Inversion of multi-source time domain electromagnetic data

导电性是与矿藏、石油和天然气、水和环境污染物相关的关键物理性质。已经设计了电磁测量来获取对未知电导率敏感的数据。通常，由接地偶极子或电流环源引起的电场和/或磁场被记录下来，然后实际困难是恢复导致这些测量数据的3D导电结构。UBC地球物理反演设施之前的研究已经产生了在仅使用几个发射源的情况下同时反演频率(FE)和时间域(TEM)数据的方法和软件程序。这些成果被用于赞助公司。然而，瞬变电磁数据的三维反演是一项具有计算挑战性的任务，它随着发射机位置的增加而增加，而且目前的技术不适用于使用许多发射机位置的调查。在这项研究中，我们建议开发稳健和高效的三维反演方法，用于与航空或地面测量相关的多源瞬变电磁数据集。这项研究的一个重要组成部分是制定处理这些大型数据集的工作流程和高效战略。我们建议开发实用软件，并与行业赞助商密切合作，以实现这一目标。最终代码将分发给赞助商。首先，麦克斯韦方程将以电场和磁场的形式重新表述，随后的正演模拟矩阵将被分解。我们之前开发的拟牛顿方法将被实施。该代码的测试版将在赞助商提供的合成和现场数据集上进行测试，包括使用多个来源的地面和航空调查。工作流程和实用程序将随着测试的进展而开发。最终的反转代码(从MatLab转换成Fortran)将交付给赞助商，包括运行代码、文档和工作流元素的图形用户界面。

Electrical conductivity is a key physical property associated with mineral deposits, oil and gas, water, and environmental contaminants. Electromagnetic surveys have been devised to acquire data that are sensitive to the unknown conductivity. Typically, the electric and/or magnetic fields caused by a grounded dipole or a current loop source are recorded, and the practical difficulty is then to recover the 3D conductivity structure that caused these measured data. Previous research at the UBC Geophysical Inversion Facility has resulted in methodologies and software programs for inverting both frequency (FEM) and time domain (TEM) data when only a few transmitter sources are employed. The outcomes are in use at sponsoring companies. However, 3D inversion of TEM data is a computationally challenging task that increases with the number of transmitter locations, and current techniques are not viable for surveys using many transmitter locations. In this research we propose to develop robust and efficient 3D inversion methods for multi-source TEM data sets associated with airborne or ground surveys. An important  component of the research is to develop workflows and efficient strategies for handling these large data sets. We propose to develop practical software and work closely with industry sponsors to achieve this goal. Final codes will be distributed to the sponsors. First, Maxwell's equations will be reformulated in terms of electric and magnetic fields and the subsequent forward modeling matrix will be decomposed. The quasi-Newton methodology we developed earlier will be implemented. The beta-version of the code will be tested on synthetic and field data sets provided by sponsors, including ground and airborne surveys with multiple sources. Workflow and utility programs will be developed as testing progresses. A final inversion code (converted from MatLab into Fortran) will be delivered to sponsors, including GUI interfaces to run the codes, documentation, and workflow elements.