Stochastic inversion of two-layer soil model parameters from electromagnetic induction data

Soil electrical conductivity and magnetic susceptibility are connected to a number of soil properties such as water content, salinity and clay content. Electromagnetic induction (EMI) sensors for geoelectric characterization and mapping of upper soil layers typically consist of a transmitter and several spatially distributed receiver coils. In this paper, we develop a stochastic approach to the inverse problem of determination of electrical conductivity and magnetic susceptibility of two-layered soil, and thickness of its upper layer. As a forward model, we use an analytical truncated-region EMI model with one transmitter and several receiver coils placed horizontally above the soil. For solving the stochastic inversion problem we employ Markov Chain Monte Carlo (MCMC) approach. We illustrate the application of the inversion procedure on a synthetic single-frequency data set obtained from the model of an EMI sensor. Furthermore, we investigate the measurement uncertainty requirements for the sensor. The model and the stochastic inversion approach are suitable for design of EMI sensors and off-line analysis of the EMI data.