Three-dimensional magnetotelluric inversion in practice-the electrical conductivity structure of the San Andreas Fault in Central California
Three-dimensional magnetotelluric inversion in practice-the electrical conductivity structure of the San Andreas Fault in Central California
复制标题
DOI：
10.1093/gji/ggt234
复制
发表时间：
2013-10-01
影响因子：
2.8
通讯作者：
Ritter, Oliver
中科院分区：
文献类型：
--
作者：
Tietze, Kristina;Ritter, Oliver
基于合成的3-D数据，我们探索了模型空间，并测试了各种反演设置的影响。试验表明，在完全阻抗张量的反演中，明显的区域二维结构的恢复依赖于坐标系。由于标准的3-D MT反演程序不考虑数据分量之间的相互依赖关系，如果数据不与区域走向对齐，2-D地下结构可能消失。先验模型和数据加权，即阻抗张量和/或垂直磁场传递函数的单个分量对解的支配程度，是三维反演结果的关键控制因素。如果与先前模型的偏差受到严重惩罚，正则化很容易导致错误和误导性的三维反演模型，特别是在存在强烈电导率对比的情况下。“好的”总体均方根不匹配通常是没有意义的或具有误导性的，因为存在大量的3D反演结果，所有这些结果都具有类似的“可接受”不匹配，但产生的导电性结构图像明显不同。只有在频率-空间域系统地评估数据失配，才能恢复可靠和有意义的三维反演模型。
Based on synthetic 3-D data we explore the model space and test the impacts of a wide range of inversion settings. The tests showed that the recovery of a pronounced regional 2-D structure in inversion of the complete impedance tensor depends on the coordinate system. As interdependencies between data components are not considered in standard 3-D MT inversion codes, 2-D subsurface structures can vanish if data are not aligned with the regional strike direction. A priori models and data weighting, that is, how strongly individual components of the impedance tensor and/or vertical magnetic field transfer functions dominate the solution, are crucial controls for the outcome of 3-D inversion. If deviations from a prior model are heavily penalized, regularization is prone to result in erroneous and misleading 3-D inversion models, particularly in the presence of strong conductivity contrasts. A 'good' overall rms misfit is often meaningless or misleading as a huge range of 3-D inversion results exist, all with similarly 'acceptable' misfits but producing significantly differing images of the conductivity structures. Reliable and meaningful 3-D inversion models can only be recovered if data misfit is assessed systematically in the frequency-space domain.