Laboratory calibration of the CS655 soil moisture sensor
MetadataShow full item record
Calibration of a frequency domain reflectometer (model CS655, Campbell Scientific, Inc. Logan, UT) is presented using five distinct soil types and three different calibration methods. Frequency domain reflectometers estimate soil water content (SWC) using electromagnetic properties of the surrounding media. Few, if any, sensors directly measure volumetric water content. Instead, a sensor’s output must be converted either by a universal or a user-specific calibration equation. These sensors are used for a variety of applications using a factory-supplied equation with an error of 0.03 m³m⁻³. Soil specific properties such as clay content and salinity can affect their performance in field situations. A site or soil specific calibration can provide more accurate measurements albeit at greater time and expense. For this research, three calibration methods on five central Texas soils were evaluated to determine soil-specific calibration equations. First, a standard calibration method was performed by packing soil cores with soil at progressively higher SWC, inserting the probe vertically, and taking repeated measurements. Next, an upward infiltration method was used to slowly introduce water at the bottom of the soil core performed on soil cores with vertically inserted probes. Lastly, a downward infiltration method was performed by introducing known amounts of water to the top of the soil core with a vertically inserted probe and allowing infiltration and redistribution between subsequent water additions. The data from all three methods were fitted to a third-order polynomial, based on the relationship between the dielectric permittivity and the SWC. Overall, the CS655 performance across all five soils improved from a root mean square error (RMSE) of 0.065 and 0.042 m³m⁻³ using standard and downward calibrations, respectively, to 0.026 and 0.024 m³m⁻³ using a site-wide calibration. Results further indicate that the soil-specific calibration curves provide better fits than the commonly-used Topp’s equation, and that the coefficients in the soil-specific curve differ significantly (p<0.05). The research presented here improves our understanding of the CS655 sensor, and the calibration curve needed to improve field-based measurements currently occurring across the Texas Soil Observation Network.