Paddy fields in China contributed to one third of the global cropland ammonia (NH3) emission inventory, while rice accounted for half of cereal consumption, necessitating exhaustive considerations of the balance between NH3 emissions abatement and food demand. The concept of yield-scaled emission intensity (emissions per unit crop production) has the potential to guide sustainable intensification strategies, yet its application to NH3 emissions remains poorly understood. Here, by constructing novel crop-specific models for single rice production and NH3 emissions in the Middle and Lower Yangtze River Basin (LYRB) as a case study, the relationships between fertilizer N application and yield-scaled NH3 were estimated. Contrary to our hypothesis of a tipping point, our results showed that yield-scaled NH3 curves could not directly identify optimal nitrogen (N) rates. However, the benefit of lower N fertilizer rate on NH3 abatement consistently outweighed the risk of yield loss. The exponential relationships between yield-scaled NH3 and N surplus allowed us to estimate the N surplus criterion as 15.6 kg N ha−1 (or 190 kg N ha−1 fertilizer N rate) for the LYRB. Under the N surplus criterion, NH3 emissions can be reduced by 23–27% without severely impacting rice yield, compared to the N rate required for the highest yield. Moreover, five major controlling factors for yield-scaled NH3 were estimated by random forest models, ranked in order of importance as N rate, total N, K rate, mean annual precipitation, and soil organic carbon. Among the agricultural practices (irrigation, tillage, and fertilizer management), deep placement was the most effective measure to reduce yield-scaled NH3, showing 48% reduction potential, followed by proper N splitting frequency (43%). Overall, this study highlights the efficacy of N application optimization and targeted farm management in mitigating NH3 emission while maintaining crop productivity.