Abstract: Emission datasets of nitrogen oxides (NOx) at high horizontal resolutions (e.g., 0.05∘×0.05∘) are crucial for understanding human influences at fine scales, air quality studies, and pollution control. Yet high-resolution emission data are often missing or contain large uncertainties especially for the developing regions. Taking advantage of long-term satellite measurements of nitrogen dioxide (NO2), here we develop a computationally efficient method of estimating NOx emissions in major urban areas at the 0.05∘×0.05∘ resolution. The top-down inversion method accounts for the nonlinear effects of horizontal transport, chemical loss, and deposition. We construct a two-dimensional Peking University High-resolution Lifetime-Emission-Transport model (PHLET), its adjoint model (PHLET-A), and a satellite conversion matrix approach to relate emissions, lifetimes, simulated NO2, and satellite NO2 data. The inversion method is applied to the summer months of 2012–2015 in the Yangtze River Delta (YRD; 29–34∘ N, 118–123∘ E) area, a major polluted region of China, using the NO2 vertical column density data from the Peking University Ozone Monitoring Instrument NO2 product (POMINO). A systematic analysis of inversion errors is performed, including using an independent test based on GEOS-Chem simulations. Across the YRD area, the summer average emissions obtained in this work range from 0 to 15.3 kg km−2 h−1, and the lifetimes (due to chemical loss and deposition) range from 0.6 to 3.3 h. Our emission dataset reveals fine-scale spatial information related to nighttime light, population density, road network, maritime shipping, and land use (from a Google Earth photo). We further compare our emissions with multiple inventories. Many of the fine-scale emission structures are not well represented or not included in the widely used Multi-scale Emissions Inventory of China (MEIC).