Airborne desert dust influences radiative transfer, atmospheric chemistry and dynamics, as well as nutrient transport and deposition. It directly and indirectly affects climate on regional and global scales. We present two versions of a parameterization scheme to compute desert dust emissions, incorporated into the atmospheric chemistry general circulation model EMAC (ECHAM5/MESSy2.41 Atmospheric Chemistry). One uses a globally uniform soil particle size distribution, whereas the other explicitly accounts for different soil textures worldwide. We have tested these schemes and investigated the sensitivity to input parameters, using remote sensing data from the Aerosol Robotic Network (AERONET) and dust concentrations and deposition measurements from the AeroCom dust benchmark database (and others). The two schemes are shown to produce similar atmospheric dust loads in the N-African region, while they deviate in the Asian, Middle Eastern and S-American regions. The dust outflow from Africa over the Atlantic Ocean is accurately simulated by both schemes, in magnitude, location and seasonality. The modelled dust concentrations and deposition fluxes compare well with observations at (island) stations in the Atlantic Ocean and Asia, and are underestimated in the Pacific Ocean where annual means are relatively low (<1 μg m<sup>−3</sup>). The two schemes perform similarly well, even though the total annual source differs by ~50%, indicating the importance of transport and deposition processes (being the same for the two schemes). Our results emphasize the need to represent arid regions individually and explicitly in global models according to their unique land characteristics and meteorological conditions.