2021年5月21日云南漾濞地震及5月22日青海玛多地震的支持向量机震级估算

2021年5月21日及5月22日，云南漾濞县与青海玛多县分别发生破坏性地震，主震震级分别为M_s6.4级与M_s7.4级。本文基于机器学习中的支持向量机方法，以多类型特征参数为输入建立地震预警震级估算模型SVM-M，离线模拟云南漾濞M_s5.6级前震、M_s6.4级主震以及青海玛多M_s7.4级主震的连续震级估算。结果表明:对于云南漾濞M_s5.6级前震，支持向量机方法在首台触发后1s可估算震级为5.6级，且随着首台触发时间的增加，估算震级一直在实际震级附近波动；对于云南漾濞M_s6.4级主震和青海玛多M_s7.4级主震，随着首台触发时间的增加，支持向量机方法对于大震低估问题得到了有效的改善，且震级估算结果逐渐接近实际震级。同时，这3次地震的震级估算离线模拟表明:引入震源距的支持向量机方法（SVM-M1模型）对于震级估算有更好的稳定性，且在地震预警系统的震级估算中有着潜在的应用前景。

Magnitude estimation via support vector machine for the earthquakes occur in Yangbi,Yunnan on May 21, 2021 and in Maduo,Qinghai on May 22, 2021

On May 21 and May 22, 2021, destructive earthquakes occurred in Yangbi County, Yunnan Province and Maduo County, Qinghai, respectively. The magnitude of the main shock for the earthquakes are M_s6.4 and M_s7.4, respectively. In this paper, based on the support vector machine (SVM) method in machine learning, the SVM-M model of earthquake early warning magnitude estimation is established with multiple types of characteristic parameters as input. The continuous magnitude estimation of M_s5.6 foreshock, M_s6.4 main earthquake occurred in Yangbi County, Yunnan and M_s7.4 main earthquake occurred in Maduo County, Qinghai are simulated off-line, and the estimated magnitude fluctuates near the catalog magnitude with the increase of the trigger time of the first station. For the M_s6.4 main earthquake occurred in Yangbi County, Yunnan Province and the M_s7.4 main earthquake occurred in Maduo County, Qinghai Province, with the increase of the trigger time of the first station, the SVM method has effectively improved the problem of underestimation of large earthquakes, and the magnitude estimation is gradually close to the catalog magnitude. Meanwhile, the off-line simulation of these three earthquakes shows that the SVM method (SVM-M1 model) with source distance has better stability for magnitude estimation, and has a potential application prospect in the magnitude estimation of earthquake early warning system.