Abstract: Three years of blowing snow and meteorological observations have been collected along a 7 m mast at site D17 in coastal Adélie Land, Antarctica. This is a region particularly exposed to katabatic winds. The atmospheric surface layer is often close to saturation because of the sublimation of the airborne snow particles. A systematic dry bias results in atmospheric models that ignore blowing snow and its moistening effects, and in meteorological analyses that use such model. The Crocus snow-pack model, including a parameterization for the erosion of surface snow by wind, reproduces the observed march of snow accumulation and ablation if the observed meteorology is used as input. Because of subsaturation, a 2.5 fold increase in surface sublimation is obtained if analyzed surface air meteorology is used. The sublimation obtained in the Crocus model poorly agrees with the moisture fluxes evaluated using the profile method along the mast. Moisture gradients are very weak, particularly when blowing snow saturates the air, to a point where measurement accuracy is an issue. Using the profile method, the measurement uncertainties are strongly amplified in case of strong wind. In such conditions, a single level bulk parameterization with surface energy balance closure as in the Crocus model is preferred. At D17, more than half of the total snow fall is removed by erosion and sublimation, both at the surface and, mainly, of airborne snow particles.