基于机器学习的长宁地震三维断层面几何特征建模  

Geometry features modeling of three-dimensional fault plane of Changning earthquake based on machine learning

作　　者：龚丽文 张怀 陈石[4,5] David A.Yuen 陈丽娟 Brennan Brunsvik 尹光耀 Gong Liwen;Zhang Huai;Chen Shi;David AYuen;Chen Lijuan;Brennan Brunsvik;Yin Guangyao(College of Earth and Planetary Sciences,University of Chinese Academy of Sciences,Beijing 100049,China;Key Laboratory of Computational Geodynamics,University of Chinese Academy of Sciences,Beijing 100049,China;Chongqing Earthquake Agency,Chongqing 401147,China;Institute of Geophysics,China Earthquake Administration,Beijing 100081,China;Beijing Baijiatuan Earth Science National Observation and Research Station,Beijing 100095,China;Department of Applied Physics and Applied Mathematics,Columbia University,NY 10027,USA;Department of Earth Science,University of California Santa Barbara,CA 93106,USA;School of Geosciences,University of Louisiana at Lafayette,LA 70504,USA)

机构地区：[1]中国科学院大学地球与行星学院,中国北京100049 [2]中国科学院大学计算地球动力学重点实验室,中国北京100049 [3]重庆市地震局,中国重庆401147 [4]中国地震局地球物理研究所,中国北京100081 [5]北京白家疃地球科学国家野外观测科学研究站,中国北京100095 [6]Department of Applied Physics and Applied Mathematics,Columbia University,NY 10027,USA [7]Department of Earth Science,University of California Santa Barbara,CA 93106,USA [8]School of Geosciences,University of Louisiana at Lafayette,LA 70504,USA

出　　处：《地震学报》2023年第6期1040-1054,共15页Acta Seismologica Sinica

基　　金：国家杰出青年科学基金项目(41725017);中国地震局地球物理研究所基本科研业务费专项(DQJB19A0121,DQJB21R30);国家重点研发计划课题项目(2020YFA0713401)共同资助.

摘　　要：结合长宁地区大量的地震精定位数据和其它研究成果,利用监督分类和聚类分析等机器学习算法,基于地震簇的形态特征和地震震源机制解,编写了一套自动化提取三维破裂面形态特征的程序,获取了长宁地区地震破裂面的精细结构,可为相关研究提供可参考的发震构造模型。结果显示通过聚类分析最终获取了四个地震簇,结合对应的震源机制解节面信息,最终拟合出四条破裂面,其中:长宁背斜上的破裂面沿狮子滩背斜下部的高速体呈NW−SE方向展布,破裂面平直,倾角较陡,倾向SE;建武向斜内部的三条破裂面,主要分布在向斜两翼,规模较小,走向分别为NW,NNE和NNW,从外部包围了建武向斜核部的高速体,破裂面的展布方向与该地区三个主要震源机制解节面的产状一致,其中新城镇附近的NNW向破裂面切割深度较深,约为20 km,且倾向ENE,倾角约为70°。此外,结合地质构造背景和速度结构等反演结果推断,地震破裂面主要存在于先期形成的构造薄弱带或断裂带,例如背斜的核部和向斜的翼部因节理面贯通所形成的薄弱带以及高速体周围的软弱带,在构造应力的加载和工业开采下更容易微破裂成核,形成典型的发震构造。In recent years,the seismicity of the Changning area in the Sichuan Province has increased significantly.Seismogenic models and seismogenic structures on the background of structural loading coupled with human activities have gradually become the focus of research in the field.Using abundant and accurate hypocenters in the Changning area,we established a program for automatically extracting morphological fault features by using machine learning algorithms including supervised classification and clustering.The method provides a reliable,detailed model of seismogenic faults for relative researches.As a result,four earthquake clusters were identified by clustering analysis,and four fracture planes were fit based on the distribution of hypocenters.The fracture plane on the Changning anticline spreads in NW-SE direction along a high velocity body beneath the Shizitan anticline.The fracture plane is straight with steep dip angle,and inclines SE.The three fracture planes in the inner part of the Jianwu syncline are mainly distributed in small scale on the limbs of the syncline with strike of NW,NNE,and NNW,respectively.They are also distributed in the periphery of the high-velocity body at the core of the Jianwu syncline,and their spreading directions are consistent with the strike of nodal planes of three main focal mechanism solutions in this area.Among these fracture planes,the Xincheng fracture plane extends deep to about 20 km and dips ENE with dip angle 70°.Based also on the geological tectonic settings and velocity structure,the fracture planes mainly exist in weak tectonic zones,such as the nucleus of the anticline and the limbs of the syncline.In particular,the fragile zone around the high-velocity body is more likely to rupture and nucleate under the loading of tectonic stress and industrial mining,forming new seismogenic structures.

关 键 词：断层形态 地震精定位 聚类分析 监督分类 特征提取 

分 类 号：P315.63[天文地球—地震学]