Numerical simulations are performed to investigate quantitatively density effects of a contaminant plume on its movement in time and space. The numerical model used is a characteristics/finite difference method. The density of the solute is described by an equation of state similar to sea water or lead nitrate. The results of a large number of simulations do not allow us to establish a general empirical law for predicting the fate of a density-dependent plume. Whether density effects are important or not depends on a variety of hydraulic parameters of the aquifer model, such as the permeability (isotropic or anisotropic), flow gradient, the hydrodynamic dispersivity, and the time-scale considered. Following some theoretical ideas of previously derived analytical scaling laws for buoyant plumes, the concept of a near-source and a far-source region of the plume has been found very useful. The modeled plume centerlines compare favorably with the predictions of these scaling laws in the intermediate-source region, but deviate significantly in the far-source region where diffusion processes become important or necking instabilities at the plume edges occur. It is found that the combined influence of the various model parameters above can be very different, depending on whether the plume is still within the near-source or has already extended into the far-source region. In spite of the large dependency of the density effects on these model parameters, in some cases variable density might have an effect on the plume migration for already moderate density contrasts of~0.3% (which corresponds to a concentration of~4000 ppm for a leachate of salt or lead nitrate). Hydrodynamic dispersion tends to reduce the density effects. For sufficiently small values of the dispersivity, convective instabilities and fingering phenomena occur at the horizontal plume boundaries. For a lighter-than-water contaminant, such as acetone or methylethylketone, significant buoyant movements occur only for (unrealistically) high concentration of ~ 20,000 ppm. In conclusion the results of the simulations do not provide a generally applicable guideline for the evaluation of density-dependence transport. Therefore, the individual numerical modeling of a particular site will often be indispensable.