The physics of the two processes (saltation and sandblasting) leading to fine mineral dust emissions by wind erosion in arid or semi-arid areas has been detailed and modeled. The combination of these two models has led to a physically explicit Dust Production Model (DPM). In this work, sensitivity tests are performed with the DPM to determine the nature of the main soil parameters that control dust emissions by sandblasting. It is found that the soil roughness length and the dry size distribution of the soil aggregates constituting the loose wind erodible fraction of the topsoil have the greatest influence on the soil potential for mineral dust production. Contrary to what is often assumed, soil texture is not a relevant parameter. In the light of these new findings, results of vertical flux measurements performed over a wide variety of sources in Niger and the US south west (14 soils) have been reanalyzed. Results show (1) that for the tested soils the DPM, and hence sandblasting, explain all dust emissions, and (2) that 13 of the 14 soils that had been selected a priori for their high potential for dust emissions contained a fine soil-aggregate component. This is consistent with the sensitivity tests indicating that the presence of such a component could enhance dust emissions by one order of magnitude. Finally, it can be concluded that most of the apparent scatter in the experimental results was in large part due to an inappropriate choice of soil parameters to interpret them.