It may be thought that treebelts acting as windbreaks will modify the process of atmospheric diffusion and, consequently, the conditions of localization and intensity of the deposition of pollutants emitted into the atmosphere. That action may be strictly local in the case of a single windbreak, or extended to a regional scale by using networks of windbreaks, which create new roughness elements in the landscape. Consequences of that modification of land surface roughness are studied theoretically. Analytical developments are simplified by using restrictive hypotheses, relative to the definition of the concerned surface (which is supposed to be homogeneous on a sufficiently large extent, both upwind and downwind of the emission source) and the adoption of near-neutral atmospheric stability conditions. The high regional roughness so created tends to increase the turbulent transfer of momentum and, following the dispersive power of the atmosphere. The resulting effect is quantified in two independent ways, using separately the eddy diffusivity method and the statistical theory method. Results are in good agreement for a common case (continuous cross-wind infinite line source at ground level). For the case of elevated sources, modified expressions of Sutton's generalized diffusion coefficients are established with the following form ▪ The increase in roughness tends to bring the point of maximum concentration nearer the foot of stack, and the extent of the zone of high surface concentrations is reduced. These theoretical findings are expected to be valid for high roughness regions only when the restrictive hypotheses previously defined are verified. The extension to a more general case is merely outlined in the text. This study just takes into account the effect of treebelts for a regional scale. In fact, near the ground, the local influence of each windbreak would have to be superimposed to this regional influence. Ground concentrations defined in this study are only hypothetical values, from which real ground concentrations could be determined by considering local effects as defined by the experimental study of Caput et al. (1972).