Abrupt but smaller than expected changes in surface air quality attributable to COVID-19 lockdowns

Abstract

The COVID-19 lockdowns led to major reductions in air pollutant emissions. Here, we quantitatively evaluate changes in ambient NO₂, O₃, and PM_2.5 concentrations arising from these emission changes in 11 cities globally by applying a deweathering machine learning technique. Sudden decreases in deweathered NO₂ concentrations and increases in O₃ were observed in almost all cities. However, the decline in NO₂ concentrations attributable to the lockdowns was not as large as expected, at reductions of 10 to 50%. Accordingly, O₃ increased by 2 to 30% (except for London), the total gaseous oxidant (O _x = NO₂ + O₃) showed limited change, and PM_2.5 concentrations decreased in most cities studied but increased in London and Paris. Our results demonstrate the need for a sophisticated analysis to quantify air quality impacts of interventions and indicate that true air quality improvements were notably more limited than some earlier reports or observational data suggested.

Fig. 1. Concept of detrending air pollutant levels.

C_2016–2019 and C₂₀₂₀ are the average concentrations of an air pollutant in the second and third weeks before the lockdown start date or equivalent in 2016–2019 and 2020, respectively; C_{i,2016–2019} and C_i,2020 are the daily average concentrations of an air pollutant in the ith day starting in the second week after the lockdown start date or equivalent in 2016–2019 and 2020, respectively. The vertical dashed line represents lockdown start date. P_2016–2019 and P₂₀₂₀ are the percentage changes in air pollutant levels after versus before the lockdown began or equivalent in 2016–2019 and 2020, respectively (see Eq. 2 in the main text for definition). C_{2020Business as usual} is the hypothetical concentration for the ith day starting in the second week after the lockdown date under “business-as-usual” (i.e., no lockdown) conditions. This is calculated from the prelockdown concentration (C₂₀₂₀) assuming the same percentage change as in 2016–2019 (P_2016–2019, as the “business-as-usual” change). The detrended percentage change P* (i.e., the change in air pollutant concentration arising from lockdown effects alone) is given by P₂₀₂₀ − P_2016–2019.

Fig. 2. Observed and deweathered daily NO₂ and O₃ concentrations in selected cities before and after the lockdown start dates or equivalent in 2020 versus 2018.

Columns correspond to (A) NO₂ at roadside sites, (B) NO₂ at urban background sites, and (C) O₃ at urban background sites; rows show different cities as indicated. Fine and heavy lines indicate observed and deweathered concentrations, respectively. Data are shown from December to May, shown as day of year (DOY; 1 January = 1), where the vertical dashed lines represent lockdown date. The sudden drop in deweathered NO₂ and corresponding increase in deweathered O₃ are apparent in Beijing, Wuhan, and Paris, whereas London and Los Angeles show more gradual changes. The saw-like shape in the deweathered data in some cities captures the weekly cycles of NO₂ and, to a lesser extent, O₃, particularly in western cities. Results from other cities/sites are shown in fig. S1. No data are available for roadside sites in Wuhan.

Fig. 3. Box plots of percentage change (P_dew) in deweathered concentrations of air pollutants in 2020 versus 2016–2019.

Rows represent (A) NO₂, (B) O₃, (C) PM_2.5, and (D) O_x. Lower and upper box boundaries represent the 25th and 75th percentiles, respectively; line and triangle inside boxes represent median and mean values, respectively; lower and upper error lines represent 1.5 * IQR (interquartile range) below the third quartile and above the first quartile, respectively. Number of samples for P_dew in 2020 and 2016–2019 is usually 28 and 112, respectively.

Fig. 4. Observed and deweathered O_x (i.e., NO₂+ O₃) mixing ratios in the 5 weeks before and after the lockdown start dates in the studied cities in 2020.

The six rows (from top to bottom) show results from roadside observed (A) and deweathered (B), urban background observed (C) and deweathered (D), and rural observed (E) and deweathered (F) mixing ratios. Deweathered O_x shows little change before and after the lockdown dates in 2020 and is similar across all urban background sites (all close to 40 ppb). Error bars (included for all points) represent 1 SD (n = 14). Transition period refers to the 2 weeks with the lockdown start date in the middle.

Fig. 5. Observed and deweathered daily PM_2.5 concentrations in the selected cities before and after the lockdown start dates or equivalent in 2020 versus 2018.

Columns correspond to (A) roadside, (B) urban background, and (C) rural sites; rows show different cities as indicated. Fine and heavy lines indicate observed and deweathered concentrations, respectively. Data are shown from December to May, shown as day of year (1 January = 1), where the vertical dashed lines represent lockdown start date. Results from other cities/sites are shown in fig. S2. No data are available for roadside sites in Wuhan.