A particle based model for soil water dynamics: how to match and step beyond Richards' equation?
Abstract. Within this study we propose a stochastic approach to simulate soil water dynamics in the unsaturated zone by using a non-linear, space domain random walk of water particles. Soil water is represented by particles of constant mass, which travel according to the Itô form of the Fokker Planck equation. The model concept builds on established soil physics by estimating the drift velocity and the diffusion term based on the soil water characteristics. A naive random walk, which assumes all water particles to move at the same drift velocity and diffusivity, overestimated depletion of soil moisture gradients compared to a Richards' solver. This is because soil water and hence the corresponding water particles in smaller pore size fractions, are, due to the non-linear decrease of soil hydraulic conductivity with decreasing soil moisture, much less mobile. After accounting for this subscale variability of particle mobility, the particle model and a Richards' solver performed similarly during simulated wetting and drying circles in three distinctly different soils. The particle model typically produced slightly smaller top soil water contents during wetting and was faster in depleting soil moisture gradients during subsequent drainage phases. Within a real world benchmark the particle model matched observed soil moisture response to a moderated rainfall event even slightly better than the Richards' solver. The proposed approach is hence a promising, easy to implement alternative to the Richards equation. This is particularly also because it allows one to step beyond the assumption of local equilibrium during rainfall driven conditions. This is demonstrated by treating infiltrating event water particles as different type of particle which travel initially, mainly gravity driven, in the largest pore fraction at maximum velocity, and yet experience a slow diffusive mixing with the pre-event water particles within a characteristic mixing time.