A new approach for separating mixed model parameters: application to simultaneous inversion of earthquake source parameters
Weijian Mao1,2
1 Center for Computational and Exploration Geophysics, Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China 2 State Key Laboratory of Geodesy and Earth’s Dynamics, Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China
A new approach for separating mixed model parameters: application to simultaneous inversion of earthquake source parameters
Weijian Mao1,2
1 Center for Computational and Exploration Geophysics, Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China 2 State Key Laboratory of Geodesy and Earth’s Dynamics, Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China
摘要A method for simultaneous determination of mixed model parameters, which have different physical dimensions or different responses to data, is presented. Mixed parameter estimation from observed data within a single model space shows instabilities and trade-offs of the solutions. We separate the model space into N-subspaces based on their physical properties or computational convenience and solve the N-subspaces systems by damped least-squares and singular-value decomposition. Since the condition number of each subsystem is smaller than that of the single global system, the approach can greatly increase the stability of the inversion. We also introduce different damping factors into the subsystems to reduce the tradeoffs between the different parameters. The damping factors depend on the conditioning of the subsystems and may be adequately chosen in a range from 0.1 % to 10 % of the largest singular value. We illustrate the method with an example of simultaneous determination of source history, source geometry, and hypocentral location from regional seismograms, although it is applicable to any geophysical inversion.
Abstract:A method for simultaneous determination of mixed model parameters, which have different physical dimensions or different responses to data, is presented. Mixed parameter estimation from observed data within a single model space shows instabilities and trade-offs of the solutions. We separate the model space into N-subspaces based on their physical properties or computational convenience and solve the N-subspaces systems by damped least-squares and singular-value decomposition. Since the condition number of each subsystem is smaller than that of the single global system, the approach can greatly increase the stability of the inversion. We also introduce different damping factors into the subsystems to reduce the tradeoffs between the different parameters. The damping factors depend on the conditioning of the subsystems and may be adequately chosen in a range from 0.1 % to 10 % of the largest singular value. We illustrate the method with an example of simultaneous determination of source history, source geometry, and hypocentral location from regional seismograms, although it is applicable to any geophysical inversion.
基金资助:Innovation Project of Chinese Academy of Sciences
通讯作者:
Weijian Mao
E-mail: wjmao@whigg.ac.cn
引用本文:
Weijian Mao. A new approach for separating mixed model parameters: application to simultaneous inversion of earthquake source parameters[J]. 《地震学报》英文版, 2014, 27(2): 189-196.
Weijian Mao. A new approach for separating mixed model parameters: application to simultaneous inversion of earthquake source parameters. Earthquake Science, 2014, 27(2): 189-196.