5:15 PM - 6:45 PM
[SSS10-P13] Estimation of heterogeneity parameters of subsurface soil using seismic observation records between adjacent sites
Keywords:subsurface soil, heterogeneity, spatial variation, correlation length
Heterogeneities in the subsurface soil cause spatial variations in ground motions between adjacent points on the surface and affect the response of buildings. This study investigates the heterogeneity parameters of S-wave velocity in the subsurface soil by using the ground motion records obtained in the Chiba seismometer array of the Institute of Industrial Science, University of Tokyo.
Table 1 shows the S-wave velocity structure and coefficient of variation of the subsurface soil in the Chiba seismometer array. The coefficient of variation was estimated from the fluctuation of N-value (Katayama et al., 1990) and the relation between N-value and S-wave velocity (Ota et al., 1975). We used observation records of seismic events with magnitude (MJ) 5.0 or more, epicenter depth of 80 km or more, apparent incident angle of 40 degrees or less, and maximum acceleration of 100 cm/s2 to estimate the heterogeneity parameters.
Following Ritter et al.(1998), we defined the average of the waveforms of observation records between two adjacent stations as the coherent component u, the residual waveform subtracting u from the original records as the fluctuation component uf and the spatial variation v2 as the ratio of the power spectrum between uf and u. ln(v2+1) of each separation distance between observation stations is plotted as solid lines in Fig. 1. The higher the frequency, the larger ln(v2+1). Furthermore, ln(v2+1) decreases as the distance between observation stations decreases. With reference to Ritter et al.(1998), Tokumitsu et al.(2022) present the relation between the heterogeneity pattern of the subsurface soil and the spatial variation of ground motion between adjacent points, assuming the anisotropic heterogeneity pattern to be Gaussian. In this study, the horizontal and vertical correlation lengths of S-wave velocity a1, a3 are determined such that the sum of squares of the residuals between ln(v2+1) estimated from the relation (Tokumitsu et al., 2022) and ln(v2+1) of the observation records shown in Fig. 1 are minimized. Table 1 shows the estimates for a1 and a3. a1 is estimated to be larger than a3 in all layers. This trend is consistent with previous work indicating that the correlation length in the vertical direction is greater than that in the horizontal direction.
To verify the estimation results of a1 and a3, we conducted numerical simulations using 3D FEM heterogeneous soil models. Fig. 2 shows an image of the models and heterogeneity patterns constructed from the estimates of a1 and a3. An impulsive plane wave enters at the bottom of the model, and ln(v2+1) between adjacent points was calculated from the response waves obtained at every 1m on 100m lines at the center of the surface shown in Fig.2. Simulations are performed five times with different heterogeneous patterns. ln(v2+1) from the simulations are shown as dotted lines in Fig. 1. Although ln(v2+1) are estimated to be smaller than the observation records (solid lines), they show the same trend that ln(v2+1) increases with increasing separation distance as the observation records.
Acknowledges
We used observation records in the Chiba seismic array published by the Assoc. for Earthquake Disaster Prevention. GMT is used to draw some of the figures.
Table 1 shows the S-wave velocity structure and coefficient of variation of the subsurface soil in the Chiba seismometer array. The coefficient of variation was estimated from the fluctuation of N-value (Katayama et al., 1990) and the relation between N-value and S-wave velocity (Ota et al., 1975). We used observation records of seismic events with magnitude (MJ) 5.0 or more, epicenter depth of 80 km or more, apparent incident angle of 40 degrees or less, and maximum acceleration of 100 cm/s2 to estimate the heterogeneity parameters.
Following Ritter et al.(1998), we defined the average of the waveforms of observation records between two adjacent stations as the coherent component u, the residual waveform subtracting u from the original records as the fluctuation component uf and the spatial variation v2 as the ratio of the power spectrum between uf and u. ln(v2+1) of each separation distance between observation stations is plotted as solid lines in Fig. 1. The higher the frequency, the larger ln(v2+1). Furthermore, ln(v2+1) decreases as the distance between observation stations decreases. With reference to Ritter et al.(1998), Tokumitsu et al.(2022) present the relation between the heterogeneity pattern of the subsurface soil and the spatial variation of ground motion between adjacent points, assuming the anisotropic heterogeneity pattern to be Gaussian. In this study, the horizontal and vertical correlation lengths of S-wave velocity a1, a3 are determined such that the sum of squares of the residuals between ln(v2+1) estimated from the relation (Tokumitsu et al., 2022) and ln(v2+1) of the observation records shown in Fig. 1 are minimized. Table 1 shows the estimates for a1 and a3. a1 is estimated to be larger than a3 in all layers. This trend is consistent with previous work indicating that the correlation length in the vertical direction is greater than that in the horizontal direction.
To verify the estimation results of a1 and a3, we conducted numerical simulations using 3D FEM heterogeneous soil models. Fig. 2 shows an image of the models and heterogeneity patterns constructed from the estimates of a1 and a3. An impulsive plane wave enters at the bottom of the model, and ln(v2+1) between adjacent points was calculated from the response waves obtained at every 1m on 100m lines at the center of the surface shown in Fig.2. Simulations are performed five times with different heterogeneous patterns. ln(v2+1) from the simulations are shown as dotted lines in Fig. 1. Although ln(v2+1) are estimated to be smaller than the observation records (solid lines), they show the same trend that ln(v2+1) increases with increasing separation distance as the observation records.
Acknowledges
We used observation records in the Chiba seismic array published by the Assoc. for Earthquake Disaster Prevention. GMT is used to draw some of the figures.