| 基于支持向量机的2021年2月13日日本福岛近海Mj7.3级地震震级估算 |
| |
| 引用本文： | 朱景宝,宋晋东,李山有.基于支持向量机的2021年2月13日日本福岛近海Mj7.3级地震震级估算[J].世界地震工程,2021,0(2):074-81. |
| |
| 作者姓名： | 朱景宝 宋晋东 李山有 |
| |
| 作者单位： | 中国地震局工程力学研究所, 中国地震局地震工程与工程振动重点实验室, 黑龙江 哈尔滨 150080 |
| |
| 摘 要： | 2021年2月13日日本福岛县近海发生Mj7.3级地震，触发了日本气象厅地震预警系统，系统在首台触发后5.6s发出震级为Mj6.3级的预警第1报，首台触发后10s对公众发布警报、预警震级为Mj6.4级。基于多类型特征参数输入的机器学习支持向量机震级估算模型（SVM-M），利用2021年2月13日日本福岛县近海Mj7.3级地震获取的日本K-net强震动观测数据，分析SVM-M模型在该次地震中首台触发初期（首台触发后1~10s）的震级估算效能。结果表明:SVM-M震级估算模型，在首台触发后1s即可给出Mj6.3级的震级估算结果，与日本气象厅在首台触发后5.6s发布的预警第1报震级相同；随着时间窗的增加，首台触发后5s和10s，SVM-M模型的震级估算结果分别是Mj6.7级和Mj6.6级，均大于日本气象厅首台触发后10s对公众发布警报的预警震级。该次地震的离线模拟结果表明:SVM-M模型可在地震发生初期有效提高地震预警震级确定的准确性和时效性。
|
| 关 键 词： | 地震预警 支持向量机 福岛县近海地震 震级估算 机器学习 |
Magnitude estimation for the February 13, 2021 Mj7.3 earthquake near the coast of Fukushima Japan based on support vector machine |
| |
| ZHU Jingbao,SONG Jindong,LI Shanyou.Magnitude estimation for the February 13, 2021 Mj7.3 earthquake near the coast of Fukushima Japan based on support vector machine[J].World Information On Earthquake Engineering,2021,0(2):074-81. |
| |
| Authors: | ZHU Jingbao SONG Jindong LI Shanyou |
| |
| Institution: | Institute of Engineering Mechanics, China Earthquake Administration; Key Laboratory of Earthquake Engineering and Engineering Vibration of China Earthquake Administration, Harbin 150080,China |
| |
| Abstract: | The earthquake early warning (EEW) system of the Japan Meteorological Agency was triggered by the Mj7.3 earthquake off the coast of Fukushima, Japan on February 13, 2021. The system issued first alarm with a magnitude of Mj6.3 at 5.6 seconds after the first station was triggered, and a magnitude of Mj6.4 was issued to the public at 10 seconds after the first station was triggered. In this paper, the support vector machine magnitude estimation model (SVM-M) based on several types of characteristic parameters inputs via machine learning is used to analyze the magnitude estimation performance at the initial trigger stage of the first station (1-10 seconds after the first station is triggered) in this earthquake by using the Japanese K-net strong motion observation data obtained from the Mj7.3 earthquake off the coast of Fukushima, Japan on February 13, 2021.The results show that the SVM-M model can obtain the magnitude of Mj6.3 at 1 second after the first station is triggered, which is the same as the magnitude of the first alarm issued by the Japan Meteorological Agency at 5.6 seconds after the first station is triggered. With the increase of the time window, 5 seconds and 10 seconds after the first station was triggered, the estimated magnitude of the SVM-M model was Mj6.7 and Mj6.6 respectively, both larger than the magnitude of the alarm issued to the public at 10 seconds after the first station was triggered by the Japan Meteorological Agency. The off-line simulation results of this earthquake show that the SVM-M model can effectively improve the accuracy and timeliness of EEW magnitude determination in the early stage of earthquake occurrence. |
| |
| Keywords: | earthquake early warning support vector machine earthquake near the coast of Fukushima magnitude estimation machine learning |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《世界地震工程》浏览原始摘要信息 |
| 点击此处可从《世界地震工程》下载免费的PDF全文 |
