Calibration of a novel low-cost soil water content sensor based on a ring oscillator

Wireless sensor networks are becoming more popular to monitor spatial and temporal variability of soil water content. The aim of this paper is to calibrate and test the newly developed SPADE sensor for wireless sensor network applications. During calibration, the measured sensor output is related to dielectric permittivity using an empirical calibration in liquids with well-defined permittivities. Using these calibration measurements, we have evaluated sensor-to-sensor variability and compared the accuracy of a universal calibration between sensor output and permittivity with a sensor-specific calibration. Our results showed that the sensor-to-sensor variability of the SPADE sensor is larger than sensor noise and that a sensor-specific calibration can improve the accuracy of soil water content estimation as compared to a single universal calibration. To quantify the effect of temperature on the sensor output, we have derived a temperature correction function in a temperature range from 5 to 40°C. The results showed an underestimation for low temperatures (5 to 25°C) and an overestimation for high temperatures (25 to 40°C). A second-order polynomial function was fitted to the measurements with good agreement (R2 = 0.9831, RMSE = 0.1489). The transferability of the temperature correction function from reference liquids to soils was verified using saturated coarse sand and a silty loam soil sample. The temperature-corrected SPADE sensor measurements in soil showed good agreement with TDR measurements and corresponded well with theoretical predictions from a dielectric mixing model.