Monitoring and describing the spatiotemporal variability of dust aerosols is crucial to understand their multiple effects, related feedbacks and impacts within the Earth system. This study describes the development of the MIDAS (ModIs Dust AeroSol) dataset. MIDAS provides columnar daily dust optical depth (DOD) at 550 nm at global scale and fine spatial resolution (0.1° x 0.1°) over a 15-year period (2003-2017). This new dataset combines quality filtered satellite aerosol optical depth (AOD) retrievals from MODIS-Aqua at swath level (Collection 6.1, Level 2), along with DOD-to-AOD ratios provided by MERRA-2 reanalysis to derive DOD on the MODIS native grid. The uncertainties of MODIS AOD and MERRA-2 dust fraction with respect to AERONET and LIVAS, respectively, are taken into account for the estimation of the total DOD uncertainty. MERRA-2 dust fractions are in very good agreement with those of LIVAS across the “dust belt”, in the Tropical Atlantic Ocean and the Arabian Sea; the agreement degrades in North America and the Southern Hemisphere where dust sources are smaller. MIDAS, MERRA-2 and LIVAS DODs strongly agree when it comes to annual and seasonal spatial patterns, with collocated global DOD averages of 0.033, 0.031 and 0.029, respectively; however, deviations in dust loading are evident and regionally dependent. Overall, MIDAS is well correlated with AERONET-derived DODs (R=0.89), only showing a small positive bias (0.004 or 2.7%). Among the major dust areas of the planet, the highest R values (> 0.9) are found at sites of N. Africa, Middle East and Asia. MIDAS expands, complements and upgrades existing observational capabilities of dust aerosols and it is suitable for dust climatological studies, model evaluation and data assimilation.