Annual Variation of PM 2.5 Chemical Composition in Ho Chi Minh City, Vietnam Including the COVID-19 Outbreak Period

PM 2.5 was continuously collected in Ho Chi Minh City (HCMC), Vietnam, during the period from September 2019 to August 2020, which included the period of socioeconomic suppression caused by restrictions imposed in the face of the coronavirus disease of 2019. The concentrations of PM 2.5 mass, water-soluble ions (WSIs), organic carbon (OC), elemental carbon (EC), and water-soluble organic carbon (WSOC) were determined to evaluate the seasonal variations in PM 2.5 , the effect of socioeconomic suppression on PM 2.5 , and potential PM 2.5 sources in HCMC. The PM 2.5 mass concentration during the sampling period was 28.44 ± 11.55 µ g m – 3 (average ± standard deviation). OC, EC, and total WSIs accounted for 30.7 ± 6.6%, 9.7 ± 2.9%, and 24.9 ± 6.6% of the PM 2.5 mass, respectively. WSOC contributed 46.4 ± 10.1% to OC mass. NO 3 – , SO 42 – , and NH 4+ were the dominant species in WSIs (72.7 ± 17.7 % of the total WSIs’ mass). The concentrations of PM 2.5 mass and total WSIs during the rainy season were lower than those during the dry season, whereas the concentrations of carbonaceous species during the rainy season were higher. The concentrations of PM 2.5 mass and chemical species during the socioeconomic suppression period significantly decreased by 45% – 61% compared to the values before this period. The OC/EC ratio (3.28 ± 0.61) and char-EC/soot-EC (4.88 ± 2.72) suggested that biomass burning, coal combustion, vehicle emissions, cooking activities are major PM 2.5 sources in HCMC. Furthermore, the results of a concentration-weighted trajectory analysis suggested that the geological sources of PM 2.5 were in the local areas of HCMC and the northeast provinces of Vietnam (where coal-fired power plants are located).


INTRODUCTION
Fine particulate matter (PM2.5) is a major cause of air pollution in Southeast Asia, and high PM2.5 concentrations are present here because of several complexed sources such as industry, automobiles, motorcycles, and biomass burning (Fujii et al., 2019;Koplitz et al., 2017;Nguyen et al., 2020;Yin et al., 2019). From 1998 to 2015, Southeast Asia (especially the Indo-China Peninsula) recorded the fastest urbanization rate and an increasing PM2.5 trend compared to other areas such as Europe, Africa, and America (Yang et al., 2018). Furthermore, in Southeast Asia, the largest increases in SO2 and NOx emission from coal combustion by 2030 are expected to occur in Indonesia and Vietnam (Koplitz et al., 2017).
Several studies on PM2.5 were conducted in northern Vietnam (especially Hanoi) to evaluate their impacts on air quality and human health (Co et al., 2014;Hien et al., 2002;Luong et al., 2021;Lee et al., 2016;Popovicheva et al., 2016;Thuy et al., 2018;Tran et al., 2018). Most studies revealed the significant contribution of biomass burning (crop residue burning) to ambient PM2.5 in Hanoi. Ho Chi Minh City (HCMC), located in southern Vietnam, is the most developed and populous city in Vietnam. In addition, the climate of HCMC is different from that of Hanoi, that is, HCMC is tropical monsoon and Hanoi is humid subtropical climate, resulting in different status of atmospheric pollution by PM2.5. Few studies on PM2.5 were conducted in HCMC, and only reports based on intensive field observations (Hien et al., 2019;Huong Giang and Kim Oanh, 2014;Huy et al., 2020) and emission inventories (Nguyen et al., 2021;Nguyen et al., 2022) are available. The results of these studies suggested that transportation and biomass burning were the main PM2.5 sources in HCMC. However, no data based on long-term field observations of PM2.5 are currently available. Data analysis based on long-term field observations is crucial to fully evaluate the PM2.5 characteristics in HCMC.
Short-term and long-term exposure to PM2.5 cause adverse effects on human health, such as respiratory and cardiovascular disorders, and lung cancer mortality (WHO, 2005). PM2.5-bound chemical species, especially polycyclic aromatic hydrocarbons, have been considered as the major carcinogen to human (Ali-Taleshi et al., 2021). Luong et al. (2020) confirmed that the daily hospital admissions for acute lower respiratory infections among children in HCMC were associated with the ambient PM2.5 concentration. HCMC has a high-density population of 4,097 individuals per square kilometer, whereas Hanoi has 2,300 individuals per square kilometer. Hence, it is crucial for policymakers to identify the major PM2.5 sources to mitigate PM2.5 air pollution in HCMC.
In this study, we chemically characterized PM2.5 based on ground-based samplings in HCMC throughout one year. To the best of our knowledge, this is the first of such study focusing on HCMC. Furthermore, the effects of the COVID-19 lockdown and the suppression of socioeconomic activities in HCMC on the ambient PM2.5 mass concentrations and their chemical constituents were discussed.

Sample Collection
A field study was conducted on the rooftop (49 m above the ground level) at Vietnam National University HCMC-University of Science in Vietnam from September 2019 to August 2020. The sampling site faces one of the main streets in HCMC and is surrounded by more than 10 industrial zones dominated by machinery and equipment, textile and apparel, fabricated metal products, rubber and plastic products, chemicals and chemical products, and food processing units within a 50 km radius from the site. Heavy traffic congestion occurred during peak hours (6:30-8:30 am and 4:00-6:00 pm) within a radius of 10 km from the site.
The sampling system had two sets of PM2.5 samplers (PM2.5 IMPACT Sampler, SKC). PM2.5 samples were collected on polytetrafluoroethylene (PTFE; pore size, 2.0 µm) and quartz fiber filters (diameter, 47 mm) for 24 h at a flow rate of 10 L min -1 . After sampling, the PTFE and quartz fiber filters were stored in a refrigerator at 4°C and a freezer at -20°C, respectively, until the analyses were conducted. Field blank samples were also collected and undergone the same analysis process with samples and were subtracted from loaded samples for blank correction.
Quartz fiber filters were used to identify carbonaceous contents: organic carbon (OC), elemental carbon (EC), and water-soluble organic carbons (WSOC). A filter aliquot (~2 cm 2 ) was extracted with 15 mL of ultrapure water by ultrasonication for 20 min. The extracted solutions were filtered through a PTFE syringe filter (pore size, 0.45 µm) and analyzed by total organic carbon analyzer (TOC-L, Shimazu) to analyze WSOC, and the limit of WSOC detection were 0.01 µg m -3 . Meanwhile, OC and EC were quantified using the OC-EC Carbon Analyzer (Lab OC-EC Aerosol Analyzer, Sunset Laboratory), which employs the thermal optical reflectance method following the IMPROVE_A protocol (Chow et al., 2007). Carbon fractions were quantified at 140°C, 280°C, 480°C, and 580°C for OC1, OC2, OC3, and OC4, respectively, in helium and at 580°C, 740°C, and 840°C for EC1, EC2, and EC3, respectively, in helium:oxygen (98:2) gas. OC and EC were calculated as shown in the following equations (Chow et al., 2007): here, OP (i.e., pyrolyzed OC) is defined as the amount of carbon content measured after the introduction of oxygen until the reflectance returns to its initial value corresponding to the beginning of the analysis. Noted that the limits of detection were 0.66 µg m -3 and 0.05 µg m -3 for OC and EC, respectively. To distinguish the primary organic carbon (POC) and the secondary organic carbon (SOC) in OC, we applied the EC tracer method (Turpin and Huntzicker, 1995). We used the OC-to-EC mass ratio (OC/EC) to estimate the SOC content. POC and SOC were calculated as shown in the following equations: here, OCtot is the total OC concentration, and (OC/EC)pri is calculated by averaging three lowest OC/EC values in the data set (ChooChuay et al., 2020;Bhowmik et al., 2021).

CWT Model
To interpret the geological origins of PM2.5, we applied the weighted concentration-weighted trajectory (WCWT) model by using TrajStat software developed by Wang et al. (2009). The sampling site was set as the starting point (10°45′43.6′′N; 106°40′52.8′′E), and backward trajectories over 72 h were considered every 6 h (00:00, 06:00, 12:00, and 18:00 UTC) based on 1° × 1° Global Data Assimilation System data for a height of 500 m above ground level. A detailed description of the WCWT model and the criteria that we applied are available in the work of Hsu et al. (2003).

Data Categorization
HCMC is strongly affected by a monsoon-influenced tropical climate, resulting in two distinct seasons: the dry season (December-April) and the rainy season (May-November). Since late 2019, the COVID-19 pandemic spread worldwide, and Vietnam recorded early cases since January 2020. Subsequently, the Vietnam government declared various directives that restricted the immigration, movement, assembly, and national lockdown of the citizens from February to April 2020 (during the dry season in 2020). Therefore, we divided the data into four stages: Stage 1 represents the rainy season before the assembly restriction and lockdown period (ARL), Stage 2 represents the dry season before ARL, Stage 3 represents the dry season during ARL, and Stage 4 represents the rainy season following ARL.
Anion equivalent (AE) and cation equivalent (CE) were used to determine the acidity of PM2.5 based on the following equations (Han et al., 2010 A good correlation (r = 0.96, p < 0.01) between AE and CE during the sampling period is shown in Fig. 5(a). The ratio of AE to CE (AE/CE) in HCMC during the sampling periods was slightly higher than the unity line, indicating an acidic condition attributed to cation deficiency.
The relationships between SO4 2-, NO3 -, and NH4 + during the sampling period are shown in Fig. 5(b) and

Diagnostic ratios of chemical species
Biomass burning has been evaluated as a main source of PM2.5 in Southeast Asia (ChooChuay et al., 2020;Fujii et al., 2014;Kim Oanh et al., 2018;Tham et al., 2019;Thuy et al., 2018)  study, the high correlation of K + , which is generally known as biomass burning tracer (Andreae et al., 1998), to PM2.5, OC, and EC during 4 stages ( Table 2) indicated the presence of biomass burning source in HCMC. In this section, the potential emission sources in HCMC, including biomass burning, are discussed. OC and EC mass concentrations varied significantly over the whole sampling periods, however, OC/EC showed little fluctuation ( Fig. 2(a) and Fig. 3), suggesting the OC and EC sources did not change. Literature reviews have shown that OC/EC was in the range of 0.3-7.6 for coal combustion and 4.1-14.5 for biomass burning (Watson et al., 2001), 1.0-4.2 for vehicular exhaust (Schauer et al., 2002), 1.87-9.96 (with an average of 3.52 ± 1.41) for gasoline exhaust (ChooChuay et al., 2020), and 0.22-0.93 for diesel engine exhaust (Cui et al., 2020), and 1.4-4.6 for barbecue cooking . In this study, OC/EC ranged from 1.98 to 5.85 (3.28 ± 0.61), suggesting that OC and EC sources were derived from transportation, biomass burning, barbecue cooking, and coal combustion.
The mass ratio NO3 -/SO4 2has been used as an indicator of stationary (such as coal combustion which highly emitted SO2) and mobile sources (such as vehicle emission which mainly discharge NOx) for the particle pollution (Luong et al., 2021;Qiao et al., 2019). The NO3 -/SO4 2for stationary

WCWT analysis
To reduce the ambient PM2.5 mass concentrations in HCMC, it is crucial to identify the geological origins of PM2.5. For this purpose, we conducted a backward trajectory cluster analysis and WCWT analysis. The potential emission source areas for PM2.5 and the major chemical species (OC, SO4 2-, and NH4 + ) based on WCWT analyses are shown in Fig. 6. The color bar shows the target concentration (µg m -3 ), indicating the strong and weak source areas based on the pollutant concentrations (Ali-Taleshi et al., 2021;Jain et al., 2021).
As shown in Fig. 6, all pollutants were mainly emitted from local sources in and around HCMC and were scattered in the northeast direction. The patterns of the PM2.5 and chemical species maps were similar, indicating that they had similar sources. The highest concentration grid of PM2.5 (> 40 µg m -3 ) was in the center of HCMC, indicating the impact of local sources. OC, SO4 2-, and NH4 + maps suggest larger emission areas than PM2.5 map, indicating long-range transport of aerosols into HCMC. The results of 72 h air-mass back trajectories at 500 m above the ground level are shown in Fig. 7. During the dry season, the main trajectories (~99%) came from the ocean to the northeast of HCMC. During the rainy season, the main trajectory passed through south Thailand (34.74%), and another came from the ocean to the northeast of HCMC (29.58%). There is a coal-fired power plant region that is located roughly 300 km to the northeast of HCMC. This emision source was suggested as the long-range transport source of PM2.5 dominant species during the sampling period.

Seasonal Variations of PM2.5
The differences in the PM2.5 mass concentrations and chemical species between the dry and rainy seasons are discussed in this section. The spread of COVID-19 in HCMC since Stage 3 toppled the lives of the residents here. Hence, we focus on the seasonal variations of PM2.5 concentrations by distinguishing the data for Stages 1 and 2 (i.e., the period unaffected by the spread of COVID-19) from those for Stages 3 and 4 (i.e., the period affected by the spread of COVID-19).
The rainfall level during the rainy season was significantly higher than that during the dry season ( Fig. 1(b)), meanwhile, meteorological conditions showed stable variations during the sampling period ( Fig. 1(c)). Hence, the variation of PM2.5 mass concentration and chemical compositions in HCMC attributed to the wet deposition during the rainy season, regradless of temperature, relative humidity, sunshine hours, and wind speeds. The average concentrations of PM2.5 mass, WSIs (except SO4 2and K + ), WSOC, and SOC were higher during the dry season than during the rainy season ( Fig. 1(b), Fig. 3, and Fig. 4), indicating the effect of washout during rainy season. In contrast, the average concentrations of OC, EC, POC, char-EC, and soot-EC were higher during the rainy season than the dry season (Fig. 3). The reduction of EC, including char-EC and soot-EC, was explained by the dry deposition in the absence of precipitation during the dry seasons (Matsuda et al., 2012). As mentioned above, SOC concentration was higher during the dry seasons, accompanied with the higher SOC/OC in Stage 2 (37%) and Stage 3 (42%) than in Stage 1 (26%) and Stage 4 (31%), suggested the intensive photochemical activities during the dry seasons in HCMC.
K + concentrations were strongly correlated with OC and EC (r > 0.7, p < 0.01) during the dry season (Table 2), indicating that biomass burning made a major contribution. The PM2.5 mass and C2O4 2concentrations during the dry season were also strongly correlated (r = 0.83, p < 0.01). C2O4 2also showed higher correlations with OC, EC, SO4 2-, NO3 -, NH4 + , and K + during the dry season (Table 2). C2O4 2originates from primary emissions (biomass burning and vehicular exhaust), and secondary sources via the photooxidation of volatile organic compounds (Wang et al., 2007;Thepnuan et al., 2019). Thus, the photochemical and burning activities were considered to have decreased during the rainy season.
Despite the notable decreases of OC (~45%) and EC (~50%), OC/EC did not change significantly from Stage 2 to Stage 3 ( Fig. 2(a) and Fig. 3), indicating the similar emission sources during ARL. In addition, POC and SOC were also reduced by 52% and 39%, respectively, indicating the strong effect of ARL on the emissions of POC and precursor gases to form SOC (Fig. 2(b) and Fig. 3). The average WSOC concentration decreased by ~46% during ARL, along with the WSOC/OC in Stages 3 and 4 experienced larger fluctuations than that in Stages 1 and 2 (Fig. 2(c) and Fig. 3), indicating the unstable formation of WSOC based on the reduction in human activities. The average char-EC concentration also decreased by ~57% from Stage 2 (3.08 µg m -3 ) to Stage 3 (1.39 µg m -3 ), whereas soot-EC slightly decreased by ~15% from Stage 2 (0.45 µg m -3 ) to Stage 3 (0.38 µg m -3 ) ( Fig. 2(d) and Fig. 3). Therefore, the mean char-EC/soot-EC in Stage 2 (6.92) reduced by ~ 50% compared to Stage 3 (3.52), indicating the significant reduction of smoldering combustion in Stage 3. During the ARL, the local volume of transportation was reduced by Directive No. 16/CT-TTg (SRV, 2020), meanwhile industrial activities did not decrease. We inferred that soot-EC in HCMC was produced from industrial activities and was not affected by ARL or seasonal variation.
The average concentration of the total WSIs decreased by ~45% (Fig. 4). The significant decline in NO3 -(~57%) confirmed that the traffic volume and vehicular emission in HCMC decreased during the ARL. Wang et al. (2021) reported a similar trend of NO3related to the reduction of traffic emission, and they found that NO3decreased from 9.7 µg m -3 (before COVID-19 outbreak) to 3.9 µg m -3 (during COVID-19 outbreak). NH4 + mass concentration, which is regarded to originate from sewage and garbage (Huy et al., 2017), notably declined by 53% in Stage 3. In 2020, the municipal garbage in Vietnam and NH4 + concentrations in the Saigon River, HCMC decreased possibly because of the drop of tourists during the COVID-19 epidemic (The Vietnam National Environment Status Report 2016-2020; VEA, 2021). SO4 2concentration reduced by 42%, suggesting a decrease in coal combustion in Stage 3 (EVN, 2020a(EVN, , 2020b(EVN, , 2020cLuong et al., 2021). The concentration of the total WSIs continued to decrease in Stage 4 probably because of wet deposition during the rainy season ( Fig. 5(a)). However, the NO3concentration remained unchanged, suggesting the rebound of transportation in Stage 4.

CONCLUSION
In this study, PM2.5 were collected from September 2019 to August 2020 at a site located at a site located at the center of HCMC, Vietnam. The PM2.5 mass concentration and chemical composition, including WSIs and carbonaceous compounds (OC, EC, and WSOC), were quantified to understand the impact of seasonal variation and socioeconomic suppression and to identify possible emission sources. The main conclusions are as follows: 1. In most cases, the 24 h PM2.5 mass concentrations did not exceed the Vietnam National