Consistent SPH simulations of the anisotropic dispersion of a contaminant plume

TL;DR

This paper demonstrates that consistent SPH simulations can accurately model anisotropic contaminant dispersion in porous media, reducing numerical artifacts through increased resolution and neighbor count, but some negative concentrations persist.

Contribution

It introduces a method to restore SPH consistency for anisotropic dispersion simulations by increasing resolution and neighbors, ensuring convergence.

Findings

Full convergence achieved with increased neighbors and resolution

Negative concentrations are reduced but not eliminated

A balance between neighbors and smoothing length is necessary

Abstract

Solute transport through heterogeneous porous media is governed by fuid advection, molecular diffusion and anisotropic dispersion. The dispersion is assumed to obey Fick's law and the dispersion coefficient is defined as a second rank tensor. However, this problem has revealed to be a very difficult one because independently of the numerical methods employed, the solutions are seen to exhibit artificial oscillations and negative concentrations when the dispersivity becomes anisotropic. Here we report consistent SPH simulations of the anisotropic dispersion of a Gaussian contaminant plume in porous media using the open source code DualSPHysics. Consistency of the SPH method is restored by increasing the spatial resolution along with the number of neighbours within the compact support of the interpolating kernel. The solution shows that as the number of neighbours is increased with resolution, full convergence of the numerical solutions is guaranteed regardless of the dispersivity. However, despite the restored consistency, negative concentrations, albeit at a lower level, are still present. This suggests that a compromise between the number of neighbours and the size of the smoothing length must be guaranteed such that sufficient implicit numerical diffusion remains to damp out unphysical oscillations.