High-resolution spatial–temporal root zone soil moisture (RZSM) information collected at different scales is useful for a variety of agricultural, hydrologic, and climate applications. RZSM can be estimated using remote sensing, empirical equations, or process-based simulation models. Machine learning (ML) approaches for evaluating RZSM across numerous spatial–temporal scales are less generalizable than process-based models. However, data-driven ML approaches offer a unique opportunity to develop complex models of soil moisture without making assumptions about the processes governing soil water dynamics in a given study region. In this study, comparisons were made between two models, pySEBAL and EFSOIL, which were based on evaporation fraction (EF) and soil properties, and a data-driven model based on the Random Forest (RF) ensemble algorithm. These approaches were evaluated to demonstrate their capabilities for RZSM estimation. The EF obtained from Landsat images was used after validation with eddy covariance measurements as the major input to all three models, along with other meteorological and soil physical properties. The RF model was trained using in situ soil moisture data from Time Domain Reflectometry (TDR) sensors installed in a vineyard from 2018 to 2020. The predictor variables comprised of meteorological, soil properties, EF, and a vegetation index. The results reveal that there was a strong correlation between the in situ measured soil moisture and the RF predicted soil moisture at all sensor locations. Due to the complexity of the physical processes involved in soil water flow, the empirical models pySEBAL and EFSOIL were unable to reliably predict RZSM values at all monitored locations. The high RZSM predicted by pySEBAL demonstrated the presence of possible bias in the model’s algorithm used to estimate soil moisture. We also demonstrated that ML based on the RF algorithm may be used to predict spatially distributed RZSM when a few soil moisture ground measurements are combined with remote sensing to produce soil moisture maps.

In situ soil moisture monitoring networks are critical to the development of soil moisture remote sensing missions as well as agricultural and environmental management, weather forecasting, and many other endeavors. These in situ networks utilize a variety of sensors and installation practices, which confounds the development of a unified reference database for satellite calibration and validation programs. As part of the Soil Moisture Active Passive Mission, the Marena, Oklahoma, In Situ Sensor Testbed (SMAP-MOISST) was initiated to perform inter-comparisons and study sensor limitations. Soil moisture sensors that are deployed in major monitoring networks were included in the study, along with new and emerging technologies, such as the Cosmic Ray Soil Moisture Observing System (COSMOS), passive/active distributed temperature sensing (DTS), and global positioning system reflectometers (GPSR). Four profile stations were installed in May of 2010, and soil moisture was monitored to a depth of 1 m on an hourly basis. The four stations were distributed within a circular domain of approximately 600 m diameter, adequate to encompass the sensing range of COSMOS. The sensors included in the base station configuration included the Stevens Water Hydra Probe, Campbell Scientific 616 and 229, Decagon EC-TM, Delta-T Theta Probe, Acclima, and Sentek EnviroSMART capacitance system. In addition, the Pico TRIME system and additional time-domain reflectometry (TDR) systems were deployed when available. It was necessary to apply site-specific calibration to most sensors to reach an RMSE below 0.04 m³ m^-3. For most sensor types, a single near surface sensor could be scaled to represent the areal-average of a field domain by simple linear regression, resulting in RMSE values around 0.03 m³ m^-3.