Applying geostatistical electrical resistivity tomography and a water content estimation model for loess spatial mapping

Accurate and rapid quantification of the spatial distribution of loess volumetric water content (θ) is critical for geological disaster mitigation but remains challenging at the scale of our interest. This study aimed to characterize the spatial distribution of θ in a loess slope in Baota District, Yan’an City, using geostatistical electrical resistivity tomography (GERT) combined with laboratory measurements. Thirty undisturbed loess samples were collected from 0 to 500 cm depth to analyze their electrical resistivity (ρ) and θ non-linear responses. Afterwards, the moisture distribution of the slope was investigated using the slope resistivity field derived from GERT data. The results revealed that significant vertical heterogeneity in θ exists within the 0–500 cm loess layer, while dry density increased progressively from 1.37 g·cm⁻³ to 1.54 g·cm⁻³ with depth. Laboratory electrical resistivity measurements identified a critical electrical resistivity threshold (79.4 Ω.m), which marked the transition of pore water from continuous to discontinuous states. A piecewise model for θ estimation was then developed based on this threshold. Compared with traditional global models, the piecewise model reduced θ estimation errors in high ρ regions. The mean absolute error (MAE) decreased by 19.3% and the root mean squared error (RMSE) by 15.5%, partially addressing the underestimation issue. Applying a piecewise model to ρ profiles from GERT yielded θ distributions moderately consistent with borehole data (RMSE = 1.98%), successfully identifying a vertical infiltration interface at 150–200 cm depth. The accuracy of this model outperformed regional empirical models and time domain reflectometry (TDR) interpolation. The proposed method provides a practical solution for geological hazard mitigation in loess slopes.