2026 年 91 巻 843 号 p. 595-603
This study proposes a method to simplify spatial correlation of ground-motion uncertainty in probabilistic seismic risk assessment for building portfolios. Portfolios were clustered using k-means, assuming perfect correlation within clusters and no correlation between clusters, allowing approximate representation without generating correlated normal random numbers. Seismic risk assessments for multiple portfolios in the Kanto region were compared with reference cases incorporating spatial correlation based on inter-building distances. The method produced comparable estimates, particularly when average intra-cluster distances were around 20 km. Combined with portfolio aggregation, it reduced computational cost while maintaining sufficient accuracy, indicating potential for application to larger portfolios.