In The Science of the total environment
To avoid the spread of COVID-19, China implemented strict prevention and control measures, resulting in dramatic variations in urban and regional air quality. With the complex effect from long-term emission mitigation and meteorology variation, an accurate evaluation of the net effect from lockdown on air quality changes has not been fully quantified. Here, we combined machine learning algorithm and Theil-Sen regression technique to eliminate meteorological and long-term trends effects on air pollutant concentrations and precisely detect concentrations changes those ascribed to lockdown measures in North China. Our results showed that, compared to the same period in 2015-2019, the adverse meteorology during the lockdown period (January 25th to March 15th) in early 2020 increased PM2.5 concentration in North China by 9.8 %, while the reduction of anthropogenic emissions led to a 32.2 % drop. Stagnant meteorological conditions have a more significant impact on the ground-level air quality in the Beijing-Tianjin-Hebei Region than that in Shanxi and Shandong provinces. After further striping out the effect of long-term emission reduction trend, the lockdown-derived NO2, PM2.5, and O3 shown variety change trend, and at -30.8 %, -27.6 %, and +10.0 %, respectively. Air pollutant changes during the lockdown could be overestimated up to 2-fold without accounting for the influences of meteorology and long-term trends. Further, with pollution reduction during the lockdown period, it would avoid 15,807 premature deaths in 40 cities. If with no deteriorate meteorological condition, the total avoided premature should increase by 1146.
Lv Yunqian, Tian Hezhong, Luo Lining, Liu Shuhan, Bai Xiaoxuan, Zhao Hongyan, Zhang Kai, Lin Shumin, Zhao Shuang, Guo Zhihui, Xiao Yifei, Yang Junqi
Air quality, COVID-19, Disease burden, Long-term trends, Meteorological parameters, Random forest