Mapping depth-to-clay using fitted multiple depth response curves of a proximal EMI sensor

As an alternative for the depth response approximations based on the theoretical Maxwellʹs equations, a procedure was proposed to fit depth response curves for different coil configurations. A 39 ha study area was selected in the Belgian loess belt, where loess material was situated on a Tertiary substrate. A survey with the DUALEM-21S electromagnetic induction instrument was carried out to map the depth-to-clay (zclay). The depth response curves were fitted both for the vertical and perpendicular coil configurations using 85 depth observations of zclay. The resulting depth response curves R(zclay) were: R p, s ( z clay ) = 0 .8135 ⋅ e -1 .4131 ⋅ z clay s for the perpendicular coil configuration (with s as the intercoil spacing) and R v, s ( z clay ) = 0 .9802 ⋅ e -0 .8102 ⋅ z clay s for the vertical coil configuration. A set of 4 equations based on the developed depth response functions was used to model zclay at each of the 209 400 measurement points. These zclay predictions were validated using geo-electrical imaging. With two multi-electrode resistivity arrays, zclay was 1D-inverted at 95 locations along two transects, assuming a two-layered soil system. A coefficient of determination of 0.95, with a root mean-squared estimation error of 0.22 m, was found between the predicted and 1D-inverted depths. This procedure allowed the accurate 3D-reconstruction of the paleolandscape before the deposition of the loess. Flow lines were modelled on this paleosurface, revealing past or subsurface stream patterns not visible on the present relief.