key: cord-0952771-hfdilxv3
authors: Kant, Rahul; Trivedi, Avani; Ghadai, Bibhutimaya; Kumar, Vinod; Mallik, Chinmay
title: Interpreting the COVID effect on atmospheric constituents over the Indian region during the lockdown: chemistry, meteorology, and seasonality
date: 2022-03-14
journal: Environ Monit Assess
DOI: 10.1007/s10661-022-09932-7
sha: 6f2fb04db8e7f1e9f6d8a9570728c6e7cadda51f
doc_id: 952771
cord_uid: hfdilxv3

Most of the published articles which document changes in atmospheric compositions during the various lockdown and unlock phases of COVID-19 pandemic have made a direct comparison to a reference point (which may be 1 year apart) for attribution of the COVID-mediated lockdown impact on atmospheric composition. In the present study, we offer a better attribution of the lockdown impacts by also considering the effect of meteorology and seasonality. We decrease the temporal distance between the impacted and reference points by considering the difference of adjacent periods first and then comparing the impacted point to the mean of several reference points in the previous years. Additionally, we conduct a multi-station analysis to get a holistic effect of the different climatic and emission regimes. In several places in eastern and coastal India, the seasonally induced changes already pointed to a decrease in PM concentrations based on the previous year data; hence, the actual decrease due to lockdown would be much less than that observed just on the basis of difference of concentrations between subsequent periods. In contrast, northern Indian stations would normally show an increase in PM concentration at the time of the year when lockdown was effected; hence, actual lockdown-induced change would be in surplus of the observed change. The impact of wind-borne transport of pollutants to the study sites dominates over the dilution effects. Box model simulations point to a VOC-sensitive composition.

The year 2020 will forever be etched in the annals of time in memory of our (human beings) fight against coronavirus, and it came with one of the most impactful changes in all our lives -the lockdown. As the novel coronavirus began settling itself in India and the number of positive cases reached nearly 500, a nationwide lockdown was announced by the honorable Abstract Most of the published articles which document changes in atmospheric compositions during the various lockdown and unlock phases of COVID-19 pandemic have made a direct comparison to a reference point (which may be 1 year apart) for attribution of the COVID-mediated lockdown impact on atmospheric composition. In the present study, we offer a better attribution of the lockdown impacts by also considering the effect of meteorology and seasonality. We decrease the temporal distance between the impacted and reference points by considering the difference of adjacent periods first and then comparing the impacted point to the mean of several prime minister of India on 24 March 2020. In his order for the phase I of the Lockdown of 21 days, the prime minister mandated a restriction on all nonessential travel and services. The phase 2 of the lockdown was initiated from 14 April 2020, extending the ongoing nationwide lockdown until 3 May. In this second phase, all commercial and non-commercial activities were paused. After nearly 5 weeks of total nationwide lockdown, some relief came in phase 3 of the lockdown during 4-17 May 2020, characterized by partial reopening. Phase 4 of the nationwide lockdown was announced on 17 May 2020 and extended until 31 May. The country was split into 3 zones: red (high coronavirus cases and a high doubling rate), orange (comparatively fewer cases than red zone), and green (without any cases during the past 21 days). Normal movement was permitted in green zones with buses limited to 50% capacity. Orange zones would allow only private and hired vehicles but no public transportation. Complete lockdown was maintained in red zones which were further divided into containment and buffer zones. The fourth phase instituted a slow reopening with several relaxations. The curb on anthropogenic activities during the lockdown including restrictions on movement of people and vehicles had its repercussions on the concentrations of air pollutants. In addition, several industries were shut down, vehicles on the road disappeared, and power plants were operating with reduced load, which resulted in a significant decrease in the concentration of atmospheric pollutants as seen in many parts of India. As rising levels of air pollutants have been a major problem in India with repercussions on human health (Lelieveld et al., 2015) and economy (Lal et al., 2017) , understanding the lockdown-mediated change in atmospheric pollutant concentrations would be helpful to frame policy decisions.

Emissions of pollutants comprising particulate matter (aerosols) and gases (NO 2 and CO) play a vital role in the environment and human health (Xu et al., 2020) . The most commonly identified sources of primary air pollutants comprising particulate matter (aerosols) and gases (NO 2 and CO) are vehicular emissions, manufacturing, power generation, manufacturing industries, construction activities, road dust, waste burning and combustion of oil, coal, and cooking activities in the households. Figure 1 shows the source apportionment for PM 2.5 for 11 cities of India based on data from UrbanEmissions.info. UrbanEmissions.info was a program to build an emission inventory for the following pollutants (i) PM, (ii) SO 2 , (iii) NOx, (iv) CO, (v) CO 2 , and (vi) NMVOCs (Guttikunda et al., 2019) . The emission inventory was built at a spatial resolution of 1 × 1 km 2 which included anthropogenic sources, large and small scale power generations, industries, domestic, open waste burning, and open fires and non-anthropogenic sources, such as sea salt, dust storms, biogenic, and lightning. While emissions from vehicles contribute the maximum to PM 2.5 in Guwahati (35%), Thiruvananthapuram (60%), Jaipur (38%), and Pune (35%), industrial emissions dominate the PM 2.5 sources in Nagpur (83%), Ahmedabad (63%), Kolkata (54%), Amritsar (32%), and Hyderabad (28%), while road dust is the major PM 2.5 contributor in Jodhpur (42%). Guttikunda et al. (2014) observed that one of the major sources of PM 10 is road dust, and it can account for up to 30-40% of the PM 10 pollution in most cities. Figure 1 also shows the sources of PM 10 in 11 cities of India based on data from UrbanEmissions.info. It can be seen that the major contribution to PM 10 is road dust and construction activities. In Guwahati, Hyderabad, Thiruvananthapuram, Pune, and Jodhpur, road dust constitutes more than 60% in PM 10 . In Kolkata, 37% of PM 10 comes from road dust and 38% from industrial emissions. The two major sources of PM 10 in Nagpur are industrial emissions and road dust. They contribute 72% and 16%, respectively. In Ahmedabad, the major sources of PM 10 are road dust (45%) and industrial emissions (39%). In Delhi, the road dust and construction activities contribute 45% of the PM 10 , followed by 17% from burning of waste (agricultural and domestic waste) and 14% from vehicular emissions (Guttikunda et al., 2014) .

The restriction on transport sector during the lockdown in India, in addition to shutdown of factories and industries, induced a remarkable decrease in PM concentrations (Devara et al., 2020; Mahato et al., 2020; Mitra et al., 2020; Peshave & Peshave, 2020; Rahaman et al., 2021; Ramasamy et al., 2020; Sharma et al., 2020) . In Delhi, 14% of PM 10 is contributed Page 3 of 31 274

Vol.: (0123456789) by vehicles (Guttikunda et al., 2014) , so then we can expect some decrease in the concentration of PM 10 during the lockdown just due to reduction in vehicular emissions neglecting effects of dilution, oxidation, and other seasonal/meteorological impacts. Further, the restriction on vehicular transport has an additional impact on road dust resuspension in addition to reduced tail-pipe emissions.

Similarly, 45% of total NO x emissions in India is contributed by coal burning in thermal power plants (Garg et al., 2001) . Road transport contributes another 32% of NO x emissions (Garg et al., 2001) . As thermal power plants were running at reduced capacities and road transport came to a total halt, a reduction in NO x is envisaged. However, O 3 being a secondary pollutant, changes in O 3 would be much more difficult to unravel from direct observations, and the dependency on hydrocarbons, NO x , and meteorological factors needs to be investigated. Coincidentally, xylenes were found the largest contributor to the O 3 formation followed by toluene in Delhi (Hoque et al., 2008) .

Some observations pertaining to COVID-induced changes in atmospheric constituents Pathakoti et al. (2020) studied the impact of nationwide lockdown on air pollution and observed that the aerosols/particulate matter over the country decreased by ~ 24% from the 5-year mean level with a marked reduction over the Indo-Gangetic plains (IGP) region. During the lockdown period, the world's most polluted city, Ghaziabad, showed a reduction of 57% in PM 10 and 48% in PM 2.5 compared to the average levels of 2019 . also observed reductions of 57% in PM 2.5 and 58% in PM 10 in Patiala city.

Using the ratio of NO 2 and HCHO vertical column densities measured from MAX-DOAS in Mohali, it was found that the peak daytime O 3 production regime is sensitive to both NO x and VOCs in winter but strongly sensitive to NO x during summer (Kumar, Beirle, et al., 2020; Kumar, Pratap, et al., 2020) . Chen et al. (2020) pointed out that O 3 production is less sensitive to solar radiation in summer compared to winter. Reducing NO x alone increases O 3 , such that a 50% reduction in NO x emissions leads to a 10-50% increase in surface O 3 . In contrast, reducing VOC emissions can reduce O 3 efficiently, such that a 50% reduction in VOC emissions leads to a 60% reduction in ozone for Delhi . The sensitivity of atmospheric oxidants on NO x levels was also elaborately studied during the Cyprus Photochemical Experiment (Mallik et al., 2018) . Kumar (2020) found 37% reduction for NO 2 over India in comparison to the average value of 2017-2019. Jain et al. (2021) having studied the phase-wise variations in atmospheric constituents during the COVID-19 lockdown over a tropical rural site (Gadanki in southern India) point out that trace gases, viz., NO, NO 2 , CO, SO 2 , CO 2 , and CH 4 , whose emission sources are dominated by predominantly anthropogenic origin have shown a reduction of over 50% due to COVID lockdown-induced emission reductions. While a plethora of publications point to the reductions in primary pollutants as a result of lockdownmediated reduced emissions, we feel that comparison of direct year-to-year concentration change to estimate the lockdown impact is inappropriate. Therefore, we took up the following study analyzing data over twenty-four stations covering different parts of India ( Fig. 2) (ii) Particular matter of aerodynamic diameter less than 10 μm (PM 10 ) (iii) Sulfur dioxide (SO 2 ) (iv) Nitrogen dioxide (NO 2 

The measurement techniques for O 3 , CO, NO x , SO 2 , PM 2.5 , and PM 10 are available from the technical specifications for continuous ambient air quality monitoring (CAAQM) (CPCB, 2019, https:// cpcb. nic. in/). The ultraviolet photometric O 3 gas analyzers work on the Beer Lambert's principle on absorption of radiation at 254.7 nm by atmospheric O 3 . The detection limit of the instrument is 1 ppb with a response time of 30 s or less. The CO instruments are based on gas filter correlation technology and operate on the principle of infrared absorption at 4.67 μm vibration-rotation band of CO (Nedelec et al., 2003) . It has a detection limit of 100 ppbv at a 60 s response time. The zero noise of the instrument is 20 ppbv root-mean-square (RMS) at 30 s averaging time. The NOx instruments are based on the detection of chemiluminescence produced by the oxidation of nitric oxide (NO) by O 3 molecules, which peak at 630 nm radiation (Navas et al., 1997) . The method is specific to NO only. NO 2 is measured by converting it into NO using a molybdenum convertor and then measuring total NOx as NO. Unfortunately, the reduction of NO 2 to NO is not specific for NO 2 , and other nitrogen species are also reduced to NO and act as interferences in the NO 2 measurements. The detection limits of these instruments are around 1 ppb at a response time of 120 s or less. The PM 10 measurements are based on the principle of β-ray attenuation. The particulate matter in ambient air is sampled through the instrument at a flow rate of 16 l/min and collected on fiberglass filter tape. Comparison of measurements of β-ray radiation by scintillation/G.M. counter before and after sampling gives a measure of the amount of PM 10 . The PM 2.5 measurements are similar to PM 10 , but the particle size cutoff is in the range of 0-2.5 μm.

For this analysis, twenty-four stations are selected representing the different emission and climatic regimes across India. Initially, thirty-two stations were selected to derive a pan-Indian representativeness. However, based on data availability, the final number of stations is reduced to twenty-four such that at least five criteria pollutant data are available for the lockdown period and the corresponding time in previous year. The study locations span north, east, west, and south India including the IGP and north-east. Together, the stations represent most of the major emission sources of PM and NO 2 across India. The stations also represent various climatic regimes ranging from the arid regions (e.g., Jodhpur) to tropical wet (Kolkata), from hilly (Aurangabad) to coastal (Visakhapatnam). The details of these stations are provided in Table 1 , and henceforth, the stations are represented by their 3-letter station codes only.

In order to remove the outliers, the raw data were filtered station-wise and species-wise to remove values above 95 percentile and below 5 percentiles at every 4-month interval. Since we are concerned with the average variation of the pollutants, the kind of filtering helps to remove the bias due to extreme events (meteorological or chemical) and errors related to instruments, sampling, and human effects. Additionally, data was also checked manually for inconsistencies. The weekly/ fortnightly means before and after lockdown of 2020 were compared to the corresponding difference of the average of the data for the year 2015 to 2019 for this analysis. The data availability is shown in Fig. 3 . A robust regression analysis was performed to identify the dependence of PM 2.5 and PM 10 on the planetary boundary layer (PBL) height for each of the stations. Hourly resolution PBL data at each measurement stations were taken from linear interpolation of 0.25° ERA5 reanalysis dataset (Hersbach et al., 2020) . The PBL in ERA5 is estimated using the bulk Richardson number following the methods proposed by Siedel et al. (2012) . The dependence was tested for different levels of significance using a two-tailed Student's t test.

A zero-dimensional atmospheric box model F0AM version 4.1 (Wolfe et al., 2016) was set up for calculating the O 3 concentrations as a result of atmospheric photochemical processes for an example site over Ahmedabad. Ahmedabad was selected because of availability of VOC data. The model employs Master Chemical Mechanism (MCM) 3.3.1 chemistry (Jenkin et al., 2015) . The MCM setup included a total of 1363 species and 4205 chemical reactions (Kumar et al., 2018) . The rate constants used in the model are taken from the reviewed rate constants published by Atkinson et al. (2006) . The model was constrained with hourly averaged concentrations of NO, NO 2 , and CO from CPCB database and temperature, pressure, and relative humidity from wunderground.com for Ahmedabad site for 24 March 2020 as prelockdown and 31 March 2020 as lockdown period. Individual days were taken for comparing the simulation instead of the weekly mean because the average precursor ratios may not represent the true precursor composition for observed O 3 . Therefore, days with relatively high O 3 were selected for the simulation. It should be noted that our purpose with these simulations is not to predict the observed O 3 concentrations but to get an idea of the VOC-NO x regimes operating during the period. The VOCs values used in the simulation are taken from published literature over Ahmedabad listed in Table 4 . Hydrogen (H 2 ) mixing ratios are held constant at 550 ppb, respectively. The model calculated concentrations of secondary species at the end of each hour were used as the initial concentration of the model run for the second hour. The simulations are performed in steady-state conditions with a spin-up period of 3 days.

Meteorology during lockdown/unlock periods A major objective of this paper is to decouple the emission changes during lockdown from meteorological impacts on atmospheric pollutant concentrations. Crilley et al. (2021) found that though the lockdown has brought down the local emission sources, the chemical processes in the atmosphere and weather events independently contribute to the observed changes in the pollutant levels.

During winter in northern India, the winds are north-easterly and westerly with thick fog, low wind speeds, and low boundary layer height which can degrade air quality further (Tiwari et al., 2018) . The year 2020 was the eighth warmest year in India since 1901. The annual mean land surface air temperature averaged over the country was + 0.29 °C above normal of the average of 1981-2010. However, this value was very much lower than the highest warning year over India during 2016 (+ 0.71 °C anomaly from normal).The monsoon and post-monsoon seasons of 2020 showed mean temperature anomalies + 0.43 °C and + 0.53 °C.

One significant meteorological feature of 2020 was the higher numbers of western disturbances that continued even in the summer season. Due to the consecutive arrival of western disturbances in northern India, the ventilation and dispersion of pollutants can result in a better air quality. The wind system showed a major change in northern India and IGP during the first half of March bringing in the transition from winter to pre-monsoon. Also, Bhawar et al. (2021) observed that the transport of dust from West Asia was lower compared to the previous years during the lockdown period, in spring of 2020. This led to the reduction of dust aerosols over the Indian region. The cyclone Amphan was the first pre-monsoon (20 May) super cyclone of the century . The year 2020 was also the third to record the highest precipitation during the last 30 years. The country received 109% rainfall of long period average (LPA) with June (118%), August (127%), and September (104%) witnessing above normal rainfall. Weather conditions and pollutant levels have a strong linkage that may obscure the variation in emission levels over different cities (Radaideh, 2017) . Meteorological variability was found to account for 40-70% of ozone variability and 20-50% of particulate matter variability in Southwestern USA (Wise & Comrie, 2005) .

Ventilation coefficient is another parameter used as an indicator of atmospheric dispersive capacity. It is the product of mixing layer height multiplied by average wind speed. It is an important factor for the determination of pollution potential over a region (Chan et al., 2012) . If VC < 6000 m 2 s −1 , the air pollution potential is considered to be high. In winter, the VC values are lower due to stable atmospheric conditions and lower wind speeds ranging 2-3 m/s. Similarly, VC is higher in summer due to unstable atmospheric conditions and increasing the wind speed. The coinciding phases of lockdown implementation and transition from winter to summer, i.e., lower to higher VC, can also have effects on the concentration of atmospheric constituents and are analyzed here.

Estimating the role of emissions from weekly changes in atmospheric constituents The vertical line in black color represents the start of the lockdown period, i.e., the 25 March 2020, and the vertical line with blue color represents the beginning of unlock phase, i.e., 1 June 2020. The horizontal lines represent the permissible levels of pollutants based on the National Ambient Air Quality Standards given by CPCB (24 h averages of 60 µg −3 and 100 µg −3 for PM 2.5 and PM 10 , respectively, while these are 80 µg −3 (24 h) and 100 µg −3 (8 h) for NO 2 and O 3 , respectively).

A significant decrease in PM 2.5 concentrations during lockdown compared to the previous period is observed over in northern India, i.e., Jalandhar (47%), Gobindgarh (53%), Noida (59%), Patiala and Sonipat (67%), and Delhi (33%). Mahato et al. (2020) have also estimated a similar reduction of PM 2.5 , viz., 39%, in Delhi during the said time period. Chikara and Kumar (2020) also observed an appreciable decrease in the concentration of various pollutants (PM 2.5 , PM 10 ) in Delhi, Mumbai, and Kolkata due to lockdown (Table 2) . PM 2.5 concentrations have also decreased in western India; the decrease recorded just in comparison to similar period in previous year is as follows: Jaipur (49%), Jodhpur (44%), Kota (56%), Ahmedabad (65%), Pune (37%), Aurangabad (52%), and Solapur (29%). Navinya et al. (2020) also mention a decline in PM 2.5 of about 68% over Ahmedabad. Similarly, a large decrease in PM 2.5 of about 58% is observed over Nagpur in Central India. However, other stations in this region do not exhibit such a large decline, viz., Chandrapur (17%) and Hyderabad (9%).

Mixed signals are observed in IGP with a 52% decrease in Varanasi and negligible decrease in Kolkata (1.5%) and, surprisingly, a large increase in Guwahati (75%). Singh and Chauhan (2020) observed that due to the dominance of westerly winds from arid and semi-arid regions, a lower boundary layer (due to lower temperatures) prevails over Delhi and central IGP cities compared to other major cities like Mumbai, Hyderabad, and Kolkata. In March, the average concentration of PM 2.5 in Delhi and central IGP remains higher in comparison to other regions.

Stations in coastal India also show variable features, decrease in Visakapatnam (30%), Tirupati (33%), and Thiruvananthapuram (49%), while an increase is observed in Rajamundry (7%). About 60% of India's mean population-weighted PM 2.5 concentrations come from anthropogenic source sectors, while the remainder are from other sources, wind-blown dust and extraregional sources (Venkataraman et al., 2018) . As mentioned in Fig. 1 as well as studied by Venkataraman et al. (2018) , leading contributors to PM 2.5 are residential biomass combustion, emissions from power plant and industrial coal combustion, and anthropogenic dust. Like PM 2.5 , a rapid decrease in PM 10 concentrations is observed in northern India in comparison to previous period, viz., Jalandhar (68%), Gobindgarh (55%), Noida (64%), Patiala (74%), Sonipat (35%), and Delhi (56%). The total emission of PM 10 from different sources was estimated using ISCS3 model as follows: industrial point sources (26%), vehicles (21%), domestic fuel burning (19%), paved and unpaved road dust (15%), and the rest as other sources (Behera & Sharma, 2010) . Mahato et al. (2020) observed a 60% reduction in PM 10 during lockdown compared to the last year (i.e., 2019). In western India, PM 10 also shows a decrease like PM 2.5 , viz., Jaipur (56%), Jodhpur (50%), Kota (49%), Pune (62%), Aurangabad (54%), and Solapur (48%). In Central India, Nagpur (61%) shows a large decrease. However, Chandrapur which had shown only 17% decrease in PM 2.5 now shows a 43% decrease w.r.t. PM 10 . The coastal stations also show decrease in PM 10 during the lockdown, viz., Visakhapatnam (30%), Tirupati (50%), Thiruvananthapuram (28%), and Rajamundry (11%). Navinya et al. (2020) have also observed a decline of 71% in PM 10 over Delhi.

From Fig. 2 , it is observed that the PM 2.5 and PM 10 concentrations in the stations of northern India have started to increase in the later part of the lockdown. The regulation for movement of residents relaxed after the end of the first phase of the lockdown which might be the primary reason for the increase in emissions, which could be attributed to automobiles, and a consequent increase in the concentrations of atmospheric pollutants. During the second phase, the PM 2.5 and PM 10 concentrations in Chandigarh increased by 7.7% and 22.3% respectively, as compared to the first phase (Mor et al., 2021) . Mor et al. (2021) also observed that the air temperature in Chandigarh during the first, second, and third phase of lockdown increased by 4.5 °C, 3.3 °C, and 1.6 °C, respectively, compared to the prelockdown period due to the onset of the summer season. Therefore, a slight decrease in pollutant levels during the lockdown period can be attributed to higher temperature. The increase in temperature increases the vertical mixing of pollutants in the troposphere (Ravindra et al., 2019) .

From Fig. 2 , an increase in PM 2.5 is observed over Jaipur and Jodhpur in between the lockdown and unlock periods in the months of late March and April. This could be attributed to the prevailing upper air cyclonic circulation and western disturbances that caused several dust storms with gust winds and thunderstorms over different parts of Rajasthan reducing the temperature to markedly below normal values (https://m. daily hunt. in/ news/ uae/ engli sh/ gplus+ engli sh-epaper-gpls/3+ fire+ incid ents+ repor ted+ across+ guwah ati-newsid-n2633 10294).

Studying the simultaneous changes in PM 2.5 and PM 10 over the different study locations, it is observed that in general, PM 10 and PM 2.5 changes over most of the locations in the pandemic year compared to the previous year(s) average are almost going hand in hand for the weekly time series. However, there are some exceptions like Delhi, where the PM 2.5 percentage change shows a stronger increase in the initial phase of lockdown compared to PM 10 . This feature is also visible in Chandrapur and to some extent in Pune. This indicates that the lockdown measures were able to subdue the sustained natural tendency for PM increase during this period, both from emission sources and atmospheric chemical means. Here, we would like to point out that a source apportionment study of PM 2.5 and PM 10 for Delhi NCR indicates that biomass burning (BB) contributes 12% and 15% to PM 10 and PM 2.5 , respectively, during summer. The contribution of BB is slightly higher during winter with 14% and 22% influence on PM 10 and PM 2.5 , respectively (ARAI and TERI (2018). Guo et al. (2017) observed that during 2015, SOA contributed a miniscule of 7% and 3% to PM 2.5 over Jaipur and Delhi, respectively. Behera and Sharma (2010) have estimated that SOA contributes about 18% mass in winter and 12% mass in summer to PM 2.5 in Kanpur city in IGP. In Delhi and nearby regions, SOA was found to contribute 16 ± 6 µg −3 (5.8 ± 2.6% of PM2.5 mass) in summer (Nagar et al., 2017) . The oxygenated organic aerosols (OOA) over Delhi are roughly 1.7 times lower during spring compared to winter, with a distinct diurnal variation exhibiting around 15 µg −3 during peak photochemical periods, while values increase to over 30 µg −3 during night (Bhandari et al., 2020) . Organic aerosol, which contributes 55-75% of PM 1 over Ahmedabad, was measured to be about 7.5 ± 8.2 µg −3 during early October, out of which OOA was found to constitute about 58% . During lockdown, the light volatile (LV) and semivolatile (SV) components together constituted about 74% of the organic aerosol over Ahmedabad (Dave et al., 2021) . From this paper, it is important to note that the decrease of hydrocarbon like organic aerosols during lockdown over prelockdown was much larger compared to the decrease in volatiles and semi-volatiles. The daytime peak in LV-OOA was about 4.5 µg −3 before lockdown, while it decreased to around 3.5 µg −3 during lockdown, but the night-time values were identical. Figure 5 shows the dependence of PM 2.5 and PM 10 on PBL height. The association has been tested for significance using Student's t test. The dependence of PM 2.5 on PBL is significant at 95% confidence level over Jalandhar, Guwahati, and Kota; at 90% in Gobindhgarh, Delhi, and Solapur; and at 99% confidence level in Jaipur. The relationship between PM 10 and PBL is significant at 95% confidence level over Jalandhar, Guwahati, Thiruvananthapuram, and Tirupati; at 90% over Patiala and Kota; and at 99% in Jaipur and Solapur. It is further observed that PM changes are positively correlated with the PBL changes except for Visakhapatnam (Fig. 5) . Positive association of PBL with PM indicates that dilution effects are negligible compared to the import of pollutants by air masses. Of course for a site with strong marine influence, viz., Visakhapatnam, the role of PBL dilution comes into picture as increasing PBL height results in lower concentrations. Moreover, Visakhapatnam is bounded by the Eastern Ghats on three sides along with warm and humid climate, thus restricting the dispersion of particulate matter. But for all other sites, ventilation coefficient needs to be considered as transport effects dominate over dilution impacts. Figure 6 shows the weekly change of PM 2.5 and PM 10 for the selected locations. In the 1st phase of lockdown, the capital city of India experienced a decrease of around 30% in the concentrations of PM 2.5 . Noida, Jaipur, Jodhpur, and Kota experienced a decrease of around 50% in PM 2.5 concentration. Some stations in central India (e.g., Nagpur and Aurangabad) also saw a decrease of more than 50% in the initial phase of the lockdown. Some of the stations in Coastal India, Visakhapatnam and Tirupati, also recorded a decrease of 30% and more. A decrease of 50-60% is observed in the cities of Punjab and Haryana. Varanasi recorded a decrease of 50% in PM 2.5 concentration. The lowest change in the concentration of PM 2.5 during 1st phase of lockdown was seen in Kolkata, and it was estimated to be around 1.5%. Stations in northern India, i.e., Delhi, Gobindgarh, Patiala, and Noida, all experienced a decrease of more than 50% in the concentration of PM 10 during the 1st phase of lockdown. A similar reduction in PM 10 concentration was seen in all the stations in Western and Central India. All the stations in coastal India experienced a reduction of around 30-40%. An increase of only 2% in PM 10 concentration was seen in Kolkata.

Guwahati was the only among the selected stations which showed an increase of around 70% in both PM 2.5 and PM 10 concentrations. An increase in fire counts around Guwahati was observed during March 2020 from MODIS (Fig. 7) . The fire events were observed both around the city and within the city. Among the many fire events within the city was a major fire during 30 March 2020 near Lalmati near Games Village, Guwahati, and fires during March 19 at Guwahati's Fancy Bazaar area (Web Ref 1, Web Ref 2). Further, according to Guttikunda et al. (2014) , 80% of the households here have non-gas cookstove. Increased residential emissions in Guwahati might also add to an increase in the concentration of PM from fires. Therefore, the impact of lockdown, viz., closure of vehicular traffic in the 1st phase, is not seen in Guwahati, instead a spike is observed because of possible enhancement in fires and residential activities. Hari Hence, Patiala experiences a positive change after a delay when compared to the normal years. 

One of the largest impacts of the lockdown was the restriction in movement of vehicles leading to large decrease in NO 2 concentrations. In northern India, 30% decrease in NO 2 was observed over Delhi, while Jalandhar (20%), Patiala (80%), Gobindgarh and Noida (20%), and Sonipat (60%) all show a decline in the initial phase of lockdown compared to same period in previous years. Chikara and Kumar (2020) . Satellite-based seasonal variations of tropospheric NO 2 concentrations show a maximum during winter-summer month and a minimum during the monsoon seasons, with the change between maxima and minima being 2-4 times in various regions of India (Ghude et al., 2008) . Mallik and Lal (2014) estimated a 20-40% enhancement in NO 2 in winter compared to pre-monsoon months. Thus, the lockdown-induced decrease in NO 2 is comparable to the seasonal change in NO 2 over Delhi. Chimurkar et al. (2020) show that the northern Indian region showed nearly 100% decrease for NO 2 during lockdown compared to prelockdown, while this value was 38% in 2019. During the lockdown period starting from 22 March 2020, a sudden drop in tropospheric NO 2 concentrations over IGP was observed (Singh & Chauhan, 2020) . NO 2 reductions were also observed in our study locations of IGP: Varanasi (5%) and Kolkata (10%). Strangely enough, Guwahati showed an increase of 40% in NO 2 during the lockdown period. NO 2 reductions were also observed in Western India: Jaipur (60%), Jodhpur (20%), Kota (30%), Ahmedabad (50%), Solapur (80%), and Pune (40%) show a decrease, while a 50% increase was observed in Aurangabad. NO 2 decreases were also observed in coastal India: Visakhapatnam (20%), Tirupati (80%), Rajahmundry (20%), and Thiruvananthapuram (20%). Chimurkar et al. (2020) have also observed a decline of 63% in NO 2 over Nagpur in Central India.

In contrast to other pollutants, O 3 concentrations showed varied features not only in India but across the world. The mixed signal was observed in O 3 concentrations in northern India: Delhi (50%) and Jalandhar (40%) showed an increase, while Gobindgarh (20%), Noida (20%), and Sonipat (60%) show a decrease in O 3 . Most of the stations in western India show an increase in O 3 : Jaipur (20%), Jodhpur (20%), Kota (60%), Ahmedabad (40%), Aurangabad (20%), and Solapur (15%). However, Pune shows a 20% decline in O 3 . For most places during the lockdown, an increase in concentration of O 3 can be related to a corresponding decrease in NO x under VOC-limited conditions . For stations in coastal India, decrease in O 3 was observed: Visakhapatnam (20%), Tirupati (20%), Thiruvananthapuram (10%), and Rajahmundry (20%). point out that O 3 also showed a mixed variation with a mild increase in IGP and a decrease in the south. Das et al. (2021) point out that average concentration of O 3 increased by 28% during lockdown in comparison to 30 days average prior to lockdown. Figure 8 shows the difference between the mean of the data from 11 to 24 March 2020 and 25 March to 7 April 2020. In Fig. 8 (left panel) , the black bar shows the change in PBL. The red and blue bar represents PM 2.5 and PM 10 , respectively. In all the stations in northern India, the decrease in PM concentration is seen coinciding with the increase in PBL. Similar behavior is seen in Western India except for Solapur. However, regression analysis (Fig. 5) clearly indicates that PBL dilution may not be the major cause of decrease in PM levels as import of pollutants could have a much stronger effect. In Solapur, PM 2.5 increases along with an increase in PBL indicating import of PM 2.5 to the site. The increase in PM concentration is observed in Varanasi and Guwahati with an increase in PBL. A similar change can be observed in Chandrapur and Talcher. This could be because of the transport of particles (VC is very high here). The decrease in PM concentration is seen with decreasing PBL in Visakhapatnam and Thiruvananthapuram. These regions may have an overwhelming influence of the sea which led to the dilution of the PM concentration. In Fig. 8 (right panel) , the black bar shows the change in VC. The red and blue bar represents NO 2 and O 3 , respectively. In coastal India with a decrease of VC, NO 2 concentration also decreases, and O 3 concentration increases. In the IGP region of Varanasi with an increase of VC, NO 2 also increases, whereas O 3 decreases. This means that the increase in NO 2 is brought about by atmospheric transport, while this same NO x -rich air influx leads to decreased O 3 due to titration effects. But for Guwahati, variation is the opposite, and with an increase of VC, NO 2 concentration decreases and O 3 increases. In western India, Jaipur, Jodhpur, Pune, and Ahmedabad show the same variation of decrease of NO 2 and O 3 with a decrease in VC, but in Kota and Aurangabad, O 3 concentration increases with decreasing VC.

Although published literature deals with a comparison of the lockdown phase with similar phases during previous years to estimate the impact of lockdown, it is observed that the concentration of aerosols in some stations of northern India and northwestern India started to decrease even before the imposition of lockdown. The reasons for this decrease can be specific to different pollutants, but an overarching effect of reduction/closure of emission sources must be normalized to the impact of change in air mass trajectories (Meteorology during lockdown/unlock periods section, Fig. 10 ) to make a real estimate of the lockdown-induced changes. Further, the change in prevailing wind direction is not a sharp point in space and must be estimated judiciously station-wise. Also, fairly widespread rainfall/thunderstorm activity was observed over Western Himalayan region and Northwest India in the month of March 2020 IMD reports (a, b). The precipitation occurring due to the western disturbances removes some of the aerosols over the region by wet deposition, and the wet conditions reduce the introduction of fresh aerosols into the atmosphere. One way to resolve this cocktail of effects is to derive an expected change in concentration of pollutants and compare it with the observed change in the lockdown year; the difference would indicate the actual lockdown-induced change. For this analysis, the expected change is calculated based on the difference of subsequent fortnights before and after the imposition of lockdown. Fortnight is taken instead of week to subdue effects of sudden changes on a particular day, and a longer averaging period would make the mean more robust. The difference of subsequent fortnights is calculated for 2015-2019 (depending on data availability) for each station and then averaged to produce the final expected change station-wise which is then compared to the observed change in the lockdown year: Actual Change = Observed Change -Expected Change Table 2 shows a comparison of lockdown effects on air pollutants from different studies in India. As can be observed from Table 2 , there have been two popular approaches to nail the impact of lockdown. One approach is to compare data from the start of lockdown (week, fortnight, month) to the period before lockdown from the same year. The second approach is to compare the data from the lockdown period to the data from the same period of previous year/years. The problem with the first approach is that the effect of meteorology will not be accounted for in the difference between subsequent weeks as it is highly likely that the impact of temperature, pbl, and moisture have continued to change during the subsequent periods which are being compared. Similarly, the second approach is also problematic as the comparing periods are too distant in time to eliminate the impact of other factors apart from lockdowninduced emission changes. Our approach of estimating the actual lockdown impact aims to improve upon the previous approaches while still maintaining the simplicity in approach and attribution. First, we compare the percentage changes between subsequent weeks/fortnights. This takes care of the fact that we are not comparing too distant times. Next, we make an average of percentage changes during 2015-2019 to ensure we are not taking 1 year but accounting for the effect of different meteorological changes over years. Ideally, a much longer average would be preferable here, but availability of data from different stations limits us to this period only. Next, we compare the average difference of subsequent weeks/fortnights to the similar difference during lockdown year, hoping that the effect of meteorology will be taken care by the average percentage change and the difference between average and the lockdown year will be mainly the lockdown impact.

In Fig. 9 , the blue bar shows the difference between the mean of the data for 11-24 March 2015-2019 and 25 March to 7 April 2015-2019, and the red bar shows us the difference between the mean of the data for 11-24 March 2020 and 25 March to7 April 2020. Here, the blue bar gives us an understanding about the expected change, and the red bar shows the observed change. The actual change can be calculated by taking the difference of the observed change and the expected change. Table 3 also shows the change in values between fortnight prelockdown and the fortnight of the lockdown.

In case of PM 2.5 , for all the stations in northern India, a negative change is observed during the lockdown fortnight. This is also evinced in Figs. 2 and 3 , as well as several earlier studies for different stations in India. Here, we would like to point out that for stations like Delhi, Noida, and Patiala, a decrease during this time is visible even before the lockdown for 2020 as well as the mean of previous years (Fig. 6) . Surprisingly, for these stations, our calculations point to an expected positive change. This is corroborated by the fact that this is the season of widespread biomass burning in this region, but the observed change in all these stations is negative. This means the actual lockdown-induced change is much larger than what a simple difference of corresponding periods shows, a method that has been used in many earlier studies. Similarly, in a dust-dominated western India, viz., Jaipur, Jodhpur, Kota, and Ahmedabad, a positive change was expected, but observed change was negative. The expected change in PM 2.5 over Talcher, Eastern India (east of Chotta Nagpur Plateau), was − 27%, but we observed a positive change of 22%. Thus, the actual lockdown-induced change in Talcher gives us an increase of 49%, much higher than the observed value. Similar behavior is seen in Rajamundry. In coastal India, in most cases, expected changes are negative (due to increasing marine influence) and aligned with the observed change. In Thiruvananthapuram, no change was expected, but the actual change was seen to be a decrease of 27%. In Guwahati, the expected change was negative (− 41%), but the observed change was a mere 3%. Thus, the actual lockdown-induced change in PM 2.5 over Guwahati cumulates to + 44%.

In the case of PM 10 , for all the stations in northern India, a positive change was expected, but the observed changes in all the stations were negative. Thus, the accentuated COVID-induced lockdown effect was actually much more significant in this region. In the dust-dominated western India, Jaipur, Jodhpur, Kota, and Ahmedabad, as well as Solapur, a positive change was expected, but observed change was negative. Dust is a major contributor to both PM 2.5 and PM 10 (Emission sources section, Fig. 2 ). Increasing wind speeds in March compared to February also increase the amount of wind-blown dust, accumulating to the total PM loading. In biomass burning-dominated Guwahati valley in the east of IGP, the expected change in PM 10 was − 40%, but the observed change was 4%; hence, the actual change in Guwahati is + 44%. The negative expected change would be an effect of boundary layer dilution, while observed increase in PM (both 2.5 and10) could be due to an increase in residential activities including residential cooking along with increased fire events (Fig. 7) . The expected change in Talcher in Eastern India (east of Chotta Nagpur Plateau) was − 28%, but we observed a positive change of 35%; thus, the actual change in Talcher gives us an enhanced lockdown impact of 63%. However, in coastal India, change in air masses with higher marine influence would decrease the relative amount of wind-blown dust. In Thiruvananthapuram, only 3% of positive change was expected, but the observed change was a decrease of 13%, and the total change was − 16%. This can be associated with complete change in air mass trajectories after the start of lockdown such that the northwesterly trajectories coming from along west coast of India were replaced by completely marine trajectories from the east (Fig. 5) .

The expected NO 2 change is positive for the northern part of India, but the observed change is negative (Fig. 9) . The expected increase in NO 2 during this period is contributed to some extent due to change in air masses connecting the study locations to pollution plumes (Fig. 10) . The observed decrease in NO 2 is very much on the expected lines as local traffic emission is the major contributor to NO 2 in cities across India. For the IGP region, expected change for Kolkata and Guwahati was negative due to lower influence from IGP and increasing marine influence . The observed change for Kolkata and Guwahati was − 64.03 and − 34.45, respectively, but it must be noted that the whole − 64.03 and − 34.45% decrease cannot be attributed to lockdown, only the difference, i.e., 64.03-47.09 = − 16.94% is the actual lockdown influence for Kolkata and 34.45-21.29 = 13.16% for Guwahati. For Varanasi, the expected change was positive (2.59%), and observed change was positive (69.63%), indicating impact of lockdown. In western India for Jaipur and Kota, expected change and observed change are both negative. However, for Ahmedabad, Jodhpur expected change was positive, while the observed changes were negative. Similar to PM for coastal India, expected and observed changes in NO 2 are negative and aligned. This can be attributed to increasing marine influence as well as dilution. But it would be important to note that the actual change in this case would be much lower than the observed change.

For stations like Nagpur, Chandrapur, Solapur, and Hyderabad, the observed and the actual changes for PM 2.5 were positive. For PM 10 , Chandrapur in the Central India experienced a positive actual change. Pandey and Vinoj (2021) observed that reduction in wind speed, because of converging northwesterly and southeasterly over Central India during the lockdown period, provided a conducive environment for the stagnation of pollutants over the region which led to the increase in AOD (+ 10%, + 20%, and + 18% from Terra, Aqua, and MERRA2, respectively). Madineni et al. (2021) also found that long-range transport and stagnant conditions over Central India led to the increase in AOD during lockdown. Also, local biomass burning and fires associated with agricultural activities led to the enhancement of aerosol concentration over Central India (Bhawar et al., 2021) .

An aerosol source apportionment study in Varanasi by Kumar et al. (2020a) , Kumar et al. (2020b) ) showed that during the months of post-monsoon and winter periods from October to February, the particles are mainly from biofuel and vehicular emissions. With the imposition of lockdown, the main emission source was cut down leading to the observed reduction in ambient concentration of particulate matters. In Varanasi, during March to May, coarse-mode particles dominate, whereas during the months of August and September, transported particles mix along with local emissions.

For coastal regions like Visakhapatnam and Trivandrum, the influence of synoptic features and mesoscale circulations, especially the land and sea breeze circulations, is vital to the advection and dispersion of the pollutants (Remiszewska et al., 2007) . Rajeevan et al. (2019) showed the relationship between wind direction and the aerosol concentration over Trivandrum in the presence of an active sea breeze circulation.

The largest intricacies regarding the impact of lockdown on atmospheric pollutants were observed for O 3 . Because O 3 is a secondary pollutant and its concentrations depend on non-linear chemical interactions, the changes in O 3 were not unidirectional due to lockdown-induced reductions in NO 2 and VOCs. Out of 21 stations analyzed for O 3 , the observed change was positive for only 10 stations, and in the remaining stations, the observed changes were negative. Among these 10 stations, 6 stations are from central and coastal India, including Kolkata close to the Bay of Bengal. Among these 10 stations, 5 stations showed an observed positive change of greater than 25%, these being Delhi, Guwahati, Hyderabad, Talcher, and Tirupati. When we compare the observed change to the actual change, in 8 out of 10 stations, the observed positive changes translate into actual positive changes. Expected O 3 percentage change is positive in the northern part of India due to production from precursors; a significant amount of which can be sourced to biomass burning (Kumar et al., 2011) , but observed changes are negative for most part of northern India except Gobindgarh. But for Delhi expected and observed change, both were positive. For IGP region, expected change for Kolkata was − 24.12, but observed change showed 17.48% enhancement, and for Varanasi, expected change was positive (17.03%), while actual change aligned with observed negative change (− 63.71%). In western India for Ahmedabad, Jodhpur, and Pune, expected changes and observed changes, both were negative, and for Kota, expected and observed change, both were positive. For coastal India, Rajamundry and Tirupati expected and observed changes, both were positive, and for Thiruvanantpuram, both were negative. For Visakhapatnam, expected change was negative, but the observed is positive.

The largest actual positive change in O 3 is over Guwahati where 85% enhancement in O 3 occurred. This can be a result of higher O 3 production due to increase in precursors from increased fire counts (Fig. 10) . The actual change is also above + 40% in Gobindgarh, Kolkata, Talcher, and Tirupati. For Gobindgarh, a positive change in O 3 is associated with a positive change in NO x . However, over Kolkata and Tirupati, positive changes in O 3 are associated with negative changes in NO x . This would be possible if the decrease in NO x over had been sufficient to reach the NO x -sensitive (limited) region where O 3 is not much sensitive to VOCs but increases with increasing NO and increasing HO x . On the other hand, if NO x did not decrease sufficiently and we are in the VOCsensitive regime, O 3 would increase with decreasing NO x and/or increasing VOCs. It is highly likely that most of the O 3 increases occurred in VOC-sensitive regimes. Soni et al. (2021) studied the changes in O 3 in Ahmedabad due to lockdown using a photochemical box model and found that O 3 production was enhanced during the lockdown period due to lower NO x conditions and higher solar irradiance.

The model simulations are described in Table 4 . The average O 3 concentration during peak photochemical period (1300-1500 IST) is also shown in this table. Simulation 1 is a base simulation with only observed concentrations of NO, NO 2 , and CO for 24 March 2020 with CH 4 fixed at 1.85 ppmv. Meteorological parameters for simulation 1 are based on the observed meteorology for the same day over Ahmedabad. Simulation 2 takes all input values of simulation 1 but adds in the effect of C2-C5 anthropogenic VOCs. In this case, we observed an increased O 3 concentration by 20.7%.

Simulation 3 adds in more traffic-related aromatic hydrocarbons (benzene, xylene, toluene); however, these increase O 3 concentration by a further 30.9%, cumulating to 57.9% enhancement over simulation 1. The addition of 1 ppbv isoprene in simulation 4 increases the average peak O 3 by only 8.7% further. However, the addition of pinenes in simulation 5, based on Tripathi and Sahu (2020) PTR-MS measurements over Ahmedabad, increases O 3 by 1.9%. The addition of PAN, H 2 O 2 , and HCHO in simulation 6 increases O 3 further by 192.4% and takes O 3 closest to the observed prelockdown average peak O 3 of 120 μg −3 , the simulation values being only 15.3% lower compared to observed values. The greater O 3 production from biogenic VOCs is due to their higher OH reactivity potentially leading to greater OH recycling (HO 2 → OH) and hence NO → NO 2 → O 3 .

To understand the effect of NO x in this air mass composition over our example city Ahmedabad, we changed the NO x values (maintaining the NO to NO 2 ratio) in our best obtained simulation 6. Decreasing NO x by a factor of 2 in simulation 7 causes a (14.9%) increase in O 3 . Surprisingly, increasing NO x by a factor of 2 decreased O 3 by 13.8% pointing to a VOC-sensitive composition. Simulations 9 is the same as simulation 6 but with anthropogenic hydrocarbons (C2-C5, BXT) reduced by 0.5; this led to a decrease of O 3 by 9.2%. Simulations 10 and 11 are the same as simulation 9, but additionally, biogenic hydrocarbons (isoprene, apinene, bpinene) reduced by 50% and 25%, respectively, leading to 0.15 and 5% increases, respectively, in O 3 . If anthropogenic hydrocarbons are not decreased, but only biogenics are decreased, decrease of O 3 is insignificant. This Reducing anthropogenic VOCs by 0.5 reduces O 3 by 4.4% in simulation 9a. Reducing biogenic VOCs by 50% and 25% in simulations 10 and 11, respectively, with anthropogenics still kept reduced by 50% reduces O 3 by insignificantly. Similarly, reducing only biogenics by 0.5 and 0.25 without reducing anthropogenic changes O 3 negligibly in simulations 12a and 13a. Between 1 and 11a, simulation 9a is the closest to observations on 31 March 2020 with only a difference of 1% on the lower side, while simulation 6a was 3.4% more than observed O 3 . However, simulation 13a with reduced biogenic VOCs is also quite closer to observed O 3 , the difference being 3.3% only. Reducing NO x in a reduced VOC scenario (anthropogenic × 0.5, biogenic × 0.75) reduces O 3 by 7.7% in simulation 14a, while increasing NO x by 2 has a decrease of 5.2% in simulation 15a. It is highly plausible that the reduced biogenic composition represents more appropriate lockdown conditions due to reduced emissions and dilution in an increased PBL. This could be the reason that the reduced VOC composition resulted in a much closer representation to the observed lockdown O 3 .

Lockdowns were a hallmark of the year 2020 having several ramifications including changes in concentrations of atmospheric pollutants. The lockdown-induced changes in atmospheric constituents have been documented at local and regional scales based on surface and satellite measurements. However, the lockdown period in India also coincided with a seasonal change from winter to pre-monsoon. A major shortcoming in several published literature for the Indian region was the overlooking of this seasonal change leading to a spurious attribution of the lockdown effect on atmospheric constituents. A major objective of this paper is to decouple the emission changes during lockdown from meteorological impacts on atmospheric pollutant concentrations. PM, NO 2 , and O 3 from 24 urban regions spanning different emission and climatic regimes in India are studied to estimate the actual impact of lockdown. The actual lockdown-induced change is calculated from the difference of expected and observed changes. For PM, the expected change in North India is positive; meaning in absence of any external forcing like lockdown, the PM concentrations would be enhanced during this period of the year (first fortnight starting with lockdown) compared to the previous period (last fortnight ending before lockdown) as a result of increased biomass burning emissions. However, the lockdown brought down the pollutant levels as outdoor emissions were curbed. The observed decrease is though not the actual impact of lockdown. If the natural increase of concentrations during this period is taken into account, the actual lockdown effect would be much stronger than the observed effect. Similarly, in the dust-dominated western India, a positive change was expected due to enhanced dust loading during this period, but observed change was negative, indicating an accentuated COVID-induced lockdown effect than was observed. In coastal India, in most cases, expected changes are negative (due to increasing marine influence) and aligned with the observed change, so actual lockdown-induced change would be much milder than observed.

For species like NO x which have strong local sources and short residence times, the reduction is much more dramatic. The observed decrease in NO 2 is very much on the expected lines as local traffic emissions are the major contributor to NO 2 in cities across India. For eastern endpoint of IGP, e.g., for Kolkata, the expected change was a decrease (− 47.09%) due to lower influence from IGP as well as reduction in local emissions and increasing marine influence. The observed change for Kolkata is also a negative (− 64.03%), but it must be noted that the whole 64% decrease cannot be attributed to lockdown, only the difference, i.e., − 16.94%, is the actual lockdown influence. An enhancement in O 3 was observed at many stations immediately after lockdown, viz., Patiala, Delhi, and Guwahati. This can happen in a VOCsensitive regime, where a decrease in NO x would lead to enhancement in O 3 for the same VOC level. VOC-sensitive regimes generally occur in urban areas where the rate of NO x production is much larger compared to the OH production rate. In other places like Kolkata, Talcher, Jalandhar, Hyderabad, and Tirupati, O 3 was already increasing before the lockdown. This is aligned with the expected positive percentage changes in O 3 . In the Northern part of India, the positive expected change is a result of production from biomass burning, but observed changes are negative for most part of northern India due to lower levels of NO 2 except Delhi and Patiala, where NO 2 levels are higher compared to other north Indian stations selected for this study.

Box model simulations for an example station (Ahmedabad) using available NO x and CO data and approximating hydrocarbon data from different years showed that anthropogenic VOCs (C2-C5, BZT) increase O 3 by 57.9% during prelockdown and 41.3% during lockdown. However, C2-C5 VOCs have a lower contribution in lockdown O 3 enhancement compared to BZT. Biogenic hydrocarbons like isoprene and pinenes do not have much impact on increasing O 3 during either prelockdown or lockdown. The addition of PAN, HCHO, and H 2 O 2 also increases O 3 significantly both during prelockdown and during lockdown.

Using simulation 6 and 6a as base for prelockdown and lockdown, i.e., keeping the original level of VOCS, increasing NO x by a factor of 2 causes significant O 3 decrease during both prelockdown (13.8%) and lockdown (7.1%), while decreasing NO x by a factor of 2 increases prelockdown O 3 by 14.9% and lockdown O 3 by 9.8%, indicating slightly higher response to NO x during prelockdown period. Similarly, decreasing anthropogenic VOCs by a factor of 2 reduces prelockdown O 3 by 9.2% and lockdown O 3 by 4.3%. Reducing biogenic VOCs by half reduces O 3 negligibly both during prelockdown and lockdown. Increasing NO x by 2 and decreasing NO x by 0.5 at a reduced VOC composition caused prelockdown O 3 to change by − 14.7% and 7.6%, respectively, while lockdown O 3 changed by − 7.2% and 9.8%, respectively. The fact that increasing NO x promotes O 3 reduction (opposing effect) both for the prelockdown (24 March) and lockdown (31 March) for Ahmedabad example while changing VOCs had an accompanying impact on O 3 points to a VOCsensitive composition. In the absence of actual hydrocarbon measurements, detailed simulation of chemistry to understand the O 3 changes during the lockdown 274 Page 28 of 31

Vol:. (1234567890) period would be a challenging task, and any attribution of causes to the O 3 results remain speculative at best at this point. In the end, it can be concluded that the actual lockdown-induced changes in atmospheric pollutants are different from what is represented by simple dent in a time series. For pollutants like PM and O 3 in North India, the natural enhancement due to biomass burning which camouflaged the severity of lockdown-induced decreases. For coastal stations like Visakhapatnam, the lockdown effect is much milder due to prior dilution and ventilation effects.

Source apportionment of PM2.5 & PM10 concentrations of Delhi NCR for identification of major sources. The Energy and Resources Institute

Evaluated kinetic and photochemical data for atmospheric chemistry: Volume II &ndash; gas phase reactions of organic species

Assessment of air quality changes in the four metropolitan cities of India during COVID-19 pandemic lockdown

Reconstructing primary and secondary components of PM2.5 composition for an urban atmosphere

Sources and atmospheric dynamics of organic aerosol in New Delhi, India: Insights from receptor modeling. Atmospheric Chemistry and

Radiative impacts of aerosols during COVID-19 lockdown period over the Indian region

COVID-19 lockdown and its impact on tropospheric NO2 concentrations over India using satellite-based data. Heliyon

Characteristics of ventilation coefficient and its impact on urban air pollution

Avoiding high ozone pollution in Delhi

COVID-19 lockdown: Impact on air quality of three metro cities in India

Changes in air quality during the COVID-19 lockdown in India. ISEE Conference Abstracts

Distribution of volatile organic compounds over Indian subcontinent during winter: WRF-chem simulation versus observations

Technical Specifications for Continuous Ambient Air Quality Monitoring (CAAQM) Station (REAL TIME)

National Ambient Air Quality Monitoring NAAQMS/45/2019-2020, Central Pollution Control Board, Ministry of Environment, Forest and Climate Change

Importance of meteorology and chemistry in determining air pollutant levels during COVID-19 lockdown in Indian cities

Environmental Science: Processes and Impacts

Short term unwinding lockdown effects on air pollution

Influence of air pollution on coronavirus (COVID-19): Some evidences from studies at AUH, Gurugram

Effect of COVID-19 lockdown on the concentration and composition of NR-PM2.5 over Ahmedabad, a big city in western India

Sub-region (district) and sector level SO2 and NOx emissions for India: Assessment of inventories and mitigation flexibility

Detection of surface emission hot spots, trends, and seasonal cycle from satellite-retrieved NO2 over India

Source apportionment of PM2.5 in North India using source-oriented air quality models

Nature of air pollution, emission sources, and management in the Indian cities

Air pollution knowledge assessments (APnA) for 20 Indian cities. Urban Climate

Reviewing the crop residual burning and aerosol variations during the COVID-19 pandemic hit year 2020 over

The ERA5 global reanalysis

Spatial and temporal variation of BTEX in the urban atmosphere of Delhi

relea se_ pdf/ Press_ Relea se_ 05_ March. pdf IMD Reports (b) Public Weather Bulletin. India Meteorological Department Press Release

Phase-wise analysis of the COVID-19 lockdown impact on aerosol, radiation and trace gases and associated chemistry in a tropical rural environment

The MCM v3.3.1 degradation scheme for isoprene

Influence of super cyclone "Amphan" in the Indian subcontinent amid COVID-19 pandemic

Effect of meteorological parameters on spread of COVID-19 in India and air quality during lockdown. Science of the Total Environment

Influences of the springtime northern Indian biomass burning over the central Himalayas

Assessment of atmospheric aerosols over Varanasi: Physical, optical and chemical properties and meteorological implications

Long-term MAX-DOAS measurements of NO2, HCHO, and aerosols and evaluation of corresponding satellite data products over Mohali in the Indo-Gangetic Plain

Large unexplained suite of chemically reactive compounds present in ambient air due to biomass fires

Impact of lockdown measures during COVID-19 on air quality-A case study of India

Covid-19 and air pollution in Indian cities: World's most polluted cities

Loss of crop yields in India due to surface ozone: An estimation based on a network of observations

The contribution of outdoor air pollution sources to premature mortality on a global scale

Natural processes dominate the pollution levels during COVID-19 lockdown over India

Effect of lockdown amid COVID-19 pandemic on air quality of the megacity Delhi

Oxidation processes in the eastern Mediterranean atmosphere: Evidence from the modelling of 

Acknowledgements We thank the reviewers for their constructive comments which have greatly improved the quality of the MS. We thank the editor for getting the MS reviewed in difficult times of the pandemic.

The first author acknowledges the support from Harithasree Sreedevan in identification of events.

The air quality data was downloaded from the CPCB website: https:// app. cpcbc cr. com/ ccr/#/ caaqm-dashb oard/ caaqm-landi ng/ caaqm-compa rison-data. The PM sources are based on analysis presented in UrbanEmissions.info. PBL height is taken from ERA5 reanalysis. While the raw data is freely available on the above mentioned sites, the datasets generated during and/or analyzed during the current study are available from the corresponding author on request. O 3 simulations are carried out using FOAM box model: https:// sites. google. com/ site/ wolfe gm/ models.

Competing interests The authors declare no competing interests.