Reliability of 3D finite-element and finite-difference inversion of magnetotelluric data including topography for geothermal exploration: Case study in Okuaizu geothermal field

To investigate the performance of the three-dimensional (3D) inversion of magnetotelluric (MT) data for geothermal exploration (where accurate numerical modeling is essential for addressing rough topographies), we utilized two inversion codes (FEMTIC and WSINV3DMT) for the 3D inversion of MT data obtained from the Okuaizu geothermal area, northern Japan. FEMTIC is a finite-element (FEM) inversion code. It can incorporate tetrahedral elements (Tetra) or deformed nonconforming hexahedral elements (DHexa) to construct a 3D mesh. Meanwhile, WSINV3DMT is a finite-difference (FDM) inversion code. It uses rectangular cells to discretize the 3D domain. We prepared an identical subset of the MT data and set an identical noise floor to run the Tetra, DHexa, and WSINV3DMT inversions. The three inversions yielded similar 3D models. These displayed resistivity anomalies related to the cap rock and geothermal reservoir in the area. However, there are several significant differences in the model details, particularly between the FEMTIC and WSINV3DMT inversions. Numerical experiments on 3D synthetic data based on the inversion results of the field data were then conducted for DHexa and WSINV3DMT to examine the factors causing these differences. We set two low-resistivity anomalies (shallow and deep) embedded in a homogeneous earth with real topography. The experiments revealed that the DHexa inversion effectively recovered the two anomalies. However, the WSINV3DMT inversion may have failed to recover these. In particular, the deep anomaly was reconstructed ineffectively owing to numerical errors when we included a rough topographic variation in the model. Therefore, we considered that the inversion results of field data using an FDM code may have unreliable anomalies for an MT dataset obtained in a rough terrain environment. The inverted models of the field data in Okuaizu by Tetra and DHexa showed good agreement with existing borehole logging data and the geothermal conceptual model of the area.