key: cord-0832789-0n4r9nfm
authors: Sathe, Yogesh; Gupta, Pawan; Bawase, Moqtik; Lamsal, Lok; Patadia, Falguni; Thipse, Sukrut
title: Surface and Satellite Observations of Air Pollution in India during COVID-19 Lockdown: Implication to Air Quality
date: 2020-12-29
journal: Sustain Cities Soc
DOI: 10.1016/j.scs.2020.102688
sha: c376520901d876c7222778b5b27d0120bd9d170e
doc_id: 832789
cord_uid: 0n4r9nfm

The strict nationwide lockdown imposed in India starting from 25(th) March 2020 to prevent the spread of COVID-19 disease reduced the mobility and interrupted several important anthropogenic emission sources thereby creating a temporary air quality improvement. This study conducts a multi-scale (national-regional-city), multi-species, and multi-platform analysis of air pollutants and meteorological data by synergizing surface and satellite observations. Our analysis suggests a significant reduction in surface measurements of nitrogen dioxide (NO(2)) (46-61%) and fine particulate matter (PM(2.5)) (42-60%) during the lockdown period that are also corroborated by the reduction in satellite observed aerosol optical depth (AOD) (3-56%) and tropospheric NO(2) column density (25-50%) data over multiple cities. Other species, namely coarse particulate matter (PM(10)) (24-62%), ozone (22-56%) also showed a substantial reduction whereas carbon monoxide (16-46%), exhibited a moderate decline. In contrast, sulfur dioxide (SO(2)) levels did not show any defined reduction trend but rather increased in Mumbai, Bengaluru, and Kolkata. The temporary air quality improvement achieved by the painful natural experiment of this pandemic has helped demonstrate the importance of reducing emissions from other sectors along with transportation and industry to achieve the national air quality targets in the future.

The Novel Corona Virus Disease initially identified in Wuhan of China in 2019 have caused significant disruption with more than 16.56 million infections worldwide (until July 29 th , 2020; WHO, 2020a) . To control the exponential spread of COVID-19, Government of India announced a nation-wide lockdown starting midnight of 25 th March 2020 and it was extended till 17 th May 2020 in a phase-wise manner, putting its 1.3 billion citizens inside their homes. There was a strict ban on the movement of people during this lockdown and activities such as transportation (including road, rail and air), construction, and industries were completely restricted, except essential services such as medical facilities, electricity, supply of vegetables and fruits. Power plants were also functional during the lockdown period but at a reduced scale due to the decline in the electricity demand in industrial and commercial units (CRISIL, 2020) . The effect of lockdown measures was soon reflected in the mobility with notable changes in mobility index statistics (Table S1 , Google, 2020) . The mobility to outside places such as parks and workplaces significantly reduced while the residential category showed an upsurge in all the Indian states as a result of nation-wide lockdown.

Although these restrictions were primarily meant to flatten the COVID-19 infection curve, they created a unique experiment to assess the effect of anthropogenic activities on air pollution. Similar to India, many other countries have imposed certain limitations and observed lockdown with a varying magnitude of restrictions and implementation. The reduction in anthropogenic activities worldwide has fortified several researchers around the globe to investigate the environmental effects of the lockdown. A study by Tobías et al. (2020) observed a significant reduction in NO2 (47.0% to 51.0%), BC (~45.0%) and PM10 (27.8% to 31.0%) concentrations in Barcelona, Spain during the lockdown period (March 14 th to March 30 th , 2020) compared to the pre-lockdown period (February 16 th to March 13 th , 2020); however, ozone (O3) concentrations exhibited an opposite trend and reported an increase ranging from 28.5% to J o u r n a l P r e -p r o o f 57.7%. A significant reduction of traffic-related pollutants such as PM10, PM2.5, BC, benzene, CO, and NOx, is reported in the city of Milan in Italy during the lockdown period (Collivignarelli et al., 2020) . United States (U.S.) counties also observed an overall decrease in NO2 (25.5%) and PM2.5 (11.3%) concentrations (Berman and Ebisu, 2020) . Whereas, Bekbulat et al. (2020) showed that the PM2.5 and ozone pollution levels during the lockdown did not drop consistently across U.S. attempted to simulate the impact of lowered human activities on air quality changes using chemical transport model such as WRF-CAMx in Yangtze River Delta Region in China and reported a considerable decrease in pollutant concentrations of PM2.5, NO2 and SO2 compared to the year 2019. Liu et al. (2020) analysed NO2 tropospheric vertical column density (TVCD) from the Ozone Monitoring Instrument (OMI) over China and reported an abrupt decline in NO2 TVCD up to 48% during the lockdown period.

Space agencies such as European Space Agency (ESA) and National Aeronautics and Space Administration (NASA) have initially reported a significant reduction in NO2 and aerosol levels over India (ESA, 2020a; NASA, 2020) in their respective early satellite observations releases. A number of peer-reviewed studies have recently reported change in pollutants over India using multiple surface, satellite and model analysis. The extent of the analysis published in these studies varies and there still is a need to provide a comprehensive analysis supported by high-quality data sets. Sharma et al. (2020) estimated the effect of restricted human activities in India and reported an improvement in Air quality index (15% to 44%) in different parts of India. The study only reported the first twenty days (until April 14 th , 2020) of changes and focused on 22 major cities using ground data alone. Similarly, Jain and Sharma (2020) reported similar changes in five major cities for the first twelve days of the lockdown and only considered 2019 data as baseline concentrations. Navinya et al. (2020) extended this period until May 3, and covering 17 cities and using 2019 only data for baseline calculations. Singh J o u r n a l P r e -p r o o f and Chauhan (2020) used data primarily from U.S. embassy locations in five cities on PM2.5, and analyzed OMI NO2 data at coarser resolution (0.25 degree) and consider period until April 14, 2020. The study also compared 2020 values with 2019 to estimate changes. Similarly, Kumar et al. (2020) used data from U.S. embassy locations in India from 25 March to 11 May for 2015-2020 to investigate PM2.5 reduction during the lockdown period. A more recent study by Pandey and Vinoj (2020) utilized satellite and AERONET measurements of aerosols along with MERRA-model reanalysis data and concluded that the populated Indian regions showed a decline up to ~40% while central India showed an unexpected increase up to 20% during 24 th March to 22 nd April 2020. The study used 1x1 degree coarse resolution MODIS AOD data and concluded that an increase in AODs in central India is partially associated with high relative humidity. All these early studies reported changes in pollutant concentration for partial lockdown period and primarily used data from selected cities only. Studies which used satellite data are limited in using coarser resolution data. Also, most studies compared only 2019 and 2020 measurements to estimate changes.

The emissions sources seem common in major cities, but their magnitude is highly variable in space and time. In addition to emissions, meteorology, physical and chemical processes in the atmosphere significantly influence the air quality in a given region. Hence, any changes in air quality require analysis at different scales vis-à-vis changes in emissions, meteorology and atmospheric chemistry. The ground measurements of air pollutants have been traditionally used to investigate the changes in air pollution levels. Although the ground measurements have an advantage of high temporal resolution, they lack the spatial coverage and typically represent near-surface area surrounding the monitoring site. Conversely, the satellite measurements (polar-orbiting satellites) have an advantage of greater spatial coverage but provides an instantaneous measurement at a particular location. When used together, both these J o u r n a l P r e -p r o o f 6 measurement techniques can complement each other in terms of spatial and temporal limitations.

Therefore, in this study, we attempt to fill some of the gaps in earlier published studies as outlined above. We specifically address these gaps by a) extending the analysis to the three phases of complete lockdown period (i.e. March 25 to May 17) in India; b) using three years of data (i.e. 2017-2019) to form a baseline; c) using a large network of stations in six major cities and for all the criteria pollutants, d) using high-resolution pixel-level satellite observations; e) extending the analysis from national to regional to local scales, and f) finally by discussing the impact of lockdown on air quality changes and its implication for the longterm air quality management policies in India.

In this study, we analysed the impact of lockdown measures at different spatial scales i.e. national, regional and local (city, and within the city). The national and regional scale analysis was aimed at determining the changes in atmospheric loading of aerosols and columnar NO2 observations using satellite data products during the lockdown period. The national scale analysis considered entire India as study domain while the regional analysis was performed over three selected areas (see boundaries in Fig. 1(C) ) i.e. Indo-Gangetic plain (IGP), Central Southern India (CSI) and South India (SI), which observed a significant change as shown in satellite retrieved aerosols and NO2 measurements. However, a synergistic approach utilizing satellite and surface measurements was used to assess the impact of lockdown measures at local scale. Six metropolitan cities in India ( Fig. 1 

Air pollutants including PM2.5, PM10, NO2, SO2, O3 and CO, are routinely monitored in 6 selected cities using continuous ambient air quality monitors. We obtained the hourly concentrations data from the Central Pollution Control Board (CPCB) database (CPCB, 2020a) . A total of 51 monitoring stations with at least one year of valid measurements from 1 st January to 17 th May from 2017 to 2019 were selected to define the baseline conditions. The data availability for the same time period during the present year i.e. 2020 was also ensured before finalizing these stations. The number of monitoring stations with valid data varied significantly in each city (Table S2 , and S3). The measurement errors in data obtained from the CPCB monitoring stations were typically reported to be less than 5% (CPCB, 2011; Hama et al., 2020) . We obtained the hourly concentrations data from CPCB screened it to eliminate invalid and outliers (abrupt changes) points to quality control the data. The quality-assured hourly data are used to calculate 24-hourly and/or 8-hourly mean concentrations for further analysis.

The surface meteorological parameters are routinely monitored at the international airports in selected cities and are reported to the Integrated Surface Database (Smith et al., 2011; NOAA, 

MODIS aboard two NASA's Earth Observing Satellites (EOS) Terra and Aqua have been providing aerosol observations since 2000 and 2003, respectively. Aerosol optical depth (AOD) at 550 nm retrieved using MODIS observations has been used to address various climate and air quality research questions. The operational AOD data from MODIS is currently retrieved using multiple algorithms (Levy et al., 2013; Gupta et al., 2016; Hsu et al., 2013; Lyapustin et al., 2018) ; in this analysis, we used a high resolution (0.1x0.1 degree) level 3

AODs, which combined Dark Target (DT), Deep Blue (DB) retreivals from MODIS-Aqua satellite. The validation study over India shows the correlation between MODIS and AERONET larger than 0.9 with a mean bias of 0.05 and 70% retrieval within predefined uncertainty envelope . We have also analysed the MODIS-Terra data for consistency check and arrived at very similar results; MODIS-Terra analysis is not presented in this paper. The daily gridded high-resolution AOD data were then used to understand the impacts of lockdown on the atmospheric loading of aerosols in the study region.

The Dutch and Finnish-built Ozone Monitoring Instrument (OMI) on-board NASA's Aura satellite is a nadir-viewing ultraviolet-visible spectrometer, measuring the solar-backscattered radiation in the 270-500 nm range (Levelt et al., 2006 (Levelt et al., , 2018 . Launched in July 2004, the OMI/Aura has been providing data from October 2004 to present from a sun-synchronous polar

The ground resolution of the L2 product varies across swath ranging from 13 km x 24 km (along-track x across-track) at swath center to 24 km x 160 km at the outermost edges of the swath. We used data with cloud fraction less than 30% that are not affected by the OMI row anomaly (Dobber et al., 2008) . To create L3 product, a day's worth of clear-sky (effective cloud fraction < 0.3) good quality L2 data are mapped into a regular 0.1° latitude by 0.1° longitude grid. NO2 column data in any grid cell is an area-weighted average of all L2 data that have any overlap with that grid cell.

To quantify the impact of lockdown on atmospheric loading of pollutants, we performed following analysis: a) day-to-day variability using time-series, b) mean percentage changes in pollutants during the lockdown, c) spatial distribution of anomalies at national, regional, and local scales, d) change in air quality indices and its implications. We chose to use the past three years of data (2017 to 2019) for the same period to derive baseline conditions assuming it to represent atmospheric loading of pollutants in the year 2020 in absence of lockdown. Although longer time-series of data are available, to avoid the effect of the temporal trend in pollution, we generated baseline using data from these three years only. Thus, the baseline period is defined as March 25-May 17, 2017 to 2019 (referred as "baseline") and March 25-May 17, 2020 is considered as lockdown period (referred as "lockdown"). Air Quality Indices (AQI) in each city were also analysed during the baseline and lockdown periods. AQI is a measure that relates air quality to human health exposure and is derived by translating the weighted concentrations of individual pollutants (Ott, 1978; CPCB, 2015) .

We first analyzed satellite data of AOD and columnar NO2 to understand large scale changes over the region. Figs. 1 (A-C) each shows the AOD for baseline, the lockdown and their differences. AOD is a measure of light extinction by scattering and absorption from aerosol particles as the light passes through the earth's atmosphere. The AOD is always measured at a specific wavelength and represents columnar loading of atmospheric aerosols from the surface to top-of-the-atmosphere. There have been many studies (Gupta et al., 2006; Liu et al., 2009; van Donkellar et al., 2019; Sathe et al., 2019; Yang et al.,2019; Christopher and Gupta 2020) over the years to demonstrate the relationship between AOD and surface mass concentration of PM2.5. In this study, we have not used AOD to derive PM2.5 but used it as a surrogate to assess the relative change in particle pollution in the atmosphere due to lockdown. The white areas in the Figs. 1 (A-C) are areas with missing data from the satellite due to various retrieval limitations such as clouds, snow/ice on the surface or complex topography. The baseline map ( Fig. 1(A) ) represents the typical aerosol distribution over India for the season. The mean AOD ranges from 0.05 to ≥ 1 with elevated AODs (>0.5) in the IGP, CSI, and over the seas adjoining India. Contrary to the previous 3 years, in the year 2020, the maximum AOD is ≤ 0.75 ( Fig.   1 (B)). The IGP region and the SI have strikingly low AOD values of ≤ 0.25. However, CSI maintained high seasonal mean AOD ranging from 0.3 -0.75 in 2020 as well. These high values are partially related to crop residual burning. After the harvesting of the wheat crop, crop residue burning is typically observed during March to May (pre-monsoon) in some of Indian states when fields are prepared for monsoon crops (Hays et al., 2005; Vadrevu et al., 2011; Liu et al., 2018) .

. 1(C) shows the AOD anomaly (2020 minus the baseline). Compared to the baseline, the AOD in the year 2020 dropped across the country except over CSI. A drastic decrease in AOD is seen over the IGP and SI regions with AOD anomaly exceeding -0.20 with a 25% drop in J o u r n a l P r e -p r o o f regional mean AOD. The mean percentage decrease in AOD over SI is 30% while the mean change in CSI is insignificant (~1%). Satellite data shows fires (not shown here) in CSI throughout the months of March and April (www.firesandaq2020.info). The elevated AODs in this region can be attributed to these burning activities in the region. A more recent study explained these elevated AODs partially associated with an effect of relative humidity on hygroscopic aerosols (Pandey and Vinoj, 2020) . In addition, the north-western part of India and neighbouring countries demonstrated an overall decline in AOD values with some positive random anomalies. The overall changes observed in the atmospheric aerosols loading were mainly caused by a) reduced emissions due to lockdown, b) change in meteorological conditions, c) emissions and transport of dust and smoke.

Figs. 1(D-F) show average OMI tropospheric NO2 column density during the lockdown period compared with average NO2 columns over the same period of the baseline. Tropospheric NO2 data exhibit a strong spatial variation with higher values over cities, industrial areas, and power plants. While the spatial patterns remain consistent, NO2 data during lockdown are significantly lower than the baseline values. Fig. 1 (F) shows changes in tropospheric NO2 columns between these two periods. The nation-wide decline in NO2 of about 19% is consistent with the strict lockdown that affect various activities linked to NOx emissions. The observed differences could arise from i) reduced NOx emissions during lockdown; ii) inter-annual variation in NOx emissions during 2017-2020; and iii) changes in NO2 columns driven by changes in weather that affect NO2 lifetime, transport, and concentrations. We further discuss these in subsequent sections.

The prevailing meteorological conditions have a substantial influence on air quality and it is considered as one of the key drivers of air quality after emissions in a given region. For example, change in winds can either disperse the pollutant to the larger area (i.e., high winds) or it can stagnant pollutant in the smaller area (i.e., calm winds). Similarly, depending on wind direction, pollution can transport towards cities (or human settlements or monitors) or away from cities. Boundary layer height is another important parameter, which impacts the concentration of a pollutant at the surface, usually, well-mixed boundary layers (deeper layer) dilute the pollution in larger volume thus reducing the mass concentration at the surface and vice-versa. Chemical and physical processes in the atmosphere help in the production of secondary pollutants such as sulphate and nitrate aerosols and it can be affected by temperature and relative humidity. Certain pollutant's lifetime in the atmosphere also gets affected by meteorological conditions. For instance, NO2 has a much longer lifetime in colder temperatures because of lower concentrations of OH and RO radicals. The longer wintertime NOx lifetime and shallow planetary boundary layer (PBL) causes the build-up of NO2 concentrations Therefore, it is important to account for meteorology while change in mass concentrations are studied to evaluate reduction in emissions due to limited human activities during the lockdown period. The meteorological parameters observed during the lockdown period were compared against the baseline period. Fig. 3 shows the frequency distribution while Table S5 summarizes the descriptive statistics of meteorological parameters. The meteorological parameters during this period of year typically represent spring-to-summer time conditions with moderate to high temperatures and wind speeds, increased mixing depth and lower frequency of temperature inversions, which are in turn favourable for pollutant dispersion in the atmosphere. The change in meteorological conditions from winter to spring, usually favour lower pollution levels in spring as compared to winter months. This impact of change in weather conditions reflects decreased pollutant concentration during spring from winter (Fig. 4) . In general, no significant J o u r n a l P r e -p r o o f change is observed in meteorological parameters during the lockdown compared to the baseline period. The mean Relative humidity (RHUM) was observed to increase from 34.2% to 45.4% in Delhi and from 42.1% to 50.7% in Pune during the lockdown period compared to baseline.

In addition to RHUM, the dry bulb temperatures (TEMP) in Delhi were also lower by ~2.3 0 C in the lockdown compared to the baseline period, but changes in other cities were not significant. The wind-rose diagrams (Fig. S8) during the baseline period with some variability. This suggests that meteorology may not be a dominating factor in the reduced pollution levels seen during the lockdown. We have also noted certain specific observed changes in pollutant concentrations throughout the paper where weather (i.e. rain) had a significant impact. It is noteworthy that similarity of histograms suggests that the overall impact of meteorology for the entire study period could be similar during the baseline and lockdown years, but the day-to-day and diurnal variation may be different in the two periods; this can create different effects. This study does not aim to separate the impact of meteorology and lockdown on pollutant concentrations, but rather uses the meteorological parameters for the qualitative interpretation of changes in pollutants. A more advanced analysis with model simulations may be needed to isolate individual effects.

Figs. 4 and 5 shows the day-to-day evolution of surface and satellite measurements starting from 1 st January to 17 th May for the previous (2017-19) and present (2020) (Fig. 4) shows that most of the primary pollutants exhibit strong seasonality and an overall decreasing trend from January to May. One of the causes for the pollution build-up during January -February is meteorological conditions i.e. low temperature, high relative humidity, and low mixing height; these favour the formation and accumulation of secondary particles (Pan et al., 2015; Schnell et al., 2018) . These conditions then gradually change to typical summer-time unstable conditions associated with higher wind speeds that are conducive for dispersion of pollutants.

PM2.5 (Figs. 4(A-B) ) show an overall declining trend from January to May in all the cities (except Bengaluru). Satellite-based AOD showed similar trends as surface PM2.5 observations ( Fig. 5(A) ), except that AOD exhibited more day-to-day variability and trends are less steep.

The trends are stronger in surface measurements. These differences can arise from the fact that AOD is column value and averaged over a larger area as compared PM2.5, which are at the surface and represent a point measurement. In Bengaluru, PM2.5 concentrations changed substantially during the lockdown period, in comparison to the baseline conditions, with daily PM2.5 concentrations plummeting up to 13.5 µg m -3 . These were the lowest values observed in the city over the last 4 years during this specific period. The AOD also decreased to about half of its previous year's value. Although, Delhi recorded the high particulate concentrations in J o u r n a l P r e -p r o o f the month of January, the PM2.5 concentrations during year 2020 were found to be lower (by ~14%) than baseline before the imposition of lockdown (Table S7) showed an obvious decreasing trend with a considerable reduction during the lockdown period.

The AOD over these cities are generally lower than in January but with no significant trend.

Baring Mumbai and Pune, there is a reduction in AOD, especially during the first phase of lockdown. From mid-January till May, there is an overall decrease in AOD during the baseline years. However, over both Mumbai and Pune, the AOD in 2020 is lower than previous years in both March and May with an increase in AOD after 1 st week of May. This increasing trend (in PM2.5 and AOD) towards the end of the lockdown period could be due to the relaxations permitted during second (15 th April to 3 rd May 2020) and third (4 th May to 17 th May 2020) phases of the lockdown in certain parts of the country.

As shown in (Figs. 6(C-D)), with the exception of Mumbai all the cities showed an obvious reduction in both PM2.5 and PM10 concentrations during the lockdown period. The transport sector is the major contributor to PM2.5 in cities such Bengaluru, Delhi, and Pune, and accounts for 35-40% of PM2.5 emissions while industries contribute 49-70% in remaining cities (i.e., Chennai, Kolkata and Mumbai) (Table S6) . Besides, a ban on the operation of industries except J o u r n a l P r e -p r o o f essential goods manufacturing and construction/demolition activities led to further reductions in PM2.5 concentrations during the lockdown period. The decline in PM2.5 during the lockdown was in range of 41% (Kolkata) to 60% (Chennai). The satellite-based aerosol measurements (i.e. AOD, (Fig. 6(A) ) showed similar results with the highest change (56%) for Bengaluru and lowest (3%) for Mumbai. It is important to note that AODs are spatially averaged over larger regions and represent an entire column of the atmosphere under ambient conditions as compared to PMs representing the surface mass concentration of certain size particles, under controlled conditions. Mumbai showed a moderate increase in PM2.5 concentrations (42%) compared to the baseline period. Furthermore, the satellite retrieved mean AOD values (a proxy for surface PM2.5) over Mumbai also remained substantially unchanged during the lockdown period (AOD reduction: ~3%, Fig. 6(A) ). This phenomenon can be explained by three possible causes: i) the lack of data in years 2017 and 2018 to form the "baseline" conditions, thereby missing the inter-annual variations, ii) an increase in household coal combustion as people spent more time indoors and there was a likely increase in the cooking and heating activities during the lockdown period and iii) the photochemical oxidation of SO2 to sulphate particles by hydroxide free radical due to the presence of high OH concentration during summer (Khoder, 2002; Megaritis et al., 2013) . The results from Mumbai are unexpected and require further analysis with more station data and meteorological variables. Kumar et al. (2020) also observed the least PM2.5 reduction (~12%) at U.S. Embassy monitoring station in Mumbai during the lockdown period compared to other Indian cities such as Delhi, Chennai, Kolkata and Hyderabad and cited coal/biomass combustion as the primary reason.

The coarse particulates i.e. PM10, exhibited an overall decreasing trend from January to May owing to seasonal changes during both present and previous years, except for Bengaluru that showed a slightly increasing trend. The concentrations in all cities experienced a significant reduction (Fig. 6 ) ranging from 24% (Mumbai) and 62% (Pune). The re-suspended and construction dust are the major contributors to PM10 emissions in these cities with a share ranging from 28 to 61% (Table S6) . Hence, the main reason for decline could be the reduction in Vehicular Kilometres Travelled (VKT) which not only prevents the exhaust emissions but also reduces non-exhaust emissions (tyre, brake and clutch wear) and re-suspended road dust. Further, the absence of construction activities during the lockdown period also resulted in lowered PM10 concentrations.

Both surface and columnar (OMI) NO2 concentrations in cities such as Bengaluru, Delhi, Kolkata and Pune exhibited strong seasonality with an overall downward trend from January to May during previous and present years (Figs. 4(C) and 5(B)). The seasonality in NO2 is primarily driven by changes in lifetime and mixing layer depth. For example, in winter NOx lifetime is long and mixing layer depth is shallow causing the build-up of NO2 concentrations while in summer the situation reverses. The strict lockdown measures (starting 25 th March, 2020) led to a substantial reduction in NO2 concentrations (Fig. 6(E) ) in all six cities (46-61%), with Delhi showing the maximum reduction. The transport sector is a major contributor to NOx emissions in cities such as Delhi (85%), Bengaluru (42%), Kolkata (25%), Mumbai (29%) and

Pune (29%) while the industries accounted for 28-58% of total NOx emissions in cities such as Bengaluru, Kolkata, Mumbai and Pune (Refer Table S6 ). Hence, the NO2 reductions in these J o u r n a l P r e -p r o o f cities can be explained by the restrictions on the movement of people and ban on industrial operations imposed during the lockdown period. Furthermore, the NO2 reductions in Indian cities are comparable to the reductions observed in European cities (52%) and Wuhan (57%) in China (Sicard et al., 2020) . These results are consistent with the reductions in tropospheric NO2 columns (25% to 50%) observed by OMI (Fig. 6(B) ).

The ozone (O3) concentrations showed different day-to-day behaviour and did not follow a consistent trend across different cities throughout the analysis period ( Fig. 4(E) ), but concentrations throughout 2020 were comparatively lower than the previous year's mean except for Kolkata and Pune. Kolkata reported higher than baseline ozone concentrations throughout the year 2020 while Pune had higher ozone concentrations till 3 rd week of March 2020. During the lockdown a significant reduction in mean O3 concentrations (22 to 56%) was observed (Fig. 6(G) ) as compared to the baseline years in different cities except for Kolkata, which showed an enormous increase (66%). It is important to note that ozone is secondary pollutant formed as a result of complex and non-linear interplay between NOx, Volatile Organic Compounds (VOCs) and solar radiation (Kroll, et al., 2020; Menut et al. 2020) . In polluted environments, carbon monoxide (CO) can also produce ozone through similar reactions with OHradicals as the VOCs. The reduction in emissions of NOx and VOCs in cities during the lockdown could be assumed as the primary reason for the overall reduction in ozone concentrations, but the meteorology and chemical interactions need further investigation. Further, a slight increase in VOC emissions can be expected especially in COVID-19 like situation when the community disinfection practices such as spraying of sodium hypochlorite solution (Chatterjee, 2020) , frequent use of sanitizers, etc. were followed rigorously within the cities to control the spread of the disease. This situation typically represents a NOx-limited chemical regime and an overall decrease in O3 concentrations can be J o u r n a l P r e -p r o o f expected. In Kolkata, however, a contradictory scenario was observed, where NOx emissions were reduced, consistent with NO2 reduction of 47%, during the lockdown period. However, there is a possibility of the existence of a VOC-limited chemical regime, thereby enhancing the in O3 formation.

Similar to O3, SO2 (Fig. 4(D) ) in the cities also showed a mixed behaviour, and SO2 levels concentrations. Both Chennai and Delhi have more than 85% of SO2 emissions originating from the industrial sector, which includes thermal power-plants (Table S6 ). The ban on the operation of industries and partial reduction in power-plant operations can explain the SO2 reduction in these cities. As discussed earlier, the mean SO2 concentrations in Pune remained more or less constant (Percentage change: ~+4%) during the two periods while Bengaluru and Kolkata experienced a moderate increase of 31% and 42%, respectively. These three cities had seen higher SO2 concentrations throughout the year 2020 as compared to previous years mean.

In Bengaluru and Kolkata, the inclined SO2 concentrations could be attributed to primary emissions from thermal power-plants and background SO2 contributions from surrounding regions (ESA, 2020b) . Surprisingly, Mumbai showed the highest increase in mean SO2 concentrations (81%) during the lockdown period in comparison to the baseline period, and the concentrations during the lockdown period ranged from a minimum of 20.69 µg m -3 to a peak value of 90.07 µg m -3 . The examination of time-series revealed a substantial increase in daytime SO2 concentrations during 26 th April to 14 th May 2020 at Bandra, Mumbai. A slight J o u r n a l P r e -p r o o f increase was observed in PM2.5 and CO concentrations during this period. There can be multiple reasons for this increase: i) more than 90% of SO2 emissions in Mumbai region are thought to be originated from the industrial sector and some heavy industries such as power and refineries were allowed to operate at reduced capacity during the second and third phases of the lockdown (Hindustan Times, 2020) ; ii) idling ships were spotted in the ports near Mumbai (Hindustan Times, 2020); iii) baseline concentrations could be lower due to noninclusion of 2017-18 SO2 data (Section 3.3.2); and iv) people spent more time indoors during the lockdown period, thereby increasing the residential sector emissions due to more cooking and heating activities (Kumar et al., 2020) .

With the exception of Pune, all the cities showed a moderate reduction of 16 to 46% in CO concentrations during the lockdown period in comparison to the baseline period. The reduction in CO can be attributed to the limitation of combustion activities in sectors such as transport, industries, and residential (Refer Table S6 ), wherein CO is a primary pollutant resulting from an incomplete combustion process. However, Pune showed a large increase of 66% in CO concentrations as compared to the baseline years. In the view of the reduction in anthropogenic activities during the lockdown, such an increase in CO concentrations seems rather unexpected and suggests suspicious CO measurements, maybe due to a faulty calibration of the sensor and needs further investigation.

Air quality varies from location to location in and around the city and depends on sources, topography, land use type, and meteorological conditions. To evaluate changes in different parts of the city, we choose Delhi, where there are 34 ground monitors located, to map the J o u r n a l P r e -p r o o f spatial distribution of three selected pollutants namely PM2.5, NO2, and O3. Fig. 7 provides maps of mean concentrations for baseline (A, C, E) and lockdown (B, D, F) periods. The PM2.5 values for the baseline period (Fig. 7(A) ) varied from 60 to 121 g m -3 whereas it dropped to 24 -74 g m -3 during the lockdown. The average PM2.5 reduction during the lockdown period ranged from 18% (at ITO station) to 80% (at Shadipur station) in different parts of the city. Shadipur station is surrounded by numerous small and medium scale industries. Industries are mostly associated with heavy-duty diesel vehicle (HDDV) traffic for transportation of raw and processed materials to and fro this location. Hence, shut down of industrial operations along with associated HDDV traffic could explain the reason for maximum PM2.5 reduction during the lockdown period. ITO square, on the other hand, is dominated by huge daily traffic flow and is surrounded by Government offices which were allowed to operate partially during the second and third phases of the lockdown. Typically, PM2.5 values were higher in the northern part of the city and lower in the southern part of the city.

NO2 concentrations showed high values in the centre of the city and decreased as one moves away from the centre to the outside of the city. This may be primarily due to the presence of dominating sources of NO2 such as vehicles and thermal power plants in the inner part of the city. The transportation sector alone contributes to more than 85% of NOx emissions in Delhi city (ARAI and TERI, 2018) . Traffic restrictions imposed during lockdown had reduced NOx emissions in Delhi city and it is reflected in lowering the NO2 concentrations. NO2 reduction over stations ranged from 20% to 80% across the city.

Ozone concentrations exhibited high spatial variability and showed an overall mixed change in concentrations. For example, 17 stations showed an increase while remaining 17 stations showed reduction in ozone concentrations during the lockdown period in comparison to the baseline years. Maximum increase of 244% in O3 concentrations was observed at ITO square (Percentage reduction in NO2 > 73%) while maximum reduction of 85% was observed at J o u r n a l P r e -p r o o f Jahangirpuri (percentage increase in NO2 ~4%). The major source of NO and NO2 (i.e., NOx) in cities is vehicular emissions, and their ratio is about 20:1. NO plays a critical role in ozone chemistry including its destruction in the atmosphere. In the absence of NO, O3 concentration can go up; a probable effect was seen over Delhi with a net increase in O3 during the lockdown.

To examine the footprints of lockdown induced emission reduction at regional scales, we analyzed the satellite-derived AOD and NO2 data. The area-averaged daily AOD and NO2 values during January-May of 2020 were compared to the baseline period. From the spatial analysis in section 3.1, we choose to perform regional analysis over three regions (box enclosed with solid black lines in Fig. 1(C) ) that show large AOD and NO2 anomalies. Figs. 8 shows 7day moving average time-series starting January 1 through May 17 for previous years and 2020.

Home to millions, the IGP has the highest population density in the country, as well as house to several industries, power plants, and major agricultural areas. Emissions from all of these sources contribute to regional air pollution. IGP is also unique for its geography and topography with the Himalayas on the north side, Chota Nagpur Plateau in the south, and Iranian Plateau in the west. NO2 column values tend to be higher than 2020 even prior to the lockdown period, suggesting that there were some declines in NO2 during the period of our study (2017) (2018) (2019) (2020) although prior J o u r n a l P r e -p r o o f studies (e.g. Duncan et al., 2016; Krotkov et al., 2017) suggested a considerable positive trend in NOx emissions in India. During the lockdown, there was an additional decline in NO2 of 15% in SI, 23% in CSI, and 28% in IGP, which are consistent with the trends observed at the ground sites discussed in the previous section.

This section discusses the air quality indices (AQI) calculated in six cities during the baseline and lockdown periods and its implication to short and long-term national air quality goals and future directions. The AQI is multi-pollutant AQI and details on the method and the formulation are provided in the supplementary material (Refer SM19). Figs. 9 (A-F) shows the distribution of six AQI categories calculated using the monitoring data during the baseline and the lockdown periods. The AQIs in these cities were mainly driven by PM10 followed by PM2.5 and occasionally by CO and O3. This was also evident from the exceedance of National Ambient Air Quality Standards (NAAQS) in these cities (not presented here).

An examination of AQI during the baseline period revealed that air qualities in these cities were mainly distributed in Satisfactory (60-84%) and Moderate (15-38%) AQI classes, except for Delhi. Delhi's air quality during the same period was mainly distributed between Moderate (39%), Poor (38%), and Very Poor (21%), and seldom attained the Satisfactory (1%) level. The air quality situation improved significantly during the lockdown period when the combined proportion of Good and Satisfactory AQI classes in Bengaluru (100%), Chennai (100%), Pune (96%), Kolkata (92%) and Mumbai (87%) were substantially higher as compared to the baseline period. Although Delhi's air quality very rarely made to the Good (4%) category, it attained Satisfactory (48%) and Moderate (46%) levels for majority of cases during the lockdown period. (MoEFCC, 2019) . The NCAP is primarily aimed at reducing the national level PM2.5 and PM10 concentrations by 20-30% by the year 2024, as compared to 2017. PM2.5 has been a matter of serious concern in India due to its complex sources and health effects. It is important to note that, when emissions from vehicles, industry and construction activities were switched off during the lockdown, the PM2.5 levels improved by 40-60% in different cities. Hence, to reduce PM2.5 below the targeted levels, the remaining sources such as thermal power plants, diesel generators, residential biomass burning, open burning of waste, and trans-boundary contribution should also be given a due consideration.

The positive impact on air quality due to the reduction of air pollutants in cities is likely temporary and will last till the lockdown restrictions are in place. Substantial improvements in the air quality are evident by this unique experiment, wherein some of the sources were turnedoff in real-time, provided a guiding example for governments and policymakers to take long term preventive measures in the future.

The ongoing pandemic created a unique natural experiment for the first time in the modern industrial era. The natural experiment provided an opportunity to environmental and Earth science community to assess the impact of reduced human activities on many different aspects of the Earth-atmosphere system including air, water, and land pollutions.

In response to covid-19 pandemic, India implemented a nation-wide complete lockdown that lasted for 53 days (March 25 to May 17, 2020) in three different phases. The lockdown created a unique environmental condition where emissions of air pollutants from human activities were either completely shut down or limited. Air pollution measurements from surface and satellite monitors showed a sharp decline in many parts of the country. In this study, we provided a J o u r n a l P r e -p r o o f comprehensive analysis of surface and satellite measurements of air pollutants for the lockdown period comparing them with observations during the recent years of 2017-2019. We also quantified the decline in air pollutants at varying spatial (community, city, regional, and national) and temporal (hourly, daily, and seasonal) scales. We used data from two space sensors MODIS and OMI, and 52 ground stations located in 6 major cities in India.

The followings are important findings of our analysis:

-Surface concentrations of PM2.5, PM10, NO2, SO2, CO, and O3 from ground monitors, and AOD and NO2 columns from satellite sensors consistently showed lower values during the lockdown as compared to the baseline period.

-Satellite AOD data recorded the highest decline in the IGP region, where population density and level of pollution are highest in the region. The observed AODs were among the lowest in the region in the two decade long data record.

-Periodic dust storms and biomass burning in northern India and neighbouring countries offset some of the declines in atmospheric aerosols concentrations in the IGP and larger northern India.

-Central Southern India reported an increase in AOD during the lockdown period as compared to the baseline period. These increases were mostly associated with increases in biomass burning and favourable meteorological conditions in the region.

-Surface measurements show similar meteorological conditions in both lockdown and baseline years. Overall decline in pollution was driven by reduction in emissions.

-PM2.5, PM10 and AOD exhibited the lowest decline (41%, 24%, and 3%) in Kolkata and Mumbai (PM10 and AOD) whereas the highest declines of 60%, 57%, and 56% were reported in Chennai, Delhi, and Bengaluru cities, respectively.

-NO2 from the ground stations showed the lowest decline of 46% in Chennai city and highest (61%) in Delhi city whereas OMI reported 28% decline over the IGP region.

J o u r n a l P r e -p r o o f -O3 concentrations in the cities observed an overall decline during the lockdown period (22 to 56%), except in Kolkata, which recorded a significant increase (66%). These changes were mainly due to complex photochemical reactions with limiting substances such as NOx and VOCs. Several individual stations have also reported an increase in O3 during the lockdown.

-Diurnal profile of pollutants during lockdown exhibited similar patterns as the baseline period but with a substantially reduced magnitudes.

-The station-level data measured in various communities in Delhi showed spatial variability in pollutant concentration and varying scale of reduction during the lockdown period PM2.5 (18 to 80%) and NO2 (23 to 82%). The complex O3 chemistry resulted in a mixed effect with 18 stations recording an increase (7 to 244%) while remaining showing a decline (11 to 85%).

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

for previous years i.e. 2017-19 (Std. deviation) and the blue line indicates 7-day moving average for the present year i.e. 2020. 

We would also like to thank National Centers for Environmental Information (NCEI) of National Oceanic and Atmospheric Administration (NOAA), USA for making available surface meteorological observations through

Source Apportionment of PM2.5 & PM10 of Delhi NCR for Identification of Major Sources: Final

Ozone Air Pollution Levels Have Not Dropped Consistently Across the US Following Societal Covid Response

Changes in U.S. air pollution during the COVID-19 pandemic

Use of Hypochlorite Solution as Disinfectant during COVID-19

Outbreak in India: From the Perspective of Human Health and Atmospheric Chemistry

Global Distribution of Column Satellite Aerosol Optical Depth to Surface

Lockdown for CoViD-2019 in Milan: What are the effects on air quality

Air quality monitoring, emission inventory and source apportionment study for Indian cities: National Summary Report. Central Pollution Control Board

National Air Quality Index, Control of Urban Pollution Series

CPCB (2020b). IMPACT OF LOCKDOWN (25th March to 15th April) ON AIR QUALITY

Central Pollution Control Board, Ministry of Environment, Forest and Climate Change

COVID-19 corollaries

The high-resolution solar reference spectrum between 250 and 550 nm and its application to measurements with the Ozone Monitoring Instrument

A space-based, high-resolution view of notable changes in urban NOx pollution around the world

5P/Sulphur_dioxide_concentrations_drop_over_India_during_COVID-19

Satellite remote sensing of particulate matter and air quality assessment over global cities

High Resolution Gridded Level 3 Aerosol Optical Depth Data from MODIS, to be submitted to Geoscience Data Journal

Enhanced Deep Blue aerosol retrieval algorithm: The second generation

Social and Travel Lockdown Impact Considering Coronavirus Disease (COVID-19) on Air Quality in Megacities of India: Present Benefits, Future Challenges and Way Forward

Atmospheric conversion of sulfur dioxide to particulate sulfate and nitrogen dioxide to particulate nitrate and gaseous nitric acid in an urban area

The complex chemical effects of COVID-19 shutdowns on air quality

The version 3 OMI NO2 standard product

Temporary reduction in fine particulate matter due to 'anthropogenic emissions switch-off' during COVID-19 lockdown in Indian cities

Nitrogen Dioxide Standard Product with Improved Surface and Cloud Treatments (in review)

The Ozone Monitoring Instrument

The Collection 6 MODIS aerosol products over land and ocean

Air quality changes during the COVID-19

lockdown over the Yangtze River Delta Region: An insight into the impact of human activity pattern changes on air pollution variation

Impact of city lockdown on the air quality of COVID-19-hit Wuhan city

Estimating Regional Spatial and Temporal Variability of PM2.5 Concentrations Using Satellite Data, Meteorology, and Land Use Information

Abrupt decline in tropospheric nitrogen dioxide over China after the outbreak of COVID-19

Seasonal impact of regional outdoor biomass burning on air three Indian cities: Delhi, Bengaluru, and Pune

MODIS Collection 6 MAIAC algorithm

Air pollution characteristics and their relationship with emissions and meteorology in the Yangtze River Delta region during 2014-2016

Impact of lockdown measures to combat Covid-19 on air quality over Western Europe

Ministry of Environment, Forest & Climate Change

Airborne Particle Levels Plummet in Northern India

Examining effects of the COVID-19 national lockdown on ambient air quality across urban India. Aerosol and Air Quality Research

Environmental indices theory and practice

A multi-model evaluation of aerosols over South Asia: common problems and possible causes

Surprising increase in aerosol amid widespread decline in pollution over India during the COVID19 Lockdown

Temperature and recent trends in the chemistry of continental surface ozone

Application of Moderate Resolution Imaging Spectroradiometer (MODIS) Aerosol Optical Depth (AOD) and Weather Research Forecasting (WRF) model meteorological data for assessment of fine particulate matter (PM2.5) over India

Exploring the relationship between surface PM2.5 and meteorology in Northern India

Atmospheric Chemistry and Physics: From Air Pollution to Climate Change

Effect of restricted emissions during COVID-19 on air quality in India

Amplified ozone pollution in cities during the COVID-19 lockdown

Impact of lockdown on air quality in India during COVID-19 pandemic

The Integrated Surface Database: Recent Developments and Partnerships

Changes in air quality during the lockdown in Barcelona (Spain) one month into the SARS-CoV-2 epidemic

Spatial patterns in vegetation fires in the Indian region

MODIS derived fire characteristics and aerosol optical depth variations during the agricultural residue burning season