3.1 Spatial and temporal variations of air pollutants
The average hourly concentrations of PM2.5, PM10, NO2, and O3 at 36 air pollutant monitoring stations were calculated during the study period to obtain the daily variation trend map of pollutant concentration (Fig. 2). The changes in pollutant concentration across different periods within four functional areas are shown in Table 1. Notably, there are evident temporal and spatial similarities in the characteristics of PM2.5, PM10, and NO2.
Figure 2. Time changes in PM2.5, PM10, NO2, and O3 concentrations in Chongqing during the study period (shaded areas are P1-P3 for the study period);The red line represents the average pollutant.
Table 1
Spatial variations in PM2.5, PM10, NO2, and O3 concentrations in Chongqing during the study period(unit: µg/m3)
Period
|
Area
|
PM2.5
|
PM10
|
NO2
|
O3
|
---|
P1
|
Urban Function Area(UFA)
New Area for Urban Development(NAU)
Northeast Chongqing Ecological Conservation Development Area (NCA)
Southeast Chongqing Ecological Protection Area(SCA)
|
44.8
48.9
33.3
30.0
|
65.0
65.1
43.2
40.5
|
33.9
25.7
23.7
19.2
|
61.4
63.5
47.4
65.4
|
P2
|
UFA
NAU
NCA
SCA
|
58.1
62.6
63.0
55.0
|
81.7
83.0
78.1
68.5
|
41.7
33.9
31.9
25.1
|
29.2
33.4
39.9
52.1
|
P3
|
UFA
NAU
NCA
SCA
|
72.9
80.3
79.9
62.5
|
102.3
108.3
99.4
72.7
|
39.8
32.4
33.2
22.2
|
46.9
52.6
38.5
55.1
|
During the P1 period, a series of lockdown measures resulted in average concentrations of PM2.5, PM10, NO2, and O3 at 44.1 µg/m³, 61.0 µg/m³, 29.2 µg/m³, and 60.5 µg/m³, respectively. While Chongqing generally exhibited low levels of PM2.5, PM10, and NO2 concentrations. The UFA and NAU regions showed significantly higher levels compared to the NCA and SCA regions due to their higher population density, industrial concentration, and anthropogenic activities. Specifically, concentrations of PM2.5 were elevated by 11.5–18.9 µg/m³ while those of PM10 were increased by 11.8–24.6 µg/m³ along with an increase in NO2 levels ranging from 2.0-14.7 µg/m³.
During the P2 period, the average concentrations of PM2.5, PM10, NO2, and O3 were 60.5 µg/m³, 81.3 µg/m³, 37 µg/m³, and 33.1 µg/m³, respectively. compared to the P1 period, there was an increase in concentrations of PM2.5 by 37.2%, PM10 by 33.3%, and NO2 by 26.7%. This may be attributed to a series of anthropogenic activities such as increased motor vehicle usage and industrial production following the liberalization of restrictions. However, O3 concentrations decreased significantly by 45.3% during this period. Surprisingly, the O3 concentration of SCA is significantly higher than that of other areas in Chongqing. The elevated O3 levels observed in SCA may be explained by analyzing its precursors using the EMKA curve (Jin and Holloway, 2015). It is possible that lower emission of NOx weaken the “titration reaction” between O3 and NOx resulting higher O3 concentrations being retained rather than removed through chemical reactions facilitated by high levels of NOx.
During the P3 period, the average concentrations of PM2.5, PM10, NO2, and O3 were 76.0 µg/m³, 102.3 µg/m³, 35.6 µg/m³, and 48.1 µg/m³, respectively. The relaxation of restrictive policies during the Spring Festival coincided with increased transportation and human activities which led to rapid surge in emission sources. Compared to the P1 period, there was a significant increase in concentrations of PM2.5 (72.3%), PM10 (67.7%), and NO2 (21.9%). Unlike the relatively low and high levels of NOx during the P1 and P2 period, NOx concentration remained relatively stable during the P3 period, while O3 decreased compared to the P2 period across all districts of Chongqing. At the same time, varying degrees of increases were observed for PM2.5, PM10, and NO2 concentrations. Compared to the SCA, notably higher pollutant concentrations including PM2.5 (+ 17.4 µg/m³), PM10 (+ 26.7 µg/m³), and NO2 (+ 11.0 µg/m³) were found in NCA possibly attributed to higher population and more frequent human activities over recent years. This indicated that policy relaxation resulted in diverse variations characteristics of spatial-temporal distribution of pollutants. Overall findings underscoring detrimental effects on air quality due to liberalization measures, emphasized significant impact exerted by anthropogenic activities on overall air quality.
Comparable findings were observed in other regions where lockdown and normalized restrictions were alternated (Table S3). In Jiangsu Province, China, the PM10 concentration increased by 23.2%, NO2 concentration increased by 16.6%, and CO concentration increased by 1.4% after entering normalized restriction (Bhatti et al., 2022). Similarly, in Wuhan, NO2 and PM10 concentrations increased by 55.5% and 5.9%, respectively (Sulaymon et al., 2021). Pollutant levels in Tianjin, Shijiazhuang and Baoding have returned to pre-lockdown levels (Ren et al., 2023). A study in the southwest coastal region of India found that the changes in PM2.5, PM10, NO2 and O3 concentrations relative to the lockdown period ranged from + 11.5% to + 38.3%, from + 1.3% to + 36.0%, from − 5.1% to + 46.1% and from − 21.2% to + 4.1%, respectively (Thomas et al., 2023). The study conducted in Campania, Italy revealed an increase in pollutant concentrations at all stations following the lifting of lockdown measures with significant changes observed for NO2 concentrations ranging between + 32% to + 63% (Cardito et al., 2023).
Studies have demonstrated a robust association between O3 and PM2.5, exhibiting a negative correlation on an hourly scale but a positive correlation on a daily scale, primarily attributed to variations in concentrations and sensitivities towards NOx and VOC emissions (Huang et al., 2021). Consequently, it is imperative to focus on coordinating the control measures for both pollutants, aligning with the current direction of atmospheric prevention and control in Chongqing. Furthermore, there exists a significant relationship between air pollutants and meteorological conditions such as rainfall, which can directly mitigate atmospheric particulate matter through scavenging processes resulting in concentration reductions (Gao et al., 2019). As shown in Fig. 2, there was a substantial decrease in particulate matter concentrations on December 25 and 27 in the P2 period, mainly due to rainfall in these days, with rainfall amounts of 3.1 mm and 8.2 mm, which led to a reduction of PM2.5 by 48.6 µg/m³; a similar situation occurred in the P3 period, with 4.93 mm of rainfall on January 14, which led to a reduction of PM10 by 130.8 µg/m³. Therefore, it is crucial to analyze the impact of meteorological conditions on pollutant concentrations while exploring their correlations and significance (as described in 3.3 below), ultimately facilitate further optimization of input parameters for prediction models.
3.2 Potential source contribution factors of pollutants
In order to analyze the potential sources of pollutants in Chongqing, it is necessary to further investigate the characteristics of air mass transport during the study period. The Meteoinfo software was utilized for calculating 24-hour backward trajectories passing through Chongqing City, computed hourly throughout the P1-P3 period, resulting in a cumulative total of 2208 trajectories (P1: 720, P2: 744, P3: 744). Based on the spatial consistency of various air masses, the trajectories of air masses moving in different directions were clustered and the results are shown in Fig. S3. The length of these trajectories can be used as an indicator for determining their movement speed. Longer trajectories correspond to fast-moving air masses that facilitate pollutant dispersion, while shorter trajectories indicate slow-moving air streams with poor diffusion conditions. During the study period, predominant air masses originated from Sichuan, Chongqing, and Guizhou. Among them, local air masses from Chongqing accounted for the largest proportion ranging from 60.5–75.54% among all the clustered trajectories. However, most of these trajectories were relatively short due to Chongqing’s topographic features characterized by surrounded mountains and closed terrain leading to low wind speeds that are unfavorable for pollutant dispersion. The remaining air masses primarily originated from northeastern Sichuan (12.92%-25.54%), southern Shaanxi (3.33%-6.32%), northeastern Guizhou (6.99%-8.33%) and southern Guizhou (5.69%-7.36%), with their trajectory directions mainly aligned with northwest and northeast direction consistent with the prevailing northerly winds observed in winter.
To further analyze the potential pollutant source areas in Chongqing, we calculated the weighted potential pollution source contribution factors (WPSCF) for different pollutant concentrations, as shown in Fig. 3. During P1, the potential source ranges of PM2.5, PM10, and NO2 were similar, however, their pollution levels varied with PM2.5 exhibiting higher pollution than PM10. Dazhou and Nanchong in Sichuan province, along with NAU, emerged as the main potential source areas for PM2.5. Meanwhile, Dazhou was identified as the primary potential source area for PM10 while NAU remained significant for NO2 emissions. O3 exhibited a wide range of potential sources including Hanzhong in Shaanxi province, Bazhong, Nanchong and Guang'an in Sichuan province, along with NAU being recognized as major contributors. Throughout the P1 period, high-value regions associated with potential pollutant sources were relatively dispersed indicating less influence from regional transmission. During P2, the potential source contribution ranges of PM2.5, PM10, and NO2 were more concentrated. Guang'an and UFA were the main potential source areas for PM2.5 and PM10. NCA and UFA were the main potential source areas for NO2. Dazhou, Guang'an, NCA, and NAU were the main potential source areas for O3, whose WPSCF values were basically less than 0.6 indicating that O3 was weakly affected by the other areas. During P3, potential sources of pollutants were more polluted. Among them, the high-potential sources of PM2.5 and PM10 were concentrated at the junction between Southern Chongqing and Northern Guizhou provinces along with NAU. In comparison to the P2 period, the main potential source region of NO2 shifted towards south, resulting in a significant increase in the number of high-value areas. The junction between Southern Sichuan, Western Chongqing and NAU, and the junction between Southern Chongqing and Northern Guizhou were the main potential sources of O3.
In general, the potential sources of PM2.5, PM10, and NO2 exhibit an expanding trend and shift from northeast to southeast. However, they are primarily concentrated in the UFA and NAU regions due to their high population density and industrial area distribution. On the other hand, the potential source area for O3 are more dispersed with a majority located around the UFA rather than its center. This could be attributed to lower NOx emissions in surrounding areas resulting in elevated levels of O3 pollution.
3.3 Influence of meteorological conditions on pollutants
In order to highlight the changes in pollutant concentrations following the liberalization and to accurately predict the predictors of the pollutant model, we selected key meteorological parameters, including atmospheric pressure, maximum wind speed and direction, wind speed, temperature, relative humidity, and rainfall. We computed the correlation between these meteorological parameters as well as their correlation with pollutant concentrations and assessed the significance of these correlations. The heat map of correlation coefficients is presented in Fig. 4. The results showed that all meteorological parameters except for maximum wind direction exhibited p-values less than 0.01, indicating a significant correlation with pollutant concentrations. The positive correlation between O3 and temperature was particularly significant (Mantel's r ≥ 0.4, p < 0.01), suggesting that higher light or UV intensities can accelerate O3 precursor reactions and exacerbate O3 pollution. (Porter and Heald, 2019). Additionally, we observed a significant negative correlation between wind speed and particulate matter (r < 0.2, p < 0.01). Higher wind speeds promote dilution of particulate matter in the air while facilitating diffusion of pollutants (Reiminger et al., 2020). Conversely, weak winds hinder air pollutant dispersion leading to increased particulate matter concentration. Furthermore, the significant positive correlation between PM2.5 and temperature (r ≥ 0.4, p < 0.01) suggests that elevated temperatures may enhance chemical reactions in the atmosphere resulting in conversion of gaseous pollutants into solid particulate matter thereby increasing PM2.5 concentration (Le et al., 2023).
3.4 Relative contribution of emissions and meteorology to pollutants
Anthropogenic emissions and meteorological conditions are the primary determinants of air quality. Meteorological conditions may mask the true impact of emissions on air quality. To investigate the relative contribution of emissions and meteorology to pollution levels, we use the 'rmweather' package for meteorological normalization. Figure 5 displays the values of the pollutant concentrations after meteorological normalization. The results showed that, except for the relatively large fluctuation in O3 concentration, the concentrations of other three pollutants remained relatively stable during the lockdown period (P1) under restrictive policies. The stability can be attributed to industrial moratoriums, motor vehicle restrictions, and people staying indoors. However, with the relaxation of these policies (P2) and even during subsequent Chinese holidays (P3), normalized values of these pollutants exhibited varying degrees of increase, suggesting an elevation in anthropogenic emission levels since policy liberalization. Through statistical analysis, we quantified the individual contributions of emissions and meteorology to air quality during different study periods (P1-P3). Our findings reveal that anthropogenic emissions contributed to a 33% increase in PM2.5 concentration while unfavorable meteorological conditions accounted for a 40.2% increase in PM2.5 concentration. Similar patterns were observed for PM10; emission-related contribution was estimated at 26.7%, whereas unfavorable meteorological contributions contributed by 43.3%. It is worth noting that although the meteorological conditions had a higher contribution rate than emissions alone, there was a substantial increase in emissions following policy liberalization compared to those recorded during restrictive periods, highlighting adverse effects associated with such policy changes on air quality.
3.5 Air Pollutant Analysis and Forecast
To establish an accurate model for predicting air pollutant concentrations following a sudden relaxation of restrictive policies, we employed meteorological and pollutant data in the Random Forest algorithm while continuously optimizing the model parameters. Compared to the initial R2 values ranging from 0.60 to 0.72, higher R2 values were achieved in this study ranging from 0.70 to 0.89, indicates strong agreement between the predicted and observed values. In general, an R2 value above 0.70 along with lower RMSE and MAE suggest improved performance of our optimized model. It is worth noting that the RMSE values for NO2, PM2.5, and PM10 models shown in Fig. 6 are lower than those reported by other studies utilizing the same modeling technique (Lv et al., 2023). Furthermore, the RMSE for PM2.5 is consistently around 13 which outperforms XGBoost model for predicting PM2.5 (Fan et al., 2020 ; Gui et al., 2020), whereas the RMSE for PM10 is approximately 16 with MAE around 11.8. Both results surpass those obtained through XGBoost or ANN models as well as other prediction models (Zhang et al., 2023b, Masood and Ahmad, 2021). Therefore, our model demonstrates satisfactory performance in prediction.
The comparison between the predicted and measured concentrations of pollutants (Fig. 7) indicates that the differences between the predicted and measured concentrations fall within a controllable range, demonstrating its strong prediction performance. Upon complete lifting of the restrictive policies, the disparity between measured and predicted monthly average concentrations becomes more pronounced. Most measured values for PM2.5, PM10, and NO2 exceed their corresponding predicted values. Conversely, the predicted values for all three phases of O3 are higher than the measured values. The mean differences between the measured and predicted values for these pollutants are calculated to be -3.98 µg/m3, -4.87 µg/m3, -3.31 µg/m3, and 1.3 µg/m3, respectively. These discrepancies can primarily be attributed to the assumption that pollutant emissions remain at restricted levels after policy relaxation, a pattern consistent with findings from previous studies on pollutant concentration trends during periods of restriction (Hu et al., 2021, Li et al., 2022). Consequently, our model demonstrates superior capability in predicting air pollutant concentrations following policy liberalization while offering valuable insights into future air pollution control strategies under similar scenarios.