Changes in air pollutants during the COVID-19 lockdown in Beijing: Insights from a machine-learning technique and implications for future control policy

Spatiotemporal impact of COVID-19 on Taiwan air quality in the absence of a lockdown: Influence of urban public transportation use and meteorological conditions

“…This implied that the meteorological conditions in 2020 might have been unfavorable for the dispersion and transportation of air pollutants. Similar findings have also been reported in neighboring areas, such as cities in China ( Hu et al 2021 ; Bai et al 2022 ).…”

“…To ascertain the impact of COVID-19 on air quality in 2020, different base years, e.g., single year (2019) ( Naqvi et al, 2021 ; Mesas-Carrascosa et al, 2020 ), two years averaged (2018–2019) ( J. Hu et al, 2021 ; Tian et al, 2021 ), and five years averaged (2015–2019) ( Nakada and Urban, 2020 ; Zangari et al, 2020 ), were proposed. Continual improvements in the overall air quality conditions in Taiwan have been observed owing to effective pollution control strategies implemented by the government ( Tsai et al, 2021 ), and coal use for power generation is gradually being reduced and replaced by liquefied natural gas since the end of 2017 ( TEPA, 2021 ).…”

Observed coupling between air mass history, secondary growth of nucleation mode particles and aerosol pollution levels in Beijing

“…71 A machine learning study using air mass back trajectories, meteorological parameters and time as predictor variables also found varying inuences of the lockdown measures, with the observed values exceeding the predicted ones on average by more than 40% during an early period of the lockdown, while during later stages the observed PM concentrations fell below the predicted ones by more than 30% on average. 49 When comparing the situation only during the most stringent restrictions in February 2020 with February 2019, we expect the PM 2.5 in 2020 to be higher by 16.2 mg m À3 based on our predictive model. Since the observed increase is only 8.0 mg m À3 , this would give us an estimate of the PM 2.5 reduction caused by the reduced emission in Feb 2020 to be around 8 mg m À3 .…”

“… 71 A machine learning study using air mass back trajectories, meteorological parameters and time as predictor variables also found varying influences of the lockdown measures, with the observed values exceeding the predicted ones on average by more than 40% during an early period of the lockdown, while during later stages the observed PM concentrations fell below the predicted ones by more than 30% on average. 49 …”

“…running chemical transport models with constant emissions, 47 parametrising the effects of meteorological variables using multiple regression analysis 48 or using machine-learning techniques with meteorological inputs. 49 While all of the aforementioned approaches are certainly sensible and have their intended uses, machine-learning techniques and regression modelling can be criticized for lacking physical connections, and while chemical transport modelling is likely the most comprehensive method, it naturally relies on the correct description of all the relevant processes and is computationally very expensive.…”