Sublimation influences the water storage in snow covers and glaciers, which is important for water use and projections of the sea level rise. Yet, it is challenging to quantify sublimation for large areas or in conditions of snow transport. In-situ measurements only provide local information and can be affected by errors in conditions of snow transport. Therefore, models are crucial. Large-scale models suggest that sublimation of drifting and blowing snow is a relevant term in the surface mass balance of the Antarctic Ice Sheet but the uncertainties are large because the underlying processes are strongly simplified. On the contrary, small-scale models such as large-eddy simulation (LES) with Lagrangian particles represent these processes with a high level of detail and provide unique insights. Indirect information on sublimation can be obtained by measuring stable water isotopes (SWIs) as their relative abundance is influenced by phase changes. Yet, SWIs have so far only provided qualitative insights in the sublimation process because the effects of snow processes on SWIs are still incompletely understood. The goals of this work are to (a) better understand the moisture transport through the near-surface atmosphere in conditions of snow transport using unique measurements and LES simulations, (b) leverage the LES simulations to improve the parameterization approach of large-scale models, and (c) better understand the effect of sublimation and other drivers on SWIs in air masses sampled at the coast of Antarctica. We compare sublimation assessments based on the Monin-Obukhov bulk parameterization and eddy-covariance measurements during a snow transport event at the S17 site, Antarctica. For specific situations, the vertical profile of the vapor and heat fluxes are modeled using LES simulations and a simple one-dimensional model inspired by large-scale models. It is shown that snow transport violates an assumption of the Monin-Obukhov bulk measurements, leading to a significant underestimation of the vapor and heat fluxes in absolute magnitude. More reliable fluxes are obtained with the eddy-covariance technique after removing blowing-snow-induced artifacts from the raw data. To improve large-scale models, we propose to use (a) a high vertical resolution near the surface with at least one grid level in the lowest ~0.1 m of the atmosphere; (b) prognostic profiles of near-surface humidity and temperature; (c) an empirical correction for the sublimation of drifting and blowing snow in the lowest 0.3 m of the atmosphere, which should be further validated in other weather conditions; and (d) increased particle sizes at the top of the saltation layer as long as the height of this layer is underestimated. To explain SWI variations in water vapor, we develop a simple model for the isotopic composition along backward trajectories. Apart from ocean evaporation and isotopic distillation during cloud formation, sublimation of surface snow can be an important driver of the vapor isotopic composition, especially for marine air masses arriving at the ice sheet or free-tropospheric air masses descending towards the surface. Further work is needed to address remaining questions concerning for example the effect of snow transport on SWIs or the impact of improved parameterizations on large-scale estimates of sublimation.