1D Inversion of Multi-component and Multi-frequency Low-induction Number EM Device (PROMIS) for Near-surface Exploration

Most of the low-induction number ElectroMagnetic sensors measure only the vertical magnetic field component for a vertical magnetic dipole source. We propose here to present the results of a sensitivity analysis of a slingram device, named PROMIS (Iris Instruments), which allows the measurement of ten frequencies ranging from 110 Hz to 56 kHz and the three components of the magnetic field. The main objective is to analyze how the radial component H-r, which varies during a profiling acquisition above a horizontally-layered ground, could improve the inversion process of the usual H-z, by essentially decreasing the number of equivalences as well as improving the convergence. First results coming from separated sensitivity analysis of H-r and H-z have shown that H-r is more sensitive to the variation of the thickness of a near-surface conductive layer. Analysis of the RMS error in the parameters' space (resistivity and thickness of the conductive layer) indicates less dispersion associated to equivalences for H-r rather than for H-z. The inversion process involving both H-r and H-z components have been performed using several methods (gradient and Monte Carlo types) including a Marquardt-Levenberg-like approach. The sensitivity analysis shows a "narrow banana" distribution of equivalences for the parameters' pair thickness/resistivity of the conductive layer, which leads to convergence problems that cannot be solved easily by the lone use of a damping factor. In order to improve significantly the convergence rate, we have implemented a two step inversion process that includes an intermediary random search step, associated to the Marquardt-Levenberg technique. This helps to circumvent the issue of equivalences in the "narrow banana" zone. Synthetic tests demonstrate clearly that, at least for thicknesses larger than few meters, the joint inversion of H-r and H-z measured with the PROMIS improves the estimation of the physical parameters.