Dissecting the spatio-temporal variability of soil hydraulic properties in an agricultural eroded area

The combined effect of anthropogenic and climatic stressors deeply influences the hydrological behavior of agricultural areas, especially on hillslopes. Tillage induces an abrupt change in the soil's hydraulic functioning, which can be dynamically recovered in time due to natural consolidation, alternation of wetting and drying cycles, and other biophysical factors. Heavy rainfall can accelerate the recovery process, but also induce erosion events in tilled soils, further exacerbating the spatial variability of the topsoil hydraulic properties. To better understand the mechanisms driving the spatio-temporal variability of soil hydraulic properties in agricultural areas, we combine the modified hydrological model HYDRUS with transient soil moisture observations from two hillslopes in the Czech Republic exposed to tillage and erosion. In particular, the Bayesian inference is used to calibrate two alternative HYDRUS implementations at five different locations along the hillslopes. The first model assumes static soil hydraulic properties, while the second simulates their dynamic change induced by tillage and natural consolidation (due to rainfall). The Watanabe-Akaike Information Criterion (WAIC) is used to compare the two models by considering not only the fitting accuracy, but also the predictive uncertainty. The results show that both models can reproduce soil moisture observations satisfactorily at different depths and locations. While the dynamic model exhibits slightly better fitting, this is compensated by larger predictive uncertainty compared to the static model. This is confirmed by the WAIC values, which are similar for the two models. An in-depth analysis indicates that the dynamic recovery of soil hydraulic properties happens during the first few rainfall events (confirming what was observed in other studies) and suggests that higher resolution measurements are needed to better estimate recovery factors. Finally, the spatial variability of the estimated soil hydraulic parameters hints at a possible role of overland flow fluxes along the hillslope as a heterogeneity-generating factor.