Nitrogen (N) additions to cropland soils are the largest source of anthropogenic nitrous oxide (N2O) emissions and are an important contributor to global greenhouse gas radiative forcing. Progress in understanding controls on N2O fluxes from soils is demonstrated in increasingly sophisticated emissions estimates with improved spatial and source resolution. These methods build upon ongoing field, laboratory, and modeling advances that are restricted to just a handful of countries. Thus, burgeoning new knowledge is of limited utility for improving N2O emissions estimates for the rest of the world where prospects for near-term advances are constrained by the limited breadth of observations and availability of model driver data. Here, we use Bayesian inversion to leverage information from recent national-level N2O emission inventories and reduce uncertainty by up to 65% for estimates of regional and global direct cropland N2O emissions. Our estimates for the proportion of N inputs lost as N2O vary by a factor of two between regions and depart from existing default emission factors, yet regional emissions estimates based on these factors are consistent with global, regional, and local observations. Improved regional emission factors will enhance national greenhouse gas inventories in information-poor countries and guide efforts to reduce agricultural N2O emissions.