On magnitude dependence in earthquake clustering

庄建仓 教授

日本统计数理研究所

2026.05.08（星期五）16:00，理科二号楼2821

摘要：

      A central open question in statistical seismology is whether earthquake magnitudes exhibit dependence within triggering sequences. In standard clustering models such as the Epidemic Type Aftershock Sequence (ETAS) model, magnitudes are assumed to be independent and identically distributed, implying no predictability beyond the Gutenberg - Richter law. However, a number of recent studies have reported apparent correlations between successive magnitudes, suggesting the possibility of additional predictability and underlying physical memory effects. Using both synthetic catalogs and high-resolution real seismic data, we find that apparent magnitude correlations can arise out much of the observed effect can be explained by data incompleteness and missing small events rather than clear evidence of intrinsic physical dependence. When these effects are taken into account, the additional predictive value of past magnitudes appears to be limited, even when evaluated with flexible learning models.

报告人简介：

庄建仓，统计学博士，地球物理学硕士，现任日本情报与系统研究机构统计数理研究所(The Institute of Statistical Mathematics, ISM)教授。他是该国立研究所自1944年创立以来首位获聘为副教授及正教授的外国籍研究人员。同时，他还兼任英国伦敦数学实验室中国地震局地球物理研究所、中国地震预测研究所、中国地震局地震实验场、南方科技大学风险风险分析预测与管控研究院等多所大学或者国立研究所的客座教授/客座研究员。研究领域涵盖统计地震学、时空点过程建楼地重酒训统计据科学、机器理贝斯推地震预测、统计数据科学、机器学习与贝叶斯推断领域的国际学术期刊上发表论文160篇以上，其中包括Journal of American Statistical Association》《Journal the Royal Statistical Society B》《Journal of Geophysical Research》《Geophysical Researcher Letters》《Nature Communications》《Journal of Geodesy》《Physical Review Letters》等顶尖期刊,并担任多个国际学术期刊的编委。他也是日本统计学会、国际数理统计学会 (IMS)、日本地震学会、美国地球物理联合会 (AGU) 等学术组织的活跃成员。