Air Quality in the New Delhi Metropolis under COVID-19 Lockdown

Air pollution is one of the most important sustainability concerns faced by most of the developed nations of the world [1]. It is identified in the Sustainable Development Goals (SDG): Goal 3 (Good Health and Well-being) [2] and Goal 11 (Sustainable Cities and Communities) [3]. Guégana et al. in [2] discusses SDG: Goal 3 (Good Health and Well-being) and the need for integrative thinking in public and animal health for developing countries. Furthermore, Vaidya and Chatterji in [3] discuss SDG 11 and the new urban agenda for addressing global sustainability at an urban scale. The occurrences of forest fires that happen at sporadic intervals also lead to increased air pollution [4]. The recent onset of COVID-19 (C19) and its associated lockdown led to an increased awareness of this pressing issue. Jia et al. in  [5] showed insignificant impact on air quality due to lockdown in Memphis, US using ground based monitoring data, in comparison to previous months. Similar observations were observed in the city of New York, US, where Zangari et al. observed that there was no significant air quality changes in New York because of the COVID-19 pandemic [6]. On the other hand, a city like Delhi, India known for severe air pollution issues showed a large contrast for noticeable observations in our study both with satellite and ground based data.

The lockdown induced by the COVID-19 pandemic observed a decrease in atmospheric pollutants around the globe. Owing to the complete halt of vehicular traffic and reduction in industrial activities, we observed a positive impact on the amount of atmospheric pollutants and the overall air quality. The lockdown because of COVID-19 pandemic has provided the environmental researchers an opportunity to analyze the atmospheric pollutants in finer scale. Bourdrel et al. in [7] attempted to provide a relationship between air quality and chronic diseases in the context of COVID-19. Similar work on the relationship between air pollution and COVID-19 infections and mortality was performed by Ali and Islam in [8]. Furthermore, Nigam et al. in [9] identified the positive effects of COVID-19 in the industrial cities of Ankleshwar and Vapi in India. They noted a massive decrease in the concentration of NO₂ in the atmosphere, owing to the halt in industrial activities. Karaer in [10] analyzed the 67 Florida Counties to present the strong correlation between the drop in NO2 concentrations (from Sentinel-5p) and Vehicle Mile Traveled (VMT) estimates from cell phone mobility records. Kaloni et al. performed a systematic analysis on the impact of COVID-19 on the air pollution of Dublin city [11]. Franch-Pardo et al. in [12] provided a systematic review of 221 scientific articles published in 2020 that used spatial and GIS technologies to understand the dynamics of COVID-19. Doolette et al. investigated the atmosphere from the context of nanoparticle fertilizer in precision agriculture [13]. Such recent works reveal that there is a promising opportunity for researchers to analyze the atmospheric pollutants, in the context of COVID-19 pandemic.

In this paper, we focused our case study on the city of New Delhi in India. The primary reasons of choosing New Delhi was because this city has a historical problem of poor air quality, and therefore it provided us an opportune moment in performing such atmospheric analysis. The city of New Delhi has its long-lasting history with high pollution levels [14]. Traditionally, government introduced several measures including plying of odd and even numbered vehicles on alternative days, and introduction of CNG vehicles [15, 16]. However, there was an insignificant improvement in Air Quality Index (AQI) [17]. The AQI level stayed in severe and very poor category [18]. However, with the introduction of lockdown owing to C19 pandemic, we observed a significant improvement in AQI level, and reached the moderate category during the lockdown .

In this paper,¹¹1In the spirit of reproducible research, code used to obtain plots and results is shared at https://github.com/dkaloni/LockdownAnalysis. we discussed the impact of the lockdown induced by the C19 pandemic on the atmospheric air quality. The novel contributions of this paper are as follows:

• 1.

  Systematic analysis of the impact of lockdown on air quality;

• 2.

  Subjective and objective validation of improved air quality from satellite and ground-based data; and

• 3.

  Release of open-source code and datasets for further benchmarking by the community.

The rest of the paper is arranged as follows. Section 2 discusses the TROPOMI instrument of Sentinel-5P data that is used for our analysis and the data format of files obtained from the sources. Section 3 discusses with the methodology used to obtain large sized satellite data and resolving it to usable form. Section 4 briefly explains the change in nitrogen dioxide over the Central India region, followed by discussion on concentration of harmful gases in the atmosphere of Delhi using satellite data. Section 5 correlate satellite data with the ground-based monitoring station data and thereby confirming our findings with regard to harmful gases. Finally, Section 6 concludes the paper and discusses the future work.

Copernicus, an European Union’s Earth Observation Programme consists of several satellites. The Sentinel-5P is one of the Earth Orbit Polar Satellite System collecting data on trace gases concentration, climate and air quality.

This study was mainly focused on Delhi and involved processing data specifically for the Delhi region. We systematically analyzed the data obtained from both satellite and ground-based observations. The data obtained from TROPOMI were stored in NetCDF format. The NetCDF stands for Network Common Data Form, and is the format of file used for multidimensional data to study atmospheric features. The Sentinel-5P data uses NetCDF files to share and store files in an array-like structure stacked together. It is generally used to share climate data and for related purposes. The NetCDF4 file contains both the data and the metadata for the product.

To subset out New Delhi from the global TROPOMI dataset, we performed a sequence of methodological steps in order to subset out the data of New Delhi from the global dataset. The selection of New Delhi region was performed by providing the range of latitude and longitude of the region. We subset out the Delhi region by providing the latitude range from $28.40^{\circ}$ to $28.90^{\circ}$ and longitude range from $76.80^{\circ}$ to $77.35^{\circ}$. The data points included in this region of interest were thereby used for the subsequent computation of pollutant concentration. The measurements were stored and archived in individual NetCDF files. Each NetCDF file indicated the concentration of a trace gas for the selected day of the study. We then combined all the obtained files into a single file by adding a new dimension of time to the dataset for comparative study.³³3The NCO is a set of command line utilities used to operate NetCDF files efficiently, available at http://nco.sourceforge.net/. This combined file provided us with data points for the area spread across the time dimension.

Our obtained combined NetCDF file provided us the estimate of the average daily concentration of the various atmospheric pollutants. We processed these files to obtain the pollutant concentrations of the different pollutant gases. We replaced the NaN values of the data grid to $0$, and thereby calculated the sum of the all the measurement values in the particular grid. We divided the resultant sum by the number of non-zero values to have a daily average for the region. This enabled us to calculate the average concentration of a particular pollutant over the selected region of New Delhi across a 15-days period. These steps were subsequently repeated for the different gases in the tropospheric column.

In this section, we delved further into the concentration of the individual pollutants obtained from the TROPOMI dataset. We compared the average concentration of such pollutant with that of the previous year data to quantify the improvement in the air quality.

The ground-based pollutant data was collected for ITO station of Central Pollution Control Board, Ministry of Environment, India. It is one of the largest atmosphere monitoring station in India. The data is freely accessible from https://app.cpcbccr.com/ccr. We collected and analyzed data for a period of two years for the years 2019 and 2020. We collected the Air Quality Index (AQI), Particulate Matter 10 (PM10), Particulate Matter 2.5 (PM2.5), and other pollutant gases present in the atmosphere. We specifically plotted the daily concentration data for the months of February, March and April across the years because the lockdown came into effect during these months.

The particulate matter data collected from these ground-based monitoring stations indicated an improvement in the PM concentration [18]. The PM2.5 concentration in Delhi before the lockdown was in the range of $60\mbox{ }\mu$g/m³ to $125\mbox{ }\mu$g/m³. However, after the lockdown period, the concentration clearly dropped by about 50% to the range of $25\mbox{ }\mu$g/m³ to $60\mbox{ }\mu$g/m³. A similar drop was visible in the concentration of PM10, which was recorded to be $175\mbox{ }\mu$g/m³ on 20-Mar-2020. This reduced to $100\mbox{ }\mu$g/m³ on 22-Mar-2020 and $60\mbox{ }\mu$g/m³ on 26-Mar-2020, following the similar decreasing trend for the period of lockdown [36].

In addition to this particulate matter concentration, we observed similar trend in the concentration of harmful gases in the atmosphere, which is in line with our analysis on Sentinel-5P satellite data. In Fig. 4, we observed that the concentration of NO₂ significantly changed throughout the month of Mar-2020. The NO₂ concentration in the air had a peak of $61.47\mbox{ }\mu$g/m³ on 20-Mar-2020 (just two days before janta curfew), which dropped to $40.43\mbox{ }\mu$g/m³ on 22-Mar-2020 (date of janta curfew) and continued to drop to about $17.25\mbox{ }\mu$g/m³ on 31-Mar-2020. Consequently, the average concentration of NO₂ saw a significant drop from $61.47\mbox{ }\mu$g/m³ to $17.25\mbox{ }\mu$g/m³ during the month of Mar-2020. We observed similar trend in the reduction of NO₂ concentration when compared between the same days of 2019 and 2020.

Similarly, Sulfur Dioxide (SO₂) concentration in the air had a peak of $24.16\mbox{ }\mu$g/m³ on 20-Mar-2020 (just two days before janta curfew), which dropped to $17.48\mbox{ }\mu$g/m³ on 22-Mar-2020 (day of janta curfew) and continued to drop to $7.67\mbox{ }\mu$g/m³ on 28-Mar-2020. The ground based data showed an abnormal increase in Sulfur Dioxide concentration for the first half of April which subsequently dropped noticeably in later April. Carbon Monoxide (CO) is generally produced from incomplete combustion of fuels containing carbon. Ground based collected data showed that the concentration of CO dropped to 2.94 mg/m³ on 20-Mar-2020, and further dropped to 0.33 mg/m³ on 22-Mar-2020 (day of janta curfew) indicating a drop of approximately 88%. This decrease continued till the end of Mar-2020 and first week of Apr-2020.

We also obtained the daily average ozone concentration in the atmosphere from the ground-based pollutant data. We observed from the ground-based data that O₃ concentration kept on rising post lockdown. The minimum concentration for month of Mar-2020 was $15.41$ $\mu$g/m³ on 14-Mar-2020 and had a peak of $45.06$ $\mu$g/m³ on 29-Mar-2020 which was one of the initial day of lockdown. These values kept on increasing and reached maximum value of $106.23$ $\mu$g/m³(for April) on 30-Apr-2020. This trend is in agreement with the satellite-based O₃ concentration analysis.

Our study indicated that the restrictions put in place because of COVID-19 had a clear positive impact on the pollutant concentration in the atmosphere. Owing to these restrictions, there was a reduced human movement, reduced vehicular emissions, and less industrial activities. Such studies can pave the path for the government (both at national and state level) to enact laws in dramatically reducing emission of toxic air pollutants, and provide healthcare protections to the citizens. These steps can include reducing toxic air pollution from the industrial sources, enacting stringent standards for vehicles and engines, and educating the general public via community-driven voluntary programmes. Such laws were enacted by the Government of India via the Motor Vehicles Act, 1988, that required petrol driven vehicles to require PUC (Pollution Under Control) certificate for low emission of carbon monoxide in the atmosphere. Therefore, our study will greatly assist the environmentalists in better understanding the atmosphere.

Delhi being one of the most polluted cities observed a drastic and noticeable drop in harmful gases like Nitrogen Dioxide, Carbon Monoxide, and Sulfur Dioxide. An interesting phenomena noticed in the atmosphere of Delhi was the increase in ozone concentration as the lockdown was imposed which was also linked to the improvement in the atmospheric conditions comparatively. Obtaining satellite data from Sentinel-5P depicted the scenario from a different perspective, where ground based station data verified the obtained results exactly. Remote location with lack of ground based monitoring stations will benefit from such remote sensing techniques for atmospheric monitoring without any additional setup cost. The COVID-19 lockdown brought a rare opportunity for researchers to understand the environment much better and thereby attempt to influence the policy makers to implement stricter environment laws.

Our future work include the short-term forecasting of pollutant data from historical data. We intend to use the popular LSTM-based models [37, 39] to learn the underlying pattern of such time-series pollutant data. Furthermore, we also plan to mine COVID-19 related texts in different languages [40], and provide an universal knowledge bank. We also intend to understand the impact of other real-life affecting factors, including population size, urban/rural divide, level of industrialization, income, and pulmonary diseases, on the air quality of a megacity. Such studies will enable us in proposing relevant recommendations to city planners and environmental agencies.

This research was conducted with the financial support of Science Foundation Ireland under Grant Agreement No. 13/RC/2106_P2 at the ADAPT SFI Research Centre at University College Dublin. ADAPT, the SFI Research Centre for AI-Driven Digital Content Technology, is funded by Science Foundation Ireland through the SFI Research Centres Programme.