基于多输入高斯过程回归的震级快速估算方法  

Rapid magnitude estimation based on multi-input Gaussian process regression

作　　者：赵庆旭 王延伟 莫红艳[1] 曹振中[1] Zhao Qingxu;Wang Yanwei;Mo Hongyan;Cao Zhenzhong(Guangxi Key Laboratory of Geomechanics and Geotechnical Engineering,Guilin University of Technology,Guangxi Guilin 541004,China)

机构地区：[1]桂林理工大学广西岩土力学与工程重点实验室,广西桂林541004

出　　处：《地震学报》2024年第5期806-824,共19页Acta Seismologica Sinica

基　　金：国家自然科学基金项目(51968016,51968015);广西岩土力学与工程重点实验室主任基金(桂科能19-Y-21-8)共同资助.

摘　　要：为充分利用初至地震波中与震级相关的信息,提高震级估算精度,本文提出了一种震级快速估算方法(GPR),该方法将初至地震波在时域、频域和时频域中的10个特征参数输入高斯过程回归模型实现震级估算。利用日本的大量地表强震记录对GPR方法进行训练和测试,并与最大卓越周期τ_(p)^(max)方法和位移幅值P_(d)方法进行了对比。结果表明,GPR方法在有震源距和无震源τ_(p)^(max)距两种情况下,估算震级的准确性均显著好于方法和P_(d)方法。此外,利用智利的地表强震记τ_(p)^(max)录对日本数据训练的GPR进行泛化能力测试的结果显示,GPR方法较方法和P_(d)方法具有更好的泛化能力。利用GPR方法对我国的三次典型震例进行震级估算,验证该方法是合理且可靠的,表明GPR方法不会受到地域差异的影响,可以有效提高地震预警系统估算震级的准确度。Accurate and rapid magnitude estimation is of paramount importance for earthquake early warning systems(EEWs).Traditional magnitude estimation methods based on a single characteristic parameter of the initial seismic wave are widely used in EEWs.However,these empirical formulae,established by a single characteristic parameter,fail to fully exploit the information related to magnitude contained in the initial seismic wave,significantly limiting the effectiveness of magnitude estimation.To improve the accuracy of magnitude estimation in EEWs,this paper proposes a Gaussian process regression(GPR)based method that can estim-ate magmtudes in both scenanios:with and without bypocental distance.The proposed meth-od,GPR-M,uses multiple charactenistic parameters from the time domain,frequency do-main,and time fequency domain as inputs,while GPR-M-R incorporates hypocental dis-tance.Both methods estimate magnitude by integpating various aspects of information fom the intial seismic wave.The study utilized 33698 vertical acceleration records fom the Japanese Kiban-Kyoshin Network(KiK-net)for training and testing,and 5353 vertical acceleration records from the Chilean Simulation Based Earthquake Risk and Resilience of Interdependent Systems and Networks(SIBER RISK)for generalization testing.Additonally,the method's practical application was validated using three typical earthquake cases in China,with M_(s)5.4,M_(s)6.4,and M_(s)8.0.The perfomance of the GPR method was coupared with the widely adopted τ_(p)^(max)and P.methods.The test results from the Japamese records indicate that for imitial seis-mic waves of3 to 10 s,both GPR-M and GPR-MR outperform the p and Pa methods in magnitude estimation.Specifially,the standard deviation of estiation erors for the GPR-M method is reduced by approximately 52.53%to 61.20%compared with the rmr method,while the GPR-M-R metbod reduces the standard deviation of estimation enors by about 37.72%to 41.21%compared with the Pa method For larger earthquakes(Mw≥6.5),the magnitude sat-uration p

关 键 词：地震预警 震级估算 机器学习 高斯过程回归 

分 类 号：P315.9[天文地球—地震学]