Evaluation and Diagnosis of Regional Ammonia Emission Inventory in the Pearl River Delta Using Multisite NH₃ Observations and Model Simulations

Ammonia (NH3) has attracted increasing attention for its reduction potential in fine particulate matter mitigation, yet current NH3 emission inventories involve substantial uncertainties. Previous bottom-up NH3 inventories are usually constrained by satellite observations, deposition measurements, or isotopic analysis and still lack careful validation at fine regional scales. This study develops a novel diagnostic framework combining multisite NH3 observations across the Pearl River Delta (PRD) with the Community Multiscale Air Quality (CMAQ) model simulations and machine learning techniques to evaluate and refine a regional NH3 inventory. Our analysis indicates that the inventory overestimates agricultural emissions, particularly during the wet period, while underestimating nonagricultural sources. Underrepresented precipitation effects were a key driver of overestimated agricultural emissions (∼19% during the wet period). Conversely, a natural experiment during the Spring Festival holiday provided strong evidence that vehicle emissions are a key underestimated nonagricultural source. Adjusting the inventory based on these findings (agricultural sources reduced 39% (31%) during wet (dry) periods, nonagricultural sources increased 70%) improved NH3(g) simulations across the PRD. Our study highlights the value of multisite observations in validating NH3 inventory and the critical need to better characterize underestimated (e.g., vehicles) and missing sources (e.g., urban landscaping) in current inventories.