A hybrid grid-based finite-element approach for three-dimensional magnetotelluric forward modeling in general anisotropic media

Magnetotelluric (MT) forward modeling often requires the consideration of the deviation generated by anisotropic structures to avoid misleading the detection results, which is implemented by the flexible unstructured finite-element (FE) method based on tetrahedrons. However, the unstructured FE method needs a large number of small elements near the sharp boundary of the electrical structure to produce sufficiently stable grids, and this leads to a sharp increase in the degrees of freedom (DoFs) of the FE system. To this end, we develop an FE approach based on a hybrid grid. It uses the stretched prismatic elements to divide the regions around the abrupt interfaces, which can significantly reduce the number of elements needed to capture the corresponding changes of the physical fields compared to using tetrahedrons. The remaining regions are still divided with tetrahedrons in this hybrid grid in order to keep the total elements needed for the entire computational region at a minimum. By showing numerical examples of the sea- and land-based electrical anisotropy models, as well as a real anisotropic inversion model, the advantages of our method are tested by evaluating the MT response curves, the number of elements, the DoFs, the computational time, and the consumed memory. The results show that the prismatic elements are suitable for discretizing the near-surface region, seawater layer, and electrical anisotropic blocks. This approach can maintain the high accuracy of the numerical solution and reduce the number of elements required for discretizing these regions and the DoFs, thus requiring less computer memory. With this hybrid grid, the computational efficiency of the FE method can be improved for MT forward modeling for a complicated model, which combined with the lower memory consumption is very suitable to implement inversion.

Automated summary

This paper proposes a finite element approach based on a hybrid grid for handling deviations in magnetotelluric forward modeling. By using a combination of prismatic and tetrahedral elements, this method can reduce the number of elements and degrees of freedom while maintaining high accuracy, thus improving computational efficiency.