Many boundary layer processes simulated within a Mars General Circulation Model (MGCM), including the description of the processes controlling dust rising from the Martian surface, are highly sensitive to the aerodynamic roughness length z0. On the basis of rock-size frequency distributions inferred from different Martian landing sites and Earth analog sites, we have first established that lognormal-modeled rock-size frequency distributions are able to reproduce correctly the observed Martian rock populations. We have validated the hypothesis that the rock abundance ζ of a given area could be estimated at a first order from its thermophysical properties, namely its thermal inertia I and its albedo α. We have demonstrated the possibility of using rock abundance ζ to estimate the roughness density λ on Mars and to retrieve subsequently the aerodynamic roughness length by using semi-empirical relationships based on terrestrial wind-tunnel and field measurements. By combining our methodology with remote sensing measurements of the Thermal Emission Spectrometer aboard Mars Global Surveyor, we have derived a global map of the aeolian aerodynamic roughness length with a 1/8° × 1/8° resolution over the entire Martian surface. Contrary to what is often assumed, the Martian aeolian aerodynamic roughness length is spatially highly heterogeneous. At the fullest resolution, the Martian aerodynamic roughness length varies from 10−3 cm to 2.33 cm. About 84% of the Martian surface seems to be characterized by an aeolian aerodynamic roughness length value lower than 1 cm, the spatially uniform value that most of the MGCMs simulations have assumed recently. Since the aerodynamic roughness length z0 is a key parameter in deriving the erosion threshold wind velocities, we anticipate a significant impact of our findings on the efficiencies for lifting dust in future MGCMs.< Réduire