基于非结构网格及自适应有限元的海洋CSEM和MT三维联合反演方法研究

Marine Magnetotellurics (MT) performs well at imaging large-scale regional electrical structures of the interior earth, whereas marine Controlled-Source Electromagnetic (CSEM) method has a better resolution towards relatively shallow local electrical structures. Therefore, it is limited to reflect the real underlying structures in inversion when only one single type of data is involved. The traditional 3-D inversion of marine CSEM method, which is usually based on structured grids, cannot simulate the submarine undulating terrain and arbitrarily complex geo-electric structures effectively and accurately. .To address these problems, this project intends to carry out a research on 3-D joint inversion of marine CSEM and MT data to reduce the non-uniqueness and uncertainty of inversion results. We adopt a novel goal-oriented adaptive edge-based finite-element algorithm based on unstructured tetrahedral grids to improve the accuracy of numerical simulation, and a parallel scheme based on MPI is developed to improve computational efficiency as well. Conjugated Gaussian Newton optimization algorithm, which does not need to store the Jacobian matrix directly, is employed to reduce the memory requirements in the process of inversion. Also, the optimal data weighting method is employed to ensure the stability of the joint inversion. The results of both separate inversions and joint inversion will be performed and compared respectively with synthetic marine CSEM and MT data. And then a joint inversion will be performed with realistic data based on appropriate a priori constraints. The incorporation of multiple marine EM methods provides a new approach to marine EM surveying and data interpretation, which effectively increases the reliability of marine electromagnetic exploration.

海洋MT主要用于探测深部大尺度的区域电性结构；海洋CSEM主要用于探测浅部小尺度的局部电性结构。当反演中采用单独一种数据时，其结果难以反映真实地质构造，同时传统的海洋电磁法三维反演通常基于结构化网格，这难以有效模拟实际的海底起伏地形和复杂的地电结构。针对这些问题，本项目拟开展海洋CSEM和MT三维联合反演方法研究，以降低反演中解的非唯一性。采用非结构四面体网格和面向目标的自适应矢量有限元算法提高正演数值精度，同时采用MPI并行技术提高计算效率。在求解反演问题中，采用不直接存储雅可比矩阵的高斯牛顿变种算法，降低内存需求，同时选取最优的数据加权方法，确保联合反演的稳定性。针对理论合成数据，分别利用单一数据进行单独反演和两种数据开展联合反演，添加合适的先验约束条件，最终实现对实测海洋电磁数据的联合反演，为海洋电磁探测提供可靠的数据解释手段和理论方法技术，提高海洋电磁勘探的可靠性。

本报告是国家自然科学基金青年科学基金项目“基于非结构网格及自适应有限元的海洋CSEM和MT三维联合反演方法研究”结题报告。根据项目任务书要求，我们开展了海洋CSEM和MT三维联合反演方法研究工作，取得了一些创新性结果：实现了电阻率各向异性介质海洋CSEM和MT三维自适应矢量有限元正演，分析了不同网格细化比例对数值模拟结果的影响，并通过一维理论模型验证了算法的有效性，进一步模拟了三维各向异性模型CSEM和MT响应，较详细地讨论了电阻率各向异性对海洋CSEM和MT响应的影响；基于MPI技术实现了并行计算，提升了正演计算效率；采用伴随方法推导并计算了相应的灵敏度矩阵计算;实现了海洋CSEM和MT三维联合反演；.发表带项目资助标注的论文5篇，其中SCI论文4篇，EI论文1篇；在读硕士生3名。