Assessing Spatial Variability of Soil Moisture Across an Erosion-Prone Agricultural Hillslope

Soil moisture is a critical state variable in hydrological systems, acting as both an initial and a boundary condition for physically based hydrological models. Its spatial and temporal variability strongly influences the partitioning of rainfall into infiltration, overland flow and subsurface runoff, which regulates the magnitude, timing and threshold behaviour of extreme events such as flash floods and soil degradation. However, the extensive and multiscale variability of soil moisture has challenged hydrological scientists for over two decades. A common approach to address this issue is to perform distributed point sampling of soil moisture and apply geostatistical methods to analyze spatial relationships and patterns, perform interpolations and provide uncertainty estimates for predictions.

In this study, we aim to quantify the spatial variability of soil moisture at the hillslope scale, as this variability is a key factor controlling hydrological responses and erosion dynamics. The research area is an agricultural hillslope in the Attert River Basin, Luxembourg, where severe erosion occurs year-round on agricultural parcels due steep slopes and extreme rainfall events. A nested cluster sampling design was implemented to cover as much area as possible and to represent a wide range of distance classes to perform geostatistical analysis.

Two soil moisture campaigns were conducted under wet and dry conditions. Soil moisture was measured at 110 cluster points using Time Domain Reflectometry (TDR), which records dielectric permittivity and converts it into volumetric water content using general onboard calibration equations, selected according to soil texture. While these factory calibrations are widely used, they can introduce errors when applied to soils with specific hydraulic properties or textures. Therefore, 15 soil samples (3 per cluster) were collected for gravimetric determination of soil moisture to validate the TDR measurements.

During both campaigns, the TDR measurements revealed a negative bias compared to the gravimetric measurements. Empirical variogram models were fitted for both datasets, with and without the data correction for the bias. The wet case, in comparison to the dry case, exhibited a shorter effective range (~145 m) and a higher nugget-to-sill ratio (~0.4), indicating weaker spatial correlation and a larger relative contribution of small-scale variability. In contrast, the dry case showed a longer effective range (~190 m) and a lower nugget-to-sill ratio (~0.3) reflecting stronger spatial organization and more coherent soil moisture patterns. These differences arise because under wet conditions, increased hydraulic connectivity and redistribution promote local-scale variability and reduce large-scale spatial organization. On the other hand, drier conditions enhance the influence of soil texture, rooting depth and evapotranspiration patterns that operate over larger spatial scales.