How Does COVID-19 Lockdown Impact Air Quality in India?

: Air pollution is a severe environmental problem in the Indian subcontinent. Largely caused by the rapid growth of the population, industrialization, and urbanization, air pollution can adversely affect human health and environment. To mitigate such adverse impacts, the Indian government launched the National Clean Air Programme (NCAP) in January 2019. Meanwhile, the unexpected city-lockdown due to the COVID-19 pandemic in March 2020 in India greatly reduced human activities and thus anthropogenic emissions of gaseous and aerosol pollutants. The NCAP and the lockdown could provide an ideal ﬁeld experiment for quantifying the extent to which various levels of human activity reduction impact air quality in the Indian subcontinent. Here, we study the improvement in air quality due to COVID-19 and the NCAP in the India subcontinent by employing multiple satellite products and surface observations. Satellite data shows signiﬁcant reductions in nitrogen dioxide (NO 2 ) by 17% and aerosol optical depth (AOD) by 20% during the 2020 lockdown with reference to the mean levels between 2005–2019. No persistent reduction in NO 2 nor AOD is detectable during the NCAP period (2019). Surface observations show consistent reductions in PM 2.5 and NO 2 during the 2020 lockdown in seven cities across the Indian subcontinent, except Mumbai in Central India. The increase in relative humidity and the decrease in the planetary boundary layer also play an important role in inﬂuencing air quality during the 2020 lockdown. With the decrease in aerosols during the lockdown, net radiation ﬂuxes show positive anomalies at the surface and negative anomalies at the top of the atmosphere over most parts of the Indian subcontinent. The results of this study could provide valuable information for policymakers in South Asia to adjust the scientiﬁc measures proposed in the NCAP for efﬁcient air pollution mitigation.


Introduction
The fast growth in the population has led to rapid industrialization and urbanization in developing countries during the past several decades, which has resulted in considerable increases in anthropogenic emissions of gases and particulates and, consequently, exacerbates air quality [1,2]. As the world's second most populous country, India has experienced a severe air pollution problem in the past two decades [3], which can have significant harmful impacts on human health [4,5]. The World Health Organization (WHO) reported that nine of the world's top ten most polluted cities were from India [6]. Moreover, 99.5% of the 640 districts in India exceeded the WHO guideline for annual mean Particulate Matter (PM 2.5 ) concentration (i.e., 10 µg/m 3 ) in 2016 [7]. Deteriorating air quality may have caused about 1.54 million premature mortality per year in India alone [2].
To mitigate and prevent air pollution, the Indian government launched the National Clean Air Programme (NCAP) in January 2019 [8]. The NCAP aims to augment the For carbonaceous aerosols, their relative changes were positive over the high-altitude Himalayan region, which was caused by the enhanced formation of secondary OC through photochemical reactions involving biogenic emissions [27].
At present, most studies are mainly focused on the changes in air pollution levels for specific cities during the lockdown period over India, and an overview of the changes in aerosols and their effects over the entire Indian subcontinent are still lacking (e.g., radiative forcing). Meanwhile, such large-scale changes in aerosol characteristics have the potential to modulate the radiation budget through direct and indirect radiative effects and subsequently impact the regional climate, so detailed investigations are needed [18,28]. Further, meteorological conditions are known to significantly influence regional air quality, and changes in aerosol characteristics also reversely impact meteorology, but such impacts during COVID-19 remain under-investigated. Additionally, most previous studies focused on a shorter period, i.e., starting from 16 March to 14 April 2020, which did not coincide with the actual lockdown period. In this study, we utilize multiple satellite products and surface observations to investigate and compare the changes in air quality over the Indian subcontinent due to NCAP and the 2020 lockdown. The entire lockdown period (March 24-June 30) was considered here. By contrasting aerosol changes during 2020 and 2019, we aim to quantify the following: (1) the contribution of human activities to the total column and surface aerosol abundances, then (2) the potential impacts of meteorological conditions on air quality, and (3) radiation responses to COVID-19 emission reductions.

OMI
The Ozone Monitoring Instrument (OMI) onboard the Aqua satellite is a nadir-viewing solar backscatter. It measures solar irradiance and Earth radiance from 270 to 500 nm (ultraviolet (UV) to visible (VIS)) at high spectral and spatial resolutions with daily global coverage [29]. The entering light is split into the following two channels using a scrambler: the UV channel with the range 270-380 nm and the VIS channel that covers 350-500 nm. OMI provides the column tropospheric amounts of trace gases (i.e., NO 2 ) and ozone. Generally, NO 2 has a short lifetime and is primarily emitted from anthropogenic sources, including industries, powerplants, transportation, and residential combustion [30,31]. It serves as a key precursor for both secondary aerosol formation and ozone production [31][32][33]. NO 2 is often used to indicate surface air quality during specific events, such as the 2008 Olympic Games in Beijing [34], and the 2014 Asia-Pacific Economic Cooperation summit in Beijing [35]. Here, the level-3 daily global gridded tropospheric NO 2 at 0.25 • × 0.25 • from OMNO2d is used as a surrogate for human activities. The OMI tropospheric NO 2 has been shown to correlate well with ground-based and in-situ NO 2 measurements and bottom-up emission inventories [36]. In this study, tropospheric NO 2 retrieved from OMI was used to quantify the contribution of human activities.

MODIS
The MODIS instruments onboard the Aqua and Terra satellites observe the Earth system in 36 spectral bands ranging from 0.4 to 14.4 µm and provide a nearly global coverage within 1 to 2 days owing to their wide swath of 2330 km [37]. MODIS AOD is retrieved from the deep blue (DB) algorithm over bright land (e.g., desert) and dark target (DT) algorithms over both vegetated lands and waters [38][39][40][41][42]. In MODIS collection 6.1, the DB algorithm is updated to produce a dynamic surface reflectance dataset depending on the normalized difference vegetation index (NDVI); the DT algorithm is updated to reduce the biases in urban areas based on a surface reflectance model [43]. In MODIS collection 6.1, a "merged" dataset of AOD at 550 nm is produced by combining the DT and DB retrievals to increase the data spatial coverage [44,45]. In this study, the merged daily AOD data with the resolution of 1 • × 1 • between 2005-2020 are used to analyze aerosol changes.

CERES
The Clouds and the Earth's Radiant Energy System (CERES) instruments measure Earth's radiation budget and cloud properties in 15 shortwave (SW, 0.2 to 4.0 µm) and 12 longwave (LW, 2.850 µm to 1 cm) spectral bands. The radiation fluxes at the top of atmosphere (TOA) and the Earth's surface are assessed using delta-two stream radiation transfer model [46] under clear-sky (without clouds and aerosols) and cloudy-sky (with clouds and aerosols). The TOA radiation fluxes are derived using satellite-derived aerosol and cloud properties together with a radiative transfer model and observed radiances together with angular distribution models [47]. The surface fluxes are computed with cloud properties derived from geostationary satellites (GEO) and MODIS [47]. In this study, daily radiation flux data from 2005 to 2020 are used to quantify radiation response to emission reductions. Flux datasets are derived from Level 3 SYN1deg products with a spatial resolution of 1 • × 1 • [48]. Here, we have considered radiation fluxes at the surface, in the atmosphere (computed as a residual term), and at the TOA.

Surface Observation
Surface observational data are retrieved from 230 operational stations in Indian National Air Monitoring Network (https://app.cpcbccr.com/AQI_India/, accessed on 31 November 2021). Generally, the air quality is continuously monitored by sophisticated instruments. Therefore, the chemical method is used to measure SO 2 and NO 2, and the high-volume sampler is being widely used for particulate matter measurement. In this study, hourly surface concentrations (units: µg/m 3 ) of PM 2.5 , PM 10 , NO 2 , NH 3 , SO 2 , CO, and O 3 were used. Based on the availability of measurements from 2015 to 2020, observations from seven cities were selected for analysis-Ahmedabad, Bengaluru, Hyderabad, Mumbai, Lucknow, Chennai, and Delhi to quantify the contribution of human activities to surface air pollution.

ERA5
The fifth generation ECMWF reanalysis (ERA5) is the latest generation of atmospheric reanalyses of the global climate. It can provide dozens of commonly used atmospheric and land-surface variables with temporal coverage from 1950 to now [49]. Many studies have proved that ERA5 can provide reasonable temporal and spatial variability of meteorological fields (i.e., wind and precipitation) on a large scale by assimilating remote sensing data, atmospheric sounding data, and ground-based observations. Also, ERA5 with reasonable temporal and spatial variability can be used as the background input of the proposed correction-downscaling model. In this study, variables of the precipitation, wind speed, planetary boundary layer and relative humidity from ERA5 are used to analyze the impact of meteorological conditions on air pollution levels in India. Although the meteorological field from ERA5 is shown reasonable in the spatiotemporal distributions, we also further compare the wind speed, planetary boundary layer, and relative humidity with the corresponding variables from the Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2), and the precipitation from the Global Precipitation Measurement (GPM).

Method
India reported the first COVID case on 30 January 2020, and the number of confirmed cases increased to 519 on March 24 (Table 1). To curb the spread of COVID-19 in the Indian subcontinent, a strict lockdown was enforced. The first phase was officially announced on 24 March 2020 and lasted for 21 days from 25 March to 14 April 2020. The lockdown was further extended to 30 June 2020. Here, the following two periods are defined: the prelockdown period (1 January-23 March) and the lockdown period (24 March-30 June). As the city locks down, the aerosol mass changes obviously due to the reduction of anthropogenic emissions. To quantify aerosol changes in the Indian subcontinent, daily anomalies of AOD and tropospheric NO 2 are calculated from 1 January to 30 June in 2019 and 2020 based on the fifteen-year climatology mean (2005-2019). Corresponding daily anomalies of meteorological fields, and radiation fluxes are also calculated. The anomalies in 2019 and 2020 represent NCAP-induced and city-lockdown-induced changes, respectively. The anomalies during 2019 and 2020 are further compared for two periods-pre-lockdown and lockdown, to investigate the potential impacts of COVID-19 lockdown on air quality and radiation. Note that although pre-lockdown and lockdown periods are defined for the year of 2020, these two periods are also selected for 2019 for comparison purposes. Additionally, the Indian subcontinent is divided into the following three sub-regions: North, Central, and South (Figure 1b), based mainly on population density. lockdown period (1 January-23 March) and the lockdown period (24 March-30 June). As the city locks down, the aerosol mass changes obviously due to the reduction of anthropogenic emissions. To quantify aerosol changes in the Indian subcontinent, daily anomalies of AOD and tropospheric NO2 are calculated from 1 January to 30 June in 2019 and 2020 based on the fifteen-year climatology mean (2005-2019). Corresponding daily anomalies of meteorological fields, and radiation fluxes are also calculated. The anomalies in 2019 and 2020 represent NCAP-induced and city-lockdown-induced changes, respectively. The anomalies during 2019 and 2020 are further compared for two periods-prelockdown and lockdown, to investigate the potential impacts of COVID-19 lockdown on air quality and radiation. Note that although pre-lockdown and lockdown periods are defined for the year of 2020, these two periods are also selected for 2019 for comparison purposes. Additionally, the Indian subcontinent is divided into the following three subregions: North, Central, and South (Figure 1b), based mainly on population density.

Changes in Satellite Retrieved Tropospheric NO 2 and AOD
Dramatic reductions in air pollution levels during the 2020 lockdown can be observed by satellite retrieved concentrations of NO 2 in the troposphere. The main sources of tropospheric NO 2 are transportation and industrial activities, thus tropospheric NO 2 is a good indicator of anthropogenic emissions [30,31]. The spatial distributions of climatological (2005-2019) tropospheric NO 2 are also shown in Figure 1a,b for the pre-lockdown and lockdown periods. The spatial distributions of tropospheric NO 2 concentrations generally follow population density as follows: the highest concentrations over the Indo-Gangetic Plain (IGP) are found in the North Indian subcontinent, followed by the Central Indian and lower concentrations over the South Indian subcontinent.
An analysis of tropospheric NO 2 anomalies strongly suggests significant reductions in anthropogenic emissions in 2020 during the lockdown period (Figure 1c-f). In 2019, tropospheric NO 2 anomalies did not show spatially consistent differences in any of the three aforementioned sub-regions during either pre-lockdown or lockdown periods, implying that NCAP did not make substantial differences in air quality during the first half of 2019. However, in 2020, tropospheric NO 2 anomalies during the pre-lockdown period showed negative values in large areas of Central and East India, suggesting reductions in anthropogenic emissions likely due to NCAP. Moreover, tropospheric NO 2 in 2020 showed larger (~2 times) reductions during the lockdown period than during the pre-lockdown period across most regions of the Indian subcontinent, with the largest reductions in IGP, followed by the Central and South subcontinent. A total reduction of 7.1% (-0.006 DU) in tropospheric NO 2 concentrations was detected over the entire three sub-regions in comparison with the 2005-2019 climatology.
The time series of tropospheric NO 2 during the pre-lockdown and lockdown periods over the three sub-regions is shown in Figure 2a-c (left). To better show the trend in tropospheric NO 2 , we calculated the de-trended tropospheric NO 2 over the three sub-regions. Tropospheric NO 2 levels in 2020 are the smallest due to the reductions in anthropogenic emissions during the lockdown period (light-red dots), in which the regional mean tropospheric NO 2 is 0.067, 0.076, and 0.045 DU ( Figure S3). In comparison with the averaged tropospheric NO 2 of 2015-2019 ( Figure S3), tropospheric NO 2 over the three sub-regions changed by -0.011 (-13.9%, North), -0.016 (-18.5%, Central), and -0.011 DU (-23.1%, South) during the lockdown, respectively. Also, the interannual variability of regional mean tropospheric NO 2 during the pre-lockdown period over the three sub-regions is similar to that of the lockdown period. However, tropospheric NO 2 in 2020 over the three sub-regions changed by 0.002 (2.9%, North), -0.012 (-12.5%, Central), and -0.008 (-14.9%, South) DU during the pre-lockdown, which may be caused by the reductions in anthropogenic emissions due to NCAP. Further, the time series of daily mean tropospheric NO 2 over the three sub-regions are shown in Figure 3a-c (left). The daily mean tropospheric NO 2 in 2019 is close to climatology during most of the days in the three sub-regions. However, the daily mean tropospheric NO 2 in 2020 is far below climatological levels during the lockdown period in all three sub-regions, despite that it is close to climatological levels during the pre-lockdown period over the North and South Indian subcontinents, with generally lower levels in the Central Indian subcontinent. Also, the reductions in tropospheric NO 2 are more apparent during the first three to four weeks of the lockdown. It is worth noting that daily mean tropospheric NO 2 in 2020 during the lockdown period in all three sub-regions is possible within the fifteen-year spread of tropospheric NO 2 (gray shadings), which is the minimum and maximum tropospheric NO 2 levels between 2005-2019, and they can include larger uncertainties of tropospheric NO 2 . With the reductions in anthropogenic emissions during COVID-19 lockdown, tropospheric NO 2 in the 2020 lockdown period is expected to decrease significantly. Moreover, the tropospheric NO 2 and AOD are almost out of phase in all presented regions during the lockdown period, which could be partially attributed to the influences of natural aerosol loadings (i.e., dust aerosol) on AOD [50].
shadings), which is the minimum and maximum tropospheric NO2 levels between 2005 2019, and they can include larger uncertainties of tropospheric NO2. With the reduction in anthropogenic emissions during COVID-19 lockdown, tropospheric NO2 in the 202 lockdown period is expected to decrease significantly. Moreover, the tropospheric NO and AOD are almost out of phase in all presented regions during the lockdown period which could be partially attributed to the influences of natural aerosol loadings (i.e., dus aerosol) on AOD [50]. In addition to tropospheric NO2 concentrations, the satellite-retrieved AOD is an other useful variable for investigating persistent haze issues, which can well represent th aerosol loadings. Figure 4a,b shows the climatology of the averaged AOD during the pre lockdown and lockdown periods. It shows that the high AOD occurs over North Indi along the southern slope of the Himalayas, which is similar to the spatial feature of trop ospheric NO2 climatology. In order to better understand the improvement in air qualit due to city-lockdown, AOD anomalies are illustrated in Figure 4c-f. Significantly, positiv AOD anomalies are generally observed over the Indian subcontinent in 2019 and 202 during the pre-lockdown period. The regional mean of AOD anomalies over the thre Indian sub-regions is about 0.03 (North), 0.07 (Central), and 0.001 (South) in 2019, respec tively ( Figure S3). However, during the lockdown period, the AOD anomalies in 2019 tur into negative values, especially over the Indus plain. The reduction of aerosol loading ove In addition to tropospheric NO 2 concentrations, the satellite-retrieved AOD is another useful variable for investigating persistent haze issues, which can well represent the aerosol loadings. Figure 4a,b shows the climatology of the averaged AOD during the pre-lockdown and lockdown periods. It shows that the high AOD occurs over North India along the southern slope of the Himalayas, which is similar to the spatial feature of tropospheric NO 2 climatology. In order to better understand the improvement in air quality due to city-lockdown, AOD anomalies are illustrated in Figure 4c-f. Significantly, positive AOD anomalies are generally observed over the Indian subcontinent in 2019 and 2020 during the pre-lockdown period. The regional mean of AOD anomalies over the three Indian sub-regions is about 0.03 (North), 0.07 (Central), and 0.001 (South) in 2019, respectively ( Figure S3). However, during the lockdown period, the AOD anomalies in 2019 turn into negative values, especially over the Indus plain. The reduction of aerosol loading over the Indus plain is mainly induced by the decrease in dust aerosol loading, as shown in a previous study [50]. In 2020, however, the decrease in aerosol loading is mainly observed over the Indo Gangetic Plain and the Ganges Delta, which are mainly induced by anthropogenic pollution emissions. Those larger negative AOD anomalies are closely related to the reduction of emissions from the traffic and manufacturing sectors, which are expected to be substantially impacted by the city lockdown [18,19]. The consistent positive anomalies of tropospheric NO 2 and AOD during the lockdown period of 2020 implies that the dramatic reductions in tropospheric NO 2 and other anthropogenic emissions played more important roles than the variations of meteorological conditions in leading to negative AOD anomalies. Quantitatively, in 2020, AOD was reduced by about 22.6% (0.05, North), 2.9% (0.11, Central), and 20.1% (0.02, South) over the three Indian sub-regions during the lockdown period, which is about 1.7 (North), 1.6 (Central), and 20 (South) times of that in 2019, respectively. pogenic pollution emissions. Those larger negative AOD anomalies are closely related the reduction of emissions from the traffic and manufacturing sectors, which are expecte to be substantially impacted by the city lockdown [18,19]. The consistent positive anom lies of tropospheric NO2 and AOD during the lockdown period of 2020 implies that th dramatic reductions in tropospheric NO2 and other anthropogenic emissions played mo important roles than the variations of meteorological conditions in leading to negativ AOD anomalies. Quantitatively, in 2020, AOD was reduced by about 22.6% (0.05, North 2.9% (0.11, Central), and 20.1% (0.02, South) over the three Indian sub-regions during th lockdown period, which is about 1.7 (North), 1.6 (Central), and 20 (South) times of that 2019, respectively.  (Figure 3d-f). The reason is attributed to the more contribution of natural sea salt aerosol and dust emitted from the Thar Desert and eastward tran ported to AOD [50], and enhancement of the photolysis of NO2 due to the warmer an  (Figure 3d-f). The reason is attributed to the more contributions of natural sea salt aerosol and dust emitted from the Thar Desert and eastward transported to AOD [50], and enhancement of the photolysis of NO 2 due to the warmer and more humid conditions [51,52]. During the pre-lockdown period in 2020, AOD was almost located within the climatological ranges of the three sub-regions. However, during the lockdown period in 2020, a significant decrease in AOD was observed over the North and South Indian subcontinent, especially in the first few weeks of the lockdown period. Over the Central Indian subcontinent, close-to-climatology AOD is found in the first few weeks during the lockdown period. This unexpected high AOD is induced by positive anomalies in mid-tropospheric relative humidity [18]. Overall, the decreases of AOD induced by city-lockdown are the most significant over North Indian subcontinent, where population is the densest and thus anthropogenic emissions are the highest. Also, daily AOD within the fifteen-year spread is naturally due to the larger range of AOD variation in the past fifteen years. located within the climatological ranges of the three sub-regions. However, d lockdown period in 2020, a significant decrease in AOD was observed over the N South Indian subcontinent, especially in the first few weeks of the lockdown per the Central Indian subcontinent, close-to-climatology AOD is found in the first f during the lockdown period. This unexpected high AOD is induced by positive a in mid-tropospheric relative humidity [18]. Overall, the decreases of AOD in city-lockdown are the most significant over North Indian subcontinent, where p is the densest and thus anthropogenic emissions are the highest. Also, daily AO the fifteen-year spread is naturally due to the larger range of AOD variation in fifteen years.  However, th reduction in the tropospheric NO2 is over Ahmedabad (city 1, 0.1 DU and 40%), by city 2 (0.037 DU and 30%) and city 5 (0.036 DU and 33%) ( Figure S4a). For creases are detected in all cities except for city 4, with the highest decrease b Ahmedabad (city 1, 70 µg/m 3 , and 58%). These observations are generally consis the changes in satellite retrieved AOD, which also show a significant decreas (Figure 5c). The highest reduction in column AOD is observed over Delhi (city 7 25%), followed by city 1 (0.13 and 19%) and city 2 (0.13 and 37%) ( Figure S4a). G PM2.5 concentrations decrease by about 39 µg/m 3 (52%) over North India, 20 µg

Changes in Gaseous Emissions near the Surface
Observed surface concentrations of NO 2 , PM 2.5 , O 3 , SO 2 , and CO from surface stations are also utilized to demonstrate the impact of COVID-19 on air quality. Due to data availability, data from seven stations operated by the Pollution Control Board (CPCB, https://www.cpcb.nic.in/, accessed on 31 November 2021) are selected across India from 2015 to 2020 ( Figure 5). Data anomalies during the lockdown period (with reference to 2015-2019) are shown in Figure 5. For NO 2 , reductions are observed in all cities, with the highest reduction over Bengaluru (city 2, 22.8 µg/m 3 and a 48.3% reduction), followed by city 1 (14.3 µg/m 3 and 39.8%) and city 7 (15.4 µg/m 3 and 50.7%). However, the highest reduction in the tropospheric NO 2 is over Ahmedabad (city 1, 0.1 DU and 40%), followed by city 2 (0.037 DU and 30%) and city 5 (0.036 DU and 33%) ( Figure S4a). For PM 2.5 , decreases are detected in all cities except for city 4, with the highest decrease being over Ahmedabad (city 1, 70 µg/m 3 , and 58%). These observations are generally consistent with the changes in satellite retrieved AOD, which also show a significant decreasing trend (Figure 5c). The highest reduction in column AOD is observed over Delhi (city 7, 0.16 and 25%), followed by city 1 (0.13 and 19%) and city 2 (0.13 and 37%) ( Figure S4a). Generally, PM 2.5 concentrations decrease by about 39 µg/m 3 (52%) over North India, 20 µg/m 3 (45%) over Central India, and 27 µg/m 3 (41%) over South India, respectively. As for O 3 , reductions are observed in all cities except for cities 1, 2, and 6. The increases in O 3 in these three cities may be related to the variations of VOCs and other factors that can influence the production and/or consumption of tropospheric O 3 . For SO 2 and CO, the largest reductions were seen in Bengaluru, a pattern consistent with the reductions in NO 2 and PM 2.5 . While over other cities, the changes in SO 2 and CO are much smaller. Overall, the surface observed changes in NO 2 and PM 2.5 are consistent with the reductions in satellite retrieved NO 2 and AOD. Additionally, the relative changes of meteorological conditions within the same time periods show that precipitation (except Bengaluru and Delhi), wind speed, and planetary boundary layer (except Hyderabad) decrease, while the relative humidity increases (except Hyderabad) ( Figure S4e). Moreover, the largest reduction in precipitation is observed over Hyderabad (city 3, 1.46 mm/day and 40%), followed by city 4 (1.31 mm/day and 23%). For the wind speed, the largest reduction is found over Mumbai (city 4, 0.83 m/s and 25%), followed by city 5 (0.58 m/s and 18%) and city 3 (0.56 m/s and 36%). For the planetary boundary layer and relative humidity, the largest reduction and increment are found over Delhi (city 7; 167 m and 21%; 10.3% and 20%), followed by city 1 (130 m and 17%; 9.7% and 19%), respectively. over Central India, and 27 µg/m 3 (41%) over South India, respectively. As for O tions are observed in all cities except for cities 1, 2, and 6. The increases in O3 in th cities may be related to the variations of VOCs and other factors that can influ production and/or consumption of tropospheric O3. For SO2 and CO, the large tions were seen in Bengaluru, a pattern consistent with the reductions in NO2 a While over other cities, the changes in SO2 and CO are much smaller. Overall, th observed changes in NO2 and PM2.5 are consistent with the reductions in satellite NO2 and AOD. Additionally, the relative changes of meteorological conditions w same time periods show that precipitation (except Bengaluru and Delhi), wind sp planetary boundary layer (except Hyderabad) decrease, while the relative hum creases (except Hyderabad) ( Figure S4e). Moreover, the largest reduction in prec is observed over Hyderabad (city 3, 1.46 mm/day and 40%), followed by cit mm/day and 23%). For the wind speed, the largest reduction is found over Mum 4, 0.83 m/s and 25%), followed by city 5 (0.58 m/s and 18%) and city 3 (0.56 m/s a For the planetary boundary layer and relative humidity, the largest reduction a ment are found over Delhi (city 7; 167 m and 21%; 10.3% and 20%), followed by c m and 17%; 9.7% and 19%), respectively.

Potential Impacts of Meteorological Fields on Air Quality
Meteorological conditions could affect air pollution levels [18,53]. For exam ticulate matter levels in northern China during the city-lockdown period increase icantly and even led to several severe haze formations [53]. The unexpected air p was induced by the anomalously high humidity and uninterrupted emissions f rochemical facilities and power plants, whose emissions could promote aerosol h neous chemistry. The unexpected AOD increase in the central part of India is li

Potential Impacts of Meteorological Fields on Air Quality
Meteorological conditions could affect air pollution levels [18,53]. For example, particulate matter levels in northern China during the city-lockdown period increased significantly and even led to several severe haze formations [53]. The unexpected air pollution was induced by the anomalously high humidity and uninterrupted emissions from petrochem-ical facilities and power plants, whose emissions could promote aerosol heterogeneous chemistry. The unexpected AOD increase in the central part of India is likely due to the simultaneous increase in relative humidity and decrease in wind speed [18]. Here, the climatology of meteorological conditions is analyzed to study the potential impacts of meteorological fields on air quality over the Indian subcontinent, based on ERA5. Firstly, we evaluated ERA5 data against GPM and MERRA-2. The results show that ERA5 can well represent the precipitation and wind filed at 10 m over the India-subcontinent (Figures S5  and S6). Generally, the climatology of precipitation is mainly distributed over the Northeastern and Southwestern Indian subcontinents, with a maximum value of 40 mm/day during the lockdown period ( Figure 6). Also, fractional changes ((lockdown minus 2005-2019 climatology)/(climatology average)) in precipitation, circulations, boundary layer height, and relative humidity during the 2020 lockdown period are shown in Figure 7. Compared to climatology for the year 2020, there is a significant increase in precipitation over the Indian subcontinent, with a maximum increase of 100%. The mean fractional changes over North, Central, and South India are 9.6%, 9.7%, and 0.3%, respectively, which indicate the reductions of aerosol particles by wet scavenging. Simultaneously with the precipitation, the fractional changes in wind speed show a decreasing trend of 30-50% over the whole Indian subcontinent, which is consistent with the observed results in Pandey and Vinoj [18]. In general, the prevailing wind at 10 m during the lockdown period over South India is westerly, which turns northwesterly over North India ( Figure 6). The mean wind speeds are 1.3, 2.0, and 1.7 m/s over North, Central, and South India. The reductions in wind speed could provide a conducive environment for the stagnation of air-pollutants and increase fire counts over this region [18]. Here, the correlations between AOD, precipitation, and wind speed are also examined, and the day-to-day variation of AOD is associated with the two aforementioned meteorological variables. Clearly, the decreased (increased) AOD is consistent with the increased (decreased) precipitation a few days ago and the increased (decreased) wind during the lockdown period speed ( Figure S7). meteorological fields on air quality over the Indian subcontinent, based on ERA5. Firstly, we evaluated ERA5 data against GPM and MERRA-2. The results show that ERA5 can well represent the precipitation and wind filed at 10 m over the India-subcontinent (Figures S5 and S6). Generally, the climatology of precipitation is mainly distributed over the Northeastern and Southwestern Indian subcontinents, with a maximum value of 40 mm/day during the lockdown period ( Figure 6). Also, fractional changes ((lockdown minus 2005-2019 climatology)/(climatology average)) in precipitation, circulations, boundary layer height, and relative humidity during the 2020 lockdown period are shown in Figure 7. Compared to climatology for the year 2020, there is a significant increase in precipitation over the Indian subcontinent, with a maximum increase of 100%. The mean fractional changes over North, Central, and South India are 9.6%, 9.7%, and 0.3%, respectively, which indicate the reductions of aerosol particles by wet scavenging. Simultaneously with the precipitation, the fractional changes in wind speed show a decreasing trend of 30-50% over the whole Indian subcontinent, which is consistent with the observed results in Pandey and Vinoj [18]. In general, the prevailing wind at 10 m during the lockdown period over South India is westerly, which turns northwesterly over North India ( Figure 6). The mean wind speeds are 1.3, 2.0, and 1.7 m/s over North, Central, and South India. The reductions in wind speed could provide a conducive environment for the stagnation of air-pollutants and increase fire counts over this region [18]. Here, the correlations between AOD, precipitation, and wind speed are also examined, and the day-to-day variation of AOD is associated with the two aforementioned meteorological variables. Clearly, the decreased (increased) AOD is consistent with the increased (decreased) precipitation a few days ago and the increased (decreased) wind during the lockdown period speed ( Figure S7). Humidity is an important atmospheric parameter and can strongly influence AOD through chemical reaction processes [54]. The relative humidity at 2 m is significantly higher over the Indian coastal zones than in the inland areas (Figure 6), and the maximum mainly attributed to the reduction of anthropogenic emissions during the lockdown pe riod. It is worth noting that aerosols can reduce planetary boundary layer height via rad ative effects due to the positive feedback to the meteorology. Also, the day-to-day varia tion of AOD over the three sub-regions is significantly associated with variations in rela tive humidity and planetary boundary layer height during the lockdown period (Figur S7). To illustrate the correlations between daily AOD and the four meteorology variable in the three sub-regions during the lockdown period, their scatter plots and correlation are provided in Figure 8. The orthogonal linear regression is used to calculate the correla tion coefficients between AOD and meteorological variables, and the statistical signif cance of the coefficients is evaluated using the two-tailed Student's t-test. No statisticall significant correlation between daily AOD and precipitation and wind speed was detecte at the 95% confidence level. Relatively, AOD over North and Central India is highly cor related with relative humidity (positive). However, over South India, the correlation be tween AOD and relative humidity is not a statistically significant correlation. Over Centra Humidity is an important atmospheric parameter and can strongly influence AOD through chemical reaction processes [54]. The relative humidity at 2 m is significantly higher over the Indian coastal zones than in the inland areas ( Figure 6), and the maximum relative humidity along the coastal regions can reach over 90%. Compared to climatology, the fractional changes show an increasing tendency. Notably, the increase in relative humidity is prominent over North and Central India. Compared to climatology, the fractional changes in relative humidity over North, Central, and South India show increasing variations of 9.6%, 9.7%, and 0.3%, respectively. Pandey and Vinoj [18] have reported that the increase in AOD that occurred over Central India is due to the increase in relative humidity. The larger negative anomalies of tropospheric NO 2 and AOD that occurred over North India were also because of the higher relative humidity (Figure 7). Similar to the increase in relative humidity, the planetary boundary layer in the Indian subcontinent generally declined during the lockdown period, which could favor the accumulation of pollutants ( Figure 6). Even so, the AOD shows a decreasing tendency. The reason is mainly attributed to the reduction of anthropogenic emissions during the lockdown period. It is worth noting that aerosols can reduce planetary boundary layer height via radiative effects due to the positive feedback to the meteorology. Also, the day-to-day variation of AOD over the three sub-regions is significantly associated with variations in relative humidity and planetary boundary layer height during the lockdown period ( Figure S7).
To illustrate the correlations between daily AOD and the four meteorology variables in the three sub-regions during the lockdown period, their scatter plots and correlations are provided in Figure 8. The orthogonal linear regression is used to calculate the correlation coefficients between AOD and meteorological variables, and the statistical significance of the coefficients is evaluated using the two-tailed Student's t-test. No statistically significant correlation between daily AOD and precipitation and wind speed was detected at the 95% confidence level. Relatively, AOD over North and Central India is highly correlated with relative humidity (positive). However, over South India, the correlation between AOD and relative humidity is not a statistically significant correlation. Over Central India, AOD has a highly negative correlation with the planetary boundary layer, but the correlation is not statistically significant over the other two regions. These results indicate that AOD over North and Central India could be conditioned on the relative humidity and planetary boundary layer, though how to disentangle these multiple effects is challenging and beyond the capability of this analysis.
Remote Sens. 2022, 14, x FOR PEER REVIEW 13 of India, AOD has a highly negative correlation with the planetary boundary layer, but th correlation is not statistically significant over the other two regions. These results indica that AOD over North and Central India could be conditioned on the relative humidit and planetary boundary layer, though how to disentangle these multiple effects is cha lenging and beyond the capability of this analysis. Figure 8. Relationships between daily mean AOD with precipitation, wind speed, radiative humid ity and planetary boundary layer height over North, Central, and South India during the lockdow period. Correlation coefficients (r) are calculated via the orthogonal linear regression, and the st tistical significance (p) is using the two-tailed Student's t-test. Inhere, the statistical significance lev at 95% (p < 0.05) is marked with the asterisk (red color).

Radiation Response to COVID-19 Emission Reductions
The considerable reductions in anthropogenic emissions due to COVID-19 lockdow could be strong enough to cause changes in the regional radiation budget. The change (lockdown minus 2005-2019 climatology) in radiation fluxes at clear sky conditions fo shortwave, longwave, and net at the top of the atmosphere (TOA), in the atmosphere, an at the surface are shown in Figure 9. For SW radiation at the surface, positive anomalie are observed over the northeastern (10-15 W m −2 ) and western (5-10 W m −2 ) Indian sub continents (Table 2), which is consistent with the spatial patterns of decreases in AO ( Figure 4). Less SW radiation is absorbed by the atmosphere, which is likely caused b reductions in absorbing aerosols such as black carbon. Negative radiation anomalies the TOA indicate a higher planetary albedo (i.e., more scattering), which can also be a tributed to reductions in absorbing aerosols in the atmosphere. For longwave radiation a the surface, changes in radiation fluxes have much weaker magnitudes and opposite d rections at the surface and in the atmosphere, but the same direction of change at the TOA For net radiation, it generally follows the changes of shortwave radiation with a slightl smaller magnitude. Figure 8. Relationships between daily mean AOD with precipitation, wind speed, radiative humidity and planetary boundary layer height over North, Central, and South India during the lockdown period. Correlation coefficients (r) are calculated via the orthogonal linear regression, and the statistical significance (p) is using the two-tailed Student's t-test. Inhere, the statistical significance level at 95% (p < 0.05) is marked with the asterisk (red color).

Radiation Response to COVID-19 Emission Reductions
The considerable reductions in anthropogenic emissions due to COVID-19 lockdown could be strong enough to cause changes in the regional radiation budget. The changes (lockdown minus 2005-2019 climatology) in radiation fluxes at clear sky conditions for shortwave, longwave, and net at the top of the atmosphere (TOA), in the atmosphere, and at the surface are shown in Figure 9. For SW radiation at the surface, positive anomalies are observed over the northeastern (10-15 W m −2 ) and western (5-10 W m −2 ) Indian subcontinents (Table 2), which is consistent with the spatial patterns of decreases in AOD (Figure 4). Less SW radiation is absorbed by the atmosphere, which is likely caused by reductions in absorbing aerosols such as black carbon. Negative radiation anomalies at the TOA indicate a higher planetary albedo (i.e., more scattering), which can also be attributed to reductions in absorbing aerosols in the atmosphere. For longwave radiation at the surface, changes in radiation fluxes have much weaker magnitudes and opposite directions at the surface and in the atmosphere, but the same direction of change at the TOA. For net radiation, it generally follows the changes of shortwave radiation with a slightly smaller magnitude.

Conclusions and Discussion
In this study, we investigate the impacts of COVID-19 and the NCAP on the ity in the Indian subcontinent by using multiple satellite retrievals from OMI and and surface observation datasets from the Indian National Air Monitoring Netw results show a significant reduction of 10% and 15% in tropospheric nitrogen (NO2) and aerosol optical depth (AOD) during the 2020 lockdown period com their climatological levels (2005-2019) due to restricted human activities in many in India. The surface observations further demonstrate that PM2.5 and NO2 have si reduced by 33-50% and 41-58% during the 2020 lockdown period in seven cities India subcontinent, except for Mumbai in Central India. However, no significan tion has been detected for tropospheric NO2 and AOD during the same period Moreover, the potential impacts of meteorological fields on air quality, such as p tion, wind speed, relative humidity, and planetary boundary layer height, have a examined. The increase in relative humidity could, respectively, enhance aeroso scopic growth and induce a more stable boundary layer, which favors the accum of pollutants. However, total AOD over the Indian subcontinent decreases by m 20% compared to climatology. These findings suggest air pollution levels over th

Conclusions and Discussion
In this study, we investigate the impacts of COVID-19 and the NCAP on the air quality in the Indian subcontinent by using multiple satellite retrievals from OMI and MODIS and surface observation datasets from the Indian National Air Monitoring Network. The results show a significant reduction of 10% and 15% in tropospheric nitrogen dioxide (NO 2 ) and aerosol optical depth (AOD) during the 2020 lockdown period compared to their climatological levels (2005-2019) due to restricted human activities in many regions in India. The surface observations further demonstrate that PM 2.5 and NO 2 have significant reduced by 33-50% and 41-58% during the 2020 lockdown period in seven cities over the India subcontinent, except for Mumbai in Central India. However, no significant reduction has been detected for tropospheric NO 2 and AOD during the same period in 2019. Moreover, the potential impacts of meteorological fields on air quality, such as precipitation, wind speed, relative humidity, and planetary boundary layer height, have also been examined. The increase in relative humidity could, respectively, enhance aerosol hygroscopic growth and induce a more stable boundary layer, which favors the accumulation of pollutants. However, total AOD over the Indian subcontinent decreases by more than 20% compared to climatology. These findings suggest air pollution levels over the Indian subcontinent are significantly influenced by anthropogenic emissions. Further, the analysis of radiation demonstrates an overall increase in surface net radiation over the Indian subcontinent, with decreases in atmosphere-absorbed net radiation and decreases in radiation entering the TOA. The increase in surface net radiation is consistent with a surface temperature increase (0.04-0.07 K) over the South Indian subcontinent in March-May of 2020 [55,56]. These results provide valuable information for policymakers on the effectiveness of the measures proposed in the NCAP in improving air quality compared to the nationwide shutdown due to COVID-19.
However, such city-lockdown induced reductions in air pollutants are conflicted with economic development. Further observation-based analysis and model simulations are needed to investigate more economically efficient measures to mitigate aerosol pollution in the Indian subcontinent and to address the associated climatic and economic impacts.  Figure S3: Same as Figure S1, but for total column AOD. Data are from MODIS; Figure S4: The changes (lockdown minus climatology) of the tropospheric NO 2 , AOD, precipitation, wind, relative humidity, and planetary boundary layer at 7 sites in 2020 (light-red bar); Figure S5