Influence of Meteorological Conditions and Fire Hotspots on PM0.1 in Northern Thailand during Strong Haze Episodes and Carbonaceous Aerosol Characterization

Received: March 27, 2021
Revised: July 13, 2021
Accepted: August 23, 2021

 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:

Sresawasd, C., Chetiyanukornkul, T., Suriyawong, P., Tekasakul, S., Furuuchi, M., Hata, M., Malinee, R., Tekasakul, P., Dejchanchaiwong, R. (2021). Influence of Meteorological Conditions and Fire Hotspots on PM_0.1 in Northern Thailand during Strong Haze Episodes and Carbonaceous Aerosol Characterization. Aerosol Air Qual. Res. 21, 210069. https://doi.org/10.4209/aaqr.210069

HIGHLIGHTS

• Understanding of physicochemical characteristicsof PM_0.1 in upper SEA.
• PM_0.1 concentration during haze periods was 3 times higher than non-haze periods.
• OC and EC levels during haze period were 4 times higher than non-haze periods.
• Lower OC and EC correlation for haze period suggested complex pattern of PM source.
• Good OC/EC and hotspots correlation indicated more contribution of biomass burning.

Keywords: Ultrafine particles, Carbonaceous aerosols, Forest fire, Source identification, ASEAN haze

1 INTRODUCTION

The current ambient air quality standard of particulate matters focuses on PM_2.5 mass concentrations. However, ultrafine particles or nanoparticles (PM_0.1) has recently become of more concern, because they are more carcinogenic and can cause greater health effect than larger particles due to larger surface area (Sharma and Balasubramanian, 2018; Ajdary et al., 2018; Schraufnagel, 2020). Besides traveling deep into the lung mainly in the alveolar region (Schraufnagel, 2020), PM_0.1 can translocate from the cardiopulmonary system to the central nervous system by crossing the blood-brain barrier, and potentially cause brain damage (Cheng et al., 2016; Jew et al., 2019; Morris-Schaffer et al., 2019; Hahad et al., 2020; Potter et al., 2021). A recent study in Thailand showed that PM_0.1 represented more than 5.7–14.8% of total suspended particulates (TSP) in atmospheric air (Chomanee et al., 2020).

Southeast Asia (SEA) is strongly affected by atmospheric haze pollution (Wiriya et al., 2013; Li and Shao, 2009; Pongpiachan et al., 2014a; Pongpiachan et al., 2015; Chuang et al., 2016; Li et al., 2016; Chomanee et al., 2020; Dejchanchaiwong et al., 2020; ChooChuay et al., 2020a, b, c; Dejchanchaiwong and Tekasakul, 2021). In the upper SEA, the haze is observed during the first quarter of every year and studied, mostly in Thailand (Wiriya et al., 2013, 2016; Thepnuan et al., 2019; Pani et al., 2019), but recently, air quality affected by PM₁₀ and PM_2.5 was studied in Laos (Nguyen et al., 2019), Myanmar (Sricharoenvech et al., 2020) and Vietnam (Le, 2020). Previous investigations showed that aerosol particles in northern Thailand originated from biomass burning both transboundary and domestic sources (Wiriya et al., 2013). Domestic sources of haze phenomena in northern Thailand were predominantly from forest fires, which occurred mostly in deciduous dipterocarp forest national parks, whereas the agricultural waste burning was minor. The total burnt area, in northern Thailand was about 1 Mha during Jan 1 to Apr 16, 2019: more than 90% of the burnt area was in the forest and only 5% in agricultural areas (GISTDA, 2020; Earthdata, 2020). Forest fires were triggered by human activities, including forest foraging (collecting wild plants and pursuing wild animals), land clearing, resource collection, accidents, land tenure conflicts, logging, etc. Forest fires in northern Thailand adversely affected the livelihood of people, leading to significant poverty.

Biomass burning is an important source of PM_0.1—accounting for more than 30% of the smoke particles (Hata et al., 2014)—though the dominant size of PM from biomass burning was in accumulation mode particles (PM_0.5-1.0) (Tekasakul et al., 2008; Hata et al., 2014; Phairuang et al., 2019; Dejchanchaiwong et al., 2020; Chomanee et al., 2020). Biomass burning emissions contain a significant number of carbonaceous aerosols, i.e., organic carbon (OC) and elemental carbon (EC), released into the atmosphere (Chow et al., 2004; Chow et al., 2005; Chow et al., 2007; Chuang et al., 2016; Lee et al., 2016; Pani et al., 2019). Carbon was a significant component of fine particles, accounting for 20–60% of PM_2.5 mass in urban areas (Phairuang et al., 2019). Elemental carbon was the primary emission, generated from carbon fuel-based combustion. Organic carbon can be formed by direct emission and through secondary pathways (Chu et al., 2005; Han et al., 2009). Carbonaceous aerosols are formed from various emission sources. Hence, the OC/EC ratio is an important identifier of PM emission sources (Chu et al., 2005; Pongpiachan et al., 2014b; Han et al., 2016; Haque et al., 2019). Recently, Pani et al. (2019) reported that the OC and EC concentrations for PM_2.5 in Chiang Mai, Thailand, in March 2015 were OC 46.6 ± 19.0 µg m^–3 (42.9 ± 5.0%) and EC 9.5 ± 4. µg m^–3 (8.9 ± 2.7%). The OC/EC ratios for PM_2.5 ranged from 4.9 ± 1.7 to 5.6 ± 0.8 with mean 5.2 ± 1.3, which suggested the main contribution from biomass burning. Similarly, Thepnuan et al. (2019) also described the source emission of PM in Chiang Mai, Thailand, where the annual average OC and EC concentrations, during 2014–2015, were 3.8 ± 2.5 and 1.1 ± 1.1 µg m^–3 for PM_0.1 and 2.3 ± 1.6 and 1.4 ± 1.1 µg m^–3 for PM_0.5-1.0 (Phairuang et al., 2019).

Recently, not only black carbon (BC) or EC, but also brown carbon (BrC), which is a fraction of OC, have been identified as light-absorbing carbonaceous aerosols in PM_2.5 (Pani et al., 2018, 2020, 2021; Tao et al., 2020; Zhang et al., 2021). Radiative forcing (RF) of biomass burning aerosols during dry season in Chiang Mai was –45.3 to –103.4 at the surface and –1.7 to +6.2 W m^–2 at top of atmosphere corresponding to low - high biomass burning emissions events respectively (Pani et al., 2018). A higher negative RF at the surface indicated a significant decrease in solar radiation, that was enhanced by atmospheric absorption. This agrees with results from a global chemical transport model (GEOS-Chem) coupled with the rapid radiative transfer model for GCMs (RRTMG) models to study contributions of BC and BrC to light absorption (Zhang et al., 2021). Contributions of BC and BrC were 60–67% and 33–40%, respectively, at 400 nm, consistent with the observation by Pani et al. (2021). Instantaneous radiative forcing of carbonaceous PM_2.5 in Chiang Mai, during the dry season, were about 4–9 times higher than those in rainy seasons. Also, Tao et al. (2020) reported that contributions of EC and OC to aerosol absorption coefficients ranged from 49 ± 5 to 79 ± 3% and 21 ± 3 to 51 ± 5%, respectively, in the 370–880 nm region. Hence, the strong light absorption in the ultraviolet and short visible wavelength was closely related with the biomass burning produced carbonaceous aerosols, including BC and BrC. All of these studies were based on PM_2.5. Thus, it is important to study these effects on finer particles, especially PM_0.1, because of their recently discovered adverse effects on health as mentioned earlier. However, investigations of PM_0.1 and its carbonaceous compositions to understand the levels of air pollution, sources, meteorological conditions and impacts in upper SEA, especially during the heavy haze episodes in northern Thailand, are limited and are needed to better understand these phenomena.

We measured PM_0.1 and carbonaceous aerosols, in particular the EC and OC behaviors, in northern Thailand, during 2017–2019. Atmospheric aerosol samples were collected in Chiang Mai during April 2017, March to April 2018 and March 2019, all during heavy haze periods in Chiang Mai, and compared to the normal period of May 2017 to February 2018. The physical characteristics of PM_0.1, were studied. OC and EC were analyzed and correlated with the hotspots in the path of the air mass transport by backward trajectories (BT) during the period and the OC/EC ratios were calculated to identify possible sources of PM_0.1. Influences from meteorological parameters were evaluated.

 
2 METHODS

 
2.1 Aerosol Sampling and Physical Characteristics Analysis

Chiang Mai is the largest and one of the most famous cities in the upper SEA: it attracts global tourists, as well as local business persons. It is located in the Chiang Mai-Lamphun basin: the geography consists of high mountains along north-south corridor, covered with forest (69.7%), and agriculture (23.7%) (Land Development Department, 2018). Chiang Mai is typically influenced by transboundary and domestic aerosol transports. Forest fires and agriculture waste burning in northern Thailand are the main sources of haze pollution. Also, aerosol transported from open biomass burning area in some parts of Myanmar and Laos affect the northern Thailand air quality (Thepnuan et al., 2019; Pani et al., 2019). The geography, as well as temperature inversions in the winter, causes air pollution, especially PM, to remain in the region for a prolonged time. Chiang Mai University (CMU), located in the center of Chiang Mai city, Thailand (18°48'09.0''N, 98°57'12.3''E), was selected as a sampling site to confirm the results of physical and chemical characteristics of nanoparticles (PM_0.1) in the Chiang Mai urban area, as shown in Fig. 1. Samples were taken on the 5^th floor of a building, ~16 m above ground to study atmospheric air quality with minimum disturbance from possible local sources (U.S. EPA, 2016). In the present study, we designated the "haze period" as March to April, when significant numbers of hotspots in upper SEA were observed, resulting in high PM_2.5 concentration in the dry season, as suggested by Pani et al. (2019) and Thepnuan et al. (2019). The remaining months were labelled the "non-haze period."

Fig. 1. Sampling locations in Chiang Mai, Thailand.

Ambient aerosol was sampled using a cascade air sampler capable of segregating particle sizes down to 100 nm, so called PM_0.1 sampler or nanosampler. The operating air flow was 40 L min^–1 and sampling time was 24 hr, during the haze periods, and 48 hr, during non-haze periods. Particles were separated into six size ranges: < 0.1 µm (PM_0.1), 0.1–0.5 µm (PM_0.1‑0.5), 0.5–1.0 µm (PM_0.5-1), 1.0–2.5 µm (PM_1-2.5), 2.5–10 µm (PM_2.5-10) and > 10 µm (PM_>10). A 55-mm quartz fiber filter (2500QAT-UP, Pallflex, USA) was used for particle collection in each stage. The stage with an inertial filter (IF) used a stainless-steel filter pack (SUS304, fiber diameter, 9.8 µm) to collect PM_0.1-0.5. Before and after collection, the quartz fiber filters were stored in a desiccator with controlled relative humidity 50 ± 5% at room temperature for 72 hrs (Dejchanchaiwong et al., 2020). The filter was then weighed by a 0.01 mg resolution balance (CP225D, Sartorius, Germany).

Five samples during April 2017 and March to April 2018 and seven samples during May 2017–February 2018 were collected to represent 2017–2018 haze and 2017–2018 non-haze periods.

Seven samples at the same site during the strong haze periods in March 2019 were collected. Sampling details are shown in Table S1. In the following text, ‘haze periods’ refers to the strong haze periods in years 2017, 2018 and 2019 and ‘non-haze periods’ refers to the reference, normal periods in years 2017 and 2018.

 
2.2 Organic and Elemental Carbon Analysis

The carbonaceous components in PM were investigated using a OCEC carbon analyzer (Model 5, Sunset Laboratory, USA), following the Interagency Monitoring of Protected Visual Environments Thermal/Optical Transmittance (IMPROVE_TOT) protocol. OC fractions, including OC1, OC2, OC3 and OC4, were taken at temperatures 120, 250, 450 and 550°C, respectively, in a non-oxidizing helium (He), while the EC fractions, EC1, EC2 and EC3, were taken at 550, 700, 800°C, respectively, in a 2% O₂ and 98% He. The pyrolyzed carbon fraction (PyC) occurred during the transition between EC and OC formation (Chow et al., 2004). Due to a limitation of the carbon aerosol analyzer, samples, collected by the inertial filter, could not be analyzed. Thus, we analyzed only the PM_0.1 stage to study the nanoparticles in the atmosphere. PM_0.1 samples, collected uniformly on a backup filter, were punched as 10 × 15 mm² rectangles using a cutter. Mass concentration of carbon was calculated from the carbon mass on a punched-out sample, along with the air sampling flow rate, following Phairuang et al. (2019).

Quality assurance (QA) and quality control (QC) measures were strictly followed. Field and laboratory blank filters were analyzed. The carbon concentration in the blank filters was subtracted from the samples. Before each carbon analysis, the total carbon (TC) value was calibrated by the reference, sucrose standard (3.206 µg C µL^–1) (Sunset Laboratory Inc., USA). The detection limit for the OC/EC carbon analyzer was 0.2 µg C cm^–2 for both OC and EC analysis. A linear correlation between the sucrose standard solution concentrations and measured concentration was found with R² = 0.995. Coefficient of variation of the instrument did not exceed 10% for TC to ensure precision.

The OC was typically derived from primary emission sources and photochemical reactions. Concentrations of primary organic carbon (POC) and secondary organic carbon (SOC) were estimated following the EC-tracer method based on the minimum OC/EC ratio. Since EC and POC are generally emitted from carbon fuel-based combustion processes, which are exclusively associated with primary emissions (Pongpiachan et al., 2014b; Pani et al., 2019), EC is a good tracer of primary combustion generated OC. The concentrations of POC and SOC were calculated following Eq. (1):

2.3 Backward Trajectories and Fire Hotspots-Air Mass Transport Mapping

Aerosol transport from the sources to receptors was simulated to generate 24-hour backward trajectories (BT) at 1,000 m above ground level (AGL) using the Hybrid Single-Particle Lagrangian Integrated Trajectory Model version 4 (HYSPLIT4) (National Oceanic and Atmospheric Administration, 2020). The sampling site (18.79°N, 98.95°E) was chosen to represent the receptor in Chiang Mai. Hotspots from open biomass burning were obtained from the NASA VIIRS 375 m active fire data (Earthdata, 2020). The VIIRS sensor was on the Suomi NPP and NOAA-20 satellites, which are on the same polar orbit, about 50 minutes apart. Each satellite crosses over Thailand twice a day during daylight and the night hours. The observed hotspots could be limited by the orbital times. The present study attempted to evaluate the effects of the fire hotspots on the air quality by assuming the hotspots occurring in the paths of air mass to the receptor influenced PM concentrations. The boundaries of these hotspots were built by connecting each adjacent end point of the air mass trajectories, as shown in Fig. S1. They were calculated by the Geographic Information System (GIS) method to create a shapefile of the area bounding the targeted hotspots. The number of hotspots inside the boundary shapefile was then counted.

 
3 RESULTS AND DISCUSSION

 
3.1 PM_0.1 Mass Concentration

The PM_0.1 mass concentrations in ambient air at the Chiang Mai (CMU sites) during the haze and non-haze periods are shown in Fig. 2. Concentrations during the haze periods ranged from 6.6 to 19.1 µg m^–3 with a mean of 12.1 ± 4.0 µg m^–3, while during the non-haze periods, they were 3.4 to 6.2 µg m^–3 with a mean of 4.4 ± 1.0 µg m^–3. The values during the haze episode were about 3 times higher than in the non-haze periods. The highest PM_0.1 concentration, 19.1 µg m^–3, occurred during 14–15 March 2019. Contribution of PM_0.1 was 13.0 ± 3.9% (8.8–20.1%) of the total suspended particulate (TSP) mass concentrations during the haze periods, and 12.9 ± 2.4% (8.8–15.4%) during the non-haze periods. The PM_0.1 mass fraction during the haze period in northern Thailand was slightly higher, but the difference was not considered significant. This agrees with measured PM_0.1 fractions during the haze period in southern Thailand (Chomanee et al., 2020). PM_2.5 concentrations obtained from the sampling are also shown in Fig. 2. The values were 46.8, 38.9–65.7 and 65.0–114.4 µg m^–3 during the 2017, 2018 and 2019 haze periods, respectively. They were 12.5–34.4 µg m^–3 during the 2017 non-haze period. The average contributions of PM_0.1 in PM_2.5 were 14.1%, 24.2 ± 2.3% and 13.7 ± 2.4% during the 2017, 2018 and 2019 haze periods, and 22.5 ± 7.0% during the non-haze period. The contributions in the haze periods (2017–2019) were quite different. The change of concentration of submicron particles made a large difference in the PM_0.1/PM_2.5 ratio, because the peak of the particle size distribution, during the haze period, as a result of biomass burning, was usually below 1 micron (Dejchanchaiwong et al., 2020). Ambient PM_2.5 concentrations were compared to those monitored by the Thailand Pollution Control Department (PCD), located in Mueang District, Chiang Mai Province (18.79°N, 98.98°E), 3.6 km SE of the sampling site. PM_2.5 concentrations from the sampling were in good agreement with those obtained at the PCD station. The highest PM_2.5 concentration was 114.4 µg m^–3 observed on 16–17 March 2019 compared to 151.8 µg m^–3 at the PCD station. However, the PCD station was in a high traffic city center area, explaining the difference.

Fig. 2. PM_0.1 and PM_2.5 concentrations during the haze (April 2017, March to April 2018 and March 2019) and non-haze periods (May 2017–Febuary 2018).

3.2 Correlation of PM Concentration with Meteorological Conditions and Fire Hotspots

Meteorological parameters, including precipitation, relative humidity (RH), temperature, wind speed (WS) and wind direction (WD), were obtained from the Thai Meteorological Department, Thailand, as shown in Table S2. Precipitation, during the non-haze period, ranged from 0 to 59.3 mm with mean of 20.1 ± 24.4 mm, and in the haze period from 0–8.5 mm, with mean of 3.1 ± 4.1 mm. Temperatures, during the non-haze and haze periods, were 28.0 ± 2.3°C and 25.3 ± 5.2°C, while RH were 73.5 ± 6.6% and 56.6 ± 7.4%, respectively. Fire hotspots were detected by the VIIRS satellite obtained from Fire Information for Resource Management System (Earthdata, 2020) and counted by the GIS method described in Sec 1.3.

Linear correlations between PM_0.1 concentration and meteorological parameters and number of fire hotspots are shown in Table 1. The best correlations for the entire period were PM_0.1 with relative humidity and number of hotspots (R² = 0.68 and 0.53, respectively). The positive slope of PM_0.1 versus number of hotspots indicated a strong influence from the biomass burning, which occurred in the dry season, as suggested by the negative correlation between PM_0.1 and RH correlation. On the other hand, correlations between PM_0.1 mass concentrations and precipitation, temperature and wind speed were not strong. High wind speeds and precipitation favored the dispersion of atmospheric pollutants and reduced PM concentrations, while low wind speeds and low precipitation helped pollution levels to rise. Typically, temperature inversion plays an important role in high accumulation of air pollution during the winter months. In Chiang Mai, it usually occurs from November to March (Wiriya et al., 2013). During the 2018 haze periods, temperature inversion, from ground to 606 m, was observed only on 21 March. In 2019, it occurred from ground to 476–658 AGL during 12–25 March. Atmospheric temperature data was taken from the Department of Atmospheric Science, University of Wyoming, radiosonde database (University of Wyoming, 2021). This caused low levels of planetary boundary layer (PBL) and the aerosols, including PM_0.1, to be trapped and difficult to ventilate.

 
3.3 OC and EC Characteristics during Haze and Non-haze Periods in PM_0.1 Particles

The average OC, EC, total carbon (TC), POC and SOC concentrations, TC/PM, EC/TC, OC/EC and effective carbon ratio (ECR) ratios (defined in Sec. 3.4) for PM_0.1 during the haze and non-haze periods in Chiang Mai are listed in Table 2. Average TC mass concentrations in PM_0.1 were 8.2 ± 3.2 µg m^–3 during the haze periods, approximately 4 times higher than the value of 2.4 ± 1.1 µg m^–3 during the non-haze periods. The high TC/PM portion was found in PM_0.1 during both the haze (0.7 ± 0.2) and non-haze (0.5 ± 0.3) periods. This agrees with previous studies by Phairuang et al. (2019) and Dejchanchaiwong et al. (2020), indicating that PM_0.1 particles were the significant carcinogenic emission during both periods. The average OC concentration in PM_0.1 during the haze periods was 6.8 ± 2.7 µg m^–3, nearly 4 times higher than that during the non-haze periods, where average OC concentration was 1.9 ± 0.9 µg m^–3. Average EC concentrations in PM_0.1 during the haze periods were 1.4 ± 0.5 µg m^-3, approximately 3 times higher than the non-haze periods. During haze periods, contributions to PM_0.1 mass concentration were OC (57.1 ± 14.2%) and EC (11.7 ± 3.3%), whereas during the non-haze periods, they were 40.7 ± 24.4% and 11.0 ± 6.3%, respectively. It is significant to note that carbonaceous aerosols were a large component of PM_0.1, accounting for 51.8 ± 30.57%–68.7 ± 16.9% of PM_0.1 mass.

 
3.4 Primary and Secondary Organic Carbon in PM_0.1

The fractional contributions of POC and SOC in total OC during the haze and non-haze periods are shown Fig. S2. The average POC and SOC concentrations in PM_0.1, during the haze period, were 5.2 ± 2.0 µg m^–3 (78.4 ± 13.9% in OC), and 1.6 ± 1.2 µg m^–3 (21.6 ± 13.9% in OC) respectively. On the other hand, the average POC and SOC during the non-haze periods were 1.6 ± 0.8 µg m^–3 (87.1 ± 12.7% in OC) and 0.3 ± 0.3 µg m^–3 (12.9 ± 12.7% in OC), respectively. The fractional contribution of SOC to total OC increased during the haze periods, compared to the non-haze periods, because the increasing PM concentration provided more organic precursors to react with the volatile organic compounds (VOCs) and scattered the solar radiation, leading to increasing SOC formation (Ramirez et al., 2018, Pani et al., 2019). However, this increase in SOC was not sufficient to overcome the dominance of POC, which confirmed the primary source of aerosol during both periods in the region, because POC is linked to primary emissions (Pani et al., 2019). This also agrees with results of OC and EC analysis from PM_2.5 in Poland (Błaszczak et al., 2020).

 
3.5 Effective Carbon Ratio (ECR)

The POC and SOC are produced with different mechanisms, i.e., different interactions with solar radiation (Ramirez et al., 2018). The POC and EC or black carbon (BC) commonly originate from the same emission sources, i.e., biomass burning, cooking and vehicles exhaust, and generally absorb solar radiation, whereas SOC mainly originates from photochemical reaction and scatters it (Safai et al., 2014). However, BrC, which is a fraction of OC, has recently been found to strongly absorb light in the ultraviolet and short visible wavelength regions (Pani et al., 2018; Pani et al., 2020; Pani et al., 2021). The effective carbon ratio (ECR) was estimated to understand the relation between carbonaceous aerosols and climate impact (Safai et al., 2014; Ramirez et al., 2018):

In the present study, the average ECRs in PM_0.1 were 0.3 ± 0.2 during the haze and 0.1 ± 0.1 during the non-haze periods, as shown in Table 2. A higher ECR value during the haze period was observed due to a higher SOC concentration. SOC was previously believed to be light scattering carbonaceous aerosol (Safai et al., 2014; Singh et al., 2016; Ramirez et al., 2018). The presence of BrC in SOC could reduce the strong light scattering characteristics. However, an increase of ECR as a result of higher SOC along with high PM_0.1 concentration during the haze period should lead to higher potential for global warming. However, lower ECRs due to the lower SOC concentrations during the non-haze period did not indicate significant atmospheric warming because of low PM concentration.

 
3.6 OC/EC Ratios of PM_0.1 Particles

The average OC/EC ratios in the present and previous studies in Thailand are shown in Table 3. Most studies used thermal optical reflectance (TOR). OC/EC ratios results, using IMPROVE_TOT methods in this study, were corrected to OC/EC ratios, based on the IMPROVE_TOR equation of Han et al. (2016), who compared OC and EC measurements in Xi’an, China, and are also shown in Table 3. Ratios of IMPROVE_TOT to IMPROVE_TOR were 1.22 for OC concentration (R² = 0.99) and 0.6 for EC concentration (R² = 0.95). Hence, discrepancies between TOR and TOT for OC and EC were smaller, when measured by the IMPROVE protocol. EC measured, using the TOT method, was lower than the TOR method. This is because PyC calculated by TOT was delayed and split the OC and EC more than values derived from reflectance measurements (Chow et al., 2004; Han et al., 2009; Han et al., 2016). In addition, Han et al. (2016) reported that OC/EC ratios in PM_2.5 of ambient air, by IMPROVE_TOT (OC/EC: 2.3–6.7), were approximately 2 times higher than those using IMPROVE_TOR (OC/EC: 1.4–3.7), while average OC/EC ratios for biomass combustion emission by IMPROVE_TOT (OC/EC: 4.6–19.3) were higher than those by IMPROVE_TOR (OC/EC 3.1–10.8). This agrees with Khan et al. (2012), who reported that the average OC/EC ratios for IMPROVE_TOT were higher than with IMPROVE_TOR, with average TOT/TOR ratio of 1.2 for diesel and 1.4 for cooking stove combustion emission.

In the present study, average OC/EC ratios in PM_0.1, based on the IMPROVE_TOT method, during haze periods, ranged from 3.6-6.8, with mean of 5.0 ± 1.0 and ranged from 3.0–4.2 (mean 3.7 ± 0.5), during non-haze periods. Results showed that the corrected OC/EC ratios in PM_0.1, based on IMPROVE_TOR, during the haze periods, ranged from 1.8-3.3 (mean 2.5 ± 0.5,) versus from 1.5–2.1 (mean 1.8 ± 0.3), during the non-haze periods. Phairuang et al. (2019) also reported PM_0.1 OC/EC ratios, in Chiang Mai, were about 3.4 and suggested the main contribution was from biomass burning. Khan et al. (2012) reported that the average OC/EC ratios based on the IMPROVE_TOT method in PM_2.5 were 0.1–0.97 for diesel exhaust. Additionally, Zhu et al. (2010) reported PM_0.1 OC/EC ratios, based on IMPROVE_TOR, were 0.68, due to motor vehicle fuel combustion. It can be seen that both the OC/EC ratios, exceeding 1.0, based on IMPROVE_TOT and IMPROVE_TOR methods, can be attributed to the biomass burning emission. Hence, the high values of OC/EC based on IMPROVE_TOT and IMPROVE_TOR methods in PM_0.1 during haze periods in the present study can be explained by increasing contribution from biomass burning emissions. This is in agreement with the studies of PM emission sources in Chiang Mai, Thailand (Pongpiachan et al., 2013; Pani et al., 2019). The vehicular emissions dominated EC during the non-haze period because of the generally high concentration of traffic in Chiang Mai city. This was confirmed by the low OC/EC ratios.

 
3.7 Distributions of Carbonaceous Components in PM_0.1

Eight carbon fractions are often used for source identification of the carbonaceous aerosols. OC3 and EC1-PyC are the most abundant fractions of OC in biomass burning (Cao et al., 2005; Chuang et al., 2013; Pani et al., 2019). OC2 represents the secondary organic carbon (SOC) formation, as well as coal combustion and biomass burning, and is also the major fraction of OC in motor vehicle exhaust (Chow et al., 2004). Moreover, PyC represents incomplete combustion during cooking and biomass burning (Cao et al., 2005). EC2 and EC3 are good indicators of hot-gas evaporation from motor vehicles (Phairuang et al., 2019). Eight carbonaceous fractions and concentrations in PM_0.1 during the haze and non-haze periods from this study are shown in Fig. S3. During the haze periods, the order of OC and EC fraction in PM_0.1 was PyC > OC3 > OC2 > OC4 > EC2 > (EC1 – PyC)> EC3 > OC1, while, during the non-haze periods, it was OC2 > OC3 > OC4 > PyC > EC2 > EC3 > (EC1 – PyC) > OC1.

OC3 and OC2 were the most abundant fractions of OC in PM_0.1, during both haze and non-haze periods, indicating the significant contribution of biomass burning to carbonaceous aerosols. The OC1 fraction in PM_0.1 mostly contains volatile organic substances, which can rapidly evaporate during long-range transport (Lee et al., 2016). Our results agree with a previous study of OC and EC distributions to PM_0.1 (Phairuang et al., 2019)

 
3.8 Correlations between OC and EC, and OC/EC and Number of Hotpots

Correlation analysis is usually used to determine whether the pollution sources of OC and EC are consistent (Han et al., 2009; Zhang et al., 2011; He et al., 2015; Guo, 2016; Wang et al., 2017). Correlations between OC and EC in PM_0.1, during the haze and non-haze periods, are shown in Fig. 3. Strong OC and EC correlations during the non-haze periods (R² = 0.91) suggested impacts from a combination of common and consistent sources, particularly motor vehicle exhaust and biomass burning, which was dominant in the area (Cao et al., 2007). However, the correlation was lower during the haze periods (R² = 0.80), implying a more complex pattern of emission sources (Cao et al., 2007). The OC vs. EC slope during the haze periods (2.28) was higher than during the non-haze periods (1.84). This indicated a larger contribution of OC in the carbonaceous aerosol from primary emissions of various biomass burnings during the haze periods: the major emission source at those times in the north of Thailand is fire in deciduous forests. The other major source was agricultural waste (Wiriya et al., 2013, 2016; Thepnuan et al., 2019; Pani et al., 2019). The intercepts between OC & EC relations were very small and acceptably close to the ideal zeroes, when no sources were present.

Fig. 3. Correlation OC vs. EC in PM_0.1 during non-haze and haze periods.

Influence from biomass burning can be confirmed by the number of fire hotspots in the path of the wind to the receptor. During the haze periods, OC/EC ratio increased linearly with hotspot numbers and at nearly twice the slope in the non-haze periods, see Fig. 4. The positive correlation was stronger for the haze periods (R² = 0.42) than the non-haze period (R² = 0.26). This suggested larger contributions from biomass burning when the hotspots were increased. The intercepts from both correlations, 1.77 and 1.97, were very close and confirmed the good agreement, when no hotspots were present. The low OC/EC ratios indicated the PM_0.1 from other sources, particularly emission from diesel engines, because Chiang Mai is a congested city with significant traffic problems (Pongpiachan et al., 2014a; Phairuang et al., 2019; Thepnuan et al., 2019). The hotspots-air mass transport mapping technique used here allowed an appropriate evaluation of the biomass burning source contribution to the PM_0.1 concentration at the receptor.

Fig. 4. Correlation between OC/EC ration in PM_0.1 and number of hotspots during non-haze and haze periods.

 
3.9 Backward Trajectories

Backward trajectory simulations for Chiang Mai during the haze and the non-haze periods are shown in Fig. S4. High PM_0.1 concentrations during the haze periods were correlated with a trajectory from south to north and west to east. Wind simulation on April 3, 2017 (Fig. S4(a)) showed that the east and west wind moved towards Chiang Mai mostly from northern Thailand and also some parts of Laos and Myanmar, where partial hotspots were observed. Backward trajectories, on March 2018 and 2019, showed west and southwest bound air masses, passing through some hotspots in Myanmar, to Chiang Mai (Figs. S4(b) and S4(d)). Also, south wind flowed to Chiang Mai from the west and northern Thailand, where high numbers of biomass burning hotspots, from agriculture residues, were observed, along with some parts of Myanmar (Fig. S4(c)). Open biomass burning in the forest has been a major local source affecting air quality in northern Thailand. It contributed 90% of about 1 Mha of burnt area during the 2019 dry season, whereas the remainder was from agricultural waste burning (GISDA, 2020). Simulation during the non-haze periods, on the other hand, showed that the air mass moved from southwest to Chiang Mai in August 2017 carrying clean air from Indian Ocean and no hotspots occurred during this period (Fig. S4(e)). Wind trajectories in October 2017 clearly confirmed that that there was no significant domestic and transboundary haze transport (Fig. S4(f)). Hence, the 48-hrs backward trajectories during the haze periods showed aerosol transported from open biomass burning areas, in Thailand and some parts of Myanmar, affected the Chiang Mai air quality, especially its PM concentrations due to east to west winds. This confirms previous studies of potential PM sources in Chiang Mai (Wiriya et al., 2013; Wiriya et al., 2016; Thepnuan et al., 2019; Pani et al., 2019; Phairuang et al., 2019).

 
4 CONCLUSIONS

Physical characteristics and carbonaceous components in PM_0.1, during the haze and non-haze periods in Chiang Mai, Thailand were investigated to identify and characterize sources of haze pollution. We conclude:

• The PM₁ level during the haze periods was significantly higher: the average PM_0.1 concentration during the haze periods (12.1 ± 4.0 µg m^–3) was about 3 times as high as the non-haze periods (4.4 ± 1.0 µg m^–3). PM_0.1 concentration was strongly negatively correlated with relative humidity and positively correlated with hotspot numbers.
  
  
• OC concentrations during the haze and non-haze periods were 6.8 ±7 µg m^–3 and 1.9 ± 0.9 µg m^–3, whereas EC concentrations were 1.4 ± 0.5 and 0.5 ± 0.2 µg m^–3, respectively, i.e., OC and EC concentrations were about 3 and 4 times higher during the haze periods. The OC fraction in PM_0.1 was also affected by long-range transport and secondary organic aerosol formation, because the contribution of SOC to total OC during the haze periods also increased relative to the non-haze periods.
  
  
• Strong OC and EC correlation (R² = 0.91), during non-haze periods, indicated a combination of common and consistent sources, whereas the lower correlation (R² = 0.80) during haze period suggested a more complex pattern of emission source. The OC/EC ratio and number of hotspots correlation was stronger for the haze periods (R² = 0.42) than the non-haze period (R² = 0.26).
  
  
• The high PM₁ concentration and OC/EC ratio indicated more contribution from various biomass burning to the PM_0.1 during the haze periods. The 48-hr backward trajectories confirmed this. An increase of OC and EC, the light-absorbing carbonaceous aerosols, may also significantly impact regional air quality and climate in Chiang Mai and upper SEA.

This study enhanced knowledge of carbonaceous aerosol in PM_0.1 over upper SEA and emission control strategies on PM_0.1 of ambient air. This should be of concern because PM_0.1 has adverse effects on humans and the atmospheric environment.

 
ACKNOWLEDGMENTS

This research was financially supported by the Biodiversity-based Economy Development Office (BEDO) of Thailand under grant number ENG590707S. This work was also supported by the Interdisciplinary Graduate School of Energy Systems, Prince of Songkla University under grant number 1-2018/08. We would like to thank Chevron Thailand Exploration and Production, Ltd. for manufacturing a nice and sharp filter cutter. Additionally, we would like to thank the Faculty of Environment, Kasetsart University for providing the OCEC carbon analyzer. We acknowledge Dr. Worradorn Phairuang for his carbon analysis consultation, and are also thankful to John Morris for English language editing.