Field use and calibration of a TDR-based probe for monitoring water content in a high-clay landslide soil in Austria

Dielectric-based sensors are widely used for field monitoring of soil volumetric water content (θv), including in situ applications in ecological monitoring programs. However, sensor response depends strongly on the location-specific soil properties, which in turn affects measurement accuracy and data processing. Published general or manufacturers' calibrations often misrepresent the θv-sensor output relationship, requiring soil-specific calibration. We report on use of the CS615 Water Content Reflectometer (WCR) to monitor the soil water dynamics in a creeping flow at a landslide site (Bad Goisern, Austria), and on the soil-specific adjustment of measurement errors. Extraordinary soil conditions (high clay and water contents) caused anomalous overestimation of θv via the manufacturer's standard calibration. Further, a laboratory calibration had to be aborted due to the intractable soil material. However an in situ field calibration and an ex situ field-soil calibration successfully provided relations between θv and the probe output (multivibrator period, τ). The calibration was performed as a two-stage procedure according to the inverse regression method. Linear (LR) and multiple (MR) regression models and polynomial (P2, P3) relations were generated via regression analyses. Bias, mean squared error (MSE) and mean deviation (MD) were used to evaluate the quality of θv estimation using the inverse prediction function. LR and MR models provided better data adjustment than polynomial functions. Best results were derived from MR models including as additional variables temperature (T) and porosity (P), and subset-specific (S) to sensor position in the field (model MR TP S). Measurement error was reduced from 0.068 ± 0.122 m3 m− 3 (MSE ± σ for the standard calibration) to 0.001 ±0.002 m3 m− 3 (MSE ± σ for the MR TP S model). Restricted sample size and moisture range impaired the statistical analyses of both field soil calibrations. Deviations of sensor response specific to soil layer and sensor position were observed and statistically confirmed. However, reasonable location-specific calibration functions were obtained for both the entire water content range and the site-specific high moisture range. Our results indicate an especially anomalous, soil pH-dependent response of the WCR (which operates in the lower frequency range 15 to 45 MHz) in a smectite-dominated soil, partly consistent with the findings of Ishida and Makino (1999) for the dielectric behaviour of montmorillonite suspensions. Unfavourable soil conditions, especially high moisture levels combined with high clay contents, demonstrated the limits of WCR-application. Our findings strongly support media-specific sensor calibration over general calibrations, especially for soils with extraordinary and challenging properties.