Analysis

Comparative study of AQI
In the present investigation, the AQI level was at its highest peak on year starting among all of the four studied sites i.e., 443 in site 1, 298 in site 2, 292 in site 3, and 166 in site 4. Initial data indicates Delhi was in the hazardous range while poor air quality in other states. Although irregular declining pattern was observed in the AQI level for all of the studied locations, a significant reduction within the pollutant level can be seen after comparing initial and final values. A remarkable drop falls of 44%, 59%, 59%, and 6% in mean concentration of AQI which was observed during COVID-19 pandemic confinement for sites 1, 2, 3, and 4 respectively as shown in Fig. 3.
Fig. 3 Comparative AQI levels during pre-lockdown and lockdown period at 17:00 IST among four different air quality monitoring stations of the CPCB for four major metropolitan cities in India (site 1—ITO, Delhi, site 2—Worli, Mumbai, site 3—Jadavpur, Kolkata, and site 4—Manali Village, Chennai)

Comparative study of air pollutants

Site 1—ITO, Delhi
Delhi, India’s capital, is a massive metropolitan state in the northern area of the country and is among one of the most polluted capitals in the globe. Due to overpopulation and other responsible factors for urbanization, the pessimistic anthropogenic impact on the environment is at maximum. But, COVID-19 pandemic confinement facilitates the environment to retain its health which can be observed as a significant reduction in the air pollutant level in Delhi. At site 1—ITO, Delhi, during confinement period, the mean concentrations of PM2.5, PM10, NO2, NH3, and SO2 significantly plummeted by 49%, 33%, 29%, 63%, and 24% respectively due to reduction in anthropogenic activities including traffic and manufacturing industries. Besides, due to high temperature and insolation during the confinement period, mean ozone concentration was highly elevated by 109% as shown in Table 1.
Table 1 Air quality assessment—variations and change (%) of average concentrations for different air pollutants during the pre and COVID-19 pandemic confinement, 2020 among populous sites of four major metropolitan cities in India
Pollutants Pre-lockdown values Lockdown Variation and % change (pre-lockdown and lockdown)
Site 1 Site 2 Site 3 Site 4 Site 1 Site 2 Site 3 Site 4 Site 1 Site 2 Site 3 Site 4
AQI 238 151 144 68 134 62 59 64 − 104 (44%) − 89 (59%) − 86 (59%) − 4 (6%)
PM2.5 238 132 135 56 122 36 36 26 − 116 (49%) − 96 (73%) − 99 (73%) − 30 (54%)
PM10 150 116 122 60 100 61 45 49 − 50 (33%) − 54 (47%) − 77 (63%) − 10 (17%)
NO2 44 48 55 9 31 7 11 10 − 13 (29%) − 41 (86%) − 43 (79%) 1 (7%)
NH3 10 2 8 14 4 1 2 9 − 6 (63%) − 1 (58%) − 6 (74%) − 4 (30%)
SO2 19 12 11 14 14 5 9 9 − 4 (24%) − 7 (58%) − 2 (15%) − 6 (39%)
CO 53 28 33 25 84 13 22 35 31 (59%) − 15 (55%) − 11 (32%) 9 (37%)
O3 35 85 29 36 73 34 51 65 38 (109%) − 51 (60%) 22 (77%) 29 (80%)
PM2.5 in μg/m3, PM10 in μg/m3, CO in μg/m3, NH3 in μg/m3, NO2 in μg/m3, SO2 in μg/m3, and O3 in μg/m3
AOI air quality index

Site 2—Worli, Mumbai
Mumbai, the sixth most populous city in the world, is located on India’s west coast and is the capital of Maharashtra. It is the financial, entertainment, and commercial center of India. During COVID-19 pandemic confinement, the second most populated city of India i.e., Mumbai has moved from poor to a satisfactory level of air quality. As initially at site 2, the values of the pollutants which were scattered around 200–300 μg/m3 before confinement fallen to less than 60 μg/m3 during the confinement period (Fig. 4). The mean concentration of PM2.5, PM10, NO2, NH3, SO2, and CO, significantly reduced with a percentage of 73, 47, 86, 58, 58, 55, and 60 respectively due to shutdown of navigation activities and other industrial sectors with automobile transportation (Table 1). The drastic decline in nitrogen oxide levels over Mumbai is the result of reduced carbon-emission hotspots, industrial and coal combustion-dominated areas. A decrease in the concentration of urban ground-level ozone was recorded by 60% due to high reduction in nitrogen oxide concentration in the atmosphere.
Fig. 4 The concentration of air pollutants (PM2.5 in μg/m3, PM10 in μg/m3, CO in μg/m3, NH3 in μg/m3, NO2 in μg/m3, SO2 in μg/m3, and O3 in μg/m3) during pre-lockdown and lockdown period at 17:00 IST among four different air quality monitoring stations of the CPCB for four major metropolitan cities in India (site 1—ITO, Delhi, site 2—Worli, Mumbai, site 3—Jadavpur, Kolkata, and site 4—Manali Village, Chennai)

Site 3—Jadavpur, Kolkata
After Delhi and Mumbai, Kolkata is the third populous metropolitan area in the nation. Kolkata is the educational, cultural, and commercial center of the eastern part of the country and is the capital of West Bengal. The concentration of PM2.5, PM10, NO2, NH3, SO2, and CO at site 3 significantly dropped steeply from 242, 205, 85, 10, 9, and 49 μg/m3 as on January 1, 2020 to 20, 28, 9, 1, 7, and 22 μg/m3 during COVID-19 pandemic confinement on May 31, 2020, respectively. Also, the mean concentration levels of PM2.5, PM10, NO2, NH3, SO2, and CO significantly reduced by 73%, 63%, 79%, 74%, 15%, and 32% due to decline in fossil fuel consumption, biomass burning, and other anthropogenic activities as observed from Fig. 4, while ozone levels were significantly raised by 77% with total variation of + 22 μg/m3 during confinement period as similar to Delhi due to high winds, intermittent rains and thunderstorms, and high temperature and heatwaves.

Site 4—Manali Village, Chennai
Chennai, the capital of Indian state of Tamil Nadu, is the fourth urban agglomeration in the nation and is the 36th largest urban area by population in the world. It is located on the Coromandel Coast off the Bay of Bengal and is center for the cultural, economical, and educational activities of south India. Similar to all other studied sites, the air quality of site 4—Manali Village, Chennai also confirmed improvement in terms of reduction in pollutant level during the confinement period. The mean concentrations of PM2.5, PM10, NH3, and SO2 were reduced by 54%, 17%, 30%, and 39% respectively as shown in Fig. 4, while due to fuel and coal burning, vehicular emissions, and continuous functioning of power plants in the neighborhood of site 4, there was no significant reduction in NO2 (+ 1 μg/m3), CO (+ 9 μg/m3), and ozone levels (+ 29 μg/m3) (https://www.cag.org.in/blogs/air-quality-chennai-during-lockdown-do-we-have-clues-mitigate-air-pollution).

Pearson correlation analysis
The Pearson correlation coefficient was determined by constructing a heatmap for the concentration of various pollutants (pre and during pandemic confinement) among populous sites of four metropolitan cities of India, viz. ITO, Delhi, Worli, Mumbai, Jadavpur, Kolkata, and Manali Village, Chennai.

Site 1—ITO, Delhi
At this site, the perfect positive correlation was observed between AQI and PM2.5, a strong positive correlation between AQI-PM10 and PM2.5-PM10, whereas a negative correlation was observed for ozone with AQI and other pollutants. The correlation coefficient between AQI-PM2.5, AQI-PM10, and PM2.5-PM10 was found as 0.98, 0.82, and 0.77 respectively, showing a significantly higher positive relationship. This indicate the changes in PM2.5 and PM10 concentrations have a great influence on AQI content; i.e., an increase in their concentration will directly elevate the air quality index. Besides, AQI-ozone, PM2.5-ozone, and PM10-ozone confirmed low negatively correlated variables, i.e., − 0.31, − 0.38, and − 0.18 respectively indicating the higher values of AQI, PM2.5, and PM10 will lower down the ozone concentration. A feeble correlation exists between AQI-NH3 (0.46), AQI-NO2 (0.38), AQI-SO2 (0.28), and AQI-CO (0.11) showing mild effect on AQI (Fig. 5 (a)).
Fig. 5 Pearson’s correlation heatmap for air pollutants during the pre and COVID-19 pandemic confinement, 2020 among populous sites of four major metropolitan cities in India

Site 2—Worli, Mumbai
Product-moment correlation coefficient analysis for site 2 demonstrates the positive correlation between all of the studied pollutants as shown in Fig. 5 (b). The highest correlations were confirmed between AQI-PM2.5, with a correlation of 0.97, AQI-PM10, with 0.94, and PM2.5-PM10, with 0.91 which demonstrates PM2.5 and PM10 are the most significant dominating factors in elevating the AQI. A correlation value of 0.80, 0.74, 0.72, and 0.86 between AQI-NO2, AQI-NH3, AQI-SO2, and AQI-CO indicates a significant positive relationship, while moderate correlation was determined between CO and ozone concentration (0.53).

Site 3—Jadavpur, Kolkata
A significant positive correlation was observed between the prominent pollutants PM2.5, PM10, NO2, NH3, and CO with AQI, i.e., 0.96, 0.95, 0.86, 0.70, and 0.70 respectively in site 3 as shown in Fig. 5 (c). This implies the studied pollutants had a great impact on air quality among monitoring station of Jadavpur, Kolkata, whereas ozone shows a negative correlation with AQI (− 0.25), and other studied pollutants i.e., PM2.5 (− 0.32), PM10 (− 0.36), NO2 (− 0.48), NH3 (− 0.50), and CO (− 0.35). This indicates mean O3 concentration will significantly increase with a decrease in the mean AQI, PM2.5, PM10, NO2, NH3, and CO concentrations.

Site 4—Manali Village, Chennai
Pearson’s correlation heatmap for Manali Village, Chennai demonstrates significant positive correlations for PM2.5 (0.69) and PM10 (0.73) with AQI, while other pollutants exhibit a moderate or negative correlation. The lowest values of correlation coefficient were found for the pairs AQI-NO2 (0.26), AQI-NH3 (0.04), and AQI-CO (0.33) indicating mild association between these variables; i.e., the effect of concentration of NO2, NH3, and CO on air quality is minimal. However, the approximately zero correlation between AQI-SO2 (0.009) and AQI-ozone (0.01) indicates no linear relationship, but there may be some other strong non-linear relationship between the two variables (Fig. 5 (d)). In other words, we can say that the simple linear function cannot describe its relationship in depth.

Inferential t-statistic (Welch’s two-sample t test)
In the present study, the significant impact of COVID-19 pandemic confinement on air quality in studied locations was determined by right-tailed, Welch’s two-sample t test. The complete data set was divided into two groups, pre-confinement (A) and during confinement (B) to assess if there is a statistically significant effect of confinement on AQI. Independent random samples of sizes n1, n2 were drawn by using a random number table from both the groups and applied t test using the R-software. This inferential statistic was used to test the following hypothesis:H0: No significant difference between the means of two groups i.e., no significant effect of COVID-19 pandemic confinement on AQI (μ1 = μ2).
HA: Significant difference between the means of two groups i.e., air quality is significantly improved during COVID-19 pandemic confinement (μ1 > μ2), where μ1 and μ2 are the population means of the two groups.
From Table 2, we can observe that the t-statistic (5.91), which when compared with critical t value (1.67) at 5% level of significance (α), rejected the null hypothesis and confirmed the significant reduction in the AQI for site 1. The p value was also found to be very small, suggesting that the COVID-19 pandemic confinement reduced AQI (45%). The p value revealed it is “unlikely” that we would observe such an extreme test statistic t* in the direction of HA if the null hypothesis was true. Therefore, the initial assumption that the null hypothesis is true must be incorrect. That is, since the p value, 0.00000015, is very less than α = 0.05, we reject the null hypothesis H0 : μ1 = μ2 in favor of the alternative hypothesis HA : μ1 > μ2. However, if we lowered our willingness to make a type I error to α = 0.01 instead, the significant rejection of the null hypothesis is again observed. This is due to reduction in anthropogenic activities including fuel and coal burning, vehicular emissions, and manufacturing industries.
Table 2 Welch’s two-sample t test analysis
Site 1 Site 2 Site 3 Site 4
Sample A Sample B Sample A Sample B Sample A Sample B Sample A Sample B
Mean 241.65 134.25 159.12 65.77 144.86 57.45 75.78 63.20
Observations 36 35 36 35 36 35 36 35
Hypothesized mean difference 0 0 0 0
Degree of freedom 62 38 40 42
95% confidence interval (71.05, 143.75) (69.45, 116.99) (63.14, 111.65) (1.09, 24.05)
t-statistic 5.91 7.94 7.28 2.20
P (T ≤t) one-tail 0.00000015 0.0000000014 0.0000000074 0.03
t Critical one-tail 1.67 1.68 1.68 1.68
The same behavior can be observed from the data of Table 2 for 2nd, 3rd, and 4th studied locations where the much lowered p values exhibited the statistically significant effect of COVID-19 pandemic confinement in lowering the sample mean AQI by 58%, 60%, and 17% respectively.