Characterization of haze episodes and factors contributing to their formation using a panel model.

A haze episode is a complex pollution process with high levels of fine particulate matter smaller than 2.5 μm (PM2.5). Understanding factors contributing to their formation is crucial to mitigate PM2.5 pollution, which varies substantially on the daily and city scales. In this study, we attempted to introduce the dynamic panel model that uses the group deviation method to generate unbiased estimates of contributions from different factors by eliminating time-invariant confounding variables. Taking 25 cities in the Yangtze Delta Region (YDR), China, as a case study and we analyzed how natural factors (e.g., wind) and anthropogenic emissions (e.g., sulfur dioxide (SO2)) together contribute to PM2.5 pollution. Results showed that there was significant lag effect on PM2.5 concentration, and approximately 45% of the PM2.5 remained from one day to the next. On the contrary, present day's emission had little effect on its PM2.5 concentration. It suggested that daily variation of PM2.5 concentration was largely affected by natural factors, while the long term PM2.5 pollution such as annual concentration was more determined by anthropogenic emissions. The unbiased estimates of this simple dynamic panel model could well predict the annual changes of PM2.5 concentration with an uncertainty of less than 2% on city scale. Reducing SO2 and nitrogen oxide (NOx) emissions could mitigate PM2.5 pollution to some extent in the YDR; however, to achieve the clean air standard, more pollutants such as ammonia should be added to the emission reduction list. The analyses provide an alternative method to easily quantify contributing factors and their variability to air pollution. It could be helpful to better understand the confounding factors on the assessment of air pollution governance despite the panel model still need to be improved on aspects such as long-range transportation.