[S-TT43] Seismic Big Data Analysis Based on the State-of-the-Art of Bayesian Statistics

[STT43-02] Extraction of Spatial Features of Seismic Intensity Based on Field Reconstruction: Towards Constraining the Source Locations of Historical Earthquakes

*Motoko Ishise¹, Takahiro Shiina², Ryoichi Nakamura³, Shigeki Nakagawa⁴ (1.Yamagata University, 2.National Institute of Advanced Industrial Science and Technology, 3.Naka Earthquake Research, 4.Earthquake Research Institute, the University of Tokyo)

Keywords:Distribution of seismic intensity, seismic intensity field reconstruction, hypocenter location estimation, historical earthquake, basis function expansion, Kanto region

The spatial distribution of seismic intensity varies depending on the earthquake’s location, radiation pattern, and underground structure. Typically, seismic intensity is expected to display a concentric distribution, with the strongest intensity above the epicenter. However, non-concentric seismic intensity has often been observed for earthquakes beneath the Kanto region, Japan (Nakanishi & Horie, 1980 JPE; Nakamura et al., 2007 Historical Earthquakes). Recent research indicates that such non-concentric seismic intensity is primarily caused by three-dimensional heterogeneity in seismic attenuation structure (Nakamura et al., 2023 Zisin). This implies that the spatial pattern of the seismic intensity provides valuable information for investigating earthquake locations in the Kanto region.
Based on the above knowledge, we have initiated the development of a method to identify the source locations of historical earthquakes by comparing the seismic intensity recorded in historical documents with that of present-day earthquakes. The only available information on historical earthquakes is their seismic intensity, which is often estimated from the building collapse rate and degree of building damage. In our first report (Ishise et al., 2023 SSJ), we compiled the spatial pattern of seismic intensity for the present-day earthquakes in the Kanto region and demonstrated that these patterns vary depending on the source locations. In the second report (Ishise et al., 2024, JpGU), we introduced a method for evaluating the spatial characteristics and similarities of seismic intensity distributions. This approach is based on the reconstruction of a continuous seismic intensity field using basis functions. The proposed method assesses the similarity of seismic intensity distributions by utilizing the coefficients of the basis functions that present the reconstructed seismic intensity field. By incorporating the field reconstruction technique, the method enables a unified evaluation of spatial similarities in seismic intensity distributions, even when the arrangement of the observation points changes over time over extended periods.
In this study, we refined the methodology proposed by Ishise et al. (2024 JpGU) by incorporating UMAP (McInnes et al., 2018), which is a dimensionality reduction technique, for clustering the characteristics of seismic intensity distribution. The refines method was applied the seismic intensity onbervations of 900 earthquakes that occurred in the Kanto region.
As a result, the seismic intensity distribution data were classified into approximately seven clusters, including earthquakes occurring in southern Ibaraki prefecture and northern Chiba prefecture. Additionally, earthquakes occurred in the Pacific coastal area were categorized into three distinct groups: off Ibaraki prefecture, off Choshi, and off Boso Peninsula. These findings suggest that the spatial characteristics of seismic intensity distributions can potentially constrain earthquake source locations without relying on subsurface structural information.