Percolation analysis of urban air quality: A case in China

• Published in: Physica A 2020-03, Vol.541, p.123312, Article 123312
• Main Authors: Du, Ruijin, Li, Jingjing, Dong, Gaogao, Tian, Lixin, Qing, Ting, Fang, Guochang, Dong, Yujuan
• Format: Article
• Language: English
• Subjects: Air pollution forecasting; Air quality index; Correlation network; Percolation analysis
• ISSN: 0378-4371; 1873-2119
• DOI: 10.1016/j.physa.2019.123312

Air pollution has caused widespread environmental and public health problems and aroused significant attention around the world. Based on the daily air quality index (AQI) data of 35 major cities in China, the cross-correlation functions of time lags between cities are calculated and a sequence of time-evolving directed and weighted AQI correlation networks is built. The probability distribution of correlations is separated into positive and negative parts. The probability distribution of time lag exhibits that the effect of time lag is clear for cities with negative correlations and not for cities with positive correlations. Further, percolation theory technique is put forward to analyze the behavior of connected clusters in the correlation networks. The results show that abrupt phase transition usually occurs between three to six weeks ahead of the peak or valley point of the evolution of AQIs mean for highly polluted region, which suggests that this event can make an alarm. The method and results presented not only improve the understanding of the climate effects and correlated effects of AQIs, but also facilitate the study of air pollution forecasting and warning. •A sequence of time-evolving AQI correlation networks is constructed.•The time delay effect is clear for cities with negative correlations.•Abrupt transition occurs 3-6 weeks ahead of the peak point of the air pollution.•Percolation method can be used to estimate signals about extremely polluted weather.