2.1 Data
Satellite-measured air pollutants data was collected from the satellite Sentinel-5 Precursor of the European Space Agency(ESA). The satellite hosts the TROPOspheric Monitoring Instrument (TROPOMI) instrument. The TROPOMI instrument is a space-borne, nadir-viewing imaging spectrometer covering wavelength bands between the ultraviolet and the shortwave infrared. The instrument has a daily synoptic coverage of the globe with 7km X 3km spatial resolution and a 2600 km swath width. The criteria pollutants considered are Ozone(O3),Carbon monoxide (CO), Sulphur Dioxide (SO2) and Nitrogen Dioxide (NO2).
The images acquired by the Sentinel-2 satellite were also downloaded for the preparation of the NDVI map. NDVI is a dimensionless index that can be substituted for surface vegetation coverage and its change. Plant foliage absorbs the particles leading to reduced ground-level PM2.5 concentrations. So NDVI is used as a proxy to account for the same in the model.
The Aerosol Optical Depth(AOD) data from the Moderate Resolution Imaging Spectro-radiometer (MODIS) onboard AQUA and TERRA satellites under the Earth Observing System (EOS) series of satellites of NASA was collected from the USGS Earth Explorer website. The data product MCD19A2 will be used, which is the combined AOD data from both AQUA and TERRA satellites and is obtained by applying the Multi-angle Implementation of Atmospheric Correction (MAIAC) algorithm, resulting in a data product of spatial resolution of 1km X 1km. The high resolution (1 km) MAIAC AOD is suitable for predicting the spatial distribution of PM2.5 in urban areas where the input parameters vary a lot with space. It has been used to estimate ground PM2.5 concentrations (Jung et al., 2017) in the US (Lee et al., 2016), China(Lin et al., 2016), (Xiao et al., 2021) and many other areas as well. Previous studies indicated that an inherent disadvantage of many AOD products is their coarse spatial resolution (Chang et al., 2014)(Chang et al., 2014)(Chang et al., 2014), (Hu et al., 2014), which can be avoided by the use of MAIAC-derived MODIS AOD data product.
The Digital Elevation Model data from the Cartosat-1 satellite, i.e., CartoDEM data was downloaded from BHUVAN, the georepository platform of the Indian Space Research Organization (ISRO). The data product CartoDEM has a spatial resolution of 10m X 10m.
Population Data data was collected from the website of Census of India. The last census data that was conducted in 2011 has been considered.
For creating the district boundary shapefile of the study area, the digitally scanned toposheets were downloaded from the Survey of India website. Other ancillary data like Google Earth and Open Street Maps were also used for analytical purposes.
The station level air temperature, relative humidity, and atmospheric pressure data were downloaded from the CPCB website. The gridded precipitation data was downloaded from the IMD, Pune website. Aerosols increase due to moisture absorption, which changes their distribution and optical properties influencing the PM2.5–AOD relationship (Yao et al., 2018). This will be accounted for by using the relative humidity as a parameter in the model. The result of the model depends on various dispersion conditions, such as meteorological and topography conditions for especially large areas and daily PM2.5 concentrations (Fang et al., 2016). So, daily meteorological data including precipitation (mm), temperature (K), and pressure (hPa) was considered as input parameters.
The ground level measured pollutant data was collected from the Central Pollution Control Board (CPCB) official website for the automatic stations, and for the manual stations, it was collected from the West Bengal Pollution Control Board (WBPCB) website. The WBPCB monitors nine manual stations in Howrah, namely Amta, Bandhaghat, Bator, Sankrail, Howrah Municipality Corporation (HMC), Dhulagarh, Ghusuri, Uluberia, Bagnan. The stations are located near the main market area or big traffic junctions. The data for the Bagnan manual station was not available for any date throughout the study period, and also the data for lockdown phases 1, 2, 3, and 4 of the year 2020 was also not available for any of the manual stations. There are three CPCB monitoring sites in Howrah at Belur Math, Ghusuri, Padmapukur. Data from all the three sites will be collected. The criteria pollutants that are considered in this study are Sulphur dioxide (SO2), Nitrogen dioxide (NO2), Ozone (O3), PM2.5, PM10, and Carbon monoxide (CO).
2.2 Study Period
The study period starts on 11th March 2020, which is exactly 14 days before the government of India declared the 1st lockdown (25th march 2020) to control the spread of the Covid 19 disease and ends on 31st July 2020, which marks the end of the unlock 2.0. The extra 14 days are considered to account for the pre-lockdown period of the year 2020. To ascertain the change in air quality (if any) is related to the COVID-19 lockdown, the same period of the previous year, 2019 is also considered to compare the air quality data with the no-lockdown period (Naqvi et al., 2021). As the air quality depends on meteorological factors, that is to say, it has seasonal variability, the consideration of the same period of the previous year for comparison can be justified (Liu, et al., 2020), (Bedi et al., 2020). Even though the amount of restrictions on different activities has been eased by the government, the lockdown period of 2021 for the Howrah district as imposed by the government of West Bengal to control the 2nd wave of the Coronavirus, starting from 16th May 2021to 31st July 2021 is to be considered in the study.
1.3 Analysis
After the collection of primary and secondary data, a bifurcated methodology has been adopted that has been shown in Fig. 1, and the steps are described in the subsequent section.
Starting with the left branch of the methodology, the study area was digitized by georeferencing the topo sheets obtained from SOI. The Normalized Differential Vegetation Index(NDVI) map was made using ArcMap software using the data products of Sentinel 2.
Following that, data integration was done for the model input as all the collected data were in different forms, and then the Geographical Weighted Regression (GWR) model was run. GWR is a spatial statistical modeling technique for spatially heterogeneous processes that allows the relationships between a response and a set of covariates to vary across geographic space. In general, the PM2.5-AOD relationship is known to be complicated, varying both spatially and temporally. In this study, to find the phase wise variation of satellite based PM 2.5, a GWR model was used and the local coefficient of determination parameter (R2) for satellite based PM 2.5 and ground measured PM2.5 was calculated to assess the effectiveness of the method. In order to calibrate the parameters in the GWR model, a local inverse distance weighted algorithm is usually employed (Bai et al., 2016). A spatial weight matrix was constructed using spatiotemporal distances between observations.
All other pollutant variation was studied using Google Earth Code Editor by importing Sentinel 5P data products for the study area. The variation of these satellite-observed pollutants was then plotted, and mean concentration maps were created for the different phases of the study period.
For the remaining branch of the methodology, the ground-measured pollutants data were aggregated, and the subindices of pollutants were calculated to eventually find the Air Quality Index (AQI) using the CPCB method of calculation that is followed in India.
The AQI can be calculated by evaluating the values of Sub-indices for each pollutant. The Sub-indices at a monitoring location for individual pollutants are calculated using their daily average concentration value (8-hourly in the case of CO and O3) and health breakpoint concentration range. The worst sub-index is the AQI for that location. The general equation, as mentioned in CPCB, 2014, for the sub-index (Ii) for a given pollutant concentration (Cp) is calculated as follows:
$$\text{I}\text{i}=\left[\left\{\frac{\text{I}\text{HI} - \text{I}\text{LO}}{\text{B}\text{HI} -\text{B}\text{LO}}\right\}\text{*} \left(\text{C}\text{p}-\text{B}\text{LO}\right)\right]+ \text{I}\text{LO}$$
Where,
BHI= Breakpoint concentration greater or equal to given concentration.
BLO= Breakpoint concentration smaller or equal to given concentration.
IHI =AQI value corresponding to BHI
ILO = AQI value corresponding to BLO
Cp = Pollutant concentration
The AQI category is decided based on ambient concentration values of air pollutants and their likely health breakpoints. AQ sub-index and health breakpoints are developed for eight pollutants (PM10, PM2.5, NO2, SO2, CO, O3, NH3, and Pb) for which short-term (up to 24 hours) NAAQS are prescribed. The categories are divided into Good, Satisfactory, Moderate, Poor, Very Poor, and Severe according to the range of concentration of the pollutants. Several studies have been conducted using this Indian National Air Quality Standard (D. Kumar et al., 2021), (Mahato & Ghosh, 2020), (Markandeya et al., 2021), (Mor et al., 2021) and many others.
Furthermore, owing to the normality of the data, a Pearson correlation test was done to find the relation between meteorological variables and the temporal variation of AQI to study the seasonal variation of the AQI, as the meteorological conditions affect the ground-level pollutant concentration (Sathe et al., 2021). Then using IDW interpolation, the ground measured AQI in the control stations of CPCB was interpolated for the study area to see the spatial variation of the ground measured pollutants over different phases of the study period. The IDW interpolation method was considered as it has been proven to be a very efficient method to show the variation of continuous data over space (Bai et al., 2016). Inverse distance weighting is a type of deterministic method of multivariate interpolation with a known set of scattered points. The interpolated values are calculated using weights assigned in the known points. It generates high-resolution maps with smaller errors in space and time (Ramos et al., 2016) (Das et al., 2017).
2.4 Study Area
Howrah is home to over 4.85 million inhabitants, a sizeable population with a growth rate of 13.31%, having a significant economy and industry based around engineering, with a history of this dating back hundreds of years. Moreover, even though the forest is an important resource for raw products, its presence in the district is insignificant (MSME, 2020). Nowadays, the area sees itself subjected to various pollutants along with rapid urbanization and growth. PM 2.5 being the major source of pollution, the area is also affected a great deal by NO2, SO2, and O3 from various sources, including congested traffic to factories. In 2019, Howrah recorded a PM 2.5 reading of 55.9 µg/m³ as its yearly average, a considerably high value that placed it within the ‘unhealthy’ range. This reading also positioned it in 38th place out of all the cities that were ranked in India, indicating that Howrah has a long way to go to get its air quality to more appreciable and safe levels. With such record background, it became evident to study the effect of Covid-19 lockdown on its air quality to see whether restrictions bring down the AQI to a safe level or not.