Comparing one- and two-dimensional EMI conductivity inverse modeling procedures for characterizing a two-layered soil

Abstract

Apparent electrical conductivity (σa) measured with electromagnetic induction (EMI) instruments is used widely as a proxy to map lateral variations in soil properties. To account for the vertical changes in soil properties, EMI inversion techniques have been developed. Improvements in EMI sensing instrumentation allow simultaneous recording of σa measurements with different depth responses, facilitating 1D and 2D inversions. In this study, different inversion procedures of EMI data were evaluated on their effectiveness to characterize the depth of the interface between two contrasting soil layers, as well as their respective conductivities. A 1D-laterally constrained inversion procedure was compared with a non-constrained, robust 1D-inversion procedure. Both procedures make use of low induction number (LIN)-approximated depth response curves and provided similar results, although calibration with soil data was essential to attribute absolute values to the inversion data. Aforementioned 1D procedures were then compared to a procedure wherein the full solution of Maxwell's equations, which describe the electromagnetic signal response into the soil, is applied. All approaches rendered similar results. This shows that despite their limitations, the cumulative approximations of the EMI signal response can be used as a valuable and effective alternative to the full solution of Maxwell equations within EMI inversion procedures, especially when the measurements are mainly situated within the LIN measurement range. In a final step, 2D inversions of the EMI data were compared to 2D-inverted electrical resistivity tomography (ERT) data. The general patterns of the resulting inverted soil models were largely comparable and consistent with the observed soil information. To conclude, the different inversion procedures revealed analogous results which were largely comparable and consistent with the soil information. However absolute values could impossibly be obtained without any prior knowledge about the vertical distribution of the soil model. Therefore, the implementation of a thorough calibration based on soil observations was essential to guide the inversion results to a realistic outcome.