Watershed hydrological processes controlled by subsurface structures that have hierarchical organization across scales, but there is a lack in multiscale model validation. In this study, using a comprehensive dataset collected in the forested Shale Hills catchment, we tested the series HYDRUS codes (i.e., HYDRUS-1D at the pedon scale, HYDRUS-2D at the hillslope scale, and HYDRUS-3D at the catchment scale) that included a hierarchical multi-dimensional modeling approach for water flow simulation in the vadose zone. There is good agreement between 1D simulations and measurements of soil moisture profiles controlled by soil hydraulic parameters and precipitation characteristics; however, short-term fluctuations in preferential flow were poorly captured. Notably, 2D and 3D simulations (Nash–Sutcliffe efficiency, ), which accounting subsurface preferential flow controlled by slope positions and shallow fractured bedrock, provided better results than 1D simulations (). Our modeling approach also illustrated that the studied watershed was characterized by weathered and un-weathered fracture bedrocks, which routed water through a network of subsurface preferential flow pathways to the first-order stream. Furthermore, a dual-porosity or anisotropy model produced more accurate predictions than a single-porosity or isotropy model due to a more realistic representation of local soil characteristics and layered structure. Our multi-dimensional modeling approaches credited with diagnosing and presenting the dominant hydrological processes and the interactions within soil-landscape features across one sloped catchment.

Preferential flow processes are controlled by subsurface structures with the hierarchical organization across scales, but there is a lack of multiscale model validation using the field data. In this study, using a comprehensive dataset collected in the forested Shale Hills catchment, we tested and validated preferential flow occurrence by 2-dimension HYDRUS-2D at the hillslope scale, and in comparison, with 1-dimension HYDRUS-1D at the pedon scale and 3-dimension HYDRUS-3D at the catchment scale. There was good agreement between the 1D simulations and measurements of soil moisture in the soil profile, which was mainly affected by the vertical change in porosity/permeability with depth and precipitation characteristics; however, short-term fluctuations due to preferential flow were poorly captured. Notably, 2D and 3D simulations, accounting for preferential flow controlled by slope positions and shallow fractured bedrock, provided better results than the 1D simulations. Furthermore, a dual-porosity or anisotropic model provided more accurate predictions of soil moisture than a single-porosity or isotropic model due to a more realistic representation of local soil and fractured shale structure, which is also the premise of preferential flow (PF) occurrence. Consequently, our study reflected the central importance of multi-dimensional model approaches while highlighting the quantification of the soil structure and fractured nature of the bedrocks itself is essential to the simulation of preferential flow. The multi-dimensional modeling approaches can provide the mechanic presentation of PF pathways to the first-order stream and the necessity of the 3D simulation with detailed information to identify the dominant hydrological process.

Modeling and prediction of soil hydrologic processes require the identification of soil moisture spatial-temporal patterns and effective methods allowing the data observations to be used across different spatial and temporal scales. This work presents a methodology for the combination of spatially- and temporally-extensive soil moisture data obtained in the Shale Hills Critical Zone Observatory (CZO) from 2004 to 2010. The soil moisture data sets were decomposed into spatial Empirical Orthogonal Function (EOF) patterns, and their relationship with various geophysical parameters was examined to determine the dominant factors contributing to the profiled soil moisture variability. The EOF analyses indicated that one or two EOFs of soil moisture could explain 76-89% of data variation. The primary EOF pattern had high values clustered in the valley region, and conversely low values located in the sloped hills. We suggest a novel approach to integrate the spatially-extensive manually measured datasets with the temporally-extensive automated monitored datasets based on the EOF analyses. Given the data accessibility, the current data merging framework has provided the methodology for the coupling of the mapped and monitored soil moisture datasets, as well as the conceptual coupling of slow and fast pedologic and hydrologic functions. This successful coupling implies that a combination of different extensive moisture data has provided interesting insights into our understanding of hydrological processes at multiple scales.