The Role of Sulfur Emission from the Petroleum Industry on Ultrafine Particle Number Concentration in Singapore

Received: July 3, 2022
Revised: September 6, 2022
Accepted: December 19, 2022

 Copyright The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are cited.

Cite this article:

Hou, S., Li, W., Yang, L., Chen, G., Zhang, Y., Kuwata, M. (2023). The Role of Sulfur Emission from the Petroleum Industry on Ultrafine Particle Number Concentration in Singapore. Aerosol Air Qual. Res. 23, 220265. https://doi.org/10.4209/aaqr.220265

HIGHLIGHTS

• A year-long observation of ultrafine particle was conducted in Singapore.
• Both primary and secondary sources of ultrafine particles were identified.
• Ultrafine particle number concentration increased for SO₂-rich air masses.

Keywords: Singapore, Ultrafine particles, New particle formation, Sulfur dioxide

1 INTRODUCTION

Aerosol particles in the atmosphere affect atmospheric phenomena and the environment. For example, a subset of aerosol particles can serve as cloud condensation nuclei (CCN), affecting cloud formation processes (Kreidenweis et al., 2019). Aerosol particles deposit on the surface of the human respiratory system, affecting health (Schraufnagel, 2020). The diameter of aerosol particles (d_p) typically ranges from 1 nm to 10 µm (Kreidenweis et al., 2019). Aerosol particles are classified into coarse (2.5 µm ≤ d_p ≤ 10 µm), fine (d_p ≤ 2.5 µm), ultrafine (d_p ≤ 100 nm), and Aitken (d_p ≤ 50 nm) modes, depending on their diameter range (Ibald-Mulli et al., 2002; Kittelson et al., 2022). Coarse and fine mode particles are dominant in mass concentrations, while ultrafine particles can occupy a significant portion of the number concentration (Du et al., 2021; Kulmala et al., 2021; Wehner et al., 2004).

Ultrafine particles are important for determining the number concentration of CCN (Fan et al., 2018; Kuwata and Kondo, 2008). According to the Köhler theory, the CCN activity of aerosol particles is controlled by the diameter and chemical composition (Farmer et al., 2015). That is, larger particles that contain a higher fraction of water-soluble materials can be highly CCN-active. The number size distribution of aerosol particles is known to be more important than the chemical composition in determining the CCN number concentration (Dusek et al., 2006). The threshold diameter for CCN activation is typically larger than 30 nm for high supersaturation (Kuwata et al., 2008; Pöhlker et al., 2018).

Ultrafine particles affect human health because they can easily reach the lungs due to their small size (Habre et al., 2018; Schraufnagel, 2020). The mechanisms for the health effects of ultrafine particles include enhanced oxidation stress and effects on the immune system (Leikauf et al., 2020). Epidemiological studies have demonstrated that ultrafine particles in the urban atmosphere can induce enhanced mortality and asthma (Habre et al., 2018; Ibald-Mulli et al., 2002). The importance of the regulatory control of ultrafine particle number concentration has been discussed, although the development of consistent measurement techniques is required for this purpose (Baldauf et al., 2016).

Ultrafine particles have both primary and secondary sources (Brines et al., 2015). The major primary emission sources include combustion and traffic (Kittelson et al., 2022). New particle formation is an important secondary formation process for ultrafine particles (Nieminen et al., 2018). Both primary and secondary sources are important in urban areas (Presto et al., 2021).

New particle formation is induced by organic and inorganic precursors (Gordon et al., 2017; Mohr et al., 2017). In urban areas, SO₂ is the most important precursor for forming new particles following oxidation (Dunn et al., 2004; Zhang et al., 2004). The formation of sulfuric acid through the oxidation of SO₂ dominantly occurs by chemical reactions with OH radicals, which means that the phenomenon is typically observed during daytime (Lu et al., 2019).

Singapore is a highly urbanized city-state located in a tropical area. Traffic emissions are one of the important sources of aerosol particles in Singapore (Zhang et al., 2017). In addition, the emission of SO₂ is intense in Singapore due to the existence of petroleum industries, such as oil refineries (Li and Tartarini, 2020). The combination of high insolation in a tropical environment and SO₂ emission is favorable for the production of ultrafine particles (Brines et al., 2015). In fact, a previous study that conducted number size distribution measurements in Singapore for seven weeks recorded a frequent occurrence of nucleation events (Betha et al., 2013).

This study aims to determine the sources and controlling factors of the ultrafine particle number concentration in Singapore. We conducted a year-long continuous observation of the number size distribution in Singapore in 2018 and 2019, especially focusing on ultrafine particles. In addition to the number size distribution, the concentrations of CO, CO₂, CH₄, and SO₂ were also monitored to elucidate the sources of ultrafine particles.

 
2 METHOD

 
2.1 Observation Site

The observation was conducted on the campus of Nanyang Technological University between May 2018 and April 2019. The campus is located in the western part of the country (1°20′N, 103°40′E) (Fig. 1). The observation site was located on the rooftop of a five-story building. The distance between the central area of Singapore and the observation site was 15–20 km.

Fig. 1. Map of Singapore and its surrounding region, including Johor Bahru in Malaysia. The locations of the observation site (Nanyang Technological University) and the National University of Singapore are indicated. Jurong island and Pulau Bukom are located at south/southeast of the observation site. Petroleum industry such as oil refining is highly active at these islands.

A highway (Pan Island Expressway) is located southeast of the university campus. The distance between the highway and the observation site was approximately 600 m. Jurong Island, which is a cluster for the petrochemical industry, is located 6 km south of the campus. Oil refineries at Pulau Bukom are situated 15 km southeast of the observation site. According to the National Environmental Agency of Singapore, the SO₂ emission rate from the oil refineries in Singapore was 65 × 10³ ton year^–1 in 2018, corresponding to 91% of total emissions from the country (https://www.nea.gov.sg/our-services/pollution-control/air-pollution/air-quality, last accessed September 1, 2022). Intense SO₂ emissions from Jurong Island and Pulau Bukom were also confirmed by a previous satellite observation study (Li and Tartarini, 2020). The Johor Bahru district in Malaysia is located approximately 15 km north of the observation site.

Seasonal and diurnal variations of meteorological conditions in Singapore have been documented well in literature. Briefly, the northeast monsoon (NE) (November–February), southwest monsoon (SW) (May–August), pre-NE (September–October), and pre-SW (March–April) (Chow and Roth, 2006). The major wind directions for each season are northeast (NE), south to southeast (SE), south to southwest (Pre-NE), and northwest to northeast (Pre-SW). The mean monthly precipitation is relatively stable during February and October (150–200 mm month^–1), while rainfall during November, December, and January is higher than in other periods (250 mm month^–1) (Chow and Roth, 2006).

As a coastal city, Singapore experiences sea and land breezes (Li et al., 2013). During daytime, especially in the afternoon, wind blows from the Singapore strait (south). During nighttime, wind blows from the Malay Peninsula (north). As Singapore is located close to the equator, sunrise and sunset occur at approximately 7:00 (local time, GMT + 8) and 19:00, respectively, throughout the year. The planetary boundary layer (PBL) starts to develop at around 8:00. The PBL height is a maximum of 600–700 m at 12:00–16:00. Afterward, the PBL height starts decreasing (Li et al., 2013).

 
2.2 Measurement of Particle Number Size Distributions

Particle number size distributions were measured using the scanning mobility particle sizer (SMPS: TSI) (Wang and Flagan, 1990). The SMPS was connected to a particle sampling line equipped with a PM_2.5 cyclone (URG) and a Nafion dryer (MD-700, Permapure). Dry compressed air was provided to the Nafion tubing for drying aerosol flow.

During the period from May 2018 to March 2019, a combination of the differential mobility analyzer (DMA) (model 3081, TSI) and condensational particle counter (CPC) (model 3772, TSI) was employed. The setup measured the number size distributions of the particles for the range of 7 < d_p < 300 nm. In April 2019, the CPC was changed to model 3775 (TSI), quantifying the size distributions for the range of 14 < d_p < 700 nm. In both cases, the sheath-to-sample flow ratio of the DMA was set to 10:1. The particles were charge-neutralized using a ⁸⁵Kr neutralizer (model 3077A, TSI) before entering the DMA. The data were acquired and analyzed using Aerosol Instrument Manager (TSI) ver. 9.0. Size selection using DMA was calibrated using standard polystyrene latex particles (Thermo Scientific).

The data for the range of 10 < d_p < 190 nm were used for the analysis. This data selection ensured that multiple charge corrections were applied to the data during May 2018 and March 2019. The data from April 2019 suggested that particles in the size range covered 96% of the total number concentration (Fig. S1). The counting efficiencies of both CPCs (models 3772 and 3775) were almost equivalent to the size range (Hermann et al., 2007).

 
2.3 Measurement of Gas Species

In addition to the number size distribution, the concentrations of CO, CO₂, CH₄, and SO₂ were monitored at the same observation site (Yang et al., 2021). CO was measured using a nondispersive infrared absorption technique (APMA-370, Horiba). The concentrations of CO₂ and CH₄ were quantified using the cavity-enhanced absorption technique (ultraportable greenhouse gas analyzer, Los Gatos Research). The SO₂ concentration was monitored using the ultraviolet fluorescence technique (Serinus 50, Ecotech). A Nafion dryer was installed on the sampling line for CO, CO₂, and CH₄, reducing the interference of water vapor in infrared absorption. All gas analyzers were calibrated using standard gas cylinders (Taiyo Nippon Sanso). All data were analyzed using Igor Pro (WaveMetrics).

 
3 RESULTS

 
3.1 General Characteristics of the Data for the Entire Observation Period

Fig. 2 summarizes the concentrations of the (a) particle number (10–190 nm: N_10-190nm), (b) CO, (c) CO₂, (d) CH₄, and (e) SO₂ during the entire observation period. The corresponding data are summarized in Table 1. The background concentrations of CO, CO₂, and CH₄ were calculated using three percentile data for ± 3 weeks of a specific data point. The values of the background concentrations were 119 ± 12 ppb (CO), 412 ± 2 ppm (CO₂), and 1.90 ± 0.03 ppm (CH₄). The background concentration of CO₂ was slightly higher than the average surface atmospheric concentration for 2018 (407.38 ± 0.1 ppm) (Friedlingstein et al., 2019). The background concentration of CO₂ was higher in boreal summer than in boreal winter, while an opposite trend was observed for the background concentration of CH₄. The background concentrations were not calculated for SO₂ because they were negligible.

Fig. 2. Time series data for (a) N_10-190nm, (b) CO, (c) CO₂, (d) CH₄, and (e) SO₂. Background concentrations for CO, CO₂, and CH₄ are indicated as red solid lines.

The observation period was separated into June–July–August (JJA), September–October–November (SON), December–January–February (DJF), and March–April–May (MAM) to facilitate analyzing the seasonal variation. JJA corresponds to SW, while the data for DJF provide the characteristics of NE. The data for SON and MAM predominantly represent Pre-NE and Pre-SW, respectively.

The average value of N_10-190nm for the entire observation period was approximately 1.1 × 10⁴ cm^–3 (Table 1). The average concentration of N_10-190nm was the highest during JJA (1.4 × 10⁴ cm^–3) and the lowest during DJF (6.4 × 10³ cm^–3). In addition to the seasonal variation, shorter time fluctuations in the particle number concentrations were also observed. For example, the values of N_10-190nm occasionally exceeded 2 × 10⁴ cm^–3.

Table 1 also presents the particle number concentrations with diameter ranges of 10–20 nm (N_10-20nm), 20–30 nm (N_20-30nm), 30–50 nm (N_30-50nm), 50–100 nm (N_50-100nm), 100–190 nm (N_100-190nm), and 10–100 nm (N_10-100nm). The concentrations of particles for all size ranges were consistently low during DJF. The ultrafine particle number concentrations were the highest during JJA, and N_100-190nm was the highest during MAM.

The observed concentrations were comparable to previously reported ultrafine particle number concentrations in Singapore. See et al. (2006) measured the particle number size distribution (8 nm < d_p < 20 µm) during March 2001 and March 2002 on the campus of the National University of Singapore. The campus of the university is located approximately 10 km southwest of the observation site of the present study (Fig. 1). The average concentration during non-biomass burning periods was 2.88 × 10⁴ ± 1.42 × 10⁴ cm^–3. Betha et al. (2013) reported that the mean particle number concentrations (5.6 nm < d_p < 100 nm) during July–August 2008 and January–February 2009 were 1.5 × 10⁴ cm^–3 and 1.3 × 10⁴ cm^–3, respectively. The observation was also conducted on the campus of the National University of Singapore.

The observed particle number concentrations were also compared to data from other cities in the world. The particle number concentrations (17.5 nm < d_p < 100 nm) in Barcelona (Spain), Madrid (Spain), Brisbane (Australia), Rome (Italy), and Los Angeles (USA) were reported as 7,500 ± 7,000 cm^–3, 7,000 ± 8,000 cm^–3, 6,000 ± 7,000 cm^–3, 5,000 ± 3,000 cm^–3, and 12,000 ± 7,000 cm^–3, respectively (Brines et al., 2015). Hofman et al. (2016) conducted an atmospheric observation of number size distributions in Amsterdam (the Netherlands), Antwerp (Belgium), Leicester (UK), and London (UK) during 2013 and 2015 and obtained mean concentrations of 9,070, 13,481, 8,623, and 8,353 cm^–3, respectively.

The seasonal variations in CO and CO₂ concentrations were similar. The concentrations of CO and CO₂ were high during SON and MAM. The corresponding values for DJF were lower than those for SON and MAM and higher than those for JJA. In the case of CH₄, the concentrations during SON, DJF, and MAM were comparable. The observed CH₄ concentration was the lowest in JJA. 

The seasonal variation of SO₂ was different from that of other gas species. The concentration of SO₂ was the highest in JJA and the lowest in DJF. The concentrations during SON and MAM were intermediate.

The seasonal variations in gas species and particle number concentrations were also compared. The seasonal variations in N_10-20nm, N_20-30nm, N_30-50nm, and N_50-100nm were similar to those for SO₂ in that the maximum/minimum concentrations were observed during JJA/DJF. Conversely, the corresponding trend for N_100-190nm had some similarities with that for CO and CO₂. The concentrations of all species were high during SON and MAM. These results suggest that the major emission sources of particles depend on the size range.

 
3.2 An Example of Diurnal Variation for Number Size Distribution

Fig. 3 shows an example of the (a) particle number size distribution, (b) CO, and (c) SO₂ concentrations for a day (May 14, 2018). During 0:00 and 6:00, the mode diameter was stable at 50–60 nm. Starting at 6:00, both particle number and CO concentrations began to increase. This increase was likely due to enhanced traffic in the morning, as previously observed on the campus of the National University of Singapore (Rivellini et al., 2020). The CO concentration started decreasing at around 9:00, probably due to the enhancement of PBL height (Li et al., 2013). At the same time, the SO₂ concentration started increasing and was especially pronounced in the afternoon. During this period, the particle number concentration was also high, especially for the Aitken mode. This type of enhancement in the number concentration of the Aitken mode particles is sometimes called an “apple”-type new particle formation, although unambiguous identification is challenging (Vakkari et al., 2011; Vana et al., 2008). The SO₂ concentration started decreasing at around 18:00, while the CO concentration started increasing. The number size distribution of particles was relatively stable in the evening, with a mode diameter of 50 nm.

Fig. 3. (a) Particle number size distribution, (b) CO, and (c) SO₂ concentrations on May 14, 2018.

3.3 Diurnal Variations for Each Season

Fig. 4 shows the diurnal variations in the number size distributions for each season. Throughout all seasons, the number size distributions during 0:00–4:00 and 4:00–8:00 were similar, although there were some differences in the total number concentrations. The mode diameters for the period exhibited slight seasonal variations: 51 nm (JJA), 62 nm (SON), 69 nm (DJF), and 66 nm (MAM). Starting from 8:00, an increase in the particle number concentration was observed at around d_p ~20 nm for all seasons. The enhancement in the number concentration of this mode peaked in the afternoon (12:00–16:00). The particle number concentration was the maximum during this period. A decrease in the particle number concentration, especially for particles with a diameter of 20 nm, was observed during 16:00–20:00. The number size distributions of particles during 20:00–24:00 were similar to those during the early morning for all seasons.

Fig. 4. Diurnal variations for number size distributions during (a) JJA, (b) SON, (c) DJF, and (d) MAM. The data were averaged for 0:00–4:00 (blue solid line), 4:00–8:00 (green solid line), 8:00–12:00 (brown solid line), 12:00–16:00 (red dashed line), 16:00–20:00 (brown dashed line), and 20:00–24:00 (purple dashed line).

Fig. 5 shows the diurnal variations for N_10-20nm, N_20-30nm, N_30-50nm, N_50-100nm, N_100-190nm, ΔCO, ΔCO₂, ΔCH₄, and SO₂ for JJA, SON, DJF, and MAM. The values of ΔX (X = CO, CO₂, and CH₄) were calculated by subtracting the background concentrations (X_background) from the measured values (X_measured) (i.e., ΔX = X_measured – X_background). The values of ΔCO were the lowest during daytime and higher during nighttime. In addition, peaks associated with morning traffic were observed at around 8:00. A slight enhancement in the particle number concentration due to morning traffic was also observed. The trend was especially clear for MAM.

Fig. 5. Diurnal variations of observed parameters for (a, e, i, m, q) JJA, (b, f, j, n, r) SON, (c, g, k, o, s) DJF, and (d, h, l, p, t) MAM. The data include (a, b, c, d) number concentration of particles, (e, f, g, h) ΔCO, (i, j, k, l) ΔCO₂, (m, n, o, p) ΔCH₄, and (q, r, s, t) SO₂ concentration.
 

In the case of ΔCH₄, the diurnal variations were generally similar to those for ΔCO, except for the absence of the morning traffic peak. Diurnal variations in SO₂ concentration contrasted with that of CH₄. Namely, SO₂ concentrations were generally low during nighttime, while the corresponding value was elevated during daytime. This trend was especially clear for DJF and MAM when the dominant wind blows from the north.

The enhancements in SO₂ coincided with a relatively small increase in ΔCO₂. For example, both ΔCO₂ and SO₂ increased at 13:00 during MAM. The particle number concentration also peaked around the time period. The result suggests that SO₂ emitted from combustion sources most likely plays an important role in enhancing the ultrafine particle concentration.

 
4 DISCUSSION

 
4.1 Categorization of Air Masses Using ΔCH₄ and SO₂

Fig. 6(a) shows the relationship between ΔCH₄ and SO₂ during the entire observation period. The data are color-coded by local time. The figure exhibits an L-shaped relationship, which means that the value of ΔCH₄ is extremely low when the SO₂ concentration is high and vice versa. The SO₂-dominant air masses were predominantly observed in the afternoon to early evening. Conversely, CH₄-dominant (or SO₂-depleted) air was mainly observed after sunset or the early morning.

Fig. 6. Relationships between (a) ΔCH₄ and SO₂, and (b) N_10-190nm and SO₂.

The linear lines in Fig. 6(a) correspond to ΔCH₄/SO₂ = 0.001 (solid line), 0.01 (dotted line), and 0.1 (dash-dotted line). The data points for SO₂-dominant air masses are located around the line for ΔCH₄/SO₂ = 0.001. Most of the data points for CH₄-dominant air are found in the region of ΔCH₄/SO₂ > 0.1. The data for ΔCH₄/SO₂ = 0.01 are likely influenced by the two types of air.

The observed relationship is likely induced by the diurnal variation in wind direction. Factories for the petroleum industry in Singapore, which includes oil refineries, are located south of the observation site (Velasco and Roth, 2012). As a result, the sea breeze carries chemical species from the source to the observation site. The wind direction also explains why the SO₂ concentrations were high/low during JJA (SW) and DJF (NE) (Fig. 2). Conversely, the values of ΔCH₄ were high when the land breeze blows to the observation site from the north (Li et al., 2016). Although the exact emission source of CH₄ was difficult to identify, Fig. 6(a) suggests that ΔCH₄/SO₂ can still be used as an indicator for classifying air masses that are intensively affected by emissions from the coastal region and islands in the Singapore Strait.

The effects of sea breeze on atmospheric chemical composition have also been reported by previous studies in Singapore. For example, Rivellini et al. (2020) conducted in situ chemical composition measurements of aerosol particles using the soot particle aerosol mass spectrometer (SP-AMS) in May and June 2017 on the campus of the National University of Singapore (Fig. 1). They found that the sulfate concentration was high during daytime when the sea breeze dominates. The particles were highly acidic during the corresponding time period. These results are consistent with our present study, which found that sulfur concentration in Singapore is high during daytime.

Fig. 6(b) compares N_10-190nm to the SO₂ concentration. The data are colored by ΔCH₄/SO₂. The value of N_10-190nm was positively associated with the SO₂ concentration. For example, the SO₂ concentration needs to be pronounced (> 10 ppb) for N_10-190nm to exceed 4 × 10⁴ cm^–3. The value of ΔCH₄/SO₂ was negatively associated with SO₂ concentration. The SO₂ concentration was lower than 5 ppb for ΔCH₄/SO₂ > 0.1 (ppm ppb^–1). On the other hand, the concentration of SO₂ was high (> 10 ppb) for a smaller value of the metric (ΔCH₄/SO₂ < 0.001 (ppm ppb^–1)).

Fig. S2 shows the relationship between ΔCO and ΔCO₂. The figure is color-coded by ΔCH₄/SO₂. In general, both CO and CO₂ concentrations were high when CH₄-dominant air masses were observed, as the concentrations of all of these chemical species increased during nighttime (Fig. 5). The concentrations of both CO and CO₂ were relatively low when SO₂-dominant air masses were observed.

 
4.2 Relationships between Particle Number Concentration and ΔCO

Fig. 7 plots (a) N_10-20nm, (b) N_20-30nm, (c) N_30-50nm, (d) N_50-100nm, and (e) N_100-190nm with respect to ΔCO. The figure is color-coded by ΔCH₄/SO₂. In the case of small (d_p < 30 nm) particles, the particle number concentration exhibited an L-shaped relationship with ΔCO. The particle number concentration was low for CH₄-dominant air masses (ΔCH₄/SO₂ > 0.1), while it was high under SO₂ dominant conditions (ΔCH₄/SO₂ < 0.001).

Fig. 7. Relationships between (a) N_10-20nm, (b) N_20-30nm, (c) N_30-50nm, (d) N_50-100nm, and (e) N_100-190nm with ΔCO. The data points are color-coded by ΔCH₄/SO₂.

The particle number concentration was positively associated with ΔCO for particles in the size range of d_p > 30 nm (N_30-50nm, N_50-100nm, and N_100-190nm) for CH₄-dominant air masses (ΔCH₄/SO₂ > 0.1).

This result suggests that incomplete combustion sources, such as vehicle emissions, contribute to the particle number concentration. In the case of N_30-50nm, the lower envelope of the relationship between N_30-50nm and ΔCO was represented well by of N_30-50nm (# cm^–3) = 2–4 × ΔCO (ppb). In the case of N_50-100nm and N_100-190nm, the following relationships were obtained by the least square fitting for the CH₄-dominant data points (ΔCH₄/SO₂ > 0.1): N_50-100nm (# cm^–3) = 5.90 × ΔCO (ppb) + 1323 (r² = 0.42), and N_100-190nm (# cm^–3) = 4.14 × ΔCO (ppb) + 677 (r² = 0.52).

Enhancements in the particle number concentrations for SO₂-dominant air masses (ΔCH₄/SO₂ < 0.001) were also observed for the size range, although the magnitude decreased for larger particles. In the case of SO₂-dominant air masses, N_10-20nm was significantly pronounced. Fig. S3 shows the correlations between the particle number concentration and SO₂, confirming the important role of SO₂ as a regulator of ultrafine particles.

 
4.3 The Importance of SO₂ Source on Ultrafine Particles

Fig. 8 compares (a) N_10-20nm and (b) ultrafine particle number concentration (N_10-100nm) to ΔCH₄/SO₂. A cumulative occurrence frequency for ΔCH₄/SO₂ is shown in Fig. 8(c). The cumulative occurrence frequency values for ΔCH₄/SO₂ were 0.15 (ΔCH₄/SO₂ = 0.001), 0.54 (ΔCH₄/SO₂ = 0.01), and 0.90 (ΔCH₄/SO₂ = 0.1). This result suggests that at least half of the whole data points were influenced by intense SO₂ emission from the coastal area of Singapore (Fig. 6).

Fig. 8. Relationships between (a) N_10-20nm and (b) N_20-100nm with ΔCH₄/SO₂, along with (c) the histogram for ΔCH₄/SO₂.

The values of N_10-20nm were negatively correlated with ΔCH₄/SO₂. The concentration of CO did not affect this relationship. N_10-100nm also had a negative correlation in the range of ΔCH₄/SO₂ < 0.01, while the number concentration of ultrafine particles did not depend on ΔCH₄/SO₂ for the region of ΔCH₄/SO₂ > 0.01. For this range, ΔCO played a more important role in determining N_10-100nm. Considering that more than half of the data points belonged to the region of ΔCH₄/SO₂ < 0.01, enhancements in the Aitken mode particles induced by sulfur emission from the petroleum industry likely played the dominant role in determining the ultrafine particle number concentrations.

The present study was conducted during a time period when Singapore was not affected by biomass burning activities in surrounding countries, such as Indonesia (Ohashi et al., 2021). By conducting an atmospheric observation on the campus of the National University of Singapore in 2009, Betha et al. (2014) found that the particle number concentration during the wildfire haze period increased twice. In addition, new particle formation was suppressed during the wildfire haze periods, which was likely due to the enhanced condensation sink. Consistently, our previous study also demonstrated no indication of new particle formation during the wildfire haze period in 2015, even though sulfate was still an important chemical component of aerosol particles (Budisulistiorini et al., 2018; Chen et al., 2018). The impact of regional wildfire events on ultrafine particle number concentration will need to be evaluated by conducting long-term observations in the future.

 
5 CONCLUSIONS AND ATMOSPHERIC IMPLICATIONS

The number size distribution of aerosol particles for the size range of 10 nm ≤ d_p ≤ 190 nm was continuously measured in Singapore between May 2018 and April 2019. The concentrations of CO, CO₂, CH₄, and SO₂ were also measured simultaneously. The number concentrations of the measured particles were the highest during JJA and the lowest during DJF. JJA corresponds to the SW monsoon period, while DJF is part of the NE monsoon season.

The diurnal variation of the number concentration for ultrafine (d_p ≤ 100 nm) particles exhibited maximum values during the afternoon when sea breeze from the south develops. The enhancement was especially pronounced for the Aitken mode, coinciding with an elevated SO₂ concentration. The petroleum industry, including oil refineries, was located south of the observation site. The area was the dominant emission source of SO₂ in Singapore (Li and Tartarini, 2020).

The ratio of ΔCH₄ and SO₂ (ΔCH₄/SO₂) was empirically found to be a good indicator for classifying SO₂-dominant and depleted air masses in the dataset. In the case of SO₂-dominant air masses, the ultrafine particle number concentration was correlated with the effect of the SO₂ source. Conversely, CO concentration was an important regulator for the ultrafine number concentration for the SO₂-depleted air mass. The SO₂-dominant air masses, which originated from intense SO₂ emission sources, played a dominant role in regulating the number concentration of ultrafine particles.

Regulations to reduce SO₂ emission from coal combustion in North America during the last few decades have reduced both ultrafine particle and PM_2.5 concentrations (Presto et al., 2021). The present study suggests that the implementation of a similar idea in the petrochemical industry may also be beneficial in reducing ultrafine particle number concentrations.

 
ACKNOWLEDGMENTS

We would like to thank Drs. Jing Chen and Wen-Chien Lee for having useful discussions with us. This work was supported by the Singapore National Research Foundation (NRF) under its Singapore National Research Fellowship scheme (National Research Fellow Award, NRF2012NRF-NRFF001-031) and the National Natural Science Foundation of China (42175121 and 4215061048).