From 1D-Multi-Layer-Conductivity-Inversion to Pseudo-3D-Imaging of Quantified Electromagnetic Induction Data Acquired at a Heterogeneous Test Site

Electromagnetic induction (EMI) systems enable the non-invasive spatial characterization of soil structural andhydrogeological variations, since the measured apparent electrical conductivity (ECa) can be related to changes insoil moisture, soil water, clay content and/or salinity. Due to the contactless operation, ECa maps of relatively largeareas, i.e. field to (small) catchment scale, can be measured in reasonably short times. A multi-configuration EMIsystem with one electromagnetic field transmitter and various receivers with different offsets provide simultaneousECa measurements that are representative of different sensing depths. Unfortunately, measured ECa values canonly be considered as qualitative values due to external influences like the operator, cables or other metal objects.Of course, a better vertical characterization of the subsurface is possible when quantitative measurement valuescould be obtained. To obtain such quantitative ECa values, the measured EMI apparent conductivities are calibratedusing a linear regression approach with predicted apparent conductivities obtained from a Maxwell-basedfull-solution forward model using inverted electrical resistivity tomography (ERT) data as input. These calibratedapparent conductivities enable a quantitative multi-layer-inversion to resolve for the electrical conductivity of certainlayers. To invert for a large scale three-layer model, a one-dimensional (1D) shuffled-complex-evolution inversionscheme was parallelized and run on JUROPA – one of the supercomputers of the Forschungszentrum Jülich.This novel inversion routine was applied to calibrated electromagnetic induction data acquired at the Selhausentest site (Germany), which has a size of about 190 x 70 m. The test site is weakly inclined and a distinct gradientin soil texture is present with considerably higher gravel content at the upper part of the field. Parallel profiles withapproximately three meter distance were measured using three different coil offsets in HCP and VCP measurementmodes. This resulted in six high spatial resolution data sets of approximately 60000 measurements with differentsensing depths. A 5 m block-kriging was applied to all six data sets to re-grid the sampling points on the sameregular grid. For each grid node, the six measured apparent conductivities were used in a three-layer inversion.The three-layer inversion results of electrical conductivity thus obtained were used to derive a three-dimensional(3D) model of subsurface heterogeneity, which clearly indicated lateral and vertical conductivity changes of thesubsurface that are related to changes in soil texture and soil water content