We present and analyze a simpli ed, scale-separated, anelastic uid model which is designed to assess the in uence of weak compressibility in the di usive transport of a passive scalar. Our model incorporates a slowly varying, density induced, anisotropy into a xed, rapidly varying (small scale) uid ow. This anisotropy is physically motivated through the anelastic mass balance which retains vertical density variations occurring over the atmospheric scale height (8 km). Consequently, these steady ows are non-solenoidal over large scales and approximately incompressible on small scales. We apply homogenization methods to calculate the large scale, e ective mixing experienced by a passive scalar di using in the presence of this small scale ow. Over large scales, the evolution of the scalar eld is governed by an e ective, variable coe cient, anelastic mixing equation. The variable coe cients entering this equation are shown to depend non-trivially upon both the large scale anisotropy as well as the structure of the small scale uid ow. We establish that anelastic e ects produce anisotropic mixing properties not shared by the analogous incompressible closure. Speci cally, the mixing equation possesses exact non-trivial bounded steady states, whereas the incompressible regime has only constant (i.e., spatially homogeneous) steady states. Furthermore, the anelastic mixing equation is shown numerically to possess local regions of trapped (Figure 3) and focused (Figures 6,7) contaminants, behavior not possible in the analogous incompressible model. These results imply that anelastic e ects, which occur naturally in the atmosphere, provide mechanisms which locally reduce mixing and generate inhomogeneities in large scale concentration elds.