Development of advanced electromagnetic induction based on total-field forward and inverse modeling for soil electrical property determination

(eng) The development of non-invasive geophysical techniques for digital soil mapping is important to support agricultural and environmental management activities as soil properties determine in particular crop production, groundwater recharge, runoff, river discharge, contaminant transport and climate feedbacks. In that respect, electromagnetic induction (EMI) has demonstrated to be very efficient for providing high-resolution maps of soil electrical conductivity, a surrogate mainly for water content, clay content, and salinity. Yet, existing EMI techniques still suffer from strong limitations for quantitative estimates due to calibration issues and the strong simplifying assumptions that are usually applied for data processing. The main objective of this thesis was to develop and apply advanced EMI measurement protocols and processing algorithms for quantitative retrieval of soil electrical conductivity. A novel EMI system was emulated with vector network analyzer technology and a homemade, zero-offset, off-ground loop antenna. EMI processes were described by combining a loop antenna model based on global transmission and reflection transfer functions with three-dimensional Green’s functions for wave diffusion in planar layered media. The proposed approach was successfully validated in laboratory conditions for measurements taken at different heights over a copper plane. Yet, a field validation was not possible due to the too low dynamic range of the EMI prototype. The proposed approach was then adapted for commercial EMI systems for the specific case of half-space media and was validated over water subject to different salinities. The method was then applied in field conditions along a transect. Comparison of the EMI-derived electrical conductivity with integrated values from a capacitive sensor showed satisfactory results, but discrepancies were observed and attributed to local heterogeneities and soil layering. Finally, we analyzed different strategies through numerical experiments for combining EMI and ground-penetrating radar (GPR) data in an integrated inverse modeling framework for data fusion. Although such data fusion permitted to significantly reduce uniqueness issues in the inverse problems dealt with, designing a generalized approach appeared not to be straightforward especially due to the contrasted sensitivities of both techniques in the full parameter space.