Abstract. Even though desert dust is the most abundant aerosol by mass in Earth's atmosphere, the relative contributions of the world\u2019s major dust source regions to the global dust cycle remain poorly constrained. This problem hinders accounting for the potentially large impact of regional differences in dust properties on clouds, the Earth's energy balance, and terrestrial and marine biogeochemical cycles. Here, we constrain the contribution of each of the world\u2019s main dust source regions to the global dust cycle. We use an analytical framework that integrates an ensemble of global model simulations with observationally informed constraints on the dust size distribution, extinction efficiency, and regional dust aerosol optical depth. We obtain a data set that constrains the relative contribution of each of nine major source regions to size-resolved dust emission, atmospheric loading, optical depth, concentration, and deposition flux. We find that the 22\u201329\u2009Tg (one standard error range) global loading of dust with geometric diameter up to 20\u2009\u03bcm is partitioned as follows: North African source regions contribute ~50\u2009% (11\u201315\u2009Tg), Asian source regions contribute ~40\u2009% (8\u201313\u2009Tg), and North American and Southern Hemisphere regions contribute ~10\u2009% (1.8\u20133.2\u2009Tg). Current models might on average be overestimating the contribution of North African sources to atmospheric dust loading at ~65\u2009%, while underestimating the contribution of Asian dust at ~30\u2009%. However, both our results and current models could be affected by unquantified biases, such as due to errors in separating dust aerosol optical depth from that produced by other aerosol species in remote sensing retrievals in poorly observed desert regions. Our results further show that each source region's dust loading peaks in local spring and summer, which is partially driven by increased dust lifetime in those seasons. We also quantify the dust deposition flux to the Amazon rainforest to be ~10\u2009Tg/year, which is a factor of 2\u20133 less than inferred from satellite data by previous work that likely overestimated dust deposition by underestimating the dust mass extinction efficiency. The data obtained in this paper can be used to obtain improved constraints on dust impacts on clouds, climate, biogeochemical cycles, and other parts of the Earth system.