The random walk particle tracking (RWPT) method is compared with one of the Eulerian methods in investigating pollutant transport in depth-averaged flows. As a distinguished representative of Eulerian model with high performance, the MacCormack scheme with Total Variation Diminishing modification (TVD-MacCormack) is selected for comparison. Solute concentration is simulated. First, both numerical models have been tested by two idealized cases and compared against analytical solutions. Numerical dissipation is observed for TVD-Mac model where the concentration changes abruptly, especially under the circumstances of low resolution and disalignment between the flow direction and grid orientation. On the contrary, simulations by the random walk model achieve higher accuracy for both cases and is free of fictitious oscillations in the vicinity of sharp concentration gradients. Then, the solute oscillation along a one-dimensional hypothetical tidal estuary is simulated, showing RWPT accurately conserving mass and suffering less numerical diffusion compared with the Eulearian methods. Finally, the process of pollutant transport at the Yangzte River reach is predicted by RWPT. The longitudinal dispersion coefficient DL is calculated accordingly. It is compared against that by theoretical/empirical formulae and the difference is in acceptable level, indicating the validity of RWPT in solving complex natural problems.