Probabilistic inversion of time-lapse EMI data for spatiotemporal monitoring of soil moisture

Low frequency electromagnetic induction (EMI) systems have proven to be useful in mapping soil apparent electrical conductivity and soil moisture. Nevertheless, obtaining depth profile water content has not been fully explored using EMI. In this study, we performed time-lapse EMI measurements using CMD mini-Explorer sensor along a 10 m transect in a corn farm during a 6 day period. Meanwhile, reference data were also regularly measured at the end of the profile in an excavated pit using 5TE capacitance sensors. In order to derive a timelapse, depth-specific subsurface image of true conductivity, we applied a probabilistic optimization approach, DREAM(ZS), on the measured EMI data. The uncertainties in measured apparent electrical conductivity, as well as inaccuracies in the inverted data, introduced some discrepancies between estimated and reference values in time and space. Moreover, the difference in measurement footprints of the 5TE and CMD Mini-Explorer sensors also led to the differences between reference and estimated data. The obtained depth profile permitted to accurately monitor spatiotemporal distribution and variation of soil water content encountered because of root water uptake and evaporation. The time-lapse monitoring approach, developed using DREAM(ZS) appears to be pertinent for accurate retrieval of spatiotemporal distribution of soil water content.