1: \begin{abstract}

2: Solute transport through heterogeneous porous media is governed by fluid advection,

3: molecular diffusion and anisotropic dispersion. The dispersion is assumed to obey

4: Fick's law and the dispersion coefficient is defined as a second rank tensor. 

5: However, this problem has revealed to be a very difficult one because independently

6: of the numerical methods employed, the solutions are seen to exhibit artificial

7: oscillations and negative concentrations when the dispersivity becomes anisotropic.

8: Here we report consistent SPH simulations of the anisotropic dispersion of a

9: Gaussian contaminant plume in porous media using the open source code DualSPHysics.

10: Consistency of the SPH method is restored by increasing the spatial resolution

11: along with the number of neighbours within the compact support of the interpolating

12: kernel. The solution shows that as the number of neighbours is increased with

13: resolution, full convergence of the numerical solutions is guaranteed regardless of

14: the dispersivity. However, despite the restored consistency, negative concentrations,

15: albeit at a lower level, are still present. This suggests that a compromise between

16: the number of neighbours and the size of the smoothing length must be guaranteed

17: such that sufficient implicit numerical diffusion remains to damp out unphysical

18: oscillations.

19: \keywords{Particle methods \and Stability and convergence of numerical methods \and 

20: Advection-diffusion \and Anisotropic dispersion \and Porous media}

21: % \PACS{PACS code1 \and PACS code2 \and more}

22: % \subclass{MSC code1 \and MSC code2 \and more}

23: \end{abstract}

24: