Improved nuclear magnetic resonance-based Green-Ampt infiltration model incorporating dynamic permeability of clay

The traditional Green-Ampt model does not accurately represent the infiltration behavior of clay soils. Infiltration in clay is influenced by low hydraulic conductivity, strong capillary forces, and a gradual transition zone between saturated and unsaturated zones. These factors often lead to overestimated infiltration rates and underestimated infiltration durations. Therefore, it is necessary to improve the model to better reflect the characteristics of clay infiltration and enhance its predictive accuracy and practical applicability. This study conducts hydraulic characterization tests, one-dimensional soil column rainfall infiltration experiments, and numerical analysis on a representative clay sampled from Wuhan, China, to investigate infiltration behaviors under varying rainfall intensities and initial moisture conditions. The study reveals that the proportion of the transition layer within the wetting layer decreases with increasing wetting front depth, following a power-law function. Under the same initial moisture content, this proportion tends to converge to a stable value regardless of rainfall intensity. In contrast, under the same rainfall intensity, a higher initial moisture content leads to a larger proportion of the transition layer at a given wetting front depth. Based on the NMR curve, the unsaturated permeability coefficients corresponding to different volumetric water contents of clay can be obtained quickly, accurately, and at low cost. By utilizing the unsaturated permeability coefficient prediction model based on the nuclear magnetic resonance (NMR) curve, the study refines the computational method for the equivalent permeability coefficient in the wetting layer during clay rainfall infiltration, and proposes an improved clay Green-Ampt infiltration model that considers the saturated-unsaturated differentiation layer and the dynamic variation of its equivalent permeability coefficient under continuous rainfall conditions. The computational results of the improved model were compared with measured infiltration data, numerical simulations, and predictions from the traditional GA model. The results indicate that the improved model effectively captures the dynamic variation between the transition layer and wetting layer and provides more accurate predictions of wetting front depth in clay, with an accuracy approximately 68.36% higher than that of the traditional GA model.