Abstract. China has been conducting a series of actions on air quality improvement for the past decades, and air pollutant emissions have been changing swiftly across the country. Provinces are an important administrative unit for air quality management in China; thus a reliable provincial-level emission inventory for multiple years is essential for detecting the varying sources of pollution and evaluating the effectiveness of emission controls. In this study, we selected Jiangsu, one of the most developed provinces in China, and developed a high-resolution emission inventory of nine species for 2015–2019, with improved methodologies for different emission sectors, best available facility-level information on individual sources, and real-world emission measurements. Resulting from implementation of strict emission control measures, the anthropogenic emissions were estimated to have declined 53 %, 20 %, 7 %, 2 %, 10 %, 21 %, 16 %, 6 %, and 18 % for sulfur dioxide (SO2), nitrogen oxides (NOx), carbon monoxide (CO), non-methane volatile organic compounds (NMVOCs), ammonia (NH3), inhalable particulate matter (PM10), fine particulate matter (PM2.5), black carbon (BC), and organic carbon (OC) from 2015 to 2019, respectively. Larger abatement of SO2, NOx, and PM2.5 emissions was detected for the more developed region of southern Jiangsu. During the period from 2016 to 2019, the ratio of biogenic volatile organic compounds (BVOCs) to anthropogenic volatile organic compounds (AVOCs) exceeded 50 % in the month of July, indicating the importance of biogenic sources for summer O3 formation. Our estimates in annual emissions of NOx, NMVOCs, and NH3 were generally smaller than the national emission inventory, MEIC (the Multi-resolution Emission Inventory for China), but larger for primary particles. The discrepancies between studies resulted mainly from different methods of emission estimation (e.g., the procedure-based approach for AVOC emissions from key industries used in this work) and inconsistent information of emission source operation (e.g., the penetration and removal efficiencies of air pollution control devices). Regarding the different periods, more reduction of SO2 emissions was found between 2015 and 2017 and of NOx, AVOCs, and PM2.5 between 2017 and 2019. Among the selected 13 major measures, the ultra-low-emission retrofit in the power sector was the most important contributor to the reduced SO2 and NOx emissions (accounting for 38 % and 43 % of the emission abatement, respectively) for 2015–2017, but its effect became very limited afterwards as the retrofit had been commonly completed by 2017. Instead, extensive management of coal-fired boilers and the upgrade and renovation of non-electrical industry were the most important measures for 2017–2019, accounting collectively for 61 %, 49 %, and 57 % reduction of SO2, NOx, and PM2.5, respectively. Controls on key industrial sectors were the most effective for AVOC reduction in the two periods, while measures relating to other sources (transportation and solvent replacement) have become more important in recent years. Our provincial emission inventory was demonstrated to support high-resolution air quality modeling for multiple years. Through scenario setting and modeling, worsened meteorological conditions were found from 2015 to 2019 for PM2.5 and O3 pollution alleviation. However, the efforts on emission controls were identified to largely overcome the negative influence of meteorological variation. The changed anthropogenic emissions were estimated to contribute 4.3 and 5.5 µg m−3 of PM2.5 concentration reduction for 2015–2017 and 2017–2019, respectively. While O3 was elevated by 4.9 µg m−3 for 2015–2017, the changing emissions led to 3.1 µg m−3 of reduction for 2017–2019, partly (not fully though) offsetting the meteorology-driven growth. The analysis justified the validity of local emission control efforts on air quality improvement and provided a scientific basis to formulate air pollution prevention and control policies for other developed regions in China and worldwide.