人工地震目录的评估及其在青藏高原东北缘的应用

如何使用古地震序列数据评估数值模拟得到的长期人工地震目录的准确度是地震数值预报研究所面临的重要问题.为探讨上述问题，本文开发了计算人工地震目录与古地震序列匹配度的平均绝对误差法和余弦相似度法，并将其应用于青藏高原东北缘地区，获得了与该地区古地震序列数据匹配较好的人工地震目录.基于匹配较好的目录，本文还计算了海原断裂及香山天景山断裂发生大地震后，大地震在区域四条主要断裂的迁移概率.研究结果表明：（1）以上两种方法均可用于评估人工地震目录与古地震序列的匹配度；（2）与未进行匹配的人工地震目录的迁移概率计算结果比较，通过古地震数据匹配的人工地震目录的计算结果更优，即更接近于基于古地震数据计算的地震迁移概率；（3）当大地震在海原断裂上发生后，海原断裂再次发生大地震的概率最大，约为47%，其次是香山天景山断裂，约为23%~27%.本文对评估基于动力学数值模型产生的长期人工合成地震目录的准确度做出了调查与探索，为区域地震数值预报研究提供了参考.

The evaluation of synthetic seismic catalog and its application in northeastern Tibetan Plateau

It is an important problem in numerical earthquake prediction how the paleoseismic data are used to verify the accuracy of long-term synthetic seismic catalogs obtained from numerical modeling. To investigate this problem, we develop the mean absolute error method and the cosine similarity method to calculate the matching degree between the synthetic seismic catalogs and the paleoseismic sequence data. We apply these methods to the northeastern Tibetan Plateau, and obtain the synthetic seismic catalogs that match well with the paleoseismic sequence data. From the well-matched catalogs, we calculate the probability of earthquake migration on the four major faults after one big earthquake occurring on the Haiyuan fault or Xiangshan-Tianjingshan fault. The results in this study show that these two methods can be applied to evaluate the matching degree between the synthetic seismic catalogs and the paleoseismic sequence data. The results of probability of earthquake migration from the well-matched catalogs are better than those from the raw catalogs, or they are closer to those from the paleoseismic data. Our calculation shows that when a big earthquake occurs on the Haiyuan fault, the probability of the next big earthquake on the Haiyuan fault is the highest, about 47%, and that on the Xiangshan-Tianjingshan fault is the second highest, about 23%~27%. This paper makes an investigation and exploration for evaluating the accuracy of the long-term synthetic seismic catalogs calculated by the geodynamic numerical modeling and can provide insights into regional numerical earthquake prediction.