Version of Record: https://www.sciencedirect.

com/science/article/pii/S0959652621028043
Manuscript_14c49a05c2e46bbcfb261cf25195d040

1 Google Earth Engine Based Spatio-temporal Analysis of Air Pollutants Before and
2 During the First Wave COVID-19 Outbreak over Turkey via Remote Sensing

3 Fatemeh Ghasempour 1, Aliihsan Sekertekin 2,*, Senol Hakan Kutoglu 1

4 1 Department of Geomatics Engineering, Bulent Ecevit University, Zonguldak 67100, Turkey
5 2 Department of Geomatics Engineering, Cukurova University, 01950 Ceyhan/Adana, Turkey

7 Abstract

8 Air pollution is one of the vital problems for the sustainability of cities and public health.
9 The lockdown caused by the COVID-19 outbreak has become a natural laboratory,
10 enabling to investigate the impact of human/industrial activities on the air pollution. In this
11 study, we investigated the spatio-temporal density of TROPOMI-based nitrogen dioxide
12 (NO2) and sulphur dioxide (SO2) products, and MODIS-derived Aerosol Optical Depth
13 (AOD) from January 2019 to September 2020 (also covering the first wave of the COVID-
14 19) over Turkey using Google Earth Engine (GEE). The results showed a significant
15 decrease in NO2 and AOD, while SO2 unchanged and had slightly higher concentrations in
16 some regions during the lockdown compared to 2019. The relationship between air
17 pollutants and meteorological parameters during the lockdown showed that air temperature
18 and pressure were highly correlated with air pollutants, unlike precipitation and wind speed.
19 Moreover, Purchasing Managers' Index (PMI) data, indicator of economic/industrial
20 activities, also provided poor correlation with air pollutants. TROPOMI-based NO2 and
21 SO2 were compared with station-based pollutants for three sites (suburban, urban, and
22 urban-traffic classes) in Istanbul, revealing 0.83, 0.70 and 0.65 correlation coefficients for
23 NO2, respectively, while SO2 showed no significant correlation. Besides, AOD data were
24 validated using two AERONET sites providing 0.86 and 0.82 correlation coefficients.
25 Overall, the satellite-based data provided significant outcomes for the spatio-temporal
26 evaluation of air quality, especially during the first wave of the COVID-19 lockdown.

27 Keywords: Remote Sensing, Google Earth Engine, Air Pollution, COVID-19, TROPOMI,
28 Turkey

* Corresponding Author
E-mail: [email protected]
1

© 2021 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
29 1) INTRODUCTION

30 Air pollution is one of the most important and vital challenges in developing countries,
31 which occurs as a result of rapid urban development, the expansion of the manufacturing
32 industry and the use of fossil fuels in factory, and residential activities ( He et al., 2017; Xu
33 and Lin, 2017; Perera, 2017). In recent decades, the issue of pollution and environmental
34 degradation have been among the main concerns at the domestic, regional and international
35 levels (Eckhoff, 2009). Air pollution occurs when large volumes of harmful particles or
36 substances such as gases, particulate matters, and biological molecules enter the Earth's
37 atmosphere. It can also increase global warming or disease in humans such as Asthma,
38 Bronchitis, shortness of breath, heart attacks and various respiratory allergies (Arya, 1999;
39 Eckhoff, 2009). Among the air pollutant gases, the most critical pollutants in the world's
40 metropolitan areas are carbon monoxide (CO), ozone (O3), sulfur dioxide (SO2), nitrogen
41 dioxide (NO2), Particulate Matters (PM2.5 and PM10). These pollutants are generally emitted
42 from both stationary and mobile resources. Stationary resources involve domestic,
43 residential and industrial activities, while transportation and vehicles that pollute the air by
44 generating pollutants from exhaust fumes and particulate matter are among the mobile
45 sources of the air pollution. In fact, except for SO2, which is mainly emitted from
46 residential and industrial sources, the other pollutants are generally emitted much more by
47 the vehicles.

48

49 Satellite-based remote sensing technology offers an effective way of long-term spatio-

50 temporal monitoring of air quality under various scales. The use of satellite imagery to
51 monitor air pollution dates back to the 1970s with the utilization of the Advanced Very
52 High Resolution Radiometer (AVHRR), Landsat, and the Geostationary Operational
53 Environmental Satellite (GOES). In addition, Nimbus-7, European Remote-sensing
54 Satellite-2(ERS-2), Terra, ENVironmental SATellite (ENVISAT), Aqua, Aura,
55 Meteorological OPerational satellite (MetOP), and Sentinel-5 precursor (Sentinel-5P) are
56 among the other satellites that have been frequently used for air quality monitoring since
57 1978. Concerning the literature, many researchers have studied monitoring and retrieval of
58 air pollutants such as NO2, SO2, PMX, CH4, CO, and O3 using these remote sensing
59 satellites (Laat et al., 2020; Lin et al., 2019; Ruoying et al., 2019; Salmabadi and Saeedi,
60 2019; Wang et al., 2019; Feng et al., 2019; Hajiloo et al., 2019; Wu et al., 2019; Borsdorff
2
61 et al., 2019; O’Brien et al., 2016; Tzortziou et al., 2018; Xu et al., 2018). The
62 TROPOspheric Monitoring Instrument (TROPOMI) onboard Sentinel-5P, launched on 13
63 October 2017, is the first Copernicus mission and the latest satellite-based sensor for air
64 quality monitoring(Butz et al., 2012). It has been providing high spatio-temporal resolution
65 data for monitoring the air quality parameters, namely, Absorbing Aerosol Index (AAI or
66 UVAI), O3, CO, NO2, SO2, CH4, cloud characteristics and formaldehyde (HCHO) (El
67 Khoury et al., 2019).

68

69 In addition to the aforementioned satellite-based air pollution studies, in late 2019,

70 monitoring and investigating the effect of coronavirus (COVID-19) lockdown on air
71 pollution has attracted the attention of the scientists as a milestone for air quality
72 management all around the world. Ogen (2020) investigated the relationship between the
73 COVID-19 mortality and long-term exposure to NO2 pollution via Sentinel-5P in Italy,
74 Spain, France and Germany. The results indicated that the long-term exposure to NO2 may
75 be one of the most important contributors to fatality caused by the COVID-19 in these
76 regions (Ogen, 2020). Another study was about monitoring the effect of NO2 concentration
77 on the atmosphere during the lockdown in Spain (Mesas-Carrascosa et al., 2020). They
78 concluded that there was a significant correlation between the level of the population’s
79 activity and the reduction in NO2 values. Metya et al. (2020) used ARIS-derived CO, and
80 NO2, and OMI-derived SO2 in India and China during January-April, 2020. Their results
81 indicated that the tropospheric NO2 levels reduced on an average by 17% over India and
82 25% over China. Besides, approximately 17% reduction was observed in the boundary
83 layer SO2 especially over the Eastern sector of India, and the 6.5% reduction of CO was
84 determined over north-central China. Nichol et al. (2020) investigated regional air quality
85 indicators over the Beijing–Tianjin–Hebei (BTH) region and other parts of China during
86 the winter of 2019–2020, when the COVID-19 outbreak in China caused reduction in
87 transport and economic activity from November 2019 to February 2020. They stated that,
88 in March and April 2020, Ozone Monitoring Instrument (OMI) satellite-derived NO2
89 significantly decreased across China. They also highlighted that the increase in PM2.5
90 unlike the decrease in NO2 in BTH and across China is likely due to enhanced production
91 of secondary particulates. Ghahremanloo et al. (2020) utilized SO2, CO, NO2 and HCHO
92 from Sentinel-5P, and daily AOD from the Himawari-8 satellite over BTH, Wuhan, Seoul,

3
 93 and Tokyo during February 2019 and 2020. The comparison analysis between 2019 and
94 2020 indicated the vast pollution reduction in Wuhan with a decline of about 83%, 11%,
95 71%, 4% and 62% in NO2, HCHO, SO2, CO, and AOD, respectively. The finding of
96 Kanniah et al. (2020) determined a notable decrease in tropospheric NO2 during the
97 lockdown period over the Southeast Asia region. Furthermore, some studies also stated that
98 TROPOMI-based NO2 observations were decreased during the lockdown period in some of
99 the big European cities, highly populated Chinese cities and North of Italy, respectively
100 (Cersosimo et al., 2020; Fan et al., 2020; Vîrghileanu et al., 2020). Apart from the satellite-
101 based air pollutant analyses, some researchers also proved the reduction in station-based air
102 pollutants due to the COVID-19 pandemic control strategies (Nakada and Urban, 2020;
103 Othman and Latif, 2021; Chen et al., 2021).

104 The main objective of this study is to monitor and investigate the spatial-temporal patterns
105 of Sentinel-5P derived air pollutants, namely, NO2 and SO2, and MODIS-derived AOD
106 before and during the first wave of the COVID-19 covering the period from January 2019
107 to September 2020 over Turkey. The COVID-19 lockdown period in Turkey mainly covers
108 the period from March 10 to May 31, 2020 for the first wave, and partly June since the
109 restrictions were gradually loosened in June. Thus, we assumed the lockdown period as of
110 March, April and May in 2020. Although there was a partial lockdown instead of a full
111 lockdown in Turkey for the first wave, it was observed that human activities greatly
112 reduced in this period due to the warnings to control the spread of the virus. As it is clear
113 from the given literature, many studies have been published about the effect of the COVID-
114 19 outbreak on air quality. These studies can be categorized into four distinctive groups
115 including (i) station-based studies, (ii) remote sensing based studies, (iii) combination of
116 station-based and remote sensing based studies, and (iv) remote sensing based studies with
117 meteorological/climatological parameters (Karuppasamy et al., 2020; Broomandi et al.,
118 2020; Elshorbany et al., 2021; Mesas-Carrascosa et al., 2020; Metya et al., 2020;
119 Ghahremanloo et al., 2020; Nakada and Urban, 2020; Rojas et al., 2021; Vîrghileanu et al.,
120 2020; Zhang et al., 2020c). Concerning the originality of this research, this study includes a
121 combined analysis of all four approaches mentioned above with also an addition of
122 Purchasing Managers' Index (PMI), which is an indicator of economic/industrial activities.
123 Furthermore, satellite-based pollutants and meteorological parameters were visualized,
124 analyzed and extracted using Google Earth Engine (GEE) open-source geospatial analysis
125 platform, which enables working on JavaScript programming language.
4
126 2) STUDY AREA

127 Turkey, with 81 provinces and a population of 83.15 million (https://www.nufusu.com/), is

128 a unique country that is lying partly in Asia and partly in Europe at geographical
129 coordinates of 36°- 42° N and 26°- 45°E. The territory of Turkey has a length of 1,660 km
130 from east to west and an area of 780043 square kilometers. The average altitude of the
131 country is about 1132 meters from mean sea level and there are four different types of
132 climate containing Mediterranean climate, Black Sea Climate, Marmara climate and
133 Continental climate. Ankara, the capital of Turkey, and Istanbul, the country's economic,
134 cultural and historical center are the most populous cities in Turkey. Figure 1 represents the
135 study area and the ground stations via Digital Elevation Model (DEM) obtained from
136 Shuttle Radar Topography Mission (SRTM).

137

138 Figure 1. Illustration of the study area, Turkey, and ground stations via SRTM-DEM.

139

140 3) MATERIALS AND METHODOLOGY

141 3.1) Data Collection and Preprocessing

142 In addition to several studies about air pollution monitoring during the COVID-19
143 pandemic lockdowns (Ghahremanloo et al., 2020; Kanniah et al., 2020; Mesas-Carrascosa
144 et al., 2020; Metya et al., 2020; Nichol et al., 2020; Ogen, 2020; Tobías et al., 2020), we
145 have investigated the impact of the COVID-19 lockdown on the air quality over the whole
5
146 Turkey for the period from January 2019 to September 2020. As shown in Table 1, we
147 utilized six types of data sets in this study: (a) Hourly NO2 and SO2 concentrations from the
148 Turkey Air Quality Monitoring Center (TAQMC), (b) Daily satellite image data of
149 Sentinel-5P NO2 and SO2 products from the GEE, (c) Daily satellite image data of MODIS
150 AOD product from the GEE platform, (d) ERA5 Reanalysis climatic and meteorological
151 data, (e) AErosol RObotic NETwork (AERONET) data, and (f) Purchasing Managers'
152 Index (PMI). The connections between the data sets and their usages are presented in the
153 following sections.

154 Table 1. List of all datasets used in this study

Temporal
Category Variables Units Data Source
Resolution

Satellite tropospheric_NO2_column_number_d
molec/cm2 https://developers.googl
image data of ensity daily e.com/earth-

Sentinel-5P engine/datasets/
SO2_column_number_density molec/cm2

Satellite https://developers.googl
image data of Optical_Depth_047 * daily e.com/earth-

MODIS engine/datasets/

Pollutant
Concentratio
n (Turkey https://www.havaizleme
NO2, SO2 μgr/cm3 hourly
Air Quality .gov.tr/

Monitoring
Center)

v_component_of_wind_10m m/s

u_component_of_wind_10m m/s https://developers.googl

ERA5 monthly e.com/earth-
mean_2m_air_temperature k engine/datasets/

total_precipitation m

6
 surface_pressure pa

AOD_500nm
https://aeronet.gsfc.nasa
AERONET * hourly
.gov/
AOD_440nm

https://www.investing.c
om/economic-
PMI PMI data * monthly calendar/turkish-
manufacturing-pmi-
1305

155

156 3.1.1) Satellite remote sensing data

157 In this study, we have employed satellite image collections to analyze the spatio-temporal
158 distribution of air pollutants, namely, NO2, SO2, and AOD before and during the COVID-
159 19 pandemic lockdown. The daily Sentinel-5P and MODIS data were used to produce the
160 maps of monthly spatial patterns of the air pollutants. We acquired the tropospheric vertical
161 column density of offline level 3 NO2 (COPERNICUS/S5P/OFFL/L3_NO2) and the
162 vertical column density at the ground level of offline level 3 SO2
163 (COPERNICUS/S5P/OFFL/L3_SO2) from the TROPOMI instrument on board the
164 Sentinel-5P satellite. The TROPOMI is an advanced multispectral imaging spectrometer
165 with nadir-viewing and wavelengths of Ultraviolet-Visible (UV-VIS, 270nm to 495nm),
166 Near Infrared (NIR, 675nm to 775nm), and Shortwave Infrared (SWIR, 2305nm-2385nm).
167 Besides, it is tracking the unique effects of atmospheric gases in different parts of the
168 electromagnetic spectrum. It uses passive remote sensing techniques to reach the target
169 above the atmosphere by measuring the solar radiation reflected by the Earth. The sensor
170 works in a cross direction, with ~ 2600 km swath width on the ground surface. The nadir
171 view for all spectral bands is 7x3.5 km2, except SWIR bands with 7x7 km2 and UV band
172 with 7x28 km2 (Veefkind et al., 2012). However, in this study, all these Sentinel-5P derived
173 products are delivered in 0.01 arc degree (~1.11 km) spatial resolution from the GEE. In
174 addition to Sentinel-5P data, we collected the daily AOD images of the MODIS
175 (MODIS/006/MCD19A2_GRANULES) onboard the Terra and Aqua satellites with 1 km
176 spatial resolution. Google Earth Engine, with a powerful open-source processing system in
177 JavaScript programming language, enables users to perform calculations, analyze and
7
178 visualize large volumes of data without the need for powerful systems. The system supports
179 a variety of widely used satellite data such as Landsat, Sentinel and MODIS that are
180 available free of charge.

181

182 3.1.2) Turkey Air Quality Monitoring Center (TAQMC) data

183 Hourly pollutants’ concentration data can be gathered from the Turkish air quality real-time
184 release platform (https://www.havaizleme.gov.tr/) administered by the Turkey Air Quality
185 Monitoring Center (TAQMC), one of the major institutions under the Environment and
186 Urban Ministry. The extensive monitoring stations provide continuous hourly pollutant
187 measurements. In this study, we used the ground-based measurements of NO2 and SO2
188 concentrations that were obtained from three hotspots monitoring stations as urban traffic
189 (Beşiktaş), urban background (Başakşehir), and suburban (Şile) stations over the most
190 populous cities in İstanbul. Ground-based NO2 and SO2 values were analyzed together with
191 Sentinel-5P derived NO2 and SO2 data over the related test sites. A time interval of ±0.5 h
192 was used to match the datasets of TAQMC observations with the satellite pass time. Thus,
193 the hourly pollutant concentrations from the TAQMC were used for the comparison of the
194 data of the ground stations and the satellite acquisitions.

195
196 3.1.3) Aerosol Robotic Network (AERONET) data

197 The Aerosol Robotic NETwork (AERONET) is a globally ground-based remote sensing
198 aerosol network established and distributed by NASA and PHOTONS. For more than 25
199 years, the AERONET has been using Cimel Electronique Sun/Sky radiometers to prepare
200 extremely accurate ground truth measurements. In this study, we have used the version 3
201 level 2 AOD data that have high-quality ground-based assurance for the validation of
202 MODIS AOD products. The AERONET dataset is divided into 3 quality levels including
203 raw data from level 1, cloud-screened data from level 1.5, and quality assurance data from
204 level 2 (Smirnov et al., 2000). The AERONET version 3 level 2 data may be downloaded
205 from the AERONET website (https://aeronet.gsfc.nasa.gov/The AOD in the AERONET
206 version 2 was semi-automatically controlled in near real-time quality using a cloud-
207 screening method, while the AERONET Version 3 algorithm provides fully automated
208 cloud screening and quality control (Giles et al., 2019).
8
209
210 3.1.4) ERA5 reanalysis data

211 In addition to the previous data sets, we employed daily meteorological/climatological data
212 from the latest ERA5 reanalysis produced by the European Center for Medium-range
213 Weather Forecasts (ECMWF) under the Copernicus Climate Change Service. Although the
214 main reason for air pollution is the increase of pollutant concentration in the atmosphere,
215 the local and regional meteorological/climatological conditions have also affected the air
216 quality (Ghahremanloo et al., 2020; Kanniah et al., 2020; Nichol et al., 2020). Thus, in this
217 study, we also investigated the relationship between satellite-derived air pollutants and
218 ERA5’s meteorological/climatological parameters, namely, the eastward-northward
219 components of wind speed (WS), air temperature (Temp.), surface pressure (Pres.) and total
220 precipitation (Prec.) with a spatial resolution of 0.25°x 0.25°.

221

222 3.1.5) Purchasing Manager Index data

223 The Purchasing Managers' Index (PMI) is one of the most important indicators of the
224 economic activity that examines companies 'purchasing managers' tendency to purchase
225 goods and services. It is one of the indices that can explain the growth forecasts of
226 countries in the best way (Eren, 2014) (http://www.bireyselyatirimci.com/pmi-endeksi-
227 nedir/). On the other hand, faster expansion in industrial activities or any reduction in PMI
228 may cause to increase or decrease in air pollutants. Thus, we examined whether there was
229 any correlation between PMI and satellite-derived pollutants. Besides, this analysis may
230 explain if industrial activities have a high impact on air pollutants by expecting the
231 pollutants should reduce in COVID-19 lockdown days.

232
233 3.2) Methodology

234 In this study, we focused on the following analyses: 1) monthly NO2, SO2 and AOD images
235 and their statistics were extracted for 2019 and 2020. 2) Hourly TROPOMI-based NO2
236 values were compared with three station-based NO2 in Istanbul. 3) The performance of
237 MODIS AOD was validated using AOD from two ground-based stations. 4) The
238 relationship between monthly average meteorological/climatological parameters and

9
239 monthly average satellite-based pollutants was investigated. Finally, 5) The relationship
240 between monthly average satellite-based pollutants and (PMI) was examined. Figure 2
241 represents the workflow used to analyze pollutant concentration and related variables,
242 which involve five source data sets, namely, (i) Sentinel-5P TROPOMI-based NO2 and
243 SO2 products, and MODIS-based AOD product; (ii) meteorological parameters (e.g.
244 Temp., Pres., Prec., WS) from ERA5, and (iii) AERONET AOD data, (iv) NO2 and SO2
245 ground measurements from TAQMS stations, and (v) PMI dataset. In this study, we used
246 free image collection of TROPOMI and MODIS instruments in GEE datasets. GEE is an
247 open-source geospatial analysis platform that enables users to globally or locally visualize
248 and analyze changes, map trends, and quantify differences on the Earth's surface. Many
249 studies related to land cover changes (Huang et al., 2017), crop mapping (Clemente et al.,
250 2020), disaster management (DeVries et al., 2020), surface water change (Xia et al., 2019),
251 and natural resource management (Aksoy et al., 2019; Leinonen et al., 2018) were carried
252 out successfully and efficiently on the GEE platform. Concerning the analyses, a JavaScript
253 program was developed in GEE to acquire, correct and visualize the data from image
254 collections. The codes used in this paper can be found in the supplementary material. In this
255 stage, the selected images are processed using the following steps:

256 (a) filtered the imaging time over the clipped region of interest (whole Turkey),

257 (b) removed cloudy pixels using a conditional (cloud_fraction > 0.5) on TROPOMI-based
258 products (Fioletov et al., 2020; Ialongo et al., 2020; Lorente et al., 2019; Lorente et al.,
259 2017; Van Geffen et al., 2018) and using best quality condition (for QA band AOD = 0) on
260 MODIS-derived AOD data,

261 (c) extracted daily images by mosaicking the overlapping scenes over the whole study zone,

262 (d) extracted hourly AOD pixel value, NO2 and SO2 column density pixel values from
263 images covering the ground stations (TAQMS stations and AERONET sites) for the
264 validation processes,

265 (e) calculated the monthly images by averaging the daily images for preparing monthly
266 spatio-temporal maps and also for calculating the relationship between meteorological
267 parameters in the lockdown months. Furthermore, for investigating the relationship
268 between PMI and air pollutants.

10
269 Hourly data were only used for the validation step. TROPOMI-based NO2 and SO2 data
270 were compared with TAQMS ground-measurements for three test sites (urban, urban-traffic
271 and suburban) in the city of Istanbul. To evaluate MODIS AOD 470nm data, we applied
272 AERONET data (AOD440 and AOD500) with a time tolerance of ± 0.5 h. Since
273 AERONET does not provide AOD 470 nm, the mean value of AERONET AOD 440 nm
274 and 500 nm values were calculated for AOD 470 nm retrieval relevant to MODIS AOD
275 470nm (Equation 1).

AOD 500nm +AOD(440nm)

AOD 470nm = (1)
2

276 We utilized four statistical metrics below, namely, correlation coefficient (R), Root Mean
277 Square Error (RMSE), Standard Deviation of Error (STD Error), and Bias to examine the
278 relationship between MODIS AOD and AERONET AOD.

∑ yi -yi xi -xi
R=
2 (2)
∑ yi -yi ∑ xi -xi 2

2
∑ xi -yi (3)
RMSE=
n

2
∑ yError -yError (4)
SDT of Error=
n

∑ xi -yi
Bias= (5)
n

279 Where y , x , y , and x are the MODIS-derived AOD, AERONET-based AOD, their mean
280 values, respectively. Thus, y refers to the difference between MODIS-derived AOD
281 and AERONET-based AOD, and refers to the mean of differences.

282 In the last stage, to investigate if the meteorological/climatological parameters play an

283 important role in the results of pollutants in the lockdown period, the monthly
284 meteorological data of ERA5 with the selected information layer were used to extract the

11
285 correlation coefficients. Furthermore, the relationship between monthly pollutants and PMI
286 was also investigated to find out the effect of PMI on pollutants.

287

288 Figure 2. The workflow illustrating the operations and relations between the methods for
289 the evaluation of air pollutants and related variables.

290 To summary the innovation in the methodology; i) a JavaScript program in the GEE was
291 developed to automate the image analysis, ii) satellite-based pollutants were compared and
292 analyzed with ground based observations from different land cover types, iii) the
293 relationship between meteorological parameters/PMI and satellite-based pollutants was
294 analyzed.

295 4) RESULTS AND DISCUSSION

296 4.1) Spatio-Temporal Distribution of Tropospheric NO2 Column Density over Turkey

297 The lockdown in Turkey due to the outbreak of COVID-19, mainly from March 10, 2020 to
298 May 31, 2020, decreased transportation activities that lead to lower energy consumptions
299 and less oil request. Moreover, it provided a good opportunity to study its impacts on air
300 quality by comparing some pollutant concentrations. Although there are many key
12
301 parameters that may change the air pollution concentration, this study is conducted based
302 on the assumption of the used parameters and corresponding analyses. One of the important
303 trace gases in the atmosphere is NO2. While some nitrogen oxides naturally occur, some of
304 them are due to human activities. The major part of NO2 produced by human activities is
305 related to the burning of fuel in stationary sources and resulting from combustion in
306 vehicles and industrial enterprises (Ghahremanloo et al., 2020; Kanniah et al., 2020).

307 Figure 3 shows the monthly comparisons of spatial tropospheric NO2 images over Turkey
308 from January 2019 to September 2020. There are no-data pixels in the east-direction of
309 images such as January and February that result from cloudy pixel conditions and quality
310 assurance (QA) limitation of the algorithms (Ialongo et al., 2020; Lorente et al., 2017;
311 Nichol et al., 2020; Omrani et al., 2020; Zheng et al., 2019). As seen in Figure 3, these
312 images display a general reduction in NO2 distribution when comparing January-Jun 2019
313 with January-June 2020. In particular, the spatial distribution of NO2 decreased
314 significantly during the lockdown period (March-May 2020) compared to the same period
315 of 2019. In order to support these visual interpretations, the area information in Table 2,
316 where NO2 > 4*1015 molec/cm2, was extracted. The area information from January to June
317 2019 varied from 23800.3 km2 to 3476.6 km2, respectively, whereas it ranged from
318 23509.2 km2 to 356.1 km2 for the same period of 2020. However, after the lockdown
319 restrictions are relaxed in July, the NO2 concentration started to increase in July, August
320 and September 2020 compared to the same months of 2019. The area information in Table
321 2 also reveals this increase. This observable increase was probably due to the rising vehicle
322 emissions and industrial activities after the lockdown period. The tropospheric NO2
323 reduction obtained during the lockdown period in Turkey coincides with similar studies
324 conducted over different countries(Fan et al., 2020; Mesas-Carrascosa et al., 2020; Metya et
325 al., 2020; Nichol et al., 2020; Tobías et al., 2020; Vîrghileanu et al., 2020).

326 Concerning the general evaluation of the spatial variability of NO2 pollution over Turkey,
327 Marmara and Aegean regions show the highest spatial variability of NO2, whereas the
328 Mediterranean, Southeastern Anatolia, and central Anatolia partly revealed high NO2
329 distribution.

330

13
331

332 Figure 3. The monthly images of tropospheric NO2 column density over Turkey.

14
333

334 Figure 3 (Continued). The monthly images of tropospheric NO2 column density over
335 Turkey.

336 In addition to the spatio-temporal comparisons of 2019 and 2020 images, it would be better
337 to provide quantitative analyses for this period as in Figure 4 and Table 2. Figure 4 reveals
338 the monthly average TROPOMI-based NO2 column density over Turkey in 2019 and 2020.
339 As a general evaluation, the trend of monthly NO2 averages shows the maximum amount in
340 the winter, while the minimum amount in the summer. In Figure 4, a decrease was observed
341 in monthly average NO2 values from January to July in 2020, and these values were also
342 lower than 2019 monthly averages. Considering the lockdown period, the monthly average
343 NO2 values declined in 2020 compared to the same period of 2019. On the other hand, after

15
344 the lockdown period, the monthly average NO2 values increased dramatically in 2020,
345 which were also higher than in 2019. Table 2 shows the results of the statistical parameters,
346 namely, average, standard deviation (S.D), minimum (min), maximum (max), and percent
347 of change rates (PCR) obtained from the monthly NO2 images. The average value of NO2
348 column density dropped from 1.66*1015 molec/cm2 to 1.05*1015 molec/cm2 from January
349 to June, and then increased to 1.79*1015 molec/cm2 in December in 2019. On the other
350 hand, in 2020, the average NO2 values started from 1.52*1015 molec/cm2 in January, then
351 slightly decreased to 0.92*1015 molec/cm2 in May, and finally reached 1.28*1015
352 molec/cm2 in September. The statistical results in Table 2 demonstrate that a descending
353 trend was observed in 2020 in comparison with 2019. In addition, the monthly PCR
354 between 2019 and 2020 for March, April, May and June were calculated as -16.5%, -
355 10.8%, -12.49% and -18.56%, respectively, proving the descending trend in the lockdown.
356 It is obvious that the lockdown of the COVID-19 outbreak and staying at home have caused
357 a significant reduction in tropospheric NO2 column density.

2019 2020
TROPOMI-based NO2
column density [1015

2.00
molec/cm2 ]

1.80 Lockdown Period

1.60
1.40
1.20
1.00
0.80
January

September
June

August
February

April

May

July

October

November

December
March

Month
358

359 Figure 4. The monthly average TROPOMI-based NO2 column density over Turkey in 2019
360 and 2020.

361 Table 2. Average, standard deviation, maximum, minimum, and percentage of change rate
362 values of the TROPOMI NO2 column density calculated during January-September 2019
363 and 2020 over Turkey.

Month Jan. Feb. Mar.

Year 2019 2020 2019 2020 2019 2020

16
Average 1.6581 1.5213 1.4235 1.3764 1.4557 1.2148

S.D 1.0930 1.1745 1.1187 0.9953 0.8698 0.7736

Min -1.4851 -2.1240 -1.4960 -1.0804 -0.7089 -1.0800

Max 13.1696 18.6903 39.3789 22.9577 12.4227 9.7954

Area (km2)
23800.3 23509.2 18291.8 16129.8 13388.6 9027.1
for NO2>4

PCR -8.25 -3.31 -16.55

Month Apr. May Jun.

Year 2019 2020 2019 2020 2019 2020

Average 1.1364 1.0129 1.0530 0.9214 1.2372 1.0076

S.D 0.7060 0.4808 0.5037 0.3791 0.3785 0.3999

Min -0.6486 -1.3173 -0.6213 -0.8127 -0.6144 -0.5564

Max 12.5657 6.2744 9.1033 5.9866 7.5727 6.5480

Area (km2)
5733.7 1644.4 3476.6 356.1 808.2 903.8
for NO2>4

PCR -10.86 -12.49 -18.56

Month Jul. Aug. Sep.

Year 2019 2020 2019 2020 2019 2020

Average 1.1591 1.2772 1.2227 1.2437 1.1642 1.2789

S.D 0.4211 0.4093 0.3626 0.4532 0.5225 0.5005

Min -0.4497 -0.2739 0.0937 0.1062 0.0304 0.2295

Max 6.8161 7.9473 7.7835 7.2345 7.2871 7.4934

17
 Area (km2)
1270.2 1689.8 618.6 1920.0 4225.0 3246.3
for NO2>4

PCR 10.19 1.71 9.85

364

365 4.2) Spatio-Temporal Distribution of SO2 Column Density over Turkey

366 SO2 is one of the tropospheric trace gases that is mostly produced by volcanoes and
367 industrial activities. Due to the presence of sulfur compounds in oil and coal, their
368 combustion leads to the production of SO2. We analyzed the spatio-temporal variations of
369 the monthly average SO2 column density over Turkey as in Figure 5. No-data pixels are
370 available from January to April in both 2019 and 2020 data sets. The main reasons are
371 cloud fraction condition, the algorithm limitations for producing the data and quality
372 assurance contrast. Due to noise on the SO2 values, negative vertical column concentrations
373 are perceived as low SO2 emissions. According to the Sentinel-5P product readme file
374 (https://sentinel.esa.int/web/sentinel/technical-guides/sentinel-5p/products-algorithms), it is
375 recommended to filter outliers as concentration values lower than -0.001 mol/m2.

376 Since the number of no-data pixels is too many in Figure 5 for January, February and
377 March of 2019 and 2020 images, the proper comparison is not possible for these months.
378 However, concerning the concentrations of SO2 values in these months, 2020 datasets had
379 significantly higher concentrations than 2019. The area information in Table 3, where SO2
380 > 60*1015 molec/cm2, was extracted to observe the areal changes. The area information
381 from January to September 2019 varied from 38167.5 km2 to 307.9 km2, respectively,
382 while it ranged from 152073.3 km2 to 317.0 km2 for the same period of 2020. In 2020, the
383 area information of September image was the only image that showed less area than in
384 2019. The lockdown months, except March, showed no significant decrease compared to
385 2019. On the contrary, the spatial distribution of SO2 values in the 2020 lockdown period
386 had slightly higher concentrations than the same period of 2019. These results revealed that
387 the lockdown had no effect on reducing the pollution caused by SO2. On the other hand,
388 after the lockdown restrictions are relaxed in July, the monthly SO2 distributions were
389 identical for the following months in both 2019 and 2020. Concerning the literature, some
390 studies also revealed similar results as in our study that the column density of SO2 remained

18
391 unchanged or increased in some regions of the related countries (Ghahremanloo et al.,
392 2020; Tobías et al., 2020).

393

394 Figure 5. The monthly images of SO2 column density over Turkey.

395

19
396

397 Figure 5 (Continued). The monthly images of SO2 column density over Turkey.

398 Figure 6 represents the monthly average TROPOMI-based SO2 column density over
399 Turkey in 2019 and 2020. As a general evaluation, the trend of monthly SO2 averages
400 shows the maximum amount in the winter, while the minimum amount in the summer. In
401 Figure 6, a decrease was observed in monthly average SO2 values from January to
402 September in 2020, and these values were close to or higher than 2019 monthly averages.
403 Concerning the lockdown period, the monthly average SO2 values were slightly higher in
404 2020 compared to the same period of 2019. Table 3 provides the results of the statistical
405 parameters, namely, average, S.D, min, max and PCR obtained from the monthly SO2
406 images. The average value of SO2 column density dropped from 23.51*1015 molec/cm2 to

20
407 3.48*1015 molec/cm2 from February to July, and then increased to 72.46*1015 molec/cm2 in
408 December 2019. On the other hand, in 2020, the average SO2 values started from
409 50.50*1015 molec/cm2 in January, then slightly decreased to 2.40*1015 molec/cm2 in
410 September. The statistical results in Table 3 (average and PCR) demonstrate that an
411 ascending trend was observed when comparing 2020 with 2019. The only negative monthly
412 average SO2 trend between 2020 and 2019 was observed in September as -66.71%. The
413 results of the monthly average SO2 values showed that the lockdown of COVID-19
414 outbreak and staying at home had no significant reduction on tropospheric SO2 column
415 density.

416

2019 2020
TROPOMI-based SO2 column
density [1015 molec/cm2 ]

80.80
70.80 Lockdown Period
60.80
50.80
40.80
30.80
20.80
10.80
0.80
June

August
January

February

April

May

July

September

October

November
March

December
Month
417

418 Figure 6. The monthly TROPOMI-based SO2 column density over Turkey in 2019 and
419 2020.

420 Table 3. Average, standard deviation, maximum, minimum, and percentage of change rate
421 values of the TROPOMI SO2 column density calculated during January-September 2019
422 and 2020 over Turkey.

Month Jan. Feb. Mar.

Year 2019 2020 2019 2020 2019 2020

Average 17.7540 50.4978 23.5050 33.1631 13.8294 15.9301

21
S.D 32.2796 52.6739 20.4712 31.2283 13.8729 17.9774

Min -60.2049 -60.2131 -60.2036 -60.1869 -59.9939 -60.1687

Max 378.6225 971.4098 618.0464 628.6281 284.0248 455.0226

Area(km2)
for 38167.5 152073.3 22561.5 93253.3 2661.4 8845.8
SO2>60

PCR 184.43 41.09 15.19

Month Apr. May Jun.

Year 2019 2020 2019 2020 2019 2020

Average 5.3751 9.3876 4.4431 4.9014 3.4746 4.7309

S.D 9.9822 11.5643 6.3090 7.0158 4.8672 5.7577

Min -55.6041 -59.8140 -24.3164 -27.1609 -31.1326 -21.0456

Max 190.2 287.1 56.1 187.7 54.7 70.2

Area (km2)
for 978.1 2149.7 0.0 54.8 0.0 19.4
SO2>60

PCR 74.65 10.31 36.16

Month Jul. Aug. Sep.

Year 2019 2020 2019 2020 2019 2020

Average 4.4094 4.6210 2.7875 3.2505 7.2199 2.4038

S.D 6.4622 6.2672 6.6889 6.0039 6.9321 6.2358

Min -19.2326 -32.0067 -41.5177 -46.8179 -29.2360 -28.8736

Max 108.5817 111.0524 101.9469 108.5692 109.0529 92.9221

22
 Area (km2)
for 523.5 410.3 307.9 317.0 486.8 176.5
SO2>60

PCR 4.80 16.61 -66.71

423
424 4.3) Spatio-Temporal Distribution of MODIS-based AOD over Turkey

425 Aerosols are suspensions of solid particles or liquid droplets in a gas, and they generally
426 refer to dust particles such as ash, mist, smog, etc. Human-made aerosols are commonly
427 found near large human habitats. The telemetry parameter used in this connection is the
428 optical depth of the atmosphere. The aerosol optical depth (AOD) index is a measure of an
429 aerosol that is calculated based on the ratio of solar waves propagated and absorbed in the
430 atmosphere along the wavelength and the path traveled. This index is a quantitative
431 parameter to indicate the density and concentration of particles in the atmosphere and is
432 generally calculated along a vertical path (Filonchyk and Yan, 2019; Wei et al., 2019).

433 In order to provide the best quality AOD data, masking operation was applied to the AOD
434 Quality Assessment (AOD_QA) band for generating monthly AOD images in Figure 7.
435 Although the number of no-data pixels is too many in Figure 7 for January, February,
436 March and April, proper comparison can be made since the spatial presence of the AOD
437 data are identical in both 2019 and 2020 images. Concerning the spatial distribution of
438 AOD in January and February 2020 images have lower AOD concentration than 2019
439 images. However, in March, the AOD distribution is higher in 2020 compared to the 2019
440 image. In the following lockdown month, April, a significant decrease in monthly AOD
441 distribution was observed in 2020 compared to the same month of 2019, indicating the
442 effect of lockdown was significant on AOD reduction. Since April 2020 was the month
443 with the tightest restrictions. In the last month of lockdown (May 2020) and June 2020,
444 almost identical AOD distribution was observed as in May-June 2019, while the density of
445 AOD distributions varied from region to region. After the lockdown period, the monthly
446 AOD distribution was higher in July 2020 and September 2020, whereas lower in August
447 2020 compared to the same months of 2019. On the other hand, the area information in
448 Table 4, where AOD > 0.4, was extracted to observe the areal changes. The area
449 information from January to April 2019 varied from 394.6 km2 to 1026.8 km2, respectively,

23
450 while it ranged from 722.4 km2 to 2121.9 km2 for the same period of 2020. Then, from
451 April 2019 to July 2019, the area information was between 1616.5 km2 and 1223.1 km2,
452 respectively, which was higher than the same period of 2020 (940.2 – 820.5 km2). After the
453 restrictions were relaxed in June, the area information of July, August and September 2020
454 was higher than the same months of 2019. The results of some studies also supported our
455 findings by showing a decreasing trend in AOD in different test sites during the lockdown
456 period (Ghahremanloo et al., 2020; Kanniah et al., 2020; Otmani et al., 2020).

24
457

458 Figure 7. The monthly images of MODIS-derived AOD concentration over Turkey.

25
459

460 Figure 7(Continued). The monthly images of MODIS-derived AOD concentration over
461 Turkey.

462 Figure 8 and Table 4 present comparative and quantitative statistical results of monthly
463 AOD values over the study area for both 2019 and 2020. As a general evaluation of Figure
464 8, the trend of monthly AOD averages from April to October is higher than the yearly
465 averages, while the rest of the months reveal lower monthly average AOD than the yearly
466 average. The monthly average AOD values in Figure 8 coincided with the results of the
467 spatio-temporal analysis presented above. We can see a significant decrease in April 2020
468 compared to April 2019. Besides, a sharp increase from 0.096 in March 2019 to 0.0186 in
469 April 2019 can be seen clearly in the spatial distribution map in Figure 7. In 2020, the

26
470 lowest and highest values ranged from 0.096 to 0.186 in February and September,
471 respectively. The statistical results of monthly AOD images in Table 4 display the PCR
472 about -29.84% in April and 20.62% in September considering the comparison between
473 2019 and 2020.

474

2019 2020
MODIS-derived AOD (470 nm)

0.3
0.25 Lockdown Period
0.2
0.15
0.1
0.05
0
January

April

September
June

August
February

May

July

October

November

December
March

Month
475

476 Figure 8. The monthly average MODIS-derived AOD over Turkey in 2019 and 2020.

477

478 Table 4. Average, standard deviation, maximum, minimum, and percentage of change rate
479 values of the MODIS-derived AOD calculated during January-December 2019 and 2020
480 over Turkey.

Month Jan. Feb. Mar.

Year 2019 2020 2019 2020 2019 2020

Mean 0.1298 0.1155 0.1155 0.0889 0.0955 0.1245

S.D 0.0484 0.0447 0.0429 0.0459 0.0407 0.0494

Min 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

Max 0.6040 2.1270 0.5980 1.1300 1.4710 0.8880

27
Area (km2)
for 394.6 722.4 426.0 1191.3 1026.8 2121.9
AOD>0.4

PCR -10.98 -23.07 30.32

Month Apr. May Jun.

Year 2019 2020 2019 2020 2019 2020

Mean 0.1859 0.1304 0.1757 0.1761 0.1736 0.1652

S.D 0.0636 0.0517 0.0761 0.0568 0.0765 0.0680

Min 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

Max 1.1970 4.0000 4.0000 4.0000 4.0000 4.0000

Area (km2)
for 1616.5 940.2 4262.5 1,467.8 1223.1 820.5
AOD>0.4

PCR -29.84 0.24 -4.80

Month Jul. Aug. Sep.

Year 2019 2020 2019 2020 2019 2020

Mean 0.1444 0.1622 0.1695 0.1283 0.1514 0.1826

S.D 0.0619 0.0854 0.0712 0.0897 0.0700 0.0946

Min 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

Max 4.0000 4.0000 4.0000 4.0000 4.0000 4.0000

Area (km2)
for 378.0 1728.6 930.9 514.5 748.6 1050.4
AOD>0.4

PCR 12.38 -24.34 20.62

28
481
482 4.4) Quantitative Analysis of TROPOMI-based and Station-based NO2 and SO2 for
483 Three Test Sites

484 This quantitative analysis relies on geostatistical data processing of the collected data
485 including TROPOMI-based NO2 column density (molec/cm2), TROPOMI-based SO2
486 vertical column density, related and ground station-based NO2 concentrations (µg/cm3). In
487 Istanbul, we selected the statistical data from three ground stations (Figure 1) as traffic-
488 urban (Beşiktaş), urban (Başakşehir), and suburban (Şile) stations. The monthly trendlines
489 revealed the decrease in the tropospheric NO2 column density from January to September
490 for all selected hotspots (Figure 9a-c) in 2019 and 2020. Furthermore, the ground station
491 measurements also confirmed this downward trendline in 2019 and 2020 for all test sites
492 (Figure 9d-f). In the suburban location, due to no data in February (Figure 9c), we could not
493 calculate the satellite image data for 2019 and 2020. The other important outcome of this
494 analysis was that the monthly NO2 levels were higher in 2019 than 2020 for all three sites
495 and both data sets. In the suburban site, TROPOMI-based NO2 (Figure 9c) and station-
496 based NO2 (Figure 9f) showed identical behavior in 2019 and 2020. However, in the urban
497 test sites, the difference between 2019 and 2020, especially in the lockdown months
498 (March-May), are obviously seen for both datasets (Figure 9a-9b and Figure 9d-9e). As in
499 the NO2 results, a decreasing trend in the SO2 column density was observed from January
500 to September for all selected hotspots (Figure 9g-i) in 2019 and 2020. However, the ground
501 station-based SO2 provided opposite trends in 2019 and 2020 for urban and suburban test
502 sites (Figure 9j-l). In these sites, the ground-based SO2 concentration has an increasing
503 trend in 2020, while a decreasing trend is observed in 2019, showing that the lockdown
504 restrictions were not effective in decreasing the ground-based SO2 levels as in the satellite-
505 based results.

506

29
507

508 Figure 9. Comparative graphs of the monthly TROPOMI-based NO2 (a–c) and SO2 (g-i)
509 with ground-based measurements (d–f, j-l) for three stations between January and
510 September in 2019 and 2020.

511

512 Indeed, TROPOMI-based NO2 and SO2, and station-based measurements do not present the
513 same quantity of NO2 and SO2, since the TROPOMI measures the total vertical amount of
514 atmospheric NO2 and SO2, whereas ground stations provide only near-surface distribution
515 of these pollutants. Even so, we checked if these data sets (satellite and ground
516 measurements) have any relationship in this case study. In order to validate the TROPOMI-
517 based vertical column product, it is required to have information on the vertical profile of
518 the pollutant from the ground stations or other airborne systems. Concerning the total
519 column of a pollutant, the vertical sensitivity of the retrieval is described by the total
520 column-averaging kernel (Borsdorff et al., 2014). However, when the ground measurement

30
521 is not a vertical profile, the application of the total column-averaging kernel becomes more
522 difficult (Borsdorff et al., 2018). Only few ground stations are operational for providing the
523 vertical profile of air pollutants over the world. The Total Carbon Column Observing
524 Network (TCCON) and The Infrared Working Group (IRWG), having ground-based solar
525 Fourier-transform spectrometers (FTS), are critical networks that provide regular ground-
526 based measurements of the vertically integrated column amounts of some trace gases
527 (Landgraf et al., 2018). Even so, some researchers claimed that they obtained from medium
528 to high correlation coefficients between satellite and ground-based NO2 measurements
529 (Cersosimo et al., 2020; Pinardi et al., 2018; Vîrghileanu et al., 2020). In these studies,
530 Vîrghileanu et al. (2020) had the correlation coefficients ranging from 0.597 to 0.865, and
531 the correlation coefficients from the study of Cersosimo et al. (2020) varied between 0.50
532 and 0.90. Moreover, Pinardi et al. (2018) provided very good coherence for daily and
533 monthly comparisons between TROPOMI tropospheric NO2 vertical column density and
534 ground-based NO2 data about R equal to 0.89. Besides, some other experiments also
535 revealed that the satellite-based NO2 column density had a strong correlation with ground-
536 based station data with a correlation coefficient almost above 0.8 (Bucsela et al., 2008;
537 Huijnen et al., 2009; Kramer et al., 2008). Therefore, we also compared the TROPOMI-
538 based NO2 and SO2 with ground-based measurements in three test sites (urban, urban-
539 traffic and suburban) in Istanbul, and we obtained the correlation coefficients between 0.65
540 and 0.83 for NO2, which coincide with the results of the above-mentioned studies.
541 However, the SO2 analyses did not respond well with R-value ranging from 0.004 to 0.15.

542 Considering the relationship between satellite-based and ground-based measurements, we

543 applied the hourly measurement with a time shift of about ±30 minutes from 1 January
544 2019 to 30 September 2020 in three test sites. The NO2 results in Figure 10 illustrate the
545 correlation coefficients of 0.83, 0.70, and 0.65 that were obtained from suburban (Figure
546 10a), urban (Figure 10b), and urban traffic stations (Figure 10c), respectively. On the other
547 hand, the SO2 results in Figure 10 reveal the correlation coefficients of 0.15, 0.004, and
548 0.014 that were obtained from suburban (Figure 10d), urban (Figure 10e), and urban traffic
549 stations (Figure 10f), respectively. These results showed that the more the ground-based
550 NO2 concentration was observed, the lower the correlation coefficient was obtained
551 between satellite column density and ground-based measurements. Concerning the SO2
552 comparisons, no significant relationship was found between satellite-based and ground-
553 based measurements.
31
554

555 Figure 10. Coherent of the hourly satellite-based tropospheric NO2column number density
556 derived from TROPOMI and the hourly ground-based measurements of NO2 concentrations
557 for tree sites as follows: suburban station, Şile, İstanbul (a); urban station, Başakşehir,
558 İstanbul (b); urban traffic station, Beşiktaş, İstanbul (c).

559

560 4.5) Validation of MODIS-Derived AOD and AERONET AOD

561 We have considered the AERONET AOD hourly data as reference data for validating the
562 MODIS-derived AOD 470 nm. There are five AERONET sites over turkey, while two of

32
563 them are appropriate in this study due to the time interval considered. First, because of lack
564 of enough corresponding image data over the mentioned AERONET sites, we used a
565 distance buffer around sites almost 1.5 km and 6 km on ERDEMLI site and TUZ_GÖLÜ_3
566 site, respectively (henceforth mentioned with ERDEMLI* and TUZ_GÖLÜ_3*). Due to
567 the absence of AERONET data in Tuz_Gölü_3* site, only July-August 2018 data sets were
568 used for this site, whereas, for the ERDEMLI* site, the data were available from July 2018
569 to March 2019. The results of the validation analysis between MODIS-derived AOD and
570 AERONET-based AOD are illustrated in Figure 11. The findings show that MODIS-
571 derived AOD is well correlated with the AERONET AOD in both ERDEMLI* and
572 TUZ_GÖLÜ_3* sites with the R-value equal to 0.86 and 0.82, and RMSE equal to 0.065
573 and 0.097, respectively. According to some studies, these correlation coefficient values are
574 much similar with other resulted thought varies methods (Aldabash et al., 2020; Zhang et
575 al., 2020b). It is also worth noting that estimated surface reflectance’s errors and AOD
576 conversion algorithms in aerosol loading can be influence over obtained results (Grey et al.,
577 2006; Gupta et al., 2016).

578

579

580 Figure 11. Scatterplots between MODIS-derived AOD 470nm and AERONET AOD
581 470nm.

33
582

583 4.6) Relationships Between Monthly Average Satellite-based Air Pollutants and
584 Climatological/Meteorological Parameters

585 It is worth emphasizing that the air quality does not depend only on the concentration of the
586 pollutants, but also on the local and regional climatological and meteorological conditions
587 (Ji et al., 2020). Researchers have investigated the spatial and temporal relationship
588 between air pollutants and key meteorological factors (Athanassiadou et al., 2010; Banerjee
589 et al., 2011). In addition, some studies have shown that meteorological conditions
590 controlled the spatial and temporal impacts on air quality (Ji et al., 2020; Qiao et al., 2019).
591 Hence, these factors and parameters should be thoroughly investigated to find out the
592 impact of human and industrial activities on the pollutant concentration in the atmosphere.
593 Therefore, we examined the correlation between each pollutant and key
594 meteorological/climatological parameters for the lockdown period.
595 Meteorological/climatological parameters include the monthly average of air temperature
596 (Temp.) from the ground to 2 m height in Kelvin, the monthly sum of total precipitation
597 (Prec.) in meters, the monthly average of Surface pressure (Pres.) in pascal, and the
598 monthly average of wind speed (WS) in 10 meters of ground in m/s. Table 5 provides the
599 monthly average values of all parameters and the correlation coefficients between them.
600 Table 5 shows significant correlations between all pollutants and air temperature in 2019
601 and 2020 ranging from 0.68 to 0.98. While the air temperature revealed strong negative
602 correlations with NO2 and SO2, it had a high positive correlation with AOD for the
603 lockdown period. Concerning the surface pressure, the correlations with all pollutants
604 showed identical behaviors as in air temperature providing negative correlations with NO2
605 and SO2, whereas positive correlation with AOD. Besides, the absolute value of these
606 correlations varied from 0.55 to 0.99 in 2019 and 2020. The precipitation and air pollutants
607 did not have high correlations presenting almost zero correlation with AOD in 2019 and
608 2020, while moderate positive correlations with NO2 and SO2 in 2020. Considering the
609 wind speed and air pollutant relations, low correlation coefficients were observed in 2019
610 for all variables; however, moderate negative correlation and high positive correlation was
611 obtained from NO2 (-0.56)- SO2 (-0.65) and AOD (0.92), respectively, in 2020. Hence,
612 these results are indicating the main contribution of meteorological parameters in
613 decreasing air pollution. According to Ghahremanloo et al. (2020), while there was a

34
614 relatively high correlation between daily NO2 concentration and daily meteorological
615 parameters, there was a relatively high correlation over four region of China in 2019, but
616 the correlations have reduced in 2020 during the quarantine. However, these results indicate
617 the lockdown period times were the major reasons for decreasing in NO2 density.

618

619 Table 5. Monthly average values for air temperature, surface pressure, total precipitation,
620 wind speed, and air pollutants during the lockdown period over Turkey. The abbreviations
621 Cor., Temp., Pres., Prec., and WS refer to correlation, air temperature (K), surface pressure
622 (hPa), total precipitation (m) and wind speed (m/s), respectively.

2019 2020
Variables
Mar. Apr. May Mar. Apr. May

NO2 1.46 1.14 1.05 1.22 1.01 0.92

SO2 13.83 5.38 4.44 15.93 9.39 4.90

AOD 0.10 0.19 0.18 0.12 0.13 0.18

Temp. (K) 278.77 282.40 289.92 280.43 283.22 288.72

Cor.(NO2,Temp.) -0.87 -0.92

Cor.(SO2,Temp.) -0.81 -0.96

Cor.(AOD,Temp.) 0.68 0.98

Pres. (hPa) 890.35 890.42 890.77 890.34 890.31 891.74

Cor.(NO2,Pres.) -0.77 -0.73

Cor.(SO2,Pres.) -0.70 -0.80

Cor.(AOD,Pres.) 0.55 0.99

Prec. (m) 0.07 0.09 0.05 0.09 0.07 0.08

Cor.(NO2,Prec.) 0.29 0.55

35
 Cor.(SO2,Prec.) 0.19 0.46

Cor.(AOD,Prec.) 0.01 -0.01

WS (m/s) 0.35 0.13 0.65 0.43 0.41 0.50

Cor.(NO2,WS) -0.28 -0.56

Cor.(SO2,WS) -0.18 -0.65

Cor.(AOD,WS) -0.01 0.92

623

624 4.7) Relationships Between Monthly Average Satellite-based Air Pollutants and PMI
625 Data

626 The PMI of a certain month is released on the first business day of the following month,
627 and it mainly provides the earliest information about the economic performance, which is
628 usually measured by Gross Domestic Product (GDP) (Tsuchiya, 2012). The PMI is also
629 taken into account as a good indicator of industrial production, which is one of the most
630 important indices for assessing economic conditions (Tsuchiya, 2012). Since industrial
631 activities are sometimes shown as one of the main factors causing air pollution, we
632 investigated if there is any correlation between PMI and satellite-derived pollutants. To the
633 best of our knowledge, only one study investigated NO2 pollution and PMI relationship
634 during the February 2020 COVID-19 lockdown in China (Diamond and Wood, 2020).
635 Diamond and Wood (2020) stated that the PMI and ln(NO2) showed a pronounced and
636 unprecedented decline in February 2020 followed by a rapid recovery. In Turkey, almost
637 more than 80% of economic activities are managed in Marmara and Aegean region.
638 Therefore, in this analysis, the monthly PMI values were compared with the monthly
639 average air pollutants of the Marmara and the Aegean regions. Figure 12 presents the
640 scatter plot (a) and bar charts (b-d) between monthly PMI and monthly average air
641 pollutants. The time interval of this analysis covers from January 2019 to September 2020.
642 It is clear from Figure 12a that air pollutants and PMI does not have high correlation
643 coefficients with NO2 (R=-0.08), SO2 (R=-0.13) and AOD (R=0.08) proving that industrial
644 activities do not have a high impact on air pollution on their own. Although Figure 12b and
645 12c illustrate a decreasing trend in NO2 and SO2 associated with PMI values in the

36
646 lockdown period, it is clear that this trend is similar in both 2019 and 2020, and the long-
647 term variations in the PMI do not prove any relationship with air pollutants. Unlike NO2
648 and SO2 trend, AOD has an increasing trend related to the PMI in the lockdown period, but
649 the same trend is also observed for the same period of the previous year. Besides, the long-
650 term variations in the PMI again do not prove any relationship with AOD.

651

652 Figure 12. The scatter plot (a) and bar charts (b-d) between monthly PMI and monthly
653 average air pollutants.

654

655 4.8 Discussion on the Effect of COVID-19 on Environmental Sustainability and Air
656 Quality

657 So far, many studies have emphasized the effect of the COVID-19 on the sustainable
658 development of the environment (Shulla et al., 2021; Rume and Islam, 2020; Mukherjee et
659 al., 2020; Praveena and Aris, 2021; Kumar et al., 2020). Rupani et al. (2020) published a
660 detailed review about the COVID‑19 and its natural environmental impacts, which is also
661 illustrated in Figure 13. In addition, Bherwani et al. (2021) analyzed the impact of the
662 COVID‑19 on Sustainable Development Goals (SDGs) of the United Nations in Indian
663 subcontinent with a focus on air quality.

37
664

665
666 Figure 13. COVID-19 and its environmental impacts (Rupani et al., 2020).

667

668 All these studies revealed that the measures taken to control the spread of the virus had both
669 negative and positive effects on the environment globally (Figure 14). Concerning the
670 positive effects, there were noticeable improvements in air quality, and reduction in
671 Greenhouse Gases (GHGs) emissions, water pollution and environmental noise. However,
672 the pandemic had also negative sides such as the increase in medical waste and household
673 plastics/waste, and careless and overusing of masks, gloves, and disinfectants.

38
674
675 Figure 14. Negative and positive environmental effects of COVID-19 pandemic (Rume and
676 Islam, 2020).

677

678 These environmental effects of the COVID-19 are assumed not to last long time. Therefore,
679 this pandemic provides an opportunity to develop a proper strategy for a better and long-
680 term sustainable environmental management, and to determine sustainable solutions for
681 further improvement in the air quality after the pandemic. Within this context, Figure 15,
682 for which detailed information is provided by Rume and Islam (2020), illustrates some
683 possible strategies for global environmental sustainability.

39
684
685 Figure 15. Some possible strategies for global environmental sustainability (Rume and
686 Islam, 2020).

687

688

689 5) CONCLUSION

690 In this study, the spatial-temporal patterns of Sentinel-5P TROPOMI-derived air pollutants,
691 namely, NO2 and SO2, and MODIS-derived AOD are monitored and investigated between
692 January 2019 and September 2020 including first wave lockdown period of COVID-19
693 outbreak over Turkey. The GEE platform is used for the retrieval and processing of satellite
694 imagery, and data analysis.

695 According to the obtained results, NO2 characteristically reach minimum values in the
696 spring, then starting to increase in the summer and reaches maximum in the wintertime.
697 Unlike NO2, SO2 becomes minimum in the whole spring and summer seasons. On the
698 contrary, AOD is higher between the months of April and October than the others are. The
699 NO2 results from Sentinel-5P are highly correlated with the results from the ground
700 stations. Besides, MODIS-derived AOD is very consistent with AERONET results. In
701 addition, all the results have significant correlations with some meteorological parameters,
702 especially pressure and temperature. Interestingly, the correlations with all parameters are
703 higher in 2020.

40
704 The results and analyses show that temporal variations are more representative in NO2 for
705 the lockdown period of COVID19 outbreak. As seen from Fig. 3 and Fig. 4, NO2
706 concentrations in the lockdown period of 2020 are significantly smaller than in the same
707 period of 2019. After the lockdown, the concentrations leap higher levels than 2019,
708 especially starting from July. The NO2 concentrations are mainly affected by transportation,
709 industrial and agricultural activities, heating in the wintertime, air conditioning need in the
710 summertime. PMI analyses show that there are weak correlation between industrial
711 activities and NO2 concentration. The lockdown period covers the springtime in which
712 heating, air conditioning agricultural needs are minimal. Therefore, it is considered that
713 decreasing in transportation and industrial activities play the key role in decreasing NO2
714 concentration during the lockdown period. The increase of the concentrations after the lock
715 down coincides with the summer time. As seen from Fig 3, spatial distribution of the NO2
716 concentrations remarkably grows around the southern regions rather than the northern
717 regions. They are the main touristic and agriculture regions in Turkey. In summer time,
718 human population in touristic regions dramatically increases, so transportation and air-
719 conditioning need also dramatically increase in these regions, especially in the
720 southwestern part. For agriculture regions, irrigation, fertilization and stubble burning
721 happen at the highest level in the summer time. As for the wintertime, it is understood the
722 combustion increases due to house heating, which is the main parameter affecting the NO2
723 concentration.

724 Unlike the decreasing trend of NO2 in the lockdown period, the same significant decrease
725 are not observed in SO2 vertical column density and AOD during the whole lockdown
726 period. Although, the AOD distribution significantly decreased only in April 2020, which
727 the tightest restrictions were applied due to the COVID-19 outbreak, same period of 2019.
728 These results reveal the lockdown had no effect on reducing effect on the vertical column
729 density of SO2. As in NO2 analyses, the relationship between SO2 and AOD, and
730 meteorological/climatological parameters presents that air temperature and surface pressure
731 have strong correlations with air pollutants, whereas precipitation and wind speed are not
732 effective as well as air temperature and surface pressure. PMI correlation results also
733 present almost no relationship with SO2 and AOD showing economic/industrial activities
734 do not have a high impact on the behavior of the air pollutants on their own. compared to
735 the April 2019, the spatial distribution of SO2 in the 2020 lockdown period had slightly
736 higher concentrations.
41
737 This study is a first attempt to monitor and analyze the effects of the whole COVID-19
738 lockdown period as well as on air pollution using satellite-derived air pollutants and their
739 relationships with various parameters. The analyses in this study show that air pollution
740 variation due to human activity can be monitored most effectively through NO2
741 concentration using satellite imageries. In this way, GEE provides a powerful and very
742 effective platform for researchers to analyze these parameters. Processing Sentinel-5P data
743 in this platform, it is found out that NO2 concentration significantly decreased during the
744 first wave lockdown period of COVID-19 outbreak. The further analyses of the results with
745 other parameters, shows that the transportation and industrial activities played the key role
746 in decreasing the NO2 level. As for the summer time after the lockdown, agricultural and
747 touristic activities caused to higher NO2 levels. Thus, as a future work, the authors plan to
748 investigate the relationship between air pollutants and with various parameters such as
749 traffic density, touristic and agricultural statistics as well as with fatality cases of COVID-
750 19 outbreak on the city level in Turkey.

751

752 Acknowledgements

753 The authors would like to thank the GEE team and platform, and their collaborators
754 (NASA, ESA, USGS, NOAA and so on) for providing the satellite imagery free of charge.
755 The authors also thank NASA and Turkish Ministery of Environment and Urban Planning
756 for enabling free access to AERONET and Air Quality Monitoring Center, respectively.
757 We also would like to thank the anonymous reviewers for their useful and constructive
758 comments.

759 References

760 AbdelSattar A. Monitoring Air Pollution Using Satellite Data. Proceedings of the
761 International Conference on Industrial Engineering and Operations Management
762 Riyadh, Saudi Arabia, November 26-28, 2019, pp. 772-780.
763 Aksoy S, Gorucu O, Sertel E. Drought Monitoring using MODIS derived indices and
764 Google Earth Engine Platform. 2019 8th International Conference on Agro-
765 Geoinformatics (Agro-Geoinformatics). IEEE, 2019, pp. 1-6.
766 Aldabash M, Bektas Balcik F, Glantz P. Validation of MODIS C6. 1 and MERRA-2 AOD
767 Using AERONET Observations: A Comparative Study over Turkey. Atmosphere
768 2020; 11: 905.
769 Arya SP. Air pollution meteorology and dispersion. Vol 6: Oxford University Press New
770 York, 1999.

42
771 Athanassiadou M, Baker J, Carruthers D, Collins W, Girnary S, Hassell D, et al. An
772 assessment of the impact of climate change on air quality at two UK sites.
773 Atmospheric Environment 2010; 44: 1877-1886.
774 Banerjee T, Singh S, Srivastava R. Development and performance evaluation of statistical
775 models correlating air pollutants and meteorological variables at Pantnagar, India.
776 Atmospheric Research 2011; 99: 505-517.
777 Bherwani, H., Gautam, S., & Gupta, A. (2021). Qualitative and quantitative analyses of
778 impact of COVID-19 on sustainable development goals (SDGs) in Indian
779 subcontinent with a focus on air quality. International Journal of Environmental
780 Science and Technology, 18(4), 1019-1028.
781 Borsdorff T, Hasekamp O, Wassmann A, Landgraf J. Insights into Tikhonov regularization:
782 application to trace gas column retrieval and the efficient calculation of total column
783 averaging kernels. Atmospheric Measurement Techniques 2014; 7: 523.
784 Borsdorff T, Hu H, Hasekamp O, Sussmann R, Rettinger M, Hase F, et al. Mapping carbon
785 monoxide pollution from space down to city scales with daily global coverage.
786 Atmospheric Measurement Techniques 2018; 11: 5507-5518.
787 Borsdorff T, Pandey S, Hasekamp O, Aben I, Houweling S, Landgraf J. Carbon monoxide
788 air pollution on sub-city scales and along arterial roads detected by the
789 Tropospheric Monitoring Instrument. Atmospheric Chemistry & Physics 2019; 19.
790 Broomandi P, Karaca F, Nikfal A, Jahanbakhshi A, Tamjidi M, Kim JR. Impact of COVID-
791 19 event on the air quality in Iran. Aerosol and Air Quality Research 2020; 20:
792 1793-1804.
793 Bucsela E, Perring A, Cohen R, Boersma K, Celarier E, Gleason J, et al. Comparison of
794 tropospheric NO2 from in situ aircraft measurements with near‐real‐time and
795 standard product data from OMI. Journal of Geophysical Research: Atmospheres
796 2008; 113.
797 Butz A, Galli A, Hasekamp O, Landgraf J, Tol P, Aben I. TROPOMI aboard Sentinel-5
798 Precursor: Prospective performance of CH4 retrievals for aerosol and cirrus loaded
799 atmospheres. Remote Sensing of Environment 2012; 120: 267-276.
800 Cersosimo A, Serio C, Masiello G. TROPOMI NO2 Tropospheric Column Data: Regridding
801 to 1 km Grid-Resolution and Assessment of their Consistency with In Situ Surface
802 Observations. Remote Sensing 2020; 12: 2212.
803 Clemente J, Fontanelli G, Ovando G, Roa Y, Lapini A, Santi E. Google Earth Engine:
804 Application Of Algorithms For Remote Sensing Of Crops In Tuscany (Italy). 2020
805 IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LAGIRS).
806 IEEE, 2020, pp. 195-200.
807 DeVries B, Huang C, Armston J, Huang W, Jones JW, Lang MW. Rapid and robust
808 monitoring of flood events using Sentinel-1 and Landsat data on the Google Earth
809 Engine. Remote Sensing of Environment 2020; 240: 111664.
810 Diamond MS, Wood R. Limited regional aerosol and cloud microphysical changes despite
811 unprecedented decline in nitrogen oxide pollution during the February 2020
812 COVID‐19 shutdown in China. Geophysical research letters 2020; 47:
813 e2020GL088913.
814 Eckhoff RK. Understanding dust explosions. The role of powder science and technology.
815 Journal of Loss Prevention in the Process Industries 2009; 22: 105-116.
816 El Khoury E, Ibrahim E, Ghanimeh S. A look at the Relationship Between Tropospheric
817 Nitrogen Dioxide and Aerosol Optical Thickness Over Lebanon using Spaceborne
818 data of the Copernicus Programme. 2019 Fourth International Conference on
819 Advances in Computational Tools for Engineering Applications (ACTEA). IEEE,
820 2019, pp. 1-6.
821 Elshorbany YF, Kapper HC, Ziemke JR, Parr SA. The Status of Air Quality in the United
822 States during the COVID-19 Pandemic: A Remote Sensing Perspective. Remote
823 Sensing 2021; 13: 369.

43
824 Eren O. Forecasting the Relative Direction of Economic Growth by Using the Purchasing
825 ManagersIndex. Iktisat Isletme ve Finans 2014; 29: 55-72.
826 Fan C, Li Y, Guang J, Li Z, Elnashar A, Allam M, et al. The impact of the control measures
827 during the COVID-19 outbreak on air pollution in China. Remote Sensing 2020; 12:
828 1613.
829 Feng Y, Chen D, Zhang X. Atmospheric aerosol pollution across China: a spatiotemporal
830 analysis of satellite-based aerosol optical depth during 2000–2016. International
831 Journal of Digital Earth 2019; 12: 843-857.
832 Filonchyk M, Yan H. Urban Air Pollution Monitoring by Ground-Based Stations and
833 Satellite Data,©: Springer, 2019.
834 Fioletov V, McLinden CA, Griffin D, Theys N, Loyola DG, Hedelt P, et al. Anthropogenic
835 and volcanic point source SO₂ emissions derived from TROPOMI on board
836 Sentinel-5 Precursor: first results. 2020.
837 Ghahremanloo M, Lops Y, Choi Y, Mousavinezhad S. Impact of the COVID-19 outbreak
838 on air pollution levels in East Asia. Science of The Total Environment 2020:
839 142226.
840 Giles DM, Sinyuk A, Sorokin MG, Schafer JS, Smirnov A, Slutsker I, et al. Advancements
841 in the Aerosol Robotic Network (AERONET) Version 3 database–automated near-
842 real-time quality control algorithm with improved cloud screening for Sun
843 photometer aerosol optical depth (AOD) measurements. Atmospheric
844 Measurement Techniques 2019; 12.
845 Grey WM, North PR, Los SO, Mitchell RM. Aerosol optical depth and land surface
846 reflectance from multiangle AATSR measurements: global validation and
847 intersensor comparisons. IEEE Transactions on Geoscience and Remote Sensing
848 2006; 44: 2184-2197.
849 Gupta P, Levy RC, Mattoo S, Remer LA, Munchak LA. A surface reflectance scheme for
850 retrieving aerosol optical depth over urban surfaces in MODIS Dark Target retrieval
851 algorithm. Atmospheric Measurement Techniques 2016; 9: 3293-3308.
852 Hajiloo F, Hamzeh S, Gheysari M. Impact assessment of meteorological and
853 environmental parameters on PM 2.5 concentrations using remote sensing data
854 and GWR analysis (case study of Tehran). Environmental Science and Pollution
855 Research 2019; 26: 24331-24345.
856 He W, Wang Y, Zuo J, Luo Y. Sectoral linkage analysis of three main air pollutants in
857 China's industry: Comparing 2010 with 2002. Journal of Environmental
858 Management 2017; 202: 232-241.
859 Huang H, Chen Y, Clinton N, Wang J, Wang X, Liu C, et al. Mapping major land cover
860 dynamics in Beijing using all Landsat images in Google Earth Engine. Remote
861 Sensing of Environment 2017; 202: 166-176.
862 Huijnen V, Eskes H, Amstrup B, Bergstrom R, Boersma K, Elbern H, et al. Comparison of
863 OMI NO2 tropospheric columns with an ensemble of global and European regional
864 air quality models. Atmos. Chem. Phys. Discuss 2009; 9: 22271-22330.
865 Ialongo I, Virta H, Eskes H, Hovila J, Douros J. Comparison of TROPOMI/Sentinel-5
866 Precursor NO2 observations with ground-based measurements in Helsinki.
867 Atmospheric Measurement Techniques 2020; 13.
868 Ji M, Jiang Y, Han X, Liu L, Xu X, Qiao Z, et al. Spatiotemporal Relationships between Air
869 Quality and Multiple Meteorological Parameters in 221 Chinese Cities. Complexity
870 2020; 2020.
871 Kanniah KD, Zaman NAFK, Kaskaoutis DG, Latif MT. COVID-19's impact on the
872 atmospheric environment in the Southeast Asia region. Science of The Total
873 Environment 2020: 139658.
874 Karuppasamy MB, Seshachalam S, Natesan U, Ayyamperumal R, Karuppannan S,
875 Gopalakrishnan G, et al. Air pollution improvement and mortality rate during
876 COVID-19 pandemic in India: global intersectional study. Air Quality, Atmosphere &
877 Health 2020; 13: 1375-1384.
44
878 Kramer LJ, Leigh RJ, Remedios JJ, Monks PS. Comparison of OMI and ground‐based in
879 situ and MAX‐DOAS measurements of tropospheric nitrogen dioxide in an urban
880 area. Journal of Geophysical Research: Atmospheres 2008; 113.
881 Kumar, A., Malla, M. A., & Dubey, A. (2020). With Corona outbreak: Nature started hitting
882 the reset button globally. Frontiers in public health, 8: 569353.
883 Laat Ad, Vazquez-Navarro M, Theys N, Stammes P. Analysis of properties of the 19
884 February 2018 volcanic eruption of Mount Sinabung in S5P/TROPOMI and
885 Himawari-8 satellite data. Natural Hazards and Earth System Sciences 2020; 20:
886 1203-1217.
887 Landgraf J, aan de Brugh J, Borsdorff T, Houweling S, Hasekamp O. Algorithm Theoretical
888 Baseline Document for Sentinel-5 Precursor: Carbon Monoxide Total Column
889 Retrieval. ESA, 2018.
890 Leinonen U, Koskinen J, Makandi H, Mauya E, Käyhkö N. OPEN FORIS AND GOOGLE
891 EARTH ENGINE LINKING EXPERT PARTICIPATIONWITH NATURAL
892 RESOURCE MAPPING AND REMOTE SENSING TRAINING IN TANZANIA.
893 International Archives of the Photogrammetry, Remote Sensing & Spatial
894 Information Sciences 2018; 42.
895 Lin C-A, Chen Y-C, Liu C-Y, Chen W-T, Seinfeld JH, Chou CC-K. Satellite-derived
896 correlation of SO2, NO2, and aerosol optical depth with meteorological conditions
897 over East Asia from 2005 to 2015. Remote Sensing 2019; 11: 1738.
898 Lorente A, Boersma K, Eskes H, Veefkind J, Van Geffen J, de Zeeuw M, et al.
899 Quantification of nitrogen oxides emissions from build-up of pollution over Paris
900 with TROPOMI. Scientific reports 2019; 9: 1-10.
901 Lorente A, Boersma KF, Yu H, Dörner S, Hilboll A, Richter A, et al. Structural uncertainty in
902 air mass factor calculation for NO2 and HCHO satellite retrievals. Atmospheric
903 Measurement Techniques 2017; 10: 759-782.
904 Martin RV. Satellite remote sensing of surface air quality. Atmospheric environment 2008;
905 42: 7823-7843.
906 Mesas-Carrascosa F-J, Pérez Porras F, Triviño-Tarradas P, García-Ferrer A, Meroño-
907 Larriva JE. Effect of Lockdown Measures on Atmospheric Nitrogen Dioxide during
908 SARS-CoV-2 in Spain. Remote Sensing 2020; 12: 2210.
909 Metya A, Dagupta P, Halder S, Chakraborty S, Tiwari YK. COVID-19 Lockdowns Improve
910 Air Quality in the South-East Asian Regions, as Seen by the Remote Sensing
911 Satellites. Aerosol and Air Quality Research 2020; 20.
912 Mukherjee A, Babu SS, Ghosh S. Thinking about water and air to attain Sustainable
913 Development Goals during times of COVID-19 Pandemic. Journal of Earth System
914 Science 2020; 129: 1-8.
915 Nakada LYK, Urban RC. COVID-19 pandemic: Impacts on the air quality during the partial
916 lockdown in São Paulo state, Brazil. Science of the Total Environment 2020; 730:
917 139087.
918 Nichol JE, Bilal M, Ali M, Qiu Z. Air Pollution Scenario over China during COVID-19.
919 Remote Sensing 2020; 12: 2100.
920 O’Brien DM, Polonsky IN, Utembe S, Rayner P. Remote sensing CO2, CH4 and CO
921 emissions in a polluted urban environment. 2016.
922 Ogen Y. Assessing nitrogen dioxide (NO2) levels as a contributing factor to the
923 coronavirus (COVID-19) fatality rate. Science of The Total Environment 2020:
924 138605.
925 Omrani H, Omrani B, Parmentier B, Helbich M. Spatio-temporal data on the air pollutant
926 nitrogen dioxide derived from Sentinel satellite for France. Data in brief 2020; 28:
927 105089.
928 Otmani A, Benchrif A, Tahri M, Bounakhla M, El Bouch M, Krombi Mh. Impact of Covid-19
929 lockdown on PM10, SO2 and NO2 concentrations in Salé City (Morocco). Science
930 of the total environment 2020; 735: 139541.

45
931 Perera FP. Multiple threats to child health from fossil fuel combustion: impacts of air
932 pollution and climate change. Environmental health perspectives 2017; 125: 141-
933 148.
934 Pinardi G, Vigouroux C, Langerock B, De Mazière M, Granville J, Lambert J-C, et al.
935 Sentinel-5 Precursor NO2 and HCHO validation using NDACC and complementary
936 UV-Vis DOAS systems. EGUGA 2018: 15991.
937 Praveena SM, Aris AZ. The impacts of COVID-19 on the environmental sustainability: a
938 perspective from the Southeast Asian region. Environmental Science and Pollution
939 Research 2021; 1-8.
940 Qiao Z, Wu F, Xu X, Yang J, Liu L. Mechanism of spatiotemporal air quality response to
941 meteorological parameters: a national-scale analysis in China. Sustainability 2019;
942 11: 3957.
943 Rojas JP, Urdanivia FR, Garay RA, García AJ, Enciso C, Medina EA, et al. Effects of
944 COVID-19 pandemic control measures on air pollution in Lima metropolitan area,
945 Peru in South America. Air Quality, Atmosphere & Health 2021: 1-9.
946 Rume T, Islam SDU. Environmental effects of COVID-19 pandemic and potential
947 strategies of sustainability. Heliyon 2020; 6: e04965.
948 Ruoying C, Jun Z, Wen L, Jingjing J. Temporal and Spatial Distribution of SO2 in the
949 Process of Haze in North China Based on Remote Sensing Data. International
950 Journal of Environmental Monitoring and Analysis 2019; 7: 27.
951 Rupani, P. F., Nilashi, M., Abumalloh, R. A., Asadi, S., Samad, S., & Wang, S. (2020).
952 Coronavirus pandemic (COVID-19) and its natural environmental impacts.
953 International Journal of Environmental Science and Technology, 1-12.
954 Salmabadi H, Saeedi M. Monitoring of SO2 column concentration over Iran using satellite-
955 based observations during 2005-2016. Pollution 2019; 5: 257-268.
956 Shulla K, Voigt BF, Cibian S, Scandone G, Martinez E, Nelkovski F, Salehi P. Effects of
957 COVID-19 on the Sustainable Development Goals (SDGs). Discover Sustainability
958 2021; 2: 15.
959 Smirnov A, Holben B, Eck T, Dubovik O, Slutsker I. Cloud-screening and quality control
960 algorithms for the AERONET database. Remote sensing of environment 2000; 73:
961 337-349.
962 Tobías A, Carnerero C, Reche C, Massagué J, Via M, Minguillón MC, et al. Changes in air
963 quality during the lockdown in Barcelona (Spain) one month into the SARS-CoV-2
964 epidemic. Science of the Total Environment 2020: 138540.
965 Tsuchiya Y. Is the purchasing managers’ index useful for assessing the economy’s
966 strength? A directional analysis. Economics Bulletin 2012; 32.
967 Tzortziou M, Parker O, Lamb B, Herman JR, Lamsal L, Stauffer R, et al. Atmospheric
968 Trace Gas (NO2 and O3) variability in South Korean coastal waters, and
969 implications for remote sensing of coastal ocean color dynamics. Remote Sensing
970 2018; 10: 1587.
971 Van Geffen J, Eskes H, Boersma K, Maasakkers J, Veefkind J. TROPOMI ATBD of the
972 total and tropospheric NO2 data products (issue 1.2. 0). Royal Netherlands
973 Meteorological Institute (KNMI), De Bilt, the Netherlands, s5P-KNMI-L2-0005-RP
974 2018.
975 Veefkind J, Aben I, McMullan K, Förster H, De Vries J, Otter G, et al. TROPOMI on the
976 ESA Sentinel-5 Precursor: A GMES mission for global observations of the
977 atmospheric composition for climate, air quality and ozone layer applications.
978 Remote Sensing of Environment 2012; 120: 70-83.
979 Vîrghileanu M, Săvulescu I, Mihai B-A, Nistor C, Dobre R. Nitrogen Dioxide (NO2)
980 Pollution Monitoring with Sentinel-5P Satellite Imagery over Europe during the
981 Coronavirus Pandemic Outbreak. Remote Sensing 2020; 12: 3575.
982 Wang G, Deng X-J, Wang C-L, Zhang X-Y, Yan H-H, Chen D-H, et al. A New and Detailed
983 Assessment of the Spatiotemporal Characteristics of the SO2 Distribution in the

46
 984 Pearl River Delta Region of China and the Effect of SO2 Emission Reduction.
985 Aerosol and Air Quality Research 2019; 19: 1900-1910.
986 Wei J, Li Z, Peng Y, Sun L. MODIS Collection 6.1 aerosol optical depth products over land
987 and ocean: validation and comparison. Atmospheric Environment 2019; 201: 428-
988 440.
989 Wu X, Zhang X, Chuai X, Huang X, Wang Z. Long-Term Trends of Atmospheric CH4
990 Concentration across China from 2002 to 2016. Remote Sensing 2019; 11: 538.
991 Xia H, Zhao J, Qin Y, Yang J, Cui Y, Song H, et al. Changes in Water Surface Area during
992 1989–2017 in the Huai River Basin using Landsat Data and Google Earth Engine.
993 Remote Sensing 2019; 11: 1824.
994 Xu J, Heue K, Coldewey-Egbers M, Romahn F, Doicu A, Loyola D. Full-Physics Inverse
995 Learning Machine for satellite remote sensing of ozone profile shapes and
996 tropospheric columns. International Archives of the Photogrammetry, Remote
997 Sensing and Spatial Information Sciences 2018; 42: 3.
998 Xu R, Lin B. Why are there large regional differences in CO2 emissions? Evidence from
999 China's manufacturing industry. Journal of Cleaner Production 2017; 140: 1330-
1000 1343.
1001 Zhang G, Xu H, Wang H, Xue L, He J, Xu W, et al. Exploring the inconsistent variations in
1002 atmospheric primary and secondary pollutants during the G20 2016 Summit in
1003 Hangzhou, China: 2 implications from observation and model 3. Atmos. Chem.
1004 Phys. 1e25. https://doi. org/10.5194/acp-2019-1061 2020a.
1005 Zhang T, Cao H, Song H, Shi H. Validation of MODIS C6 Aerosol Optical Depth in China.
1006 IOP Conference Series: Materials Science and Engineering. 780. IOP Publishing,
1007 2020b, pp. 042045.
1008 Zhang Z, Arshad A, Zhang C, Hussain S, Li W. Unprecedented temporary reduction in
1009 global air pollution associated with COVID-19 forced confinement: A continental
1010 and city scale analysis. Remote Sensing 2020c; 12: 2420.
1011 Zheng Z, Yang Z, Wu Z, Marinello F. Spatial Variation of NO2 and Its Impact Factors in
1012 China: An Application of Sentinel-5P Products. Remote Sensing 2019; 11: 1939.
1013

47