In recent years, China has experienced serious problem of air pollution. Three-dimensional air quality models are widely used in a lot of air quality researches and model performance need to be evaluated and improved. In this study, EOF analysis was used to systematically identify regional model bias of CMAQ. EOF analysis result shows that CMAQ is able to reproduce the annual cycle of PM_2.5, PM₁₀, NO₂, SO₂ and O₃ as well as diurnal cycle of PM_2.5, PM₁₀, NO₂ and O₃ in China. However, CMAQ fails to reproduce diurnal cycle of SO₂ and always leads observational diurnal cycle by 1 hour for other pollutants. EEOF analysis was further conducted to identify the dominant weather system controlling pollutants variation. Lag correction analysis, spatial plot of first two PCs and weather ch...[ Read more ]

In recent years, China has experienced serious problem of air pollution. Three-dimensional air quality models are widely used in a lot of air quality researches and model performance need to be evaluated and improved. In this study, EOF analysis was used to systematically identify regional model bias of CMAQ. EOF analysis result shows that CMAQ is able to reproduce the annual cycle of PM_2.5, PM₁₀, NO₂, SO₂ and O₃ as well as diurnal cycle of PM_2.5, PM₁₀, NO₂ and O₃ in China. However, CMAQ fails to reproduce diurnal cycle of SO₂ and always leads observational diurnal cycle by 1 hour for other pollutants. EEOF analysis was further conducted to identify the dominant weather system controlling pollutants variation. Lag correction analysis, spatial plot of first two PCs and weather charts reveal that the first two modes represent the passage of mid-latitude cyclone, which is the dominant weather system affecting southern China during winter, although the frequencies of the propagation mode are quite different in two winters. As climate change may affect mid-latitude cyclone frequency, its impact on PM_2.5 concentration and possible mechanism were also investigated. In addition, an operational data fusion algorithm has been developed to integrate historical and real-time ground observations with regional model CMAQ and high resolution local model ADMS to improve accuracy of model simulations at both observation time and future 48 hours forecast. For data at observation time, data fusion program improves IOA of %AR from about 0.8 to over 0.96 at all sites. For further 48 hours forecast, IOA of %AR is improved from 0.8 to over 0.9 in first 6 hours forecast and drops to CMAQ forecast skill after 20 hours at general sites, while IOA of %AR is improved from about 0.75 to over 0.9 in first 6 hours forecast and drops to about 0.85 after 20 hours at roadside sites, which is consistently better than CMAQ.