This study evaluates the spatiotemporal variability of dust emission in the Arabian Peninsula and quantifies the emission sensitivity to the land-cover heterogeneity by using the Community Land Model version 4 (CLM43) at three different spatial resolutions. The land-cover heterogeneity is represented by the CLM4-default plant function types (PFTs) and the Moderate Resolution Imaging Spectroradiometer (MODIS) land cover types, respectively, at different grids. We area-average surface vegetation data and use the default nearest neighbor method to interpolate meteorological variables. We find that using MODIS data leads to a slightly higher coverage of vegetated land than the default PFT data; the former also gives more dust emission than the latter at 25- and 50-km grids as the default PFT data have more gridcells favoring less dust emission. The research highlights the importance of using proper data-processing methods or dust emission thresholds to preserve the dust emission accuracy in land models. © 2016 Elsevier Ltd.