3:00 PM - 3:15 PM
[SSS10-12] Study on the Performance Evaluation Method of Strong Motion Observation Record Database
Keywords:Seismic Hazard Assessment, Strong Motion Database, Ground Motion Prediction Equation, Performance Evaluation
A database is being developed to serve as a basis for the construction of a Ground-Motion Model (GMM) for seismic hazard assessment (Fujiwara et al., 2021). This database consists of a database of observation records (Morikawa et al., 2021) and a Physics-Based Simulation (PBS) database (Iwaki et al., 2021), which consists of seismic ground motion simulations based on geophysical theoretical models using appropriately adjusted subsurface models. In order to determine the effectiveness of the PBS database, it is necessary to quantify the performance of the PBS database. In this study, we used the PGA of inland crustal earthquakes in the prototype version of the strong motion database (Morikawa et al., 2020) to evaluate the performance of the database of observation records.
As in Morikawa et al. (2021), we discretized the space of earthquake magnitude-to-fault shortest distance in the database into small regions, and used the estimation accuracy of the probability distribution of the seismic intensity of the observation records in each small region as an index to evaluate the performance of the database. To evaluate the probability distribution of the seismic intensity in each subregion, we used the method of Morikawa et al. (2008) as a reference to correct for the difference in the shaking intensity of each observation point. In this study, we modified the method so that we can evaluate the site coefficients without assuming the functional form of the attenuation characteristics. In this study, however, we modified the method to evaluate the site coefficient without assuming the functional form of the attenuation characteristics. We evaluated the empirical attenuation characteristics without assuming the functional form of the distance attenuation characteristics by calculating the median of the seismic intensity for each distance range according to the earthquake scale, and evaluated the site characteristics of each observation point considering the within-event residuals. The probability distributions of seismic intensity for small areas with few observation records were estimated by including data from surrounding small areas. Differences in earthquake magnitudes and distances were corrected using the empirical attenuation characteristics described above, and site coefficients and within-event residuals were also corrected.
We evaluated the performance of the database of observation records by using the Wasserstein distance to evaluate the difference between the probability distribution included in the small region and the probability distribution with the above correction including data around the small region. In the region where the Wasserstein distance is large, it is difficult to estimate the probability distribution with the current database of observation records. This suggests that the addition of the PBS database would be effective. In addition, it is possible to examine the effectiveness of the addition of the PBS database by confirming the increase or decrease of the Wasserstein distance with the addition of the PBS database. In the future, we will consider the improvement of the performance evaluation method of the observation record database by non-ergodic analysis taking into account the heterogeneity of the attenuation characteristics.
Acknowledgments: This study is supported by the Gran-in-Aid for Scientific Research (KAKENHI Number
20H00292).
As in Morikawa et al. (2021), we discretized the space of earthquake magnitude-to-fault shortest distance in the database into small regions, and used the estimation accuracy of the probability distribution of the seismic intensity of the observation records in each small region as an index to evaluate the performance of the database. To evaluate the probability distribution of the seismic intensity in each subregion, we used the method of Morikawa et al. (2008) as a reference to correct for the difference in the shaking intensity of each observation point. In this study, we modified the method so that we can evaluate the site coefficients without assuming the functional form of the attenuation characteristics. In this study, however, we modified the method to evaluate the site coefficient without assuming the functional form of the attenuation characteristics. We evaluated the empirical attenuation characteristics without assuming the functional form of the distance attenuation characteristics by calculating the median of the seismic intensity for each distance range according to the earthquake scale, and evaluated the site characteristics of each observation point considering the within-event residuals. The probability distributions of seismic intensity for small areas with few observation records were estimated by including data from surrounding small areas. Differences in earthquake magnitudes and distances were corrected using the empirical attenuation characteristics described above, and site coefficients and within-event residuals were also corrected.
We evaluated the performance of the database of observation records by using the Wasserstein distance to evaluate the difference between the probability distribution included in the small region and the probability distribution with the above correction including data around the small region. In the region where the Wasserstein distance is large, it is difficult to estimate the probability distribution with the current database of observation records. This suggests that the addition of the PBS database would be effective. In addition, it is possible to examine the effectiveness of the addition of the PBS database by confirming the increase or decrease of the Wasserstein distance with the addition of the PBS database. In the future, we will consider the improvement of the performance evaluation method of the observation record database by non-ergodic analysis taking into account the heterogeneity of the attenuation characteristics.
Acknowledgments: This study is supported by the Gran-in-Aid for Scientific Research (KAKENHI Number
20H00292).