Performance assessment of factory and field calibrations for electromagnetic sensors in a loam soil

Accurate continuous measurements of temperature (T), apparent electrical conductivity (EC_a), apparent dielectric permittivity (ε_ra), and volumetric water content (θ_v) are extremely valuable to irrigation management and other agronomic decisions. The performance of eight electromagnetic (EM) sensors (TDR315, CS655, HydraProbe2, 5TE, EC5, CS616, Field Connect, AquaCheck), were analyzed through a field study in a loam soil. T, EC_a, and ε_ra were compared in reference to overall average among all sensors, and θ_v in reference to a neutron moisture meter (NMM). The reported T and EC_a difference among the sensors were within 1 °C and 1 dS m⁻¹, respectively, at 0.15 and 0.76 m depths. Among the single-sensor probes, the range of depth-combined (0.15 and 0.76 m) RMSD for factory calibration varied from 0.039 m³ m⁻³ (5TE) to 0.157 m³ m⁻³ (CS616). In comparison to single-sensor probes, RMSD of Field Connect at combined depths (0.30 and 0.51 m) was moderate (0.083 m³ m⁻³), and RMSD of AquaCheck at combined depths (0.30 and 0.61 m) was high (0.163 m³ m⁻³). Regression calibrations improved θ_v accuracy substantially beyond factory calibrations, as RMSD of the evaluated sensors except Field Connect was below 0.025 m³ m⁻³ using regression calibrations. The betterment in θ_v accuracy gained by using offset calibrations was smaller and less consistent than the improvements gained by using regression calibrations. The lower and upper bounds of the 95% confidence interval for mean RMSD of most sensors were below 0.02 and 0.04 m³ m⁻³, respectively, when using depth-specific offset calibrations. The relative success of offset calibrations for certain sensors in this study is encouraging and may signal new opportunities. Because much of the uncertainty in sensor-reported θ_v for the sensors under evaluation was systematic, future work should aim to develop universal calibrations or facilitate site-specific calibrations.