Relative dispersion is a widely used measure to characterize mixing properties of atmospheric passive tracers. Numerical models allow for a large number of particles. In this thesis, the Lagrangian transport model FLEXPART has been used to generate particle trajectories. The advection velocity field was provided by the European Centre for Medium-Range Weather Forecasts (ECMWF). 40000 particles were deployed along latitude lines, at 30N and 30S, 60N and 60S, and at two heights, (6km and 12km). Trajectories on all these locations were made in two different months, January and September 2009. The initial particle separation was determined by the grid spacing for the winds,1o × 1o. In addition to relative dispersion, we examine probability density functions (PDFs) of pair separations. To interpret the results, we also consider the second order (Eulerian) structure functions. Up to a few hundred kilometers, an exponential dispersion regime was detected by the relative dispersion, while the separation distributions (PDFs) indicated that the nonlocal dispersion is likely to extend to the deformation radius. The structure functions are consistent with exponential growth, and suggest the large scale behavior is probably diffusive.