抄録
The convection-dispersion equation (CDE) is widely used to predict solute transport in soils. The mechanical dispersion is described with the mathematically identical equation to the Fick’s law for molecular diffusion. Physically unrealistic backward solute mixing may occur for the CDE when a high concentration gradient exists in a soil. We propose a convective random-walk model (CRWM) which only allow solute particles to move in the convective direction. Solute particles move randomly according to an asymmetric probability density function (pdf) having identical mean and variance to the normal Gaussian pdf for the CDE. The CRWM is firstly applied to a solute leaching for a Dirac delta initial distribution. The CRWM can avoid backward mixing as was observed for the CDE during the early stages of leaching. The solute distribution for the CRWM converges to the distribution for the CDE regardless of the shape of the pdf for sufficiently large travel distances because of the central limit theorem. A stochastic input method for the CRWM assuming an additional hypothetical soil outside of the boundary is employed to describe solute input at the boundary in order not to overestimate the concentration near the soil surface. We then apply the CRWM to predict a salt accumulation process on the surface in a soil column having a shallow salty water table. All the flow and transport parameter values are independently determined. Molecular diffusion and salt crystallization at the surface are included for the model prediction. A good agreement between measured and predicted solute profiles is observed for the CRWM, whereas the CDE overestimates downward solute movement due to backward mixing.
