Lagrangian particle models have recently become a very important tool for studying air pollution dispersion. Recent advances in computer technology make it possible to link Lagrangian particle dispersion models to numerical mesoscale meteorological models suitable to simulate atmospheric flow in complex terrain. However, the principal difficulty with the Lagrangian approach to problem of turbulent diffusion is that relevant turbulent characteristics are difficult to measure and parameterize. In the present study, Lagrangian turbulent statistics including velocity autocorrelations and Lagrangian integral time scales are derived from Large Eddy Simulations (LES). RAMS (Regional Atmospheric Modeling System) in a LES formulation is used to perform simulations of the convective boundary layer over homogenous terrain. Lagrangian particle model is called on each time step of the RAMS simulation. Tracer particles are being released continuously from randomly selected points within the modeling domain. Finally, both Lagrangian and Eulerian statistics are derived from the stored particle simulation output. The obtained results for the Lagrangian time scales are compared with different formulation used in applied Lagrangian particle dispersion models. Currently, the above LES simulations and particle analysis are being extended for atmospheric boundary layer over nonhomogenous (patchy) terrain