After a decade of research on improving the description of surface and soil features in desert regions to accurately model mineral dust emissions, we now emphasize the need for deeper evaluating the accuracy of modeled 10‐m surface wind speeds U10. Two mesoscale models, the Lokal‐Modell (LM) and the Regional Atmospheric Modeling System (RAMS), coupled with an explicit dust emission model have previously been used to simulate mineral dust events in the Bodélé region. We compare LM and RAMS U10, together with measurements at the Chicha site (BoDEx campaign) and Faya‐Largeau meteorological station. Surface features and soil schemes are investigated to correctly simulate U10 intensity and diurnal variability. The uncertainties in dust emissions computed with LM and RAMS U10 and different soil databases are estimated. This sensitivity study shows the importance of accurate computation of surface winds to improve the quantification of regional dust emissions from the Bodélé