17:15 〜 18:45
[SSS10-P20] A study on the source and site response effects on the acceleration response spectra of strong motions recorded at S-net sites by comparing with a land-based ground-motion model
キーワード：海底地震計、S-net、サイト応答特性、震源特性、地震動予測式、特異値分解
S-net is a seafloor observation network for earthquakes and tsunamis installed along the Japan Trench after 2011 Tohoku earthquake (Aoi et al., 2020). The analysis of strong motions recorded on the seafloor near plate boundaries may provide an insight into earthquakes that cannot be obtained through land-based observations. In this study, we attempt to examine whether there are the source and site response effects on the acceleration response spectra unique to S-net observation records by comparing with an existing ground-motion model (GMM) based on land observation records.
We use the records from earthquakes to generally meet the following criteria: (a) Mw > 5.4; (b) epicenter in the sea; (c) the closest distance from hypocenter less than 1,100 km at all observation stations; (d) the closest distance to the nearest station less than 100 km. The data set of hypocenter and CMT solution is taken from the seismological bulletin of Japan compiled by JMA.
The tilt and rotation angles of 3-component acceleration seismograms are corrected following Takagi et al. (2019).
The observation seismographs are buried approximately 1 m bellow the seafloor at a water depth shallower than 1.5 km, and are not buried at deeper depth. Since 2 horizontal components of records observed at non-buried stations show different amplitude spectra due to the natural vibration of cylindrical pressure vessels (Sawazaki and Nakamura, 2020), we focus only on the direction parallel to the cable’s long axis, where the natural vibration of a cylindrical pressure vessel is shown on the higher frequency side in this study.
Referring to Annaka (2002), we estimate the source and site response effects by calculating correction coefficients for an existing GMM in comparison to the observation record as follows:
αi(T) + βj(T) = log{Oij(T) / Cij(T)},
where T is period. Oij and Cij are observed and calculated 5% damped acceleration response spectra for earthquake i and station j, respectively. αi and βj are source and site correction coefficients, respectively. The above regression equation is solved by the least squares method using singular value decomposition with the constraint Σαi(T) = 0. The GMM proposed by Annaka and Nozawa (1988) is used. Since this GMM does not consider tectonic and fault types of the earthquake, it makes clearer to show the difference in source effect between the types. Because of the GMM for the relatively stiff ground, the estimated site response effects include the amplification factor of surface soft sediments.
The relationship between installation water depth and βj indicates that βj at non-buried stations shows larger trend and variability than that at buried stations regardless of the period. Note that this result includes the effect of the natural vibration of a cylindrical pressure vessel.
The tectonic and fault types of each earthquake were determined based on the CMT solution. It is indicated that some periodic characteristics of the average αi for each tectonic type are similar to the findings obtained from the land-based observation records as follows: e.g, αi’s on short-period side of intraslab earthquakes are larger than 1. In addition, the relationship between αi and earthquake source depth indicates followings for intraslab normal fault earthquakes that occur beneath the outer rise regardless of the period: (a) negative correlation was shown in the case of using hypocenter depth listed in the JMA catalog; (b) strong positive correlation was observed in the case of using centroid depth obtained from the CMT solution. It can be pointed out that the land-based GMM may not be able to adequately represent the effect of source depth of outer rise earthquakes far away from the land area. Since the GMM excluding earthquakes occurring far from land observation stations may not be able to adequately estimate the ground motions induced by the earthquakes near ocean trenches, we will further study the S-net observation records to solve this issue.
We use the records from earthquakes to generally meet the following criteria: (a) Mw > 5.4; (b) epicenter in the sea; (c) the closest distance from hypocenter less than 1,100 km at all observation stations; (d) the closest distance to the nearest station less than 100 km. The data set of hypocenter and CMT solution is taken from the seismological bulletin of Japan compiled by JMA.
The tilt and rotation angles of 3-component acceleration seismograms are corrected following Takagi et al. (2019).
The observation seismographs are buried approximately 1 m bellow the seafloor at a water depth shallower than 1.5 km, and are not buried at deeper depth. Since 2 horizontal components of records observed at non-buried stations show different amplitude spectra due to the natural vibration of cylindrical pressure vessels (Sawazaki and Nakamura, 2020), we focus only on the direction parallel to the cable’s long axis, where the natural vibration of a cylindrical pressure vessel is shown on the higher frequency side in this study.
Referring to Annaka (2002), we estimate the source and site response effects by calculating correction coefficients for an existing GMM in comparison to the observation record as follows:
αi(T) + βj(T) = log{Oij(T) / Cij(T)},
where T is period. Oij and Cij are observed and calculated 5% damped acceleration response spectra for earthquake i and station j, respectively. αi and βj are source and site correction coefficients, respectively. The above regression equation is solved by the least squares method using singular value decomposition with the constraint Σαi(T) = 0. The GMM proposed by Annaka and Nozawa (1988) is used. Since this GMM does not consider tectonic and fault types of the earthquake, it makes clearer to show the difference in source effect between the types. Because of the GMM for the relatively stiff ground, the estimated site response effects include the amplification factor of surface soft sediments.
The relationship between installation water depth and βj indicates that βj at non-buried stations shows larger trend and variability than that at buried stations regardless of the period. Note that this result includes the effect of the natural vibration of a cylindrical pressure vessel.
The tectonic and fault types of each earthquake were determined based on the CMT solution. It is indicated that some periodic characteristics of the average αi for each tectonic type are similar to the findings obtained from the land-based observation records as follows: e.g, αi’s on short-period side of intraslab earthquakes are larger than 1. In addition, the relationship between αi and earthquake source depth indicates followings for intraslab normal fault earthquakes that occur beneath the outer rise regardless of the period: (a) negative correlation was shown in the case of using hypocenter depth listed in the JMA catalog; (b) strong positive correlation was observed in the case of using centroid depth obtained from the CMT solution. It can be pointed out that the land-based GMM may not be able to adequately represent the effect of source depth of outer rise earthquakes far away from the land area. Since the GMM excluding earthquakes occurring far from land observation stations may not be able to adequately estimate the ground motions induced by the earthquakes near ocean trenches, we will further study the S-net observation records to solve this issue.