Measurements of atmospheric dust have long influenced our understanding of dust sources and dust model calibration. However, assessing dust emission magnitude and frequency may reveal different dust source dynamics and is critical for informing land management. Here we use MODIS (500 m) albedo-based daily wind friction estimates to produce a new dust emission climatology of North America (2001–2020), calibrated by the novel use of dust point sources from optical satellite observations (rather than being tuned to dust in the atmosphere). Calibrated dust emission occurred predominantly in the biomes of the Great Plains (GP) and North American Deserts (NAD), in broad agreement with maps of aerosol optical depth and dust deposition but with considerably smaller frequency and magnitude. Combined, these biomes produced 7.2 Tg y-1 with contributions split between biomes (59.8% NAD, 40.2% GP) due to the contrasting conditions. Dust emission is dependent on different wind friction conditions on either side of the Rocky Mountains. In general, across the deserts, aerodynamic roughness was persistently small and dust sources were activated in areas prone to large wind speeds; desert dust emissions were wind speed limited. Across the Great Plains, large winds persist, and dust emission occurred when vegetation cover was reduced; vegetated dust emissions were roughness limited. We found comparable aerodynamic roughness exists across biomes/vegetation classes demonstrating that dust emission areas are not restricted to a single biome, instead they are spread across an ‘envelope’ of conducive wind friction conditions. Wind friction dynamics, describing the interplay between changing vegetation roughness (e.g., due to climate and land management) and changing winds (stilling and its reversal), influence modelled dust emission magnitude and frequency and its current and future climatology. We confirm previous results that in the second half of the 21st century the southern Great Plains is the most vulnerable to increased dust emission and show for the first time that risk is due to increased wind friction (by decreased vegetation roughness and / or increased wind speed). Regardless of how well calibrated models are to atmospheric dust, assuming roughness is static in time and / or homogeneous over space, will not adequately represent current and future dust source dynamics.