Air quality change due to COVID-19 lockdown in India and its perception by public

In this study, air quality data for 100 days recorded at 193 stations distributed throughout India were analyzed to understand the changes in air quality following the country-wide lockdown imposed from 25 th March to 17 th May 2020 to contain the spread of the COVID-19 pandemic. The responses from a nationwide online survey conducted to obtain public perceptions of air quality improvement were also analyzed. On average, an approximately 40% improvement in the air quality index was observed, contributed by a reduction in 40% of PM 10 , 44% of PM 2.5 , 51% of NO 2 and 21% of SO 2 . There was a signicant difference between the levels of all the pollutants before and after the lockdown (p<0.05), except ozone. The Pearson correlation coecient showed that the correlation between PM 10 and PM 2.5 with ozone was signicant after the lockdown period, indicating that a signicant portion of the particulates present in the atmosphere after the lockdown period is secondary. The survey for public perception showed that 60% of the 1750 respondents perceived improvement in air quality. On a scale of 1 to 5, the respondents from Delhi perceived the highest improvement, from 2.2 to 4.5. An odds ratio of 17 indicated a very high dependence of perception on actual air quality. PM 10 levels had the highest inuence in shaping public perceptions of air quality. The results from Google Trends analysis showed that media had no inuence on shaping the perception of improvement in air quality. nonparametric (independent sample Kruskalwallis) test was conducted to establish differences in opinion among different zones in India. In the case of visibility, the opinion on improvement was found to be similar on pairwise comparison across all the zones in the country, except in the case of Southern zonal council vs. Northern zonal council and Southern zonal council vs. Central zonal council, with p value of < 0.05. Similarly, in the case of perception on improvement of health, significant differences in responses were obtained from the North Eastern vs. Central and Western zonal council and Northern vs. Central zonal council, with a p value of <0.05. In the case of indoor air quality (IAQ), a significant difference in opinion was obtained from the northern involved analysis of perception of the general public on improvement in air quality during COVID 19 lockdown. No approval is required from any committee to conduct such studies in India. No personal information was collected from any responders during our perception study that would reveal the identity of the respondent. Further, no individual responses are presented in the but only results of aggregate analysis.


Introduction
Pollution has become the major environmental cause of many diseases and premature deaths worldwide (Landrigan et al. 2017). The escalated levels of air pollution can be attributed mainly to rapid strides in industrialization and increased vehicular tra c. Other sources of air pollution include burning fossil fuels such as coal, wood and dry grass and construction activities (Gurjar et al. 2016). The Central Pollution Control Board (CPCB) of India identi es 17 different categories of industries in the country, viz., aluminum/copper/zinc smelters, thermal power plants, integrated iron and steel plants, petrochemical industries, caustic soda plants, cement plants, oil re neries, distilleries, dye and dye intermediate manufacturers, tanneries, drug and pharmaceutical industries, pulp and paper mills, sugar factories, pesticide manufacturing units and fertilizer plants. The air pollutants emitted from these numerous industries include ambient particulate matter (PM) of various sizes, metals, gases, and many organic compounds. More than 200 million vehicles (MORTH 2019) on roads also contribute signi cantly to air pollution in India. It is estimated that the emissions from the transport sector account for almost 56% and 70%, respectively, of the total PM 2.5 and PM 10 load in the air environment of the country. Of the total PM 2.5 contribution from the sector, 70% is from diesel-operated vehicles (Guttikunda et al. 2019).
Exposure to air pollutants triggers asthma, wheezing, rhinitis, eczema (Norbäck et al. 2019) and allergic disease (Brandt et al. 2015). Additionally, changes in several neurobehavioral functions in children and evidence of depression and cognitive impairment in the elderly have been found as after-effects of continuous exposure to polluted air (Costa et al. 2020). Studies have indicated that poor outdoor and indoor air quality increases mortality in some of the major cities in India (Nagpure et al. 2014;Lelieveld et al. 2015). Largely, air quality is associated with human activities. However, it is impossible to restrict human activities to affect improvements in air quality, although nonessential activities can be restricted to a certain extent (Bao and Zhang 2020). However, there can be instances when even essential activities are restricted as a response to unusual situations.
At the end of 2019, a novel infectious disease was reported in Wuhan, China (Huang et al. 2020). Later, this pandemic was identi ed as from the corona virus family and was named COVID-19 (Chen et al., 2020). WHO con rmed the transmission of COVID-19 through respiratory droplets among humans (WHO, 2020). The corona virus outbreak became an international crisis with its spread to many countries, resulting in the deaths of many and interruption of normal life. Nationwide lockdowns had to be implemented in most of the affected countries to contain the spread of the disease. The rst con rmed case of COVID-19 infection in India was reported from the state of Kerala on 30 January 2020. The country observed a jump in corona virus cases by the 4 th of March. On 22 March 2020, a 14-hour voluntary public curfew was imposed in India, followed by lockdowns in some districts as well as major cities where COVID-19 positive cases were identi ed. The Prime Minister of India ordered a nationwide lockdown for 21 days with effect from 25 th March 2020 (Ministry of Home Affairs 2020). This was further extended up to the 3 rd of May, with some conditional relaxations in areas where limited cases were reported. Even after these restrictions, the number of COVID-19 positive cases was on the rise, and a third phase of lockdown was implemented up to 17 th May. Several countries across the world went into lockdown due to the tremendous increase in the number of positive cases and the deaths associated with it. As a result of the series of lockdowns in India, industrial activities, transportation by all modes, and almost all other polluting activities decreased drastically. This had a positive impact on the environment, with pollution levels falling signi cantly in most cities of India (Mahato et al. 2020;Sharma et al. 2020). The same trends were reported in many other countries of the world, such as China (Bao and Zhang 2020;Muhammad et al. 2020;Wang et al. 2020b;Zambrano-monserrate et al. 2020;, Italy (Cristina et al. 2020;Muhammad et al. 2020;Zambrano-monserrate et al. 2020), Spain, France (Muhammad et al. 2020;Zambrano-monserrate et al. 2020), USA (Muhammad et al. 2020), Germany (Zambrano-monserrate et al. 2020), Brazil (Dantas et al. 2020), Kazakhstan (Kerimray et al. 2020), etc.
The negative impact of COVID 19 was felt in the industrial, trade and commercial sectors of the country (ICRA 2020a, b, c). It is expected that the country will go into the worst recession in recent times along with most other countries of the world. Perhaps the only area where the shutdown had a positive impact was the air environment. The shutdown resulted in the drastic reduction of pollutants discharged into the air environment.
Several studies have shown the in uence of the economic activities of a society on its air quality. NASA and the European Space Agency recently reported that there is a reduction in nitrogen-dioxide (NO 2 ) levels in China after the economic slowdown following complete lockdown of the country associated with COVID-19 infections (NASA Earth observatory 2020). A similar effect was observed in history earlier during the collapse of the Soviet Union in 1991. There was a signi cant reduction in greenhouse gases attributed to the reduction in meat consumption following the economic slowdown (Schierhorn et al. 2019). A study from California showed that economic indicators had a statistically signi cant effect on air pollution levels (Davis 2012). Similarly, a study conducted in New Jersey showed that economic activity levels can be used as a potential marker for assessing exposure to tra c-related pollutants in the absence of monitoring data (Davis et al. 2010).
The reduced tra c and restricted industrial activities have probably resulted in the reduction in PM 2.5 , carbon monoxide (CO) and NO 2 concentrations reported from different countries of the world during COVID-19-induced lockdown. The control on construction-related activities could also have contributed to the decrement observed in PM 2.5 and PM 10 concentrations. The decrease in ambient sulfur-di-oxide (SO 2 ) concentration during the control period was proportional to the decreased emission from industrial activities (Dantas et al. 2020;Mahato et al. 2020;Wang et al. 2020b). Although there was a decrease in air pollution in many countries of the world during lockdown, restricted anthropogenic activities were not su cient alone to explain the reduction in the level of pollutants in air. A study from China reported that due to the partial effect of unfavorable meteorological conditions, the reduction ratios of PM 2.5 concentrations, as a result of lockdown, were smaller than the reduction ratios of precursor emissions (Wang et al. 2020a), indicating the in uence of weather on ambient pollutant concentrations. Even though the major air pollutants, such as PM 2.5 , PM 10 , CO, NO 2 , SO 2 and ammonia (NH 3 ), saw a large reduction in their concentrations, the concentration of ozone (O 3 ) increased during the lockdown period in many parts of the world (Cristina et al. 2020;Dantas et al. 2020;Mahato et al. 2020;Wang et al. 2020b). The reason behind this reverse trend of O 3 concentration was identi ed as the decreased PM concentrations. Reduced PM in air results in increased photochemical activities and thus higher O 3 production by giving way for more sunlight to pass through the atmosphere (Dang and Liao 2017;Li et al. 2018). This may also be due to the favorable conditions for ozone formation, such as high temperatures and solar radiation indices (Escudero et al. 2019;Dantas et al. 2020). The decrease in NOx concentration in the atmosphere (Monks et al. 2015) and reduced utilization of O 3 by NO were also probably the reasons behind the increase in O 3 concentrations (Gorai et al. 2017) during the control period. This also shows that meteorological conditions are an important parameter. The simultaneous control of PM 2.5 and O 3 is quite di cult, and it requires measures such as proper adjustment of industrial structure and energy structure . A study from China found that the largest decrease in concentration during the COVID-19 lockdown period was for NO 2 and the least decrease was for SO 2 (Wang et al. 2020b).
There are many advantages for the public's perception of pollution being strongly correlated to actual levels of pollution. On the one hand, it prevents people from taking unnecessary health risks and generates public opinion against polluting emissions, forcing regulatory agencies to take corrective measures. On the other hand, it allows industries to judiciously make use of the assimilative capacity of the environment, resulting in net economic bene ts to society. However, there are various factors other than the actual levels of pollution that shape the perception of pollution. The perception of air quality has been found to be correlated with factors such as gender, education, age, health status, resident area, etc. (Guo et al. 2016;Oltra and Sala 2016). Howe et al. (2003) reported that aged responders had a more negative perception of pollution than younger responders, attributable, possibly, to their bad environmental experience during their younger ages. In another study, subjects older than 40 years, living in an urban area, having a college-level education and poor child health conditions perceived the air quality around the area as worse (Guo et al. 2016). The gender of the subjects considered was found to be associated with perception by Elliott et al. (1999). They found that most of the women reported that poor air quality will lead to adverse health effects compared to the men respondents. The respondents with respiratory indications (nocturnal shortness of breath, phlegm, rhinitis, etc.) reported higher levels of annoyance for degraded air quality (Sunyer et al. 2007). Factors such as health status, smoking status, and exposure time were also reported to have a signi cant impact on air quality perception (Pantavou et al. 2018).
However, in general, there is a strong correlation between perceived and actual air quality, although the perception varies with respect to age, gender, residence type, education, income, health condition and indoor house environment (Guo et al. 2016). Residents living in the proximity of industries and heavy tra c regions had negative perceptions of the air quality and health risks associated with it (Howel et al. 2003;Brody et al. 2004;Kohlhuber et al. 2006). Nikolopoulou et al. (2009) observed a good correlation between perceived air quality and PM concentrations in the study area. As the PM concentration increased, there was an increase in the number of 'poor air quality' votes and a decrease in the number of 'good air quality' votes. In another study, the public could truly perceive the most in uencing factors in a neighborhood's air quality (Mally 2016).
As far as human health is concerned, proper perception of pollution has signi cant advantages. Studies have shown that perceived air pollution and perceived health risk play a prominent role in the manifestation of health symptoms and contribute to ailments (Lloyd et al. 2005;Brosschot et al. 2006). In many cases, these health symptoms act as protective mechanisms against the severe consequences of pollution (Engen 1991). Often, the sources of pollutants may be identi ed based on olfactory sensations (pleasant or unpleasant). If the source is perceived to be unpleasant, it is more likely to have a negative impact on human health (Sucker et al. 2008). In addition to the olfactory systems, the trigeminal sensory system, which is activated by both vaporous substances and particles, also plays an important role in such perceptions. The sensation generated by trigeminal chemoreception includes pungency and irritation, where the re ex action prevents the inhalation of hazardous substances. The warning features of this defense re ex include sweating, coughing, mucus release, tearing and salivary ow (Silver 1991).
Perception studies on air quality have a signi cant role in creating awareness among people on the importance of having clean air (Evans et al. 1988;Liu et al. 2017). It was found that a signi cant proportion of the population was ready to take actions for the reduction of air pollution, as they perceived the air quality to be poor (Semenza et al. 2008;Li et al. 2016). Perception studies have been useful in providing suggestions to the government for improving air quality (Lan et al. 2016;Li et al. 2016). Wang et al. (2015) found that 90% of the respondents from Shanghai, China agreed that air quality improvement is the responsibility of government and the individual citizens of the country.
In the context of the importance of public perception in air pollution management, this study analyzes the change in air quality following the COVID-19 lockdown in India and its perception by the general public. Extensive data on actual air quality from 193 monitoring stations covering the whole country were taken from the repository of the regulatory agency. The public perception of air quality was obtained through an online questionnaire survey conducted among the general public answered by 1750 respondents. The data obtained were analyzed to obtain quantitative estimates of improvements in air quality and the relationship between actual and perceived air quality. The analysis results were used to arrive at conclusions regarding factors that were most in uential in shaping perception.

Methodology
Air quality data Geographic and demographic diversity is perhaps the most striking feature of India, the second most populous country in the world. From the Himalayan Mountains in the north to the Kanyakumari cape in the south and from the Thar Desert and salt marshes of the west to the humid forests of the northeast, the Indian main land covers an area of 3,278,982 sq.km. The tropic of cancer divides the country roughly into two equal halves. The southern part of the country, being a peninsula, experiences milder variations in temperature, whereas the northern region experiences extremes in temperature (R.B. Singh 2016).
Currently, there are approximately 231 continuous air monitoring stations in the country. These are connected to the web-based system, and the data are open to access for the public (CPCB 2020a). These monitoring stations are maintained by the respective state pollution control boards. Considering the size of the country, the number of air quality monitoring stations is insufficient. The government has plans to strengthen the network in major cities in a phased manner (CPCB 2020b). For this study, the air quality data for a total of 100 days (from 7-02-2020 to 16-05-2020) recorded at 193 air quality monitoring stations were downloaded from the Central Pollution Control Board (CPCB) website (https://app.cpcbccr.com/ccr/#/caaqm-dashboard-all/caaqm-landing). The pollutants considered in this study include PM 10 , PM 2.5 , SO 2 , NO 2 and O 3 . Furthermore, the air quality index (AQI), as calculated by CPCB, was also considered in this study (CPCB 2014). The period from 7-02-2020 to 22-03-2020 is considered the pre-lockdown period, and the period from 23-03-2020 to 16-05-2020 is considered the lockdown period in this study. The collected data were subjected to analysis for 1) changes in the concentration of air pollutants with respect to different zones/states, 2) the number of days the pollution level exceeded the permissible levels before and during lockdown, and 3) changes in the pollution level before and during lockdown with respect to the type of locality (residential, traffic and industrial) and major cities (based on the population).
The type of locality was decided based on the location of the monitoring station, whether in residential areas with minimum traffic, near major roads and traffic intersections or in industrial areas.
Multivariate statistical analysis was conducted using Pearson's correlation coefficients. The correlation coefficient is often used to determine the extent of the relationship between the variables when it is compared in pairs. There are various types of correlation coefficients (Núñez-Alonso et al. 2019), in which Pearson's correlation is the most commonly used method in linear regression. In this study, it is used to measure the strength of the relationship between the selected pollutants before and after the lockdown period.

GIS analysis
To estimate the spatial variation of air quality across the country, interpolation of the air quality data from monitoring stations was carried out. The most commonly used interpolation techniques for air pollution studies  Residents from all across the country were asked to complete the questionnaire (n=1750). The survey was conducted between 17-4-2020 and 27-5-2020. A total of 16 survey questions were asked. The entire survey was divided into three sections: 1) location details, 2) perception of improvement of air quality, visibility, health effects and sources, and 3) willingness to maintain air quality. Survey questions in sections 2 and 3 were asked on the rating scale and Likert scale. For example, "Rate the air quality in your locality before the lock-down period?" 1= Poor to 5= Good.
"Is there improvement in visibility in your locality?" No improvement, Slight improvement, Moderate improvement, Significant improvement and Don't know.
"I will actively be involved in maintaining the current status of the environment" Strongly agree, Agree, Undecided, Disagree and Strongly Disagree.
The questionnaire for the online survey is generated and circulated through Google forms (included as supplementary material). The questionnaire had the option of collecting the Geolocations of responders with their permission for plotting the results easily in ArcGIS. All statistical analyses were performed using SPSS 2.0.
Descriptive statistics were used to present the respondents' demographics and the response to the questionnaire. Independent sample non-parametric tests were conducted to understand the difference in opinion of the respondents belonging to different categories. Test of significance was performed using Chi-square ( 2 ) and t-tests. Statistical significance was considered at p<0.05. The strength of the relationship between the two variables was determined using "Effect size", calculated with Cramer's V (Sullivan and Feinn 2012).
The perceptions were analyzed based on 1) location (rural or urban), 2) type of locality (residential, near traffic junctions, near industries and near hospitals), 3) different zonal councils (western zonal council (W), southern zonal council (S), northern zonal council (N), eastern zonal council (E) and central zonal council (C), and 4) major cities (selected based on the population and pollution levels). Zonal councils in India were set up vide Part-III of the States Re-organizations Act, 1956, and this classification was made to establish advisory council "to develop the habit of cooperative working" among the states in the council. Councils were created to develop healthy inter-state and Center-state relationships by solving the inter-state issues and balancing the socioeconomic development of the corresponding zones (Government of India 2019).
The perception of air quality is often influenced by the media (Murukutla et al. 2019). To find the effect of media on air quality perception, an analysis was conducted using "Google trends". Keywords such as "Air Quality", "Air Quality Index" "AQI" and "Air Pollution" were used to analyze the frequency of discussions on air quality by the media. Frequent discussions on the topic by media may unduly affect the perception.

Qualitative response analysis
The question asked in the qualitative survey part is "Please give your suggestions for maintaining the air quality after the lock-down period". Qualitative response analysis was conducted by the process of identification, examination and finally interpreting the frequently repeated keywords in the textual data and based on the frequency of repetition of the keywords. Based on this analysis, suggestions are given to maintain the air quality after the lockdown period.

Relationship between AAQ and PAQ
The perception of air quality was collected on a rating scale from 1 to 5. Similarly, the AQI and the concentrations of the individual pollutants were also converted to rating scales as per breakpoint scales proposed by CPCB (CPCB 2014) given in Table 1. Both values were subjected to a test of significance using SPSS 2.0 to establish the relationship between PAQ and AAQ. Furthermore, the odds ratio (OR) was also used to express the strength of the association between AAQ and PAQ. The odds ratio is a statistic that is used to measure the association between the exposure and the outcome Page 10/33 (Szumilas 2010). The responses of people from two states, one where there was significant improvement in air quality (Delhi) and the other with nominal improvement in air quality (Telangana), were used to calculate the odds ratio. The odds ratio was calculated as in equation (2) When OR is >1, it indicates that the perception is dependent on improvement in the air quality. The higher the value of OR, the stronger the dependence.

Results And Discussion
Actual Air Quality The overall country averages for PM 10 calculated using the interpolated values were 116 µg/m 3 and 70 µg/m 3 before and after lockdown, respectively. Statistical analysis using a t-test showed that there was a significant difference in PM 10 levels before and after lockdown, with p<0.05. Fig 1(a-c) shows PM 10 levels before and after lockdown and improvement in its levels during lockdown. The overall country averages calculated using the interpolated value for PM 2.5 are 56 µg/m 3 and 31 µg/m 3 before and after lockdown, respectively. Statistical analysis using a t-test showed that there was a significant difference in PM 2.5 levels before and after lockdown, Page 11/33 with p<0.05. Fig 1(d-f) shows PM 2.5 levels before and after lockdown and improvement in its levels during lockdown. Compared to other states in India, many of the southern states had low PM 10 and PM 2.5 levels before and after the lockdown period. However, while considering the improvement in air quality in absolute terms following lockdown, the southern states showed lower improvement compared to the other states. Though the improvement in absolute terms was less, an overall better air quality was observed in the southern states. With respect to the percentage improvement in air quality calculated with the actual data of PM 2.5 and PM 10 , West Bengal showed the highest (64%) improvement in air quality. Orissa showed the lowest improvement in air quality (10%) with respect to PM 2.5 levels. In the case of Haryana, Uttar Pradesh and Bihar, the percentage reduction of PM 2.5 was found to be higher than the percentage reduction of PM 10 . The values of PM 2.5 /PM 10 were found to be > 0.5 in North East states such as Arunachal Pradesh and Meghalaya, which shows that a major portion of PM 10 was PM 2.5 . The North East states are less industrialized and scarcely populated, and this observation points to the long-distance transport of PM 2.5 from other places. The country-wide average reduction percentages for PM 10 (44%) and PM 2.5 (51%) were significantly different (p<0.05), with PM 2.5 having a higher reduction.
The overall country averages using the interpolated value for NO 2 were 30.45 µg/m 3 and 14.64 µg/m 3 before and after lockdown, respectively. Statistical analysis using a t-test showed that there was a significant difference in NO 2 levels before and after lockdown, with p<0.05. Fig 1(g-i) shows NO 2 levels before and after lockdown and improvement in NO 2 levels during lockdown. The average NO 2 level in the country more than halved during the lockdown period. Vehicles are the major contributor of NO 2 in ambient air (Ramachandran et al. 2013; US EPA 2019), and a large reduction in NO 2 levels is expected. On March 24 th , 2020, soon after the implementation of lockdown, an approximately 60% reduction in traffic was observed. In Delhi, the morning traffic congestion in 2019 was between 50 and 80 %, but it was reduced to single digit values ranging from 0 to 6 %. Similarly, in Bombay, the traffic congestion in 2019 during morning rush hours was 60 to 80%, and it decreased to <5%. A similar trend was observed throughout the lockdown period (Tom Tom Traffic Index, 2020). Among the states, the lowest level of NO 2 before lockdown was observed in the states of Jammu and Kashmir, Punjab, Kerala, Tamil Nadu and Andhra Pradesh. Similar to other pollutants, such as PM 10 and PM 2.5 , the highest decrease percentage in NO 2 levels as a result of lockdown was observed in West Bengal. In Howrah and Kolkata (West Bengal), all the monitoring stations in the traffic zone recorded 66 to 85% reduction of NO 2 . Similar reductions were noticed in the high traffic zones of Noida, Greater Noida and Ghaziabad (52-74%), Haryana (74%), Kanpur (54-66%), Chennai (67%), Udaipur (68%), Kota (57%), Jodhpur (55%), Jaipur (66%), Mizoram (74%), Mumbai (67-88%), Nagpur (60%), Navi Mumbai (85%), Tirupati (76%), and Thiruvananthapuram (65%). Reductions were also observed in traffic zones of Delhi (50 to 75%), Madhya Pradesh (60 to 70%), and Bangalore (58-80%).
Among petrol and diesel vehicles, a higher contribution of NO 2 is from diesel vehicles (European Union 2019).
The fact that a major portion of the vehicles registered in the country are diesel vehicles (Government of India 2015) could also have contributed to the higher reduction of NO 2 . In Delhi, compressed natural gas (CNG) vehicles constitute a significant portion of the vehicles on the road. Studies have confirmed that although CNG vehicles emit fewer pollutants than petrol and diesel vehicles with respect to PM 10 , PM 2.5 and SO 2, there is not much reduction in NO 2 emissions (Narain and Krupnick 2011).
The overall country averages using the interpolated value for SO 2 were 14 µg/m 3 and 11 µg/m 3 before and after lockdown, respectively. Statistical analysis using a t-test showed that there was a significant difference in SO 2 levels before and after lockdown with p<0.05. Fig 1(j-l) shows SO 2 levels before and after lockdown and improvement in SO 2 levels during lockdown. During the lockdown period, the lowest SO 2 levels were observed in the southern states, Kerala and Tamil Nadu. India is the world's largest emitter of SO 2 , and its emissions are mainly from 45 hotspots in the country (Shagun Kapil 2019). Out of the 45 hotspots, 43 have coal-based electricity generation. In Talcher Coalfields, Orissa, an approximately 50% reduction of SO 2 was observed, although the coal mines were still operational, albeit at reduced capacity. A significant reduction in SO 2 was also observed in many industrial belts across the country. For example, in Jahangirpuri-Delhi (70%), Pusa-Delhi (50%), GIDC, Ankleshwar-Gujarat (78%), Phase-4 GIDC, Vatva-Gujarat (83%), Industrial belt near to Chhatrapati Shivaji International Airport-Mumbai (83%), Mahape, Navi Mumbai (59%), and RIICO Industrial Area III-Bhiwadi (72%).
The overall country averages using the interpolated values for O 3 were 35 µg/m 3 and 37 µg/m 3 before and after lockdown, respectively. The slight increase can be attributed to the increased photochemical activity due to the decrease in particulate matter concentration (Dang and Liao 2017;Li et al. 2018). Statistical analysis using t-test showed that there is no significant difference in O 3 level before and after lockdown with p>0.05. Fig 1 (m-o) shows O 3 levels before and after lockdown and improvement in O 3 levels during lockdown. concentrations were higher before the lockdown period.
The country experienced an overall increase in AQI levels, as shown in Fig 1(p). The highest improvement in AQI was seen in West Bengal (61.89%), followed by Arunachal Pradesh (47.80%) and Meghalaya (47.58%). The least change was observed in Orissa (3.55%). The AQI values in Orissa before and after lockdown were 132.49 and 127.78, respectively. Even before lockdown, the AQI in the southern states was in the satisfactory range. Among the major cities, the highest percentage reduction in AQI was in Jaipur, followed by Kolkata, Mumbai, Pune and New Delhi. The lowest reduction was observed in Chennai. The other major cities, such as Bangalore,

Hyderabad and Ahmedabad, had intermediate percentage reductions in AQI values. Although the lowest AQI
value before lockdown was observed in Chennai, the lowest value after lockdown was found in Jaipur.
Additionally, the highest value before and after lockdown was observed in New Delhi.
The interpolated values of pollutant concentrations were compared with NAAQS to determine the number of days when it exceeded the permissible limits (Fig 2(a & b)). It was observed that gaseous pollutants (NO 2 , SO 2 and O 3 ) were within the permissible limits before and after lockdown. In the case of West Bengal, Telangana, Meghalaya, Maharashtra, Kerala, Karnataka, Chandigarh and Andhra Pradesh, there was no day when the state average value exceeded the permissible limits as prescribed by CPCB in the case of PM 10 after the implementation of lockdown. The percentage reduction of pollution in all three categories of areas (traffic, residential, and industrial) was approximately 40% after the implementation of COVID-19 lockdown, as shown in Fig. 3. As per the 2011 census data, the most populated cities in India are Mumbai, Delhi, Bangalore, Hyderabad, Ahmedabad, Chennai, Kolkata, Surat, Pune and Jaipur. Except for Surat, data were available for all these cities.
For PM 2.5 , most of the cities showed a more than 50% reduction in concentration after the start of lockdown, as shown in Fig. 4 Table 2. The Pearson correlation is significant for a pair when p<0.05. The PM 10 -PM 2.5 pair is positively correlated before (0.880) and after (0.746) the lockdown, which suggests a common source for most of these pollutants. The coefficients for the pairs PM 10 -NO 2 and PM 2.5 -NO 2 are also significant. Motor vehicles are a common source of particulates, and NO 2 explains this correlation. The higher correlation between reductions in PM 2.5 and NO 2 , compared to the correlations of their concentrations in ambient air before and after lockdown, perhaps indicates that their reductions to a great extent can be attributed to the removal of diesel vehicles from roads due to lockdown. An interesting observation was the positive correlation of PM 10 and PM 2.5 with O 3 after the lockdown period. This probably indicates that a significant source of PM after lockdown is photochemical reactions (Mangia et al. 2015) that also result in the formation of O 3 (North Earth Observatory 2003). With most of the anthropogenic sources of primary particulate matter cut down, a good proportion of the available particulate matter might be the secondary particulate matter formed due to chemical reactions between various existing pollutants. Before the lockdown period, PM 10 /O 3 and PM 2.5 /O 3 were either uncorrelated or negatively correlated. There is no significant correlation in the case of NO 2 /SO 2 , SO 2 /O 3 and NO 2 /O 3 both before and after the lockdown period, indicating that these pollutant pairs are from different sources. .043 0.860 1 **-correlation is significant at the 0.01 level (2-tailed), BL-Before lockdown, AL-After lockdown.

Perceived air quality
On conducting 2 tests on the responses obtained from rural and urban populations, a significant difference in opinion was noticed between them; 2 (2, N= 1750) = 43.99, p < 0.05. Urban responders felt higher improvement in air quality during the lockdown (64% to 49%). The observed effect size was Cramer's V = 0.2, indicating a small effect size in the difference in perception. This shows that although there is a significant difference in the opinion, the people in both rural and urban environments have perceived improvement in air quality. Considering the different types of locality (residential, traffic, industrial), a significant difference in opinion was observed, 2 (2, N= 1395) = 18.987, p <0.05. The observed effect size in this scenario was much smaller (Cramer's V = 0.1).
The order of improvement of air quality perception among different localities was Industries > Near to traffic junctions > Residential > Near to hospital, as shown in fig 5. Nearness to hospital was considered a separate category expecting more traffic near hospitals due to the pandemic situation. More than half of the respondents in all major cities perceived improvement in air quality: Delhi (100%), Ahmedabad (85%), Chennai (83%), Mumbai (80%), Bangalore (65%), Jaipur (66%) and Hyderabad (54%).
A nonparametric (independent sample Kruskalwallis) test was conducted to establish differences in opinion among different zones in India. In the case of visibility, the opinion on improvement was found to be similar on pairwise comparison across all the zones in the country, except in the case of Southern zonal council vs.
Northern zonal council and Southern zonal council vs. Central zonal council, with p value of < 0.05. Similarly, in the case of perception on improvement of health, significant differences in responses were obtained from the North Eastern vs. Central and Western zonal council and Northern vs. Central zonal council, with a p value of <0.05. In the case of indoor air quality (IAQ), a significant difference in opinion was obtained from the northern Page 16/33 and southern zonal council, as in the case of visibility. The mean rating (on a five-point scale) of air quality perception in the case of rural areas increased from 3.5 to 4.3, whereas in the case of urban areas, it increased from 2.9 to 4.12. The improvement in the perception of air quality is shown in Fig. 6 (a). In a zone-wise comparison, people from all zones felt air quality to be 'excellent' after lockdown, as shown in Fig. 6 (b).
Responses from major cities in India were analyzed separately, and it was found that the mean perception of Ahmedabad changed from 2.4 to 4.1, Bangalore and Hyderabad changed from 2.92 to 4, Chennai changed from 2.7 to 4.2, Delhi changed from 2.2 to 4.5, Jaipur changed from 2.8 to 4.3, Kolkata and Mumbai changed from 2.6 to 4.2, and Pune and Surat changed from 3 to 4, as represented in fig 6 (c). This indicates that respondents from major cities across India perceived improvement in air quality during this lockdown period. The obtained perception scales and geolocations were directly subjected to geo-spatial analysis using ArcGIS. The maps in Fig. 7 (a-c) indicate the perception before and after lockdown and the change in perception due to the lockdown.
Most people perceived the air as moderately polluted before lockdown. From the plots obtained for perception of air quality after lockdown, it is clear that most people perceived air quality as satisfactory or good. Therefore, it can be interpreted that the lockdown has created a positive feeling among people regarding the air quality in the country.
The major sources of pollution as perceived by the respondents before lockdown were Vehicular Pollution > Road dust > Construction works > Industries > Road side burning > Burning of agricultural waste >Power plant.
However, after lockdown, the order was Household emissions> Solid waste burnings > Traffic > None >Industrial activities> others. Other factors included burning of crackers, spraying of disinfectant chemicals and smoking. The perception of sources among different localities and zonal councils is given as supplementary material.
The results from Google Trend Analysis (https://trends.google.com/trends/?geo=US) showed that terms related to air pollution ('air quality', 'air quality index', 'air pollution', 'AQI') were trending from October to December 2019, as shown in Fig. 8 (Rizwan et al. 2013;Patel 2019). The trend that declined drastically after the annual pollution episode continued decreasing to the lockdown period, albeit at a smaller rate, except for a small spike on 22/03/2020, the day when the Janata curfew was implemented. This trend can be interpreted as a lack of influence of media on the perception of air quality by the public. Similar inferences were made by Searle et al., (2020) and Szmuda et al.,(2020).

Relationship between PAQ and AAQ
Our perception survey showed that approximately 60% of the respondents perceived improvement in air quality during the COVID-19 lockdown compared to the pre-lockdown period. The analysis of the air quality monitoring data showed that from the pre-lockdown to lockdown period, the AQI improved by 40% in the country. To have a quantitative comparison, the perceived air quality was obtained on a rating scale from 1 to 5 (Poor to Good). The pollutant concentrations were converted to the same rating scale (1 to 5) as per the break points proposed by CPCB. The results from the paired t-test showed that there was a significant difference (p<0.05) in air quality perception before and after lockdown (Table 3). Similarly, there is a significant difference (p<0.05) in actual air quality before and after lockdown. However, there is no significant difference (p>0.05) between the pairs 'air quality perception before lockdown -actual air quality before lockdown' and 'air quality perception after lockdown -actual air quality after lockdown'. This shows that there is a clear association between air quality perception and actual air quality. On conducting the test of significance between air quality perception and the converted rating scale of actual air quality, interesting results were found. There was no significant difference (p>0.05) between air quality perception and PM 10 level, but there was a significant difference between the perception and levels of SO 2 , NO 2 and O 3 (P=0.00). This shows that among various pollutants, PM influenced perception most, possibly because of its contribution to visibility. Several studies have shown a clear association between PM 10 levels and visibility, with an increase in PM 10 levels resulting in lower visibility (Zhao et al. 2013;Huang et al. 2016). In earlier days, before the invention of air quality monitoring instruments, visibility was the parameter that was used to assess the air quality. Visibility impairment is caused by scattering and absorption of visible light by the suspended particulates and gaseous pollutants present in the atmosphere (Hyslop 2009;Lee et al. 2015;Majewski et al. 2015). In urban environments, visibility impairment is closely associated with pollutants emitted from anthropogenic sources such as automobile exhaust, combustion of fuel, emission from industries, etc. (Tsai et al. 2007;Deng et al. 2008;Majewski et al. 2015). It was also observed that this visibility impairment is mainly due to airborne particulate matter (Malm and Day 2001;Tsai et al. 2003). The odds ratio (OR), often used in medical statistics, represents the odds that an outcome will occur given a particular exposure compared to the odds of the outcome occurring in the absence of that exposure (Szumilas 2010). It is extensively used to analyze the relationship between exposure to pollutants and its health impacts (Baxter et al. 2010;Lee et al. 2014;Yorifuji et al. 2014;Klompmaker et al. 2019). It is also used in air quality perception studies (Malenka et al. 1993;VanderWeele and Vansteelandt 2010;Voda et al. 2020). To quantify how the change in air quality actually changed the perception, responses from a city with one of the highest reductions of pollution (Delhi -72 responses) and an area where there was only a small reduction in pollution after lockdown (Rural Telangana -39 responses) were used to calculate the odds ratio. The details of the responses are given in Table 4. No of respondents perceived significant improvement in air quality 57(N pi ) 7 (N pn ) No of respondents perceived no improvement in air quality 15 (N ni ) 32 (N nn ) The OR obtained is 17, which indicates that the perception improvement in air quality is highly dependent on the actual improvement in the air quality. Higher odds ratios indicate higher dependence between PAQ and AAQ.

Qualitative interpretation of the suggestions
The words were decoded from the suggestions based on the frequency of appearance. The suggestions given by the respondents were to lift the lockdown scientifically, to implement strict regulations with respect to traffic, industries, vehicular emissions and trash burning on road margins, living in harmony with nature, strengthening public transportation and adoption of carpooling systems, promotion of E-vehicles and bio-fuels, plantation of trees, installation of air quality monitors across the country and creating awareness among the public about the improvement in air quality levels and maintenance of the same.

Conclusion
From this study, it is evident that there is signi cant improvement in the actual and perceived air quality in India after the COVID-19-induced lockdown. Approximately 60% of the respondents perceived improvement in air quality, and there was approximately 40% improvement in the monitored air quality across the country. It is evident that the respondents perceived improvement in air quality without the in uence of media. The reduction in air pollution was investigated with respect to three different zones. Major tra c zones across the country have experienced signi cant improvement in the NO 2 level due to the decrease in vehicular load. Similarly, a signi cant reduction in SO 2 levels was observed in industrial belts and coal mines. The correlation matrix developed gave a clear association between the pollutants and the possible sources. During the lockdown period, an increased photochemical reaction was observed, which led to an increase in the levels of ozone at many locations. Along with improvements in air quality, signi cant improvements in visibility, indoor air quality and health were perceived by the respondents. The perception of improvement in air quality was in uenced mainly by the reduction in particulate matter. The odds ratio showed a very strong dependence of perception on actual air quality.
Suggestions by the public for maintaining air quality even after lifting the COVID-19 lockdown are also given in this study.

Declarations
Funding: No funds received Con icts of interest/Competing interests: The authors declare that they have no known competing interests.
Availability of data and material: Added as supplementary material Code availability: Not applicable  supplementarymaterial.docx