Key points are not available for this paper at this time.
Accurate estimation of soil unsaturated hydraulic conductivity (kθ) is critical for predicting vadose zone flow dynamics and characterizing subsurface hydrological processes. Traditional point scale tests are invasive and lack the spatiotemporal resolution required to capture field heterogeneity. This study presents an innovative framework that couples time-lapse ground penetrating radar (GPR), electrical resistivity tomography (ERT) with an improved instantaneous profile (IIP) inversion to non-destructively quantify kθ dynamics in different soils. Resulting kθ estimates were validated against laboratory soil water characteristic curve (SWCC)-based predictions from van Genuchten Mualem (VGM) and Childs–Collis-George (CCG) models. Parsimonious, logarithmic constitutive models were established linking kθ to relative permittivity (εr) and to bulk electrical conductivity (σ) for two soils, with corresponding predictive performance assessed by root mean square error (RMSE) and uncertainty summarized with the coefficient of variation (CV). Comparison between model estimated kθ with lab reference gives an overall RMSE = 0.32 mm/min for both εr and σ based functions, whilst Monte Carlo uncertainty propagation yields CV≈2.5–5.4% in the intermediate moisture range and CV≈7.1–8.6% near saturation, indicating that model confidence is highest in drained to partially saturated regimes (0.20 ≤θv ≤ 0.40 cm3/cm3), and declines near saturation (θv> 0.40 cm3/cm3) where thin-film and surface conduction effects emerge. The proposed approach provides a practical pathway to spatially explicit estimation of kθ from time-lapse geophysical data, yet field validation and joint inversion strategies are recommended to improve model transferability.
Zou et al. (Sat,) studied this question.