Current global aerosol-climate models typically have relatively coarse resolution, thus do not explicitly represent phenomena occurring at spatial scales smaller than about 100 km. The subgrid variability of surface winds is often neglected, which can lead to resolution-dependence of simulated dust and sea salt emissions, and consequently influence the model estimate of anthropogenic aerosol effect. It is thus important to quantify the impact of subgrid variability of surface winds on sea salt and dust emissions and parameterize this effect. In this study we analyzed high-resolution meteorological fields from global operational analysis at approximately 15 km resolution, and from regional model simulations with horizontal grid size down to 1km. Using these data, we characterized the spatial variability of surface wind within typical global model grid, and analyzed relationships between the wind speed variability and various factors such as the grid-mean turbulence kinetic energy, dry convective eddies, moist convective eddies, and surface elevation inhomogeneity. It was found that the contribution from moist convective eddies dominates the sub-grid variability of surface wind speeds over the oceans. Over land, subgrid variation of surface elevation also plays an important role. A parameterization based on above findings was developed and tested in CAM5. The impact on natural aerosol emissions was also evaluated.