Parameter estimation of preferential water flow in soil using particle swarm optimization inverse method: Comparison of kinematic–dispersive wave (KDW) and KDW–van Genuchten (KDW-VG) models

Abstract

Swift preferential water flow through macropores can rapidly pollute groundwater, spreading agricultural and industrial contaminants and threatening water security and ecosystems. To improve the simulation of pollutant transport in soil, a software package based on the kinematic–dispersive wave van Genuchten (KDW-VG) model combined with the particle swarm optimization (PSO) method was used to simulate preferential water flow through an unsaturated soil matrix. The KDW-VG model evolved from the KDW model, replacing the KDW model’s power-law function with the more physically robust Mualem–van Genuchten framework. However, existing models often require detailed measurements of water flux versus mobile water content, which limits their applicability under field conditions. In this research, observed data from four rainfall intensities from 55.58 to 160.49 ($\text{mm}{h}^{-1}$), were used to calibrate both the KDW and KDW-VG models. The hydrographs from a soil column with artificial macropores were recorded to calibrate both models. Using the PSO inverse method, unknown parameters were determined by minimizing the error between observed and simulated hydrographs. The finite-difference technique was used to solve both models. The results showed that the KDW-VG model fit the observations more closely, because of the replacement of the power-law function with the Mualem–van Genuchten framework. The dispersive effect was higher at lower rainfall intensities. Overall, the KDW-VG model's parameters exhibited less sensitivity to rainfall variations, which is a key advantage. This research advances computational techniques for modelling mass transfer in environmental systems, specifically addressing preferential water flow and pollutant transport. By improving the accuracy of pollutant transport models while requiring less detailed input data, the method can be applied under field conditions to provide more reliable predictions. Future work will test the model under field conditions, extend it to varied soils, and integrate realistic macropores using advanced imaging and computation.