of S-net independently for X and Y components. For the K-NET and KiK-net records, we used the geometric mean of the two horizontal components.
We set the GMPE function form as follows with reference to Si and Midorikawa (1999).
logPGVi,j=bi-log(Xsh+0.0075*100.5Mw)-kst.Xsh+Gj
A kst. is an inelastic attenuation coefficient, and is divided into two cases: the S-net or the K-NET and KiK-net. As a constraint condition for the horizontal PGV at a reference, KiK-net Kahoku, we fixed the empirical amplification factor proposed by Yamaguchi and Midorikawa (2011). For the vertical PGV, we used the horizontal reference value multiplied by the V/H ratio of the observed PGV at the reference site.
The source term and attenuation term were not significantly different from Si and Midorikawa (1999). Hereafter, we focus on the values of the site terms.
Site terms on the backarc tended to be smaller than those on the forearc (Figure. (a)). This might be because the site term was reflected by the amplitude attenuated through the low Q-zone just below the volcanic area. The site term for the horizontal motion was correlated with the depth of the layer top of the Vs >= 1400 m/s (hereafter referred to as D1400) in the J-SHIS model. Our results in the horizontal site term qualitatively correspond to the empirical relation between the site term and D1400 in Morikawa and Fujiwara (2013) (Figure (a)). The vertical site terms were also smaller on the backarc sites than on the forearc sites and correlated with D1400 (Figure (b)).
In the S-net records, the N-shape effect is not seen in the X components lower than 10 Hz (Dhakal et al., 2023), so we focus only on the X components. The horizontal site term was correlated with D1400 (Figure. (a)), and the vertical site terms were weakly correlated with D1400 (Figure. (b)). For both horizontal and vertical components, site terms on the non-buried sites were significantly larger than those on the buried sites with similar D1400 of them (Figure. (a), (b)). Both horizontal and vertical site terms were correlated with water depths (Figure. (c)). The site terms of the non-buried sites were larger than on the buried sites, because most non-buried sites have deeper water depths than buried sites.
The vertical site terms were smaller than the horizontal site terms on both the land and the seafloor (Figure. (a)), and the vertical ones to horizontal ones seemed to be different from on the sea or on the land. Comparing the histograms of the vertical to horizontal ratios of the records, the V/H ratio on the seafloor was clearly smaller than that on the land (Figure. (d)). We consider that the low vertical amplitude on the seafloor could be affected by the interference with P waves propagating in water layer (Boore and Smith, 1999). We are currently studying the effect of the water layer on the seismic motion of the seafloor.
Acknowledgments
We used the strong motion data from K-NET, KiK-net, S-net, the mechanism data from F-net, and the deep subsurface structure model from J-SHIS(NIED).