Influence Factors on Air Quality in Greater Bangkok, Thailand

7 With the outbreak of the COVID-19 pandemic around the world, many countries announced 8 lockdown measures, including Thailand. Several scientific studies have reported on 9 improvements in air quality due to the impact of these COVID-19 lockdowns. This study aims 10 to investigate the effects of the COVID-19 lockdown and its driving influence factors on air 11 pollution in Greater Bangkok, Thailand using in-situ measurements. Overall PM2.5, PM10, 12 O3, and CO concentrations presented a significant decreasing trend during the COVID-19 13 outbreak year based on three periods: the before, lockdown and after periods, for PM2.5: 0.7%, 14 15.8% and 20.7%; PM10: 4.1%, 31.7% and 6.1%; O3: 0.3%, 7.1% and 4.7%, respectively, 15 compared to the same periods in 2019. CO concentrations, especially, were increased by 16 14.7%, but decreased by 8.0% and 23.6% during the before, lockdown and after periods, 17 respectively. Meanwhile, SO2 and NO2 increased by 54.0%, 41.5% and 84.6%, and 20.1%, 18 3.2% and 26.6%, respectively, during the before, lockdown and after periods. PCA analysis 19 indicated a significant combination effect of atmospheric mechanisms that were strongly linked 20 to emission sources such as traffic and biomass burning. It has been demonstrated that the 21 COVID-19 lockdown can pause some of these anthropogenic emissions, i.e. traffic, 22 commercial and industrial activities, but not all, even low traffic emissions can’t absolutely 23 cause reductions in air pollution, since there are several primary emission sources that dominate 24 the air quality over Greater Bangkok. Finally, these findings highlight the impact of the 25 COVID-19 lockdown measures, not only on the air pollution levels, but also affects to air 26 pollution characteristics, as well. 27


29
The entire world has been battling with the Coronavirus since the first case was reported on 31 30 December 2019  Lumpur and Singapore also reported decreasing trends of NO2 (27% -30%) and of PM10, 48 PM2.5, NO2, SO2, and CO concentrations, of 26-31%, 23-32%, 63-64%, 9-20%, and 25-49 31%, respectively [9]. In addition, Nadzir, et al. [3] found that in Malaysia, CO dropped by 50 48.7%, but PM2.5 and PM10 increased up to 60% and 9.7%, respectively, as their results 51 indicated high AODs from Himawari-8, and NO2 concentrations from Aura-OMI satellite 52 sensors, associated with massive biomass burning in northern Thailand and Laos during the 53 lockdown period (March 2020) which prevented the exploration of impacts due to lockdown 54 on the air pollution in this region. Most of the research has been performed in the mega-city, 55 Stratoulias and Nuthammachot [10] analysed concentrations of air pollutants over a medium-56 sized city (Songkhla Province) in Southern Thailand and found that concentrations of PM2.

Study area 79
Greater Bangkok refers to Bangkok the capital along with the surrounding provinces, including 80 Nakhon Pathom, Nonthaburi, Pathum Thani, Samut Prakan and Samut Sakhon. Greater 81 Bangkok covers an area of 7,762 km 2 (100.20E to 100.9E, 13.0N to 14.0N) and is the center 82 of economic development and an important industrial base for the surrounding provinces ( Fig.  83 1). Some industries in Samutprakarn, Samutsakorn, and Pathumthani have already become the 84 main emission sources of atmospheric pollution from industry. 85

Ground-based air pollution monitoring, traffic index and fire spots 86
Major air pollutants and aerosols, including carbon monoxide (CO), ozone (O3), nitrogen 87 dioxide (NO2), sulfur dioxide (SO2), particulate matter with diameter lower than 10 μm (PM10) 88 and 2.5 μm (PM2.5) concentration data were collected from the Pollution Control Department 89 (PCD) of Thailand by observing 23 automatic monitoring sites [16]. The monitoring sites are 90 almost all located in the Bangkok metropolitan area, as shown in Figure 1. Data was collected 91 hourly from the period of 1 January 2019 to 20 July 2020 for both aerosols and gaseous 92 pollutants. In addition, the traffic index refers as TI [17] for the same period, as the air pollutant 93 dataset will be used to analyze the emission source from vehicles on the entire road network in 94 Bangkok with a range of 0 to 10

Meteorological dataset 104
It's not only emission sources that influence air quality, meteorological factors also 105 significantly impact the dilution and accumulation process of pollutants emitted from local 106 sources [19]. Therefore, to access the variations of air pollutants, meteorological factors must 107 be examined. In this study, the meteorological factors were achieved from the ECMWF's fifth-108 generation Reanalysis (ERA5), the European Centre for Medium-Range Weather Forecasts 109 [20]. The meteorological dataset contains total precipitation (TP), 2-meter air temperature 110 (T2M), planetary boundary layer height (BLH), relative humidity (RH), surface pressure (SP), 111 and wind speed (WS), which have a horizontal resolution of 30 km × 30 km at hourly temporal 112 resolutions. The meteorological data were picked up at the same hour of the day as the sampling 113 time for the air pollutants. 114

Data analyses methods 115
Variations of air quality regarding the COVID-19 outbreak were investigated for three different 116 periods, before-lockdown (from 1 January 2020 to 25 March 2020), lockdown (26 March 2020 117 to 31 May 2020) and after-lockdown (1 June 2020 -20 June 2020). The evaluation of impacts 118 of COVID-19 were compared with data in 2019 at the same period, which was used as a 119 baseline. The changes in the air pollutant levels were evaluated by comparing those 3 different 120 periods in year 2020 with 2019 at the same time (expressed in %) between the before, lockdown 121 and after periods. In order to access the influences between meteorological factors, the air 122 pollutants and other accompanying parameters gave different responses between the three 123 periods associated with the COVID-19 lockdown. We performed data analysis using a 124 Principal Component Analysis (PCA), which is a statistical multivariate analysis for data that 125 features a large variable set. This method enables the researcher to identify correlations and 126 patterns in a dataset by transforming them into a new smaller set of uncorrelated variables, 127 namely principal components (PCs), that still contain most of the information in the large set 128 [21]. Therefore, by applying a PCA method to air pollutant concentrations and meteorological 129 variables, a dataset could be obtained with the most significant variables, which could indicate 130 the source of the pollutants and largely explain the variations in the air pollution [22]. In this 131 study, the meteorological variables of T2M, SP, TP, RH, WS and BLH; the major air 132 pollutants: PM10, PM2.5, NO2, SO2, CO and O3 concentrations; and the anthropogenic 133 activities, TI and Fire, were taken up for analysis. The PCs created by PCA were rotated using 134 an orthogonal rotation method (varimax) to compute the explained variance matrix, the number 135 PCs were selected according to an eigenvalue greater than or equal to 1. These PCs are a linear 136 combination of the explanatory variables; therefore, Pearson correlation tests were used to 137 determine the correlation between the PCs and the original variables as a loading factor. The 138 significant variables were identified when the correlation value was greater than or equal to 139 0.3. 140 where the highest mean concentrations occurred in January during the before lockdown period 151

Results and Discussion
for Greater Bangkok, then decreasing concentrations during and after the lockdown period. 152 While SO2 concentrations show highly fluctuating time series throughout the year, especially 153 in 2020, there were higher concentration levels and more fluctuation by degree than those in 154 2019. As illustrated in Fig. 3  19 outbreak period were examined by comparing them with previous years, as shown in Table  190 1. During the before-lockdown period, TI, Fire, WS and RH all increased by 4%, 26%, 22% 191 and 14%, respectively, whereas there were decreases in BLH, TP and T2M by 42%, 99% and 192 5%, respectively. The increasing TI and Fire conditions in 2020 may cause higher NO2, SO2,193 and CO concentrations than in 2019 at the same time. During the first weeks of lockdown 194 beginning in 2020, there was a sharp decrease in the TI due to limited transportation in greater 195 Bangkok; after that, the concentrations increased gradually until the end of June (Fig. 4a). A 196 similar trend is observed for NO2 (Fig. 2c). As well, Fire (counts per day) within a 240 km. 197 radius of Bangkok city (Fig. 4b) shows a high number in the first week of the lockdown period. 198 News reports indicate there were great wildfires in northern Thailand, which produced tons of 199 aerosols and pollutants [26]. The hourly meteorological data of Greater Bangkok during the 200 study period in 2019 was compared to 2020, with results shown in Table 1

Influence factors driving the improvements in air quality 240
In order to clarify what the main influence is between expected emission sources, 241 meteorological parameters and the six pollutants during the COVID-19 outbreak will be 242 explored in this section. To obtain a better understanding and interpretation of the data, the 243 principal components (PC) were subjected to a Varimax rotation matrix. Only components with 244 an eigenvalue greater than 1 are determined as principal components (grey color), as shown in 245 Table 2. There are five major PCs in each subset period, comprising PC1, PC2, PC3, PC4 and 246 PC5. The percentage of total variance represents how much proportion of that PC largely 247 explains the variation in air quality. In each period at the same year, the percentage of total 248 variance was slightly different. However, it had some significant differences between the 249 before-lockdown and the lockdown periods. To obtain the factor loading, the Varimax rotation 250 with Kaiser Normalization (Fig. 5) was computed, a loading factor higher than 0.3 contained 251 from the output will become a principle component (PC). The results of PCA are summarized 252 in Fig. 5, presenting the significant PC contributions. A loading factor of more than 0.70 is 253 considered as strong, a range of 0.50 -0.69 is considered moderate, and a range of 0.31 -0.49 254 is considered weak. 255 256

259
During the before-lockdown period, which denotes the winter season in Thailand, the PC1 and 260 the PC2 could explain the variance by 27.4% and 18.2% for 2019, and 36.7% and 11.9% for 261 2020, respectively. The results reveal some similarities between those two years, there were 262 significant mechanisms associated with the air quality. PM10, NO2 and CO are dominant 263 pollutant parameters that associate with the particular atmospheric mechanism of low T2M, 264 BLH and WS, and high SP and RH. Hence, these atmospheric mechanisms reduced the ability 265 of the pollutants to disperse from their sources [31]. These pollutants relate to unknown 266 emissions as major and traffic-originated emissions were minor sources in 2019, while in 2020 267 the major and minor pollutants related to biomass burning and unknown emission sources, 268 respectively. These results supporting a comparison of concentrations for PM2.5, PM10 and 269 O3 in Section 3.1 between those two years are not significantly different. As mentioned before, loading factor of PM2.5 and O3. As well, in 2020 the PC2 exhibits moderate positive 277 contributions of T2M but negative contributions of O3, this is mainly affected by 278 photochemical reaction. The reaction system can produce NO2 (positive contribution) due to 279 the reaction of NO with O3 [32]. Additionally, a comparison of the PC1 also explains the 280 increase of NO2 and CO regarding higher positive contribution magnitude in 2020 than those 281 in 2019. As well, the PC5 had a higher contribution magnitude for SO2 in 2020 than in 2019, 282 resulting in increased SO2 (Section 3.1). 283 During the lockdown period, which denotes the summer season in Thailand were largely restricted to their homes, and greater Bangkok with its higher numbers of vehicles 299 should have had greater reductions in traffic emissions during the lockdown period. The 300 decreases in PM2.5, PM10 and CO concentration in 2020 strongly contributed to the increased 301 fire (PC1) and the decreased TI (PC2), suggesting that the changes in traffic emissions were 302 more responsible for the improvements air quality during the lockdown period, especially fine 303 particles, than biomass burning. On the contrary, the increases in NO2 concentrations in 2020 304 (PC2) are significantly related to biomass burning. According to during the lockdown period 305 (March 2020), there were massive forest fires in northern Thailand, which reduced the impact 306 of the lockdown on air pollution in that region. A report found an increase in some pollutants 307 during the lockdown period regarding forest fires in Malaysia [3,10]. In addition, the increased 308 SO2 concentrations were associated with unknown emission sources, which were probably 309 emitted from the industrial sector. 310 During the after-lockdown period, which denotes the rainy season in Thailand, the PC1 and the 311 PC2 could explain the variance by 25.6% and 15.8% for 2019, and 23.0% and 17.2% for 2020, 312 respectively. In 2019, there were similar contributions of air pollutants with the lockdown 313 period in 2020, as seen with the increases in O3 concentrations by the production of 314 photochemical reactions being the major mechanism (PC1), and the increased PM2.5, PM10, 315 NO2, SO2, and CO concentrations, which were minor mechanism (PC2). In 2020, PM2. the pollution from these sources can transport to Greater Bangkok, resulting in decreasing 356 magnitudes of each pollutant being lower than other countries. Furthermore, the results show 357 that after the lockdown was relieved, all pollutants except O3 tended to increase, even though 358 Greater Bangkok's people still kept to decreased mobility and social activity. This implies that 359 the COVID-19 lockdown was able to pause some anthropogenic emissions i.e. traffic, 360 commercial and industrial activities, but not all, even low traffic emissions could not absolutely 361 cause a reduction in air pollution, since several primary emission sources dominate the air 362 quality over Greater Bangkok. In addition, social distancing guideline recommend that people 363 stay at home, which causes consumption of higher electricity, resulting in electric power plant 364 increasing their production capacity and emitting more air pollution. Finally, the results 365 demonstrate that the COVID-19 lockdown measures influenced not only the air pollution 366 levels, but also affected air pollution characteristics. 367

368
All the data used in this study are available from the corresponding author upon request. 369