Through its important role in the formation of particulate matter, atmospheric ammonia affects air quality and has implications for human health and life expectancy 1,2. Excess ammonia in the environment also contributes to the acidification and eutrophication of ecosystems 3-5 and to climate change 6. Anthropogenic emissions dominate natural ones and mostly originate from agricultural, domestic and industrial activities 7. However, the total ammonia budget and the attribution of emissions to specific sources remain highly uncertain across different spatial scales 7-9. Here we identify, categorize and quantify the world's ammonia emission hotspots using a high-resolution map of atmospheric ammonia obtained from almost a decade of daily IASI satellite observations. We report 248 hotspots with diameters smaller than 50 kilometres, which we associate with either a single point source or a cluster of agricultural and industrial point sources-with the exception of one hotspot, which can be traced back to a natural source. The state-of-the-art EDGAR emission inventory 10 mostly agrees with satellitederived emission fluxes within a factor of three for larger regions. However, it does not adequately represent the majority of point sources that we identified and underestimates the emissions of two-thirds of them by at least one order of magnitude. Industrial emitters in particular are often found to be displaced or missing. Our results suggest that it is necessary to completely revisit the emission inventories of anthropogenic ammonia sources and to account for the rapid evolution of such sources over time. This will lead to better health and environmental impact assessments of atmospheric ammonia and the implementation of suitable nitrogen management strategies. Considerable effort goes into establishing spatially and temporally resolved ammonia (NH 3) bottom-up emission inventories, as these are critical drivers of models that are used to assess NH 3 distributions and impacts on the environment. Bottom-up inventories are built from activity data coupled with estimated emission factors. The correctness of these input data is their Achilles' heel, as activity data can be absent or outdated and estimated emission factors are based on specific case studies and may not be representative of either local or global conditions 11,12. When they are available, global measured atmospheric distributions of trace gas concentrations allow us to retrieve source emissions. In the past few years, satellite sounders have offered (bi) daily global NH 3 measurements 13-17 , which have a huge potential to improve our knowledge of the NH 3 emission budget 18. The first global distributions 13 and inverse modelling efforts 19 have confirmed the correctness of the location of the large source regions in the inventories, but have also revealed likely underestimates in the magnitude of their emissions, especially in the Northern Hemisphere. A regional study 20 has highlighted the advantage of averaging data to reveal smaller, localized NH 3 point...