摘要
以全球地震面波相速度变化问题为例，使用约３００００道高质量的面波记录数据集，在数据空间和模型空间的误差服从零平均Ｇａｕｓｓ随机分布的假设下，研究反演中的阻尼和参量化问题．发现最佳分块尺度随频率增高而减小；分块参量化方法比球谐展开方法引入的人为因素少，但难以分辨甚低阶横向非均匀性，除非施加极端的空间平滑；球谐函数展开则具备恢复长波结构的优点．为此提出一种混合参量反演方法：首先用球谐函数作为全球基函数，恢复相速度的低阶球谐分量，然后用此作为进一步反演的初始模型，用分块模型迭代反演，得到最终结果．
Abstract:Using about 30 000 seismic surface wave records associated with M≥6.0,and assuming that the errors in data and model spaces are both Gaussian random distributed with zero mean, we investigated the damping and parametrization selection in the global inversion.The horizontal resolution decreases with frequency increases.The block parametrization,with its local basis function, has less artifacts than the spherical harmonic expansion,and is difficult to recover low degree inhomogeneity except extreme spatial smoothing is imposed, while the spherical harmonic expansion has the advantage of covering long wavelength features.Therefore,a hybrid parametrization method is suggested: the spherical harmonic function is used first as a global basis function to retrieve low order spherical harmonic components,then,the result is used as an spherical reference model, and the iterative block parameterization inversion is performed to obtain final result.
出处
《地球物理学报》
SCIE
EI
CSCD
北大核心
1996年第6期753-763,共11页
Chinese Journal of Geophysics
基金
国家杰出青年科学基金
