Laboratory calibrations of water content reflectometers and their in-situ verification

Abstract

Like time domain reflectometers, cheaper CS615 water content reflectometers (WCRs) also measure dielectric properties of the soil to determine its volumetric water content (VWC), but are more affected by environmental factors. Quadratic equations described the laboratory data from 12 horizons of 4 soils of vastly differing properties slightly better than linear equations. Root mean squared errors (RMSE) averaged over the 3 horizons increased slightly in the order Gley (av. 1.0%), Pumice (av. 1.3%), Recent (av. 1.6%), and Allophanic soil (av. 1.9%). Using the manufacturer’s standard calibration resulted in significantly higher RMSE (av. 6.2–25.3%) and mean errors (av. –5.5% to +21.5%), with the VWC of the 2 soils of volcanic origin being underestimated. An atypical dielectric response of water stored in volcanic soils has been attributed to their low bulk density, high porosity, and large specific surface area. The in-situ verification was hampered by the variability observed between the data from duplicate WCRs. Measuring the inter-sensor variability in air and water indicated that this could account for a significant part of this variability, while small-scale variation of VWC in-situ was also observed. Nevertheless, the laboratory calibrations were usually better, or at least similarly suited, to describe the in-situ data than the manufacturer’s calibrations.

Calibration of a water content reflectometer and soil water dynamics for an agroforestry practice

“…A quadratic regression equation improved the r 2 of Menfro soil, but there was no change for Putnam soils as compared to the linear calibration. This corroborates with previous studies which showed that differences between linear versus quadratic and cubic calibrations were either small or did not improve estimated water contents (Chandler et al 2004;Stenger et al 2005;Stangl et al 2009). Table 2 shows that linear and quadratic calibrations estimate h v within ±0.028-0.040 m 3 m -3 for these three soils.…”

“…Stangl et al (2009) also noticed overestimation anomalies as high as 67%. In contrast, soil water content was underestimated for volcanic soils (Stenger et al 2005). Capacitance sensors behave the same way, water content was underestimated at lower h v and overestimated at higher h v ; as high as 80% (Kelleners et al 2004b).…”

“…Similar to the results of this study, in a soil moisture sensor comparison research, Walker et al (2004) observed that CS615 sensors predicted greater soil water content near saturation than the soil porosity. In a laboratory calibration study, Stenger et al (2005) observed 15-19% overestimation in Australia with manufacturer-provided equation. Stangl et al (2009) also noticed overestimation anomalies as high as 67%.…”

“…Root mean square error (RMSE) for linear and quadratic equations with period was 0.038 and 0.039 m 3 m -3 , respectively, for the three soils combined as compared to 0.15 m 3 m -3 for the manufacturer-provided equation. RMSE is preferred as an additional measure of quality of models (Stenger et al 2005). It is a measure of goodness of fit compared to correlation and regression coefficients, and these values indicated that period alone can be used to estimate h v with less than ±0.04 m 3 m -3 difference in measured-and estimated-h v .…”

Evaluations of the effects of soil properties and electrical conductivity on the water content reflectometer calibration for landfill cover soils

“…Only sandy soil agreed with the manufacturer's calibrations. It was also noted that high clay contents strongly affect sensor output period, which impedes data interpretation and requires soil specific calibration (Stenger et al 2005;Stangl et al 2009). Correspondingly, WCRs emit lower frequency pulse which causes output period to be more sensitive to differences in soil physical properties such as bulk dry density, fine content proportions, and soil ionic concentration.…”