COVID-19 driven changes in the air quality; a study of major cities in the Indian state of Uttar Pradesh.

Amid the COVID-19 pandemic, there has been an unprecedented cessation of outdoor anthropogenic activities leading to a significant improvement of the environment across the world. However, the positive impacts on the environment are not expected to last long as countries have started to gradually come out of lockdown and engage in aggressive measures to regain the pre-COVID-19 levels of economic activity. The present study provides for an assessment of air quality changes during the period of lockdown and unlocking across 9 major cities in the Indian state of Uttar Pradesh, including three cities (Ghaziabad, Noida, and Greater Noida) in the national capital region, which have frequently been included among the most polluted cities in the world. The pollutant load in a vertical column of air during March-July 2020 has been analyzed and compared with the corresponding period's pollution load in 2019. In addition, a detailed analysis of the ground-level changes in pollution load for Ghaziabad, Noida, and Greater Noida is also presented, along with the changes in local meteorology. A significant reduction in the total column density of NO2, CO and ground-level pollution load of PM10, PM2.5, NO2, and SO2 have been observed. In contrast, an increase in total column density of SO2 across all the cities (except Kanpur) and ground-level concentration of CO (in Noida and Greater Noida) and O3 (in Noida) was evident. The improvement in air quality (with respect to particulate matter) can primarily be attributed to the restrictions on construction and demolition activities, reduced re-suspension of roadside dust, and the restrictions on the movement of vehicles. A significant decline in the average summer temperature was recorded, and it can plausibly be attributed to lower radiative forcing due to reduced pollutant load in the atmosphere.

Introduction

Air pollution has been identified as one of the most significant causes of premature deaths worldwide, as it kills nearly 7 million people annually (World Health Organization, 2014). Air pollution has also contributed indirectly to the death toll of COVID-19, as individuals having compromised lung functionality are more likely to develop severe breathing and blood O2 saturation related complications and often need hospitalization and ventilator support (Comunian et al., 2020). Exposure to certain pollutants such as CO is further expected to aggravate the issue of low O2 saturation in blood as these preferentially bind to hemoglobin and deprive cells of O2 (Liu et al., 2020). Declining air quality is not a new threat to humanity, and the health impacts of breathing polluted air are well recognized as countries across the globe are increasingly adopting stringent air quality standards (Vahlsing and Smith, 2012).

The COVID-19 pandemic has had a fair share of attention due to its detrimental impacts on human health and socio-economic well being. Nevertheless, amid the lockdown, our outdoor environment received a well-deserved window of curtailed anthropogenic activity, which has had a favorable impact on the quality of streams and air (Zambrano-Monserrate et al., 2020). To contain the spread of COVID-19 cases, restrictions on outdoor anthropogenic activities were imposed, and such measures contributed to a reduction in energy demand across most of the industrial sectors and have also led to a significant decrease in the emission of greenhouse gases (IEA, 2020). The favorable environmental impacts of lockdown have been highlighted in multiple studies (Berman and Ebisu, 2020; Mahato et al., 2020a; Muhammad et al., 2020; Wang et al., 2020). These are primarily attributed to the restrictions on polluting anthropogenic activities, including those on non-emergency transport, industrial and agricultural activities. Like most countries, India announced the imposition of lockdown from March 25th, 2020, which continued until May 31st, 2020 in 4 phases. It was followed by the period of Unlock (ongoing) to facilitate gradual exit from lockdown in which restrictions on multiple sectors of the economy were eased. The different phases of lockdown and unlock, their time periods, and permitted activities are listed in Table 1 .

Table 1

COVID-19 lockdown and unlock in India, and allowed and restricted activities.

Phase	Acronym	Period	Restricted and allowed activities
Lockdown 1	L1	March 25th – April 14th	Restriction on all outdoor activities except essential services
Lockdown 2	L2	April 15th – May 3rd	Restriction on all outdoor activities except essential services
Lockdown 3	L3	May 4th – May 17th	Restriction on all outdoor activities except essential services, agricultural, construction, and few industrial activities
Lockdown 4	L4	May 18th – May 31st	Movement of all types of goods cargo (including rickshaws and auto-rickshaws, empty cargo vehicles, taxis and cab aggregators, Interstate movement of passenger vehicles/buses) and hospitality services were allowed
Unlock 1	U1	June 1st – June 30th	Interstate movement of vehicles allowed, special trains on selected routes, domestic air travel, all industrial and commercial activities with time restrictions, and hospitality services were allowed, night curfew from 9 p.m. till 5 a.m.
Unlock 2	U2	July 1st – July 31st	Same as unlock 1, more trains and domestic flights allowed, operation of industrial units in multiple shifts, night curfew from 10 p.m. to 5 a.m.

Recently, multiple Indian cities have featured in the list of most polluted cities in the world, and the capital city New Delhi and a few satellite cities nearby (Part of the National Capital Region: NCR) have been among the worst affected (Bhanarkar et al., 2018; Guttikunda and Mohan, 2014). The Delhi-NCR region is infamous for its poor air and often features in the list of the world’s most polluted cities. In fact, a recent report by World Economic Forum has placed Ghaziabad, Noida, and Greater Noida in the 1st, 6th , and 9th positions, respectively, in the list of most polluted cities in the world (Douglas Broom, 2020).

Continuous monitoring and dissemination of real-time air quality data in an easily understandable manner is central to the idea of maintaining a healthy outdoor environment. The pollution at the ground level governs a direct impact on human health, while the pollution in a vertical (tropospheric) column of air may provide for an analysis of overall distribution and effects on the environment (Engel-Cox et al., 2004). The satellite-based sensors offer a global range at high spatial, spectral, and temporal resolutions but usually require advanced models for source apportionment and quantification of ground-level pollution (Liu et al., 2005, Xu et al., 2018). The problem with ground-based air quality monitoring stations lies in their low spatial distribution and inadequate coverage. Nevertheless, unlike the satellite-based observations, they provide actual levels of exposure to pollutants. Therefore, both of the techniques have an inherent set of advantages and limitations. The combined use of these technologies can prove to be an ideal air quality monitoring strategy (Alvarado et al., 2019).

COVID-19 pandemic, in many ways, has proved to be an unprecedented event in the history of humanity. Qualitative and quantitative analysis of the changes in air quality attributed to the measures adopted to contain the cases of COVID-19 can aid in the estimation of background pollution load. Moreover, it can also highlight the relative influence of different sectors and can dictate future governmental interventions aimed at keeping the pollution levels in check.

Multiple studies report the short-term/long-term changes in air quality and the effect of COVD-19 dictated restrictions for Indian cities (Mahato et al., 2020a; Sharma et al., 2020; Shehzad et al., 2020; Singh and Chauhan, 2020). Unlike most of the published work on Indian cities, the present study reports the findings on COVID-19 driven changes in air quality for the lockdown and unlocking periods both. In addition, we also report both ground level and total column changes in air quality and the findings of a source apportionment study to reflect the role of different sectors of the economy. We report the changes in air quality in response to the curtailed outdoor anthropogenic activities across 9 major cities in the Indian state of Uttar Pradesh, including 3 in the NCR (Ghaziabad, Noida, and Greater Noida). For this purpose, satellite data for all the cities and ground-based air quality dataset for the NCR have been collected and analyzed. An in-depth analysis of the Ghaziabad, Noida, and Greater Noida is presented, and the relative influence of lockdown and unlock periods have been highlighted.

Material and methods

Site description

Uttar Pradesh is the most populous state of India (nearly 200 million inhabitants) and occupies 7.34% of the country’s total land area (4th largest; roughly the size of the United Kingdom) (Sarkar, 2020). The state experiences a humid subtropical climate and witnesses four distinct seasons. The cities selected for the present study include Ghaziabad (28.6692° N, 77.4538° E), Noida (28.5355° N, 77.3910° E), Greater Noida (28.4744° N, 77.5040° E), Agra (27.1767° N, 78.0081° E), Kanpur (26.4499° N, 80.3319° E), Lucknow (26.8467° N, 80.9462° E), Gorakhpur (26.7606° N, 83.3732° E), Varanasi (25.3176° N, 82.9739° E) and Prayagraj (25.4358° N, 81.8463° E) (Fig. 1 ). The first three are close to the national capital (New Delhi) and are part of the National Capital Region (NCR). The rest are major cities spread across the state of Uttar Pradesh, with Lucknow being the capital of the state.

Fig. 1

Study area in the Indian state of Uttar Pradesh.

Datasets

Satellite data

The pollution load with respect to NO2, CO, and SO2 between March 25th (beginning of the 1st phase of lockdown), 2020 till July 31st, 2020 (end of the 2nd phase of unlocking) was analyzed and compared with their levels for the corresponding period in 2019 for nine major cities across Uttar Pradesh. The Spatio-temporal distribution of pollution load was derived from the TROPMI (Tropospheric Monitoring Instrument) payload onboard Sentinel 5P satellite, a part of the European Space Agency’s Copernicus mission. The TROPMI hosts four spectrometers that operate in the ultraviolet (270–320 nm), visible (310–500 nm), near-infrared (675–775 nm), and short-wavelength infrared (2305–2385 nm) wavelengths (Kleipool et al., 2018). The level-2 data for Uttar Pradesh, having a spatial resolution of 7 × 3.5 km, was downloaded from https://s5phub.copernicus.eu/.

The NCR region is infamous for its air quality, and the same has also been sustained by the satellite-based observations in the present investigation. The air quality changes in these cities were further examined using a ground-level air quality dataset made available by pollution monitoring institutions.

AQI/IP data and instrumentation

For an in-depth analysis of air quality changes in these cities, the air quality index (AQI) and sub-indices (IP) data for individual pollutants from their respective air quality monitoring stations were used. The air quality data (daily average AQI and IP) of these cities for a total of 129 days (March 25th to July 31st) for 2019 and 2020 each were recorded from historical air quality monitoring data hosted on the website of the Central Pollution Control Board (https://app.cpcbccr.com/AQI_India/). The data for 6 criteria pollutants, including PM10, PM2.5, NO2, SO2, CO, and O3, were recorded. The measurement of these pollutants is based on the guidelines prescribed under National Ambient Air Quality Standards (NAAQS) and involves β (Beta) Attenuation (for the measurement of PM), UV-Photometric Absorption (for O3), Non-Dispersive Infrared Spectroscopy (for CO), UV Fluorescence (SO2) and Chemiluminescence (for NO2). In addition to these, these stations only record NH3, and since NH3 rarely exceeds the air quality standards, it was kept out of the scope of the study. Inadequate distribution of monitoring stations, high spatial variability in pollutant load across stations, and unavailability of scientifically sound approaches to address data gaps limit the use of spatial averaging for reporting average air quality in a city, as in the case of Delhi-NCR (Roychowdhury and Somvanshi, 2020). Since averaging was undesirable, for all the three cities, the most polluted locality (the station recording the maximum AQI) was chosen as the source of data for the reference year (2019), and data for 2020 was also recorded for the same station. In a few cases, the AQI and/or IP data for 2020 was not available for such stations, and for such cases, the data was sourced from the nearest alternate station located in an area earmarked for similar activities (same land use category). Data coverage, location specificity, and geographical coordinates for all the stations are provided in Supplementary Table S1. Hourly AQI and IP data spread across 24 h (10 p.m.–10 p.m.) was used to record the average values of AQI and IP.

The calculation of AQI in India is based on the linear segmented principle and involves the estimation of sub-indices for individual pollutants (Eq. 1), which relates the ambient concentration of pollutants to their breakpoint concentrations. The pollutant with the highest (worst) sub-index determines the AQI of the locality (Kanchan et al., 2015).where IP is the sub-index for a given pollutant, IHi and ILo are the AQI values corresponding to higher (BHi) and lower breakpoint concentration (BLo) for the pollutant, and CP is the concentration of the pollutants in μg/m3 (mg/m3 in case of CO and O3).

The breakpoint point concentration for individual pollutants, AQI categories, and the detailed methodology for the estimation of AQI and IP is provided as supplementary material (Table S2).

Meteorological data

The precipitation data (daily average) was collected from the CHIRPS (Climate Hazard Infrared Precipitation with Stations) database, which provides a high resolution (0.05°) dataset with integrated global climatology, satellite estimates, and rain gauge observations (Funk et al., 2015). The daily temperature data for the study area was collected from National Centres for Environmental Prediction’s Climate Forecast System Realanalysis (CFSR), which provides 3rd generation real analysis data at high resolution (He and Zhao, 2018). The predominant wind direction in the study area was south and south-western and the average wind speed varied between 3 and 4 ms−1. The speed and direction of wind for the study area remained unchanged during 2017–2020, as reported by Sharma et al. (2020).

Analysis of data

Prior to the analysis of data for 2019 and 2020, we performed an analysis of average air quality (AQI and IP of pollutants) data for the study area during 2016–2019 using ANOVA (data not shown). The findings of the study revealed that the difference in average air quality during 2016–2019 was insignificant (p > 0.05). Hence the use of just two years of data for quantifying the changes in air quality was a rational decision.

The statistical analysis of data was performed on Microsoft Excel (2007) and Minitab (18). Descriptive statistical analysis of AQI data (for 2019 and 2020) was carried out for the calculation of measures of central tendency (mean, median, and mode) and measures of dispersion (standard error, standard deviation, variance, range, kurtosis, skewness, and confidence level) using the inbuilt module in Minitab 18. Finally, the degree of correlation between the AQI and IP of monitored pollutants was analyzed. The correlation coefficients among the pollutants (IP) and AQI for 2019 and 2020 were also calculated. Two-tailed Student’s t-statistic for the two years of AQI and IP data of pollutants (daily difference) for all the three cities was similarly calculated, with a confidence of 95%. The premises for the t-test of AQI and IP data are listed below.

Null Hypothesis (H0): The mean AQI and IP of pollutants in 2019 (x 1) and 2020 (x 2) were the same (i.e., x 1-x 2=0).

Alternate Hypothesis (Ha): The mean AQI and IP of pollutants in 2019 (x 1) and 2020 (x 2) were not the same (i.e., x 1-x 2≠0).

Likewise, the Student’s t-test was also performed for the daily difference of temperature and precipitation and the satellite-based data for air quality. In addition, a source apportionment study by ARAI and TERI (2018) for PM pollution in the Delhi-NCR and the advisories regarding the permitted and restricted activities by the Ministry of Home Affairs, Government of India (Table 1) were used to identify the major contributors to the changes in air quality during the lockdown and unlocking periods of COVID-19.

Results and discussion

Air quality information from satellite imagery

The air quality with respect to the gaseous pollutants (NO2, CO, and SO2) was worst in the NCR region in general and Ghaziabad in particular. The average concentration of NO2 in the NCR region varied between 160-229 and 112–129 mol/km2 during the study period (March to July) in 2019 and 2020, respectively (Supplementary Table S3). Among the three gaseous pollutants, a significant reduction in average NO2 levels (p < 0.000) was observed across all the cities (Supplementary Table S4). The highest reduction of 46% was observed in Ghaziabad, which was closely followed by Noida (45%) and Greater Noida (38%). COVID-19 led lockdown (March to May) marked a characteristic reduction in the average column NO2 levels (17–46%) across all the cities. The corresponding decrease during the period of unlocking (June–July) was 2–40%, while the average reduction for the entire period was 1–44%. The NO2 level started to rise after April as restrictions were eased across sectors, including those on vehicular inter and intrastate movement of private and public passenger vehicles. Nonetheless, its concentration remained considerably lower than the values recorded in 2019.

The Spatio-temporal distribution of NO2, CO, and SO2 levels (monthly average total column density), across the state of Uttar Pradesh for both 2019 and 2020 (from March till July) is presented in Fig. 2 , Fig. 3 , and Fig. 4 , respectively.

Fig. 2

Spatio-temporal distribution of total column NO2 for the state of Uttar Pradesh.

Fig. 3

Spatio-temporal distribution of total column CO for the state of Uttar Pradesh.

Fig. 4

Spatio-temporal distribution of total column SO2 for the state of Uttar Pradesh.

CO was identified to be the most prominent air pollutant as its average concentration across cities in 2019 , and 2020 varied between 39,851-46,5455 and 39,697-42,9688 mol/km2, respectively. The highest average CO levels were recorded in Gorakhpur (49,055 mol/km2), which also recorded the highest average reduction in CO levels (11%). Noida, Ghaziabad, and Greater Noida occupied the 2nd, 3rd , and 4th place in the list of most affected cities, respectively. Like NO2, the CO levels also recorded a significant (p < 0.000) reduction during the period of lockdown, but the quantum of reduction was relatively smaller. The total reduction in average CO levels in 2020 across cities was 1–8% and the same during the period of unlocking was 1–6%. The most significant reduction was observed in April, during which only essential services were allowed.

The SO2 levels in the cities part of the NCR was significantly higher than the rest of the cities, and the maximum SO2 level was recorded in Ghaziabad. Unlike CO and NO2, the average SO2 levels recorded a significant increase (16–46%; p = 0.000–0.031) across cities except for a minor reduction in Kanpur and an insignificant increase in Greater Noida (p = 0.216) in 2020 (Supplementary Table S3 and S4).

The temporal distribution of total column NO2, CO, and SO2 levels in the NCR region is presented graphically in Supplementary Fig. S1. The NO2 levels in 2020 for all three cities in the NCR were consistently lower than the corresponding values in 2019. The CO levels followed a somewhat similar pattern. The findings of the study are in agreement with similar studies on changes in air quality during COVID-19 (Mahato et al., 2020b; Sharma et al., 2020). The lower south-eastern part of Uttar Pradesh (Sonebhadra district) was identified as the hot-spot of SO2 emission (Fig. 4). The region is host to the Rihand thermal power plant (installed capacity of 3000 MW), which is one of the largest coal-fired thermal power plants in India.

The satellite-based measurements involve the estimation of a pollutant in a vertical column of air, and these need not necessarily come from the landmass directly under the column of air. Often transport and chemical conversion of pollutants play an important role in the levels and distribution of pollutants. Advanced models are required for source apportionment and for monitoring of ground-level pollution, which is of immediate importance from a human health point of view.

Air quality changes with respect to AQI, IP, and meteorology

Meteorology

The t-test of the meteorological dataset (daily precipitation and temperature) helped identify the changes in meteorology during the study period. The average precipitation during the summer of 2019 and 2020 in Ghaziabad was 0.38 ± 1.09 and 0.48 ± 1.21 mm, respectively, and the same for the monsoon season was 4.9 ± 10.1 and 5.43 ± 9.28 mm. Although a slight increase in the average precipitation in summer (March–May) and Monsoon (June–July) both was observed for all the three cities, any statistically significant variation in seasonal during the study period was absent (p > 0.05).

The summer temperature for all the cities witnessed a significant decline (p < 0.05), which can be attributed to lower radiative forcing due to reduced pollutant load in the atmosphere (Pal et al., 2020). The highest difference between the mean summer temperatures was observed in Ghaziabad (1.507). However, the reduction in temperature during the monsoon between 2019 and 2020 was statistically insignificant for all the cities (Supplementary Table S5).

Ghaziabad

In the past 4–5 years, Ghaziabad has often featured in the list of cities that do not comply with the national ambient air quality standards. Moreover, it has also been included among the world’s most polluted cities on multiple occasions (Douglas Broom, 2020). The background concentration of PM in India is 5–10 times higher than the permissible levels in Europe and the US (Parkhi et al., 2016). Hence, their prescribed breakpoint concentrations for India have been adjusted accordingly (Kanchan et al., 2015). The non-compliance is attributed to the exceedingly high levels of particulate matter in general and PM10, in particular. For instance, the average annual concentration of PM during 2013–2018 was consistently above 250 μg m−3 (AQI poor); it, however, also includes the period of monsoon, during which the PM levels remain low, and the air quality is usually worst during winter when conditions for pollutant dispersion are unfavorable (temperature inversion) (Sharma et al., 2019). The PM2.5 levels follow closely with the concentration of PM10, as these often emanate from similar sources (ARAI and TERI, 2018).

Among the three cities, Ghaziabad was identified to be the most polluted. The average AQI in Ghaziabad from March 25th to July 31st, 2019, was 242.38 ± 96.20 with maximum and minimum values of 459 and 66, respectively. A considerable temporal variation in AQI was observed, and the days for which AQI values were <100 (satisfactory; 6.97% days) correlated well with precipitation (R2 = −0.63). In the absence of precipitation, the AQI was consistently in either moderately polluted (27.90%)/poor (33.33%)/very poor (27.13%) or severe (4.65%) category. Thus, out of the 129 days, the outdoor air was polluted on 93% of days, and in an overwhelming majority of cases, the index pollutant was PM10. The IP with respect to the ambient concentrations of PM2.5 was mostly >100 (on 74.5% days). Such high levels of PM reflect public health emergency and warrants drastic pollution abatement measures. Fig. 5 depicts the temporal variation in AQI during 2019 and 2020.

Fig. 5

Temporal variation in AQI in Ghaziabad, Noida, and Greater Noida.

The average AQI during 2020 was significantly lower (p < 0.000) than its corresponding value for 2019. The average AQI in Ghaziabad came down from 242.8 (poor) in 2019 to 130.14 (moderately polluted) in 2020. The measures of dispersion (standard error: 5.25, standard deviation: 59.73) were also lower for 2020. It suggests that the improvement in air quality was consistent. The descriptive statistics for both periods are presented in Supplementary Table S6. The t-statistic (all parameters) for all the sites was significantly (p ≤ 0.050) higher than the t-critical, and thus the H of no difference was rejected (Supplementary Table S7). The decrease in average AQI during 2019–2020 (March–July) was 46.31% while the corresponding decrease in the IP of pollutants (in descending order) followed NO2 (54.09%) > PM2.5 (49.23) > PM10 (42.70) > SO2 (36.13) > O3 (19.65%) > CO (15.57%) order (Fig. 6 ). The average AQI and IP (color-coded) of pollutants for both the periods, along with their standard deviation, are depicted graphically in supplementary Fig. S2. The average AQI and IP of PM10 improved from the polluted to moderately polluted category. The average IP of PM2.5 improved from moderately polluted to satisfactory, and the IP of NO2 and CO improved from satisfactory to good category. The IP of SO2 and O3 remained unchanged (good category), and like others, their average concentration remained lower in 2020. Out of the 129 days of restrictions, the AQI was <100 (good or satisfactory) on 35.65% days, which for the same period in 2019 was only 6.97%. It rained on 24.8 and 34.88% days in 2019 and 2020, respectively, and the degree of correlation between the rainfall and AQI was evidently strong (R2 = −0.64) in 2019. These variables were poorly correlated (R2 = −0.13) in 2020. Since there was no significant difference in the mean rainfall between the years, a weak correlation serves as an indication of a minimal effect of precipitation on AQI in 2020. The air quality improvement can primarily be attributed to the restrictions imposed to curtail the spread of COVID-19.

Fig. 6

Total Changes (%) in AQI and IP of pollutants in Ghaziabad, Greater Noida, and Noida.

The sub-indices for all the criteria pollutants witnessed similar trends. The number of days on which the air quality was poor (i.e., IP > 200) with respect to PM2.5 and PM10 levels were reduced by 53.73 and 64.64%, respectively. In 2019, the NO2 levels were in the moderately polluted category (IP = 101–200) on 14 days (10.85%), and the same in 2020 was consistently <100 (good/satisfactory). SO2 levels were within the prescribed emission standards for both periods. A significant shift in IP of CO from satisfactory to good was observed. O3 was identified to be the 3rd most significant pollutant (after PM10 and PM2.5) as its IP was 43.10% times in excess of the criteria for the good/satisfactory category (42.63% days in moderately polluted and 0.77% days in polluted category). In 2020, like other pollutants, a significant shift in the IP of O3 from moderately polluted to good/satisfactory category was recorded. The correlation between the AQIs (of 2019 and 2020) during the period of unlocking (R2 = 0.55) was significantly higher than the same for the total period of investigation (R2 = 0.245). Hence, the improvement of air quality between June and July cannot be entirely attributed to the restrictions. The correlation between the AQIs for the period of lockdown exhibited a weakly negative relation (R2 = −0.09). The negative correlation substantiates the positive impact of restrictions on air quality. The boxplot (color-coded with respect to AQI/IP) of AQI and IPs of pollutants depicting the median value, interquartile range, and the distribution of outliers for both periods is shown in Fig. 7 . Despite several outliers, the average values remained lower in 2020. A robust positive correlation (>0.99) between PM10 levels and AQI in 2019 was observed, and, on most days, the PM2.5 (R2 = 0.82) levels were also in excess of the prescribed standard (60 μg m−3). Despite a significant decrease in 2020, the degree of correlation between the AQI and the IPs of PM10 and PM2.5 was substantial. The PM2.5 and PM10 levels remained the most influential variable with an R2 of 0.79 and 0.91, respectively. Supplementary Fig. S3 depicts the plot of correlation between these parameters, and a significant degree of overlap between the AQI and IP of PM10 can easily be visualized. The R2 value for all other pollutants (with respect to AQI in 2020) increased (Table S8), suggesting that these pollutants had a more significant influence on AQI in 2020, but their overall impact was shadowed by PM10 and PM2.5.

Fig. 7

The interquartile distribution of AQI and IP of pollutants (The box shows the interquartile range, Median is the middle line of the box, whiskers represent the spread of the dataset, and circles depict the outliers).

Noida

The mean AQI in Noida during the 129 days of investigation (2019) was 198.41 ± 81.35, and like Ghaziabad, sizeable temporal variation in AQI was observed (range: 330). However, the measures of dispersion had lower values than Ghaziabad. The AQI on 88.37% days was >101 (including 42.63, 34.88, and 10.85% days in moderate, poor, and very poor category, respectively), but unlike Ghaziabad, the air quality in Noida did not witness any severe pollution events (AQI≥401). The PM2.5 and PM10 levels exceeded the good/satisfactory limits on 66.66 and 86.04% days, respectively. The PM10 levels significantly dictated the AQI as suggested by a coefficient of correlation of >0.99 while the R2 for the later and PM2.5 IP was 0.78. The SO2 levels adhered to the national ambient air quality standards as their sub-indices were always lower than the cut-off criteria for the moderately polluted category. The IP of NO2 and O3 only fell in the polluted (moderately) category on 1 and 10 days, respectively.

Fig. 5 presents the temporal variation in AQI during 2019 and 2020 at Noida. Except for a few select days, the AQI in 2020 was consistently lower than the corresponding values in 2019. The average AQI during the 129 days in 2020 was 108.60 ± 40.25. Like, Ghaziabad the measures of dispersion for the 2020 dataset were significantly lower than the corresponding figures for 2019 (Supplementary Table S6). The correlation between the AQI and rainfall for Noida was similar to the values reported for Ghaziabad for both 2019 and 2020. The restrictions greatly improved the air quality in Noida as, during 2020, there was a 426% increase in the number of days when the AQI was <100, and the corresponding decrease in the number of days when the AQI was ≥101 was 175.38%. The number of days on which the AQI with respect to PM2.5 and PM10 was ≥101 was reduced from 66.66 to 86.04% to 20.93 and 41.86%, respectively. Extreme air quality days (IP ≥ 301) with respect to PM10 were reduced from an initial 10.85% (14 days) to 0.77% (1 day), and the corresponding decrease for PM2.5 was from 1.55% (2 days) to 0%. The IP values for NO2 and SO2 were consistently in the good category. The IP of CO partially shifted from the good to satisfactory and moderately polluted categories. The number of days when the CO levels were in the moderately polluted categories increased from 2 in 2019 to 14 in 2020. Likewise, there was a partial shift in IP of O3 from good to satisfactory, but the number of days with an IP of <101 increased, which is attributed to a 20% reduction in the number of days when the AQI with respect to O3 was moderately polluted. The average reduction in the AQI of Noida was 45.26% and the same for the IP of pollutants (in descending order) was NO2 (56.58%) > PM2.5 (49.97%) > PM10 (45.40%) > SO2 (44.26%) (Fig. 6). Interestingly, there was a corresponding increase in the IP of CO (46.28%) and O3 (18.52%). An overall increase in the concentration of O3 (18.52%) and CO (46.27%) was recorded. The increase in the concentration of O3 can be attributed to higher levels of CO, as it plays an important role in the formation of tropospheric O3 (R2 = 0.412) (Table S9) (Yarragunta et al., 2019). O3 increased during the period of lockdown and decreased during unlock. It can also be attributed to higher levels of NO2 during the period of unlocking (even though a marginal increase in average reduction was noticed), leading to a higher titter value of O3, as reported in earlier studies (Hashim et al., 2020; Semple and Moore, 2020). The increase in CO was also more apparent during the period of unlocking as restrictions on many businesses and activities were eased. The boxplot of AQI and IPs of pollutants for Noida is shown in Fig. 7.

Like Ghaziabad, the correlation between the AQIs (of 2019 and 2020) during the period of unlocking (R2 = 0.41) was significantly higher than the same for the total period of investigation (R2 = 0.14). The correlation between the AQIs for the period of lockdown exhibited a weakly negative relation (R2 = −0.165). The negative correlation substantiates the positive impact of restrictions on air quality.

Greater Noida

The average AQI in Greater Noida during the 129 days of study (in 2019) was 195.68 ± 75.71, and the minimum and maximum AQIs were recorded as 43 and 378, respectively. Among the 2019 datasets for three localities, the lowest values for measures of dispersion were recorded for Greater Noida. The AQI on 91.47% days was in excess of 101 (48.83% in moderate, 31.78% in poor, and 10.85% in the very poor category). The air quality with respect to PM10 and PM2.5 was ≥101 on 85.27% (42.63% in moderate, 31.78% in poor, 10.85% in very poor) and 58.91% (43.41% in moderate, 15.50% in poor) days, respectively. The SO2 levels remained in the good/satisfactory category throughout. The IP of CO and NO2 fell in the moderately polluted category on 7 and 3 days, respectively, which was otherwise consistently <100. O3 was identified to be the 3rd most significant pollutant as its IP was ≥101 on 57.36% (55.81% in moderate and 1.55% in poor category) days.

The average AQI in Greater Noida recorded a 41.83% reduction in 2020 with a mean value of 113.82 ± 40.18. The range for the AQI dataset (2020) of Greater Noida had the lowest range (210; 24 to 234). Likewise, the other relevant measures of dispersion also recorded the lowest values (Supplementary Table S6). The t-test of the AQI and IP of pollutants revealed substantial evidence of a non-zero difference between their means for 2019 and 2020 (Supplementary Table S7). The AQI values in 2019 were ≤100 only on 11 days, while the corresponding figure in 2020 was 50. In 2019, on 14 days, the AQI values fell into the very poor category, and the restrictions in 2020 led to the complete elimination of such events as the AQI was well below the cut-off criteria for the very poor category. An increase in the number of moderately polluted days was recorded, and the same can be attributed to the reduction in days on which the AQI was poor and very poor. The correlation coefficients for the association between the AQI and rainfall were −0.59 and −0.14 for 2019 and 2020, respectively. The highest reduction in average IP was observed in the case of SO2 (58.20%), followed by NO2 (51.13%), PM10 (46.41%), PM2.5 (44.31%), and O3 (33.70%) (Fig. 6). CO, on the other hand, recorded an average increase of 14.03%, most of which occurred during the later phases of lockdown, and the increase was more pronounced during the period of unlocking. The color-coded box plot for the interquartile range for AQI and IPs of pollutants for Noida is shown in Fig. 7.

Source apportionment

Based on the analysis of meteorological factors (wind speed and direction) and four years of pollutant data, Hama et al. (2020) identified the pollution load in Delhi-NCR to be predominantly of local origin. Using a Bivariate polar plot and k-mean clustering, they suggested that the long-range pollutant dispersion (regional sources) has an insignificant effect on the air quality of Delhi-NCR. However, the contribution of regional sources is significantly higher during winter as the emissions from the burning of crop stubble in neighboring states is a dominant source of PM and gaseous pollutants in the Delhi-NCR (Beig et al., 2020).

Fig. 8 presents the findings of source apportionment and chemical speciation study for PM10 and PM2.5 pollution in the NCR by ARAI and TERI (2018). For both PM10 and PM2.5, roadside dust and construction activities were identified to be the most significant sources, accounting respectively for 38 and 31% of their total levels. Dry weather, high wind velocity, and vehicular movement support the entrainment of dust during the summer. Multiple industrial clusters characterize the landscape of the NCR region, including brick, textile, stone crusher, chemical ceramic, fertilizer, iron and steel, paper, dying industries, and these contribute 13–14% of the total PM levels. The lockdown period (especially the first 2 phases) was characterized by a stringent control on construction activities and industrial operations except for power plants and pharmaceutical operations, which marked a characteristic decrease in PM and gaseous pollutants. Although restrictions were gradually eased for several sectors after the 4th phase of the lockdown (during unlock), it did not lead to a proportionate increase in pollution levels, owing primarily to the onset of monsoonal rainfall, during which the air quality (with respect to PM) is usually improved. After the 3rd phase of the lockdown, the workforce engaged in construction and demolition activities were allowed to resume duties as these typically involve daily wage laborers for whom it was a matter of livelihood. The contribution of stubble burning towards air quality is usually seen during the winter season.

Fig. 8

Source apportionment of PM10 and PM2.5 pollution in the NCR (Modified from (ARAI and TERI, 2018)).

The contribution of vehicular emissions and biomass burning to PM levels were the 2nd and the 3rd largest, respectively. During the 1st phase of Lockdown, all modes of transportation were suspended, but gradually the ban on the transportation of goods was lifted. Since, fright vehicles do not constitute a very significant proportion of the automobile fleet, the restrictions on the movement of passenger vehicles played a role in the improvement of air quality. Since the average wind speed for both periods was similar, the restrictions on non-essential transport led to a considerable reduction in the re-suspension of roadside dust. There are designated bypass routes for large commercial vehicles, and hence their movement is not expected to contribute to air pollution in the city of Ghaziabad in any significant way. As the restrictions on non-emergency transport remain in place for most parts of the county, the contribution of re-suspended roadside dust towards air quality is primarily dictated by wind speed.

Over the past decade, the government of India has taken multiple initiatives to keep pollution from automobiles in check (Chen et al., 2019). The government decided the leapfrogging to Bharat Stage (BS; European equivalence of automotive fuel quality in India) VI by bypassing BS V and then preponed the rollout date for BS-VI fuel in the Delhi-NCR region to April 2018. The BS-VI fuel quality standard entails a stringent control on the maximum allowed concentration of contaminants; for instance, the maximum allowed concentration of sulfur in BS-VI is 80% lower than the corresponding value in BS-IV (Lathia and Dadhaniya, 2019). The judiciary has also kept an eye on air quality and has frequently sought a status report and issued notices to the governmental agencies involved in the monitoring and control of air pollution. Recently the apex court of India banned the sale and registration of BS-IV vehicles across the country with effect from April 1st , 2020. In addition, restrictions on the entry of large commercial vehicles in the city limits, phasing out of old diesel vehicles, increasing the blending percentage of biofuels, and stringent requirements on fitness certificates have contributed to reductions in vehicular pollution over the past 2–3 years. Nonetheless, the road transport sector remains to be a significant source of anthropogenic NO2, CO, and PM pollution in Delhi-NCR and the effect of such measures are expected to be seen in the long term (ARAI and TERI, 2018; Hama et al., 2020).

India is the world’s largest emitter of anthropogenic SO2, as it contributes nearly 15% of the world’s annual SO2 emission (Kapil, 2019). Coal-fired thermal power plants remain the single largest culprit, followed by the use of coal and diesel in the industrial and transportation sectors. Unlike many countries in the developed world, the utilization of flue-gas desulphurization in the power industry is virtually non-existent in India (Kapil, 2019; Solarin and Tiwari, 2020). Moreover, the contribution of adulterated fuel (by blending cheaper fractions of petroleum in the fuel) towards deteriorating air quality has also been highlighted (Badami, 2005; Goyal et al., 2006).

Unlike the satellite-based measurements, the AQI/IP study suggests that the containment measures led to a significant reduction in SO2 levels. The higher levels of SO2 in 2020 (as detected by the sensors onboard Sentinel 5P) can plausibly be attributed to regional sources, including the frequent episodes of forest fires in the neighboring state of Uttarakhand and the COVID-19 led disruptions of imported coal supply chains. The power sector was notified under essential services, and the disruptions of the supply chain prompted the power plant operators to increase the share of locally sourced coal, which is of inferior quality (Chikkatur et al., 2009; Kanitkar, 2020; Mishra, 2004). In addition, the use of unwashed coal (for the removal of sulfur) and the absence of flue-gas desulphurization could have added to the woes.

Conclusion

The findings of the study substantiate the effectiveness of lockdown in the curtailment of air pollutants across the sectors. As restrictions were eased during the period of unlocking, a slight increase in the emission levels was recorded. The most significant reduction was observed for NO2 (both ground level and total column concentration). Likewise, a significant reduction in the ground level concentration of PM10 and PM2.5 was evident. The AQI was consistently better than the corresponding values in 2019. The ground level pollution load in the NCR was mainly of local origin. The total column concentration of SO2 was higher across all the cities, while the ground level pollution load was significantly lower. Reduced vehicular movement, avoided re-suspension of roadside dust, ban on construction activities, and favorable meteorology marked a characteristic improvement in air quality during the lockdown. The restrictions had a pronounced impact on the air quality of the region, but gains started to negate with the easing of restrictions to regain pre-COVID levels of economic activity.

Author statement

Dr. Dipesh Kumar: Methodology, Formal analysis, Investigation, Writing - Original Draft, Data Curation, Writing - Review & Editing. Mr. Anil Kumar Singh: Data Curation, Visualization, Investigation. Dr. Vaibhav Kumar: Data Curation, Visualization. Ms. R Poyoja: Investigation, Data Curation. Dr. Ashok Ghosh: Supervision. Dr. Bhaskar Singh: Conceptualization, Supervision, Writing - Review & Editing, Project administration.

Declaration of competing interest

The authors have no conflict of interest to declare.