11:55 AM - 12:10 PM
[SEM14-11] Development of a 3D inversion code to consider ocean bottom magnetotelluric data including topographic distortion ~forward modeling part~
Keywords:magnetotelluric
In recent years, new experiments using Ocean Bottom ElectroMagnetometers (OBEMs) around Japan have greatly increased the amount of seafloor electromagnetic field data that can be used to estimate sub-bottom resistivity structures using the MT method. In order to analyze such data sets, this study develops a 3D inversion method for seafloor MT data.
In this study, we are developing a code suitable for seafloor EM data based on ModEM (Egbert and Kelbert, 2012), which is a three-dimensional inversion algorithm using the finite difference method. The Flattening Seafloor (FS) technique (Baba and Seama, 2002) is introduced into ModEM to incorporate the undulations of the seafloor topography with fewer computational elements. The seafloor topography corresponds to thickness variations of the ocean and of the underlying crust. With this method, these thickness variations are transformed into variations in the electrical properties of cells that have anisotropic conductivity. As a result, the seafloor topography is represented as flat, with only two layers above and below the seafloor that retain the electrical properties of the topographic undulations. This simplification of the seafloor in numerical calculation also saves computer resources and computation time, which is especially crucial for 3D inversion.
In this study, we first investigate the correlation between the thickness of the grid cells above and below the seafloor and the computed MT response. We consider the case of a one-dimensional structure that is flat and uniformly 100 Ωm below the seafloor. The thickness of the grid cells above and below the seafloor was varied, and the MT response values obtained by forward calculation were compared with the theoretical ones. As a result, for a period of 10 seconds, the difference between the forward calculated MT response value and the theoretical one is about 5% for a thickness of 5 m above and below the seafloor, but it is as large as 80% for a thickness of 200 m above and below the seafloor. In other words, it is clear that the accuracy of the forward calculation can be improved by adding thin layers above and below the seafloor when you use seafloor MT data.
We next used ModEM with the experimental FS method to compare the FS method forward calculation with (A) and without (B) a thin layer of 0.01 m above and below the seafloor, with the forward calculation with a direct grid representation of the seafloor topography(C). The original topography was a two-dimensional topography with a smooth change in water depth from 2750 to 3250m, which is the same as the one used by Baba and Seama (2002). As a result of the comparison of (A) and (C), at a period of 1000 seconds, the apparent resistivity of (A) was 20-40% smaller and the phase of (A) was 30-40° smaller than that from (C). On the other hand, in the case of (B) the MT response values were generally consistent with the results of (C). The error in MT response is about 10% in the undulating part of the topography.
Furthermore, we found that it is necessary to prepare a wide computational domain outside the two-dimensional undulating target region when creating a numerical grid for the ModEM forward calculation using the FS method. When the numerical grid is created using only the undulating region, the values at the edge of the numerical grid, mainly for the TM mode of the MT response, show anomalous values. The difference between the anomalous values and normal values that are calculated with the direct numerical grid representation of the topography is about 40-80% for a period of 1000 seconds. In order to avoid these anomalous values in this topography with a period of 1000 seconds, it was necessary to expand the computational domain horizontally by more than 300 km. This is probably due to the fact that the boundary conditions of the ModEM numerical grid are not suitable for the FS method because the boundary conditions in ModEM do not consider the anisotropic conductivity.
In this study, we are developing a code suitable for seafloor EM data based on ModEM (Egbert and Kelbert, 2012), which is a three-dimensional inversion algorithm using the finite difference method. The Flattening Seafloor (FS) technique (Baba and Seama, 2002) is introduced into ModEM to incorporate the undulations of the seafloor topography with fewer computational elements. The seafloor topography corresponds to thickness variations of the ocean and of the underlying crust. With this method, these thickness variations are transformed into variations in the electrical properties of cells that have anisotropic conductivity. As a result, the seafloor topography is represented as flat, with only two layers above and below the seafloor that retain the electrical properties of the topographic undulations. This simplification of the seafloor in numerical calculation also saves computer resources and computation time, which is especially crucial for 3D inversion.
In this study, we first investigate the correlation between the thickness of the grid cells above and below the seafloor and the computed MT response. We consider the case of a one-dimensional structure that is flat and uniformly 100 Ωm below the seafloor. The thickness of the grid cells above and below the seafloor was varied, and the MT response values obtained by forward calculation were compared with the theoretical ones. As a result, for a period of 10 seconds, the difference between the forward calculated MT response value and the theoretical one is about 5% for a thickness of 5 m above and below the seafloor, but it is as large as 80% for a thickness of 200 m above and below the seafloor. In other words, it is clear that the accuracy of the forward calculation can be improved by adding thin layers above and below the seafloor when you use seafloor MT data.
We next used ModEM with the experimental FS method to compare the FS method forward calculation with (A) and without (B) a thin layer of 0.01 m above and below the seafloor, with the forward calculation with a direct grid representation of the seafloor topography(C). The original topography was a two-dimensional topography with a smooth change in water depth from 2750 to 3250m, which is the same as the one used by Baba and Seama (2002). As a result of the comparison of (A) and (C), at a period of 1000 seconds, the apparent resistivity of (A) was 20-40% smaller and the phase of (A) was 30-40° smaller than that from (C). On the other hand, in the case of (B) the MT response values were generally consistent with the results of (C). The error in MT response is about 10% in the undulating part of the topography.
Furthermore, we found that it is necessary to prepare a wide computational domain outside the two-dimensional undulating target region when creating a numerical grid for the ModEM forward calculation using the FS method. When the numerical grid is created using only the undulating region, the values at the edge of the numerical grid, mainly for the TM mode of the MT response, show anomalous values. The difference between the anomalous values and normal values that are calculated with the direct numerical grid representation of the topography is about 40-80% for a period of 1000 seconds. In order to avoid these anomalous values in this topography with a period of 1000 seconds, it was necessary to expand the computational domain horizontally by more than 300 km. This is probably due to the fact that the boundary conditions of the ModEM numerical grid are not suitable for the FS method because the boundary conditions in ModEM do not consider the anisotropic conductivity.