Pollution source emission inventories are the basis for analyzing the causes of pollution, identifying the contribution of pollution sources, and scientifically formulating air pollution prevention strategies. The current inventory construction methods mainly focus on improving the accuracy and spatial and temporal resolution of the inventory, and there is an urgent need to look into dynamic updating methods to address the problem of lagging source emission inventories. In order to develop an effective and versatile method for the dynamic updating of source emission inventories, a meteorological normalization method based on random forests was chosen to capture the response relationship between pollutants and anthropogenic emissions. Based on the response relationship and the base emission inventory, the method for updating emission inventories based on the relationship between emissions and air quality （REQEI） was developed. Taking MEIC as the example, the REQEI_MEIC inventory was constructed using this method. The applicability and reasonableness of the REQEI_MEIC inventory was examined through the horizontal comparison method and the model validation method, so as to prove the feasibility of the inventory construction method. The results showed that the differences in pollutant emissions between the REQEI_MEIC and MEIC inventories were small, and the model validation effects of the two inventories were generally consistent. In some areas, the emissions of the REQEI_MEIC were closer to the actual emissions, and the pollutant modeling effect was more favorable. The inventory dynamic updating method can construct a relatively accurate pollutant emission inventory on the basis of greatly reducing the workload. The method quickly extrapolates the inventory through the relationship between anthropogenic source emissions and air quality response, accelerates the speed of inventory updating, is suitable for realizing the rapid dynamic updating of existing inventories, and can to a certain extent reduce the limitations imposed by the spatial and temporal limitations of emission inventories on the analysis of the causes of pollution.
Keywords: WRF-Chem model; air pollution; emissions inventory dynamic updating; machine learning; meteorological normalization.