|dc.description.abstract||Long-term soil moisture monitoring sites are a source of data both for drought forecasting and hydrological and land surface modeling validation. Uncertainties in soil moisture monitoring data are not well documented and more knowledge can improve the interpretations based on these data. Texas has 23 locations with long-term soil moisture monitoring, supported by the Soil Climate Analysis Network (SCAN) and the U.S. Climate Reference Network. Both networks use Stevens Water HydraProbes for measuring soil moisture and one manufacturer-provided calibration equation for all locations. The Texas SCAN sites contain soils with vertic properties and soils with large texture discontinuities with depth, which may create distinct sources of uncertainties in soil moisture measurement. The objectives of this study were to 1) compare the default calibration to soil specific calibrations made in the lab and in-situ, 2) assess temporal and spatial uncertainties associated with using the HydraProbe, and 3) report errors and recommend methods for reducing uncertainty in SCAN data. Calibration equations were developed in the laboratory and in-situ for an Alfisol and a Vertisol. These two soils were also monitored over 18 months to 1 m depth with HydraProbes and a neutron moisture meter. Additionally, nine SCAN locations were sampled at three soil moisture conditions, field capacity, very dry and somewhere in between. Results showed root mean squared errors (RMSE) of 0.077, 0.051, and 0.035 m3m-3 for the default, lab, and in-situ calibrations for the Alfisol and 0.167, 0.077, and 0.045 m3m-3 for the Vertisol.
The data varied from 0.045 to 0.174 m3m-3 RMSE and 0.004 to 0.120 m3m-3 bias for individual SCAN sites. Soil properties of clay, pH, CEC, exchangeable cations, or bulk density did not explain trends in SCAN errors; however, a positive linear relationship between SCAN prediction errors and soil moisture was found. This study uniquely documents temporal and spatial variability in long-term soil monitoring networks in Texas and provides some documentation of errors to modelers and land use planners using soil moisture data for model evaluation.||en