Field-scale root-zone soil moisture mapping in sandy soils with L-band radiometry and hybrid radiative transfer – machine learning modeling

Accurate field-scale estimation and mapping of root-zone soil moisture is critical for precision irrigation management and optimizing crop yield, especially in Florida’s sandy agroecosystems where low water-holding capacity and high nutrient leaching increase irrigation challenges. While microwave satellites provide soil moisture at large scale, their coarse resolution (km scale) and surface (0–5 cm) estimates limit their application for within-field irrigation decisions. In this study, we developed and evaluated a field-scale mapping framework that integrates a Utility Terrain Vehicle (UTV)-mounted dual-polarized L -band (1.4 GHz) radiometer with i) the tau-omega radiative transfer model, and ii) a hybrid (τ–ω–XGBoost) approach that integrates tau-omega outputs with extreme gradient boosting to estimate and map soil moisture at 10, 20, 30, and 40 cm depths. The framework combines temporal brightness temperature observations with ancillary variables (e.g., vegetation water content, effective soil temperature) to parameterize tau-omega model at the field-scale. The resulting estimates were then assimilated into extreme gradient boosting (XGBoost) as additional predictors and physical constraints to improve retrieval accuracy. Results indicated that tau-omega and hybrid approaches produced mean RMSE of ∼ 0.02–0.04 cm 3 cm -3 , with best performance at upper depths and vertical polarization outperforming horizontal polarization. Allowing spatial variability in surface roughness improved tau-omega model parameterization and retrieval accuracy. The hybrid model slightly outperformed the tau-omega model, especially at deeper depths (mean unbiased RMSE of 0.020 cm 3 cm −3 ). Overall, the proposed framework not only provides a mesoscale bridge between point sensors and satellite pixels for field-scale mapping of root-zone soil moisture to support irrigation management in sandy agroecosystems, but it can also benefit airborne- and satellite-based soil moisture retrievals. • Mobile dual-polarized L -band radiometry provided ∼1 m root-zone soil moisture maps. • Tau-omega and XGBoost hybrid model improved retrievals, especially at depth. • Considering spatial surface roughness variability improved model performance. • V polarization outperformed H polarization for brightness temperature and soil moisture retrievals.