Ambient air quality forecasting evaluation plays an important role in improving forecasting capability. In order to provide better support for refined air quality management, with reference to the UK air quality forecasting evaluation method, this study divided six air quality index (AQI) levels into 12 half-levels and explored AQI, PM2.5, and O3-8h concentration forecasting evaluation based on the half-level method in "2+26" cities during 2020. Comparison with the AQI range forecasting and AQI level range forecasting evaluation showed that the half-level forecasting evaluation method could combine these two indicators into one, providing feasibility and application value in operational air quality forecasting evaluation. Specific half-level forecasting evaluation showed that the forecasting effect of AQI and O3-8h concentration at the low and high levels was significantly worse than that of the middle levels in "2+26" cities. The forecasting effect of the PM2.5 concentration was relatively stable in different half-levels. The monthly variation curves of AQI, PM2.5, and O3-8h concentration forecasting accuracy exhibited a bimodal pattern, first rising and then falling and a flat pattern, respectively. The overestimate of PM2.5 concentration forecasting was significant in each month. The accuracy gaps of AQI and O3-8h concentration forecasting in different cities was relatively small; however, the PM2.5 concentration forecasting accuracy fluctuated greatly. The AQI forecasting accuracies of Beijing and Tianjin were higher than that of neighboring provinces. Additionally, the PM2.5 and O3-8h concentration forecasting effects in Beijing and Henan province were relatively the best.
Keywords: "2+26" cities; AQI; air quality; forecasting evaluation; half-level.