Recalibration of Sensors in One of The World's Longest Running Automated Soil Moisture Monitoring Networks

Abstract

Obtaining accurate soil moisture data from sensors in automated monitoring networks is critical as these data are increasingly used for research in soil hydrology, ecohydrology, and related disciplines. One of the earliest such networks is the Oklahoma Mesonet, which monitors soil matric potential using heat dissipation sensors. Various calibration equations have been proposed for those sensors, and there is a need to compare and validate the performance of those equations, especially for matric potentials < –150 kPa. A laboratory experiment was conducted in silt loam soil using a sand‐kaolin box and a pressure plate apparatus with matric potentials ranging from 0 to –1500 kPa. The calibration equations included Starks’ equation, Flint et al.'s equation, Schneider et al.'s equation, and a new logistic equation. The upper limit of the sensors was ‐9 kPa, and the sensors remained responsive at ‐1500 kPa matric potential. The logistic equation produced the lowest root mean squared error (34 kPa), followed by Flint et al.'s equation (192 kPa), Starks’ equation (295 kPa), and Schneider et al.'s equation (463 kPa). After recalibration of the coefficients in the three preexisting equations, their performances improved, with all RMSE values ≤ 251 kPa; however, the logistic equation still provided superior accuracy. The logistic equation effectively removed an ∼ 0.02 cm3 cm−3 positive bias in soil water content that resulted from use of the original parameterization of Schneider et al.'s equation. This logistic equation is recommended for use with past and future data from the Oklahoma Mesonet's heat dissipation sensors.Core Ideas Soil matric potential data from the Oklahoma Mesonet are widely used. Those data are estimated using a calibration equation shown here to be inaccurate. The logistic equation developed here produces more accurate matric potential estimates.