Calibration and large-scale inversion of multi-configuration electromagnetic induction data for vadose zone characterization

Frequency-domain electromagnetic induction (EMI) devices measure a secondary magnetic field superimposed by the transmitted primary magnetic field in current conducting media. Commercially available systems convert this magnetic field ratio into an apparent electrical conductivity (ECa), not concretely stated but probably with a linear approximation assuming low induction numbers (LIN). In the LIN-based conversion, errors were observed between the true ground electrical conductivity ($\sigma(z_{i}$)) and ECa such that the present thesis introduces an improved non-linear exact ECa conversion (EEC) approach that can be used beyond the LIN approximation. Until recently, the EMI method was used for qualitative data interpretations because quantitative ECa values were often not obtained. For example, the operator or the field setup generated additional magnetic fields being measured by the EMI device that shift the recorded ECa. To eliminate the shifts, a post-calibration is required. Here, a cross-correlation between measured and predicted EMI-ECa values resulted in calibration parameters that were applied to the EMI data such that quantitative ECa values were obtained. To predict the EMI device specific ECa values, a Maxwell-based electromagnetic forward model (EM-FM) used $\sigma(z_{i}$) obtained from inverted electrical resistivity tomography (ERT) or inverted vertical electrical sounding (VES) data. Analyzing several post-calibrations based on ERT, coefficients of determination of R$^{2}$ > 0.75 were obtained when the data range along a calibration line exceeded 3 mS/m and when the ground electrical conductivity was larger than 5 mS/m. Using derived calibrations of different test sites, universal calibration parameters were obtained that allowed postcalibrations without an ERT reference line. Combining the introduced EEC with the modeling using the EM-FM that assumes horizontal layers in a multi-layer inversion of the post-calibrated EMI data, no errors were introduced anymore such that these methods can be applied also for high electrical conductive, e.g., saline areas, where the LIN approximation is no longer valid. Large-scale EMI measurements often reflect relevant subsurface patterns, but only few researchers have attempted to resolve the vertical changes in electrical conductivity [...