3-D joint inversion of airborne gravity gradiometry and magnetic data using a probabilistic method

3-D joint inversion of airborne gravity gradiometry and magnetic data using a probabilistic method
复制标题

基于概率方法的航空重力梯度与地磁数据三维联合反演

DOI:
10.1093/gji/ggaa283
复制
发表时间:
2020-06
影响因子:
2.8
通讯作者:
M. Geng;J. Welford;C. Farquharson;A. Peace;Xiangyun Hu
M. Geng;J. Welford;C. Farquharson;A. Peace;Xiangyun Hu
中科院分区:
地球科学2区
文献类型:
--
作者:
M. Geng;J. Welford;C. Farquharson;A. Peace;Xiangyun Hu

文献摘要

相似文献

提出了一种概率方法,用于联合反演重力梯度和磁力数据,以获得密度和磁化率的 3D 地下分布。通过使用密度和磁化率的互协方差矩阵将物理性质模型的耦合纳入反演中。这使得能够合并结构相似性,例如源的方向和空间范围以及两个物理属性之间的交叉方差。用户定义的相关系数可以控制两个模型之间的相似程度。通过在联合反演中应用边缘化算法,可以将反演域划分为不同的区域,根据源的特征和相似性,每个区域可以有自己的协方差、互协方差矩阵以及相关系数。因此,可以同时恢复具有不同形状、大小和两种物理特性之间关系的源。使用三个综合示例验证了该方法的有效性,这些示例演示了互协方差矩阵的错误参数如何影响倒置结果。最后,所提出的方法成功应用于在加拿大纽芬兰中北部的巴吉尔港股票(BHS)入侵中收集的全张量梯度测量和磁数据。与独立反演生成的模型相比,使用联合反演可以更好地定义和定位主要侵入体以及浅层相关的煌斑岩脉。所解析的侵入物物理特性与 BHS 附近的煌斑岩岩脉的现场观察结果非常一致。
A probabilistic approach is presented for jointly inverting gravity gradient and magnetic data for 3-D subsurface distributions of density and magnetic susceptibility. The coupling of the physical property models is incorporated in the inversion by using a cross-covariance matrix of density and magnetic susceptibility. This enables structural similarity such as the orientation and spatial extent of the sources and cross-variance between the two physical properties to be incorporated. A user-defined correlation coefficient can control the level of similarity between the two models. By applying a marginalizing algorithm in the joint inversion, the inversion domain is allowed to be partitioned into various zones, each of which can have its own covariance, cross-covariance matrix, as well as correlation coefficient, depending upon the feature and similarity of sources. Thus, sources with different shapes, sizes and relationships between the two physical properties can be simultaneously recovered. The validity of the method is verified using three synthetic examples, which demonstrate how incorrect parameters of the cross-covariance matrix affect the inverted results. Finally, the proposed method is successfully applied to full tensor gradiometry and magnetic data collected over the Budgell Harbour Stock (BHS) intrusion in north-central Newfoundland, Canada. Compared with models generated from independent inversions, better definition and localization of the main intrusion, as well as associated lamprophyre dykes at shallow depth, are achieved by using the joint inversion. The resolved physical properties for the intrusions show good agreement with field observations of lamprophyre dykes in proximity to the BHS.