Preferential flow is common in clay or expansive clay soils, involving water bypassing a large portion of the soil matrix. Dye tracer experiment and numerical modeling are used to simulate the surface runoff and subsurface preferential flow patterns influenced by the soil fracture network of a relatively steep hillslope system (slope angle equals to 10 degrees). The result of the experiments indicates that part of the water is infiltrated through cracks, leading to the delay of the initial runoff-yielding time and reduction of the discharge of the surface runoff. The soil water flow is mainly in the matrix when the intensity of precipitation is low. With the increasing of precipitation, soil water movement may become in the form of preferential flow through cracks. In addition, the nonuniformity of soil water infiltration and the depth of the average water infiltration increase as the precipitation intensity increases. To this end, the complete coupling model was established by using the surface-matrix-crack (SMC) model to simulate water flow within discrete fracture as well as to simulate water flow in the soil matrix based on the concept of dual permeability using the traditional Richards’ equation. In this model, the “cubic law” of fluid motion in cracks within smooth parallel plates and the two-dimensional diffusion wave approximation to Saint-Venant equations with momentum term ignored (two-dimensional shallow water equations) were used. The model divides soil water infiltration into two forms and uses the overall method to calculate the exchange of water between the crack networks and matrix regions as well as the exchange water between surface runoff and infiltration water. Results indicate that the SMC model has better performance compared with the traditional equivalent continuum model when those models are used to simulate the surface runoff movement and the soil water movement in the presence of cracks.

Soil salinization limits agricultural production and the management of soil and water resources in arid irrigated regions. Regional prediction remains difficult, while soil salinity is influenced by hydrology, topography, and land cover. This study developed an explainable automated machine learning (AutoML) framework to map surface soil salinity in the Qingtongxia Irrigation District, Ningxia, China. The analysis combined 108 soil samples (0–10 cm), collected in March-April 2024, with 36 candidate input features from Sentinel-2 spectral indices, topography, soil texture, groundwater depth, and climate. Pearson correlation analysis and recursive feature elimination with cross-validation (RFECV) selected 10 key input features. Among the tested models, AutoML achieved the best validation performance, with R² = 0.78, MAE = 1.94 g/kg, and RMSE = 2.65 g/kg. The resulting 30 m prediction map captured broad regional patterns of soil salinity, with predicted values from 0.34 to 22.91 g/kg and higher salinity in northern and northeastern areas. Shapley additive explanations (SHAP) analysis highlighted brightness index, groundwater depth, clay content, elevation, and salinity index I as influential features. These findings suggest that explainable AutoML can help identify regional salinity risk and guide future sampling and model refinement.