Open AccessA Robust Calibration Method for Continental‐Scale Soil Water Content Measurements
Author(s) -
Roberti Joshua A.,
Ayres Edward,
Loescher Henry W.,
Tang Jianwu,
Starr Gregory,
Durden David J.,
Smith Derek E.,
Reguera Elizabeth,
Morkeski Kate,
McKlveen Margot,
Benstead Heidi,
SanClements Michael D.,
Lee Robert H.,
Gebremedhin Maheteme,
Zulueta Rommel C.
Publication year - 2018
Publication title -
vadose zone journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.036
H-Index - 81
ISSN - 1539-1663
DOI - 10.2136/vzj2017.10.0177
Subject(s) - water content , calibration , capacitance probe , mean squared error , soil science , environmental science , scale (ratio) , soil water , range (aeronautics) , remote sensing , capacitance , mathematics , statistics , geology , materials science , geography , geotechnical engineering , chemistry , cartography , electrode , composite material
Core Ideas We present a semiautomated framework for large‐scale calibration of capacitance soil moisture sensors. Soil‐specific calibrations are needed for accurate moisture measurements in nearly all soil types. Transparent, traceable, and complete uncertainty estimates provided for entire approach. Technological advances have allowed in situ monitoring of soil water content in an automated manner. These advances, along with an increase in large‐scale networks monitoring soil water content, stress the need for a robust calibration framework that ensures that soil water content measurements are accurate and reliable. We have developed an approach to make consistent and comparable soil water content sensor calibrations across a continental‐scale network in a production framework that incorporates a thorough accounting of uncertainties. More than 150 soil blocks of varying characteristics from 33 locations across the United States were used to generate soil‐specific calibration coefficients for a capacitance sensor. We found that the manufacturer's nominal calibration coefficients poorly fit the data for nearly all soil types. This resulted in negative (91% of samples) and positive (5% of samples) biases and a mean root mean square error (RMSE) of 0.123 cm 3 cm −3 (1σ) relative to reference standard measurements. We derived soil‐specific coefficients, and when used with the manufacturer's nominal function, the biases were corrected and the mean RMSE dropped to ±0.017 cm 3 cm −3 (±1σ). A logistic calibration function further reduced the mean RMSE to ±0.016 cm 3 cm −3 (±1σ) and increased the range of soil moistures to which the calibration applied by 18% compared with the manufacturer's function. However, the uncertainty of the reference standard was notable (±0.022 cm 3 cm −3 ), and when propagated in quadrature with RMSE estimates, the combined uncertainty of the calibrated volumetric soil water content values increased to ±0.028 cm 3 cm −3 regardless of the calibration function used.